* [PATCH v12 01/17] riscv: Rename __switch_to_aux -> fpu
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Guo Ren, Greentime Hu, Anup Patel, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Atish Patra, Oleg Nesterov, Guo Ren,
Heinrich Schuchardt, Conor Dooley, Arnaud Pouliquen,
Chris Stillson, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Vincent Chen, Heiko Stuebner, Dao Lu, Jisheng Zhang, Sunil V L,
Nick Knight, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Paul E. McKenney, Frederic Weisbecker,
Mark Rutland, Vitaly Wool, Myrtle Shah, Ruinland Tsai,
Catalin Marinas, Mark Brown, Will Deacon, Alexey Dobriyan,
Huacai Chen, Janosch Frank, Christian Brauner,
Eugene Syromiatnikov, Peter Collingbourne, Colin Cross,
Andrew Morton, Suren Baghdasaryan, Barret Rhoden,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <ren_guo@c-sky.com>
The name of __switch_to_aux is not clear and rename it with the
determine function: __switch_to_fpu. Next we could add other regs'
switch.
Signed-off-by: Guo Ren <ren_guo@c-sky.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/asm/switch_to.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index 11463489fec6..df1aa589b7fd 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -46,7 +46,7 @@ static inline void fstate_restore(struct task_struct *task,
}
}
-static inline void __switch_to_aux(struct task_struct *prev,
+static inline void __switch_to_fpu(struct task_struct *prev,
struct task_struct *next)
{
struct pt_regs *regs;
@@ -65,7 +65,7 @@ static __always_inline bool has_fpu(void)
static __always_inline bool has_fpu(void) { return false; }
#define fstate_save(task, regs) do { } while (0)
#define fstate_restore(task, regs) do { } while (0)
-#define __switch_to_aux(__prev, __next) do { } while (0)
+#define __switch_to_fpu(__prev, __next) do { } while (0)
#endif
extern struct task_struct *__switch_to(struct task_struct *,
@@ -76,7 +76,7 @@ do { \
struct task_struct *__prev = (prev); \
struct task_struct *__next = (next); \
if (has_fpu()) \
- __switch_to_aux(__prev, __next); \
+ __switch_to_fpu(__prev, __next); \
((last) = __switch_to(__prev, __next)); \
} while (0)
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 01/17] riscv: Rename __switch_to_aux -> fpu
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Guo Ren, Greentime Hu, Anup Patel, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Atish Patra, Oleg Nesterov, Guo Ren,
Heinrich Schuchardt, Conor Dooley, Arnaud Pouliquen,
Chris Stillson, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Vincent Chen, Heiko Stuebner, Dao Lu, Jisheng Zhang, Sunil V L,
Nick Knight, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Paul E. McKenney, Frederic Weisbecker,
Mark Rutland, Vitaly Wool, Myrtle Shah, Ruinland Tsai,
Catalin Marinas, Mark Brown, Will Deacon, Alexey Dobriyan,
Huacai Chen, Janosch Frank, Christian Brauner,
Eugene Syromiatnikov, Peter Collingbourne, Colin Cross,
Andrew Morton, Suren Baghdasaryan, Barret Rhoden,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <ren_guo@c-sky.com>
The name of __switch_to_aux is not clear and rename it with the
determine function: __switch_to_fpu. Next we could add other regs'
switch.
Signed-off-by: Guo Ren <ren_guo@c-sky.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/asm/switch_to.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index 11463489fec6..df1aa589b7fd 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -46,7 +46,7 @@ static inline void fstate_restore(struct task_struct *task,
}
}
-static inline void __switch_to_aux(struct task_struct *prev,
+static inline void __switch_to_fpu(struct task_struct *prev,
struct task_struct *next)
{
struct pt_regs *regs;
@@ -65,7 +65,7 @@ static __always_inline bool has_fpu(void)
static __always_inline bool has_fpu(void) { return false; }
#define fstate_save(task, regs) do { } while (0)
#define fstate_restore(task, regs) do { } while (0)
-#define __switch_to_aux(__prev, __next) do { } while (0)
+#define __switch_to_fpu(__prev, __next) do { } while (0)
#endif
extern struct task_struct *__switch_to(struct task_struct *,
@@ -76,7 +76,7 @@ do { \
struct task_struct *__prev = (prev); \
struct task_struct *__next = (next); \
if (has_fpu()) \
- __switch_to_aux(__prev, __next); \
+ __switch_to_fpu(__prev, __next); \
((last) = __switch_to(__prev, __next)); \
} while (0)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 02/17] riscv: Extending cpufeature.c to detect V-extension
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Guo Ren, Anup Patel, Greentime Hu, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Atish Patra, Oleg Nesterov, Heinrich Schuchardt,
Guo Ren, Conor Dooley, Mayuresh Chitale, Chris Stillson,
Paolo Bonzini, Qinglin Pan, Alexandre Ghiti, Arnd Bergmann,
Vincent Chen, Heiko Stuebner, Dao Lu, Jisheng Zhang,
Geert Uytterhoeven, Sunil V L, Nick Knight, Han-Kuan Chen,
Changbin Du, Li Zhengyu, Alexander Graf, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Will Deacon, Mark Brown,
Heiko Carstens, Alexey Dobriyan, Huacai Chen, Christian Brauner,
Eugene Syromiatnikov, Peter Collingbourne, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <ren_guo@c-sky.com>
Current cpufeature.c doesn't support detecting V-extension, because
"rv64" also contain a 'v' letter and we need to skip it.
Signed-off-by: Guo Ren <ren_guo@c-sky.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/uapi/asm/hwcap.h | 1 +
arch/riscv/kernel/cpufeature.c | 1 +
2 files changed, 2 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/hwcap.h b/arch/riscv/include/uapi/asm/hwcap.h
index 46dc3f5ee99f..c52bb7bbbabe 100644
--- a/arch/riscv/include/uapi/asm/hwcap.h
+++ b/arch/riscv/include/uapi/asm/hwcap.h
@@ -21,5 +21,6 @@
#define COMPAT_HWCAP_ISA_F (1 << ('F' - 'A'))
#define COMPAT_HWCAP_ISA_D (1 << ('D' - 'A'))
#define COMPAT_HWCAP_ISA_C (1 << ('C' - 'A'))
+#define COMPAT_HWCAP_ISA_V (1 << ('V' - 'A'))
#endif /* _UAPI_ASM_RISCV_HWCAP_H */
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 553d755483ed..8d4448c2d4f4 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -83,6 +83,7 @@ void __init riscv_fill_hwcap(void)
isa2hwcap['f'] = isa2hwcap['F'] = COMPAT_HWCAP_ISA_F;
isa2hwcap['d'] = isa2hwcap['D'] = COMPAT_HWCAP_ISA_D;
isa2hwcap['c'] = isa2hwcap['C'] = COMPAT_HWCAP_ISA_C;
+ isa2hwcap['v'] = isa2hwcap['V'] = COMPAT_HWCAP_ISA_V;
elf_hwcap = 0;
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 02/17] riscv: Extending cpufeature.c to detect V-extension
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Guo Ren, Anup Patel, Greentime Hu, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Atish Patra, Oleg Nesterov, Heinrich Schuchardt,
Guo Ren, Conor Dooley, Mayuresh Chitale, Chris Stillson,
Paolo Bonzini, Qinglin Pan, Alexandre Ghiti, Arnd Bergmann,
Vincent Chen, Heiko Stuebner, Dao Lu, Jisheng Zhang,
Geert Uytterhoeven, Sunil V L, Nick Knight, Han-Kuan Chen,
Changbin Du, Li Zhengyu, Alexander Graf, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Will Deacon, Mark Brown,
Heiko Carstens, Alexey Dobriyan, Huacai Chen, Christian Brauner,
Eugene Syromiatnikov, Peter Collingbourne, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <ren_guo@c-sky.com>
Current cpufeature.c doesn't support detecting V-extension, because
"rv64" also contain a 'v' letter and we need to skip it.
Signed-off-by: Guo Ren <ren_guo@c-sky.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/uapi/asm/hwcap.h | 1 +
arch/riscv/kernel/cpufeature.c | 1 +
2 files changed, 2 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/hwcap.h b/arch/riscv/include/uapi/asm/hwcap.h
index 46dc3f5ee99f..c52bb7bbbabe 100644
--- a/arch/riscv/include/uapi/asm/hwcap.h
+++ b/arch/riscv/include/uapi/asm/hwcap.h
@@ -21,5 +21,6 @@
#define COMPAT_HWCAP_ISA_F (1 << ('F' - 'A'))
#define COMPAT_HWCAP_ISA_D (1 << ('D' - 'A'))
#define COMPAT_HWCAP_ISA_C (1 << ('C' - 'A'))
+#define COMPAT_HWCAP_ISA_V (1 << ('V' - 'A'))
#endif /* _UAPI_ASM_RISCV_HWCAP_H */
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 553d755483ed..8d4448c2d4f4 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -83,6 +83,7 @@ void __init riscv_fill_hwcap(void)
isa2hwcap['f'] = isa2hwcap['F'] = COMPAT_HWCAP_ISA_F;
isa2hwcap['d'] = isa2hwcap['D'] = COMPAT_HWCAP_ISA_D;
isa2hwcap['c'] = isa2hwcap['C'] = COMPAT_HWCAP_ISA_C;
+ isa2hwcap['v'] = isa2hwcap['V'] = COMPAT_HWCAP_ISA_V;
elf_hwcap = 0;
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 02/17] riscv: Extending cpufeature.c to detect V-extension
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Guo Ren, Anup Patel, Greentime Hu, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Atish Patra, Oleg Nesterov, Heinrich Schuchardt,
Guo Ren, Conor Dooley, Mayuresh Chitale, Chris Stillson,
Paolo Bonzini, Qinglin Pan, Alexandre Ghiti, Arnd Bergmann,
Vincent Chen, Heiko Stuebner, Dao Lu, Jisheng Zhang,
Geert Uytterhoeven, Sunil V L, Nick Knight, Han-Kuan Chen,
Changbin Du, Li Zhengyu, Alexander Graf, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Will Deacon, Mark Brown,
Heiko Carstens, Alexey Dobriyan, Huacai Chen, Christian Brauner,
Eugene Syromiatnikov, Peter Collingbourne, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <ren_guo@c-sky.com>
Current cpufeature.c doesn't support detecting V-extension, because
"rv64" also contain a 'v' letter and we need to skip it.
Signed-off-by: Guo Ren <ren_guo@c-sky.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/uapi/asm/hwcap.h | 1 +
arch/riscv/kernel/cpufeature.c | 1 +
2 files changed, 2 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/hwcap.h b/arch/riscv/include/uapi/asm/hwcap.h
index 46dc3f5ee99f..c52bb7bbbabe 100644
--- a/arch/riscv/include/uapi/asm/hwcap.h
+++ b/arch/riscv/include/uapi/asm/hwcap.h
@@ -21,5 +21,6 @@
#define COMPAT_HWCAP_ISA_F (1 << ('F' - 'A'))
#define COMPAT_HWCAP_ISA_D (1 << ('D' - 'A'))
#define COMPAT_HWCAP_ISA_C (1 << ('C' - 'A'))
+#define COMPAT_HWCAP_ISA_V (1 << ('V' - 'A'))
#endif /* _UAPI_ASM_RISCV_HWCAP_H */
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 553d755483ed..8d4448c2d4f4 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -83,6 +83,7 @@ void __init riscv_fill_hwcap(void)
isa2hwcap['f'] = isa2hwcap['F'] = COMPAT_HWCAP_ISA_F;
isa2hwcap['d'] = isa2hwcap['D'] = COMPAT_HWCAP_ISA_D;
isa2hwcap['c'] = isa2hwcap['C'] = COMPAT_HWCAP_ISA_C;
+ isa2hwcap['v'] = isa2hwcap['V'] = COMPAT_HWCAP_ISA_V;
elf_hwcap = 0;
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread[parent not found: <4b6e20fb-d013-0a09-0b74-b6c46e045af3@rivosinc.com>]
* [PATCH v12 03/17] riscv: Add new csr defines related to vector extension
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Guo Ren, Guo Ren, Vincent Chen, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Anup Patel, Atish Patra, Oleg Nesterov,
Heinrich Schuchardt, Arnaud Pouliquen, Mayuresh Chitale,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L,
Han-Kuan Chen, Li Zhengyu, Alexander Graf, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Changbin Du, hasheddan, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Andrew Morton, Alexey Dobriyan, Huacai Chen, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Colin Cross,
Peter Collingbourne, Eugene Syromiatnikov, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Follow the riscv vector spec to add new csr numbers.
[guoren@linux.alibaba.com: first porting for new vector related csr]
Acked-by: Guo Ren <guoren@kernel.org>
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/asm/csr.h | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
index 0e571f6483d9..78a68f29c3ba 100644
--- a/arch/riscv/include/asm/csr.h
+++ b/arch/riscv/include/asm/csr.h
@@ -24,6 +24,12 @@
#define SR_FS_CLEAN _AC(0x00004000, UL)
#define SR_FS_DIRTY _AC(0x00006000, UL)
+#define SR_VS _AC(0x00000600, UL) /* Vector Status */
+#define SR_VS_OFF _AC(0x00000000, UL)
+#define SR_VS_INITIAL _AC(0x00000200, UL)
+#define SR_VS_CLEAN _AC(0x00000400, UL)
+#define SR_VS_DIRTY _AC(0x00000600, UL)
+
#define SR_XS _AC(0x00018000, UL) /* Extension Status */
#define SR_XS_OFF _AC(0x00000000, UL)
#define SR_XS_INITIAL _AC(0x00008000, UL)
@@ -31,9 +37,9 @@
#define SR_XS_DIRTY _AC(0x00018000, UL)
#ifndef CONFIG_64BIT
-#define SR_SD _AC(0x80000000, UL) /* FS/XS dirty */
+#define SR_SD _AC(0x80000000, UL) /* FS/VS/XS dirty */
#else
-#define SR_SD _AC(0x8000000000000000, UL) /* FS/XS dirty */
+#define SR_SD _AC(0x8000000000000000, UL) /* FS/VS/XS dirty */
#endif
#ifdef CONFIG_64BIT
@@ -297,6 +303,12 @@
#define CSR_MIMPID 0xf13
#define CSR_MHARTID 0xf14
+#define CSR_VSTART 0x8
+#define CSR_VCSR 0xf
+#define CSR_VL 0xc20
+#define CSR_VTYPE 0xc21
+#define CSR_VLENB 0xc22
+
#ifdef CONFIG_RISCV_M_MODE
# define CSR_STATUS CSR_MSTATUS
# define CSR_IE CSR_MIE
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 03/17] riscv: Add new csr defines related to vector extension
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Guo Ren, Guo Ren, Vincent Chen, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Anup Patel, Atish Patra, Oleg Nesterov,
Heinrich Schuchardt, Arnaud Pouliquen, Mayuresh Chitale,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L,
Han-Kuan Chen, Li Zhengyu, Alexander Graf, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Changbin Du, hasheddan, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Andrew Morton, Alexey Dobriyan, Huacai Chen, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Colin Cross,
Peter Collingbourne, Eugene Syromiatnikov, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Follow the riscv vector spec to add new csr numbers.
[guoren@linux.alibaba.com: first porting for new vector related csr]
Acked-by: Guo Ren <guoren@kernel.org>
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/asm/csr.h | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
index 0e571f6483d9..78a68f29c3ba 100644
--- a/arch/riscv/include/asm/csr.h
+++ b/arch/riscv/include/asm/csr.h
@@ -24,6 +24,12 @@
#define SR_FS_CLEAN _AC(0x00004000, UL)
#define SR_FS_DIRTY _AC(0x00006000, UL)
+#define SR_VS _AC(0x00000600, UL) /* Vector Status */
+#define SR_VS_OFF _AC(0x00000000, UL)
+#define SR_VS_INITIAL _AC(0x00000200, UL)
+#define SR_VS_CLEAN _AC(0x00000400, UL)
+#define SR_VS_DIRTY _AC(0x00000600, UL)
+
#define SR_XS _AC(0x00018000, UL) /* Extension Status */
#define SR_XS_OFF _AC(0x00000000, UL)
#define SR_XS_INITIAL _AC(0x00008000, UL)
@@ -31,9 +37,9 @@
#define SR_XS_DIRTY _AC(0x00018000, UL)
#ifndef CONFIG_64BIT
-#define SR_SD _AC(0x80000000, UL) /* FS/XS dirty */
+#define SR_SD _AC(0x80000000, UL) /* FS/VS/XS dirty */
#else
-#define SR_SD _AC(0x8000000000000000, UL) /* FS/XS dirty */
+#define SR_SD _AC(0x8000000000000000, UL) /* FS/VS/XS dirty */
#endif
#ifdef CONFIG_64BIT
@@ -297,6 +303,12 @@
#define CSR_MIMPID 0xf13
#define CSR_MHARTID 0xf14
+#define CSR_VSTART 0x8
+#define CSR_VCSR 0xf
+#define CSR_VL 0xc20
+#define CSR_VTYPE 0xc21
+#define CSR_VLENB 0xc22
+
#ifdef CONFIG_RISCV_M_MODE
# define CSR_STATUS CSR_MSTATUS
# define CSR_IE CSR_MIE
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 03/17] riscv: Add new csr defines related to vector extension
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Guo Ren, Guo Ren, Vincent Chen, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Anup Patel, Atish Patra, Oleg Nesterov,
Heinrich Schuchardt, Arnaud Pouliquen, Mayuresh Chitale,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L,
Han-Kuan Chen, Li Zhengyu, Alexander Graf, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Changbin Du, hasheddan, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Andrew Morton, Alexey Dobriyan, Huacai Chen, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Colin Cross,
Peter Collingbourne, Eugene Syromiatnikov, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Follow the riscv vector spec to add new csr numbers.
[guoren@linux.alibaba.com: first porting for new vector related csr]
Acked-by: Guo Ren <guoren@kernel.org>
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/include/asm/csr.h | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h
index 0e571f6483d9..78a68f29c3ba 100644
--- a/arch/riscv/include/asm/csr.h
+++ b/arch/riscv/include/asm/csr.h
@@ -24,6 +24,12 @@
#define SR_FS_CLEAN _AC(0x00004000, UL)
#define SR_FS_DIRTY _AC(0x00006000, UL)
+#define SR_VS _AC(0x00000600, UL) /* Vector Status */
+#define SR_VS_OFF _AC(0x00000000, UL)
+#define SR_VS_INITIAL _AC(0x00000200, UL)
+#define SR_VS_CLEAN _AC(0x00000400, UL)
+#define SR_VS_DIRTY _AC(0x00000600, UL)
+
#define SR_XS _AC(0x00018000, UL) /* Extension Status */
#define SR_XS_OFF _AC(0x00000000, UL)
#define SR_XS_INITIAL _AC(0x00008000, UL)
@@ -31,9 +37,9 @@
#define SR_XS_DIRTY _AC(0x00018000, UL)
#ifndef CONFIG_64BIT
-#define SR_SD _AC(0x80000000, UL) /* FS/XS dirty */
+#define SR_SD _AC(0x80000000, UL) /* FS/VS/XS dirty */
#else
-#define SR_SD _AC(0x8000000000000000, UL) /* FS/XS dirty */
+#define SR_SD _AC(0x8000000000000000, UL) /* FS/VS/XS dirty */
#endif
#ifdef CONFIG_64BIT
@@ -297,6 +303,12 @@
#define CSR_MIMPID 0xf13
#define CSR_MHARTID 0xf14
+#define CSR_VSTART 0x8
+#define CSR_VCSR 0xf
+#define CSR_VL 0xc20
+#define CSR_VTYPE 0xc21
+#define CSR_VLENB 0xc22
+
#ifdef CONFIG_RISCV_M_MODE
# define CSR_STATUS CSR_MSTATUS
# define CSR_IE CSR_MIE
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* Re: [PATCH v12 03/17] riscv: Add new csr defines related to vector extension
2022-09-21 21:43 ` Chris Stillson
@ 2023-01-23 11:24 ` Heiko Stübner
-1 siblings, 0 replies; 147+ messages in thread
From: Heiko Stübner @ 2023-01-23 11:24 UTC (permalink / raw)
To: linux-riscv
Cc: Greentime Hu, Guo Ren, Guo Ren, Vincent Chen, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Anup Patel, Atish Patra, Oleg Nesterov,
Heinrich Schuchardt, Arnaud Pouliquen, Mayuresh Chitale,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Arnd Bergmann, Jisheng Zhang, Dao Lu, Sunil V L, Han-Kuan Chen,
Li Zhengyu, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Changbin Du, hasheddan, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Andrew Morton, Alexey Dobriyan, Huacai Chen, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Colin Cross,
Peter Collingbourne, Eugene Syromiatnikov, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv, Chris Stillson
Am Mittwoch, 21. September 2022, 23:43:45 CET schrieb Chris Stillson:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> Follow the riscv vector spec to add new csr numbers.
>
> [guoren@linux.alibaba.com: first porting for new vector related csr]
> Acked-by: Guo Ren <guoren@kernel.org>
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 03/17] riscv: Add new csr defines related to vector extension
@ 2023-01-23 11:24 ` Heiko Stübner
0 siblings, 0 replies; 147+ messages in thread
From: Heiko Stübner @ 2023-01-23 11:24 UTC (permalink / raw)
To: linux-riscv
Cc: Mark Rutland, Guo Ren, kvm, Catalin Marinas, Paul Walmsley,
Palmer Dabbelt, linux-kernel, linux-mm, Guo Ren, Jisheng Zhang,
Qinglin Pan, Nicolas Saenz Julienne, Will Deacon, Ard Biesheuvel,
Vitaly Wool, Davidlohr Bueso, Yury Norov, Changbin Du,
Anup Patel, Huacai Chen, Myrtle Shah, Eugene Syromiatnikov,
linux-riscv, Tsukasa OI, Greentime Hu, Arnaud Pouliquen,
Li Zhengyu, Evgenii Stepanov, Heinrich Schuchardt, Barret Rhoden,
Albert Ou, Kees Cook, Arnd Bergmann, Frederic Weisbecker,
Suren Baghdasaryan, Dao Lu, Alexander Graf, Atish Patra,
Mayuresh Chitale, Alexandre Ghiti, Peter Collingbourne,
Janosch Frank, Chris Stillson, Christian Brauner, hasheddan,
Oleg Nesterov, Colin Cross, Alexey Dobriyan, Han-Kuan Chen,
Vincent Chen, Mark Brown, Palmer Dabbelt, Eric Biederman,
kvm-riscv, Paolo Bonzini, Andrew Morton
Am Mittwoch, 21. September 2022, 23:43:45 CET schrieb Chris Stillson:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> Follow the riscv vector spec to add new csr numbers.
>
> [guoren@linux.alibaba.com: first porting for new vector related csr]
> Acked-by: Guo Ren <guoren@kernel.org>
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 04/17] riscv: Add vector feature to compile
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Greentime Hu, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Vincent Chen, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Mark Brown, Will Deacon,
Heiko Carstens, Huacai Chen, Alexey Dobriyan, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Guo Ren <guoren@linux.alibaba.com>
This patch adds a new config option which could enable assembler's
vector feature.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/Kconfig | 15 +++++++++++++--
arch/riscv/Makefile | 1 +
2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index ed66c31e4655..e294d85bfb7d 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -432,7 +432,17 @@ config FPU
If you don't know what to do here, say Y.
-endmenu # "Platform type"
+config VECTOR
+ bool "VECTOR support"
+ depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
+ default n
+ help
+ Say N here if you want to disable all vector related procedure
+ in the kernel.
+
+ If you don't know what to do here, say Y.
+
+endmenu
menu "Kernel features"
@@ -556,6 +566,7 @@ config CMDLINE_EXTEND
cases where the provided arguments are insufficient and
you don't want to or cannot modify them.
+
config CMDLINE_FORCE
bool "Always use the default kernel command string"
help
@@ -648,7 +659,7 @@ config XIP_PHYS_ADDR
be linked for and stored to. This address is dependent on your
own flash usage.
-endmenu # "Boot options"
+endmenu
config BUILTIN_DTB
bool
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
index 3fa8ef336822..1ec17f3d6d09 100644
--- a/arch/riscv/Makefile
+++ b/arch/riscv/Makefile
@@ -50,6 +50,7 @@ riscv-march-$(CONFIG_ARCH_RV32I) := rv32ima
riscv-march-$(CONFIG_ARCH_RV64I) := rv64ima
riscv-march-$(CONFIG_FPU) := $(riscv-march-y)fd
riscv-march-$(CONFIG_RISCV_ISA_C) := $(riscv-march-y)c
+riscv-march-$(CONFIG_VECTOR) := $(riscv-march-y)v
# Newer binutils versions default to ISA spec version 20191213 which moves some
# instructions from the I extension to the Zicsr and Zifencei extensions.
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Greentime Hu, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Vincent Chen, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Mark Brown, Will Deacon,
Heiko Carstens, Huacai Chen, Alexey Dobriyan, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Guo Ren <guoren@linux.alibaba.com>
This patch adds a new config option which could enable assembler's
vector feature.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/Kconfig | 15 +++++++++++++--
arch/riscv/Makefile | 1 +
2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index ed66c31e4655..e294d85bfb7d 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -432,7 +432,17 @@ config FPU
If you don't know what to do here, say Y.
-endmenu # "Platform type"
+config VECTOR
+ bool "VECTOR support"
+ depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
+ default n
+ help
+ Say N here if you want to disable all vector related procedure
+ in the kernel.
+
+ If you don't know what to do here, say Y.
+
+endmenu
menu "Kernel features"
@@ -556,6 +566,7 @@ config CMDLINE_EXTEND
cases where the provided arguments are insufficient and
you don't want to or cannot modify them.
+
config CMDLINE_FORCE
bool "Always use the default kernel command string"
help
@@ -648,7 +659,7 @@ config XIP_PHYS_ADDR
be linked for and stored to. This address is dependent on your
own flash usage.
-endmenu # "Boot options"
+endmenu
config BUILTIN_DTB
bool
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
index 3fa8ef336822..1ec17f3d6d09 100644
--- a/arch/riscv/Makefile
+++ b/arch/riscv/Makefile
@@ -50,6 +50,7 @@ riscv-march-$(CONFIG_ARCH_RV32I) := rv32ima
riscv-march-$(CONFIG_ARCH_RV64I) := rv64ima
riscv-march-$(CONFIG_FPU) := $(riscv-march-y)fd
riscv-march-$(CONFIG_RISCV_ISA_C) := $(riscv-march-y)c
+riscv-march-$(CONFIG_VECTOR) := $(riscv-march-y)v
# Newer binutils versions default to ISA spec version 20191213 which moves some
# instructions from the I extension to the Zicsr and Zifencei extensions.
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Greentime Hu, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Chris Stillson, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Vincent Chen, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Mark Brown, Will Deacon,
Heiko Carstens, Huacai Chen, Alexey Dobriyan, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Guo Ren <guoren@linux.alibaba.com>
This patch adds a new config option which could enable assembler's
vector feature.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/Kconfig | 15 +++++++++++++--
arch/riscv/Makefile | 1 +
2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index ed66c31e4655..e294d85bfb7d 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -432,7 +432,17 @@ config FPU
If you don't know what to do here, say Y.
-endmenu # "Platform type"
+config VECTOR
+ bool "VECTOR support"
+ depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
+ default n
+ help
+ Say N here if you want to disable all vector related procedure
+ in the kernel.
+
+ If you don't know what to do here, say Y.
+
+endmenu
menu "Kernel features"
@@ -556,6 +566,7 @@ config CMDLINE_EXTEND
cases where the provided arguments are insufficient and
you don't want to or cannot modify them.
+
config CMDLINE_FORCE
bool "Always use the default kernel command string"
help
@@ -648,7 +659,7 @@ config XIP_PHYS_ADDR
be linked for and stored to. This address is dependent on your
own flash usage.
-endmenu # "Boot options"
+endmenu
config BUILTIN_DTB
bool
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
index 3fa8ef336822..1ec17f3d6d09 100644
--- a/arch/riscv/Makefile
+++ b/arch/riscv/Makefile
@@ -50,6 +50,7 @@ riscv-march-$(CONFIG_ARCH_RV32I) := rv32ima
riscv-march-$(CONFIG_ARCH_RV64I) := rv64ima
riscv-march-$(CONFIG_FPU) := $(riscv-march-y)fd
riscv-march-$(CONFIG_RISCV_ISA_C) := $(riscv-march-y)c
+riscv-march-$(CONFIG_VECTOR) := $(riscv-march-y)v
# Newer binutils versions default to ISA spec version 20191213 which moves some
# instructions from the I extension to the Zicsr and Zifencei extensions.
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-07 17:21 ` Björn Töpel
-1 siblings, 0 replies; 147+ messages in thread
From: Björn Töpel @ 2022-11-07 17:21 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Guo Ren, linux-riscv, linux-kernel
[Cropping the huge Cc:-list.]
Chris Stillson <stillson@rivosinc.com> writes:
> From: Guo Ren <guoren@linux.alibaba.com>
>
> This patch adds a new config option which could enable assembler's
> vector feature.
>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/Kconfig | 15 +++++++++++++--
> arch/riscv/Makefile | 1 +
> 2 files changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index ed66c31e4655..e294d85bfb7d 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -432,7 +432,17 @@ config FPU
>
> If you don't know what to do here, say Y.
>
> -endmenu # "Platform type"
> +config VECTOR
> + bool "VECTOR support"
> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> + default n
> + help
> + Say N here if you want to disable all vector related procedure
> + in the kernel.
> +
> + If you don't know what to do here, say Y.
> +
> +endmenu
"VECTOR" is not really consistent to how the other configs are named;
RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
Björn
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-07 17:21 ` Björn Töpel
0 siblings, 0 replies; 147+ messages in thread
From: Björn Töpel @ 2022-11-07 17:21 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Guo Ren, linux-riscv, linux-kernel
[Cropping the huge Cc:-list.]
Chris Stillson <stillson@rivosinc.com> writes:
> From: Guo Ren <guoren@linux.alibaba.com>
>
> This patch adds a new config option which could enable assembler's
> vector feature.
>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/Kconfig | 15 +++++++++++++--
> arch/riscv/Makefile | 1 +
> 2 files changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index ed66c31e4655..e294d85bfb7d 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -432,7 +432,17 @@ config FPU
>
> If you don't know what to do here, say Y.
>
> -endmenu # "Platform type"
> +config VECTOR
> + bool "VECTOR support"
> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> + default n
> + help
> + Say N here if you want to disable all vector related procedure
> + in the kernel.
> +
> + If you don't know what to do here, say Y.
> +
> +endmenu
"VECTOR" is not really consistent to how the other configs are named;
RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
Björn
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-07 17:21 ` Björn Töpel
@ 2022-11-08 0:04 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-08 0:04 UTC (permalink / raw)
To: Björn Töpel, Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Guo Ren, linux-riscv, linux-kernel,
Conor Dooley, Andy Chiu
+CC Andy, Conor
On 11/7/22 09:21, Björn Töpel wrote:
>> +config VECTOR
>> + bool "VECTOR support"
>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>> + default n
>> + help
>> + Say N here if you want to disable all vector related procedure
>> + in the kernel.
>> +
>> + If you don't know what to do here, say Y.
>> +
>> +endmenu
> "VECTOR" is not really consistent to how the other configs are named;
> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
Good point, I've changed it to RISCV_ISA_V to keep it consistent with
existing RISCV_ISA_C.
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-08 0:04 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-08 0:04 UTC (permalink / raw)
To: Björn Töpel, Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Guo Ren, linux-riscv, linux-kernel,
Conor Dooley, Andy Chiu
+CC Andy, Conor
On 11/7/22 09:21, Björn Töpel wrote:
>> +config VECTOR
>> + bool "VECTOR support"
>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>> + default n
>> + help
>> + Say N here if you want to disable all vector related procedure
>> + in the kernel.
>> +
>> + If you don't know what to do here, say Y.
>> +
>> +endmenu
> "VECTOR" is not really consistent to how the other configs are named;
> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
Good point, I've changed it to RISCV_ISA_V to keep it consistent with
existing RISCV_ISA_C.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-08 0:04 ` Vineet Gupta
@ 2022-11-08 7:56 ` Conor Dooley
-1 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-11-08 7:56 UTC (permalink / raw)
To: Vineet Gupta
Cc: Björn Töpel, Chris Stillson, Greentime Hu, Guo Ren,
Vincent Chen, Paul Walmsley, Palmer Dabbelt, Albert Ou, Guo Ren,
linux-riscv, linux-kernel, Andy Chiu
On Mon, Nov 07, 2022 at 04:04:28PM -0800, Vineet Gupta wrote:
> +CC Andy, Conor
>
> On 11/7/22 09:21, Björn Töpel wrote:
> > > +config VECTOR
> > > + bool "VECTOR support"
> > > + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> > > + default n
> > > + help
> > > + Say N here if you want to disable all vector related procedure
> > > + in the kernel.
> > > +
> > > + If you don't know what to do here, say Y.
> > > +
> > > +endmenu
> > "VECTOR" is not really consistent to how the other configs are named;
> > RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
>
> Good point, I've changed it to RISCV_ISA_V to keep it consistent with
> existing RISCV_ISA_C.
Hey Vineet, kinda randomly replying here but the wording makes it look
like you're going to take this patchset on?
If so, please check out v10 (think it was from April) as there are some
comments on that version that IIRC remain un-resolved.
Thanks,
Conor.
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-08 7:56 ` Conor Dooley
0 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-11-08 7:56 UTC (permalink / raw)
To: Vineet Gupta
Cc: Björn Töpel, Chris Stillson, Greentime Hu, Guo Ren,
Vincent Chen, Paul Walmsley, Palmer Dabbelt, Albert Ou, Guo Ren,
linux-riscv, linux-kernel, Andy Chiu
On Mon, Nov 07, 2022 at 04:04:28PM -0800, Vineet Gupta wrote:
> +CC Andy, Conor
>
> On 11/7/22 09:21, Björn Töpel wrote:
> > > +config VECTOR
> > > + bool "VECTOR support"
> > > + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> > > + default n
> > > + help
> > > + Say N here if you want to disable all vector related procedure
> > > + in the kernel.
> > > +
> > > + If you don't know what to do here, say Y.
> > > +
> > > +endmenu
> > "VECTOR" is not really consistent to how the other configs are named;
> > RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
>
> Good point, I've changed it to RISCV_ISA_V to keep it consistent with
> existing RISCV_ISA_C.
Hey Vineet, kinda randomly replying here but the wording makes it look
like you're going to take this patchset on?
If so, please check out v10 (think it was from April) as there are some
comments on that version that IIRC remain un-resolved.
Thanks,
Conor.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-08 7:56 ` Conor Dooley
@ 2022-11-08 17:17 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-08 17:17 UTC (permalink / raw)
To: Conor Dooley, Vineet Gupta
Cc: Björn Töpel, Chris Stillson, Greentime Hu, Guo Ren,
Vincent Chen, Paul Walmsley, Palmer Dabbelt, Albert Ou, Guo Ren,
linux-riscv, linux-kernel, Andy Chiu
On 11/7/22 23:56, Conor Dooley wrote:
> On Mon, Nov 07, 2022 at 04:04:28PM -0800, Vineet Gupta wrote:
>> +CC Andy, Conor
>>
>> On 11/7/22 09:21, Björn Töpel wrote:
>>>> +config VECTOR
>>>> + bool "VECTOR support"
>>>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>>>> + default n
>>>> + help
>>>> + Say N here if you want to disable all vector related procedure
>>>> + in the kernel.
>>>> +
>>>> + If you don't know what to do here, say Y.
>>>> +
>>>> +endmenu
>>> "VECTOR" is not really consistent to how the other configs are named;
>>> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
>>
>> Good point, I've changed it to RISCV_ISA_V to keep it consistent with
>> existing RISCV_ISA_C.
>
> Hey Vineet, kinda randomly replying here but the wording makes it look
> like you're going to take this patchset on?
> If so, please check out v10 (think it was from April) as there are some
> comments on that version that IIRC remain un-resolved.
Sure thing. Although I only see a few from Christoph and kernel build bot.
-Vineet
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-08 17:17 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-08 17:17 UTC (permalink / raw)
To: Conor Dooley, Vineet Gupta
Cc: Björn Töpel, Chris Stillson, Greentime Hu, Guo Ren,
Vincent Chen, Paul Walmsley, Palmer Dabbelt, Albert Ou, Guo Ren,
linux-riscv, linux-kernel, Andy Chiu
On 11/7/22 23:56, Conor Dooley wrote:
> On Mon, Nov 07, 2022 at 04:04:28PM -0800, Vineet Gupta wrote:
>> +CC Andy, Conor
>>
>> On 11/7/22 09:21, Björn Töpel wrote:
>>>> +config VECTOR
>>>> + bool "VECTOR support"
>>>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>>>> + default n
>>>> + help
>>>> + Say N here if you want to disable all vector related procedure
>>>> + in the kernel.
>>>> +
>>>> + If you don't know what to do here, say Y.
>>>> +
>>>> +endmenu
>>> "VECTOR" is not really consistent to how the other configs are named;
>>> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
>>
>> Good point, I've changed it to RISCV_ISA_V to keep it consistent with
>> existing RISCV_ISA_C.
>
> Hey Vineet, kinda randomly replying here but the wording makes it look
> like you're going to take this patchset on?
> If so, please check out v10 (think it was from April) as there are some
> comments on that version that IIRC remain un-resolved.
Sure thing. Although I only see a few from Christoph and kernel build bot.
-Vineet
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-08 17:17 ` Vineet Gupta
@ 2022-11-08 17:22 ` Conor Dooley
-1 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-11-08 17:22 UTC (permalink / raw)
To: Vineet Gupta
Cc: Conor Dooley, Björn Töpel, Chris Stillson,
Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Guo Ren, linux-riscv, linux-kernel,
Andy Chiu
On Tue, Nov 08, 2022 at 09:17:26AM -0800, Vineet Gupta wrote:
> On 11/7/22 23:56, Conor Dooley wrote:
> > On Mon, Nov 07, 2022 at 04:04:28PM -0800, Vineet Gupta wrote:
> > > +CC Andy, Conor
> > >
> > > On 11/7/22 09:21, Björn Töpel wrote:
> > > > > +config VECTOR
> > > > > + bool "VECTOR support"
> > > > > + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> > > > > + default n
> > > > > + help
> > > > > + Say N here if you want to disable all vector related procedure
> > > > > + in the kernel.
> > > > > +
> > > > > + If you don't know what to do here, say Y.
> > > > > +
> > > > > +endmenu
> > > > "VECTOR" is not really consistent to how the other configs are named;
> > > > RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
> > >
> > > Good point, I've changed it to RISCV_ISA_V to keep it consistent with
> > > existing RISCV_ISA_C.
> >
> > Hey Vineet, kinda randomly replying here but the wording makes it look
> > like you're going to take this patchset on?
> > If so, please check out v10 (think it was from April) as there are some
> > comments on that version that IIRC remain un-resolved.
>
> Sure thing. Although I only see a few from Christoph and kernel build bot.
Yup, they were minor - I just don't wanna see them get lost :)
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-08 17:22 ` Conor Dooley
0 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-11-08 17:22 UTC (permalink / raw)
To: Vineet Gupta
Cc: Conor Dooley, Björn Töpel, Chris Stillson,
Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Guo Ren, linux-riscv, linux-kernel,
Andy Chiu
On Tue, Nov 08, 2022 at 09:17:26AM -0800, Vineet Gupta wrote:
> On 11/7/22 23:56, Conor Dooley wrote:
> > On Mon, Nov 07, 2022 at 04:04:28PM -0800, Vineet Gupta wrote:
> > > +CC Andy, Conor
> > >
> > > On 11/7/22 09:21, Björn Töpel wrote:
> > > > > +config VECTOR
> > > > > + bool "VECTOR support"
> > > > > + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> > > > > + default n
> > > > > + help
> > > > > + Say N here if you want to disable all vector related procedure
> > > > > + in the kernel.
> > > > > +
> > > > > + If you don't know what to do here, say Y.
> > > > > +
> > > > > +endmenu
> > > > "VECTOR" is not really consistent to how the other configs are named;
> > > > RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
> > >
> > > Good point, I've changed it to RISCV_ISA_V to keep it consistent with
> > > existing RISCV_ISA_C.
> >
> > Hey Vineet, kinda randomly replying here but the wording makes it look
> > like you're going to take this patchset on?
> > If so, please check out v10 (think it was from April) as there are some
> > comments on that version that IIRC remain un-resolved.
>
> Sure thing. Although I only see a few from Christoph and kernel build bot.
Yup, they were minor - I just don't wanna see them get lost :)
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-07 17:21 ` Björn Töpel
@ 2022-11-13 16:16 ` Conor.Dooley
-1 siblings, 0 replies; 147+ messages in thread
From: Conor.Dooley @ 2022-11-13 16:16 UTC (permalink / raw)
To: bjorn, stillson
Cc: greentime.hu, guoren, vincent.chen, paul.walmsley, palmer, aou,
guoren, linux-riscv, linux-kernel
On 07/11/2022 17:21, Björn Töpel wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>
> [Cropping the huge Cc:-list.]
>
> Chris Stillson <stillson@rivosinc.com> writes:
>
>> From: Guo Ren <guoren@linux.alibaba.com>
>>
>> This patch adds a new config option which could enable assembler's
>> vector feature.
>>
>> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
>> Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
>> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
>> ---
>> arch/riscv/Kconfig | 15 +++++++++++++--
>> arch/riscv/Makefile | 1 +
>> 2 files changed, 14 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
>> index ed66c31e4655..e294d85bfb7d 100644
>> --- a/arch/riscv/Kconfig
>> +++ b/arch/riscv/Kconfig
>> @@ -432,7 +432,17 @@ config FPU
>>
>> If you don't know what to do here, say Y.
>>
>> -endmenu # "Platform type"
>> +config VECTOR
>> + bool "VECTOR support"
>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>> + default n
>> + help
>> + Say N here if you want to disable all vector related procedure
>> + in the kernel.
>> +
>> + If you don't know what to do here, say Y.
>> +
>> +endmenu
>
> "VECTOR" is not really consistent to how the other configs are named;
> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
It'd be RISCV_ISA_V to match the others single letter extentions, right?
The toolchain dependency check here also seems rather naive.
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-13 16:16 ` Conor.Dooley
0 siblings, 0 replies; 147+ messages in thread
From: Conor.Dooley @ 2022-11-13 16:16 UTC (permalink / raw)
To: bjorn, stillson
Cc: greentime.hu, guoren, vincent.chen, paul.walmsley, palmer, aou,
guoren, linux-riscv, linux-kernel
On 07/11/2022 17:21, Björn Töpel wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>
> [Cropping the huge Cc:-list.]
>
> Chris Stillson <stillson@rivosinc.com> writes:
>
>> From: Guo Ren <guoren@linux.alibaba.com>
>>
>> This patch adds a new config option which could enable assembler's
>> vector feature.
>>
>> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
>> Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
>> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
>> ---
>> arch/riscv/Kconfig | 15 +++++++++++++--
>> arch/riscv/Makefile | 1 +
>> 2 files changed, 14 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
>> index ed66c31e4655..e294d85bfb7d 100644
>> --- a/arch/riscv/Kconfig
>> +++ b/arch/riscv/Kconfig
>> @@ -432,7 +432,17 @@ config FPU
>>
>> If you don't know what to do here, say Y.
>>
>> -endmenu # "Platform type"
>> +config VECTOR
>> + bool "VECTOR support"
>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>> + default n
>> + help
>> + Say N here if you want to disable all vector related procedure
>> + in the kernel.
>> +
>> + If you don't know what to do here, say Y.
>> +
>> +endmenu
>
> "VECTOR" is not really consistent to how the other configs are named;
> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
It'd be RISCV_ISA_V to match the others single letter extentions, right?
The toolchain dependency check here also seems rather naive.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-13 16:16 ` Conor.Dooley
@ 2022-11-15 17:38 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-15 17:38 UTC (permalink / raw)
To: Conor.Dooley, bjorn, stillson
Cc: greentime.hu, guoren, vincent.chen, paul.walmsley, palmer, aou,
guoren, linux-riscv, linux-kernel
On 11/13/22 08:16, Conor.Dooley@microchip.com wrote:
>>> +config VECTOR
>>> + bool "VECTOR support"
>>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>>> + default n
>>> + help
>>> + Say N here if you want to disable all vector related procedure
>>> + in the kernel.
>>> +
>>> + If you don't know what to do here, say Y.
>>> +
>>> +endmenu
>> "VECTOR" is not really consistent to how the other configs are named;
>> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
> It'd be RISCV_ISA_V to match the others single letter extentions, right?
Yep.
> The toolchain dependency check here also seems rather naive.
Indeed. I can build the code just fine with gcc-11 (and gcc-12),
although my reworked patcheset doesn't include all the orig patches
including the in-kernel xor stuff.
-Vineet
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-15 17:38 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-15 17:38 UTC (permalink / raw)
To: Conor.Dooley, bjorn, stillson
Cc: greentime.hu, guoren, vincent.chen, paul.walmsley, palmer, aou,
guoren, linux-riscv, linux-kernel
On 11/13/22 08:16, Conor.Dooley@microchip.com wrote:
>>> +config VECTOR
>>> + bool "VECTOR support"
>>> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
>>> + default n
>>> + help
>>> + Say N here if you want to disable all vector related procedure
>>> + in the kernel.
>>> +
>>> + If you don't know what to do here, say Y.
>>> +
>>> +endmenu
>> "VECTOR" is not really consistent to how the other configs are named;
>> RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
> It'd be RISCV_ISA_V to match the others single letter extentions, right?
Yep.
> The toolchain dependency check here also seems rather naive.
Indeed. I can build the code just fine with gcc-11 (and gcc-12),
although my reworked patcheset doesn't include all the orig patches
including the in-kernel xor stuff.
-Vineet
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-11-15 17:38 ` Vineet Gupta
@ 2022-11-15 22:17 ` Conor Dooley
-1 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-11-15 22:17 UTC (permalink / raw)
To: Vineet Gupta
Cc: Conor.Dooley, bjorn, stillson, greentime.hu, guoren,
vincent.chen, paul.walmsley, palmer, aou, guoren, linux-riscv,
linux-kernel, ajones
On Tue, Nov 15, 2022 at 09:38:53AM -0800, Vineet Gupta wrote:
> On 11/13/22 08:16, Conor.Dooley@microchip.com wrote:
> > > > +config VECTOR
> > > > + bool "VECTOR support"
> > > > + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> > > > + default n
> > > > + help
> > > > + Say N here if you want to disable all vector related procedure
> > > > + in the kernel.
> > > > +
> > > > + If you don't know what to do here, say Y.
> > > > +
> > > > +endmenu
> > > "VECTOR" is not really consistent to how the other configs are named;
> > > RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
> > It'd be RISCV_ISA_V to match the others single letter extentions, right?
>
> Yep.
>
> > The toolchain dependency check here also seems rather naive.
>
> Indeed. I can build the code just fine with gcc-11 (and gcc-12), although my
> reworked patcheset doesn't include all the orig patches including the
> in-kernel xor stuff.
By naive here I meant that checking cc alone is probably not a
sufficient check for whether the toolchain supports the extension.
What about the assembler etc?
With Zicbom and Zihintpause we ran into problems with mixed usage, eg
binutils 2.35 + gcc 12. In his Zicboz series Drew has gone with insn
definitions - but while that's okay for something small like Zicboz,
do we want to do that for something with as many instructions as vector?
The alternative is cc-option, but that feels a lot less clean than what
Drew cooked up here:
https://lore.kernel.org/linux-riscv/20221027130247.31634-1-ajones@ventanamicro.com/
I've not checked this because I am lazy, but I am also assuming that
whoever put clang-13 in there picked it such that it doesn't require
experimental extensions flags. Mostly just writing this to remind myself
to check it at some point.
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-11-15 22:17 ` Conor Dooley
0 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-11-15 22:17 UTC (permalink / raw)
To: Vineet Gupta
Cc: Conor.Dooley, bjorn, stillson, greentime.hu, guoren,
vincent.chen, paul.walmsley, palmer, aou, guoren, linux-riscv,
linux-kernel, ajones
On Tue, Nov 15, 2022 at 09:38:53AM -0800, Vineet Gupta wrote:
> On 11/13/22 08:16, Conor.Dooley@microchip.com wrote:
> > > > +config VECTOR
> > > > + bool "VECTOR support"
> > > > + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> > > > + default n
> > > > + help
> > > > + Say N here if you want to disable all vector related procedure
> > > > + in the kernel.
> > > > +
> > > > + If you don't know what to do here, say Y.
> > > > +
> > > > +endmenu
> > > "VECTOR" is not really consistent to how the other configs are named;
> > > RISCV_ISA_V, RISCV_ISA_VECTOR, RISCV_VECTOR?
> > It'd be RISCV_ISA_V to match the others single letter extentions, right?
>
> Yep.
>
> > The toolchain dependency check here also seems rather naive.
>
> Indeed. I can build the code just fine with gcc-11 (and gcc-12), although my
> reworked patcheset doesn't include all the orig patches including the
> in-kernel xor stuff.
By naive here I meant that checking cc alone is probably not a
sufficient check for whether the toolchain supports the extension.
What about the assembler etc?
With Zicbom and Zihintpause we ran into problems with mixed usage, eg
binutils 2.35 + gcc 12. In his Zicboz series Drew has gone with insn
definitions - but while that's okay for something small like Zicboz,
do we want to do that for something with as many instructions as vector?
The alternative is cc-option, but that feels a lot less clean than what
Drew cooked up here:
https://lore.kernel.org/linux-riscv/20221027130247.31634-1-ajones@ventanamicro.com/
I've not checked this because I am lazy, but I am also assuming that
whoever put clang-13 in there picked it such that it doesn't require
experimental extensions flags. Mostly just writing this to remind myself
to check it at some point.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
2022-09-21 21:43 ` Chris Stillson
@ 2022-12-15 0:40 ` Atish Patra
-1 siblings, 0 replies; 147+ messages in thread
From: Atish Patra @ 2022-12-15 0:40 UTC (permalink / raw)
To: Chris Stillson
Cc: Guo Ren, Greentime Hu, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Oleg Nesterov, Guo Ren,
Heinrich Schuchardt, Arnaud Pouliquen, Paolo Bonzini,
Qinglin Pan, Alexandre Ghiti, Vincent Chen, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Myrtle Shah, Vitaly Wool,
Mathieu Desnoyers, Catalin Marinas, Mark Brown, Will Deacon,
Heiko Carstens, Huacai Chen, Alexey Dobriyan, Janosch Frank,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
On Wed, Sep 21, 2022 at 2:47 PM Chris Stillson <stillson@rivosinc.com> wrote:
>
> From: Guo Ren <guoren@linux.alibaba.com>
>
> This patch adds a new config option which could enable assembler's
> vector feature.
>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/Kconfig | 15 +++++++++++++--
> arch/riscv/Makefile | 1 +
> 2 files changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index ed66c31e4655..e294d85bfb7d 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -432,7 +432,17 @@ config FPU
>
> If you don't know what to do here, say Y.
>
> -endmenu # "Platform type"
> +config VECTOR
> + bool "VECTOR support"
> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> + default n
> + help
> + Say N here if you want to disable all vector related procedure
> + in the kernel.
> +
> + If you don't know what to do here, say Y.
> +
> +endmenu
>
> menu "Kernel features"
>
> @@ -556,6 +566,7 @@ config CMDLINE_EXTEND
> cases where the provided arguments are insufficient and
> you don't want to or cannot modify them.
>
> +
> config CMDLINE_FORCE
> bool "Always use the default kernel command string"
> help
> @@ -648,7 +659,7 @@ config XIP_PHYS_ADDR
> be linked for and stored to. This address is dependent on your
> own flash usage.
>
> -endmenu # "Boot options"
> +endmenu
>
> config BUILTIN_DTB
> bool
> diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
> index 3fa8ef336822..1ec17f3d6d09 100644
> --- a/arch/riscv/Makefile
> +++ b/arch/riscv/Makefile
> @@ -50,6 +50,7 @@ riscv-march-$(CONFIG_ARCH_RV32I) := rv32ima
> riscv-march-$(CONFIG_ARCH_RV64I) := rv64ima
> riscv-march-$(CONFIG_FPU) := $(riscv-march-y)fd
> riscv-march-$(CONFIG_RISCV_ISA_C) := $(riscv-march-y)c
> +riscv-march-$(CONFIG_VECTOR) := $(riscv-march-y)v
>
> # Newer binutils versions default to ISA spec version 20191213 which moves some
> # instructions from the I extension to the Zicsr and Zifencei extensions.
> --
> 2.25.1
>
Kernel boot hangs if compiled LLVM and vector enabled. Because LLVM
enables auto vectorization by default and it inserts
random vector instructions.
We need to add "-mno-implicit-float" for llvm builds to disable auto
vectorization. Thanks Vineet and Saleem for the hint :).
--
Regards,
Atish
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 04/17] riscv: Add vector feature to compile
@ 2022-12-15 0:40 ` Atish Patra
0 siblings, 0 replies; 147+ messages in thread
From: Atish Patra @ 2022-12-15 0:40 UTC (permalink / raw)
To: Chris Stillson
Cc: Mark Rutland, Guo Ren, Heiko Stuebner, kvm,
Peter Zijlstra (Intel),
Catalin Marinas, linux-kernel, Heiko Carstens, linux-mm, Guo Ren,
Jisheng Zhang, Qinglin Pan, linux-riscv, Will Deacon,
Ard Biesheuvel, Vitaly Wool, Davidlohr Bueso, Yury Norov,
Changbin Du, Anup Patel, Huacai Chen, Myrtle Shah,
Eugene Syromiatnikov, Tsukasa OI, Mathieu Desnoyers,
Greentime Hu, Arnaud Pouliquen, Li Zhengyu, Evgenii Stepanov,
Heinrich Schuchardt, Barret Rhoden, Albert Ou, Kees Cook,
Arnd Bergmann, Frederic Weisbecker, Suren Baghdasaryan, Dao Lu,
Alexander Graf, Paul Walmsley, Alexandre Ghiti,
Peter Collingbourne, Janosch Frank, Christian Brauner,
Oleg Nesterov, Colin Cross, Alexey Dobriyan, Han-Kuan Chen,
Vincent Chen, Mark Brown, Palmer Dabbelt, Eric Biederman,
kvm-riscv, Paolo Bonzini, Andrew Morton
On Wed, Sep 21, 2022 at 2:47 PM Chris Stillson <stillson@rivosinc.com> wrote:
>
> From: Guo Ren <guoren@linux.alibaba.com>
>
> This patch adds a new config option which could enable assembler's
> vector feature.
>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/Kconfig | 15 +++++++++++++--
> arch/riscv/Makefile | 1 +
> 2 files changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index ed66c31e4655..e294d85bfb7d 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -432,7 +432,17 @@ config FPU
>
> If you don't know what to do here, say Y.
>
> -endmenu # "Platform type"
> +config VECTOR
> + bool "VECTOR support"
> + depends on GCC_VERSION >= 120000 || CLANG_VERSION >= 130000
> + default n
> + help
> + Say N here if you want to disable all vector related procedure
> + in the kernel.
> +
> + If you don't know what to do here, say Y.
> +
> +endmenu
>
> menu "Kernel features"
>
> @@ -556,6 +566,7 @@ config CMDLINE_EXTEND
> cases where the provided arguments are insufficient and
> you don't want to or cannot modify them.
>
> +
> config CMDLINE_FORCE
> bool "Always use the default kernel command string"
> help
> @@ -648,7 +659,7 @@ config XIP_PHYS_ADDR
> be linked for and stored to. This address is dependent on your
> own flash usage.
>
> -endmenu # "Boot options"
> +endmenu
>
> config BUILTIN_DTB
> bool
> diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
> index 3fa8ef336822..1ec17f3d6d09 100644
> --- a/arch/riscv/Makefile
> +++ b/arch/riscv/Makefile
> @@ -50,6 +50,7 @@ riscv-march-$(CONFIG_ARCH_RV32I) := rv32ima
> riscv-march-$(CONFIG_ARCH_RV64I) := rv64ima
> riscv-march-$(CONFIG_FPU) := $(riscv-march-y)fd
> riscv-march-$(CONFIG_RISCV_ISA_C) := $(riscv-march-y)c
> +riscv-march-$(CONFIG_VECTOR) := $(riscv-march-y)v
>
> # Newer binutils versions default to ISA spec version 20191213 which moves some
> # instructions from the I extension to the Zicsr and Zifencei extensions.
> --
> 2.25.1
>
Kernel boot hangs if compiled LLVM and vector enabled. Because LLVM
enables auto vectorization by default and it inserts
random vector instructions.
We need to add "-mno-implicit-float" for llvm builds to disable auto
vectorization. Thanks Vineet and Saleem for the hint :).
--
Regards,
Atish
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Chris Stillson, Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L,
Ruinland Tsai, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Paul E. McKenney, Frederic Weisbecker,
Mark Rutland, Vitaly Wool, Myrtle Shah, Catalin Marinas,
Will Deacon, Mark Brown, WANG Xuerui, Huacai Chen,
Alexey Dobriyan, Christian Brauner, Suren Baghdasaryan,
Peter Collingbourne, Colin Cross, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch is used to detect vector support status of CPU and use
riscv_vsize to save the size of all the vector registers. It assumes
all harts has the same capabilities in SMP system.
[guoren@linux.alibaba.com: add has_vector checking]
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/vector.h | 14 +++++
arch/riscv/kernel/cpufeature.c | 19 +++++++
arch/riscv/kernel/riscv_ksyms.c | 6 +++
arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
4 files changed, 132 insertions(+)
create mode 100644 arch/riscv/include/asm/vector.h
create mode 100644 arch/riscv/kernel/vector.S
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
new file mode 100644
index 000000000000..16304b0c6a6f
--- /dev/null
+++ b/arch/riscv/include/asm/vector.h
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#ifndef __ASM_RISCV_VECTOR_H
+#define __ASM_RISCV_VECTOR_H
+
+#include <linux/types.h>
+
+void rvv_enable(void);
+void rvv_disable(void);
+
+#endif /* ! __ASM_RISCV_VECTOR_H */
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 8d4448c2d4f4..0487ab19b234 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -30,6 +30,14 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
__ro_after_init DEFINE_STATIC_KEY_ARRAY_FALSE(riscv_isa_ext_keys, RISCV_ISA_EXT_KEY_MAX);
EXPORT_SYMBOL(riscv_isa_ext_keys);
+#ifdef CONFIG_FPU
+__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
+#endif
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
+unsigned long riscv_vsize __read_mostly;
+#endif
/**
* riscv_isa_extension_base() - Get base extension word
@@ -249,6 +257,16 @@ void __init riscv_fill_hwcap(void)
if (j >= 0)
static_branch_enable(&riscv_isa_ext_keys[j]);
}
+
+#ifdef CONFIG_VECTOR
+ if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
+ static_branch_enable(&cpu_hwcap_vector);
+ /* There are 32 vector registers with vlenb length. */
+ rvv_enable();
+ riscv_vsize = csr_read(CSR_VLENB) * 32;
+ rvv_disable();
+ }
+#endif
}
#ifdef CONFIG_RISCV_ALTERNATIVE
@@ -328,3 +346,4 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
}
}
#endif
+}
diff --git a/arch/riscv/kernel/riscv_ksyms.c b/arch/riscv/kernel/riscv_ksyms.c
index 5ab1c7e1a6ed..3489d2a20ca3 100644
--- a/arch/riscv/kernel/riscv_ksyms.c
+++ b/arch/riscv/kernel/riscv_ksyms.c
@@ -15,3 +15,9 @@ EXPORT_SYMBOL(memmove);
EXPORT_SYMBOL(__memset);
EXPORT_SYMBOL(__memcpy);
EXPORT_SYMBOL(__memmove);
+
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+EXPORT_SYMBOL(rvv_enable);
+EXPORT_SYMBOL(rvv_disable);
+#endif
diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
new file mode 100644
index 000000000000..9f7dc70c4443
--- /dev/null
+++ b/arch/riscv/kernel/vector.S
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2012 Regents of the University of California
+ * Copyright (C) 2017 SiFive
+ * Copyright (C) 2019 Alibaba Group Holding Limited
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation, version 2.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/linkage.h>
+
+#include <asm/asm.h>
+#include <asm/csr.h>
+#include <asm/asm-offsets.h>
+
+#define vstatep a0
+#define datap a1
+#define x_vstart t0
+#define x_vtype t1
+#define x_vl t2
+#define x_vcsr t3
+#define incr t4
+#define status t5
+
+ENTRY(__vstate_save)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ csrr x_vstart, CSR_VSTART
+ csrr x_vtype, CSR_VTYPE
+ csrr x_vl, CSR_VL
+ csrr x_vcsr, CSR_VCSR
+ vsetvli incr, x0, e8, m8, ta, ma
+ vse8.v v0, (datap)
+ add datap, datap, incr
+ vse8.v v8, (datap)
+ add datap, datap, incr
+ vse8.v v16, (datap)
+ add datap, datap, incr
+ vse8.v v24, (datap)
+
+ REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
+ REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
+ REG_S x_vl, RISCV_V_STATE_VL(vstatep)
+ REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__vstate_save)
+
+ENTRY(__vstate_restore)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ vsetvli incr, x0, e8, m8, ta, ma
+ vle8.v v0, (datap)
+ add datap, datap, incr
+ vle8.v v8, (datap)
+ add datap, datap, incr
+ vle8.v v16, (datap)
+ add datap, datap, incr
+ vle8.v v24, (datap)
+
+ REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
+ REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
+ REG_L x_vl, RISCV_V_STATE_VL(vstatep)
+ REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
+ vsetvl x0, x_vl, x_vtype
+ csrw CSR_VSTART, x_vstart
+ csrw CSR_VCSR, x_vcsr
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__vstate_restore)
+
+ENTRY(rvv_enable)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+ ret
+ENDPROC(rvv_enable)
+
+ENTRY(rvv_disable)
+ li status, SR_VS
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(rvv_disable)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Chris Stillson, Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L,
Ruinland Tsai, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Paul E. McKenney, Frederic Weisbecker,
Mark Rutland, Vitaly Wool, Myrtle Shah, Catalin Marinas,
Will Deacon, Mark Brown, WANG Xuerui, Huacai Chen,
Alexey Dobriyan, Christian Brauner, Suren Baghdasaryan,
Peter Collingbourne, Colin Cross, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch is used to detect vector support status of CPU and use
riscv_vsize to save the size of all the vector registers. It assumes
all harts has the same capabilities in SMP system.
[guoren@linux.alibaba.com: add has_vector checking]
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/vector.h | 14 +++++
arch/riscv/kernel/cpufeature.c | 19 +++++++
arch/riscv/kernel/riscv_ksyms.c | 6 +++
arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
4 files changed, 132 insertions(+)
create mode 100644 arch/riscv/include/asm/vector.h
create mode 100644 arch/riscv/kernel/vector.S
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
new file mode 100644
index 000000000000..16304b0c6a6f
--- /dev/null
+++ b/arch/riscv/include/asm/vector.h
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#ifndef __ASM_RISCV_VECTOR_H
+#define __ASM_RISCV_VECTOR_H
+
+#include <linux/types.h>
+
+void rvv_enable(void);
+void rvv_disable(void);
+
+#endif /* ! __ASM_RISCV_VECTOR_H */
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 8d4448c2d4f4..0487ab19b234 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -30,6 +30,14 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
__ro_after_init DEFINE_STATIC_KEY_ARRAY_FALSE(riscv_isa_ext_keys, RISCV_ISA_EXT_KEY_MAX);
EXPORT_SYMBOL(riscv_isa_ext_keys);
+#ifdef CONFIG_FPU
+__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
+#endif
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
+unsigned long riscv_vsize __read_mostly;
+#endif
/**
* riscv_isa_extension_base() - Get base extension word
@@ -249,6 +257,16 @@ void __init riscv_fill_hwcap(void)
if (j >= 0)
static_branch_enable(&riscv_isa_ext_keys[j]);
}
+
+#ifdef CONFIG_VECTOR
+ if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
+ static_branch_enable(&cpu_hwcap_vector);
+ /* There are 32 vector registers with vlenb length. */
+ rvv_enable();
+ riscv_vsize = csr_read(CSR_VLENB) * 32;
+ rvv_disable();
+ }
+#endif
}
#ifdef CONFIG_RISCV_ALTERNATIVE
@@ -328,3 +346,4 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
}
}
#endif
+}
diff --git a/arch/riscv/kernel/riscv_ksyms.c b/arch/riscv/kernel/riscv_ksyms.c
index 5ab1c7e1a6ed..3489d2a20ca3 100644
--- a/arch/riscv/kernel/riscv_ksyms.c
+++ b/arch/riscv/kernel/riscv_ksyms.c
@@ -15,3 +15,9 @@ EXPORT_SYMBOL(memmove);
EXPORT_SYMBOL(__memset);
EXPORT_SYMBOL(__memcpy);
EXPORT_SYMBOL(__memmove);
+
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+EXPORT_SYMBOL(rvv_enable);
+EXPORT_SYMBOL(rvv_disable);
+#endif
diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
new file mode 100644
index 000000000000..9f7dc70c4443
--- /dev/null
+++ b/arch/riscv/kernel/vector.S
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2012 Regents of the University of California
+ * Copyright (C) 2017 SiFive
+ * Copyright (C) 2019 Alibaba Group Holding Limited
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation, version 2.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/linkage.h>
+
+#include <asm/asm.h>
+#include <asm/csr.h>
+#include <asm/asm-offsets.h>
+
+#define vstatep a0
+#define datap a1
+#define x_vstart t0
+#define x_vtype t1
+#define x_vl t2
+#define x_vcsr t3
+#define incr t4
+#define status t5
+
+ENTRY(__vstate_save)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ csrr x_vstart, CSR_VSTART
+ csrr x_vtype, CSR_VTYPE
+ csrr x_vl, CSR_VL
+ csrr x_vcsr, CSR_VCSR
+ vsetvli incr, x0, e8, m8, ta, ma
+ vse8.v v0, (datap)
+ add datap, datap, incr
+ vse8.v v8, (datap)
+ add datap, datap, incr
+ vse8.v v16, (datap)
+ add datap, datap, incr
+ vse8.v v24, (datap)
+
+ REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
+ REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
+ REG_S x_vl, RISCV_V_STATE_VL(vstatep)
+ REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__vstate_save)
+
+ENTRY(__vstate_restore)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ vsetvli incr, x0, e8, m8, ta, ma
+ vle8.v v0, (datap)
+ add datap, datap, incr
+ vle8.v v8, (datap)
+ add datap, datap, incr
+ vle8.v v16, (datap)
+ add datap, datap, incr
+ vle8.v v24, (datap)
+
+ REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
+ REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
+ REG_L x_vl, RISCV_V_STATE_VL(vstatep)
+ REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
+ vsetvl x0, x_vl, x_vtype
+ csrw CSR_VSTART, x_vstart
+ csrw CSR_VCSR, x_vcsr
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__vstate_restore)
+
+ENTRY(rvv_enable)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+ ret
+ENDPROC(rvv_enable)
+
+ENTRY(rvv_disable)
+ li status, SR_VS
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(rvv_disable)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Chris Stillson, Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L,
Ruinland Tsai, Han-Kuan Chen, Changbin Du, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Paul E. McKenney, Frederic Weisbecker,
Mark Rutland, Vitaly Wool, Myrtle Shah, Catalin Marinas,
Will Deacon, Mark Brown, WANG Xuerui, Huacai Chen,
Alexey Dobriyan, Christian Brauner, Suren Baghdasaryan,
Peter Collingbourne, Colin Cross, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch is used to detect vector support status of CPU and use
riscv_vsize to save the size of all the vector registers. It assumes
all harts has the same capabilities in SMP system.
[guoren@linux.alibaba.com: add has_vector checking]
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/vector.h | 14 +++++
arch/riscv/kernel/cpufeature.c | 19 +++++++
arch/riscv/kernel/riscv_ksyms.c | 6 +++
arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
4 files changed, 132 insertions(+)
create mode 100644 arch/riscv/include/asm/vector.h
create mode 100644 arch/riscv/kernel/vector.S
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
new file mode 100644
index 000000000000..16304b0c6a6f
--- /dev/null
+++ b/arch/riscv/include/asm/vector.h
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2020 SiFive
+ */
+
+#ifndef __ASM_RISCV_VECTOR_H
+#define __ASM_RISCV_VECTOR_H
+
+#include <linux/types.h>
+
+void rvv_enable(void);
+void rvv_disable(void);
+
+#endif /* ! __ASM_RISCV_VECTOR_H */
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 8d4448c2d4f4..0487ab19b234 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -30,6 +30,14 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
__ro_after_init DEFINE_STATIC_KEY_ARRAY_FALSE(riscv_isa_ext_keys, RISCV_ISA_EXT_KEY_MAX);
EXPORT_SYMBOL(riscv_isa_ext_keys);
+#ifdef CONFIG_FPU
+__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
+#endif
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
+unsigned long riscv_vsize __read_mostly;
+#endif
/**
* riscv_isa_extension_base() - Get base extension word
@@ -249,6 +257,16 @@ void __init riscv_fill_hwcap(void)
if (j >= 0)
static_branch_enable(&riscv_isa_ext_keys[j]);
}
+
+#ifdef CONFIG_VECTOR
+ if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
+ static_branch_enable(&cpu_hwcap_vector);
+ /* There are 32 vector registers with vlenb length. */
+ rvv_enable();
+ riscv_vsize = csr_read(CSR_VLENB) * 32;
+ rvv_disable();
+ }
+#endif
}
#ifdef CONFIG_RISCV_ALTERNATIVE
@@ -328,3 +346,4 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
}
}
#endif
+}
diff --git a/arch/riscv/kernel/riscv_ksyms.c b/arch/riscv/kernel/riscv_ksyms.c
index 5ab1c7e1a6ed..3489d2a20ca3 100644
--- a/arch/riscv/kernel/riscv_ksyms.c
+++ b/arch/riscv/kernel/riscv_ksyms.c
@@ -15,3 +15,9 @@ EXPORT_SYMBOL(memmove);
EXPORT_SYMBOL(__memset);
EXPORT_SYMBOL(__memcpy);
EXPORT_SYMBOL(__memmove);
+
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+EXPORT_SYMBOL(rvv_enable);
+EXPORT_SYMBOL(rvv_disable);
+#endif
diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
new file mode 100644
index 000000000000..9f7dc70c4443
--- /dev/null
+++ b/arch/riscv/kernel/vector.S
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2012 Regents of the University of California
+ * Copyright (C) 2017 SiFive
+ * Copyright (C) 2019 Alibaba Group Holding Limited
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation, version 2.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/linkage.h>
+
+#include <asm/asm.h>
+#include <asm/csr.h>
+#include <asm/asm-offsets.h>
+
+#define vstatep a0
+#define datap a1
+#define x_vstart t0
+#define x_vtype t1
+#define x_vl t2
+#define x_vcsr t3
+#define incr t4
+#define status t5
+
+ENTRY(__vstate_save)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ csrr x_vstart, CSR_VSTART
+ csrr x_vtype, CSR_VTYPE
+ csrr x_vl, CSR_VL
+ csrr x_vcsr, CSR_VCSR
+ vsetvli incr, x0, e8, m8, ta, ma
+ vse8.v v0, (datap)
+ add datap, datap, incr
+ vse8.v v8, (datap)
+ add datap, datap, incr
+ vse8.v v16, (datap)
+ add datap, datap, incr
+ vse8.v v24, (datap)
+
+ REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
+ REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
+ REG_S x_vl, RISCV_V_STATE_VL(vstatep)
+ REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__vstate_save)
+
+ENTRY(__vstate_restore)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ vsetvli incr, x0, e8, m8, ta, ma
+ vle8.v v0, (datap)
+ add datap, datap, incr
+ vle8.v v8, (datap)
+ add datap, datap, incr
+ vle8.v v16, (datap)
+ add datap, datap, incr
+ vle8.v v24, (datap)
+
+ REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
+ REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
+ REG_L x_vl, RISCV_V_STATE_VL(vstatep)
+ REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
+ vsetvl x0, x_vl, x_vtype
+ csrw CSR_VSTART, x_vstart
+ csrw CSR_VCSR, x_vcsr
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__vstate_restore)
+
+ENTRY(rvv_enable)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+ ret
+ENDPROC(rvv_enable)
+
+ENTRY(rvv_disable)
+ li status, SR_VS
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(rvv_disable)
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
2022-09-21 21:43 ` Chris Stillson
@ 2022-09-22 4:23 ` Samuel Holland
-1 siblings, 0 replies; 147+ messages in thread
From: Samuel Holland @ 2022-09-22 4:23 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Ruinland Tsai,
Han-Kuan Chen, linux-riscv, linux-kernel
On 9/21/22 16:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch is used to detect vector support status of CPU and use
> riscv_vsize to save the size of all the vector registers. It assumes
> all harts has the same capabilities in SMP system.
>
> [guoren@linux.alibaba.com: add has_vector checking]
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/asm/vector.h | 14 +++++
> arch/riscv/kernel/cpufeature.c | 19 +++++++
> arch/riscv/kernel/riscv_ksyms.c | 6 +++
> arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
This file is not added to the Makefile until patch 8.
> 4 files changed, 132 insertions(+)
> create mode 100644 arch/riscv/include/asm/vector.h
> create mode 100644 arch/riscv/kernel/vector.S
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-09-22 4:23 ` Samuel Holland
0 siblings, 0 replies; 147+ messages in thread
From: Samuel Holland @ 2022-09-22 4:23 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Ruinland Tsai,
Han-Kuan Chen, linux-riscv, linux-kernel
On 9/21/22 16:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch is used to detect vector support status of CPU and use
> riscv_vsize to save the size of all the vector registers. It assumes
> all harts has the same capabilities in SMP system.
>
> [guoren@linux.alibaba.com: add has_vector checking]
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/asm/vector.h | 14 +++++
> arch/riscv/kernel/cpufeature.c | 19 +++++++
> arch/riscv/kernel/riscv_ksyms.c | 6 +++
> arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
This file is not added to the Makefile until patch 8.
> 4 files changed, 132 insertions(+)
> create mode 100644 arch/riscv/include/asm/vector.h
> create mode 100644 arch/riscv/kernel/vector.S
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
2022-09-22 4:23 ` Samuel Holland
@ 2022-09-23 16:27 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-23 16:27 UTC (permalink / raw)
To: Samuel Holland
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Ruinland Tsai,
Han-Kuan Chen, linux-riscv, linux-kernel
(resending, as I forgot to set it to plain mail)
This is the way the original set of patches worked. I tried to change
them as little as possible for the rebase.
On Wed, Sep 21, 2022 at 9:23 PM Samuel Holland <samuel@sholland.org> wrote:
>
> On 9/21/22 16:43, Chris Stillson wrote:
> > From: Greentime Hu <greentime.hu@sifive.com>
> >
> > This patch is used to detect vector support status of CPU and use
> > riscv_vsize to save the size of all the vector registers. It assumes
> > all harts has the same capabilities in SMP system.
> >
> > [guoren@linux.alibaba.com: add has_vector checking]
> > Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> > Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> > Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> > Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> > Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> > ---
> > arch/riscv/include/asm/vector.h | 14 +++++
> > arch/riscv/kernel/cpufeature.c | 19 +++++++
> > arch/riscv/kernel/riscv_ksyms.c | 6 +++
> > arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
>
> This file is not added to the Makefile until patch 8.
>
> > 4 files changed, 132 insertions(+)
> > create mode 100644 arch/riscv/include/asm/vector.h
> > create mode 100644 arch/riscv/kernel/vector.S
>
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-09-23 16:27 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-23 16:27 UTC (permalink / raw)
To: Samuel Holland
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Ruinland Tsai,
Han-Kuan Chen, linux-riscv, linux-kernel
(resending, as I forgot to set it to plain mail)
This is the way the original set of patches worked. I tried to change
them as little as possible for the rebase.
On Wed, Sep 21, 2022 at 9:23 PM Samuel Holland <samuel@sholland.org> wrote:
>
> On 9/21/22 16:43, Chris Stillson wrote:
> > From: Greentime Hu <greentime.hu@sifive.com>
> >
> > This patch is used to detect vector support status of CPU and use
> > riscv_vsize to save the size of all the vector registers. It assumes
> > all harts has the same capabilities in SMP system.
> >
> > [guoren@linux.alibaba.com: add has_vector checking]
> > Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> > Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> > Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> > Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> > Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> > ---
> > arch/riscv/include/asm/vector.h | 14 +++++
> > arch/riscv/kernel/cpufeature.c | 19 +++++++
> > arch/riscv/kernel/riscv_ksyms.c | 6 +++
> > arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
>
> This file is not added to the Makefile until patch 8.
>
> > 4 files changed, 132 insertions(+)
> > create mode 100644 arch/riscv/include/asm/vector.h
> > create mode 100644 arch/riscv/kernel/vector.S
>
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
2022-09-23 16:27 ` Chris Stillson
@ 2022-09-24 18:01 ` Conor Dooley
-1 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-09-24 18:01 UTC (permalink / raw)
To: Chris Stillson
Cc: Samuel Holland, Greentime Hu, Guo Ren, Vincent Chen,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Anup Patel, Atish Patra, Oleg Nesterov, Guo Ren,
Heinrich Schuchardt, Arnaud Pouliquen, Paolo Bonzini,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Ruinland Tsai, Han-Kuan Chen, linux-riscv,
linux-kernel
On Fri, Sep 23, 2022 at 09:27:01AM -0700, Chris Stillson wrote:
> On Wed, Sep 21, 2022 at 9:23 PM Samuel Holland <samuel@sholland.org> wrote:
> >
> > On 9/21/22 16:43, Chris Stillson wrote:
> > > From: Greentime Hu <greentime.hu@sifive.com>
> > >
> > > This patch is used to detect vector support status of CPU and use
> > > riscv_vsize to save the size of all the vector registers. It assumes
> > > all harts has the same capabilities in SMP system.
> > >
> > > [guoren@linux.alibaba.com: add has_vector checking]
> > > Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> > > Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> > > Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> > > Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> > > ---
> > > arch/riscv/include/asm/vector.h | 14 +++++
> > > arch/riscv/kernel/cpufeature.c | 19 +++++++
> > > arch/riscv/kernel/riscv_ksyms.c | 6 +++
> > > arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
> >
> > This file is not added to the Makefile until patch 8.
> (resending, as I forgot to set it to plain mail)
And please don't top post either :)
> This is the way the original set of patches worked. I tried to change
> them as little as possible for the rebase.
What is your goal with the series? Are you going to work on getting the
whole thing merged, or just looking to tack your patch onto the end of
the on-going series?
There were two warnings from LKP & some comments from reviewers on v10,
I assume that you did not make the changes those reviewers requested as
the build warnings didn't get fixed.
https://lore.kernel.org/linux-riscv/cover.1652257230.git.greentime.hu@sifive.com/
I see a couple more caused by another patch in the series too:
../arch/riscv/kvm/vcpu_vector.c:134:6: error: variable 'reg_val' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:139:7: note: uninitialized use occurs here
if (!reg_val)
^~~~~~~
../arch/riscv/kvm/vcpu_vector.c:134:2: note: remove the 'if' if its condition is always true
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:134:6: error: variable 'reg_val' is used uninitialized whenever '&&' condition is false [-Werror,-Wsometimes-uninitialized]
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:139:7: note: uninitialized use occurs here
if (!reg_val)
^~~~~~~
../arch/riscv/kvm/vcpu_vector.c:134:6: note: remove the '&&' if its condition is always true
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:131:15: note: initialize the variable 'reg_val' to silence this warning
void *reg_val;
^
= NULL
2 errors generated.
Do you intend working on getting the series merged, or should I not
bother actually reviewing the individual patches?
If you do intend on getting it merged, can you please run checkpatch
with the --strict option and clear up the stuff it whines about & sort
out the Signed-off-bys in the series? Almost all the patches are missing
your sign-off, which is required as you are now the submitter. You can
also drop Greentime's signoff on any patch he is not the author of.
Give it a few days before resubmitting though, to give people a chance
at looking at the patchset first.
Thanks,
Conor.
> > > 4 files changed, 132 insertions(+)
> > > create mode 100644 arch/riscv/include/asm/vector.h
> > > create mode 100644 arch/riscv/kernel/vector.S
> >
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-09-24 18:01 ` Conor Dooley
0 siblings, 0 replies; 147+ messages in thread
From: Conor Dooley @ 2022-09-24 18:01 UTC (permalink / raw)
To: Chris Stillson
Cc: Samuel Holland, Greentime Hu, Guo Ren, Vincent Chen,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Anup Patel, Atish Patra, Oleg Nesterov, Guo Ren,
Heinrich Schuchardt, Arnaud Pouliquen, Paolo Bonzini,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Ruinland Tsai, Han-Kuan Chen, linux-riscv,
linux-kernel
On Fri, Sep 23, 2022 at 09:27:01AM -0700, Chris Stillson wrote:
> On Wed, Sep 21, 2022 at 9:23 PM Samuel Holland <samuel@sholland.org> wrote:
> >
> > On 9/21/22 16:43, Chris Stillson wrote:
> > > From: Greentime Hu <greentime.hu@sifive.com>
> > >
> > > This patch is used to detect vector support status of CPU and use
> > > riscv_vsize to save the size of all the vector registers. It assumes
> > > all harts has the same capabilities in SMP system.
> > >
> > > [guoren@linux.alibaba.com: add has_vector checking]
> > > Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> > > Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> > > Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> > > Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> > > ---
> > > arch/riscv/include/asm/vector.h | 14 +++++
> > > arch/riscv/kernel/cpufeature.c | 19 +++++++
> > > arch/riscv/kernel/riscv_ksyms.c | 6 +++
> > > arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
> >
> > This file is not added to the Makefile until patch 8.
> (resending, as I forgot to set it to plain mail)
And please don't top post either :)
> This is the way the original set of patches worked. I tried to change
> them as little as possible for the rebase.
What is your goal with the series? Are you going to work on getting the
whole thing merged, or just looking to tack your patch onto the end of
the on-going series?
There were two warnings from LKP & some comments from reviewers on v10,
I assume that you did not make the changes those reviewers requested as
the build warnings didn't get fixed.
https://lore.kernel.org/linux-riscv/cover.1652257230.git.greentime.hu@sifive.com/
I see a couple more caused by another patch in the series too:
../arch/riscv/kvm/vcpu_vector.c:134:6: error: variable 'reg_val' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:139:7: note: uninitialized use occurs here
if (!reg_val)
^~~~~~~
../arch/riscv/kvm/vcpu_vector.c:134:2: note: remove the 'if' if its condition is always true
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:134:6: error: variable 'reg_val' is used uninitialized whenever '&&' condition is false [-Werror,-Wsometimes-uninitialized]
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:139:7: note: uninitialized use occurs here
if (!reg_val)
^~~~~~~
../arch/riscv/kvm/vcpu_vector.c:134:6: note: remove the '&&' if its condition is always true
if ((rtype == KVM_REG_RISCV_VECTOR) &&
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_vector.c:131:15: note: initialize the variable 'reg_val' to silence this warning
void *reg_val;
^
= NULL
2 errors generated.
Do you intend working on getting the series merged, or should I not
bother actually reviewing the individual patches?
If you do intend on getting it merged, can you please run checkpatch
with the --strict option and clear up the stuff it whines about & sort
out the Signed-off-bys in the series? Almost all the patches are missing
your sign-off, which is required as you are now the submitter. You can
also drop Greentime's signoff on any patch he is not the author of.
Give it a few days before resubmitting though, to give people a chance
at looking at the patchset first.
Thanks,
Conor.
> > > 4 files changed, 132 insertions(+)
> > > create mode 100644 arch/riscv/include/asm/vector.h
> > > create mode 100644 arch/riscv/kernel/vector.S
> >
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-04 4:10 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 4:10 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, linux-riscv,
lkml, linux-mm
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch is used to detect vector support status of CPU and use
> riscv_vsize to save the size of all the vector registers. It assumes
> all harts has the same capabilities in SMP system.
Patch title is horrible. The meat of patch is vector state save/restore,
but no users of it yet. And then there are random unrelated snippets
thrown in same patch.
>
> [guoren@linux.alibaba.com: add has_vector checking]
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/asm/vector.h | 14 +++++
> arch/riscv/kernel/cpufeature.c | 19 +++++++
> arch/riscv/kernel/riscv_ksyms.c | 6 +++
> arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
> 4 files changed, 132 insertions(+)
> create mode 100644 arch/riscv/include/asm/vector.h
> create mode 100644 arch/riscv/kernel/vector.S
>
> diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
> new file mode 100644
> index 000000000000..16304b0c6a6f
> --- /dev/null
> +++ b/arch/riscv/include/asm/vector.h
> @@ -0,0 +1,14 @@
> +/* SPDX-License-Identifier: GPL-2.0-or-later */
> +/*
> + * Copyright (C) 2020 SiFive
> + */
> +
> +#ifndef __ASM_RISCV_VECTOR_H
> +#define __ASM_RISCV_VECTOR_H
> +
> +#include <linux/types.h>
> +
> +void rvv_enable(void);
> +void rvv_disable(void);
> +
> +#endif /* ! __ASM_RISCV_VECTOR_H */
> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> index 8d4448c2d4f4..0487ab19b234 100644
> --- a/arch/riscv/kernel/cpufeature.c
> +++ b/arch/riscv/kernel/cpufeature.c
> @@ -30,6 +30,14 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
>
> __ro_after_init DEFINE_STATIC_KEY_ARRAY_FALSE(riscv_isa_ext_keys, RISCV_ISA_EXT_KEY_MAX);
> EXPORT_SYMBOL(riscv_isa_ext_keys);
> +#ifdef CONFIG_FPU
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
> +#endif
This needs to be broken out to a FPU patch which actually uses
cpu_hwcap_fpu.
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
> +unsigned long riscv_vsize __read_mostly;
> +#endif
I would have moved all the detection code to patch 2/17 - including the
static branch definition and enable below (except for vlen stuff)
>
> /**
> * riscv_isa_extension_base() - Get base extension word
> @@ -249,6 +257,16 @@ void __init riscv_fill_hwcap(void)
> if (j >= 0)
> static_branch_enable(&riscv_isa_ext_keys[j]);
> }
> +
> +#ifdef CONFIG_VECTOR
> + if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
> + static_branch_enable(&cpu_hwcap_vector);
> + /* There are 32 vector registers with vlenb length. */
> + rvv_enable();
> + riscv_vsize = csr_read(CSR_VLENB) * 32;
> + rvv_disable();
> + }
> +#endif
> }
>
> #ifdef CONFIG_RISCV_ALTERNATIVE
> @@ -328,3 +346,4 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
> }
> }
> #endif
> +}
> diff --git a/arch/riscv/kernel/riscv_ksyms.c b/arch/riscv/kernel/riscv_ksyms.c
> index 5ab1c7e1a6ed..3489d2a20ca3 100644
> --- a/arch/riscv/kernel/riscv_ksyms.c
> +++ b/arch/riscv/kernel/riscv_ksyms.c
> @@ -15,3 +15,9 @@ EXPORT_SYMBOL(memmove);
> EXPORT_SYMBOL(__memset);
> EXPORT_SYMBOL(__memcpy);
> EXPORT_SYMBOL(__memmove);
> +
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +EXPORT_SYMBOL(rvv_enable);
> +EXPORT_SYMBOL(rvv_disable);
> +#endif
> diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
> new file mode 100644
> index 000000000000..9f7dc70c4443
> --- /dev/null
> +++ b/arch/riscv/kernel/vector.S
> @@ -0,0 +1,93 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2012 Regents of the University of California
> + * Copyright (C) 2017 SiFive
> + * Copyright (C) 2019 Alibaba Group Holding Limited
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation, version 2.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + */
> +
> +#include <linux/linkage.h>
> +
> +#include <asm/asm.h>
> +#include <asm/csr.h>
> +#include <asm/asm-offsets.h>
> +
> +#define vstatep a0
> +#define datap a1
> +#define x_vstart t0
> +#define x_vtype t1
> +#define x_vl t2
> +#define x_vcsr t3
> +#define incr t4
> +#define status t5
> +
A few words here as to when is this save/restore done.
Best to do this in the patch which actually uses this code.
> +ENTRY(__vstate_save)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> +
> + csrr x_vstart, CSR_VSTART
> + csrr x_vtype, CSR_VTYPE
> + csrr x_vl, CSR_VL
> + csrr x_vcsr, CSR_VCSR
> + vsetvli incr, x0, e8, m8, ta, ma
> + vse8.v v0, (datap)
> + add datap, datap, incr
> + vse8.v v8, (datap)
> + add datap, datap, incr
> + vse8.v v16, (datap)
> + add datap, datap, incr
> + vse8.v v24, (datap)
> +
> + REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_S x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_save)
> +
> +ENTRY(__vstate_restore)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> +
> + vsetvli incr, x0, e8, m8, ta, ma
> + vle8.v v0, (datap)
> + add datap, datap, incr
> + vle8.v v8, (datap)
> + add datap, datap, incr
> + vle8.v v16, (datap)
> + add datap, datap, incr
> + vle8.v v24, (datap)
> +
> + REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_L x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> + vsetvl x0, x_vl, x_vtype
> + csrw CSR_VSTART, x_vstart
> + csrw CSR_VCSR, x_vcsr
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_restore)
> +
> +ENTRY(rvv_enable)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> + ret
> +ENDPROC(rvv_enable)
> +
> +ENTRY(rvv_disable)
> + li status, SR_VS
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(rvv_disable)
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-11-04 4:10 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 4:10 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, linux-riscv,
lkml, linux-mm
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch is used to detect vector support status of CPU and use
> riscv_vsize to save the size of all the vector registers. It assumes
> all harts has the same capabilities in SMP system.
Patch title is horrible. The meat of patch is vector state save/restore,
but no users of it yet. And then there are random unrelated snippets
thrown in same patch.
>
> [guoren@linux.alibaba.com: add has_vector checking]
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/asm/vector.h | 14 +++++
> arch/riscv/kernel/cpufeature.c | 19 +++++++
> arch/riscv/kernel/riscv_ksyms.c | 6 +++
> arch/riscv/kernel/vector.S | 93 +++++++++++++++++++++++++++++++++
> 4 files changed, 132 insertions(+)
> create mode 100644 arch/riscv/include/asm/vector.h
> create mode 100644 arch/riscv/kernel/vector.S
>
> diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
> new file mode 100644
> index 000000000000..16304b0c6a6f
> --- /dev/null
> +++ b/arch/riscv/include/asm/vector.h
> @@ -0,0 +1,14 @@
> +/* SPDX-License-Identifier: GPL-2.0-or-later */
> +/*
> + * Copyright (C) 2020 SiFive
> + */
> +
> +#ifndef __ASM_RISCV_VECTOR_H
> +#define __ASM_RISCV_VECTOR_H
> +
> +#include <linux/types.h>
> +
> +void rvv_enable(void);
> +void rvv_disable(void);
> +
> +#endif /* ! __ASM_RISCV_VECTOR_H */
> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> index 8d4448c2d4f4..0487ab19b234 100644
> --- a/arch/riscv/kernel/cpufeature.c
> +++ b/arch/riscv/kernel/cpufeature.c
> @@ -30,6 +30,14 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
>
> __ro_after_init DEFINE_STATIC_KEY_ARRAY_FALSE(riscv_isa_ext_keys, RISCV_ISA_EXT_KEY_MAX);
> EXPORT_SYMBOL(riscv_isa_ext_keys);
> +#ifdef CONFIG_FPU
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
> +#endif
This needs to be broken out to a FPU patch which actually uses
cpu_hwcap_fpu.
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
> +unsigned long riscv_vsize __read_mostly;
> +#endif
I would have moved all the detection code to patch 2/17 - including the
static branch definition and enable below (except for vlen stuff)
>
> /**
> * riscv_isa_extension_base() - Get base extension word
> @@ -249,6 +257,16 @@ void __init riscv_fill_hwcap(void)
> if (j >= 0)
> static_branch_enable(&riscv_isa_ext_keys[j]);
> }
> +
> +#ifdef CONFIG_VECTOR
> + if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
> + static_branch_enable(&cpu_hwcap_vector);
> + /* There are 32 vector registers with vlenb length. */
> + rvv_enable();
> + riscv_vsize = csr_read(CSR_VLENB) * 32;
> + rvv_disable();
> + }
> +#endif
> }
>
> #ifdef CONFIG_RISCV_ALTERNATIVE
> @@ -328,3 +346,4 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
> }
> }
> #endif
> +}
> diff --git a/arch/riscv/kernel/riscv_ksyms.c b/arch/riscv/kernel/riscv_ksyms.c
> index 5ab1c7e1a6ed..3489d2a20ca3 100644
> --- a/arch/riscv/kernel/riscv_ksyms.c
> +++ b/arch/riscv/kernel/riscv_ksyms.c
> @@ -15,3 +15,9 @@ EXPORT_SYMBOL(memmove);
> EXPORT_SYMBOL(__memset);
> EXPORT_SYMBOL(__memcpy);
> EXPORT_SYMBOL(__memmove);
> +
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +EXPORT_SYMBOL(rvv_enable);
> +EXPORT_SYMBOL(rvv_disable);
> +#endif
> diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
> new file mode 100644
> index 000000000000..9f7dc70c4443
> --- /dev/null
> +++ b/arch/riscv/kernel/vector.S
> @@ -0,0 +1,93 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2012 Regents of the University of California
> + * Copyright (C) 2017 SiFive
> + * Copyright (C) 2019 Alibaba Group Holding Limited
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation, version 2.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + */
> +
> +#include <linux/linkage.h>
> +
> +#include <asm/asm.h>
> +#include <asm/csr.h>
> +#include <asm/asm-offsets.h>
> +
> +#define vstatep a0
> +#define datap a1
> +#define x_vstart t0
> +#define x_vtype t1
> +#define x_vl t2
> +#define x_vcsr t3
> +#define incr t4
> +#define status t5
> +
A few words here as to when is this save/restore done.
Best to do this in the patch which actually uses this code.
> +ENTRY(__vstate_save)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> +
> + csrr x_vstart, CSR_VSTART
> + csrr x_vtype, CSR_VTYPE
> + csrr x_vl, CSR_VL
> + csrr x_vcsr, CSR_VCSR
> + vsetvli incr, x0, e8, m8, ta, ma
> + vse8.v v0, (datap)
> + add datap, datap, incr
> + vse8.v v8, (datap)
> + add datap, datap, incr
> + vse8.v v16, (datap)
> + add datap, datap, incr
> + vse8.v v24, (datap)
> +
> + REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_S x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_save)
> +
> +ENTRY(__vstate_restore)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> +
> + vsetvli incr, x0, e8, m8, ta, ma
> + vle8.v v0, (datap)
> + add datap, datap, incr
> + vle8.v v8, (datap)
> + add datap, datap, incr
> + vle8.v v16, (datap)
> + add datap, datap, incr
> + vle8.v v24, (datap)
> +
> + REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_L x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> + vsetvl x0, x_vl, x_vtype
> + csrw CSR_VSTART, x_vstart
> + csrw CSR_VCSR, x_vcsr
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_restore)
> +
> +ENTRY(rvv_enable)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> + ret
> +ENDPROC(rvv_enable)
> +
> +ENTRY(rvv_disable)
> + li status, SR_VS
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(rvv_disable)
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-04 4:33 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 4:33 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Arnd Bergmann, Heiko Stuebner,
Jisheng Zhang, Dao Lu, Sunil V L, Andy Chiu, Guo Ren, lkml,
linux-riscv, linux-mm
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch is used to detect vector support status of CPU and use
> riscv_vsize to save the size of all the vector registers. It assumes
> all harts has the same capabilities in SMP system.
Patch title is horrible. The meat of patch is vector state save/restore,
but no users of it yet. And then there are random unrelated snippets
thrown in same patch.
> +#ifdef CONFIG_FPU
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
> +#endif
This needs to be broken out to a FPU patch which actually uses
cpu_hwcap_fpu.
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
> +unsigned long riscv_vsize __read_mostly;
> +#endif
I'd move this patch v2/17 as part of detection etc.
> +#ifdef CONFIG_VECTOR
> + if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
> + static_branch_enable(&cpu_hwcap_vector);
Ditto.
> + /* There are 32 vector registers with vlenb length. */
> + rvv_enable();
> + riscv_vsize = csr_read(CSR_VLENB) * 32;
Ditto.
> + rvv_disable();
But guess these needs to be added first as well, see below.
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +EXPORT_SYMBOL(rvv_enable);
> +EXPORT_SYMBOL(rvv_disable);
As suggested in prior review comment, we don't need to EXPORT these, for
now at least.
> diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
> new file mode 100644
> index 000000000000..9f7dc70c4443
> --- /dev/null
> +++ b/arch/riscv/kernel/vector.S
> @@ -0,0 +1,93 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2012 Regents of the University of California
> + * Copyright (C) 2017 SiFive
> + * Copyright (C) 2019 Alibaba Group Holding Limited
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation, version 2.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + */
SPDX hdr ...
> +
> +#include <linux/linkage.h>
> +
> +#include <asm/asm.h>
> +#include <asm/csr.h>
> +#include <asm/asm-offsets.h>
> +
> +#define vstatep a0
> +#define datap a1
> +#define x_vstart t0
> +#define x_vtype t1
> +#define x_vl t2
> +#define x_vcsr t3
> +#define incr t4
> +#define status t5
> +
A few words here as to when is this save/restore done.
Perhaps best to add this code in the patch which actually uses these.
> +ENTRY(__vstate_save)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> +
> + csrr x_vstart, CSR_VSTART
> + csrr x_vtype, CSR_VTYPE
> + csrr x_vl, CSR_VL
> + csrr x_vcsr, CSR_VCSR
> + vsetvli incr, x0, e8, m8, ta, ma
> + vse8.v v0, (datap)
> + add datap, datap, incr
> + vse8.v v8, (datap)
> + add datap, datap, incr
> + vse8.v v16, (datap)
> + add datap, datap, incr
> + vse8.v v24, (datap)
> +
> + REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_S x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_save)
> +
> +ENTRY(__vstate_restore)
> + li status, SR_VS
> + csrs CSR_STATUS, status
This is rvv_enable code duplicated inline.
> +
> + vsetvli incr, x0, e8, m8, ta, ma
> + vle8.v v0, (datap)
> + add datap, datap, incr
> + vle8.v v8, (datap)
> + add datap, datap, incr
> + vle8.v v16, (datap)
> + add datap, datap, incr
> + vle8.v v24, (datap)
> +
> + REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_L x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> + vsetvl x0, x_vl, x_vtype
> + csrw CSR_VSTART, x_vstart
> + csrw CSR_VCSR, x_vcsr
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_restore)
> +
> +ENTRY(rvv_enable)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> + ret
> +ENDPROC(rvv_enable)
> +
> +ENTRY(rvv_disable)
> + li status, SR_VS
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(rvv_disable)
I'd suggest these be made asm macros, to avoid function call overhead
and duplication as in save/restore above.
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 05/17] riscv: Add has_vector/riscv_vsize to save vector features.
@ 2022-11-04 4:33 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 4:33 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Guo Ren, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Arnaud Pouliquen, Paolo Bonzini, Arnd Bergmann, Heiko Stuebner,
Jisheng Zhang, Dao Lu, Sunil V L, Andy Chiu, Guo Ren, lkml,
linux-riscv, linux-mm
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch is used to detect vector support status of CPU and use
> riscv_vsize to save the size of all the vector registers. It assumes
> all harts has the same capabilities in SMP system.
Patch title is horrible. The meat of patch is vector state save/restore,
but no users of it yet. And then there are random unrelated snippets
thrown in same patch.
> +#ifdef CONFIG_FPU
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
> +#endif
This needs to be broken out to a FPU patch which actually uses
cpu_hwcap_fpu.
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
> +unsigned long riscv_vsize __read_mostly;
> +#endif
I'd move this patch v2/17 as part of detection etc.
> +#ifdef CONFIG_VECTOR
> + if (elf_hwcap & COMPAT_HWCAP_ISA_V) {
> + static_branch_enable(&cpu_hwcap_vector);
Ditto.
> + /* There are 32 vector registers with vlenb length. */
> + rvv_enable();
> + riscv_vsize = csr_read(CSR_VLENB) * 32;
Ditto.
> + rvv_disable();
But guess these needs to be added first as well, see below.
> +#ifdef CONFIG_VECTOR
> +#include <asm/vector.h>
> +EXPORT_SYMBOL(rvv_enable);
> +EXPORT_SYMBOL(rvv_disable);
As suggested in prior review comment, we don't need to EXPORT these, for
now at least.
> diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
> new file mode 100644
> index 000000000000..9f7dc70c4443
> --- /dev/null
> +++ b/arch/riscv/kernel/vector.S
> @@ -0,0 +1,93 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2012 Regents of the University of California
> + * Copyright (C) 2017 SiFive
> + * Copyright (C) 2019 Alibaba Group Holding Limited
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation, version 2.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + */
SPDX hdr ...
> +
> +#include <linux/linkage.h>
> +
> +#include <asm/asm.h>
> +#include <asm/csr.h>
> +#include <asm/asm-offsets.h>
> +
> +#define vstatep a0
> +#define datap a1
> +#define x_vstart t0
> +#define x_vtype t1
> +#define x_vl t2
> +#define x_vcsr t3
> +#define incr t4
> +#define status t5
> +
A few words here as to when is this save/restore done.
Perhaps best to add this code in the patch which actually uses these.
> +ENTRY(__vstate_save)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> +
> + csrr x_vstart, CSR_VSTART
> + csrr x_vtype, CSR_VTYPE
> + csrr x_vl, CSR_VL
> + csrr x_vcsr, CSR_VCSR
> + vsetvli incr, x0, e8, m8, ta, ma
> + vse8.v v0, (datap)
> + add datap, datap, incr
> + vse8.v v8, (datap)
> + add datap, datap, incr
> + vse8.v v16, (datap)
> + add datap, datap, incr
> + vse8.v v24, (datap)
> +
> + REG_S x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_S x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_S x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_S x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_save)
> +
> +ENTRY(__vstate_restore)
> + li status, SR_VS
> + csrs CSR_STATUS, status
This is rvv_enable code duplicated inline.
> +
> + vsetvli incr, x0, e8, m8, ta, ma
> + vle8.v v0, (datap)
> + add datap, datap, incr
> + vle8.v v8, (datap)
> + add datap, datap, incr
> + vle8.v v16, (datap)
> + add datap, datap, incr
> + vle8.v v24, (datap)
> +
> + REG_L x_vstart, RISCV_V_STATE_VSTART(vstatep)
> + REG_L x_vtype, RISCV_V_STATE_VTYPE(vstatep)
> + REG_L x_vl, RISCV_V_STATE_VL(vstatep)
> + REG_L x_vcsr, RISCV_V_STATE_VCSR(vstatep)
> + vsetvl x0, x_vl, x_vtype
> + csrw CSR_VSTART, x_vstart
> + csrw CSR_VCSR, x_vcsr
> +
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(__vstate_restore)
> +
> +ENTRY(rvv_enable)
> + li status, SR_VS
> + csrs CSR_STATUS, status
> + ret
> +ENDPROC(rvv_enable)
> +
> +ENTRY(rvv_disable)
> + li status, SR_VS
> + csrc CSR_STATUS, status
> + ret
> +ENDPROC(rvv_disable)
I'd suggest these be made asm macros, to avoid function call overhead
and duplication as in save/restore above.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 06/17] riscv: Reset vector register
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Vincent Chen, Han-Kuan Chen, Greentime Hu,
Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu,
Peter Zijlstra (Intel),
Sunil V L, Ruinland Tsai, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Frederic Weisbecker, Changbin Du, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Peter Collingbourne, Eugene Syromiatnikov, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <guoren@linux.alibaba.com>
Reset vector registers at boot-time and disable vector instructions
execution for kernel mode.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Co-developed-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/kernel/entry.S | 6 +++---
arch/riscv/kernel/head.S | 35 +++++++++++++++++++++++++++++------
2 files changed, 32 insertions(+), 9 deletions(-)
diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index b9eda3fcbd6d..1e9987376591 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -77,10 +77,10 @@ _save_context:
* Disable user-mode memory access as it should only be set in the
* actual user copy routines.
*
- * Disable the FPU to detect illegal usage of floating point in kernel
- * space.
+ * Disable the FPU/Vector to detect illegal usage of floating point
+ * or vector in kernel space.
*/
- li t0, SR_SUM | SR_FS
+ li t0, SR_SUM | SR_FS | SR_VS
REG_L s0, TASK_TI_USER_SP(tp)
csrrc s1, CSR_STATUS, t0
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index b865046e4dbb..2c81ca42ec4e 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -140,10 +140,10 @@ secondary_start_sbi:
.option pop
/*
- * Disable FPU to detect illegal usage of
- * floating point in kernel space
+ * Disable FPU & VECTOR to detect illegal usage of
+ * floating point or vector in kernel space
*/
- li t0, SR_FS
+ li t0, SR_FS | SR_VS
csrc CSR_STATUS, t0
/* Set trap vector to spin forever to help debug */
@@ -234,10 +234,10 @@ pmp_done:
.option pop
/*
- * Disable FPU to detect illegal usage of
- * floating point in kernel space
+ * Disable FPU & VECTOR to detect illegal usage of
+ * floating point or vector in kernel space
*/
- li t0, SR_FS
+ li t0, SR_FS | SR_VS
csrc CSR_STATUS, t0
#ifdef CONFIG_RISCV_BOOT_SPINWAIT
@@ -431,6 +431,29 @@ ENTRY(reset_regs)
csrw fcsr, 0
/* note that the caller must clear SR_FS */
#endif /* CONFIG_FPU */
+
+#ifdef CONFIG_VECTOR
+ csrr t0, CSR_MISA
+ li t1, COMPAT_HWCAP_ISA_V
+ and t0, t0, t1
+ beqz t0, .Lreset_regs_done
+
+ /*
+ * Clear vector registers and reset vcsr
+ * VLMAX has a defined value, VLEN is a constant,
+ * and this form of vsetvli is defined to set vl to VLMAX.
+ */
+ li t1, SR_VS
+ csrs CSR_STATUS, t1
+ csrs CSR_VCSR, x0
+ vsetvli t1, x0, e8, m8, ta, ma
+ vmv.v.i v0, 0
+ vmv.v.i v8, 0
+ vmv.v.i v16, 0
+ vmv.v.i v24, 0
+ /* note that the caller must clear SR_VS */
+#endif /* CONFIG_VECTOR */
+
.Lreset_regs_done:
ret
END(reset_regs)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 06/17] riscv: Reset vector register
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Vincent Chen, Han-Kuan Chen, Greentime Hu,
Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu,
Peter Zijlstra (Intel),
Sunil V L, Ruinland Tsai, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Frederic Weisbecker, Changbin Du, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Peter Collingbourne, Eugene Syromiatnikov, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <guoren@linux.alibaba.com>
Reset vector registers at boot-time and disable vector instructions
execution for kernel mode.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Co-developed-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/kernel/entry.S | 6 +++---
arch/riscv/kernel/head.S | 35 +++++++++++++++++++++++++++++------
2 files changed, 32 insertions(+), 9 deletions(-)
diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index b9eda3fcbd6d..1e9987376591 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -77,10 +77,10 @@ _save_context:
* Disable user-mode memory access as it should only be set in the
* actual user copy routines.
*
- * Disable the FPU to detect illegal usage of floating point in kernel
- * space.
+ * Disable the FPU/Vector to detect illegal usage of floating point
+ * or vector in kernel space.
*/
- li t0, SR_SUM | SR_FS
+ li t0, SR_SUM | SR_FS | SR_VS
REG_L s0, TASK_TI_USER_SP(tp)
csrrc s1, CSR_STATUS, t0
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index b865046e4dbb..2c81ca42ec4e 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -140,10 +140,10 @@ secondary_start_sbi:
.option pop
/*
- * Disable FPU to detect illegal usage of
- * floating point in kernel space
+ * Disable FPU & VECTOR to detect illegal usage of
+ * floating point or vector in kernel space
*/
- li t0, SR_FS
+ li t0, SR_FS | SR_VS
csrc CSR_STATUS, t0
/* Set trap vector to spin forever to help debug */
@@ -234,10 +234,10 @@ pmp_done:
.option pop
/*
- * Disable FPU to detect illegal usage of
- * floating point in kernel space
+ * Disable FPU & VECTOR to detect illegal usage of
+ * floating point or vector in kernel space
*/
- li t0, SR_FS
+ li t0, SR_FS | SR_VS
csrc CSR_STATUS, t0
#ifdef CONFIG_RISCV_BOOT_SPINWAIT
@@ -431,6 +431,29 @@ ENTRY(reset_regs)
csrw fcsr, 0
/* note that the caller must clear SR_FS */
#endif /* CONFIG_FPU */
+
+#ifdef CONFIG_VECTOR
+ csrr t0, CSR_MISA
+ li t1, COMPAT_HWCAP_ISA_V
+ and t0, t0, t1
+ beqz t0, .Lreset_regs_done
+
+ /*
+ * Clear vector registers and reset vcsr
+ * VLMAX has a defined value, VLEN is a constant,
+ * and this form of vsetvli is defined to set vl to VLMAX.
+ */
+ li t1, SR_VS
+ csrs CSR_STATUS, t1
+ csrs CSR_VCSR, x0
+ vsetvli t1, x0, e8, m8, ta, ma
+ vmv.v.i v0, 0
+ vmv.v.i v8, 0
+ vmv.v.i v16, 0
+ vmv.v.i v24, 0
+ /* note that the caller must clear SR_VS */
+#endif /* CONFIG_VECTOR */
+
.Lreset_regs_done:
ret
END(reset_regs)
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 06/17] riscv: Reset vector register
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Guo Ren, Vincent Chen, Han-Kuan Chen, Greentime Hu,
Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu,
Peter Zijlstra (Intel),
Sunil V L, Ruinland Tsai, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Frederic Weisbecker, Changbin Du, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Peter Collingbourne, Eugene Syromiatnikov, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Guo Ren <guoren@linux.alibaba.com>
Reset vector registers at boot-time and disable vector instructions
execution for kernel mode.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Co-developed-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Co-developed-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
---
arch/riscv/kernel/entry.S | 6 +++---
arch/riscv/kernel/head.S | 35 +++++++++++++++++++++++++++++------
2 files changed, 32 insertions(+), 9 deletions(-)
diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index b9eda3fcbd6d..1e9987376591 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -77,10 +77,10 @@ _save_context:
* Disable user-mode memory access as it should only be set in the
* actual user copy routines.
*
- * Disable the FPU to detect illegal usage of floating point in kernel
- * space.
+ * Disable the FPU/Vector to detect illegal usage of floating point
+ * or vector in kernel space.
*/
- li t0, SR_SUM | SR_FS
+ li t0, SR_SUM | SR_FS | SR_VS
REG_L s0, TASK_TI_USER_SP(tp)
csrrc s1, CSR_STATUS, t0
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index b865046e4dbb..2c81ca42ec4e 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -140,10 +140,10 @@ secondary_start_sbi:
.option pop
/*
- * Disable FPU to detect illegal usage of
- * floating point in kernel space
+ * Disable FPU & VECTOR to detect illegal usage of
+ * floating point or vector in kernel space
*/
- li t0, SR_FS
+ li t0, SR_FS | SR_VS
csrc CSR_STATUS, t0
/* Set trap vector to spin forever to help debug */
@@ -234,10 +234,10 @@ pmp_done:
.option pop
/*
- * Disable FPU to detect illegal usage of
- * floating point in kernel space
+ * Disable FPU & VECTOR to detect illegal usage of
+ * floating point or vector in kernel space
*/
- li t0, SR_FS
+ li t0, SR_FS | SR_VS
csrc CSR_STATUS, t0
#ifdef CONFIG_RISCV_BOOT_SPINWAIT
@@ -431,6 +431,29 @@ ENTRY(reset_regs)
csrw fcsr, 0
/* note that the caller must clear SR_FS */
#endif /* CONFIG_FPU */
+
+#ifdef CONFIG_VECTOR
+ csrr t0, CSR_MISA
+ li t1, COMPAT_HWCAP_ISA_V
+ and t0, t0, t1
+ beqz t0, .Lreset_regs_done
+
+ /*
+ * Clear vector registers and reset vcsr
+ * VLMAX has a defined value, VLEN is a constant,
+ * and this form of vsetvli is defined to set vl to VLMAX.
+ */
+ li t1, SR_VS
+ csrs CSR_STATUS, t1
+ csrs CSR_VCSR, x0
+ vsetvli t1, x0, e8, m8, ta, ma
+ vmv.v.i v0, 0
+ vmv.v.i v8, 0
+ vmv.v.i v16, 0
+ vmv.v.i v24, 0
+ /* note that the caller must clear SR_VS */
+#endif /* CONFIG_VECTOR */
+
.Lreset_regs_done:
ret
END(reset_regs)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* Re: [PATCH v12 06/17] riscv: Reset vector register
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-04 5:01 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 5:01 UTC (permalink / raw)
To: Chris Stillson
Cc: Guo Ren, Vincent Chen, Han-Kuan Chen, Greentime Hu,
Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Mayuresh Chitale,
Paolo Bonzini, linux-riscv, lkml, Andy Chiu
On 9/21/22 14:43, Chris Stillson wrote:
> From: Guo Ren <guoren@linux.alibaba.com>
>
> Reset vector registers at boot-time and disable vector instructions
> execution for kernel mode.
Perhaps bike-shedding, but "Reset" has a different connotation in
kernel, this is clear registers IMO. And "Reset Vector ..." sounds
totally different at first glance.
> - * Disable the FPU to detect illegal usage of floating point in kernel
> - * space.
> + * Disable the FPU/Vector to detect illegal usage of floating point
> + * or vector in kernel space.
> */
> - li t0, SR_SUM | SR_FS
> + li t0, SR_SUM | SR_FS | SR_VS
Is VS writable in implementations not implementing V hardware.
Priv spec seems to be confusing. It states
"The FS[1:0] and VS[1:0] WARL fields..."
Above implies it can be written always but will read legal values only.
But then this follows.
"If neither the v registers nor S-mode is implemented, then VS
is read-only zero. If S-mode is implemented but the v registers
are not, VS may optionally be read-only zero"
What does optionally mean for software ?
>
> REG_L s0, TASK_TI_USER_SP(tp)
> csrrc s1, CSR_STATUS, t0
> diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
> index b865046e4dbb..2c81ca42ec4e 100644
> --- a/arch/riscv/kernel/head.S
> +++ b/arch/riscv/kernel/head.S
> @@ -140,10 +140,10 @@ secondary_start_sbi:
> .option pop
>
> /*
> - * Disable FPU to detect illegal usage of
> - * floating point in kernel space
> + * Disable FPU & VECTOR to detect illegal usage of
> + * floating point or vector in kernel space
> */
> - li t0, SR_FS
> + li t0, SR_FS | SR_VS
> csrc CSR_STATUS, t0
>
> /* Set trap vector to spin forever to help debug */
> @@ -234,10 +234,10 @@ pmp_done:
> .option pop
>
> /*
> - * Disable FPU to detect illegal usage of
> - * floating point in kernel space
> + * Disable FPU & VECTOR to detect illegal usage of
> + * floating point or vector in kernel space
> */
> - li t0, SR_FS
> + li t0, SR_FS | SR_VS
> csrc CSR_STATUS, t0
Third instance of duplicated SR_FS | SR_VS. Better to add a helper
SR_FS_VS or some such macro.
>
> #ifdef CONFIG_RISCV_BOOT_SPINWAIT
> @@ -431,6 +431,29 @@ ENTRY(reset_regs)
> csrw fcsr, 0
> /* note that the caller must clear SR_FS */
> #endif /* CONFIG_FPU */
> +
> +#ifdef CONFIG_VECTOR
> + csrr t0, CSR_MISA
> + li t1, COMPAT_HWCAP_ISA_V
> + and t0, t0, t1
> + beqz t0, .Lreset_regs_done
> +
> + /*
> + * Clear vector registers and reset vcsr
> + * VLMAX has a defined value, VLEN is a constant,
> + * and this form of vsetvli is defined to set vl to VLMAX.
> + */
> + li t1, SR_VS
> + csrs CSR_STATUS, t1
> + csrs CSR_VCSR, x0
> + vsetvli t1, x0, e8, m8, ta, ma
> + vmv.v.i v0, 0
> + vmv.v.i v8, 0
> + vmv.v.i v16, 0
> + vmv.v.i v24, 0
> + /* note that the caller must clear SR_VS */
Is that actually happening ?
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 06/17] riscv: Reset vector register
@ 2022-11-04 5:01 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 5:01 UTC (permalink / raw)
To: Chris Stillson
Cc: Guo Ren, Vincent Chen, Han-Kuan Chen, Greentime Hu,
Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Mayuresh Chitale,
Paolo Bonzini, linux-riscv, lkml, Andy Chiu
On 9/21/22 14:43, Chris Stillson wrote:
> From: Guo Ren <guoren@linux.alibaba.com>
>
> Reset vector registers at boot-time and disable vector instructions
> execution for kernel mode.
Perhaps bike-shedding, but "Reset" has a different connotation in
kernel, this is clear registers IMO. And "Reset Vector ..." sounds
totally different at first glance.
> - * Disable the FPU to detect illegal usage of floating point in kernel
> - * space.
> + * Disable the FPU/Vector to detect illegal usage of floating point
> + * or vector in kernel space.
> */
> - li t0, SR_SUM | SR_FS
> + li t0, SR_SUM | SR_FS | SR_VS
Is VS writable in implementations not implementing V hardware.
Priv spec seems to be confusing. It states
"The FS[1:0] and VS[1:0] WARL fields..."
Above implies it can be written always but will read legal values only.
But then this follows.
"If neither the v registers nor S-mode is implemented, then VS
is read-only zero. If S-mode is implemented but the v registers
are not, VS may optionally be read-only zero"
What does optionally mean for software ?
>
> REG_L s0, TASK_TI_USER_SP(tp)
> csrrc s1, CSR_STATUS, t0
> diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
> index b865046e4dbb..2c81ca42ec4e 100644
> --- a/arch/riscv/kernel/head.S
> +++ b/arch/riscv/kernel/head.S
> @@ -140,10 +140,10 @@ secondary_start_sbi:
> .option pop
>
> /*
> - * Disable FPU to detect illegal usage of
> - * floating point in kernel space
> + * Disable FPU & VECTOR to detect illegal usage of
> + * floating point or vector in kernel space
> */
> - li t0, SR_FS
> + li t0, SR_FS | SR_VS
> csrc CSR_STATUS, t0
>
> /* Set trap vector to spin forever to help debug */
> @@ -234,10 +234,10 @@ pmp_done:
> .option pop
>
> /*
> - * Disable FPU to detect illegal usage of
> - * floating point in kernel space
> + * Disable FPU & VECTOR to detect illegal usage of
> + * floating point or vector in kernel space
> */
> - li t0, SR_FS
> + li t0, SR_FS | SR_VS
> csrc CSR_STATUS, t0
Third instance of duplicated SR_FS | SR_VS. Better to add a helper
SR_FS_VS or some such macro.
>
> #ifdef CONFIG_RISCV_BOOT_SPINWAIT
> @@ -431,6 +431,29 @@ ENTRY(reset_regs)
> csrw fcsr, 0
> /* note that the caller must clear SR_FS */
> #endif /* CONFIG_FPU */
> +
> +#ifdef CONFIG_VECTOR
> + csrr t0, CSR_MISA
> + li t1, COMPAT_HWCAP_ISA_V
> + and t0, t0, t1
> + beqz t0, .Lreset_regs_done
> +
> + /*
> + * Clear vector registers and reset vcsr
> + * VLMAX has a defined value, VLEN is a constant,
> + * and this form of vsetvli is defined to set vl to VLMAX.
> + */
> + li t1, SR_VS
> + csrs CSR_STATUS, t1
> + csrs CSR_VCSR, x0
> + vsetvli t1, x0, e8, m8, ta, ma
> + vmv.v.i v0, 0
> + vmv.v.i v8, 0
> + vmv.v.i v16, 0
> + vmv.v.i v24, 0
> + /* note that the caller must clear SR_VS */
Is that actually happening ?
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 06/17] riscv: Reset vector register
2022-11-04 5:01 ` Vineet Gupta
@ 2022-11-04 8:45 ` Guo Ren
-1 siblings, 0 replies; 147+ messages in thread
From: Guo Ren @ 2022-11-04 8:45 UTC (permalink / raw)
To: Vineet Gupta
Cc: Chris Stillson, Guo Ren, Vincent Chen, Han-Kuan Chen,
Greentime Hu, Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Mayuresh Chitale,
Paolo Bonzini, linux-riscv, lkml, Andy Chiu
On Fri, Nov 4, 2022 at 1:01 PM Vineet Gupta <vineet.gupta@linux.dev> wrote:
>
> On 9/21/22 14:43, Chris Stillson wrote:
> > From: Guo Ren <guoren@linux.alibaba.com>
> >
> > Reset vector registers at boot-time and disable vector instructions
> > execution for kernel mode.
>
> Perhaps bike-shedding, but "Reset" has a different connotation in
> kernel, this is clear registers IMO. And "Reset Vector ..." sounds
> totally different at first glance.
Agree, "Clear vector registers" is okay.
>
>
> > - * Disable the FPU to detect illegal usage of floating point in kernel
> > - * space.
> > + * Disable the FPU/Vector to detect illegal usage of floating point
> > + * or vector in kernel space.
> > */
> > - li t0, SR_SUM | SR_FS
> > + li t0, SR_SUM | SR_FS | SR_VS
>
> Is VS writable in implementations not implementing V hardware.
>
> Priv spec seems to be confusing. It states
>
> "The FS[1:0] and VS[1:0] WARL fields..."
>
> Above implies it can be written always but will read legal values only.
> But then this follows.
>
> "If neither the v registers nor S-mode is implemented, then VS
> is read-only zero. If S-mode is implemented but the v registers
> are not, VS may optionally be read-only zero"
>
> What does optionally mean for software ?
The read-only zero bit is safe for writing 1, but the result is still
zero. So let's keep it for easier coding.
>
> >
> > REG_L s0, TASK_TI_USER_SP(tp)
> > csrrc s1, CSR_STATUS, t0
> > diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
> > index b865046e4dbb..2c81ca42ec4e 100644
> > --- a/arch/riscv/kernel/head.S
> > +++ b/arch/riscv/kernel/head.S
> > @@ -140,10 +140,10 @@ secondary_start_sbi:
> > .option pop
> >
> > /*
> > - * Disable FPU to detect illegal usage of
> > - * floating point in kernel space
> > + * Disable FPU & VECTOR to detect illegal usage of
> > + * floating point or vector in kernel space
> > */
> > - li t0, SR_FS
> > + li t0, SR_FS | SR_VS
> > csrc CSR_STATUS, t0
> >
> > /* Set trap vector to spin forever to help debug */
> > @@ -234,10 +234,10 @@ pmp_done:
> > .option pop
> >
> > /*
> > - * Disable FPU to detect illegal usage of
> > - * floating point in kernel space
> > + * Disable FPU & VECTOR to detect illegal usage of
> > + * floating point or vector in kernel space
> > */
> > - li t0, SR_FS
> > + li t0, SR_FS | SR_VS
> > csrc CSR_STATUS, t0
>
> Third instance of duplicated SR_FS | SR_VS. Better to add a helper
> SR_FS_VS or some such macro.
Good point. But we could move it to another patch and define a new
SR_AXS for all.
#define SR_AXS (SR_FS | SR_VS | SR_XS)
>
> >
> > #ifdef CONFIG_RISCV_BOOT_SPINWAIT
> > @@ -431,6 +431,29 @@ ENTRY(reset_regs)
> > csrw fcsr, 0
> > /* note that the caller must clear SR_FS */
> > #endif /* CONFIG_FPU */
> > +
> > +#ifdef CONFIG_VECTOR
> > + csrr t0, CSR_MISA
> > + li t1, COMPAT_HWCAP_ISA_V
> > + and t0, t0, t1
> > + beqz t0, .Lreset_regs_done
> > +
> > + /*
> > + * Clear vector registers and reset vcsr
> > + * VLMAX has a defined value, VLEN is a constant,
> > + * and this form of vsetvli is defined to set vl to VLMAX.
> > + */
> > + li t1, SR_VS
> > + csrs CSR_STATUS, t1
> > + csrs CSR_VCSR, x0
> > + vsetvli t1, x0, e8, m8, ta, ma
> > + vmv.v.i v0, 0
> > + vmv.v.i v8, 0
> > + vmv.v.i v16, 0
> > + vmv.v.i v24, 0
> > + /* note that the caller must clear SR_VS */
>
> Is that actually happening ?
Yes, It's the same as FPU, see head.S _start_kernel:
ENTRY(_start_kernel)
..
/* Reset all registers except ra, a0, a1 */
call reset_regs
...
>
>
--
Best Regards
Guo Ren
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 06/17] riscv: Reset vector register
@ 2022-11-04 8:45 ` Guo Ren
0 siblings, 0 replies; 147+ messages in thread
From: Guo Ren @ 2022-11-04 8:45 UTC (permalink / raw)
To: Vineet Gupta
Cc: Chris Stillson, Guo Ren, Vincent Chen, Han-Kuan Chen,
Greentime Hu, Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Mayuresh Chitale,
Paolo Bonzini, linux-riscv, lkml, Andy Chiu
On Fri, Nov 4, 2022 at 1:01 PM Vineet Gupta <vineet.gupta@linux.dev> wrote:
>
> On 9/21/22 14:43, Chris Stillson wrote:
> > From: Guo Ren <guoren@linux.alibaba.com>
> >
> > Reset vector registers at boot-time and disable vector instructions
> > execution for kernel mode.
>
> Perhaps bike-shedding, but "Reset" has a different connotation in
> kernel, this is clear registers IMO. And "Reset Vector ..." sounds
> totally different at first glance.
Agree, "Clear vector registers" is okay.
>
>
> > - * Disable the FPU to detect illegal usage of floating point in kernel
> > - * space.
> > + * Disable the FPU/Vector to detect illegal usage of floating point
> > + * or vector in kernel space.
> > */
> > - li t0, SR_SUM | SR_FS
> > + li t0, SR_SUM | SR_FS | SR_VS
>
> Is VS writable in implementations not implementing V hardware.
>
> Priv spec seems to be confusing. It states
>
> "The FS[1:0] and VS[1:0] WARL fields..."
>
> Above implies it can be written always but will read legal values only.
> But then this follows.
>
> "If neither the v registers nor S-mode is implemented, then VS
> is read-only zero. If S-mode is implemented but the v registers
> are not, VS may optionally be read-only zero"
>
> What does optionally mean for software ?
The read-only zero bit is safe for writing 1, but the result is still
zero. So let's keep it for easier coding.
>
> >
> > REG_L s0, TASK_TI_USER_SP(tp)
> > csrrc s1, CSR_STATUS, t0
> > diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
> > index b865046e4dbb..2c81ca42ec4e 100644
> > --- a/arch/riscv/kernel/head.S
> > +++ b/arch/riscv/kernel/head.S
> > @@ -140,10 +140,10 @@ secondary_start_sbi:
> > .option pop
> >
> > /*
> > - * Disable FPU to detect illegal usage of
> > - * floating point in kernel space
> > + * Disable FPU & VECTOR to detect illegal usage of
> > + * floating point or vector in kernel space
> > */
> > - li t0, SR_FS
> > + li t0, SR_FS | SR_VS
> > csrc CSR_STATUS, t0
> >
> > /* Set trap vector to spin forever to help debug */
> > @@ -234,10 +234,10 @@ pmp_done:
> > .option pop
> >
> > /*
> > - * Disable FPU to detect illegal usage of
> > - * floating point in kernel space
> > + * Disable FPU & VECTOR to detect illegal usage of
> > + * floating point or vector in kernel space
> > */
> > - li t0, SR_FS
> > + li t0, SR_FS | SR_VS
> > csrc CSR_STATUS, t0
>
> Third instance of duplicated SR_FS | SR_VS. Better to add a helper
> SR_FS_VS or some such macro.
Good point. But we could move it to another patch and define a new
SR_AXS for all.
#define SR_AXS (SR_FS | SR_VS | SR_XS)
>
> >
> > #ifdef CONFIG_RISCV_BOOT_SPINWAIT
> > @@ -431,6 +431,29 @@ ENTRY(reset_regs)
> > csrw fcsr, 0
> > /* note that the caller must clear SR_FS */
> > #endif /* CONFIG_FPU */
> > +
> > +#ifdef CONFIG_VECTOR
> > + csrr t0, CSR_MISA
> > + li t1, COMPAT_HWCAP_ISA_V
> > + and t0, t0, t1
> > + beqz t0, .Lreset_regs_done
> > +
> > + /*
> > + * Clear vector registers and reset vcsr
> > + * VLMAX has a defined value, VLEN is a constant,
> > + * and this form of vsetvli is defined to set vl to VLMAX.
> > + */
> > + li t1, SR_VS
> > + csrs CSR_STATUS, t1
> > + csrs CSR_VCSR, x0
> > + vsetvli t1, x0, e8, m8, ta, ma
> > + vmv.v.i v0, 0
> > + vmv.v.i v8, 0
> > + vmv.v.i v16, 0
> > + vmv.v.i v24, 0
> > + /* note that the caller must clear SR_VS */
>
> Is that actually happening ?
Yes, It's the same as FPU, see head.S _start_kernel:
ENTRY(_start_kernel)
..
/* Reset all registers except ra, a0, a1 */
call reset_regs
...
>
>
--
Best Regards
Guo Ren
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 06/17] riscv: Reset vector register
2022-09-21 21:43 ` Chris Stillson
@ 2023-01-20 12:20 ` Heiko Stübner
-1 siblings, 0 replies; 147+ messages in thread
From: Heiko Stübner @ 2023-01-20 12:20 UTC (permalink / raw)
To: linux-riscv
Cc: Guo Ren, Vincent Chen, Han-Kuan Chen, Greentime Hu,
Palmer Dabbelt, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Jisheng Zhang, Dao Lu, Peter Zijlstra (Intel),
Sunil V L, Ruinland Tsai, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Frederic Weisbecker, Changbin Du, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Will Deacon, Mark Brown,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Peter Collingbourne, Eugene Syromiatnikov, Colin Cross,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv, Chris Stillson
Am Mittwoch, 21. September 2022, 23:43:48 CET schrieb Chris Stillson:
> @@ -431,6 +431,29 @@ ENTRY(reset_regs)
> csrw fcsr, 0
> /* note that the caller must clear SR_FS */
> #endif /* CONFIG_FPU */
> +
> +#ifdef CONFIG_VECTOR
> + csrr t0, CSR_MISA
> + li t1, COMPAT_HWCAP_ISA_V
> + and t0, t0, t1
> + beqz t0, .Lreset_regs_done
> +
> + /*
> + * Clear vector registers and reset vcsr
> + * VLMAX has a defined value, VLEN is a constant,
> + * and this form of vsetvli is defined to set vl to VLMAX.
> + */
> + li t1, SR_VS
> + csrs CSR_STATUS, t1
> + csrs CSR_VCSR, x0
> + vsetvli t1, x0, e8, m8, ta, ma
> + vmv.v.i v0, 0
> + vmv.v.i v8, 0
> + vmv.v.i v16, 0
> + vmv.v.i v24, 0
> + /* note that the caller must clear SR_VS */
> +#endif /* CONFIG_VECTOR */
> +
> .Lreset_regs_done:
Not sure how much they go together, but the #ifdef CONFIG_FPU block above
your new VECTOR block also jumps to the same .Lreset_regs_done, so with
the patch as is the vector-reset block is never reached in the !FPU case.
So maybe making them independent of each other might prevent issues
down the roead.
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 06/17] riscv: Reset vector register
@ 2023-01-20 12:20 ` Heiko Stübner
0 siblings, 0 replies; 147+ messages in thread
From: Heiko Stübner @ 2023-01-20 12:20 UTC (permalink / raw)
To: linux-riscv
Cc: Mark Rutland, Guo Ren, kvm, Peter Zijlstra (Intel),
Catalin Marinas, Paul Walmsley, Palmer Dabbelt, linux-kernel,
linux-mm, Guo Ren, Jisheng Zhang, Qinglin Pan,
Nicolas Saenz Julienne, Will Deacon, Ard Biesheuvel, Vitaly Wool,
Davidlohr Bueso, Yury Norov, Changbin Du, Anup Patel,
Huacai Chen, Myrtle Shah, Eugene Syromiatnikov, linux-riscv,
Tsukasa OI, Ruinland Tsai, Greentime Hu, Li Zhengyu,
Heinrich Schuchardt, Barret Rhoden, Albert Ou, Kees Cook,
Arnd Bergmann, Frederic Weisbecker, Suren Baghdasaryan, Dao Lu,
Alexander Graf, Atish Patra, Mayuresh Chitale, Alexandre Ghiti,
Peter Collingbourne, Janosch Frank, Chris Stillson,
Christian Brauner, Oleg Nesterov, Colin Cross, Alexey Dobriyan,
Han-Kuan Chen, Vincent Chen, Mark Brown, Palmer Dabbelt,
Eric Biederman, kvm-riscv, Paolo Bonzini, Andrew Morton
Am Mittwoch, 21. September 2022, 23:43:48 CET schrieb Chris Stillson:
> @@ -431,6 +431,29 @@ ENTRY(reset_regs)
> csrw fcsr, 0
> /* note that the caller must clear SR_FS */
> #endif /* CONFIG_FPU */
> +
> +#ifdef CONFIG_VECTOR
> + csrr t0, CSR_MISA
> + li t1, COMPAT_HWCAP_ISA_V
> + and t0, t0, t1
> + beqz t0, .Lreset_regs_done
> +
> + /*
> + * Clear vector registers and reset vcsr
> + * VLMAX has a defined value, VLEN is a constant,
> + * and this form of vsetvli is defined to set vl to VLMAX.
> + */
> + li t1, SR_VS
> + csrs CSR_STATUS, t1
> + csrs CSR_VCSR, x0
> + vsetvli t1, x0, e8, m8, ta, ma
> + vmv.v.i v0, 0
> + vmv.v.i v8, 0
> + vmv.v.i v16, 0
> + vmv.v.i v24, 0
> + /* note that the caller must clear SR_VS */
> +#endif /* CONFIG_VECTOR */
> +
> .Lreset_regs_done:
Not sure how much they go together, but the #ifdef CONFIG_FPU block above
your new VECTOR block also jumps to the same .Lreset_regs_done, so with
the patch as is the vector-reset block is never reached in the !FPU case.
So maybe making them independent of each other might prevent issues
down the roead.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 07/17] riscv: Add vector struct and assembler definitions
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Paul E. McKenney, Frederic Weisbecker, Vitaly Wool,
Myrtle Shah, Nick Knight, Catalin Marinas, Mark Brown,
Will Deacon, Jiaxun Yang, Janosch Frank, Alexey Dobriyan,
Huacai Chen, Christian Brauner, Andrew Morton, Colin Cross,
Eugene Syromiatnikov, Peter Collingbourne, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Add vector state context struct in struct thread and asm-offsets.c
definitions.
The vector registers will be saved in datap pointer of __riscv_v_state. It
will be dynamically allocated in kernel space. It will be put right after
the __riscv_v_state data structure in user space.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/processor.h | 1 +
arch/riscv/include/uapi/asm/ptrace.h | 17 +++++++++++++++++
arch/riscv/kernel/asm-offsets.c | 6 ++++++
3 files changed, 24 insertions(+)
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index 19eedd4af4cd..95917a2b24f9 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -39,6 +39,7 @@ struct thread_struct {
unsigned long s[12]; /* s[0]: frame pointer */
struct __riscv_d_ext_state fstate;
unsigned long bad_cause;
+ struct __riscv_v_state vstate;
};
/* Whitelist the fstate from the task_struct for hardened usercopy */
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
index 882547f6bd5c..6ee1ca2edfa7 100644
--- a/arch/riscv/include/uapi/asm/ptrace.h
+++ b/arch/riscv/include/uapi/asm/ptrace.h
@@ -77,6 +77,23 @@ union __riscv_fp_state {
struct __riscv_q_ext_state q;
};
+struct __riscv_v_state {
+ unsigned long vstart;
+ unsigned long vl;
+ unsigned long vtype;
+ unsigned long vcsr;
+ void *datap;
+ /*
+ * In signal handler, datap will be set a correct user stack offset
+ * and vector registers will be copied to the address of datap
+ * pointer.
+ *
+ * In ptrace syscall, datap will be set to zero and the vector
+ * registers will be copied to the address right after this
+ * structure.
+ */
+};
+
#endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index df9444397908..37e3e6a8d877 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -75,6 +75,12 @@ void asm_offsets(void)
OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
#endif
+ OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
+ OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
+ OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
+ OFFSET(RISCV_V_STATE_VCSR, __riscv_v_state, vcsr);
+ OFFSET(RISCV_V_STATE_DATAP, __riscv_v_state, datap);
+
DEFINE(PT_SIZE, sizeof(struct pt_regs));
OFFSET(PT_EPC, pt_regs, epc);
OFFSET(PT_RA, pt_regs, ra);
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 07/17] riscv: Add vector struct and assembler definitions
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Paul E. McKenney, Frederic Weisbecker, Vitaly Wool,
Myrtle Shah, Nick Knight, Catalin Marinas, Mark Brown,
Will Deacon, Jiaxun Yang, Janosch Frank, Alexey Dobriyan,
Huacai Chen, Christian Brauner, Andrew Morton, Colin Cross,
Eugene Syromiatnikov, Peter Collingbourne, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Add vector state context struct in struct thread and asm-offsets.c
definitions.
The vector registers will be saved in datap pointer of __riscv_v_state. It
will be dynamically allocated in kernel space. It will be put right after
the __riscv_v_state data structure in user space.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/processor.h | 1 +
arch/riscv/include/uapi/asm/ptrace.h | 17 +++++++++++++++++
arch/riscv/kernel/asm-offsets.c | 6 ++++++
3 files changed, 24 insertions(+)
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index 19eedd4af4cd..95917a2b24f9 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -39,6 +39,7 @@ struct thread_struct {
unsigned long s[12]; /* s[0]: frame pointer */
struct __riscv_d_ext_state fstate;
unsigned long bad_cause;
+ struct __riscv_v_state vstate;
};
/* Whitelist the fstate from the task_struct for hardened usercopy */
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
index 882547f6bd5c..6ee1ca2edfa7 100644
--- a/arch/riscv/include/uapi/asm/ptrace.h
+++ b/arch/riscv/include/uapi/asm/ptrace.h
@@ -77,6 +77,23 @@ union __riscv_fp_state {
struct __riscv_q_ext_state q;
};
+struct __riscv_v_state {
+ unsigned long vstart;
+ unsigned long vl;
+ unsigned long vtype;
+ unsigned long vcsr;
+ void *datap;
+ /*
+ * In signal handler, datap will be set a correct user stack offset
+ * and vector registers will be copied to the address of datap
+ * pointer.
+ *
+ * In ptrace syscall, datap will be set to zero and the vector
+ * registers will be copied to the address right after this
+ * structure.
+ */
+};
+
#endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index df9444397908..37e3e6a8d877 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -75,6 +75,12 @@ void asm_offsets(void)
OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
#endif
+ OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
+ OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
+ OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
+ OFFSET(RISCV_V_STATE_VCSR, __riscv_v_state, vcsr);
+ OFFSET(RISCV_V_STATE_DATAP, __riscv_v_state, datap);
+
DEFINE(PT_SIZE, sizeof(struct pt_regs));
OFFSET(PT_EPC, pt_regs, epc);
OFFSET(PT_RA, pt_regs, ra);
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 07/17] riscv: Add vector struct and assembler definitions
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Mark Rutland, Paul E. McKenney, Frederic Weisbecker, Vitaly Wool,
Myrtle Shah, Nick Knight, Catalin Marinas, Mark Brown,
Will Deacon, Jiaxun Yang, Janosch Frank, Alexey Dobriyan,
Huacai Chen, Christian Brauner, Andrew Morton, Colin Cross,
Eugene Syromiatnikov, Peter Collingbourne, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Add vector state context struct in struct thread and asm-offsets.c
definitions.
The vector registers will be saved in datap pointer of __riscv_v_state. It
will be dynamically allocated in kernel space. It will be put right after
the __riscv_v_state data structure in user space.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/processor.h | 1 +
arch/riscv/include/uapi/asm/ptrace.h | 17 +++++++++++++++++
arch/riscv/kernel/asm-offsets.c | 6 ++++++
3 files changed, 24 insertions(+)
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index 19eedd4af4cd..95917a2b24f9 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -39,6 +39,7 @@ struct thread_struct {
unsigned long s[12]; /* s[0]: frame pointer */
struct __riscv_d_ext_state fstate;
unsigned long bad_cause;
+ struct __riscv_v_state vstate;
};
/* Whitelist the fstate from the task_struct for hardened usercopy */
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
index 882547f6bd5c..6ee1ca2edfa7 100644
--- a/arch/riscv/include/uapi/asm/ptrace.h
+++ b/arch/riscv/include/uapi/asm/ptrace.h
@@ -77,6 +77,23 @@ union __riscv_fp_state {
struct __riscv_q_ext_state q;
};
+struct __riscv_v_state {
+ unsigned long vstart;
+ unsigned long vl;
+ unsigned long vtype;
+ unsigned long vcsr;
+ void *datap;
+ /*
+ * In signal handler, datap will be set a correct user stack offset
+ * and vector registers will be copied to the address of datap
+ * pointer.
+ *
+ * In ptrace syscall, datap will be set to zero and the vector
+ * registers will be copied to the address right after this
+ * structure.
+ */
+};
+
#endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index df9444397908..37e3e6a8d877 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -75,6 +75,12 @@ void asm_offsets(void)
OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
#endif
+ OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
+ OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
+ OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
+ OFFSET(RISCV_V_STATE_VCSR, __riscv_v_state, vcsr);
+ OFFSET(RISCV_V_STATE_DATAP, __riscv_v_state, datap);
+
DEFINE(PT_SIZE, sizeof(struct pt_regs));
OFFSET(PT_EPC, pt_regs, epc);
OFFSET(PT_RA, pt_regs, ra);
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* Re: [PATCH v12 07/17] riscv: Add vector struct and assembler definitions
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-04 5:13 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 5:13 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Arnd Bergmann, Andy Chiu, linux-riscv, lkml
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> Add vector state context struct in struct thread and asm-offsets.c
> definitions.
>
> The vector registers will be saved in datap pointer of __riscv_v_state. It
> will be dynamically allocated in kernel space. It will be put right after
> the __riscv_v_state data structure in user space.
"Vector state includes vector reg file and additional dynamic
configuration CSRs. To handle variable sized reg file context (due to
implementation defined ref size) and to enable lazy-allocation of this,
there's datap which points to appropriate location on user/kernel mode
stack as relevant..."
Something like above.
> diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
> index 19eedd4af4cd..95917a2b24f9 100644
> --- a/arch/riscv/include/asm/processor.h
> +++ b/arch/riscv/include/asm/processor.h
> @@ -39,6 +39,7 @@ struct thread_struct {
> unsigned long s[12]; /* s[0]: frame pointer */
> struct __riscv_d_ext_state fstate;
> unsigned long bad_cause;
> + struct __riscv_v_state vstate;
I think this patch should be preparatory, don't wire up the vstate in
thread_struct now. Only do it when the save/restore calls are wired up
in low level code.
> +struct __riscv_v_state {
> + unsigned long vstart;
> + unsigned long vl;
> + unsigned long vtype;
> + unsigned long vcsr;
> + void *datap;
> + /*
> + * In signal handler, datap will be set a correct user stack offset
> + * and vector registers will be copied to the address of datap
> + * pointer.
> + *
> + * In ptrace syscall, datap will be set to zero and the vector
> + * registers will be copied to the address right after this
> + * structure.
> + */
Nice.
>
> + OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
> + OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
> + OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
> + OFFSET(RISCV_V_STATE_VCSR, __riscv_v_state, vcsr);
> + OFFSET(RISCV_V_STATE_DATAP, __riscv_v_state, datap);
> +
Ok.
Also move the __vstate_{save,restore} functions from patch 5/17 here.
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 07/17] riscv: Add vector struct and assembler definitions
@ 2022-11-04 5:13 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 5:13 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Arnd Bergmann, Andy Chiu, linux-riscv, lkml
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> Add vector state context struct in struct thread and asm-offsets.c
> definitions.
>
> The vector registers will be saved in datap pointer of __riscv_v_state. It
> will be dynamically allocated in kernel space. It will be put right after
> the __riscv_v_state data structure in user space.
"Vector state includes vector reg file and additional dynamic
configuration CSRs. To handle variable sized reg file context (due to
implementation defined ref size) and to enable lazy-allocation of this,
there's datap which points to appropriate location on user/kernel mode
stack as relevant..."
Something like above.
> diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
> index 19eedd4af4cd..95917a2b24f9 100644
> --- a/arch/riscv/include/asm/processor.h
> +++ b/arch/riscv/include/asm/processor.h
> @@ -39,6 +39,7 @@ struct thread_struct {
> unsigned long s[12]; /* s[0]: frame pointer */
> struct __riscv_d_ext_state fstate;
> unsigned long bad_cause;
> + struct __riscv_v_state vstate;
I think this patch should be preparatory, don't wire up the vstate in
thread_struct now. Only do it when the save/restore calls are wired up
in low level code.
> +struct __riscv_v_state {
> + unsigned long vstart;
> + unsigned long vl;
> + unsigned long vtype;
> + unsigned long vcsr;
> + void *datap;
> + /*
> + * In signal handler, datap will be set a correct user stack offset
> + * and vector registers will be copied to the address of datap
> + * pointer.
> + *
> + * In ptrace syscall, datap will be set to zero and the vector
> + * registers will be copied to the address right after this
> + * structure.
> + */
Nice.
>
> + OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
> + OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
> + OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
> + OFFSET(RISCV_V_STATE_VCSR, __riscv_v_state, vcsr);
> + OFFSET(RISCV_V_STATE_DATAP, __riscv_v_state, datap);
> +
Ok.
Also move the __vstate_{save,restore} functions from patch 5/17 here.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 08/17] riscv: Add task switch support for vector
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Andrew Waterman, Nick Knight, Guo Ren,
Vincent Chen, Ruinland Tsai, kernel test robot, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Chris Stillson, Conor Dooley, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Mark Rutland, Frederic Weisbecker,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Will Deacon,
Mark Brown, Jiaxun Yang, Huacai Chen, Janosch Frank,
Alexey Dobriyan, Christian Brauner, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds task switch support for vector. It supports partial lazy
save and restore mechanism. It also supports all lengths of vlen.
[guoren@linux.alibaba.com: First available porting to support vector
context switching]
[nick.knight@sifive.com: Rewrite vector.S to support dynamic vlen, xlen and
code refine]
[vincent.chen@sifive.com: Fix the might_sleep issue in vstate_save,
vstate_restore]
[andrew@sifive.com: Optimize task switch codes of vector]
[ruinland.tsai@sifive.com: Fix the arch_release_task_struct free wrong
datap issue]
Suggested-by: Andrew Waterman <andrew@sifive.com>
Co-developed-by: Nick Knight <nick.knight@sifive.com>
Signed-off-by: Nick Knight <nick.knight@sifive.com>
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Co-developed-by: Ruinland Tsai <ruinland.tsai@sifive.com>
Signed-off-by: Ruinland Tsai <ruinland.tsai@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: kernel test robot <lkp@intel.com>
---
arch/riscv/include/asm/switch_to.h | 66 ++++++++++++++++++++++++++++++
arch/riscv/kernel/Makefile | 1 +
arch/riscv/kernel/process.c | 43 +++++++++++++++++++
3 files changed, 110 insertions(+)
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index df1aa589b7fd..527951c033d4 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -7,11 +7,13 @@
#define _ASM_RISCV_SWITCH_TO_H
#include <linux/jump_label.h>
+#include <linux/slab.h>
#include <linux/sched/task_stack.h>
#include <asm/hwcap.h>
#include <asm/processor.h>
#include <asm/ptrace.h>
#include <asm/csr.h>
+#include <asm/asm-offsets.h>
#ifdef CONFIG_FPU
extern void __fstate_save(struct task_struct *save_to);
@@ -68,6 +70,68 @@ static __always_inline bool has_fpu(void) { return false; }
#define __switch_to_fpu(__prev, __next) do { } while (0)
#endif
+#ifdef CONFIG_VECTOR
+extern struct static_key_false cpu_hwcap_vector;
+static __always_inline bool has_vector(void)
+{
+ return static_branch_likely(&cpu_hwcap_vector);
+}
+extern unsigned long riscv_vsize;
+extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
+extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
+
+static inline void __vstate_clean(struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
+}
+
+static inline void vstate_off(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~SR_VS) | SR_VS_OFF;
+}
+
+static inline void vstate_save(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ if ((regs->status & SR_VS) == SR_VS_DIRTY) {
+ struct __riscv_v_state *vstate = &(task->thread.vstate);
+
+ __vstate_save(vstate, vstate->datap);
+ __vstate_clean(regs);
+ }
+}
+
+static inline void vstate_restore(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ if ((regs->status & SR_VS) != SR_VS_OFF) {
+ struct __riscv_v_state *vstate = &(task->thread.vstate);
+
+ __vstate_restore(vstate, vstate->datap);
+ __vstate_clean(regs);
+ }
+}
+
+static inline void __switch_to_vector(struct task_struct *prev,
+ struct task_struct *next)
+{
+ struct pt_regs *regs;
+
+ regs = task_pt_regs(prev);
+ if (unlikely(regs->status & SR_SD))
+ vstate_save(prev, regs);
+ vstate_restore(next, task_pt_regs(next));
+}
+
+#else
+static __always_inline bool has_vector(void) { return false; }
+#define riscv_vsize (0)
+#define vstate_save(task, regs) do { } while (0)
+#define vstate_restore(task, regs) do { } while (0)
+#define __switch_to_vector(__prev, __next) do { } while (0)
+#endif
+
extern struct task_struct *__switch_to(struct task_struct *,
struct task_struct *);
@@ -77,6 +141,8 @@ do { \
struct task_struct *__next = (next); \
if (has_fpu()) \
__switch_to_fpu(__prev, __next); \
+ if (has_vector()) \
+ __switch_to_vector(__prev, __next); \
((last) = __switch_to(__prev, __next)); \
} while (0)
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 33bb60a354cd..35752fb6d145 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -55,6 +55,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
obj-$(CONFIG_FPU) += fpu.o
+obj-$(CONFIG_VECTOR) += vector.o
obj-$(CONFIG_SMP) += smpboot.o
obj-$(CONFIG_SMP) += smp.o
obj-$(CONFIG_SMP) += cpu_ops.o
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index ceb9ebab6558..e88a37fc77ed 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -124,6 +124,25 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
*/
fstate_restore(current, regs);
}
+
+ if (has_vector()) {
+ struct __riscv_v_state *vstate = &(current->thread.vstate);
+
+ /* Enable vector and allocate memory for vector registers. */
+ if (!vstate->datap) {
+ vstate->datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ if (WARN_ON(!vstate->datap))
+ return;
+ }
+ regs->status |= SR_VS_INITIAL;
+
+ /*
+ * Restore the initial value to the vector register
+ * before starting the user program.
+ */
+ vstate_restore(current, regs);
+ }
+
regs->epc = pc;
regs->sp = sp;
@@ -148,15 +167,29 @@ void flush_thread(void)
fstate_off(current, task_pt_regs(current));
memset(¤t->thread.fstate, 0, sizeof(current->thread.fstate));
#endif
+#ifdef CONFIG_VECTOR
+ /* Reset vector state */
+ vstate_off(current, task_pt_regs(current));
+ memset(¤t->thread.vstate, 0, RISCV_V_STATE_DATAP);
+#endif
}
int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
{
fstate_save(src, task_pt_regs(src));
*dst = *src;
+ dst->thread.vstate.datap = NULL;
+
return 0;
}
+void arch_release_task_struct(struct task_struct *tsk)
+{
+ /* Free the vector context of datap. */
+ if (has_vector() && tsk->thread.vstate.datap)
+ kfree(tsk->thread.vstate.datap);
+}
+
int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
{
unsigned long clone_flags = args->flags;
@@ -175,7 +208,17 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.ra = (unsigned long)ret_from_kernel_thread;
p->thread.s[0] = (unsigned long)args->fn;
p->thread.s[1] = (unsigned long)args->fn_arg;
+ p->thread.vstate.datap = NULL;
} else {
+ /* Allocate the datap for the user process if datap is NULL */
+ if (has_vector() && !p->thread.vstate.datap) {
+ void *datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ /* Failed to allocate memory. */
+ if (!datap)
+ return -ENOMEM;
+ p->thread.vstate.datap = datap;
+ memset(&p->thread.vstate, 0, RISCV_V_STATE_DATAP);
+ }
*childregs = *(current_pt_regs());
if (usp) /* User fork */
childregs->sp = usp;
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 08/17] riscv: Add task switch support for vector
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Andrew Waterman, Nick Knight, Guo Ren,
Vincent Chen, Ruinland Tsai, kernel test robot, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Chris Stillson, Conor Dooley, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Mark Rutland, Frederic Weisbecker,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Will Deacon,
Mark Brown, Jiaxun Yang, Huacai Chen, Janosch Frank,
Alexey Dobriyan, Christian Brauner, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds task switch support for vector. It supports partial lazy
save and restore mechanism. It also supports all lengths of vlen.
[guoren@linux.alibaba.com: First available porting to support vector
context switching]
[nick.knight@sifive.com: Rewrite vector.S to support dynamic vlen, xlen and
code refine]
[vincent.chen@sifive.com: Fix the might_sleep issue in vstate_save,
vstate_restore]
[andrew@sifive.com: Optimize task switch codes of vector]
[ruinland.tsai@sifive.com: Fix the arch_release_task_struct free wrong
datap issue]
Suggested-by: Andrew Waterman <andrew@sifive.com>
Co-developed-by: Nick Knight <nick.knight@sifive.com>
Signed-off-by: Nick Knight <nick.knight@sifive.com>
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Co-developed-by: Ruinland Tsai <ruinland.tsai@sifive.com>
Signed-off-by: Ruinland Tsai <ruinland.tsai@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: kernel test robot <lkp@intel.com>
---
arch/riscv/include/asm/switch_to.h | 66 ++++++++++++++++++++++++++++++
arch/riscv/kernel/Makefile | 1 +
arch/riscv/kernel/process.c | 43 +++++++++++++++++++
3 files changed, 110 insertions(+)
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index df1aa589b7fd..527951c033d4 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -7,11 +7,13 @@
#define _ASM_RISCV_SWITCH_TO_H
#include <linux/jump_label.h>
+#include <linux/slab.h>
#include <linux/sched/task_stack.h>
#include <asm/hwcap.h>
#include <asm/processor.h>
#include <asm/ptrace.h>
#include <asm/csr.h>
+#include <asm/asm-offsets.h>
#ifdef CONFIG_FPU
extern void __fstate_save(struct task_struct *save_to);
@@ -68,6 +70,68 @@ static __always_inline bool has_fpu(void) { return false; }
#define __switch_to_fpu(__prev, __next) do { } while (0)
#endif
+#ifdef CONFIG_VECTOR
+extern struct static_key_false cpu_hwcap_vector;
+static __always_inline bool has_vector(void)
+{
+ return static_branch_likely(&cpu_hwcap_vector);
+}
+extern unsigned long riscv_vsize;
+extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
+extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
+
+static inline void __vstate_clean(struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
+}
+
+static inline void vstate_off(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~SR_VS) | SR_VS_OFF;
+}
+
+static inline void vstate_save(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ if ((regs->status & SR_VS) == SR_VS_DIRTY) {
+ struct __riscv_v_state *vstate = &(task->thread.vstate);
+
+ __vstate_save(vstate, vstate->datap);
+ __vstate_clean(regs);
+ }
+}
+
+static inline void vstate_restore(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ if ((regs->status & SR_VS) != SR_VS_OFF) {
+ struct __riscv_v_state *vstate = &(task->thread.vstate);
+
+ __vstate_restore(vstate, vstate->datap);
+ __vstate_clean(regs);
+ }
+}
+
+static inline void __switch_to_vector(struct task_struct *prev,
+ struct task_struct *next)
+{
+ struct pt_regs *regs;
+
+ regs = task_pt_regs(prev);
+ if (unlikely(regs->status & SR_SD))
+ vstate_save(prev, regs);
+ vstate_restore(next, task_pt_regs(next));
+}
+
+#else
+static __always_inline bool has_vector(void) { return false; }
+#define riscv_vsize (0)
+#define vstate_save(task, regs) do { } while (0)
+#define vstate_restore(task, regs) do { } while (0)
+#define __switch_to_vector(__prev, __next) do { } while (0)
+#endif
+
extern struct task_struct *__switch_to(struct task_struct *,
struct task_struct *);
@@ -77,6 +141,8 @@ do { \
struct task_struct *__next = (next); \
if (has_fpu()) \
__switch_to_fpu(__prev, __next); \
+ if (has_vector()) \
+ __switch_to_vector(__prev, __next); \
((last) = __switch_to(__prev, __next)); \
} while (0)
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 33bb60a354cd..35752fb6d145 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -55,6 +55,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
obj-$(CONFIG_FPU) += fpu.o
+obj-$(CONFIG_VECTOR) += vector.o
obj-$(CONFIG_SMP) += smpboot.o
obj-$(CONFIG_SMP) += smp.o
obj-$(CONFIG_SMP) += cpu_ops.o
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index ceb9ebab6558..e88a37fc77ed 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -124,6 +124,25 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
*/
fstate_restore(current, regs);
}
+
+ if (has_vector()) {
+ struct __riscv_v_state *vstate = &(current->thread.vstate);
+
+ /* Enable vector and allocate memory for vector registers. */
+ if (!vstate->datap) {
+ vstate->datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ if (WARN_ON(!vstate->datap))
+ return;
+ }
+ regs->status |= SR_VS_INITIAL;
+
+ /*
+ * Restore the initial value to the vector register
+ * before starting the user program.
+ */
+ vstate_restore(current, regs);
+ }
+
regs->epc = pc;
regs->sp = sp;
@@ -148,15 +167,29 @@ void flush_thread(void)
fstate_off(current, task_pt_regs(current));
memset(¤t->thread.fstate, 0, sizeof(current->thread.fstate));
#endif
+#ifdef CONFIG_VECTOR
+ /* Reset vector state */
+ vstate_off(current, task_pt_regs(current));
+ memset(¤t->thread.vstate, 0, RISCV_V_STATE_DATAP);
+#endif
}
int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
{
fstate_save(src, task_pt_regs(src));
*dst = *src;
+ dst->thread.vstate.datap = NULL;
+
return 0;
}
+void arch_release_task_struct(struct task_struct *tsk)
+{
+ /* Free the vector context of datap. */
+ if (has_vector() && tsk->thread.vstate.datap)
+ kfree(tsk->thread.vstate.datap);
+}
+
int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
{
unsigned long clone_flags = args->flags;
@@ -175,7 +208,17 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.ra = (unsigned long)ret_from_kernel_thread;
p->thread.s[0] = (unsigned long)args->fn;
p->thread.s[1] = (unsigned long)args->fn_arg;
+ p->thread.vstate.datap = NULL;
} else {
+ /* Allocate the datap for the user process if datap is NULL */
+ if (has_vector() && !p->thread.vstate.datap) {
+ void *datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ /* Failed to allocate memory. */
+ if (!datap)
+ return -ENOMEM;
+ p->thread.vstate.datap = datap;
+ memset(&p->thread.vstate, 0, RISCV_V_STATE_DATAP);
+ }
*childregs = *(current_pt_regs());
if (usp) /* User fork */
childregs->sp = usp;
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 08/17] riscv: Add task switch support for vector
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Andrew Waterman, Nick Knight, Guo Ren,
Vincent Chen, Ruinland Tsai, kernel test robot, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Chris Stillson, Conor Dooley, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Mark Rutland, Frederic Weisbecker,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Will Deacon,
Mark Brown, Jiaxun Yang, Huacai Chen, Janosch Frank,
Alexey Dobriyan, Christian Brauner, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds task switch support for vector. It supports partial lazy
save and restore mechanism. It also supports all lengths of vlen.
[guoren@linux.alibaba.com: First available porting to support vector
context switching]
[nick.knight@sifive.com: Rewrite vector.S to support dynamic vlen, xlen and
code refine]
[vincent.chen@sifive.com: Fix the might_sleep issue in vstate_save,
vstate_restore]
[andrew@sifive.com: Optimize task switch codes of vector]
[ruinland.tsai@sifive.com: Fix the arch_release_task_struct free wrong
datap issue]
Suggested-by: Andrew Waterman <andrew@sifive.com>
Co-developed-by: Nick Knight <nick.knight@sifive.com>
Signed-off-by: Nick Knight <nick.knight@sifive.com>
Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Co-developed-by: Ruinland Tsai <ruinland.tsai@sifive.com>
Signed-off-by: Ruinland Tsai <ruinland.tsai@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: kernel test robot <lkp@intel.com>
---
arch/riscv/include/asm/switch_to.h | 66 ++++++++++++++++++++++++++++++
arch/riscv/kernel/Makefile | 1 +
arch/riscv/kernel/process.c | 43 +++++++++++++++++++
3 files changed, 110 insertions(+)
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index df1aa589b7fd..527951c033d4 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -7,11 +7,13 @@
#define _ASM_RISCV_SWITCH_TO_H
#include <linux/jump_label.h>
+#include <linux/slab.h>
#include <linux/sched/task_stack.h>
#include <asm/hwcap.h>
#include <asm/processor.h>
#include <asm/ptrace.h>
#include <asm/csr.h>
+#include <asm/asm-offsets.h>
#ifdef CONFIG_FPU
extern void __fstate_save(struct task_struct *save_to);
@@ -68,6 +70,68 @@ static __always_inline bool has_fpu(void) { return false; }
#define __switch_to_fpu(__prev, __next) do { } while (0)
#endif
+#ifdef CONFIG_VECTOR
+extern struct static_key_false cpu_hwcap_vector;
+static __always_inline bool has_vector(void)
+{
+ return static_branch_likely(&cpu_hwcap_vector);
+}
+extern unsigned long riscv_vsize;
+extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
+extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
+
+static inline void __vstate_clean(struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
+}
+
+static inline void vstate_off(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~SR_VS) | SR_VS_OFF;
+}
+
+static inline void vstate_save(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ if ((regs->status & SR_VS) == SR_VS_DIRTY) {
+ struct __riscv_v_state *vstate = &(task->thread.vstate);
+
+ __vstate_save(vstate, vstate->datap);
+ __vstate_clean(regs);
+ }
+}
+
+static inline void vstate_restore(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ if ((regs->status & SR_VS) != SR_VS_OFF) {
+ struct __riscv_v_state *vstate = &(task->thread.vstate);
+
+ __vstate_restore(vstate, vstate->datap);
+ __vstate_clean(regs);
+ }
+}
+
+static inline void __switch_to_vector(struct task_struct *prev,
+ struct task_struct *next)
+{
+ struct pt_regs *regs;
+
+ regs = task_pt_regs(prev);
+ if (unlikely(regs->status & SR_SD))
+ vstate_save(prev, regs);
+ vstate_restore(next, task_pt_regs(next));
+}
+
+#else
+static __always_inline bool has_vector(void) { return false; }
+#define riscv_vsize (0)
+#define vstate_save(task, regs) do { } while (0)
+#define vstate_restore(task, regs) do { } while (0)
+#define __switch_to_vector(__prev, __next) do { } while (0)
+#endif
+
extern struct task_struct *__switch_to(struct task_struct *,
struct task_struct *);
@@ -77,6 +141,8 @@ do { \
struct task_struct *__next = (next); \
if (has_fpu()) \
__switch_to_fpu(__prev, __next); \
+ if (has_vector()) \
+ __switch_to_vector(__prev, __next); \
((last) = __switch_to(__prev, __next)); \
} while (0)
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 33bb60a354cd..35752fb6d145 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -55,6 +55,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
obj-$(CONFIG_FPU) += fpu.o
+obj-$(CONFIG_VECTOR) += vector.o
obj-$(CONFIG_SMP) += smpboot.o
obj-$(CONFIG_SMP) += smp.o
obj-$(CONFIG_SMP) += cpu_ops.o
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index ceb9ebab6558..e88a37fc77ed 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -124,6 +124,25 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
*/
fstate_restore(current, regs);
}
+
+ if (has_vector()) {
+ struct __riscv_v_state *vstate = &(current->thread.vstate);
+
+ /* Enable vector and allocate memory for vector registers. */
+ if (!vstate->datap) {
+ vstate->datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ if (WARN_ON(!vstate->datap))
+ return;
+ }
+ regs->status |= SR_VS_INITIAL;
+
+ /*
+ * Restore the initial value to the vector register
+ * before starting the user program.
+ */
+ vstate_restore(current, regs);
+ }
+
regs->epc = pc;
regs->sp = sp;
@@ -148,15 +167,29 @@ void flush_thread(void)
fstate_off(current, task_pt_regs(current));
memset(¤t->thread.fstate, 0, sizeof(current->thread.fstate));
#endif
+#ifdef CONFIG_VECTOR
+ /* Reset vector state */
+ vstate_off(current, task_pt_regs(current));
+ memset(¤t->thread.vstate, 0, RISCV_V_STATE_DATAP);
+#endif
}
int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
{
fstate_save(src, task_pt_regs(src));
*dst = *src;
+ dst->thread.vstate.datap = NULL;
+
return 0;
}
+void arch_release_task_struct(struct task_struct *tsk)
+{
+ /* Free the vector context of datap. */
+ if (has_vector() && tsk->thread.vstate.datap)
+ kfree(tsk->thread.vstate.datap);
+}
+
int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
{
unsigned long clone_flags = args->flags;
@@ -175,7 +208,17 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.ra = (unsigned long)ret_from_kernel_thread;
p->thread.s[0] = (unsigned long)args->fn;
p->thread.s[1] = (unsigned long)args->fn_arg;
+ p->thread.vstate.datap = NULL;
} else {
+ /* Allocate the datap for the user process if datap is NULL */
+ if (has_vector() && !p->thread.vstate.datap) {
+ void *datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ /* Failed to allocate memory. */
+ if (!datap)
+ return -ENOMEM;
+ p->thread.vstate.datap = datap;
+ memset(&p->thread.vstate, 0, RISCV_V_STATE_DATAP);
+ }
*childregs = *(current_pt_regs());
if (usp) /* User fork */
childregs->sp = usp;
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* Re: [PATCH v12 08/17] riscv: Add task switch support for vector
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-04 22:08 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 22:08 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Andrew Waterman, Nick Knight, Guo Ren,
Vincent Chen, Ruinland Tsai, kernel test robot, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Conor Dooley, linux-riscv, lkml, linux-mm, Andy Chiu
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch adds task switch support for vector. It supports partial lazy
> save and restore mechanism. It also supports all lengths of vlen.
>
> [guoren@linux.alibaba.com: First available porting to support vector
> context switching]
> [nick.knight@sifive.com: Rewrite vector.S to support dynamic vlen, xlen and
> code refine]
> [vincent.chen@sifive.com: Fix the might_sleep issue in vstate_save,
> vstate_restore]
> [andrew@sifive.com: Optimize task switch codes of vector]
> [ruinland.tsai@sifive.com: Fix the arch_release_task_struct free wrong
> datap issue]
>
> Suggested-by: Andrew Waterman <andrew@sifive.com>
> Co-developed-by: Nick Knight <nick.knight@sifive.com>
> Signed-off-by: Nick Knight <nick.knight@sifive.com>
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Co-developed-by: Ruinland Tsai <ruinland.tsai@sifive.com>
> Signed-off-by: Ruinland Tsai <ruinland.tsai@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> Reported-by: kernel test robot <lkp@intel.com>
> Reported-by: kernel test robot <lkp@intel.com>
> ---
> arch/riscv/include/asm/switch_to.h | 66 ++++++++++++++++++++++++++++++
> arch/riscv/kernel/Makefile | 1 +
> arch/riscv/kernel/process.c | 43 +++++++++++++++++++
> 3 files changed, 110 insertions(+)
>
> diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
> index df1aa589b7fd..527951c033d4 100644
> --- a/arch/riscv/include/asm/switch_to.h
> +++ b/arch/riscv/include/asm/switch_to.h
> @@ -7,11 +7,13 @@
> #define _ASM_RISCV_SWITCH_TO_H
>
> #include <linux/jump_label.h>
> +#include <linux/slab.h>
> #include <linux/sched/task_stack.h>
> #include <asm/hwcap.h>
> #include <asm/processor.h>
> #include <asm/ptrace.h>
> #include <asm/csr.h>
> +#include <asm/asm-offsets.h>
>
> #ifdef CONFIG_FPU
> extern void __fstate_save(struct task_struct *save_to);
> @@ -68,6 +70,68 @@ static __always_inline bool has_fpu(void) { return false; }
> #define __switch_to_fpu(__prev, __next) do { } while (0)
> #endif
>
> +#ifdef CONFIG_VECTOR
> +extern struct static_key_false cpu_hwcap_vector;
> +static __always_inline bool has_vector(void)
> +{
> + return static_branch_likely(&cpu_hwcap_vector);
> +}
> +extern unsigned long riscv_vsize;
> +extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
> +extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
> +
> +static inline void __vstate_clean(struct pt_regs *regs)
> +{
> + regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
> +}
> +
> +static inline void vstate_off(struct task_struct *task,
> + struct pt_regs *regs)
> +{
> + regs->status = (regs->status & ~SR_VS) | SR_VS_OFF;
> +}
> +
> +static inline void vstate_save(struct task_struct *task,
> + struct pt_regs *regs)
> +{
> + if ((regs->status & SR_VS) == SR_VS_DIRTY) {
> + struct __riscv_v_state *vstate = &(task->thread.vstate);
> +
> + __vstate_save(vstate, vstate->datap);
> + __vstate_clean(regs);
> + }
> +}
> +
> +static inline void vstate_restore(struct task_struct *task,
> + struct pt_regs *regs)
> +{
> + if ((regs->status & SR_VS) != SR_VS_OFF) {
> + struct __riscv_v_state *vstate = &(task->thread.vstate);
> +
> + __vstate_restore(vstate, vstate->datap);
> + __vstate_clean(regs);
> + }
> +}
> +
> +static inline void __switch_to_vector(struct task_struct *prev,
> + struct task_struct *next)
> +{
> + struct pt_regs *regs;
> +
> + regs = task_pt_regs(prev);
> + if (unlikely(regs->status & SR_SD))
> + vstate_save(prev, regs);
> + vstate_restore(next, task_pt_regs(next));
> +}
> +
> +#else
> +static __always_inline bool has_vector(void) { return false; }
> +#define riscv_vsize (0)
> +#define vstate_save(task, regs) do { } while (0)
> +#define vstate_restore(task, regs) do { } while (0)
> +#define __switch_to_vector(__prev, __next) do { } while (0)
> +#endif
All of this needs to be moved into vector.h for better containment.
I would also wire in struct __riscv_v_state vstate in struct
thread_struct in this patch.
> diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
> index 33bb60a354cd..35752fb6d145 100644
> --- a/arch/riscv/kernel/Makefile
> +++ b/arch/riscv/kernel/Makefile
> @@ -55,6 +55,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
>
> obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
> obj-$(CONFIG_FPU) += fpu.o
> +obj-$(CONFIG_VECTOR) += vector.o
This needs to go into last patch which adds Kconfig/Makefile enabling.
> +
> + if (has_vector()) {
Would it make sense to add IS_ENABLED(CONFIG_VECTOR) inside this helper
- would help compiler remove the codegen completely for !VECTOR but
still having some build test coverage. Anyhow this is minor point and
can be added later.
> + struct __riscv_v_state *vstate = &(current->thread.vstate);
> +
> + /* Enable vector and allocate memory for vector registers. */
> + if (!vstate->datap) {
> + vstate->datap = kzalloc(riscv_vsize, GFP_KERNEL);
> + if (WARN_ON(!vstate->datap))
> + return;
> + }
> + regs->status |= SR_VS_INITIAL;
> +
> + /*
> + * Restore the initial value to the vector register
> + * before starting the user program.
> + */
> + vstate_restore(current, regs);
> + }
> +
...
> +#ifdef CONFIG_VECTOR
> + /* Reset vector state */
> + vstate_off(current, task_pt_regs(current));
> + memset(¤t->thread.vstate, 0, RISCV_V_STATE_DATAP);
> +#endif
This doesn't check has_vector() as we want to unconditionally clean
memory for security reasons ?
> }
>
> int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
> {
> fstate_save(src, task_pt_regs(src));
> *dst = *src;
> + dst->thread.vstate.datap = NULL;
has_vector() needed here ?
>
> +void arch_release_task_struct(struct task_struct *tsk)
> +{
> + /* Free the vector context of datap. */
> + if (has_vector() && tsk->thread.vstate.datap)
> + kfree(tsk->thread.vstate.datap);
> +}
> +
> int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
> {
> unsigned long clone_flags = args->flags;
> @@ -175,7 +208,17 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
> p->thread.ra = (unsigned long)ret_from_kernel_thread;
> p->thread.s[0] = (unsigned long)args->fn;
> p->thread.s[1] = (unsigned long)args->fn_arg;
> + p->thread.vstate.datap = NULL;
> } else {
> + /* Allocate the datap for the user process if datap is NULL */
> + if (has_vector() && !p->thread.vstate.datap) {
> + void *datap = kzalloc(riscv_vsize, GFP_KERNEL);
> + /* Failed to allocate memory. */
> + if (!datap)
> + return -ENOMEM;
> + p->thread.vstate.datap = datap;
> + memset(&p->thread.vstate, 0, RISCV_V_STATE_DATAP);
> + }
> *childregs = *(current_pt_regs());
> if (usp) /* User fork */
> childregs->sp = usp;
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 08/17] riscv: Add task switch support for vector
@ 2022-11-04 22:08 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-04 22:08 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Andrew Waterman, Nick Knight, Guo Ren,
Vincent Chen, Ruinland Tsai, kernel test robot, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Guo Ren, Heinrich Schuchardt,
Conor Dooley, linux-riscv, lkml, linux-mm, Andy Chiu
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch adds task switch support for vector. It supports partial lazy
> save and restore mechanism. It also supports all lengths of vlen.
>
> [guoren@linux.alibaba.com: First available porting to support vector
> context switching]
> [nick.knight@sifive.com: Rewrite vector.S to support dynamic vlen, xlen and
> code refine]
> [vincent.chen@sifive.com: Fix the might_sleep issue in vstate_save,
> vstate_restore]
> [andrew@sifive.com: Optimize task switch codes of vector]
> [ruinland.tsai@sifive.com: Fix the arch_release_task_struct free wrong
> datap issue]
>
> Suggested-by: Andrew Waterman <andrew@sifive.com>
> Co-developed-by: Nick Knight <nick.knight@sifive.com>
> Signed-off-by: Nick Knight <nick.knight@sifive.com>
> Co-developed-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Co-developed-by: Ruinland Tsai <ruinland.tsai@sifive.com>
> Signed-off-by: Ruinland Tsai <ruinland.tsai@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> Reported-by: kernel test robot <lkp@intel.com>
> Reported-by: kernel test robot <lkp@intel.com>
> ---
> arch/riscv/include/asm/switch_to.h | 66 ++++++++++++++++++++++++++++++
> arch/riscv/kernel/Makefile | 1 +
> arch/riscv/kernel/process.c | 43 +++++++++++++++++++
> 3 files changed, 110 insertions(+)
>
> diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
> index df1aa589b7fd..527951c033d4 100644
> --- a/arch/riscv/include/asm/switch_to.h
> +++ b/arch/riscv/include/asm/switch_to.h
> @@ -7,11 +7,13 @@
> #define _ASM_RISCV_SWITCH_TO_H
>
> #include <linux/jump_label.h>
> +#include <linux/slab.h>
> #include <linux/sched/task_stack.h>
> #include <asm/hwcap.h>
> #include <asm/processor.h>
> #include <asm/ptrace.h>
> #include <asm/csr.h>
> +#include <asm/asm-offsets.h>
>
> #ifdef CONFIG_FPU
> extern void __fstate_save(struct task_struct *save_to);
> @@ -68,6 +70,68 @@ static __always_inline bool has_fpu(void) { return false; }
> #define __switch_to_fpu(__prev, __next) do { } while (0)
> #endif
>
> +#ifdef CONFIG_VECTOR
> +extern struct static_key_false cpu_hwcap_vector;
> +static __always_inline bool has_vector(void)
> +{
> + return static_branch_likely(&cpu_hwcap_vector);
> +}
> +extern unsigned long riscv_vsize;
> +extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
> +extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
> +
> +static inline void __vstate_clean(struct pt_regs *regs)
> +{
> + regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
> +}
> +
> +static inline void vstate_off(struct task_struct *task,
> + struct pt_regs *regs)
> +{
> + regs->status = (regs->status & ~SR_VS) | SR_VS_OFF;
> +}
> +
> +static inline void vstate_save(struct task_struct *task,
> + struct pt_regs *regs)
> +{
> + if ((regs->status & SR_VS) == SR_VS_DIRTY) {
> + struct __riscv_v_state *vstate = &(task->thread.vstate);
> +
> + __vstate_save(vstate, vstate->datap);
> + __vstate_clean(regs);
> + }
> +}
> +
> +static inline void vstate_restore(struct task_struct *task,
> + struct pt_regs *regs)
> +{
> + if ((regs->status & SR_VS) != SR_VS_OFF) {
> + struct __riscv_v_state *vstate = &(task->thread.vstate);
> +
> + __vstate_restore(vstate, vstate->datap);
> + __vstate_clean(regs);
> + }
> +}
> +
> +static inline void __switch_to_vector(struct task_struct *prev,
> + struct task_struct *next)
> +{
> + struct pt_regs *regs;
> +
> + regs = task_pt_regs(prev);
> + if (unlikely(regs->status & SR_SD))
> + vstate_save(prev, regs);
> + vstate_restore(next, task_pt_regs(next));
> +}
> +
> +#else
> +static __always_inline bool has_vector(void) { return false; }
> +#define riscv_vsize (0)
> +#define vstate_save(task, regs) do { } while (0)
> +#define vstate_restore(task, regs) do { } while (0)
> +#define __switch_to_vector(__prev, __next) do { } while (0)
> +#endif
All of this needs to be moved into vector.h for better containment.
I would also wire in struct __riscv_v_state vstate in struct
thread_struct in this patch.
> diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
> index 33bb60a354cd..35752fb6d145 100644
> --- a/arch/riscv/kernel/Makefile
> +++ b/arch/riscv/kernel/Makefile
> @@ -55,6 +55,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
>
> obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
> obj-$(CONFIG_FPU) += fpu.o
> +obj-$(CONFIG_VECTOR) += vector.o
This needs to go into last patch which adds Kconfig/Makefile enabling.
> +
> + if (has_vector()) {
Would it make sense to add IS_ENABLED(CONFIG_VECTOR) inside this helper
- would help compiler remove the codegen completely for !VECTOR but
still having some build test coverage. Anyhow this is minor point and
can be added later.
> + struct __riscv_v_state *vstate = &(current->thread.vstate);
> +
> + /* Enable vector and allocate memory for vector registers. */
> + if (!vstate->datap) {
> + vstate->datap = kzalloc(riscv_vsize, GFP_KERNEL);
> + if (WARN_ON(!vstate->datap))
> + return;
> + }
> + regs->status |= SR_VS_INITIAL;
> +
> + /*
> + * Restore the initial value to the vector register
> + * before starting the user program.
> + */
> + vstate_restore(current, regs);
> + }
> +
...
> +#ifdef CONFIG_VECTOR
> + /* Reset vector state */
> + vstate_off(current, task_pt_regs(current));
> + memset(¤t->thread.vstate, 0, RISCV_V_STATE_DATAP);
> +#endif
This doesn't check has_vector() as we want to unconditionally clean
memory for security reasons ?
> }
>
> int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
> {
> fstate_save(src, task_pt_regs(src));
> *dst = *src;
> + dst->thread.vstate.datap = NULL;
has_vector() needed here ?
>
> +void arch_release_task_struct(struct task_struct *tsk)
> +{
> + /* Free the vector context of datap. */
> + if (has_vector() && tsk->thread.vstate.datap)
> + kfree(tsk->thread.vstate.datap);
> +}
> +
> int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
> {
> unsigned long clone_flags = args->flags;
> @@ -175,7 +208,17 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
> p->thread.ra = (unsigned long)ret_from_kernel_thread;
> p->thread.s[0] = (unsigned long)args->fn;
> p->thread.s[1] = (unsigned long)args->fn_arg;
> + p->thread.vstate.datap = NULL;
> } else {
> + /* Allocate the datap for the user process if datap is NULL */
> + if (has_vector() && !p->thread.vstate.datap) {
> + void *datap = kzalloc(riscv_vsize, GFP_KERNEL);
> + /* Failed to allocate memory. */
> + if (!datap)
> + return -ENOMEM;
> + p->thread.vstate.datap = datap;
> + memset(&p->thread.vstate, 0, RISCV_V_STATE_DATAP);
> + }
> *childregs = *(current_pt_regs());
> if (usp) /* User fork */
> childregs->sp = usp;
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 09/17] riscv: Add ptrace vector support
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Sunil V L, Nick Knight, Han-Kuan Chen, Li Zhengyu,
Changbin Du, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Paul E. McKenney, Frederic Weisbecker,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Mark Brown,
Will Deacon, Alexey Dobriyan, Huacai Chen, Christian Brauner,
Colin Cross, Peter Collingbourne, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds ptrace support for riscv vector. The vector registers will
be saved in datap pointer of __riscv_v_state. This pointer will be set
right after the __riscv_v_state data structure then it will be put in ubuf
for ptrace system call to get or set. It will check if the datap got from
ubuf is set to the correct address or not when the ptrace system call is
trying to set the vector registers.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/uapi/asm/ptrace.h | 6 +++
arch/riscv/kernel/ptrace.c | 71 ++++++++++++++++++++++++++++
include/uapi/linux/elf.h | 1 +
3 files changed, 78 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
index 6ee1ca2edfa7..2491875be80d 100644
--- a/arch/riscv/include/uapi/asm/ptrace.h
+++ b/arch/riscv/include/uapi/asm/ptrace.h
@@ -94,6 +94,12 @@ struct __riscv_v_state {
*/
};
+/*
+ * According to spec: The number of bits in a single vector register,
+ * VLEN >= ELEN, which must be a power of 2, and must be no greater than
+ * 2^16 = 65536bits = 8192bytes
+ */
+#define RISCV_MAX_VLENB (8192)
#endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */
diff --git a/arch/riscv/kernel/ptrace.c b/arch/riscv/kernel/ptrace.c
index 2ae8280ae475..cce459ff551d 100644
--- a/arch/riscv/kernel/ptrace.c
+++ b/arch/riscv/kernel/ptrace.c
@@ -27,6 +27,9 @@ enum riscv_regset {
#ifdef CONFIG_FPU
REGSET_F,
#endif
+#ifdef CONFIG_VECTOR
+ REGSET_V,
+#endif
};
static int riscv_gpr_get(struct task_struct *target,
@@ -83,6 +86,64 @@ static int riscv_fpr_set(struct task_struct *target,
}
#endif
+#ifdef CONFIG_VECTOR
+static int riscv_vr_get(struct task_struct *target,
+ const struct user_regset *regset,
+ struct membuf to)
+{
+ struct __riscv_v_state *vstate = &target->thread.vstate;
+
+ /*
+ * Ensure the vector registers have been saved to the memory before
+ * copying them to membuf.
+ */
+ if (target == current)
+ vstate_save(current, task_pt_regs(current));
+
+ /* Copy vector header from vstate. */
+ membuf_write(&to, vstate, RISCV_V_STATE_DATAP);
+ membuf_zero(&to, sizeof(void *));
+#if __riscv_xlen == 32
+ membuf_zero(&to, sizeof(__u32));
+#endif
+
+ /* Copy all the vector registers from vstate. */
+ return membuf_write(&to, vstate->datap, riscv_vsize);
+}
+
+static int riscv_vr_set(struct task_struct *target,
+ const struct user_regset *regset,
+ unsigned int pos, unsigned int count,
+ const void *kbuf, const void __user *ubuf)
+{
+ int ret, size;
+ struct __riscv_v_state *vstate = &target->thread.vstate;
+
+ /* Copy rest of the vstate except datap and __padding. */
+ ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate, 0,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(ret))
+ return ret;
+
+ /* Skip copy datap. */
+ size = sizeof(vstate->datap);
+ count -= size;
+ ubuf += size;
+#if __riscv_xlen == 32
+ /* Skip copy _padding. */
+ size = sizeof(vstate->__padding);
+ count -= size;
+ ubuf += size;
+#endif
+
+ /* Copy all the vector registers. */
+ pos = 0;
+ ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate->datap,
+ 0, riscv_vsize);
+ return ret;
+}
+#endif
+
static const struct user_regset riscv_user_regset[] = {
[REGSET_X] = {
.core_note_type = NT_PRSTATUS,
@@ -102,6 +163,16 @@ static const struct user_regset riscv_user_regset[] = {
.set = riscv_fpr_set,
},
#endif
+#ifdef CONFIG_VECTOR
+ [REGSET_V] = {
+ .core_note_type = NT_RISCV_VECTOR,
+ .align = 16,
+ .n = (32 * RISCV_MAX_VLENB)/sizeof(__u32),
+ .size = sizeof(__u32),
+ .regset_get = riscv_vr_get,
+ .set = riscv_vr_set,
+ },
+#endif
};
static const struct user_regset_view riscv_user_native_view = {
diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
index c7b056af9ef0..5a5056c6a2a1 100644
--- a/include/uapi/linux/elf.h
+++ b/include/uapi/linux/elf.h
@@ -439,6 +439,7 @@ typedef struct elf64_shdr {
#define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
#define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
#define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
+#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
#define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
#define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
#define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 09/17] riscv: Add ptrace vector support
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Sunil V L, Nick Knight, Han-Kuan Chen, Li Zhengyu,
Changbin Du, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Paul E. McKenney, Frederic Weisbecker,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Mark Brown,
Will Deacon, Alexey Dobriyan, Huacai Chen, Christian Brauner,
Colin Cross, Peter Collingbourne, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds ptrace support for riscv vector. The vector registers will
be saved in datap pointer of __riscv_v_state. This pointer will be set
right after the __riscv_v_state data structure then it will be put in ubuf
for ptrace system call to get or set. It will check if the datap got from
ubuf is set to the correct address or not when the ptrace system call is
trying to set the vector registers.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/uapi/asm/ptrace.h | 6 +++
arch/riscv/kernel/ptrace.c | 71 ++++++++++++++++++++++++++++
include/uapi/linux/elf.h | 1 +
3 files changed, 78 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
index 6ee1ca2edfa7..2491875be80d 100644
--- a/arch/riscv/include/uapi/asm/ptrace.h
+++ b/arch/riscv/include/uapi/asm/ptrace.h
@@ -94,6 +94,12 @@ struct __riscv_v_state {
*/
};
+/*
+ * According to spec: The number of bits in a single vector register,
+ * VLEN >= ELEN, which must be a power of 2, and must be no greater than
+ * 2^16 = 65536bits = 8192bytes
+ */
+#define RISCV_MAX_VLENB (8192)
#endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */
diff --git a/arch/riscv/kernel/ptrace.c b/arch/riscv/kernel/ptrace.c
index 2ae8280ae475..cce459ff551d 100644
--- a/arch/riscv/kernel/ptrace.c
+++ b/arch/riscv/kernel/ptrace.c
@@ -27,6 +27,9 @@ enum riscv_regset {
#ifdef CONFIG_FPU
REGSET_F,
#endif
+#ifdef CONFIG_VECTOR
+ REGSET_V,
+#endif
};
static int riscv_gpr_get(struct task_struct *target,
@@ -83,6 +86,64 @@ static int riscv_fpr_set(struct task_struct *target,
}
#endif
+#ifdef CONFIG_VECTOR
+static int riscv_vr_get(struct task_struct *target,
+ const struct user_regset *regset,
+ struct membuf to)
+{
+ struct __riscv_v_state *vstate = &target->thread.vstate;
+
+ /*
+ * Ensure the vector registers have been saved to the memory before
+ * copying them to membuf.
+ */
+ if (target == current)
+ vstate_save(current, task_pt_regs(current));
+
+ /* Copy vector header from vstate. */
+ membuf_write(&to, vstate, RISCV_V_STATE_DATAP);
+ membuf_zero(&to, sizeof(void *));
+#if __riscv_xlen == 32
+ membuf_zero(&to, sizeof(__u32));
+#endif
+
+ /* Copy all the vector registers from vstate. */
+ return membuf_write(&to, vstate->datap, riscv_vsize);
+}
+
+static int riscv_vr_set(struct task_struct *target,
+ const struct user_regset *regset,
+ unsigned int pos, unsigned int count,
+ const void *kbuf, const void __user *ubuf)
+{
+ int ret, size;
+ struct __riscv_v_state *vstate = &target->thread.vstate;
+
+ /* Copy rest of the vstate except datap and __padding. */
+ ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate, 0,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(ret))
+ return ret;
+
+ /* Skip copy datap. */
+ size = sizeof(vstate->datap);
+ count -= size;
+ ubuf += size;
+#if __riscv_xlen == 32
+ /* Skip copy _padding. */
+ size = sizeof(vstate->__padding);
+ count -= size;
+ ubuf += size;
+#endif
+
+ /* Copy all the vector registers. */
+ pos = 0;
+ ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate->datap,
+ 0, riscv_vsize);
+ return ret;
+}
+#endif
+
static const struct user_regset riscv_user_regset[] = {
[REGSET_X] = {
.core_note_type = NT_PRSTATUS,
@@ -102,6 +163,16 @@ static const struct user_regset riscv_user_regset[] = {
.set = riscv_fpr_set,
},
#endif
+#ifdef CONFIG_VECTOR
+ [REGSET_V] = {
+ .core_note_type = NT_RISCV_VECTOR,
+ .align = 16,
+ .n = (32 * RISCV_MAX_VLENB)/sizeof(__u32),
+ .size = sizeof(__u32),
+ .regset_get = riscv_vr_get,
+ .set = riscv_vr_set,
+ },
+#endif
};
static const struct user_regset_view riscv_user_native_view = {
diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
index c7b056af9ef0..5a5056c6a2a1 100644
--- a/include/uapi/linux/elf.h
+++ b/include/uapi/linux/elf.h
@@ -439,6 +439,7 @@ typedef struct elf64_shdr {
#define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
#define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
#define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
+#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
#define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
#define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
#define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 09/17] riscv: Add ptrace vector support
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Sunil V L, Nick Knight, Han-Kuan Chen, Li Zhengyu,
Changbin Du, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Paul E. McKenney, Frederic Weisbecker,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Mark Brown,
Will Deacon, Alexey Dobriyan, Huacai Chen, Christian Brauner,
Colin Cross, Peter Collingbourne, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds ptrace support for riscv vector. The vector registers will
be saved in datap pointer of __riscv_v_state. This pointer will be set
right after the __riscv_v_state data structure then it will be put in ubuf
for ptrace system call to get or set. It will check if the datap got from
ubuf is set to the correct address or not when the ptrace system call is
trying to set the vector registers.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/uapi/asm/ptrace.h | 6 +++
arch/riscv/kernel/ptrace.c | 71 ++++++++++++++++++++++++++++
include/uapi/linux/elf.h | 1 +
3 files changed, 78 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
index 6ee1ca2edfa7..2491875be80d 100644
--- a/arch/riscv/include/uapi/asm/ptrace.h
+++ b/arch/riscv/include/uapi/asm/ptrace.h
@@ -94,6 +94,12 @@ struct __riscv_v_state {
*/
};
+/*
+ * According to spec: The number of bits in a single vector register,
+ * VLEN >= ELEN, which must be a power of 2, and must be no greater than
+ * 2^16 = 65536bits = 8192bytes
+ */
+#define RISCV_MAX_VLENB (8192)
#endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */
diff --git a/arch/riscv/kernel/ptrace.c b/arch/riscv/kernel/ptrace.c
index 2ae8280ae475..cce459ff551d 100644
--- a/arch/riscv/kernel/ptrace.c
+++ b/arch/riscv/kernel/ptrace.c
@@ -27,6 +27,9 @@ enum riscv_regset {
#ifdef CONFIG_FPU
REGSET_F,
#endif
+#ifdef CONFIG_VECTOR
+ REGSET_V,
+#endif
};
static int riscv_gpr_get(struct task_struct *target,
@@ -83,6 +86,64 @@ static int riscv_fpr_set(struct task_struct *target,
}
#endif
+#ifdef CONFIG_VECTOR
+static int riscv_vr_get(struct task_struct *target,
+ const struct user_regset *regset,
+ struct membuf to)
+{
+ struct __riscv_v_state *vstate = &target->thread.vstate;
+
+ /*
+ * Ensure the vector registers have been saved to the memory before
+ * copying them to membuf.
+ */
+ if (target == current)
+ vstate_save(current, task_pt_regs(current));
+
+ /* Copy vector header from vstate. */
+ membuf_write(&to, vstate, RISCV_V_STATE_DATAP);
+ membuf_zero(&to, sizeof(void *));
+#if __riscv_xlen == 32
+ membuf_zero(&to, sizeof(__u32));
+#endif
+
+ /* Copy all the vector registers from vstate. */
+ return membuf_write(&to, vstate->datap, riscv_vsize);
+}
+
+static int riscv_vr_set(struct task_struct *target,
+ const struct user_regset *regset,
+ unsigned int pos, unsigned int count,
+ const void *kbuf, const void __user *ubuf)
+{
+ int ret, size;
+ struct __riscv_v_state *vstate = &target->thread.vstate;
+
+ /* Copy rest of the vstate except datap and __padding. */
+ ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate, 0,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(ret))
+ return ret;
+
+ /* Skip copy datap. */
+ size = sizeof(vstate->datap);
+ count -= size;
+ ubuf += size;
+#if __riscv_xlen == 32
+ /* Skip copy _padding. */
+ size = sizeof(vstate->__padding);
+ count -= size;
+ ubuf += size;
+#endif
+
+ /* Copy all the vector registers. */
+ pos = 0;
+ ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate->datap,
+ 0, riscv_vsize);
+ return ret;
+}
+#endif
+
static const struct user_regset riscv_user_regset[] = {
[REGSET_X] = {
.core_note_type = NT_PRSTATUS,
@@ -102,6 +163,16 @@ static const struct user_regset riscv_user_regset[] = {
.set = riscv_fpr_set,
},
#endif
+#ifdef CONFIG_VECTOR
+ [REGSET_V] = {
+ .core_note_type = NT_RISCV_VECTOR,
+ .align = 16,
+ .n = (32 * RISCV_MAX_VLENB)/sizeof(__u32),
+ .size = sizeof(__u32),
+ .regset_get = riscv_vr_get,
+ .set = riscv_vr_set,
+ },
+#endif
};
static const struct user_regset_view riscv_user_native_view = {
diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
index c7b056af9ef0..5a5056c6a2a1 100644
--- a/include/uapi/linux/elf.h
+++ b/include/uapi/linux/elf.h
@@ -439,6 +439,7 @@ typedef struct elf64_shdr {
#define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
#define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
#define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
+#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
#define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
#define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
#define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* Re: [PATCH v12 09/17] riscv: Add ptrace vector support
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-08 1:38 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-08 1:38 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Anup Patel, Atish Patra, Guo Ren, Conor Dooley, Andy Chiu,
linux-riscv, lkml, Björn Töpel, Arnd Bergmann
Trimmed CC list
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch adds ptrace support for riscv vector. The vector registers will
> be saved in datap pointer of __riscv_v_state. This pointer will be set
> right after the __riscv_v_state data structure then it will be put in ubuf
> for ptrace system call to get or set. It will check if the datap got from
> ubuf is set to the correct address or not when the ptrace system call is
> trying to set the vector registers.
>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/uapi/asm/ptrace.h | 6 +++
> arch/riscv/kernel/ptrace.c | 71 ++++++++++++++++++++++++++++
> include/uapi/linux/elf.h | 1 +
> 3 files changed, 78 insertions(+)
>
> diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
> index 6ee1ca2edfa7..2491875be80d 100644
> --- a/arch/riscv/include/uapi/asm/ptrace.h
> +++ b/arch/riscv/include/uapi/asm/ptrace.h
> @@ -94,6 +94,12 @@ struct __riscv_v_state {
> */
> };
>
> +/*
> + * According to spec: The number of bits in a single vector register,
> + * VLEN >= ELEN, which must be a power of 2, and must be no greater than
> + * 2^16 = 65536bits = 8192bytes
> + */
> +#define RISCV_MAX_VLENB (8192)
Need a line here.
> #endif /* __ASSEMBLY__ */
>
> #endif /* _UAPI_ASM_RISCV_PTRACE_H */
> diff --git a/arch/riscv/kernel/ptrace.c b/arch/riscv/kernel/ptrace.c
> index 2ae8280ae475..cce459ff551d 100644
> --- a/arch/riscv/kernel/ptrace.c
> +++ b/arch/riscv/kernel/ptrace.c
> @@ -27,6 +27,9 @@ enum riscv_regset {
> #ifdef CONFIG_FPU
> REGSET_F,
> #endif
> +#ifdef CONFIG_VECTOR
> + REGSET_V,
> +#endif
> };
>
> static int riscv_gpr_get(struct task_struct *target,
> @@ -83,6 +86,64 @@ static int riscv_fpr_set(struct task_struct *target,
> }
> #endif
>
> +#ifdef CONFIG_VECTOR
> +static int riscv_vr_get(struct task_struct *target,
> + const struct user_regset *regset,
> + struct membuf to)
> +{
> + struct __riscv_v_state *vstate = &target->thread.vstate;
> +
> + /*
> + * Ensure the vector registers have been saved to the memory before
> + * copying them to membuf.
> + */
> + if (target == current)
> + vstate_save(current, task_pt_regs(current));
> +
> + /* Copy vector header from vstate. */
> + membuf_write(&to, vstate, RISCV_V_STATE_DATAP);
> + membuf_zero(&to, sizeof(void *));
> +#if __riscv_xlen == 32
> + membuf_zero(&to, sizeof(__u32));
> +#endif
> +
> + /* Copy all the vector registers from vstate. */
> + return membuf_write(&to, vstate->datap, riscv_vsize);
> +}
> +
> +static int riscv_vr_set(struct task_struct *target,
> + const struct user_regset *regset,
> + unsigned int pos, unsigned int count,
> + const void *kbuf, const void __user *ubuf)
> +{
> + int ret, size;
> + struct __riscv_v_state *vstate = &target->thread.vstate;
> +
> + /* Copy rest of the vstate except datap and __padding. */
> + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate, 0,
> + RISCV_V_STATE_DATAP);
> + if (unlikely(ret))
> + return ret;
> +
> + /* Skip copy datap. */
> + size = sizeof(vstate->datap);
> + count -= size;
> + ubuf += size;
> +#if __riscv_xlen == 32
> + /* Skip copy _padding. */
> + size = sizeof(vstate->__padding);
> + count -= size;
> + ubuf += size;
> +#endif
> +
> + /* Copy all the vector registers. */
> + pos = 0;
> + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate->datap,
> + 0, riscv_vsize);
> + return ret;
> +}
> +#endif
> +
> static const struct user_regset riscv_user_regset[] = {
> [REGSET_X] = {
> .core_note_type = NT_PRSTATUS,
> @@ -102,6 +163,16 @@ static const struct user_regset riscv_user_regset[] = {
> .set = riscv_fpr_set,
> },
> #endif
> +#ifdef CONFIG_VECTOR
> + [REGSET_V] = {
> + .core_note_type = NT_RISCV_VECTOR,
> + .align = 16,
> + .n = (32 * RISCV_MAX_VLENB)/sizeof(__u32),
> + .size = sizeof(__u32),
> + .regset_get = riscv_vr_get,
> + .set = riscv_vr_set,
> + },
> +#endif
> };
>
> static const struct user_regset_view riscv_user_native_view = {
> diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
> index c7b056af9ef0..5a5056c6a2a1 100644
> --- a/include/uapi/linux/elf.h
> +++ b/include/uapi/linux/elf.h
> @@ -439,6 +439,7 @@ typedef struct elf64_shdr {
> #define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
> #define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
> #define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
> +#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
> #define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
> #define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
> #define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
I think we should break this one out as a seperate patch to be applied
independently, avoid merge conflicts (but this file doesn't change much
anyways. @Arnd or is it ok to carry with riscv change ?
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 09/17] riscv: Add ptrace vector support
@ 2022-11-08 1:38 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-08 1:38 UTC (permalink / raw)
To: Chris Stillson
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Anup Patel, Atish Patra, Guo Ren, Conor Dooley, Andy Chiu,
linux-riscv, lkml, Björn Töpel, Arnd Bergmann
Trimmed CC list
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch adds ptrace support for riscv vector. The vector registers will
> be saved in datap pointer of __riscv_v_state. This pointer will be set
> right after the __riscv_v_state data structure then it will be put in ubuf
> for ptrace system call to get or set. It will check if the datap got from
> ubuf is set to the correct address or not when the ptrace system call is
> trying to set the vector registers.
>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/uapi/asm/ptrace.h | 6 +++
> arch/riscv/kernel/ptrace.c | 71 ++++++++++++++++++++++++++++
> include/uapi/linux/elf.h | 1 +
> 3 files changed, 78 insertions(+)
>
> diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h
> index 6ee1ca2edfa7..2491875be80d 100644
> --- a/arch/riscv/include/uapi/asm/ptrace.h
> +++ b/arch/riscv/include/uapi/asm/ptrace.h
> @@ -94,6 +94,12 @@ struct __riscv_v_state {
> */
> };
>
> +/*
> + * According to spec: The number of bits in a single vector register,
> + * VLEN >= ELEN, which must be a power of 2, and must be no greater than
> + * 2^16 = 65536bits = 8192bytes
> + */
> +#define RISCV_MAX_VLENB (8192)
Need a line here.
> #endif /* __ASSEMBLY__ */
>
> #endif /* _UAPI_ASM_RISCV_PTRACE_H */
> diff --git a/arch/riscv/kernel/ptrace.c b/arch/riscv/kernel/ptrace.c
> index 2ae8280ae475..cce459ff551d 100644
> --- a/arch/riscv/kernel/ptrace.c
> +++ b/arch/riscv/kernel/ptrace.c
> @@ -27,6 +27,9 @@ enum riscv_regset {
> #ifdef CONFIG_FPU
> REGSET_F,
> #endif
> +#ifdef CONFIG_VECTOR
> + REGSET_V,
> +#endif
> };
>
> static int riscv_gpr_get(struct task_struct *target,
> @@ -83,6 +86,64 @@ static int riscv_fpr_set(struct task_struct *target,
> }
> #endif
>
> +#ifdef CONFIG_VECTOR
> +static int riscv_vr_get(struct task_struct *target,
> + const struct user_regset *regset,
> + struct membuf to)
> +{
> + struct __riscv_v_state *vstate = &target->thread.vstate;
> +
> + /*
> + * Ensure the vector registers have been saved to the memory before
> + * copying them to membuf.
> + */
> + if (target == current)
> + vstate_save(current, task_pt_regs(current));
> +
> + /* Copy vector header from vstate. */
> + membuf_write(&to, vstate, RISCV_V_STATE_DATAP);
> + membuf_zero(&to, sizeof(void *));
> +#if __riscv_xlen == 32
> + membuf_zero(&to, sizeof(__u32));
> +#endif
> +
> + /* Copy all the vector registers from vstate. */
> + return membuf_write(&to, vstate->datap, riscv_vsize);
> +}
> +
> +static int riscv_vr_set(struct task_struct *target,
> + const struct user_regset *regset,
> + unsigned int pos, unsigned int count,
> + const void *kbuf, const void __user *ubuf)
> +{
> + int ret, size;
> + struct __riscv_v_state *vstate = &target->thread.vstate;
> +
> + /* Copy rest of the vstate except datap and __padding. */
> + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate, 0,
> + RISCV_V_STATE_DATAP);
> + if (unlikely(ret))
> + return ret;
> +
> + /* Skip copy datap. */
> + size = sizeof(vstate->datap);
> + count -= size;
> + ubuf += size;
> +#if __riscv_xlen == 32
> + /* Skip copy _padding. */
> + size = sizeof(vstate->__padding);
> + count -= size;
> + ubuf += size;
> +#endif
> +
> + /* Copy all the vector registers. */
> + pos = 0;
> + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vstate->datap,
> + 0, riscv_vsize);
> + return ret;
> +}
> +#endif
> +
> static const struct user_regset riscv_user_regset[] = {
> [REGSET_X] = {
> .core_note_type = NT_PRSTATUS,
> @@ -102,6 +163,16 @@ static const struct user_regset riscv_user_regset[] = {
> .set = riscv_fpr_set,
> },
> #endif
> +#ifdef CONFIG_VECTOR
> + [REGSET_V] = {
> + .core_note_type = NT_RISCV_VECTOR,
> + .align = 16,
> + .n = (32 * RISCV_MAX_VLENB)/sizeof(__u32),
> + .size = sizeof(__u32),
> + .regset_get = riscv_vr_get,
> + .set = riscv_vr_set,
> + },
> +#endif
> };
>
> static const struct user_regset_view riscv_user_native_view = {
> diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
> index c7b056af9ef0..5a5056c6a2a1 100644
> --- a/include/uapi/linux/elf.h
> +++ b/include/uapi/linux/elf.h
> @@ -439,6 +439,7 @@ typedef struct elf64_shdr {
> #define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
> #define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
> #define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
> +#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
> #define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
> #define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
> #define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
I think we should break this one out as a seperate patch to be applied
independently, avoid merge conflicts (but this file doesn't change much
anyways. @Arnd or is it ok to carry with riscv change ?
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 09/17] riscv: Add ptrace vector support
2022-11-08 1:38 ` Vineet Gupta
@ 2022-11-14 20:01 ` Arnd Bergmann
-1 siblings, 0 replies; 147+ messages in thread
From: Arnd Bergmann @ 2022-11-14 20:01 UTC (permalink / raw)
To: Vineet Gupta, Chris Stillson
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Anup Patel, Atish Patra, guoren, Conor.Dooley, Andy Chiu,
linux-riscv, lkml, Björn Töpel
On Tue, Nov 8, 2022, at 02:38, Vineet Gupta wrote:
>> static const struct user_regset_view riscv_user_native_view = {
>> diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
>> index c7b056af9ef0..5a5056c6a2a1 100644
>> --- a/include/uapi/linux/elf.h
>> +++ b/include/uapi/linux/elf.h
>> @@ -439,6 +439,7 @@ typedef struct elf64_shdr {
>> #define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
>> #define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
>> #define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
>> +#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
>> #define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
>> #define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
>> #define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
>
> I think we should break this one out as a seperate patch to be applied
> independently, avoid merge conflicts (but this file doesn't change much
> anyways. @Arnd or is it ok to carry with riscv change ?
I only saw this by accident, sorry for not replying earlier. Yes, please
keep it in this patch and merge it through the riscv tree.
Arnd
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 09/17] riscv: Add ptrace vector support
@ 2022-11-14 20:01 ` Arnd Bergmann
0 siblings, 0 replies; 147+ messages in thread
From: Arnd Bergmann @ 2022-11-14 20:01 UTC (permalink / raw)
To: Vineet Gupta, Chris Stillson
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Anup Patel, Atish Patra, guoren, Conor.Dooley, Andy Chiu,
linux-riscv, lkml, Björn Töpel
On Tue, Nov 8, 2022, at 02:38, Vineet Gupta wrote:
>> static const struct user_regset_view riscv_user_native_view = {
>> diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h
>> index c7b056af9ef0..5a5056c6a2a1 100644
>> --- a/include/uapi/linux/elf.h
>> +++ b/include/uapi/linux/elf.h
>> @@ -439,6 +439,7 @@ typedef struct elf64_shdr {
>> #define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */
>> #define NT_MIPS_FP_MODE 0x801 /* MIPS floating-point mode */
>> #define NT_MIPS_MSA 0x802 /* MIPS SIMD registers */
>> +#define NT_RISCV_VECTOR 0x900 /* RISC-V vector registers */
>> #define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */
>> #define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */
>> #define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
>
> I think we should break this one out as a seperate patch to be applied
> independently, avoid merge conflicts (but this file doesn't change much
> anyways. @Arnd or is it ok to carry with riscv change ?
I only saw this by accident, sorry for not replying earlier. Yes, please
keep it in this patch and merge it through the riscv tree.
Arnd
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 10/17] riscv: Add sigcontext save/restore for vector
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Nick Knight,
Han-Kuan Chen, Changbin Du, Li Zhengyu, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Vitaly Wool, Myrtle Shah, Ruinland Tsai,
Catalin Marinas, Mark Brown, Will Deacon, Alexander Gordeev,
Janosch Frank, Huacai Chen, Alexey Dobriyan, Christian Brauner,
Suren Baghdasaryan, Colin Cross, Peter Collingbourne,
Eugene Syromiatnikov, Andrew Morton, Barret Rhoden,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds sigcontext save/restore for vector. The vector registers
will be saved in datap pointer. The datap pointer will be allocated
dynamically when the task needs in kernel space. The datap pointer will
be set right after the __riscv_v_state data structure to save all the
vector registers in the signal handler stack.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/uapi/asm/sigcontext.h | 24 ++++
arch/riscv/kernel/asm-offsets.c | 2 +
arch/riscv/kernel/signal.c | 165 ++++++++++++++++++++++-
3 files changed, 187 insertions(+), 4 deletions(-)
diff --git a/arch/riscv/include/uapi/asm/sigcontext.h b/arch/riscv/include/uapi/asm/sigcontext.h
index 84f2dfcfdbce..b8a0fd7d7cfc 100644
--- a/arch/riscv/include/uapi/asm/sigcontext.h
+++ b/arch/riscv/include/uapi/asm/sigcontext.h
@@ -8,6 +8,23 @@
#include <asm/ptrace.h>
+/* The Magic number for signal context frame header. */
+#define RVV_MAGIC 0x53465457
+#define END_MAGIC 0x0
+
+/* The size of END signal context header. */
+#define END_HDR_SIZE 0x0
+
+struct __riscv_ctx_hdr {
+ __u32 magic;
+ __u32 size;
+};
+
+struct __sc_riscv_v_state {
+ struct __riscv_ctx_hdr head;
+ struct __riscv_v_state v_state;
+} __attribute__((aligned(16)));
+
/*
* Signal context structure
*
@@ -17,6 +34,13 @@
struct sigcontext {
struct user_regs_struct sc_regs;
union __riscv_fp_state sc_fpregs;
+ /*
+ * 4K + 128 reserved for vector state and future expansion.
+ * This space is enough to store the vector context whose VLENB
+ * is less or equal to 128.
+ * (The size of the vector context is 4144 byte as VLENB is 128)
+ */
+ __u8 __reserved[4224] __attribute__((__aligned__(16)));
};
#endif /* _UAPI_ASM_RISCV_SIGCONTEXT_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index 37e3e6a8d877..80316ef7bb78 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -75,6 +75,8 @@ void asm_offsets(void)
OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
#endif
+ OFFSET(RISCV_V_STATE_MAGIC, __riscv_ctx_hdr, magic);
+ OFFSET(RISCV_V_STATE_SIZE, __riscv_ctx_hdr, size);
OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
index 38b05ca6fe66..41d9a02c7098 100644
--- a/arch/riscv/kernel/signal.c
+++ b/arch/riscv/kernel/signal.c
@@ -20,15 +20,16 @@
#include <asm/csr.h>
extern u32 __user_rt_sigreturn[2];
+static size_t rvv_sc_size;
#define DEBUG_SIG 0
struct rt_sigframe {
struct siginfo info;
- struct ucontext uc;
#ifndef CONFIG_MMU
u32 sigreturn_code[2];
#endif
+ struct ucontext uc;
};
#ifdef CONFIG_FPU
@@ -85,16 +86,155 @@ static long save_fp_state(struct pt_regs *regs,
#define restore_fp_state(task, regs) (0)
#endif
+#ifdef CONFIG_VECTOR
+static long restore_v_state(struct pt_regs *regs, void **sc_reserved_ptr)
+{
+ long err;
+ struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)(*sc_reserved_ptr);
+ void *datap;
+ __u32 magic;
+ __u32 size;
+
+ /* Get magic number and check it. */
+ err = __get_user(magic, &state->head.magic);
+ err = __get_user(size, &state->head.size);
+ if (unlikely(err))
+ return err;
+
+ if (magic != RVV_MAGIC || size != rvv_sc_size)
+ return -EINVAL;
+
+ /* Copy everything of __sc_riscv_v_state except datap. */
+ err = __copy_from_user(¤t->thread.vstate, &state->v_state,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the pointer datap itself. */
+ err = __get_user(datap, &state->v_state.datap);
+ if (unlikely(err))
+ return err;
+
+
+ /* Copy the whole vector content from user space datap. */
+ err = __copy_from_user(current->thread.vstate.datap, datap, riscv_vsize);
+ if (unlikely(err))
+ return err;
+
+ vstate_restore(current, regs);
+
+ /* Move sc_reserved_ptr to point the next signal context frame. */
+ *sc_reserved_ptr += size;
+
+ return err;
+}
+
+static long save_v_state(struct pt_regs *regs, void **sc_reserved_free_ptr)
+{
+ /*
+ * Put __sc_riscv_v_state to the user's signal context space pointed
+ * by sc_reserved_free_ptr and the datap point the address right
+ * after __sc_riscv_v_state.
+ */
+ struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)
+ (*sc_reserved_free_ptr);
+ void *datap = state + 1;
+ long err;
+
+ *sc_reserved_free_ptr += rvv_sc_size;
+
+ err = __put_user(RVV_MAGIC, &state->head.magic);
+ err = __put_user(rvv_sc_size, &state->head.size);
+
+ vstate_save(current, regs);
+ /* Copy everything of vstate but datap. */
+ err = __copy_to_user(&state->v_state, ¤t->thread.vstate,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the pointer datap itself. */
+ err = __put_user(datap, &state->v_state.datap);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the whole vector content to user space datap. */
+ err = __copy_to_user(datap, current->thread.vstate.datap, riscv_vsize);
+
+ return err;
+}
+#else
+#define save_v_state(task, regs) (0)
+#define restore_v_state(task, regs) (0)
+#endif
+
static long restore_sigcontext(struct pt_regs *regs,
struct sigcontext __user *sc)
{
long err;
+ void *sc_reserved_ptr = sc->__reserved;
/* sc_regs is structured the same as the start of pt_regs */
err = __copy_from_user(regs, &sc->sc_regs, sizeof(sc->sc_regs));
/* Restore the floating-point state. */
if (has_fpu())
err |= restore_fp_state(regs, &sc->sc_fpregs);
+
+ while (1 && !err) {
+ __u32 magic, size;
+ struct __riscv_ctx_hdr *head = (struct __riscv_ctx_hdr *)sc_reserved_ptr;
+
+ err |= __get_user(magic, &head->magic);
+ err |= __get_user(size, &head->size);
+ if (err)
+ goto done;
+
+ switch (magic) {
+ case 0:
+ if (size)
+ goto invalid;
+ goto done;
+ case RVV_MAGIC:
+ if (!has_vector())
+ goto invalid;
+ if (size != rvv_sc_size)
+ goto invalid;
+ err |= restore_v_state(regs, &sc_reserved_ptr);
+ break;
+ default:
+ goto invalid;
+ }
+ }
+done:
return err;
+
+invalid:
+ return -EINVAL;
+}
+
+static size_t cal_rt_frame_size(void)
+{
+ struct rt_sigframe __user *frame;
+ static size_t frame_size;
+ size_t total_context_size = 0;
+ size_t sc_reserved_size = sizeof(frame->uc.uc_mcontext.__reserved);
+
+ if (frame_size)
+ goto done;
+
+ frame_size = sizeof(*frame);
+
+ if (has_vector())
+ total_context_size += rvv_sc_size;
+ /* Preserved a __riscv_ctx_hdr for END signal context header. */
+ total_context_size += sizeof(struct __riscv_ctx_hdr);
+
+ if (total_context_size > sc_reserved_size)
+ frame_size += (total_context_size - sc_reserved_size);
+
+ frame_size = round_up(frame_size, 16);
+done:
+ return frame_size;
+
}
SYSCALL_DEFINE0(rt_sigreturn)
@@ -103,13 +243,14 @@ SYSCALL_DEFINE0(rt_sigreturn)
struct rt_sigframe __user *frame;
struct task_struct *task;
sigset_t set;
+ size_t frame_size = cal_rt_frame_size();
/* Always make any pending restarted system calls return -EINTR */
current->restart_block.fn = do_no_restart_syscall;
frame = (struct rt_sigframe __user *)regs->sp;
- if (!access_ok(frame, sizeof(*frame)))
+ if (!access_ok(frame, frame_size))
goto badframe;
if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
@@ -142,11 +283,20 @@ static long setup_sigcontext(struct rt_sigframe __user *frame,
{
struct sigcontext __user *sc = &frame->uc.uc_mcontext;
long err;
+ void *sc_reserved_free_ptr = sc->__reserved;
+
/* sc_regs is structured the same as the start of pt_regs */
err = __copy_to_user(&sc->sc_regs, regs, sizeof(sc->sc_regs));
/* Save the floating-point state. */
if (has_fpu())
err |= save_fp_state(regs, &sc->sc_fpregs);
+ /* Save the vector state. */
+ if (has_vector())
+ err |= save_v_state(regs, &sc_reserved_free_ptr);
+
+ /* Put END __riscv_ctx_hdr at the end. */
+ err = __put_user(END_MAGIC, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->magic);
+ err = __put_user(END_HDR_SIZE, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->size);
return err;
}
@@ -178,9 +328,10 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set,
{
struct rt_sigframe __user *frame;
long err = 0;
+ size_t frame_size = cal_rt_frame_size();
- frame = get_sigframe(ksig, regs, sizeof(*frame));
- if (!access_ok(frame, sizeof(*frame)))
+ frame = get_sigframe(ksig, regs, frame_size);
+ if (!access_ok(frame, frame_size))
return -EFAULT;
err |= copy_siginfo_to_user(&frame->info, &ksig->info);
@@ -326,3 +477,9 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
if (thread_info_flags & _TIF_NOTIFY_RESUME)
resume_user_mode_work(regs);
}
+
+void init_rt_signal_env(void);
+void __init init_rt_signal_env(void)
+{
+ rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
+}
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 10/17] riscv: Add sigcontext save/restore for vector
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Nick Knight,
Han-Kuan Chen, Changbin Du, Li Zhengyu, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Vitaly Wool, Myrtle Shah, Ruinland Tsai,
Catalin Marinas, Mark Brown, Will Deacon, Alexander Gordeev,
Janosch Frank, Huacai Chen, Alexey Dobriyan, Christian Brauner,
Suren Baghdasaryan, Colin Cross, Peter Collingbourne,
Eugene Syromiatnikov, Andrew Morton, Barret Rhoden,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds sigcontext save/restore for vector. The vector registers
will be saved in datap pointer. The datap pointer will be allocated
dynamically when the task needs in kernel space. The datap pointer will
be set right after the __riscv_v_state data structure to save all the
vector registers in the signal handler stack.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/uapi/asm/sigcontext.h | 24 ++++
arch/riscv/kernel/asm-offsets.c | 2 +
arch/riscv/kernel/signal.c | 165 ++++++++++++++++++++++-
3 files changed, 187 insertions(+), 4 deletions(-)
diff --git a/arch/riscv/include/uapi/asm/sigcontext.h b/arch/riscv/include/uapi/asm/sigcontext.h
index 84f2dfcfdbce..b8a0fd7d7cfc 100644
--- a/arch/riscv/include/uapi/asm/sigcontext.h
+++ b/arch/riscv/include/uapi/asm/sigcontext.h
@@ -8,6 +8,23 @@
#include <asm/ptrace.h>
+/* The Magic number for signal context frame header. */
+#define RVV_MAGIC 0x53465457
+#define END_MAGIC 0x0
+
+/* The size of END signal context header. */
+#define END_HDR_SIZE 0x0
+
+struct __riscv_ctx_hdr {
+ __u32 magic;
+ __u32 size;
+};
+
+struct __sc_riscv_v_state {
+ struct __riscv_ctx_hdr head;
+ struct __riscv_v_state v_state;
+} __attribute__((aligned(16)));
+
/*
* Signal context structure
*
@@ -17,6 +34,13 @@
struct sigcontext {
struct user_regs_struct sc_regs;
union __riscv_fp_state sc_fpregs;
+ /*
+ * 4K + 128 reserved for vector state and future expansion.
+ * This space is enough to store the vector context whose VLENB
+ * is less or equal to 128.
+ * (The size of the vector context is 4144 byte as VLENB is 128)
+ */
+ __u8 __reserved[4224] __attribute__((__aligned__(16)));
};
#endif /* _UAPI_ASM_RISCV_SIGCONTEXT_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index 37e3e6a8d877..80316ef7bb78 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -75,6 +75,8 @@ void asm_offsets(void)
OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
#endif
+ OFFSET(RISCV_V_STATE_MAGIC, __riscv_ctx_hdr, magic);
+ OFFSET(RISCV_V_STATE_SIZE, __riscv_ctx_hdr, size);
OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
index 38b05ca6fe66..41d9a02c7098 100644
--- a/arch/riscv/kernel/signal.c
+++ b/arch/riscv/kernel/signal.c
@@ -20,15 +20,16 @@
#include <asm/csr.h>
extern u32 __user_rt_sigreturn[2];
+static size_t rvv_sc_size;
#define DEBUG_SIG 0
struct rt_sigframe {
struct siginfo info;
- struct ucontext uc;
#ifndef CONFIG_MMU
u32 sigreturn_code[2];
#endif
+ struct ucontext uc;
};
#ifdef CONFIG_FPU
@@ -85,16 +86,155 @@ static long save_fp_state(struct pt_regs *regs,
#define restore_fp_state(task, regs) (0)
#endif
+#ifdef CONFIG_VECTOR
+static long restore_v_state(struct pt_regs *regs, void **sc_reserved_ptr)
+{
+ long err;
+ struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)(*sc_reserved_ptr);
+ void *datap;
+ __u32 magic;
+ __u32 size;
+
+ /* Get magic number and check it. */
+ err = __get_user(magic, &state->head.magic);
+ err = __get_user(size, &state->head.size);
+ if (unlikely(err))
+ return err;
+
+ if (magic != RVV_MAGIC || size != rvv_sc_size)
+ return -EINVAL;
+
+ /* Copy everything of __sc_riscv_v_state except datap. */
+ err = __copy_from_user(¤t->thread.vstate, &state->v_state,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the pointer datap itself. */
+ err = __get_user(datap, &state->v_state.datap);
+ if (unlikely(err))
+ return err;
+
+
+ /* Copy the whole vector content from user space datap. */
+ err = __copy_from_user(current->thread.vstate.datap, datap, riscv_vsize);
+ if (unlikely(err))
+ return err;
+
+ vstate_restore(current, regs);
+
+ /* Move sc_reserved_ptr to point the next signal context frame. */
+ *sc_reserved_ptr += size;
+
+ return err;
+}
+
+static long save_v_state(struct pt_regs *regs, void **sc_reserved_free_ptr)
+{
+ /*
+ * Put __sc_riscv_v_state to the user's signal context space pointed
+ * by sc_reserved_free_ptr and the datap point the address right
+ * after __sc_riscv_v_state.
+ */
+ struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)
+ (*sc_reserved_free_ptr);
+ void *datap = state + 1;
+ long err;
+
+ *sc_reserved_free_ptr += rvv_sc_size;
+
+ err = __put_user(RVV_MAGIC, &state->head.magic);
+ err = __put_user(rvv_sc_size, &state->head.size);
+
+ vstate_save(current, regs);
+ /* Copy everything of vstate but datap. */
+ err = __copy_to_user(&state->v_state, ¤t->thread.vstate,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the pointer datap itself. */
+ err = __put_user(datap, &state->v_state.datap);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the whole vector content to user space datap. */
+ err = __copy_to_user(datap, current->thread.vstate.datap, riscv_vsize);
+
+ return err;
+}
+#else
+#define save_v_state(task, regs) (0)
+#define restore_v_state(task, regs) (0)
+#endif
+
static long restore_sigcontext(struct pt_regs *regs,
struct sigcontext __user *sc)
{
long err;
+ void *sc_reserved_ptr = sc->__reserved;
/* sc_regs is structured the same as the start of pt_regs */
err = __copy_from_user(regs, &sc->sc_regs, sizeof(sc->sc_regs));
/* Restore the floating-point state. */
if (has_fpu())
err |= restore_fp_state(regs, &sc->sc_fpregs);
+
+ while (1 && !err) {
+ __u32 magic, size;
+ struct __riscv_ctx_hdr *head = (struct __riscv_ctx_hdr *)sc_reserved_ptr;
+
+ err |= __get_user(magic, &head->magic);
+ err |= __get_user(size, &head->size);
+ if (err)
+ goto done;
+
+ switch (magic) {
+ case 0:
+ if (size)
+ goto invalid;
+ goto done;
+ case RVV_MAGIC:
+ if (!has_vector())
+ goto invalid;
+ if (size != rvv_sc_size)
+ goto invalid;
+ err |= restore_v_state(regs, &sc_reserved_ptr);
+ break;
+ default:
+ goto invalid;
+ }
+ }
+done:
return err;
+
+invalid:
+ return -EINVAL;
+}
+
+static size_t cal_rt_frame_size(void)
+{
+ struct rt_sigframe __user *frame;
+ static size_t frame_size;
+ size_t total_context_size = 0;
+ size_t sc_reserved_size = sizeof(frame->uc.uc_mcontext.__reserved);
+
+ if (frame_size)
+ goto done;
+
+ frame_size = sizeof(*frame);
+
+ if (has_vector())
+ total_context_size += rvv_sc_size;
+ /* Preserved a __riscv_ctx_hdr for END signal context header. */
+ total_context_size += sizeof(struct __riscv_ctx_hdr);
+
+ if (total_context_size > sc_reserved_size)
+ frame_size += (total_context_size - sc_reserved_size);
+
+ frame_size = round_up(frame_size, 16);
+done:
+ return frame_size;
+
}
SYSCALL_DEFINE0(rt_sigreturn)
@@ -103,13 +243,14 @@ SYSCALL_DEFINE0(rt_sigreturn)
struct rt_sigframe __user *frame;
struct task_struct *task;
sigset_t set;
+ size_t frame_size = cal_rt_frame_size();
/* Always make any pending restarted system calls return -EINTR */
current->restart_block.fn = do_no_restart_syscall;
frame = (struct rt_sigframe __user *)regs->sp;
- if (!access_ok(frame, sizeof(*frame)))
+ if (!access_ok(frame, frame_size))
goto badframe;
if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
@@ -142,11 +283,20 @@ static long setup_sigcontext(struct rt_sigframe __user *frame,
{
struct sigcontext __user *sc = &frame->uc.uc_mcontext;
long err;
+ void *sc_reserved_free_ptr = sc->__reserved;
+
/* sc_regs is structured the same as the start of pt_regs */
err = __copy_to_user(&sc->sc_regs, regs, sizeof(sc->sc_regs));
/* Save the floating-point state. */
if (has_fpu())
err |= save_fp_state(regs, &sc->sc_fpregs);
+ /* Save the vector state. */
+ if (has_vector())
+ err |= save_v_state(regs, &sc_reserved_free_ptr);
+
+ /* Put END __riscv_ctx_hdr at the end. */
+ err = __put_user(END_MAGIC, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->magic);
+ err = __put_user(END_HDR_SIZE, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->size);
return err;
}
@@ -178,9 +328,10 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set,
{
struct rt_sigframe __user *frame;
long err = 0;
+ size_t frame_size = cal_rt_frame_size();
- frame = get_sigframe(ksig, regs, sizeof(*frame));
- if (!access_ok(frame, sizeof(*frame)))
+ frame = get_sigframe(ksig, regs, frame_size);
+ if (!access_ok(frame, frame_size))
return -EFAULT;
err |= copy_siginfo_to_user(&frame->info, &ksig->info);
@@ -326,3 +477,9 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
if (thread_info_flags & _TIF_NOTIFY_RESUME)
resume_user_mode_work(regs);
}
+
+void init_rt_signal_env(void);
+void __init init_rt_signal_env(void)
+{
+ rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
+}
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 10/17] riscv: Add sigcontext save/restore for vector
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Heiko Stuebner, Jisheng Zhang, Dao Lu, Sunil V L, Nick Knight,
Han-Kuan Chen, Changbin Du, Li Zhengyu, Ard Biesheuvel,
Tsukasa OI, Yury Norov, Mark Rutland, Nicolas Saenz Julienne,
Frederic Weisbecker, Vitaly Wool, Myrtle Shah, Ruinland Tsai,
Catalin Marinas, Mark Brown, Will Deacon, Alexander Gordeev,
Janosch Frank, Huacai Chen, Alexey Dobriyan, Christian Brauner,
Suren Baghdasaryan, Colin Cross, Peter Collingbourne,
Eugene Syromiatnikov, Andrew Morton, Barret Rhoden,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds sigcontext save/restore for vector. The vector registers
will be saved in datap pointer. The datap pointer will be allocated
dynamically when the task needs in kernel space. The datap pointer will
be set right after the __riscv_v_state data structure to save all the
vector registers in the signal handler stack.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/uapi/asm/sigcontext.h | 24 ++++
arch/riscv/kernel/asm-offsets.c | 2 +
arch/riscv/kernel/signal.c | 165 ++++++++++++++++++++++-
3 files changed, 187 insertions(+), 4 deletions(-)
diff --git a/arch/riscv/include/uapi/asm/sigcontext.h b/arch/riscv/include/uapi/asm/sigcontext.h
index 84f2dfcfdbce..b8a0fd7d7cfc 100644
--- a/arch/riscv/include/uapi/asm/sigcontext.h
+++ b/arch/riscv/include/uapi/asm/sigcontext.h
@@ -8,6 +8,23 @@
#include <asm/ptrace.h>
+/* The Magic number for signal context frame header. */
+#define RVV_MAGIC 0x53465457
+#define END_MAGIC 0x0
+
+/* The size of END signal context header. */
+#define END_HDR_SIZE 0x0
+
+struct __riscv_ctx_hdr {
+ __u32 magic;
+ __u32 size;
+};
+
+struct __sc_riscv_v_state {
+ struct __riscv_ctx_hdr head;
+ struct __riscv_v_state v_state;
+} __attribute__((aligned(16)));
+
/*
* Signal context structure
*
@@ -17,6 +34,13 @@
struct sigcontext {
struct user_regs_struct sc_regs;
union __riscv_fp_state sc_fpregs;
+ /*
+ * 4K + 128 reserved for vector state and future expansion.
+ * This space is enough to store the vector context whose VLENB
+ * is less or equal to 128.
+ * (The size of the vector context is 4144 byte as VLENB is 128)
+ */
+ __u8 __reserved[4224] __attribute__((__aligned__(16)));
};
#endif /* _UAPI_ASM_RISCV_SIGCONTEXT_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index 37e3e6a8d877..80316ef7bb78 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -75,6 +75,8 @@ void asm_offsets(void)
OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
#endif
+ OFFSET(RISCV_V_STATE_MAGIC, __riscv_ctx_hdr, magic);
+ OFFSET(RISCV_V_STATE_SIZE, __riscv_ctx_hdr, size);
OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
index 38b05ca6fe66..41d9a02c7098 100644
--- a/arch/riscv/kernel/signal.c
+++ b/arch/riscv/kernel/signal.c
@@ -20,15 +20,16 @@
#include <asm/csr.h>
extern u32 __user_rt_sigreturn[2];
+static size_t rvv_sc_size;
#define DEBUG_SIG 0
struct rt_sigframe {
struct siginfo info;
- struct ucontext uc;
#ifndef CONFIG_MMU
u32 sigreturn_code[2];
#endif
+ struct ucontext uc;
};
#ifdef CONFIG_FPU
@@ -85,16 +86,155 @@ static long save_fp_state(struct pt_regs *regs,
#define restore_fp_state(task, regs) (0)
#endif
+#ifdef CONFIG_VECTOR
+static long restore_v_state(struct pt_regs *regs, void **sc_reserved_ptr)
+{
+ long err;
+ struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)(*sc_reserved_ptr);
+ void *datap;
+ __u32 magic;
+ __u32 size;
+
+ /* Get magic number and check it. */
+ err = __get_user(magic, &state->head.magic);
+ err = __get_user(size, &state->head.size);
+ if (unlikely(err))
+ return err;
+
+ if (magic != RVV_MAGIC || size != rvv_sc_size)
+ return -EINVAL;
+
+ /* Copy everything of __sc_riscv_v_state except datap. */
+ err = __copy_from_user(¤t->thread.vstate, &state->v_state,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the pointer datap itself. */
+ err = __get_user(datap, &state->v_state.datap);
+ if (unlikely(err))
+ return err;
+
+
+ /* Copy the whole vector content from user space datap. */
+ err = __copy_from_user(current->thread.vstate.datap, datap, riscv_vsize);
+ if (unlikely(err))
+ return err;
+
+ vstate_restore(current, regs);
+
+ /* Move sc_reserved_ptr to point the next signal context frame. */
+ *sc_reserved_ptr += size;
+
+ return err;
+}
+
+static long save_v_state(struct pt_regs *regs, void **sc_reserved_free_ptr)
+{
+ /*
+ * Put __sc_riscv_v_state to the user's signal context space pointed
+ * by sc_reserved_free_ptr and the datap point the address right
+ * after __sc_riscv_v_state.
+ */
+ struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)
+ (*sc_reserved_free_ptr);
+ void *datap = state + 1;
+ long err;
+
+ *sc_reserved_free_ptr += rvv_sc_size;
+
+ err = __put_user(RVV_MAGIC, &state->head.magic);
+ err = __put_user(rvv_sc_size, &state->head.size);
+
+ vstate_save(current, regs);
+ /* Copy everything of vstate but datap. */
+ err = __copy_to_user(&state->v_state, ¤t->thread.vstate,
+ RISCV_V_STATE_DATAP);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the pointer datap itself. */
+ err = __put_user(datap, &state->v_state.datap);
+ if (unlikely(err))
+ return err;
+
+ /* Copy the whole vector content to user space datap. */
+ err = __copy_to_user(datap, current->thread.vstate.datap, riscv_vsize);
+
+ return err;
+}
+#else
+#define save_v_state(task, regs) (0)
+#define restore_v_state(task, regs) (0)
+#endif
+
static long restore_sigcontext(struct pt_regs *regs,
struct sigcontext __user *sc)
{
long err;
+ void *sc_reserved_ptr = sc->__reserved;
/* sc_regs is structured the same as the start of pt_regs */
err = __copy_from_user(regs, &sc->sc_regs, sizeof(sc->sc_regs));
/* Restore the floating-point state. */
if (has_fpu())
err |= restore_fp_state(regs, &sc->sc_fpregs);
+
+ while (1 && !err) {
+ __u32 magic, size;
+ struct __riscv_ctx_hdr *head = (struct __riscv_ctx_hdr *)sc_reserved_ptr;
+
+ err |= __get_user(magic, &head->magic);
+ err |= __get_user(size, &head->size);
+ if (err)
+ goto done;
+
+ switch (magic) {
+ case 0:
+ if (size)
+ goto invalid;
+ goto done;
+ case RVV_MAGIC:
+ if (!has_vector())
+ goto invalid;
+ if (size != rvv_sc_size)
+ goto invalid;
+ err |= restore_v_state(regs, &sc_reserved_ptr);
+ break;
+ default:
+ goto invalid;
+ }
+ }
+done:
return err;
+
+invalid:
+ return -EINVAL;
+}
+
+static size_t cal_rt_frame_size(void)
+{
+ struct rt_sigframe __user *frame;
+ static size_t frame_size;
+ size_t total_context_size = 0;
+ size_t sc_reserved_size = sizeof(frame->uc.uc_mcontext.__reserved);
+
+ if (frame_size)
+ goto done;
+
+ frame_size = sizeof(*frame);
+
+ if (has_vector())
+ total_context_size += rvv_sc_size;
+ /* Preserved a __riscv_ctx_hdr for END signal context header. */
+ total_context_size += sizeof(struct __riscv_ctx_hdr);
+
+ if (total_context_size > sc_reserved_size)
+ frame_size += (total_context_size - sc_reserved_size);
+
+ frame_size = round_up(frame_size, 16);
+done:
+ return frame_size;
+
}
SYSCALL_DEFINE0(rt_sigreturn)
@@ -103,13 +243,14 @@ SYSCALL_DEFINE0(rt_sigreturn)
struct rt_sigframe __user *frame;
struct task_struct *task;
sigset_t set;
+ size_t frame_size = cal_rt_frame_size();
/* Always make any pending restarted system calls return -EINTR */
current->restart_block.fn = do_no_restart_syscall;
frame = (struct rt_sigframe __user *)regs->sp;
- if (!access_ok(frame, sizeof(*frame)))
+ if (!access_ok(frame, frame_size))
goto badframe;
if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
@@ -142,11 +283,20 @@ static long setup_sigcontext(struct rt_sigframe __user *frame,
{
struct sigcontext __user *sc = &frame->uc.uc_mcontext;
long err;
+ void *sc_reserved_free_ptr = sc->__reserved;
+
/* sc_regs is structured the same as the start of pt_regs */
err = __copy_to_user(&sc->sc_regs, regs, sizeof(sc->sc_regs));
/* Save the floating-point state. */
if (has_fpu())
err |= save_fp_state(regs, &sc->sc_fpregs);
+ /* Save the vector state. */
+ if (has_vector())
+ err |= save_v_state(regs, &sc_reserved_free_ptr);
+
+ /* Put END __riscv_ctx_hdr at the end. */
+ err = __put_user(END_MAGIC, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->magic);
+ err = __put_user(END_HDR_SIZE, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->size);
return err;
}
@@ -178,9 +328,10 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set,
{
struct rt_sigframe __user *frame;
long err = 0;
+ size_t frame_size = cal_rt_frame_size();
- frame = get_sigframe(ksig, regs, sizeof(*frame));
- if (!access_ok(frame, sizeof(*frame)))
+ frame = get_sigframe(ksig, regs, frame_size);
+ if (!access_ok(frame, frame_size))
return -EFAULT;
err |= copy_siginfo_to_user(&frame->info, &ksig->info);
@@ -326,3 +477,9 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
if (thread_info_flags & _TIF_NOTIFY_RESUME)
resume_user_mode_work(regs);
}
+
+void init_rt_signal_env(void);
+void __init init_rt_signal_env(void)
+{
+ rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
+}
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* Re: [PATCH v12 10/17] riscv: Add sigcontext save/restore for vector
2022-09-21 21:43 ` Chris Stillson
@ 2022-11-09 1:27 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-09 1:27 UTC (permalink / raw)
To: Chris Stillson, Vincent Chen, Greentime Hu
Cc: Paul Walmsley, Palmer Dabbelt, Anup Patel, Atish Patra, Guo Ren,
Heiko Stuebner, Dao Lu, Conor Dooley, Björn Töpel,
lkml, linux-riscv, Al Viro, linux-arch
+CC linux-arch, Al Viro
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch adds sigcontext save/restore for vector. The vector registers
> will be saved in datap pointer. The datap pointer will be allocated
> dynamically when the task needs in kernel space. The datap pointer will
> be set right after the __riscv_v_state data structure to save all the
> vector registers in the signal handler stack.
>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/uapi/asm/sigcontext.h | 24 ++++
> arch/riscv/kernel/asm-offsets.c | 2 +
> arch/riscv/kernel/signal.c | 165 ++++++++++++++++++++++-
> 3 files changed, 187 insertions(+), 4 deletions(-)
>
> diff --git a/arch/riscv/include/uapi/asm/sigcontext.h b/arch/riscv/include/uapi/asm/sigcontext.h
> index 84f2dfcfdbce..b8a0fd7d7cfc 100644
> --- a/arch/riscv/include/uapi/asm/sigcontext.h
> +++ b/arch/riscv/include/uapi/asm/sigcontext.h
> @@ -8,6 +8,23 @@
>
> #include <asm/ptrace.h>
>
> +/* The Magic number for signal context frame header. */
> +#define RVV_MAGIC 0x53465457
> +#define END_MAGIC 0x0
> +
> +/* The size of END signal context header. */
> +#define END_HDR_SIZE 0x0
> +
> +struct __riscv_ctx_hdr {
> + __u32 magic;
> + __u32 size;
> +};
> +
> +struct __sc_riscv_v_state {
> + struct __riscv_ctx_hdr head;
> + struct __riscv_v_state v_state;
> +} __attribute__((aligned(16)));
> +
> /*
> * Signal context structure
> *
> @@ -17,6 +34,13 @@
> struct sigcontext {
> struct user_regs_struct sc_regs;
> union __riscv_fp_state sc_fpregs;
> + /*
> + * 4K + 128 reserved for vector state and future expansion.
> + * This space is enough to store the vector context whose VLENB
> + * is less or equal to 128.
> + * (The size of the vector context is 4144 byte as VLENB is 128)
At first glace it seems this only supports 128 byte V regs. Better to
add some words saying that wider reg file is handled in code too.
> + */
> + __u8 __reserved[4224] __attribute__((__aligned__(16)));
Is alignment of 8 not enough or is this future-proofing.
> diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
> index 37e3e6a8d877..80316ef7bb78 100644
> --- a/arch/riscv/kernel/asm-offsets.c
> +++ b/arch/riscv/kernel/asm-offsets.c
> @@ -75,6 +75,8 @@ void asm_offsets(void)
> OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
> #endif
>
> + OFFSET(RISCV_V_STATE_MAGIC, __riscv_ctx_hdr, magic);
> + OFFSET(RISCV_V_STATE_SIZE, __riscv_ctx_hdr, size);
> OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
> OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
> OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
> diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
> index 38b05ca6fe66..41d9a02c7098 100644
> --- a/arch/riscv/kernel/signal.c
> +++ b/arch/riscv/kernel/signal.c
> @@ -20,15 +20,16 @@
> #include <asm/csr.h>
>
> extern u32 __user_rt_sigreturn[2];
> +static size_t rvv_sc_size;
>
> #define DEBUG_SIG 0
>
> struct rt_sigframe {
> struct siginfo info;
> - struct ucontext uc;
> #ifndef CONFIG_MMU
> u32 sigreturn_code[2];
> #endif
> + struct ucontext uc;
> };
Just for the record this is NOT a userspace abi change (for SA_SIGINFO
signal handlers in userspace) since they are only provided struct
ucontext pointer. kernel is free to rearrange stuff in struct
rt_sigframe as it deems fits.
>
> #ifdef CONFIG_FPU
> @@ -85,16 +86,155 @@ static long save_fp_state(struct pt_regs *regs,
> #define restore_fp_state(task, regs) (0)
> #endif
>
> +#ifdef CONFIG_VECTOR
> +static long restore_v_state(struct pt_regs *regs, void **sc_reserved_ptr)
For ease of reading I would move save before restore.
> +{
> + long err;
> + struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)(*sc_reserved_ptr);
> + void *datap;
> + __u32 magic;
> + __u32 size;
> +
> + /* Get magic number and check it. */
> + err = __get_user(magic, &state->head.magic);
> + err = __get_user(size, &state->head.size);
> + if (unlikely(err))
> + return err;
> +
> + if (magic != RVV_MAGIC || size != rvv_sc_size)
> + return -EINVAL;
> +
> + /* Copy everything of __sc_riscv_v_state except datap. */
> + err = __copy_from_user(¤t->thread.vstate, &state->v_state,
> + RISCV_V_STATE_DATAP);
> + if (unlikely(err))
> + return err;
> +
> + /* Copy the pointer datap itself. */
> + err = __get_user(datap, &state->v_state.datap);
> + if (unlikely(err))
> + return err;
> +
> +
> + /* Copy the whole vector content from user space datap. */
> + err = __copy_from_user(current->thread.vstate.datap, datap, riscv_vsize);
> + if (unlikely(err))
> + return err;
> +
> + vstate_restore(current, regs);
> +
> + /* Move sc_reserved_ptr to point the next signal context frame. */
> + *sc_reserved_ptr += size;
> +
> + return err;
> +}
> +
> +static long save_v_state(struct pt_regs *regs, void **sc_reserved_free_ptr)
> +{
> + /*
> + * Put __sc_riscv_v_state to the user's signal context space pointed
> + * by sc_reserved_free_ptr and the datap point the address right
> + * after __sc_riscv_v_state.
> + */
> + struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)
> + (*sc_reserved_free_ptr);
> + void *datap = state + 1;
> + long err;
> +
> + *sc_reserved_free_ptr += rvv_sc_size;
> +
> + err = __put_user(RVV_MAGIC, &state->head.magic);
> + err = __put_user(rvv_sc_size, &state->head.size);
Can we copy these markers *after* the actual context is succesfully
copied. What if it fails (user stack can't grow anymore etc), then we
leave this half cooked state on stack.
Granted the process would most likely be killed anyways in such case.
> +
> + vstate_save(current, regs);
> + /* Copy everything of vstate but datap. */
> + err = __copy_to_user(&state->v_state, ¤t->thread.vstate,
> + RISCV_V_STATE_DATAP);
> + if (unlikely(err))
> + return err;
> +
> + /* Copy the pointer datap itself. */
> + err = __put_user(datap, &state->v_state.datap);
> + if (unlikely(err))
> + return err;
> +
> + /* Copy the whole vector content to user space datap. */
> + err = __copy_to_user(datap, current->thread.vstate.datap, riscv_vsize);
> +
> + return err;
> +}
> +#else
> +#define save_v_state(task, regs) (0)
> +#define restore_v_state(task, regs) (0)
> +#endif
> +
> static long restore_sigcontext(struct pt_regs *regs,
> struct sigcontext __user *sc)
> {
> long err;
> + void *sc_reserved_ptr = sc->__reserved;
> /* sc_regs is structured the same as the start of pt_regs */
> err = __copy_from_user(regs, &sc->sc_regs, sizeof(sc->sc_regs));
> /* Restore the floating-point state. */
> if (has_fpu())
> err |= restore_fp_state(regs, &sc->sc_fpregs);
> +
> + while (1 && !err) {
> + __u32 magic, size;
> + struct __riscv_ctx_hdr *head = (struct __riscv_ctx_hdr *)sc_reserved_ptr;
> +
> + err |= __get_user(magic, &head->magic);
> + err |= __get_user(size, &head->size);
> + if (err)
> + goto done;
> +
> + switch (magic) {
> + case 0:
> + if (size)
> + goto invalid;
> + goto done;
> + case RVV_MAGIC:
> + if (!has_vector())
> + goto invalid;
> + if (size != rvv_sc_size)
> + goto invalid;
> + err |= restore_v_state(regs, &sc_reserved_ptr);
> + break;
See question below. Is this 2 pass header check due to Vector or for
some future extension ?
> + default:
> + goto invalid;
> + }
> + }
> +done:
> return err;
> +
> +invalid:
> + return -EINVAL;
> +}
> +
> +static size_t cal_rt_frame_size(void)
> +{
> + struct rt_sigframe __user *frame;
> + static size_t frame_size;
> + size_t total_context_size = 0;
> + size_t sc_reserved_size = sizeof(frame->uc.uc_mcontext.__reserved);
Perhaps nit-picking, but "sc_reserved_size" and such names are
confusing. Its ok to call the mcontext field __reserved but in rest of
code lets avoid reserved - plain sctxt or some such will suffice.
> +
> + if (frame_size)
> + goto done;
> +
> + frame_size = sizeof(*frame);
> +
> + if (has_vector())
> + total_context_size += rvv_sc_size;
> + /* Preserved a __riscv_ctx_hdr for END signal context header. */
> + total_context_size += sizeof(struct __riscv_ctx_hdr);
> +
> + if (total_context_size > sc_reserved_size)
> + frame_size += (total_context_size - sc_reserved_size);
> +
> + frame_size = round_up(frame_size, 16);
> +done:
> + return frame_size;
This seems to be a one time computation, given we are checking a static
frame_size everytime, could we just calculate this once in
init_rt_signal_env() below in a now file scoped static variable ?
> +
> }
>
> SYSCALL_DEFINE0(rt_sigreturn)
> @@ -103,13 +243,14 @@ SYSCALL_DEFINE0(rt_sigreturn)
> struct rt_sigframe __user *frame;
> struct task_struct *task;
> sigset_t set;
> + size_t frame_size = cal_rt_frame_size();
>
> /* Always make any pending restarted system calls return -EINTR */
> current->restart_block.fn = do_no_restart_syscall;
>
> frame = (struct rt_sigframe __user *)regs->sp;
>
> - if (!access_ok(frame, sizeof(*frame)))
> + if (!access_ok(frame, frame_size))
> goto badframe;
>
> if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
> @@ -142,11 +283,20 @@ static long setup_sigcontext(struct rt_sigframe __user *frame,
> {
> struct sigcontext __user *sc = &frame->uc.uc_mcontext;
> long err;
> + void *sc_reserved_free_ptr = sc->__reserved;
> +
> /* sc_regs is structured the same as the start of pt_regs */
> err = __copy_to_user(&sc->sc_regs, regs, sizeof(sc->sc_regs));
> /* Save the floating-point state. */
> if (has_fpu())
> err |= save_fp_state(regs, &sc->sc_fpregs);
> + /* Save the vector state. */
> + if (has_vector())
> + err |= save_v_state(regs, &sc_reserved_free_ptr);
> +
> + /* Put END __riscv_ctx_hdr at the end. */
> + err = __put_user(END_MAGIC, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->magic);
> + err = __put_user(END_HDR_SIZE, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->size);
I really don't understand the need for this trailing header stuff. Is
this due to variable sized vector or enginnering for a yet to be
invented future extension's state.
But if so, won't we add a new ctx_hdr + magic for "foo" when that
happens. After all we were not doing this for existing FP stuff, granted
FP is all fixed size.
For Vector: magic+size tuple should be enough to identify/locate the ctx
info.
> return err;
> }
>
> @@ -178,9 +328,10 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set,
> {
> struct rt_sigframe __user *frame;
> long err = 0;
> + size_t frame_size = cal_rt_frame_size();
>
> - frame = get_sigframe(ksig, regs, sizeof(*frame));
> - if (!access_ok(frame, sizeof(*frame)))
> + frame = get_sigframe(ksig, regs, frame_size);
> + if (!access_ok(frame, frame_size))
> return -EFAULT;
>
> err |= copy_siginfo_to_user(&frame->info, &ksig->info);
> @@ -326,3 +477,9 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
> if (thread_info_flags & _TIF_NOTIFY_RESUME)
> resume_user_mode_work(regs);
> }
> +
> +void init_rt_signal_env(void);
> +void __init init_rt_signal_env(void)
> +{
> + rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
See above, should be precompute the output of cal_rt_frame_size() here too.
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: [PATCH v12 10/17] riscv: Add sigcontext save/restore for vector
@ 2022-11-09 1:27 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-11-09 1:27 UTC (permalink / raw)
To: Chris Stillson, Vincent Chen, Greentime Hu
Cc: Paul Walmsley, Palmer Dabbelt, Anup Patel, Atish Patra, Guo Ren,
Heiko Stuebner, Dao Lu, Conor Dooley, Björn Töpel,
lkml, linux-riscv, Al Viro, linux-arch
+CC linux-arch, Al Viro
On 9/21/22 14:43, Chris Stillson wrote:
> From: Greentime Hu <greentime.hu@sifive.com>
>
> This patch adds sigcontext save/restore for vector. The vector registers
> will be saved in datap pointer. The datap pointer will be allocated
> dynamically when the task needs in kernel space. The datap pointer will
> be set right after the __riscv_v_state data structure to save all the
> vector registers in the signal handler stack.
>
> Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> ---
> arch/riscv/include/uapi/asm/sigcontext.h | 24 ++++
> arch/riscv/kernel/asm-offsets.c | 2 +
> arch/riscv/kernel/signal.c | 165 ++++++++++++++++++++++-
> 3 files changed, 187 insertions(+), 4 deletions(-)
>
> diff --git a/arch/riscv/include/uapi/asm/sigcontext.h b/arch/riscv/include/uapi/asm/sigcontext.h
> index 84f2dfcfdbce..b8a0fd7d7cfc 100644
> --- a/arch/riscv/include/uapi/asm/sigcontext.h
> +++ b/arch/riscv/include/uapi/asm/sigcontext.h
> @@ -8,6 +8,23 @@
>
> #include <asm/ptrace.h>
>
> +/* The Magic number for signal context frame header. */
> +#define RVV_MAGIC 0x53465457
> +#define END_MAGIC 0x0
> +
> +/* The size of END signal context header. */
> +#define END_HDR_SIZE 0x0
> +
> +struct __riscv_ctx_hdr {
> + __u32 magic;
> + __u32 size;
> +};
> +
> +struct __sc_riscv_v_state {
> + struct __riscv_ctx_hdr head;
> + struct __riscv_v_state v_state;
> +} __attribute__((aligned(16)));
> +
> /*
> * Signal context structure
> *
> @@ -17,6 +34,13 @@
> struct sigcontext {
> struct user_regs_struct sc_regs;
> union __riscv_fp_state sc_fpregs;
> + /*
> + * 4K + 128 reserved for vector state and future expansion.
> + * This space is enough to store the vector context whose VLENB
> + * is less or equal to 128.
> + * (The size of the vector context is 4144 byte as VLENB is 128)
At first glace it seems this only supports 128 byte V regs. Better to
add some words saying that wider reg file is handled in code too.
> + */
> + __u8 __reserved[4224] __attribute__((__aligned__(16)));
Is alignment of 8 not enough or is this future-proofing.
> diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
> index 37e3e6a8d877..80316ef7bb78 100644
> --- a/arch/riscv/kernel/asm-offsets.c
> +++ b/arch/riscv/kernel/asm-offsets.c
> @@ -75,6 +75,8 @@ void asm_offsets(void)
> OFFSET(TSK_STACK_CANARY, task_struct, stack_canary);
> #endif
>
> + OFFSET(RISCV_V_STATE_MAGIC, __riscv_ctx_hdr, magic);
> + OFFSET(RISCV_V_STATE_SIZE, __riscv_ctx_hdr, size);
> OFFSET(RISCV_V_STATE_VSTART, __riscv_v_state, vstart);
> OFFSET(RISCV_V_STATE_VL, __riscv_v_state, vl);
> OFFSET(RISCV_V_STATE_VTYPE, __riscv_v_state, vtype);
> diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
> index 38b05ca6fe66..41d9a02c7098 100644
> --- a/arch/riscv/kernel/signal.c
> +++ b/arch/riscv/kernel/signal.c
> @@ -20,15 +20,16 @@
> #include <asm/csr.h>
>
> extern u32 __user_rt_sigreturn[2];
> +static size_t rvv_sc_size;
>
> #define DEBUG_SIG 0
>
> struct rt_sigframe {
> struct siginfo info;
> - struct ucontext uc;
> #ifndef CONFIG_MMU
> u32 sigreturn_code[2];
> #endif
> + struct ucontext uc;
> };
Just for the record this is NOT a userspace abi change (for SA_SIGINFO
signal handlers in userspace) since they are only provided struct
ucontext pointer. kernel is free to rearrange stuff in struct
rt_sigframe as it deems fits.
>
> #ifdef CONFIG_FPU
> @@ -85,16 +86,155 @@ static long save_fp_state(struct pt_regs *regs,
> #define restore_fp_state(task, regs) (0)
> #endif
>
> +#ifdef CONFIG_VECTOR
> +static long restore_v_state(struct pt_regs *regs, void **sc_reserved_ptr)
For ease of reading I would move save before restore.
> +{
> + long err;
> + struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)(*sc_reserved_ptr);
> + void *datap;
> + __u32 magic;
> + __u32 size;
> +
> + /* Get magic number and check it. */
> + err = __get_user(magic, &state->head.magic);
> + err = __get_user(size, &state->head.size);
> + if (unlikely(err))
> + return err;
> +
> + if (magic != RVV_MAGIC || size != rvv_sc_size)
> + return -EINVAL;
> +
> + /* Copy everything of __sc_riscv_v_state except datap. */
> + err = __copy_from_user(¤t->thread.vstate, &state->v_state,
> + RISCV_V_STATE_DATAP);
> + if (unlikely(err))
> + return err;
> +
> + /* Copy the pointer datap itself. */
> + err = __get_user(datap, &state->v_state.datap);
> + if (unlikely(err))
> + return err;
> +
> +
> + /* Copy the whole vector content from user space datap. */
> + err = __copy_from_user(current->thread.vstate.datap, datap, riscv_vsize);
> + if (unlikely(err))
> + return err;
> +
> + vstate_restore(current, regs);
> +
> + /* Move sc_reserved_ptr to point the next signal context frame. */
> + *sc_reserved_ptr += size;
> +
> + return err;
> +}
> +
> +static long save_v_state(struct pt_regs *regs, void **sc_reserved_free_ptr)
> +{
> + /*
> + * Put __sc_riscv_v_state to the user's signal context space pointed
> + * by sc_reserved_free_ptr and the datap point the address right
> + * after __sc_riscv_v_state.
> + */
> + struct __sc_riscv_v_state __user *state = (struct __sc_riscv_v_state *)
> + (*sc_reserved_free_ptr);
> + void *datap = state + 1;
> + long err;
> +
> + *sc_reserved_free_ptr += rvv_sc_size;
> +
> + err = __put_user(RVV_MAGIC, &state->head.magic);
> + err = __put_user(rvv_sc_size, &state->head.size);
Can we copy these markers *after* the actual context is succesfully
copied. What if it fails (user stack can't grow anymore etc), then we
leave this half cooked state on stack.
Granted the process would most likely be killed anyways in such case.
> +
> + vstate_save(current, regs);
> + /* Copy everything of vstate but datap. */
> + err = __copy_to_user(&state->v_state, ¤t->thread.vstate,
> + RISCV_V_STATE_DATAP);
> + if (unlikely(err))
> + return err;
> +
> + /* Copy the pointer datap itself. */
> + err = __put_user(datap, &state->v_state.datap);
> + if (unlikely(err))
> + return err;
> +
> + /* Copy the whole vector content to user space datap. */
> + err = __copy_to_user(datap, current->thread.vstate.datap, riscv_vsize);
> +
> + return err;
> +}
> +#else
> +#define save_v_state(task, regs) (0)
> +#define restore_v_state(task, regs) (0)
> +#endif
> +
> static long restore_sigcontext(struct pt_regs *regs,
> struct sigcontext __user *sc)
> {
> long err;
> + void *sc_reserved_ptr = sc->__reserved;
> /* sc_regs is structured the same as the start of pt_regs */
> err = __copy_from_user(regs, &sc->sc_regs, sizeof(sc->sc_regs));
> /* Restore the floating-point state. */
> if (has_fpu())
> err |= restore_fp_state(regs, &sc->sc_fpregs);
> +
> + while (1 && !err) {
> + __u32 magic, size;
> + struct __riscv_ctx_hdr *head = (struct __riscv_ctx_hdr *)sc_reserved_ptr;
> +
> + err |= __get_user(magic, &head->magic);
> + err |= __get_user(size, &head->size);
> + if (err)
> + goto done;
> +
> + switch (magic) {
> + case 0:
> + if (size)
> + goto invalid;
> + goto done;
> + case RVV_MAGIC:
> + if (!has_vector())
> + goto invalid;
> + if (size != rvv_sc_size)
> + goto invalid;
> + err |= restore_v_state(regs, &sc_reserved_ptr);
> + break;
See question below. Is this 2 pass header check due to Vector or for
some future extension ?
> + default:
> + goto invalid;
> + }
> + }
> +done:
> return err;
> +
> +invalid:
> + return -EINVAL;
> +}
> +
> +static size_t cal_rt_frame_size(void)
> +{
> + struct rt_sigframe __user *frame;
> + static size_t frame_size;
> + size_t total_context_size = 0;
> + size_t sc_reserved_size = sizeof(frame->uc.uc_mcontext.__reserved);
Perhaps nit-picking, but "sc_reserved_size" and such names are
confusing. Its ok to call the mcontext field __reserved but in rest of
code lets avoid reserved - plain sctxt or some such will suffice.
> +
> + if (frame_size)
> + goto done;
> +
> + frame_size = sizeof(*frame);
> +
> + if (has_vector())
> + total_context_size += rvv_sc_size;
> + /* Preserved a __riscv_ctx_hdr for END signal context header. */
> + total_context_size += sizeof(struct __riscv_ctx_hdr);
> +
> + if (total_context_size > sc_reserved_size)
> + frame_size += (total_context_size - sc_reserved_size);
> +
> + frame_size = round_up(frame_size, 16);
> +done:
> + return frame_size;
This seems to be a one time computation, given we are checking a static
frame_size everytime, could we just calculate this once in
init_rt_signal_env() below in a now file scoped static variable ?
> +
> }
>
> SYSCALL_DEFINE0(rt_sigreturn)
> @@ -103,13 +243,14 @@ SYSCALL_DEFINE0(rt_sigreturn)
> struct rt_sigframe __user *frame;
> struct task_struct *task;
> sigset_t set;
> + size_t frame_size = cal_rt_frame_size();
>
> /* Always make any pending restarted system calls return -EINTR */
> current->restart_block.fn = do_no_restart_syscall;
>
> frame = (struct rt_sigframe __user *)regs->sp;
>
> - if (!access_ok(frame, sizeof(*frame)))
> + if (!access_ok(frame, frame_size))
> goto badframe;
>
> if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
> @@ -142,11 +283,20 @@ static long setup_sigcontext(struct rt_sigframe __user *frame,
> {
> struct sigcontext __user *sc = &frame->uc.uc_mcontext;
> long err;
> + void *sc_reserved_free_ptr = sc->__reserved;
> +
> /* sc_regs is structured the same as the start of pt_regs */
> err = __copy_to_user(&sc->sc_regs, regs, sizeof(sc->sc_regs));
> /* Save the floating-point state. */
> if (has_fpu())
> err |= save_fp_state(regs, &sc->sc_fpregs);
> + /* Save the vector state. */
> + if (has_vector())
> + err |= save_v_state(regs, &sc_reserved_free_ptr);
> +
> + /* Put END __riscv_ctx_hdr at the end. */
> + err = __put_user(END_MAGIC, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->magic);
> + err = __put_user(END_HDR_SIZE, &((struct __riscv_ctx_hdr *)sc_reserved_free_ptr)->size);
I really don't understand the need for this trailing header stuff. Is
this due to variable sized vector or enginnering for a yet to be
invented future extension's state.
But if so, won't we add a new ctx_hdr + magic for "foo" when that
happens. After all we were not doing this for existing FP stuff, granted
FP is all fixed size.
For Vector: magic+size tuple should be enough to identify/locate the ctx
info.
> return err;
> }
>
> @@ -178,9 +328,10 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set,
> {
> struct rt_sigframe __user *frame;
> long err = 0;
> + size_t frame_size = cal_rt_frame_size();
>
> - frame = get_sigframe(ksig, regs, sizeof(*frame));
> - if (!access_ok(frame, sizeof(*frame)))
> + frame = get_sigframe(ksig, regs, frame_size);
> + if (!access_ok(frame, frame_size))
> return -EFAULT;
>
> err |= copy_siginfo_to_user(&frame->info, &ksig->info);
> @@ -326,3 +477,9 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
> if (thread_info_flags & _TIF_NOTIFY_RESUME)
> resume_user_mode_work(regs);
> }
> +
> +void init_rt_signal_env(void);
> +void __init init_rt_signal_env(void)
> +{
> + rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
See above, should be precompute the output of cal_rt_frame_size() here too.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* [PATCH v12 11/17] riscv: signal: Report signal frame size to userspace via auxv
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Russell King (Oracle),
Sunil V L, Ruinland Tsai, Han-Kuan Chen, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Paul E. McKenney, Mark Rutland, Frederic Weisbecker, Changbin Du,
Myrtle Shah, Vitaly Wool, Catalin Marinas, Mark Brown,
Will Deacon, Janosch Frank, Alexey Dobriyan, Huacai Chen,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
The vector register belongs to the signal context. They need to be stored
and restored as entering and leaving the signal handler. According to the
V-extension specification, the maximum length of the vector registers can
be 2^(XLEN-1). Hence, if userspace refers to the MINSIGSTKSZ to create a
sigframe, it may not be enough. To resolve this problem, this patch refers
to the commit 94b07c1f8c39c
("arm64: signal: Report signal frame size to userspace via auxv") to enable
userspace to know the minimum required sigframe size through the auxiliary
vector and use it to allocate enough memory for signal context.
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
---
arch/riscv/include/asm/elf.h | 47 +++++++++++++++-------------
arch/riscv/include/asm/processor.h | 2 ++
arch/riscv/include/uapi/asm/auxvec.h | 1 +
arch/riscv/kernel/signal.c | 8 +++++
4 files changed, 36 insertions(+), 22 deletions(-)
diff --git a/arch/riscv/include/asm/elf.h b/arch/riscv/include/asm/elf.h
index 14fc7342490b..df9047b619e8 100644
--- a/arch/riscv/include/asm/elf.h
+++ b/arch/riscv/include/asm/elf.h
@@ -77,28 +77,31 @@ extern unsigned long elf_hwcap;
#define COMPAT_ELF_PLATFORM (NULL)
#ifdef CONFIG_MMU
-#define ARCH_DLINFO \
-do { \
- /* \
- * Note that we add ulong after elf_addr_t because \
- * casting current->mm->context.vdso triggers a cast \
- * warning of cast from pointer to integer for \
- * COMPAT ELFCLASS32. \
- */ \
- NEW_AUX_ENT(AT_SYSINFO_EHDR, \
- (elf_addr_t)(ulong)current->mm->context.vdso); \
- NEW_AUX_ENT(AT_L1I_CACHESIZE, \
- get_cache_size(1, CACHE_TYPE_INST)); \
- NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY, \
- get_cache_geometry(1, CACHE_TYPE_INST)); \
- NEW_AUX_ENT(AT_L1D_CACHESIZE, \
- get_cache_size(1, CACHE_TYPE_DATA)); \
- NEW_AUX_ENT(AT_L1D_CACHEGEOMETRY, \
- get_cache_geometry(1, CACHE_TYPE_DATA)); \
- NEW_AUX_ENT(AT_L2_CACHESIZE, \
- get_cache_size(2, CACHE_TYPE_UNIFIED)); \
- NEW_AUX_ENT(AT_L2_CACHEGEOMETRY, \
- get_cache_geometry(2, CACHE_TYPE_UNIFIED)); \
+#define ARCH_DLINFO \
+do { \
+ NEW_AUX_ENT(AT_SYSINFO_EHDR, \
+ (elf_addr_t)current->mm->context.vdso); \
+ NEW_AUX_ENT(AT_L1I_CACHESIZE, \
+ get_cache_size(1, CACHE_TYPE_INST)); \
+ NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY, \
+ get_cache_geometry(1, CACHE_TYPE_INST)); \
+ NEW_AUX_ENT(AT_L1D_CACHESIZE, \
+ get_cache_size(1, CACHE_TYPE_DATA)); \
+ NEW_AUX_ENT(AT_L1D_CACHEGEOMETRY, \
+ get_cache_geometry(1, CACHE_TYPE_DATA)); \
+ NEW_AUX_ENT(AT_L2_CACHESIZE, \
+ get_cache_size(2, CACHE_TYPE_UNIFIED)); \
+ NEW_AUX_ENT(AT_L2_CACHEGEOMETRY, \
+ get_cache_geometry(2, CACHE_TYPE_UNIFIED)); \
+ /* \
+ * Should always be nonzero unless there's a kernel bug. \
+ * If we haven't determined a sensible value to give to \
+ * userspace, omit the entry: \
+ */ \
+ if (likely(signal_minsigstksz)) \
+ NEW_AUX_ENT(AT_MINSIGSTKSZ, signal_minsigstksz); \
+ else \
+ NEW_AUX_ENT(AT_IGNORE, 0); \
} while (0)
#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
struct linux_binprm;
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index 95917a2b24f9..a09141ecf6aa 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -7,6 +7,7 @@
#define _ASM_RISCV_PROCESSOR_H
#include <linux/const.h>
+#include <linux/cache.h>
#include <vdso/processor.h>
@@ -86,6 +87,7 @@ int riscv_of_parent_hartid(struct device_node *node, unsigned long *hartid);
extern void riscv_fill_hwcap(void);
extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
+extern unsigned long signal_minsigstksz __ro_after_init;
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/include/uapi/asm/auxvec.h b/arch/riscv/include/uapi/asm/auxvec.h
index 32c73ba1d531..6610d24e6662 100644
--- a/arch/riscv/include/uapi/asm/auxvec.h
+++ b/arch/riscv/include/uapi/asm/auxvec.h
@@ -33,5 +33,6 @@
/* entries in ARCH_DLINFO */
#define AT_VECTOR_SIZE_ARCH 7
+#define AT_MINSIGSTKSZ 51
#endif /* _UAPI_ASM_RISCV_AUXVEC_H */
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
index 41d9a02c7098..fb1f5012416e 100644
--- a/arch/riscv/kernel/signal.c
+++ b/arch/riscv/kernel/signal.c
@@ -478,8 +478,16 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
resume_user_mode_work(regs);
}
+unsigned long __ro_after_init signal_minsigstksz;
+
void init_rt_signal_env(void);
void __init init_rt_signal_env(void)
{
rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
+ /*
+ * Determine the stack space required for guaranteed signal delivery.
+ * The signal_minsigstksz will be populated into the AT_MINSIGSTKSZ entry
+ * in the auxiliary array at process startup.
+ */
+ signal_minsigstksz = cal_rt_frame_size();
}
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 11/17] riscv: signal: Report signal frame size to userspace via auxv
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Russell King (Oracle),
Sunil V L, Ruinland Tsai, Han-Kuan Chen, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Paul E. McKenney, Mark Rutland, Frederic Weisbecker, Changbin Du,
Myrtle Shah, Vitaly Wool, Catalin Marinas, Mark Brown,
Will Deacon, Janosch Frank, Alexey Dobriyan, Huacai Chen,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
The vector register belongs to the signal context. They need to be stored
and restored as entering and leaving the signal handler. According to the
V-extension specification, the maximum length of the vector registers can
be 2^(XLEN-1). Hence, if userspace refers to the MINSIGSTKSZ to create a
sigframe, it may not be enough. To resolve this problem, this patch refers
to the commit 94b07c1f8c39c
("arm64: signal: Report signal frame size to userspace via auxv") to enable
userspace to know the minimum required sigframe size through the auxiliary
vector and use it to allocate enough memory for signal context.
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
---
arch/riscv/include/asm/elf.h | 47 +++++++++++++++-------------
arch/riscv/include/asm/processor.h | 2 ++
arch/riscv/include/uapi/asm/auxvec.h | 1 +
arch/riscv/kernel/signal.c | 8 +++++
4 files changed, 36 insertions(+), 22 deletions(-)
diff --git a/arch/riscv/include/asm/elf.h b/arch/riscv/include/asm/elf.h
index 14fc7342490b..df9047b619e8 100644
--- a/arch/riscv/include/asm/elf.h
+++ b/arch/riscv/include/asm/elf.h
@@ -77,28 +77,31 @@ extern unsigned long elf_hwcap;
#define COMPAT_ELF_PLATFORM (NULL)
#ifdef CONFIG_MMU
-#define ARCH_DLINFO \
-do { \
- /* \
- * Note that we add ulong after elf_addr_t because \
- * casting current->mm->context.vdso triggers a cast \
- * warning of cast from pointer to integer for \
- * COMPAT ELFCLASS32. \
- */ \
- NEW_AUX_ENT(AT_SYSINFO_EHDR, \
- (elf_addr_t)(ulong)current->mm->context.vdso); \
- NEW_AUX_ENT(AT_L1I_CACHESIZE, \
- get_cache_size(1, CACHE_TYPE_INST)); \
- NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY, \
- get_cache_geometry(1, CACHE_TYPE_INST)); \
- NEW_AUX_ENT(AT_L1D_CACHESIZE, \
- get_cache_size(1, CACHE_TYPE_DATA)); \
- NEW_AUX_ENT(AT_L1D_CACHEGEOMETRY, \
- get_cache_geometry(1, CACHE_TYPE_DATA)); \
- NEW_AUX_ENT(AT_L2_CACHESIZE, \
- get_cache_size(2, CACHE_TYPE_UNIFIED)); \
- NEW_AUX_ENT(AT_L2_CACHEGEOMETRY, \
- get_cache_geometry(2, CACHE_TYPE_UNIFIED)); \
+#define ARCH_DLINFO \
+do { \
+ NEW_AUX_ENT(AT_SYSINFO_EHDR, \
+ (elf_addr_t)current->mm->context.vdso); \
+ NEW_AUX_ENT(AT_L1I_CACHESIZE, \
+ get_cache_size(1, CACHE_TYPE_INST)); \
+ NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY, \
+ get_cache_geometry(1, CACHE_TYPE_INST)); \
+ NEW_AUX_ENT(AT_L1D_CACHESIZE, \
+ get_cache_size(1, CACHE_TYPE_DATA)); \
+ NEW_AUX_ENT(AT_L1D_CACHEGEOMETRY, \
+ get_cache_geometry(1, CACHE_TYPE_DATA)); \
+ NEW_AUX_ENT(AT_L2_CACHESIZE, \
+ get_cache_size(2, CACHE_TYPE_UNIFIED)); \
+ NEW_AUX_ENT(AT_L2_CACHEGEOMETRY, \
+ get_cache_geometry(2, CACHE_TYPE_UNIFIED)); \
+ /* \
+ * Should always be nonzero unless there's a kernel bug. \
+ * If we haven't determined a sensible value to give to \
+ * userspace, omit the entry: \
+ */ \
+ if (likely(signal_minsigstksz)) \
+ NEW_AUX_ENT(AT_MINSIGSTKSZ, signal_minsigstksz); \
+ else \
+ NEW_AUX_ENT(AT_IGNORE, 0); \
} while (0)
#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
struct linux_binprm;
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index 95917a2b24f9..a09141ecf6aa 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -7,6 +7,7 @@
#define _ASM_RISCV_PROCESSOR_H
#include <linux/const.h>
+#include <linux/cache.h>
#include <vdso/processor.h>
@@ -86,6 +87,7 @@ int riscv_of_parent_hartid(struct device_node *node, unsigned long *hartid);
extern void riscv_fill_hwcap(void);
extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
+extern unsigned long signal_minsigstksz __ro_after_init;
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/include/uapi/asm/auxvec.h b/arch/riscv/include/uapi/asm/auxvec.h
index 32c73ba1d531..6610d24e6662 100644
--- a/arch/riscv/include/uapi/asm/auxvec.h
+++ b/arch/riscv/include/uapi/asm/auxvec.h
@@ -33,5 +33,6 @@
/* entries in ARCH_DLINFO */
#define AT_VECTOR_SIZE_ARCH 7
+#define AT_MINSIGSTKSZ 51
#endif /* _UAPI_ASM_RISCV_AUXVEC_H */
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
index 41d9a02c7098..fb1f5012416e 100644
--- a/arch/riscv/kernel/signal.c
+++ b/arch/riscv/kernel/signal.c
@@ -478,8 +478,16 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
resume_user_mode_work(regs);
}
+unsigned long __ro_after_init signal_minsigstksz;
+
void init_rt_signal_env(void);
void __init init_rt_signal_env(void)
{
rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
+ /*
+ * Determine the stack space required for guaranteed signal delivery.
+ * The signal_minsigstksz will be populated into the AT_MINSIGSTKSZ entry
+ * in the auxiliary array at process startup.
+ */
+ signal_minsigstksz = cal_rt_frame_size();
}
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 11/17] riscv: signal: Report signal frame size to userspace via auxv
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Arnaud Pouliquen,
Conor Dooley, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Dao Lu,
Jisheng Zhang, Russell King (Oracle),
Sunil V L, Ruinland Tsai, Han-Kuan Chen, Li Zhengyu,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Paul E. McKenney, Mark Rutland, Frederic Weisbecker, Changbin Du,
Myrtle Shah, Vitaly Wool, Catalin Marinas, Mark Brown,
Will Deacon, Janosch Frank, Alexey Dobriyan, Huacai Chen,
Christian Brauner, Evgenii Stepanov, Peter Collingbourne,
Eugene Syromiatnikov, Colin Cross, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
The vector register belongs to the signal context. They need to be stored
and restored as entering and leaving the signal handler. According to the
V-extension specification, the maximum length of the vector registers can
be 2^(XLEN-1). Hence, if userspace refers to the MINSIGSTKSZ to create a
sigframe, it may not be enough. To resolve this problem, this patch refers
to the commit 94b07c1f8c39c
("arm64: signal: Report signal frame size to userspace via auxv") to enable
userspace to know the minimum required sigframe size through the auxiliary
vector and use it to allocate enough memory for signal context.
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
---
arch/riscv/include/asm/elf.h | 47 +++++++++++++++-------------
arch/riscv/include/asm/processor.h | 2 ++
arch/riscv/include/uapi/asm/auxvec.h | 1 +
arch/riscv/kernel/signal.c | 8 +++++
4 files changed, 36 insertions(+), 22 deletions(-)
diff --git a/arch/riscv/include/asm/elf.h b/arch/riscv/include/asm/elf.h
index 14fc7342490b..df9047b619e8 100644
--- a/arch/riscv/include/asm/elf.h
+++ b/arch/riscv/include/asm/elf.h
@@ -77,28 +77,31 @@ extern unsigned long elf_hwcap;
#define COMPAT_ELF_PLATFORM (NULL)
#ifdef CONFIG_MMU
-#define ARCH_DLINFO \
-do { \
- /* \
- * Note that we add ulong after elf_addr_t because \
- * casting current->mm->context.vdso triggers a cast \
- * warning of cast from pointer to integer for \
- * COMPAT ELFCLASS32. \
- */ \
- NEW_AUX_ENT(AT_SYSINFO_EHDR, \
- (elf_addr_t)(ulong)current->mm->context.vdso); \
- NEW_AUX_ENT(AT_L1I_CACHESIZE, \
- get_cache_size(1, CACHE_TYPE_INST)); \
- NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY, \
- get_cache_geometry(1, CACHE_TYPE_INST)); \
- NEW_AUX_ENT(AT_L1D_CACHESIZE, \
- get_cache_size(1, CACHE_TYPE_DATA)); \
- NEW_AUX_ENT(AT_L1D_CACHEGEOMETRY, \
- get_cache_geometry(1, CACHE_TYPE_DATA)); \
- NEW_AUX_ENT(AT_L2_CACHESIZE, \
- get_cache_size(2, CACHE_TYPE_UNIFIED)); \
- NEW_AUX_ENT(AT_L2_CACHEGEOMETRY, \
- get_cache_geometry(2, CACHE_TYPE_UNIFIED)); \
+#define ARCH_DLINFO \
+do { \
+ NEW_AUX_ENT(AT_SYSINFO_EHDR, \
+ (elf_addr_t)current->mm->context.vdso); \
+ NEW_AUX_ENT(AT_L1I_CACHESIZE, \
+ get_cache_size(1, CACHE_TYPE_INST)); \
+ NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY, \
+ get_cache_geometry(1, CACHE_TYPE_INST)); \
+ NEW_AUX_ENT(AT_L1D_CACHESIZE, \
+ get_cache_size(1, CACHE_TYPE_DATA)); \
+ NEW_AUX_ENT(AT_L1D_CACHEGEOMETRY, \
+ get_cache_geometry(1, CACHE_TYPE_DATA)); \
+ NEW_AUX_ENT(AT_L2_CACHESIZE, \
+ get_cache_size(2, CACHE_TYPE_UNIFIED)); \
+ NEW_AUX_ENT(AT_L2_CACHEGEOMETRY, \
+ get_cache_geometry(2, CACHE_TYPE_UNIFIED)); \
+ /* \
+ * Should always be nonzero unless there's a kernel bug. \
+ * If we haven't determined a sensible value to give to \
+ * userspace, omit the entry: \
+ */ \
+ if (likely(signal_minsigstksz)) \
+ NEW_AUX_ENT(AT_MINSIGSTKSZ, signal_minsigstksz); \
+ else \
+ NEW_AUX_ENT(AT_IGNORE, 0); \
} while (0)
#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
struct linux_binprm;
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index 95917a2b24f9..a09141ecf6aa 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -7,6 +7,7 @@
#define _ASM_RISCV_PROCESSOR_H
#include <linux/const.h>
+#include <linux/cache.h>
#include <vdso/processor.h>
@@ -86,6 +87,7 @@ int riscv_of_parent_hartid(struct device_node *node, unsigned long *hartid);
extern void riscv_fill_hwcap(void);
extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
+extern unsigned long signal_minsigstksz __ro_after_init;
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/include/uapi/asm/auxvec.h b/arch/riscv/include/uapi/asm/auxvec.h
index 32c73ba1d531..6610d24e6662 100644
--- a/arch/riscv/include/uapi/asm/auxvec.h
+++ b/arch/riscv/include/uapi/asm/auxvec.h
@@ -33,5 +33,6 @@
/* entries in ARCH_DLINFO */
#define AT_VECTOR_SIZE_ARCH 7
+#define AT_MINSIGSTKSZ 51
#endif /* _UAPI_ASM_RISCV_AUXVEC_H */
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
index 41d9a02c7098..fb1f5012416e 100644
--- a/arch/riscv/kernel/signal.c
+++ b/arch/riscv/kernel/signal.c
@@ -478,8 +478,16 @@ asmlinkage __visible void do_notify_resume(struct pt_regs *regs,
resume_user_mode_work(regs);
}
+unsigned long __ro_after_init signal_minsigstksz;
+
void init_rt_signal_env(void);
void __init init_rt_signal_env(void)
{
rvv_sc_size = sizeof(struct __sc_riscv_v_state) + riscv_vsize;
+ /*
+ * Determine the stack space required for guaranteed signal delivery.
+ * The signal_minsigstksz will be populated into the AT_MINSIGSTKSZ entry
+ * in the auxiliary array at process startup.
+ */
+ signal_minsigstksz = cal_rt_frame_size();
}
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 12/17] riscv: Add support for kernel mode vector
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Conor Dooley, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu,
Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Nicolas Saenz Julienne,
Vitaly Wool, Myrtle Shah, Nick Knight, Catalin Marinas,
Will Deacon, Mark Brown, WANG Xuerui, Alexey Dobriyan,
Huacai Chen, Christian Brauner, Eugene Syromiatnikov,
Colin Cross, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Add kernel_rvv_begin() and kernel_rvv_end() function declarations
and corresponding definitions in kernel_mode_vector.c
These are needed to wrap uses of vector in kernel mode.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/vector.h | 3 +
arch/riscv/kernel/Makefile | 1 +
arch/riscv/kernel/kernel_mode_vector.c | 132 +++++++++++++++++++++++++
arch/riscv/kernel/vector.S | 9 ++
4 files changed, 145 insertions(+)
create mode 100644 arch/riscv/kernel/kernel_mode_vector.c
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
index 16304b0c6a6f..a59841cc81fb 100644
--- a/arch/riscv/include/asm/vector.h
+++ b/arch/riscv/include/asm/vector.h
@@ -10,5 +10,8 @@
void rvv_enable(void);
void rvv_disable(void);
+void kernel_rvv_begin(void);
+void kernel_rvv_end(void);
+void vector_flush_cpu_state(void);
#endif /* ! __ASM_RISCV_VECTOR_H */
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 35752fb6d145..8c238415f800 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -56,6 +56,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
obj-$(CONFIG_FPU) += fpu.o
obj-$(CONFIG_VECTOR) += vector.o
+obj-$(CONFIG_VECTOR) += kernel_mode_vector.o
obj-$(CONFIG_SMP) += smpboot.o
obj-$(CONFIG_SMP) += smp.o
obj-$(CONFIG_SMP) += cpu_ops.o
diff --git a/arch/riscv/kernel/kernel_mode_vector.c b/arch/riscv/kernel/kernel_mode_vector.c
new file mode 100644
index 000000000000..0277168af0c5
--- /dev/null
+++ b/arch/riscv/kernel/kernel_mode_vector.c
@@ -0,0 +1,132 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ * Author: Catalin Marinas <catalin.marinas@arm.com>
+ * Copyright (C) 2017 Linaro Ltd. <ard.biesheuvel@linaro.org>
+ * Copyright (C) 2021 SiFive
+ */
+#include <linux/compiler.h>
+#include <linux/irqflags.h>
+#include <linux/percpu.h>
+#include <linux/preempt.h>
+#include <linux/types.h>
+
+#include <asm/vector.h>
+#include <asm/switch_to.h>
+
+DECLARE_PER_CPU(bool, vector_context_busy);
+DEFINE_PER_CPU(bool, vector_context_busy);
+
+/*
+ * may_use_vector - whether it is allowable at this time to issue vector
+ * instructions or access the vector register file
+ *
+ * Callers must not assume that the result remains true beyond the next
+ * preempt_enable() or return from softirq context.
+ */
+static __must_check inline bool may_use_vector(void)
+{
+ /*
+ * vector_context_busy is only set while preemption is disabled,
+ * and is clear whenever preemption is enabled. Since
+ * this_cpu_read() is atomic w.r.t. preemption, vector_context_busy
+ * cannot change under our feet -- if it's set we cannot be
+ * migrated, and if it's clear we cannot be migrated to a CPU
+ * where it is set.
+ */
+ return !in_irq() && !irqs_disabled() && !in_nmi() &&
+ !this_cpu_read(vector_context_busy);
+}
+
+/*
+ * Claim ownership of the CPU vector context for use by the calling context.
+ *
+ * The caller may freely manipulate the vector context metadata until
+ * put_cpu_vector_context() is called.
+ */
+static void get_cpu_vector_context(void)
+{
+ bool busy;
+
+ preempt_disable();
+ busy = __this_cpu_xchg(vector_context_busy, true);
+
+ WARN_ON(busy);
+}
+
+/*
+ * Release the CPU vector context.
+ *
+ * Must be called from a context in which get_cpu_vector_context() was
+ * previously called, with no call to put_cpu_vector_context() in the
+ * meantime.
+ */
+static void put_cpu_vector_context(void)
+{
+ bool busy = __this_cpu_xchg(vector_context_busy, false);
+
+ WARN_ON(!busy);
+ preempt_enable();
+}
+
+/*
+ * kernel_rvv_begin(): obtain the CPU vector registers for use by the calling
+ * context
+ *
+ * Must not be called unless may_use_vector() returns true.
+ * Task context in the vector registers is saved back to memory as necessary.
+ *
+ * A matching call to kernel_rvv_end() must be made before returning from the
+ * calling context.
+ *
+ * The caller may freely use the vector registers until kernel_rvv_end() is
+ * called.
+ */
+void kernel_rvv_begin(void)
+{
+ if (WARN_ON(!has_vector()))
+ return;
+
+ WARN_ON(!may_use_vector());
+
+ /* Acquire kernel mode vector */
+ get_cpu_vector_context();
+
+ /* Save vector state, if any */
+ vstate_save(current, task_pt_regs(current));
+
+ /* Enable vector */
+ rvv_enable();
+
+ /* Invalidate vector regs */
+ vector_flush_cpu_state();
+}
+EXPORT_SYMBOL_GPL(kernel_rvv_begin);
+
+/*
+ * kernel_rvv_end(): give the CPU vector registers back to the current task
+ *
+ * Must be called from a context in which kernel_rvv_begin() was previously
+ * called, with no call to kernel_rvv_end() in the meantime.
+ *
+ * The caller must not use the vector registers after this function is called,
+ * unless kernel_rvv_begin() is called again in the meantime.
+ */
+void kernel_rvv_end(void)
+{
+ if (WARN_ON(!has_vector()))
+ return;
+
+ /* Invalidate vector regs */
+ vector_flush_cpu_state();
+
+ /* Restore vector state, if any */
+ vstate_restore(current, task_pt_regs(current));
+
+ /* disable vector */
+ rvv_disable();
+
+ /* release kernel mode vector */
+ put_cpu_vector_context();
+}
+EXPORT_SYMBOL_GPL(kernel_rvv_end);
diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
index 9f7dc70c4443..9c2de823c0d9 100644
--- a/arch/riscv/kernel/vector.S
+++ b/arch/riscv/kernel/vector.S
@@ -91,3 +91,12 @@ ENTRY(rvv_disable)
csrc CSR_STATUS, status
ret
ENDPROC(rvv_disable)
+
+ENTRY(vector_flush_cpu_state)
+ vsetvli t0, x0, e8, m8, ta, ma
+ vmv.v.i v0, 0
+ vmv.v.i v8, 0
+ vmv.v.i v16, 0
+ vmv.v.i v24, 0
+ ret
+ENDPROC(vector_flush_cpu_state)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 12/17] riscv: Add support for kernel mode vector
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Conor Dooley, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu,
Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Nicolas Saenz Julienne,
Vitaly Wool, Myrtle Shah, Nick Knight, Catalin Marinas,
Will Deacon, Mark Brown, WANG Xuerui, Alexey Dobriyan,
Huacai Chen, Christian Brauner, Eugene Syromiatnikov,
Colin Cross, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Add kernel_rvv_begin() and kernel_rvv_end() function declarations
and corresponding definitions in kernel_mode_vector.c
These are needed to wrap uses of vector in kernel mode.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/vector.h | 3 +
arch/riscv/kernel/Makefile | 1 +
arch/riscv/kernel/kernel_mode_vector.c | 132 +++++++++++++++++++++++++
arch/riscv/kernel/vector.S | 9 ++
4 files changed, 145 insertions(+)
create mode 100644 arch/riscv/kernel/kernel_mode_vector.c
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
index 16304b0c6a6f..a59841cc81fb 100644
--- a/arch/riscv/include/asm/vector.h
+++ b/arch/riscv/include/asm/vector.h
@@ -10,5 +10,8 @@
void rvv_enable(void);
void rvv_disable(void);
+void kernel_rvv_begin(void);
+void kernel_rvv_end(void);
+void vector_flush_cpu_state(void);
#endif /* ! __ASM_RISCV_VECTOR_H */
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 35752fb6d145..8c238415f800 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -56,6 +56,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
obj-$(CONFIG_FPU) += fpu.o
obj-$(CONFIG_VECTOR) += vector.o
+obj-$(CONFIG_VECTOR) += kernel_mode_vector.o
obj-$(CONFIG_SMP) += smpboot.o
obj-$(CONFIG_SMP) += smp.o
obj-$(CONFIG_SMP) += cpu_ops.o
diff --git a/arch/riscv/kernel/kernel_mode_vector.c b/arch/riscv/kernel/kernel_mode_vector.c
new file mode 100644
index 000000000000..0277168af0c5
--- /dev/null
+++ b/arch/riscv/kernel/kernel_mode_vector.c
@@ -0,0 +1,132 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ * Author: Catalin Marinas <catalin.marinas@arm.com>
+ * Copyright (C) 2017 Linaro Ltd. <ard.biesheuvel@linaro.org>
+ * Copyright (C) 2021 SiFive
+ */
+#include <linux/compiler.h>
+#include <linux/irqflags.h>
+#include <linux/percpu.h>
+#include <linux/preempt.h>
+#include <linux/types.h>
+
+#include <asm/vector.h>
+#include <asm/switch_to.h>
+
+DECLARE_PER_CPU(bool, vector_context_busy);
+DEFINE_PER_CPU(bool, vector_context_busy);
+
+/*
+ * may_use_vector - whether it is allowable at this time to issue vector
+ * instructions or access the vector register file
+ *
+ * Callers must not assume that the result remains true beyond the next
+ * preempt_enable() or return from softirq context.
+ */
+static __must_check inline bool may_use_vector(void)
+{
+ /*
+ * vector_context_busy is only set while preemption is disabled,
+ * and is clear whenever preemption is enabled. Since
+ * this_cpu_read() is atomic w.r.t. preemption, vector_context_busy
+ * cannot change under our feet -- if it's set we cannot be
+ * migrated, and if it's clear we cannot be migrated to a CPU
+ * where it is set.
+ */
+ return !in_irq() && !irqs_disabled() && !in_nmi() &&
+ !this_cpu_read(vector_context_busy);
+}
+
+/*
+ * Claim ownership of the CPU vector context for use by the calling context.
+ *
+ * The caller may freely manipulate the vector context metadata until
+ * put_cpu_vector_context() is called.
+ */
+static void get_cpu_vector_context(void)
+{
+ bool busy;
+
+ preempt_disable();
+ busy = __this_cpu_xchg(vector_context_busy, true);
+
+ WARN_ON(busy);
+}
+
+/*
+ * Release the CPU vector context.
+ *
+ * Must be called from a context in which get_cpu_vector_context() was
+ * previously called, with no call to put_cpu_vector_context() in the
+ * meantime.
+ */
+static void put_cpu_vector_context(void)
+{
+ bool busy = __this_cpu_xchg(vector_context_busy, false);
+
+ WARN_ON(!busy);
+ preempt_enable();
+}
+
+/*
+ * kernel_rvv_begin(): obtain the CPU vector registers for use by the calling
+ * context
+ *
+ * Must not be called unless may_use_vector() returns true.
+ * Task context in the vector registers is saved back to memory as necessary.
+ *
+ * A matching call to kernel_rvv_end() must be made before returning from the
+ * calling context.
+ *
+ * The caller may freely use the vector registers until kernel_rvv_end() is
+ * called.
+ */
+void kernel_rvv_begin(void)
+{
+ if (WARN_ON(!has_vector()))
+ return;
+
+ WARN_ON(!may_use_vector());
+
+ /* Acquire kernel mode vector */
+ get_cpu_vector_context();
+
+ /* Save vector state, if any */
+ vstate_save(current, task_pt_regs(current));
+
+ /* Enable vector */
+ rvv_enable();
+
+ /* Invalidate vector regs */
+ vector_flush_cpu_state();
+}
+EXPORT_SYMBOL_GPL(kernel_rvv_begin);
+
+/*
+ * kernel_rvv_end(): give the CPU vector registers back to the current task
+ *
+ * Must be called from a context in which kernel_rvv_begin() was previously
+ * called, with no call to kernel_rvv_end() in the meantime.
+ *
+ * The caller must not use the vector registers after this function is called,
+ * unless kernel_rvv_begin() is called again in the meantime.
+ */
+void kernel_rvv_end(void)
+{
+ if (WARN_ON(!has_vector()))
+ return;
+
+ /* Invalidate vector regs */
+ vector_flush_cpu_state();
+
+ /* Restore vector state, if any */
+ vstate_restore(current, task_pt_regs(current));
+
+ /* disable vector */
+ rvv_disable();
+
+ /* release kernel mode vector */
+ put_cpu_vector_context();
+}
+EXPORT_SYMBOL_GPL(kernel_rvv_end);
diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
index 9f7dc70c4443..9c2de823c0d9 100644
--- a/arch/riscv/kernel/vector.S
+++ b/arch/riscv/kernel/vector.S
@@ -91,3 +91,12 @@ ENTRY(rvv_disable)
csrc CSR_STATUS, status
ret
ENDPROC(rvv_disable)
+
+ENTRY(vector_flush_cpu_state)
+ vsetvli t0, x0, e8, m8, ta, ma
+ vmv.v.i v0, 0
+ vmv.v.i v8, 0
+ vmv.v.i v16, 0
+ vmv.v.i v24, 0
+ ret
+ENDPROC(vector_flush_cpu_state)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 12/17] riscv: Add support for kernel mode vector
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Vincent Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Chris Stillson,
Conor Dooley, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Jisheng Zhang, Dao Lu,
Peter Zijlstra (Intel),
Sunil V L, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Nicolas Saenz Julienne,
Vitaly Wool, Myrtle Shah, Nick Knight, Catalin Marinas,
Will Deacon, Mark Brown, WANG Xuerui, Alexey Dobriyan,
Huacai Chen, Christian Brauner, Eugene Syromiatnikov,
Colin Cross, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Add kernel_rvv_begin() and kernel_rvv_end() function declarations
and corresponding definitions in kernel_mode_vector.c
These are needed to wrap uses of vector in kernel mode.
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/vector.h | 3 +
arch/riscv/kernel/Makefile | 1 +
arch/riscv/kernel/kernel_mode_vector.c | 132 +++++++++++++++++++++++++
arch/riscv/kernel/vector.S | 9 ++
4 files changed, 145 insertions(+)
create mode 100644 arch/riscv/kernel/kernel_mode_vector.c
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
index 16304b0c6a6f..a59841cc81fb 100644
--- a/arch/riscv/include/asm/vector.h
+++ b/arch/riscv/include/asm/vector.h
@@ -10,5 +10,8 @@
void rvv_enable(void);
void rvv_disable(void);
+void kernel_rvv_begin(void);
+void kernel_rvv_end(void);
+void vector_flush_cpu_state(void);
#endif /* ! __ASM_RISCV_VECTOR_H */
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 35752fb6d145..8c238415f800 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -56,6 +56,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o
obj-$(CONFIG_FPU) += fpu.o
obj-$(CONFIG_VECTOR) += vector.o
+obj-$(CONFIG_VECTOR) += kernel_mode_vector.o
obj-$(CONFIG_SMP) += smpboot.o
obj-$(CONFIG_SMP) += smp.o
obj-$(CONFIG_SMP) += cpu_ops.o
diff --git a/arch/riscv/kernel/kernel_mode_vector.c b/arch/riscv/kernel/kernel_mode_vector.c
new file mode 100644
index 000000000000..0277168af0c5
--- /dev/null
+++ b/arch/riscv/kernel/kernel_mode_vector.c
@@ -0,0 +1,132 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ * Author: Catalin Marinas <catalin.marinas@arm.com>
+ * Copyright (C) 2017 Linaro Ltd. <ard.biesheuvel@linaro.org>
+ * Copyright (C) 2021 SiFive
+ */
+#include <linux/compiler.h>
+#include <linux/irqflags.h>
+#include <linux/percpu.h>
+#include <linux/preempt.h>
+#include <linux/types.h>
+
+#include <asm/vector.h>
+#include <asm/switch_to.h>
+
+DECLARE_PER_CPU(bool, vector_context_busy);
+DEFINE_PER_CPU(bool, vector_context_busy);
+
+/*
+ * may_use_vector - whether it is allowable at this time to issue vector
+ * instructions or access the vector register file
+ *
+ * Callers must not assume that the result remains true beyond the next
+ * preempt_enable() or return from softirq context.
+ */
+static __must_check inline bool may_use_vector(void)
+{
+ /*
+ * vector_context_busy is only set while preemption is disabled,
+ * and is clear whenever preemption is enabled. Since
+ * this_cpu_read() is atomic w.r.t. preemption, vector_context_busy
+ * cannot change under our feet -- if it's set we cannot be
+ * migrated, and if it's clear we cannot be migrated to a CPU
+ * where it is set.
+ */
+ return !in_irq() && !irqs_disabled() && !in_nmi() &&
+ !this_cpu_read(vector_context_busy);
+}
+
+/*
+ * Claim ownership of the CPU vector context for use by the calling context.
+ *
+ * The caller may freely manipulate the vector context metadata until
+ * put_cpu_vector_context() is called.
+ */
+static void get_cpu_vector_context(void)
+{
+ bool busy;
+
+ preempt_disable();
+ busy = __this_cpu_xchg(vector_context_busy, true);
+
+ WARN_ON(busy);
+}
+
+/*
+ * Release the CPU vector context.
+ *
+ * Must be called from a context in which get_cpu_vector_context() was
+ * previously called, with no call to put_cpu_vector_context() in the
+ * meantime.
+ */
+static void put_cpu_vector_context(void)
+{
+ bool busy = __this_cpu_xchg(vector_context_busy, false);
+
+ WARN_ON(!busy);
+ preempt_enable();
+}
+
+/*
+ * kernel_rvv_begin(): obtain the CPU vector registers for use by the calling
+ * context
+ *
+ * Must not be called unless may_use_vector() returns true.
+ * Task context in the vector registers is saved back to memory as necessary.
+ *
+ * A matching call to kernel_rvv_end() must be made before returning from the
+ * calling context.
+ *
+ * The caller may freely use the vector registers until kernel_rvv_end() is
+ * called.
+ */
+void kernel_rvv_begin(void)
+{
+ if (WARN_ON(!has_vector()))
+ return;
+
+ WARN_ON(!may_use_vector());
+
+ /* Acquire kernel mode vector */
+ get_cpu_vector_context();
+
+ /* Save vector state, if any */
+ vstate_save(current, task_pt_regs(current));
+
+ /* Enable vector */
+ rvv_enable();
+
+ /* Invalidate vector regs */
+ vector_flush_cpu_state();
+}
+EXPORT_SYMBOL_GPL(kernel_rvv_begin);
+
+/*
+ * kernel_rvv_end(): give the CPU vector registers back to the current task
+ *
+ * Must be called from a context in which kernel_rvv_begin() was previously
+ * called, with no call to kernel_rvv_end() in the meantime.
+ *
+ * The caller must not use the vector registers after this function is called,
+ * unless kernel_rvv_begin() is called again in the meantime.
+ */
+void kernel_rvv_end(void)
+{
+ if (WARN_ON(!has_vector()))
+ return;
+
+ /* Invalidate vector regs */
+ vector_flush_cpu_state();
+
+ /* Restore vector state, if any */
+ vstate_restore(current, task_pt_regs(current));
+
+ /* disable vector */
+ rvv_disable();
+
+ /* release kernel mode vector */
+ put_cpu_vector_context();
+}
+EXPORT_SYMBOL_GPL(kernel_rvv_end);
diff --git a/arch/riscv/kernel/vector.S b/arch/riscv/kernel/vector.S
index 9f7dc70c4443..9c2de823c0d9 100644
--- a/arch/riscv/kernel/vector.S
+++ b/arch/riscv/kernel/vector.S
@@ -91,3 +91,12 @@ ENTRY(rvv_disable)
csrc CSR_STATUS, status
ret
ENDPROC(rvv_disable)
+
+ENTRY(vector_flush_cpu_state)
+ vsetvli t0, x0, e8, m8, ta, ma
+ vmv.v.i v0, 0
+ vmv.v.i v8, 0
+ vmv.v.i v16, 0
+ vmv.v.i v24, 0
+ ret
+ENDPROC(vector_flush_cpu_state)
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 13/17] riscv: Add vector extension XOR implementation
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Han-Kuan Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Conor Dooley,
Arnaud Pouliquen, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Vincent Chen, Heiko Stuebner,
Dao Lu, Jisheng Zhang, Sunil V L, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Paul E. McKenney,
Nicolas Saenz Julienne, Mark Rutland, Frederic Weisbecker,
Changbin Du, Vitaly Wool, Myrtle Shah, Catalin Marinas,
Will Deacon, Mark Brown, Huacai Chen, Alexey Dobriyan,
Janosch Frank, Christian Brauner, Colin Cross,
Eugene Syromiatnikov, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds support for vector optimized XOR and it is tested in
qemu.
Co-developed-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/xor.h | 82 ++++++++++++++++++++++++++++++++++++
arch/riscv/lib/Makefile | 1 +
arch/riscv/lib/xor.S | 81 +++++++++++++++++++++++++++++++++++
3 files changed, 164 insertions(+)
create mode 100644 arch/riscv/include/asm/xor.h
create mode 100644 arch/riscv/lib/xor.S
diff --git a/arch/riscv/include/asm/xor.h b/arch/riscv/include/asm/xor.h
new file mode 100644
index 000000000000..d1f2eeb14afb
--- /dev/null
+++ b/arch/riscv/include/asm/xor.h
@@ -0,0 +1,82 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2021 SiFive
+ */
+
+#include <linux/hardirq.h>
+#include <asm-generic/xor.h>
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+#include <asm/switch_to.h>
+
+void xor_regs_2_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2);
+void xor_regs_3_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3);
+void xor_regs_4_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4);
+void xor_regs_5_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4,
+ const unsigned long *__restrict p5);
+
+static void xor_rvv_2(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2)
+{
+ kernel_rvv_begin();
+ xor_regs_2_(bytes, p1, p2);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_3(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3)
+{
+ kernel_rvv_begin();
+ xor_regs_3_(bytes, p1, p2, p3);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_4(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4)
+{
+ kernel_rvv_begin();
+ xor_regs_4_(bytes, p1, p2, p3, p4);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_5(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4,
+ const unsigned long *__restrict p5)
+{
+ kernel_rvv_begin();
+ xor_regs_5_(bytes, p1, p2, p3, p4, p5);
+ kernel_rvv_end();
+}
+
+static struct xor_block_template xor_block_rvv = {
+ .name = "rvv",
+ .do_2 = xor_rvv_2,
+ .do_3 = xor_rvv_3,
+ .do_4 = xor_rvv_4,
+ .do_5 = xor_rvv_5
+};
+
+#undef XOR_TRY_TEMPLATES
+#define XOR_TRY_TEMPLATES \
+ do { \
+ xor_speed(&xor_block_8regs); \
+ xor_speed(&xor_block_32regs); \
+ if (has_vector()) { \
+ xor_speed(&xor_block_rvv);\
+ } \
+ } while (0)
+#endif
diff --git a/arch/riscv/lib/Makefile b/arch/riscv/lib/Makefile
index 25d5c9664e57..acd87ac86d24 100644
--- a/arch/riscv/lib/Makefile
+++ b/arch/riscv/lib/Makefile
@@ -7,3 +7,4 @@ lib-$(CONFIG_MMU) += uaccess.o
lib-$(CONFIG_64BIT) += tishift.o
obj-$(CONFIG_FUNCTION_ERROR_INJECTION) += error-inject.o
+lib-$(CONFIG_VECTOR) += xor.o
diff --git a/arch/riscv/lib/xor.S b/arch/riscv/lib/xor.S
new file mode 100644
index 000000000000..3bc059e18171
--- /dev/null
+++ b/arch/riscv/lib/xor.S
@@ -0,0 +1,81 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2021 SiFive
+ */
+#include <linux/linkage.h>
+#include <asm-generic/export.h>
+#include <asm/asm.h>
+
+ENTRY(xor_regs_2_)
+ vsetvli a3, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a3
+ vxor.vv v16, v0, v8
+ add a2, a2, a3
+ vse8.v v16, (a1)
+ add a1, a1, a3
+ bnez a0, xor_regs_2_
+ ret
+END(xor_regs_2_)
+EXPORT_SYMBOL(xor_regs_2_)
+
+ENTRY(xor_regs_3_)
+ vsetvli a4, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a4
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a4
+ vxor.vv v16, v0, v16
+ add a3, a3, a4
+ vse8.v v16, (a1)
+ add a1, a1, a4
+ bnez a0, xor_regs_3_
+ ret
+END(xor_regs_3_)
+EXPORT_SYMBOL(xor_regs_3_)
+
+ENTRY(xor_regs_4_)
+ vsetvli a5, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a5
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a5
+ vxor.vv v0, v0, v16
+ vle8.v v24, (a4)
+ add a3, a3, a5
+ vxor.vv v16, v0, v24
+ add a4, a4, a5
+ vse8.v v16, (a1)
+ add a1, a1, a5
+ bnez a0, xor_regs_4_
+ ret
+END(xor_regs_4_)
+EXPORT_SYMBOL(xor_regs_4_)
+
+ENTRY(xor_regs_5_)
+ vsetvli a6, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a6
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a6
+ vxor.vv v0, v0, v16
+ vle8.v v24, (a4)
+ add a3, a3, a6
+ vxor.vv v0, v0, v24
+ vle8.v v8, (a5)
+ add a4, a4, a6
+ vxor.vv v16, v0, v8
+ add a5, a5, a6
+ vse8.v v16, (a1)
+ add a1, a1, a6
+ bnez a0, xor_regs_5_
+ ret
+END(xor_regs_5_)
+EXPORT_SYMBOL(xor_regs_5_)
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 13/17] riscv: Add vector extension XOR implementation
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Han-Kuan Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Conor Dooley,
Arnaud Pouliquen, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Vincent Chen, Heiko Stuebner,
Dao Lu, Jisheng Zhang, Sunil V L, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Paul E. McKenney,
Nicolas Saenz Julienne, Mark Rutland, Frederic Weisbecker,
Changbin Du, Vitaly Wool, Myrtle Shah, Catalin Marinas,
Will Deacon, Mark Brown, Huacai Chen, Alexey Dobriyan,
Janosch Frank, Christian Brauner, Colin Cross,
Eugene Syromiatnikov, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds support for vector optimized XOR and it is tested in
qemu.
Co-developed-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/xor.h | 82 ++++++++++++++++++++++++++++++++++++
arch/riscv/lib/Makefile | 1 +
arch/riscv/lib/xor.S | 81 +++++++++++++++++++++++++++++++++++
3 files changed, 164 insertions(+)
create mode 100644 arch/riscv/include/asm/xor.h
create mode 100644 arch/riscv/lib/xor.S
diff --git a/arch/riscv/include/asm/xor.h b/arch/riscv/include/asm/xor.h
new file mode 100644
index 000000000000..d1f2eeb14afb
--- /dev/null
+++ b/arch/riscv/include/asm/xor.h
@@ -0,0 +1,82 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2021 SiFive
+ */
+
+#include <linux/hardirq.h>
+#include <asm-generic/xor.h>
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+#include <asm/switch_to.h>
+
+void xor_regs_2_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2);
+void xor_regs_3_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3);
+void xor_regs_4_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4);
+void xor_regs_5_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4,
+ const unsigned long *__restrict p5);
+
+static void xor_rvv_2(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2)
+{
+ kernel_rvv_begin();
+ xor_regs_2_(bytes, p1, p2);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_3(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3)
+{
+ kernel_rvv_begin();
+ xor_regs_3_(bytes, p1, p2, p3);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_4(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4)
+{
+ kernel_rvv_begin();
+ xor_regs_4_(bytes, p1, p2, p3, p4);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_5(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4,
+ const unsigned long *__restrict p5)
+{
+ kernel_rvv_begin();
+ xor_regs_5_(bytes, p1, p2, p3, p4, p5);
+ kernel_rvv_end();
+}
+
+static struct xor_block_template xor_block_rvv = {
+ .name = "rvv",
+ .do_2 = xor_rvv_2,
+ .do_3 = xor_rvv_3,
+ .do_4 = xor_rvv_4,
+ .do_5 = xor_rvv_5
+};
+
+#undef XOR_TRY_TEMPLATES
+#define XOR_TRY_TEMPLATES \
+ do { \
+ xor_speed(&xor_block_8regs); \
+ xor_speed(&xor_block_32regs); \
+ if (has_vector()) { \
+ xor_speed(&xor_block_rvv);\
+ } \
+ } while (0)
+#endif
diff --git a/arch/riscv/lib/Makefile b/arch/riscv/lib/Makefile
index 25d5c9664e57..acd87ac86d24 100644
--- a/arch/riscv/lib/Makefile
+++ b/arch/riscv/lib/Makefile
@@ -7,3 +7,4 @@ lib-$(CONFIG_MMU) += uaccess.o
lib-$(CONFIG_64BIT) += tishift.o
obj-$(CONFIG_FUNCTION_ERROR_INJECTION) += error-inject.o
+lib-$(CONFIG_VECTOR) += xor.o
diff --git a/arch/riscv/lib/xor.S b/arch/riscv/lib/xor.S
new file mode 100644
index 000000000000..3bc059e18171
--- /dev/null
+++ b/arch/riscv/lib/xor.S
@@ -0,0 +1,81 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2021 SiFive
+ */
+#include <linux/linkage.h>
+#include <asm-generic/export.h>
+#include <asm/asm.h>
+
+ENTRY(xor_regs_2_)
+ vsetvli a3, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a3
+ vxor.vv v16, v0, v8
+ add a2, a2, a3
+ vse8.v v16, (a1)
+ add a1, a1, a3
+ bnez a0, xor_regs_2_
+ ret
+END(xor_regs_2_)
+EXPORT_SYMBOL(xor_regs_2_)
+
+ENTRY(xor_regs_3_)
+ vsetvli a4, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a4
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a4
+ vxor.vv v16, v0, v16
+ add a3, a3, a4
+ vse8.v v16, (a1)
+ add a1, a1, a4
+ bnez a0, xor_regs_3_
+ ret
+END(xor_regs_3_)
+EXPORT_SYMBOL(xor_regs_3_)
+
+ENTRY(xor_regs_4_)
+ vsetvli a5, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a5
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a5
+ vxor.vv v0, v0, v16
+ vle8.v v24, (a4)
+ add a3, a3, a5
+ vxor.vv v16, v0, v24
+ add a4, a4, a5
+ vse8.v v16, (a1)
+ add a1, a1, a5
+ bnez a0, xor_regs_4_
+ ret
+END(xor_regs_4_)
+EXPORT_SYMBOL(xor_regs_4_)
+
+ENTRY(xor_regs_5_)
+ vsetvli a6, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a6
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a6
+ vxor.vv v0, v0, v16
+ vle8.v v24, (a4)
+ add a3, a3, a6
+ vxor.vv v0, v0, v24
+ vle8.v v8, (a5)
+ add a4, a4, a6
+ vxor.vv v16, v0, v8
+ add a5, a5, a6
+ vse8.v v16, (a1)
+ add a1, a1, a6
+ bnez a0, xor_regs_5_
+ ret
+END(xor_regs_5_)
+EXPORT_SYMBOL(xor_regs_5_)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 13/17] riscv: Add vector extension XOR implementation
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, Han-Kuan Chen, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Conor Dooley,
Arnaud Pouliquen, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Vincent Chen, Heiko Stuebner,
Dao Lu, Jisheng Zhang, Sunil V L, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Paul E. McKenney,
Nicolas Saenz Julienne, Mark Rutland, Frederic Weisbecker,
Changbin Du, Vitaly Wool, Myrtle Shah, Catalin Marinas,
Will Deacon, Mark Brown, Huacai Chen, Alexey Dobriyan,
Janosch Frank, Christian Brauner, Colin Cross,
Eugene Syromiatnikov, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
This patch adds support for vector optimized XOR and it is tested in
qemu.
Co-developed-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Han-Kuan Chen <hankuan.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/xor.h | 82 ++++++++++++++++++++++++++++++++++++
arch/riscv/lib/Makefile | 1 +
arch/riscv/lib/xor.S | 81 +++++++++++++++++++++++++++++++++++
3 files changed, 164 insertions(+)
create mode 100644 arch/riscv/include/asm/xor.h
create mode 100644 arch/riscv/lib/xor.S
diff --git a/arch/riscv/include/asm/xor.h b/arch/riscv/include/asm/xor.h
new file mode 100644
index 000000000000..d1f2eeb14afb
--- /dev/null
+++ b/arch/riscv/include/asm/xor.h
@@ -0,0 +1,82 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2021 SiFive
+ */
+
+#include <linux/hardirq.h>
+#include <asm-generic/xor.h>
+#ifdef CONFIG_VECTOR
+#include <asm/vector.h>
+#include <asm/switch_to.h>
+
+void xor_regs_2_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2);
+void xor_regs_3_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3);
+void xor_regs_4_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4);
+void xor_regs_5_(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4,
+ const unsigned long *__restrict p5);
+
+static void xor_rvv_2(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2)
+{
+ kernel_rvv_begin();
+ xor_regs_2_(bytes, p1, p2);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_3(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3)
+{
+ kernel_rvv_begin();
+ xor_regs_3_(bytes, p1, p2, p3);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_4(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4)
+{
+ kernel_rvv_begin();
+ xor_regs_4_(bytes, p1, p2, p3, p4);
+ kernel_rvv_end();
+}
+
+static void xor_rvv_5(unsigned long bytes, unsigned long *__restrict p1,
+ const unsigned long *__restrict p2,
+ const unsigned long *__restrict p3,
+ const unsigned long *__restrict p4,
+ const unsigned long *__restrict p5)
+{
+ kernel_rvv_begin();
+ xor_regs_5_(bytes, p1, p2, p3, p4, p5);
+ kernel_rvv_end();
+}
+
+static struct xor_block_template xor_block_rvv = {
+ .name = "rvv",
+ .do_2 = xor_rvv_2,
+ .do_3 = xor_rvv_3,
+ .do_4 = xor_rvv_4,
+ .do_5 = xor_rvv_5
+};
+
+#undef XOR_TRY_TEMPLATES
+#define XOR_TRY_TEMPLATES \
+ do { \
+ xor_speed(&xor_block_8regs); \
+ xor_speed(&xor_block_32regs); \
+ if (has_vector()) { \
+ xor_speed(&xor_block_rvv);\
+ } \
+ } while (0)
+#endif
diff --git a/arch/riscv/lib/Makefile b/arch/riscv/lib/Makefile
index 25d5c9664e57..acd87ac86d24 100644
--- a/arch/riscv/lib/Makefile
+++ b/arch/riscv/lib/Makefile
@@ -7,3 +7,4 @@ lib-$(CONFIG_MMU) += uaccess.o
lib-$(CONFIG_64BIT) += tishift.o
obj-$(CONFIG_FUNCTION_ERROR_INJECTION) += error-inject.o
+lib-$(CONFIG_VECTOR) += xor.o
diff --git a/arch/riscv/lib/xor.S b/arch/riscv/lib/xor.S
new file mode 100644
index 000000000000..3bc059e18171
--- /dev/null
+++ b/arch/riscv/lib/xor.S
@@ -0,0 +1,81 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2021 SiFive
+ */
+#include <linux/linkage.h>
+#include <asm-generic/export.h>
+#include <asm/asm.h>
+
+ENTRY(xor_regs_2_)
+ vsetvli a3, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a3
+ vxor.vv v16, v0, v8
+ add a2, a2, a3
+ vse8.v v16, (a1)
+ add a1, a1, a3
+ bnez a0, xor_regs_2_
+ ret
+END(xor_regs_2_)
+EXPORT_SYMBOL(xor_regs_2_)
+
+ENTRY(xor_regs_3_)
+ vsetvli a4, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a4
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a4
+ vxor.vv v16, v0, v16
+ add a3, a3, a4
+ vse8.v v16, (a1)
+ add a1, a1, a4
+ bnez a0, xor_regs_3_
+ ret
+END(xor_regs_3_)
+EXPORT_SYMBOL(xor_regs_3_)
+
+ENTRY(xor_regs_4_)
+ vsetvli a5, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a5
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a5
+ vxor.vv v0, v0, v16
+ vle8.v v24, (a4)
+ add a3, a3, a5
+ vxor.vv v16, v0, v24
+ add a4, a4, a5
+ vse8.v v16, (a1)
+ add a1, a1, a5
+ bnez a0, xor_regs_4_
+ ret
+END(xor_regs_4_)
+EXPORT_SYMBOL(xor_regs_4_)
+
+ENTRY(xor_regs_5_)
+ vsetvli a6, a0, e8, m8, ta, ma
+ vle8.v v0, (a1)
+ vle8.v v8, (a2)
+ sub a0, a0, a6
+ vxor.vv v0, v0, v8
+ vle8.v v16, (a3)
+ add a2, a2, a6
+ vxor.vv v0, v0, v16
+ vle8.v v24, (a4)
+ add a3, a3, a6
+ vxor.vv v0, v0, v24
+ vle8.v v8, (a5)
+ add a4, a4, a6
+ vxor.vv v16, v0, v8
+ add a5, a5, a6
+ vse8.v v16, (a1)
+ add a1, a1, a6
+ bnez a0, xor_regs_5_
+ ret
+END(xor_regs_5_)
+EXPORT_SYMBOL(xor_regs_5_)
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 14/17] riscv: Fix a kernel panic issue if $s2 is set to a specific value before entering Linux
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, ShihPo Hung, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Heinrich Schuchardt, Guo Ren,
Chris Stillson, Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti,
Qinglin Pan, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Nick Knight, Han-Kuan Chen, Liao Chang,
Changbin Du, Li Zhengyu, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Paul E. McKenney, Mark Rutland, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Mark Brown, Will Deacon,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Evgenii Stepanov, Colin Cross, Eugene Syromiatnikov,
Peter Collingbourne, Andrew Morton, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Panic log:
[ 0.018707] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
[ 0.023060] Oops [#1]
[ 0.023214] Modules linked in:
[ 0.023725] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.14.0 #33
[ 0.023955] Hardware name: SiFive,FU800 (DT)
[ 0.024150] epc : __vstate_save+0x1c/0x48
[ 0.024654] ra : arch_dup_task_struct+0x70/0x108
[ 0.024815] epc : ffffffff80005ad8 ra : ffffffff800035a8 sp : ffffffff81203d50
[ 0.025020] gp : ffffffff812e8290 tp : ffffffff8120bdc0 t0 : 0000000000000000
[ 0.025216] t1 : 0000000000000000 t2 : 0000000000000000 s0 : ffffffff81203d80
[ 0.025424] s1 : ffffffff8120bdc0 a0 : ffffffff8120c820 a1 : 0000000000000000
[ 0.025659] a2 : 0000000000001000 a3 : 0000000000000000 a4 : 0000000000000600
[ 0.025869] a5 : ffffffff8120cdc0 a6 : ffffffe00160b400 a7 : ffffffff80a1fe60
[ 0.026069] s2 : ffffffe0016b8000 s3 : ffffffff81204000 s4 : 0000000000004000
[ 0.026267] s5 : 0000000000000000 s6 : ffffffe0016b8000 s7 : ffffffe0016b9000
[ 0.026475] s8 : ffffffff81203ee0 s9 : 0000000000800300 s10: ffffffff812e9088
[ 0.026689] s11: ffffffd004008000 t3 : 0000000000000000 t4 : 0000000000000100
[ 0.026900] t5 : 0000000000000600 t6 : ffffffe00167bcc4
[ 0.027057] status: 8000000000000720 badaddr: 0000000000000000 cause: 000000000000000f
[ 0.027344] [<ffffffff80005ad8>] __vstate_save+0x1c/0x48
[ 0.027567] [<ffffffff8000abe8>] copy_process+0x266/0x11a0
[ 0.027739] [<ffffffff8000bc98>] kernel_clone+0x90/0x2aa
[ 0.027915] [<ffffffff8000c062>] kernel_thread+0x76/0x92
[ 0.028075] [<ffffffff8072e34c>] rest_init+0x26/0xfc
[ 0.028242] [<ffffffff80800638>] arch_call_rest_init+0x10/0x18
[ 0.028423] [<ffffffff80800c4a>] start_kernel+0x5ce/0x5fe
[ 0.029188] ---[ end trace 9a59af33f7ba3df4 ]---
[ 0.029479] Kernel panic - not syncing: Attempted to kill the idle task!
[ 0.029907] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---
The NULL pointer accessing caused the kernel panic. There is a NULL
pointer is because in vstate_save() function it will check
(regs->status & SR_VS) == SR_VS_DIRTY and this is true, but it shouldn't
be true because vector is not used here. Since vector is not used, datap
won't be allocated so it is NULL. The reason why regs->status is set to
a wrong value is because pt_regs->status is put in stack and it is polluted
after setup_vm() called.
In prologue of setup_vm(), we can observe it will save s2 to stack however
s2 is meaningless here because the caller is assembly code and s2 is just
some value from previous stage. The compiler will base on calling
convention to save the register to stack. Then 0x80008638 in s2 is saved
to stack. It might be any value. In this failure case it is 0x80008638 and
it will accidentally cause SR_VS_DIRTY to call the vstate_save() function.
(gdb) info addr setup_vm
Symbol "setup_vm" is a function at address 0xffffffff80802c8a.
(gdb) va2pa 0xffffffff80802c8a
$64 = 0x80a02c8a
(gdb) x/10i 0x80a02c8a
0x80a02c8a: addi sp,sp,-48
0x80a02c8c: li a3,-1
0x80a02c8e: auipc a5,0xff7fd
0x80a02c92: addi a5,a5,882
0x80a02c96: sd s0,32(sp)
0x80a02c98: sd s2,16(sp) <-- store to stack
After returning from setup_vm()
(gdb) x/20i 0x0000000080201138
0x80201138: mv a0,s1
0x8020113a: auipc ra,0x802
0x8020113e: jalr -1200(ra) <-- jump to setup_vm()
0x80201142: auipc a0,0xa03
(gdb) p/x $sp
$70 = 0x81404000
(gdb) p/x *(struct pt_regs*)($sp-0x120)
$71 = {
epc = 0x0,
ra = 0x0,
sp = 0x0,
gp = 0x0,
tp = 0x0,
t0 = 0x0,
t1 = 0x0,
t2 = 0x0,
s0 = 0x0,
s1 = 0x0,
a0 = 0x0,
a1 = 0x0,
a2 = 0x0,
a3 = 0x81403f90,
a4 = 0x80c04000,
a5 = 0x1,
a6 = 0xffffffff81337000,
a7 = 0x81096700,
s2 = 0x81400000,
s3 = 0xffffffff81200000,
s4 = 0x81403fd0,
s5 = 0x80a02c6c,
s6 = 0x8000000000006800,
s7 = 0x0,
s8 = 0xfffffffffffffff3,
s9 = 0x80c01000,
s10 = 0x81096700,
s11 = 0x82200000,
t3 = 0x81404000,
t4 = 0x80a02dea,
t5 = 0x0,
t6 = 0x82200000,
status = 0x80008638, <- Wrong value in stack!!!
badaddr = 0x82200000,
cause = 0x0,
orig_a0 = 0x80201142
}
(gdb) p/x $pc
$72 = 0x80201142
(gdb) p/x sizeof(struct pt_regs)
$73 = 0x120
Co-developed-by: ShihPo Hung <shihpo.hung@sifive.com>
Signed-off-by: ShihPo Hung <shihpo.hung@sifive.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/kernel/head.S | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index 2c81ca42ec4e..c7effef23f41 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -301,6 +301,7 @@ clear_bss_done:
la tp, init_task
la sp, init_thread_union + THREAD_SIZE
XIP_FIXUP_OFFSET sp
+ addi sp, sp, -PT_SIZE
#ifdef CONFIG_BUILTIN_DTB
la a0, __dtb_start
XIP_FIXUP_OFFSET a0
@@ -318,6 +319,7 @@ clear_bss_done:
/* Restore C environment */
la tp, init_task
la sp, init_thread_union + THREAD_SIZE
+ addi sp, sp, -PT_SIZE
#ifdef CONFIG_KASAN
call kasan_early_init
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 14/17] riscv: Fix a kernel panic issue if $s2 is set to a specific value before entering Linux
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, ShihPo Hung, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Heinrich Schuchardt, Guo Ren,
Chris Stillson, Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti,
Qinglin Pan, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Nick Knight, Han-Kuan Chen, Liao Chang,
Changbin Du, Li Zhengyu, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Paul E. McKenney, Mark Rutland, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Mark Brown, Will Deacon,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Evgenii Stepanov, Colin Cross, Eugene Syromiatnikov,
Peter Collingbourne, Andrew Morton, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Panic log:
[ 0.018707] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
[ 0.023060] Oops [#1]
[ 0.023214] Modules linked in:
[ 0.023725] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.14.0 #33
[ 0.023955] Hardware name: SiFive,FU800 (DT)
[ 0.024150] epc : __vstate_save+0x1c/0x48
[ 0.024654] ra : arch_dup_task_struct+0x70/0x108
[ 0.024815] epc : ffffffff80005ad8 ra : ffffffff800035a8 sp : ffffffff81203d50
[ 0.025020] gp : ffffffff812e8290 tp : ffffffff8120bdc0 t0 : 0000000000000000
[ 0.025216] t1 : 0000000000000000 t2 : 0000000000000000 s0 : ffffffff81203d80
[ 0.025424] s1 : ffffffff8120bdc0 a0 : ffffffff8120c820 a1 : 0000000000000000
[ 0.025659] a2 : 0000000000001000 a3 : 0000000000000000 a4 : 0000000000000600
[ 0.025869] a5 : ffffffff8120cdc0 a6 : ffffffe00160b400 a7 : ffffffff80a1fe60
[ 0.026069] s2 : ffffffe0016b8000 s3 : ffffffff81204000 s4 : 0000000000004000
[ 0.026267] s5 : 0000000000000000 s6 : ffffffe0016b8000 s7 : ffffffe0016b9000
[ 0.026475] s8 : ffffffff81203ee0 s9 : 0000000000800300 s10: ffffffff812e9088
[ 0.026689] s11: ffffffd004008000 t3 : 0000000000000000 t4 : 0000000000000100
[ 0.026900] t5 : 0000000000000600 t6 : ffffffe00167bcc4
[ 0.027057] status: 8000000000000720 badaddr: 0000000000000000 cause: 000000000000000f
[ 0.027344] [<ffffffff80005ad8>] __vstate_save+0x1c/0x48
[ 0.027567] [<ffffffff8000abe8>] copy_process+0x266/0x11a0
[ 0.027739] [<ffffffff8000bc98>] kernel_clone+0x90/0x2aa
[ 0.027915] [<ffffffff8000c062>] kernel_thread+0x76/0x92
[ 0.028075] [<ffffffff8072e34c>] rest_init+0x26/0xfc
[ 0.028242] [<ffffffff80800638>] arch_call_rest_init+0x10/0x18
[ 0.028423] [<ffffffff80800c4a>] start_kernel+0x5ce/0x5fe
[ 0.029188] ---[ end trace 9a59af33f7ba3df4 ]---
[ 0.029479] Kernel panic - not syncing: Attempted to kill the idle task!
[ 0.029907] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---
The NULL pointer accessing caused the kernel panic. There is a NULL
pointer is because in vstate_save() function it will check
(regs->status & SR_VS) == SR_VS_DIRTY and this is true, but it shouldn't
be true because vector is not used here. Since vector is not used, datap
won't be allocated so it is NULL. The reason why regs->status is set to
a wrong value is because pt_regs->status is put in stack and it is polluted
after setup_vm() called.
In prologue of setup_vm(), we can observe it will save s2 to stack however
s2 is meaningless here because the caller is assembly code and s2 is just
some value from previous stage. The compiler will base on calling
convention to save the register to stack. Then 0x80008638 in s2 is saved
to stack. It might be any value. In this failure case it is 0x80008638 and
it will accidentally cause SR_VS_DIRTY to call the vstate_save() function.
(gdb) info addr setup_vm
Symbol "setup_vm" is a function at address 0xffffffff80802c8a.
(gdb) va2pa 0xffffffff80802c8a
$64 = 0x80a02c8a
(gdb) x/10i 0x80a02c8a
0x80a02c8a: addi sp,sp,-48
0x80a02c8c: li a3,-1
0x80a02c8e: auipc a5,0xff7fd
0x80a02c92: addi a5,a5,882
0x80a02c96: sd s0,32(sp)
0x80a02c98: sd s2,16(sp) <-- store to stack
After returning from setup_vm()
(gdb) x/20i 0x0000000080201138
0x80201138: mv a0,s1
0x8020113a: auipc ra,0x802
0x8020113e: jalr -1200(ra) <-- jump to setup_vm()
0x80201142: auipc a0,0xa03
(gdb) p/x $sp
$70 = 0x81404000
(gdb) p/x *(struct pt_regs*)($sp-0x120)
$71 = {
epc = 0x0,
ra = 0x0,
sp = 0x0,
gp = 0x0,
tp = 0x0,
t0 = 0x0,
t1 = 0x0,
t2 = 0x0,
s0 = 0x0,
s1 = 0x0,
a0 = 0x0,
a1 = 0x0,
a2 = 0x0,
a3 = 0x81403f90,
a4 = 0x80c04000,
a5 = 0x1,
a6 = 0xffffffff81337000,
a7 = 0x81096700,
s2 = 0x81400000,
s3 = 0xffffffff81200000,
s4 = 0x81403fd0,
s5 = 0x80a02c6c,
s6 = 0x8000000000006800,
s7 = 0x0,
s8 = 0xfffffffffffffff3,
s9 = 0x80c01000,
s10 = 0x81096700,
s11 = 0x82200000,
t3 = 0x81404000,
t4 = 0x80a02dea,
t5 = 0x0,
t6 = 0x82200000,
status = 0x80008638, <- Wrong value in stack!!!
badaddr = 0x82200000,
cause = 0x0,
orig_a0 = 0x80201142
}
(gdb) p/x $pc
$72 = 0x80201142
(gdb) p/x sizeof(struct pt_regs)
$73 = 0x120
Co-developed-by: ShihPo Hung <shihpo.hung@sifive.com>
Signed-off-by: ShihPo Hung <shihpo.hung@sifive.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/kernel/head.S | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index 2c81ca42ec4e..c7effef23f41 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -301,6 +301,7 @@ clear_bss_done:
la tp, init_task
la sp, init_thread_union + THREAD_SIZE
XIP_FIXUP_OFFSET sp
+ addi sp, sp, -PT_SIZE
#ifdef CONFIG_BUILTIN_DTB
la a0, __dtb_start
XIP_FIXUP_OFFSET a0
@@ -318,6 +319,7 @@ clear_bss_done:
/* Restore C environment */
la tp, init_task
la sp, init_thread_union + THREAD_SIZE
+ addi sp, sp, -PT_SIZE
#ifdef CONFIG_KASAN
call kasan_early_init
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 14/17] riscv: Fix a kernel panic issue if $s2 is set to a specific value before entering Linux
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Greentime Hu, ShihPo Hung, Vincent Chen, Paul Walmsley,
Palmer Dabbelt, Albert Ou, Eric Biederman, Kees Cook, Anup Patel,
Atish Patra, Oleg Nesterov, Heinrich Schuchardt, Guo Ren,
Chris Stillson, Mayuresh Chitale, Paolo Bonzini, Alexandre Ghiti,
Qinglin Pan, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Nick Knight, Han-Kuan Chen, Liao Chang,
Changbin Du, Li Zhengyu, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Paul E. McKenney, Mark Rutland, Vitaly Wool,
Myrtle Shah, Catalin Marinas, Mark Brown, Will Deacon,
Alexey Dobriyan, Huacai Chen, Janosch Frank, Christian Brauner,
Evgenii Stepanov, Colin Cross, Eugene Syromiatnikov,
Peter Collingbourne, Andrew Morton, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Greentime Hu <greentime.hu@sifive.com>
Panic log:
[ 0.018707] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
[ 0.023060] Oops [#1]
[ 0.023214] Modules linked in:
[ 0.023725] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.14.0 #33
[ 0.023955] Hardware name: SiFive,FU800 (DT)
[ 0.024150] epc : __vstate_save+0x1c/0x48
[ 0.024654] ra : arch_dup_task_struct+0x70/0x108
[ 0.024815] epc : ffffffff80005ad8 ra : ffffffff800035a8 sp : ffffffff81203d50
[ 0.025020] gp : ffffffff812e8290 tp : ffffffff8120bdc0 t0 : 0000000000000000
[ 0.025216] t1 : 0000000000000000 t2 : 0000000000000000 s0 : ffffffff81203d80
[ 0.025424] s1 : ffffffff8120bdc0 a0 : ffffffff8120c820 a1 : 0000000000000000
[ 0.025659] a2 : 0000000000001000 a3 : 0000000000000000 a4 : 0000000000000600
[ 0.025869] a5 : ffffffff8120cdc0 a6 : ffffffe00160b400 a7 : ffffffff80a1fe60
[ 0.026069] s2 : ffffffe0016b8000 s3 : ffffffff81204000 s4 : 0000000000004000
[ 0.026267] s5 : 0000000000000000 s6 : ffffffe0016b8000 s7 : ffffffe0016b9000
[ 0.026475] s8 : ffffffff81203ee0 s9 : 0000000000800300 s10: ffffffff812e9088
[ 0.026689] s11: ffffffd004008000 t3 : 0000000000000000 t4 : 0000000000000100
[ 0.026900] t5 : 0000000000000600 t6 : ffffffe00167bcc4
[ 0.027057] status: 8000000000000720 badaddr: 0000000000000000 cause: 000000000000000f
[ 0.027344] [<ffffffff80005ad8>] __vstate_save+0x1c/0x48
[ 0.027567] [<ffffffff8000abe8>] copy_process+0x266/0x11a0
[ 0.027739] [<ffffffff8000bc98>] kernel_clone+0x90/0x2aa
[ 0.027915] [<ffffffff8000c062>] kernel_thread+0x76/0x92
[ 0.028075] [<ffffffff8072e34c>] rest_init+0x26/0xfc
[ 0.028242] [<ffffffff80800638>] arch_call_rest_init+0x10/0x18
[ 0.028423] [<ffffffff80800c4a>] start_kernel+0x5ce/0x5fe
[ 0.029188] ---[ end trace 9a59af33f7ba3df4 ]---
[ 0.029479] Kernel panic - not syncing: Attempted to kill the idle task!
[ 0.029907] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---
The NULL pointer accessing caused the kernel panic. There is a NULL
pointer is because in vstate_save() function it will check
(regs->status & SR_VS) == SR_VS_DIRTY and this is true, but it shouldn't
be true because vector is not used here. Since vector is not used, datap
won't be allocated so it is NULL. The reason why regs->status is set to
a wrong value is because pt_regs->status is put in stack and it is polluted
after setup_vm() called.
In prologue of setup_vm(), we can observe it will save s2 to stack however
s2 is meaningless here because the caller is assembly code and s2 is just
some value from previous stage. The compiler will base on calling
convention to save the register to stack. Then 0x80008638 in s2 is saved
to stack. It might be any value. In this failure case it is 0x80008638 and
it will accidentally cause SR_VS_DIRTY to call the vstate_save() function.
(gdb) info addr setup_vm
Symbol "setup_vm" is a function at address 0xffffffff80802c8a.
(gdb) va2pa 0xffffffff80802c8a
$64 = 0x80a02c8a
(gdb) x/10i 0x80a02c8a
0x80a02c8a: addi sp,sp,-48
0x80a02c8c: li a3,-1
0x80a02c8e: auipc a5,0xff7fd
0x80a02c92: addi a5,a5,882
0x80a02c96: sd s0,32(sp)
0x80a02c98: sd s2,16(sp) <-- store to stack
After returning from setup_vm()
(gdb) x/20i 0x0000000080201138
0x80201138: mv a0,s1
0x8020113a: auipc ra,0x802
0x8020113e: jalr -1200(ra) <-- jump to setup_vm()
0x80201142: auipc a0,0xa03
(gdb) p/x $sp
$70 = 0x81404000
(gdb) p/x *(struct pt_regs*)($sp-0x120)
$71 = {
epc = 0x0,
ra = 0x0,
sp = 0x0,
gp = 0x0,
tp = 0x0,
t0 = 0x0,
t1 = 0x0,
t2 = 0x0,
s0 = 0x0,
s1 = 0x0,
a0 = 0x0,
a1 = 0x0,
a2 = 0x0,
a3 = 0x81403f90,
a4 = 0x80c04000,
a5 = 0x1,
a6 = 0xffffffff81337000,
a7 = 0x81096700,
s2 = 0x81400000,
s3 = 0xffffffff81200000,
s4 = 0x81403fd0,
s5 = 0x80a02c6c,
s6 = 0x8000000000006800,
s7 = 0x0,
s8 = 0xfffffffffffffff3,
s9 = 0x80c01000,
s10 = 0x81096700,
s11 = 0x82200000,
t3 = 0x81404000,
t4 = 0x80a02dea,
t5 = 0x0,
t6 = 0x82200000,
status = 0x80008638, <- Wrong value in stack!!!
badaddr = 0x82200000,
cause = 0x0,
orig_a0 = 0x80201142
}
(gdb) p/x $pc
$72 = 0x80201142
(gdb) p/x sizeof(struct pt_regs)
$73 = 0x120
Co-developed-by: ShihPo Hung <shihpo.hung@sifive.com>
Signed-off-by: ShihPo Hung <shihpo.hung@sifive.com>
Co-developed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/kernel/head.S | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index 2c81ca42ec4e..c7effef23f41 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -301,6 +301,7 @@ clear_bss_done:
la tp, init_task
la sp, init_thread_union + THREAD_SIZE
XIP_FIXUP_OFFSET sp
+ addi sp, sp, -PT_SIZE
#ifdef CONFIG_BUILTIN_DTB
la a0, __dtb_start
XIP_FIXUP_OFFSET a0
@@ -318,6 +319,7 @@ clear_bss_done:
/* Restore C environment */
la tp, init_task
la sp, init_thread_union + THREAD_SIZE
+ addi sp, sp, -PT_SIZE
#ifdef CONFIG_KASAN
call kasan_early_init
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 15/17] riscv: Add V extension to KVM ISA allow list
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Conor Dooley,
Arnaud Pouliquen, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Nick Knight, Han-Kuan Chen, Changbin Du,
Li Zhengyu, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Vitaly Wool, Myrtle Shah,
Catalin Marinas, Will Deacon, Mark Brown, Janosch Frank,
Huacai Chen, Alexey Dobriyan, Christian Brauner,
Vincenzo Frascino, Eugene Syromiatnikov, Colin Cross,
Peter Collingbourne, Andrew Morton, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
Add V extension to KVM_RISCV_ISA_ALLOWED list to enable VCPU
to support V extension.
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/hwcap.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/riscv/include/asm/hwcap.h b/arch/riscv/include/asm/hwcap.h
index 6f59ec64175e..b242ed155262 100644
--- a/arch/riscv/include/asm/hwcap.h
+++ b/arch/riscv/include/asm/hwcap.h
@@ -35,6 +35,7 @@ extern unsigned long elf_hwcap;
#define RISCV_ISA_EXT_m ('m' - 'a')
#define RISCV_ISA_EXT_s ('s' - 'a')
#define RISCV_ISA_EXT_u ('u' - 'a')
+#define RISCV_ISA_EXT_v ('v' - 'a')
/*
* Increse this to higher value as kernel support more ISA extensions.
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 15/17] riscv: Add V extension to KVM ISA allow list
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Conor Dooley,
Arnaud Pouliquen, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Nick Knight, Han-Kuan Chen, Changbin Du,
Li Zhengyu, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Vitaly Wool, Myrtle Shah,
Catalin Marinas, Will Deacon, Mark Brown, Janosch Frank,
Huacai Chen, Alexey Dobriyan, Christian Brauner,
Vincenzo Frascino, Eugene Syromiatnikov, Colin Cross,
Peter Collingbourne, Andrew Morton, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
Add V extension to KVM_RISCV_ISA_ALLOWED list to enable VCPU
to support V extension.
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/hwcap.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/riscv/include/asm/hwcap.h b/arch/riscv/include/asm/hwcap.h
index 6f59ec64175e..b242ed155262 100644
--- a/arch/riscv/include/asm/hwcap.h
+++ b/arch/riscv/include/asm/hwcap.h
@@ -35,6 +35,7 @@ extern unsigned long elf_hwcap;
#define RISCV_ISA_EXT_m ('m' - 'a')
#define RISCV_ISA_EXT_s ('s' - 'a')
#define RISCV_ISA_EXT_u ('u' - 'a')
+#define RISCV_ISA_EXT_v ('v' - 'a')
/*
* Increse this to higher value as kernel support more ISA extensions.
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 15/17] riscv: Add V extension to KVM ISA allow list
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Heinrich Schuchardt, Guo Ren, Conor Dooley,
Arnaud Pouliquen, Chris Stillson, Paolo Bonzini, Qinglin Pan,
Alexandre Ghiti, Arnd Bergmann, Heiko Stuebner, Jisheng Zhang,
Dao Lu, Sunil V L, Nick Knight, Han-Kuan Chen, Changbin Du,
Li Zhengyu, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Frederic Weisbecker, Mark Rutland, Vitaly Wool, Myrtle Shah,
Catalin Marinas, Will Deacon, Mark Brown, Janosch Frank,
Huacai Chen, Alexey Dobriyan, Christian Brauner,
Vincenzo Frascino, Eugene Syromiatnikov, Colin Cross,
Peter Collingbourne, Andrew Morton, Barret Rhoden,
Suren Baghdasaryan, Davidlohr Bueso, linux-riscv, linux-kernel,
linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
Add V extension to KVM_RISCV_ISA_ALLOWED list to enable VCPU
to support V extension.
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/hwcap.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/riscv/include/asm/hwcap.h b/arch/riscv/include/asm/hwcap.h
index 6f59ec64175e..b242ed155262 100644
--- a/arch/riscv/include/asm/hwcap.h
+++ b/arch/riscv/include/asm/hwcap.h
@@ -35,6 +35,7 @@ extern unsigned long elf_hwcap;
#define RISCV_ISA_EXT_m ('m' - 'a')
#define RISCV_ISA_EXT_s ('s' - 'a')
#define RISCV_ISA_EXT_u ('u' - 'a')
+#define RISCV_ISA_EXT_v ('v' - 'a')
/*
* Increse this to higher value as kernel support more ISA extensions.
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 16/17] riscv: KVM: Add vector lazy save/restore support
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Chris Stillson,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Dao Lu, Jisheng Zhang, Sunil V L,
Nick Knight, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Frederic Weisbecker, Mark Rutland,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Mark Brown,
Will Deacon, WANG Xuerui, Alexey Dobriyan, Huacai Chen,
Christian Brauner, Evgenii Stepanov, Eugene Syromiatnikov,
Colin Cross, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
This patch adds vector context save/restore for guest VCPUs. To reduce the
impact on KVM performance, the implementation imitates the FP context
switch mechanism to lazily store and restore the vector context only when
the kernel enters/exits the in-kernel run loop and not during the KVM
world switch.
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/kvm_host.h | 2 +
arch/riscv/include/asm/kvm_vcpu_vector.h | 65 +++++++++
arch/riscv/include/uapi/asm/kvm.h | 7 +
arch/riscv/kernel/asm-offsets.c | 7 +
arch/riscv/kvm/Makefile | 1 +
arch/riscv/kvm/vcpu.c | 32 +++++
arch/riscv/kvm/vcpu_switch.S | 69 +++++++++
arch/riscv/kvm/vcpu_vector.c | 173 +++++++++++++++++++++++
8 files changed, 356 insertions(+)
create mode 100644 arch/riscv/include/asm/kvm_vcpu_vector.h
create mode 100644 arch/riscv/kvm/vcpu_vector.c
diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
index 60c517e4d576..665ddb4cec62 100644
--- a/arch/riscv/include/asm/kvm_host.h
+++ b/arch/riscv/include/asm/kvm_host.h
@@ -17,6 +17,7 @@
#include <asm/hwcap.h>
#include <asm/kvm_vcpu_fp.h>
#include <asm/kvm_vcpu_insn.h>
+#include <asm/kvm_vcpu_vector.h>
#include <asm/kvm_vcpu_timer.h>
#define KVM_MAX_VCPUS 1024
@@ -143,6 +144,7 @@ struct kvm_cpu_context {
unsigned long sstatus;
unsigned long hstatus;
union __riscv_fp_state fp;
+ struct __riscv_v_state vector;
};
struct kvm_vcpu_csr {
diff --git a/arch/riscv/include/asm/kvm_vcpu_vector.h b/arch/riscv/include/asm/kvm_vcpu_vector.h
new file mode 100644
index 000000000000..1dcc1b2e05bb
--- /dev/null
+++ b/arch/riscv/include/asm/kvm_vcpu_vector.h
@@ -0,0 +1,65 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 SiFive
+ *
+ * Authors:
+ * Atish Patra <atish.patra@wdc.com>
+ * Anup Patel <anup.patel@wdc.com>
+ * Vincent Chen <vincent.chen@sifive.com>
+ * Greentime Hu <greentime.hu@sifive.com>
+ */
+
+#ifndef __KVM_VCPU_RISCV_VECTOR_H
+#define __KVM_VCPU_RISCV_VECTOR_H
+
+#include <linux/types.h>
+
+struct kvm_cpu_context;
+
+#ifdef CONFIG_VECTOR
+void __kvm_riscv_vector_save(struct kvm_cpu_context *context);
+void __kvm_riscv_vector_restore(struct kvm_cpu_context *context);
+void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu);
+void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa);
+void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa);
+void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx);
+void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx);
+void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu);
+#else
+static inline void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
+{
+}
+
+static inline void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+}
+
+static inline void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+}
+
+static inline void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx)
+{
+}
+
+static inline void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx)
+{
+}
+
+static inline void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu)
+{
+}
+#endif
+
+int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype);
+int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype);
+#endif
diff --git a/arch/riscv/include/uapi/asm/kvm.h b/arch/riscv/include/uapi/asm/kvm.h
index 7351417afd62..f4ba57b235a3 100644
--- a/arch/riscv/include/uapi/asm/kvm.h
+++ b/arch/riscv/include/uapi/asm/kvm.h
@@ -96,6 +96,7 @@ enum KVM_RISCV_ISA_EXT_ID {
KVM_RISCV_ISA_EXT_H,
KVM_RISCV_ISA_EXT_I,
KVM_RISCV_ISA_EXT_M,
+ KVM_RISCV_ISA_EXT_V,
KVM_RISCV_ISA_EXT_SVPBMT,
KVM_RISCV_ISA_EXT_SSTC,
KVM_RISCV_ISA_EXT_MAX,
@@ -145,6 +146,12 @@ enum KVM_RISCV_ISA_EXT_ID {
/* ISA Extension registers are mapped as type 7 */
#define KVM_REG_RISCV_ISA_EXT (0x07 << KVM_REG_RISCV_TYPE_SHIFT)
+/* V extension registers are mapped as type 7 */
+#define KVM_REG_RISCV_VECTOR (0x07 << KVM_REG_RISCV_TYPE_SHIFT)
+#define KVM_REG_RISCV_VECTOR_CSR_REG(name) \
+ (offsetof(struct __riscv_v_state, name) / sizeof(unsigned long))
+#define KVM_REG_RISCV_VECTOR_REG(n) \
+ ((n) + sizeof(struct __riscv_v_state) / sizeof(unsigned long))
#endif
#endif /* __LINUX_KVM_RISCV_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index 80316ef7bb78..2540b9146072 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -278,6 +278,13 @@ void asm_offsets(void)
OFFSET(KVM_ARCH_FP_D_F31, kvm_cpu_context, fp.d.f[31]);
OFFSET(KVM_ARCH_FP_D_FCSR, kvm_cpu_context, fp.d.fcsr);
+ /* V extension */
+
+ OFFSET(KVM_ARCH_VECTOR_VSTART, kvm_cpu_context, vector.vstart);
+ OFFSET(KVM_ARCH_VECTOR_VL, kvm_cpu_context, vector.vl);
+ OFFSET(KVM_ARCH_VECTOR_VTYPE, kvm_cpu_context, vector.vtype);
+ OFFSET(KVM_ARCH_VECTOR_VCSR, kvm_cpu_context, vector.vcsr);
+ OFFSET(KVM_ARCH_VECTOR_DATAP, kvm_cpu_context, vector.datap);
/*
* THREAD_{F,X}* might be larger than a S-type offset can handle, but
* these are used in performance-sensitive assembly so we can't resort
diff --git a/arch/riscv/kvm/Makefile b/arch/riscv/kvm/Makefile
index 019df9208bdd..b26bc605a267 100644
--- a/arch/riscv/kvm/Makefile
+++ b/arch/riscv/kvm/Makefile
@@ -17,6 +17,7 @@ kvm-y += mmu.o
kvm-y += vcpu.o
kvm-y += vcpu_exit.o
kvm-y += vcpu_fp.o
+kvm-y += vcpu_vector.o
kvm-y += vcpu_insn.o
kvm-y += vcpu_switch.o
kvm-y += vcpu_sbi.o
diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
index d0f08d5b4282..76941937e745 100644
--- a/arch/riscv/kvm/vcpu.c
+++ b/arch/riscv/kvm/vcpu.c
@@ -19,6 +19,7 @@
#include <linux/kvm_host.h>
#include <asm/csr.h>
#include <asm/hwcap.h>
+#include <asm/switch_to.h>
const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
KVM_GENERIC_VCPU_STATS(),
@@ -51,6 +52,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
RISCV_ISA_EXT_h,
RISCV_ISA_EXT_i,
RISCV_ISA_EXT_m,
+ RISCV_ISA_EXT_v,
RISCV_ISA_EXT_SVPBMT,
RISCV_ISA_EXT_SSTC,
};
@@ -79,6 +81,7 @@ static bool kvm_riscv_vcpu_isa_enable_allowed(unsigned long ext)
return true;
}
+//CMS FIXME
static bool kvm_riscv_vcpu_isa_disable_allowed(unsigned long ext)
{
switch (ext) {
@@ -121,6 +124,8 @@ static void kvm_riscv_reset_vcpu(struct kvm_vcpu *vcpu)
kvm_riscv_vcpu_fp_reset(vcpu);
+ kvm_riscv_vcpu_vector_reset(vcpu);
+
kvm_riscv_vcpu_timer_reset(vcpu);
WRITE_ONCE(vcpu->arch.irqs_pending, 0);
@@ -171,6 +176,15 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
cntx->hstatus |= HSTATUS_SPVP;
cntx->hstatus |= HSTATUS_SPV;
+ if (has_vector()) {
+ cntx->vector.datap = kmalloc(riscv_vsize, GFP_KERNEL);
+ if (!cntx->vector.datap)
+ return -ENOMEM;
+ vcpu->arch.host_context.vector.datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ if (!vcpu->arch.host_context.vector.datap)
+ return -ENOMEM;
+ }
+
/* By default, make CY, TM, and IR counters accessible in VU mode */
reset_csr->scounteren = 0x7;
@@ -201,6 +215,9 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
/* Free unused pages pre-allocated for G-stage page table mappings */
kvm_mmu_free_memory_cache(&vcpu->arch.mmu_page_cache);
+
+ /* Free vector context space for host and guest kernel */
+ kvm_riscv_vcpu_free_vector_context(vcpu);
}
int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
@@ -539,6 +556,9 @@ static int kvm_riscv_vcpu_set_reg(struct kvm_vcpu *vcpu,
KVM_REG_RISCV_FP_D);
else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_ISA_EXT)
return kvm_riscv_vcpu_set_reg_isa_ext(vcpu, reg);
+ else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_VECTOR)
+ return kvm_riscv_vcpu_set_reg_vector(vcpu, reg,
+ KVM_REG_RISCV_VECTOR);
return -EINVAL;
}
@@ -562,6 +582,9 @@ static int kvm_riscv_vcpu_get_reg(struct kvm_vcpu *vcpu,
KVM_REG_RISCV_FP_D);
else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_ISA_EXT)
return kvm_riscv_vcpu_get_reg_isa_ext(vcpu, reg);
+ else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_VECTOR)
+ return kvm_riscv_vcpu_get_reg_vector(vcpu, reg,
+ KVM_REG_RISCV_VECTOR);
return -EINVAL;
}
@@ -818,6 +841,9 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
kvm_riscv_vcpu_host_fp_save(&vcpu->arch.host_context);
kvm_riscv_vcpu_guest_fp_restore(&vcpu->arch.guest_context,
vcpu->arch.isa);
+ kvm_riscv_vcpu_host_vector_save(&vcpu->arch.host_context);
+ kvm_riscv_vcpu_guest_vector_restore(&vcpu->arch.guest_context,
+ vcpu->arch.isa);
vcpu->cpu = cpu;
}
@@ -834,6 +860,12 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
kvm_riscv_vcpu_timer_save(vcpu);
+ kvm_riscv_vcpu_guest_vector_save(&vcpu->arch.guest_context,
+ vcpu->arch.isa);
+ kvm_riscv_vcpu_host_vector_restore(&vcpu->arch.host_context);
+
+ csr_write(CSR_HGATP, 0);
+
csr->vsstatus = csr_read(CSR_VSSTATUS);
csr->vsie = csr_read(CSR_VSIE);
csr->vstvec = csr_read(CSR_VSTVEC);
diff --git a/arch/riscv/kvm/vcpu_switch.S b/arch/riscv/kvm/vcpu_switch.S
index d74df8eb4d71..730dc9b8c644 100644
--- a/arch/riscv/kvm/vcpu_switch.S
+++ b/arch/riscv/kvm/vcpu_switch.S
@@ -406,3 +406,72 @@ __kvm_riscv_fp_d_restore:
csrw CSR_SSTATUS, t2
ret
#endif
+
+#ifdef CONFIG_VECTOR
+
+#define vstatep a0
+#define datap a1
+#define x_vstart t0
+#define x_vtype t1
+#define x_vl t2
+#define x_vcsr t3
+#define incr t4
+#define status t5
+
+ENTRY(__kvm_riscv_vector_save)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ li a2, KVM_ARCH_VECTOR_DATAP
+ add datap, a0, a2
+ ld datap, (datap)
+ csrr x_vstart, CSR_VSTART
+ csrr x_vtype, CSR_VTYPE
+ csrr x_vl, CSR_VL
+ csrr x_vcsr, CSR_VCSR
+ vsetvli incr, x0, e8, m8, ta, ma
+ vse8.v v0, (datap)
+ add datap, datap, incr
+ vse8.v v8, (datap)
+ add datap, datap, incr
+ vse8.v v16, (datap)
+ add datap, datap, incr
+ vse8.v v24, (datap)
+
+ REG_S x_vstart, KVM_ARCH_VECTOR_VSTART(vstatep)
+ REG_S x_vtype, KVM_ARCH_VECTOR_VTYPE(vstatep)
+ REG_S x_vl, KVM_ARCH_VECTOR_VL(vstatep)
+ REG_S x_vcsr, KVM_ARCH_VECTOR_VCSR(vstatep)
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__kvm_riscv_vector_save)
+
+ENTRY(__kvm_riscv_vector_restore)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ li a2, KVM_ARCH_VECTOR_DATAP
+ add datap, a0, a2
+ ld datap, (datap)
+ vsetvli incr, x0, e8, m8, ta, ma
+ vle8.v v0, (datap)
+ add datap, datap, incr
+ vle8.v v8, (datap)
+ add datap, datap, incr
+ vle8.v v16, (datap)
+ add datap, datap, incr
+ vle8.v v24, (datap)
+
+ REG_L x_vstart, KVM_ARCH_VECTOR_VSTART(vstatep)
+ REG_L x_vtype, KVM_ARCH_VECTOR_VTYPE(vstatep)
+ REG_L x_vl, KVM_ARCH_VECTOR_VL(vstatep)
+ REG_L x_vcsr, KVM_ARCH_VECTOR_VCSR(vstatep)
+ vsetvl x0, x_vl, x_vtype
+ csrw CSR_VSTART, x_vstart
+ csrw CSR_VCSR, x_vcsr
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__kvm_riscv_vector_restore)
+#endif
diff --git a/arch/riscv/kvm/vcpu_vector.c b/arch/riscv/kvm/vcpu_vector.c
new file mode 100644
index 000000000000..37bf4ffd47dd
--- /dev/null
+++ b/arch/riscv/kvm/vcpu_vector.c
@@ -0,0 +1,173 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 SiFive
+ *
+ * Authors:
+ * Atish Patra <atish.patra@wdc.com>
+ * Anup Patel <anup.patel@wdc.com>
+ * Vincent Chen <vincent.chen@sifive.com>
+ * Greentime Hu <greentime.hu@sifive.com>
+ */
+
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/kvm_host.h>
+#include <linux/uaccess.h>
+#include <asm/hwcap.h>
+
+#ifdef CONFIG_VECTOR
+extern unsigned long riscv_vsize;
+void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
+{
+ unsigned long isa = vcpu->arch.isa;
+ struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
+
+ cntx->sstatus &= ~SR_VS;
+ if (riscv_isa_extension_available(&isa, v))
+ cntx->sstatus |= SR_VS_INITIAL;
+ else
+ cntx->sstatus |= SR_VS_OFF;
+
+ memset(cntx->vector.datap, 0, riscv_vsize);
+}
+
+static void kvm_riscv_vcpu_vector_clean(struct kvm_cpu_context *cntx)
+{
+ cntx->sstatus &= ~SR_VS;
+ cntx->sstatus |= SR_VS_CLEAN;
+}
+
+void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+ if ((cntx->sstatus & SR_VS) == SR_VS_DIRTY) {
+ if (riscv_isa_extension_available(&isa, v))
+ __kvm_riscv_vector_save(cntx);
+ kvm_riscv_vcpu_vector_clean(cntx);
+ }
+}
+
+void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+ if ((cntx->sstatus & SR_VS) != SR_VS_OFF) {
+ if (riscv_isa_extension_available(&isa, v))
+ __kvm_riscv_vector_restore(cntx);
+ kvm_riscv_vcpu_vector_clean(cntx);
+ }
+}
+
+void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx)
+{
+ /* No need to check host sstatus as it can be modified outside */
+ __kvm_riscv_vector_save(cntx);
+}
+
+void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx)
+{
+ __kvm_riscv_vector_restore(cntx);
+}
+
+void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu)
+{
+ kfree(vcpu->arch.guest_reset_context.vector.datap);
+ kfree(vcpu->arch.host_context.vector.datap);
+}
+#else
+#define riscv_vsize (0)
+#endif
+
+static void *kvm_riscv_vcpu_vreg_addr(struct kvm_vcpu *vcpu,
+ unsigned long reg_num,
+ size_t reg_size)
+{
+ struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
+ void *reg_val;
+ size_t vlenb = riscv_vsize / 32;
+
+ if (reg_num < KVM_REG_RISCV_VECTOR_REG(0)) {
+ if (reg_size != sizeof(unsigned long))
+ return NULL;
+ switch (reg_num) {
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vstart):
+ reg_val = &cntx->vector.vstart;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vl):
+ reg_val = &cntx->vector.vl;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vtype):
+ reg_val = &cntx->vector.vtype;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vcsr):
+ reg_val = &cntx->vector.vcsr;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(datap):
+ default:
+ return NULL;
+ }
+ } else if (reg_num <= KVM_REG_RISCV_VECTOR_REG(31)) {
+ if (reg_size != vlenb)
+ return NULL;
+ reg_val = cntx->vector.datap
+ + (reg_num - KVM_REG_RISCV_VECTOR_REG(0)) * vlenb;
+ } else {
+ return NULL;
+ }
+
+ return reg_val;
+}
+
+int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype)
+{
+ unsigned long isa = vcpu->arch.isa;
+ unsigned long __user *uaddr =
+ (unsigned long __user *)(unsigned long)reg->addr;
+ unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
+ KVM_REG_SIZE_MASK |
+ rtype);
+ void *reg_val;
+ size_t reg_size = KVM_REG_SIZE(reg->id);
+
+ if ((rtype == KVM_REG_RISCV_VECTOR) &&
+ riscv_isa_extension_available(&isa, v)) {
+ reg_val = kvm_riscv_vcpu_vreg_addr(vcpu, reg_num, reg_size);
+ }
+
+ if (!reg_val)
+ return -EINVAL;
+
+ if (copy_to_user(uaddr, reg_val, reg_size))
+ return -EFAULT;
+
+ return 0;
+}
+
+int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype)
+{
+ unsigned long isa = vcpu->arch.isa;
+ unsigned long __user *uaddr =
+ (unsigned long __user *)(unsigned long)reg->addr;
+ unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
+ KVM_REG_SIZE_MASK |
+ rtype);
+ void *reg_val = NULL;
+ size_t reg_size = KVM_REG_SIZE(reg->id);
+
+ if ((rtype == KVM_REG_RISCV_VECTOR) &&
+ riscv_isa_extension_available(&isa, v)) {
+ reg_val = kvm_riscv_vcpu_vreg_addr(vcpu, reg_num, reg_size);
+ }
+
+ if (!reg_val)
+ return -EINVAL;
+
+ if (copy_from_user(reg_val, uaddr, reg_size))
+ return -EFAULT;
+
+ return 0;
+}
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 16/17] riscv: KVM: Add vector lazy save/restore support
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Chris Stillson,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Dao Lu, Jisheng Zhang, Sunil V L,
Nick Knight, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Frederic Weisbecker, Mark Rutland,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Mark Brown,
Will Deacon, WANG Xuerui, Alexey Dobriyan, Huacai Chen,
Christian Brauner, Evgenii Stepanov, Eugene Syromiatnikov,
Colin Cross, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
This patch adds vector context save/restore for guest VCPUs. To reduce the
impact on KVM performance, the implementation imitates the FP context
switch mechanism to lazily store and restore the vector context only when
the kernel enters/exits the in-kernel run loop and not during the KVM
world switch.
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/kvm_host.h | 2 +
arch/riscv/include/asm/kvm_vcpu_vector.h | 65 +++++++++
arch/riscv/include/uapi/asm/kvm.h | 7 +
arch/riscv/kernel/asm-offsets.c | 7 +
arch/riscv/kvm/Makefile | 1 +
arch/riscv/kvm/vcpu.c | 32 +++++
arch/riscv/kvm/vcpu_switch.S | 69 +++++++++
arch/riscv/kvm/vcpu_vector.c | 173 +++++++++++++++++++++++
8 files changed, 356 insertions(+)
create mode 100644 arch/riscv/include/asm/kvm_vcpu_vector.h
create mode 100644 arch/riscv/kvm/vcpu_vector.c
diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
index 60c517e4d576..665ddb4cec62 100644
--- a/arch/riscv/include/asm/kvm_host.h
+++ b/arch/riscv/include/asm/kvm_host.h
@@ -17,6 +17,7 @@
#include <asm/hwcap.h>
#include <asm/kvm_vcpu_fp.h>
#include <asm/kvm_vcpu_insn.h>
+#include <asm/kvm_vcpu_vector.h>
#include <asm/kvm_vcpu_timer.h>
#define KVM_MAX_VCPUS 1024
@@ -143,6 +144,7 @@ struct kvm_cpu_context {
unsigned long sstatus;
unsigned long hstatus;
union __riscv_fp_state fp;
+ struct __riscv_v_state vector;
};
struct kvm_vcpu_csr {
diff --git a/arch/riscv/include/asm/kvm_vcpu_vector.h b/arch/riscv/include/asm/kvm_vcpu_vector.h
new file mode 100644
index 000000000000..1dcc1b2e05bb
--- /dev/null
+++ b/arch/riscv/include/asm/kvm_vcpu_vector.h
@@ -0,0 +1,65 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 SiFive
+ *
+ * Authors:
+ * Atish Patra <atish.patra@wdc.com>
+ * Anup Patel <anup.patel@wdc.com>
+ * Vincent Chen <vincent.chen@sifive.com>
+ * Greentime Hu <greentime.hu@sifive.com>
+ */
+
+#ifndef __KVM_VCPU_RISCV_VECTOR_H
+#define __KVM_VCPU_RISCV_VECTOR_H
+
+#include <linux/types.h>
+
+struct kvm_cpu_context;
+
+#ifdef CONFIG_VECTOR
+void __kvm_riscv_vector_save(struct kvm_cpu_context *context);
+void __kvm_riscv_vector_restore(struct kvm_cpu_context *context);
+void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu);
+void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa);
+void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa);
+void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx);
+void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx);
+void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu);
+#else
+static inline void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
+{
+}
+
+static inline void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+}
+
+static inline void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+}
+
+static inline void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx)
+{
+}
+
+static inline void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx)
+{
+}
+
+static inline void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu)
+{
+}
+#endif
+
+int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype);
+int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype);
+#endif
diff --git a/arch/riscv/include/uapi/asm/kvm.h b/arch/riscv/include/uapi/asm/kvm.h
index 7351417afd62..f4ba57b235a3 100644
--- a/arch/riscv/include/uapi/asm/kvm.h
+++ b/arch/riscv/include/uapi/asm/kvm.h
@@ -96,6 +96,7 @@ enum KVM_RISCV_ISA_EXT_ID {
KVM_RISCV_ISA_EXT_H,
KVM_RISCV_ISA_EXT_I,
KVM_RISCV_ISA_EXT_M,
+ KVM_RISCV_ISA_EXT_V,
KVM_RISCV_ISA_EXT_SVPBMT,
KVM_RISCV_ISA_EXT_SSTC,
KVM_RISCV_ISA_EXT_MAX,
@@ -145,6 +146,12 @@ enum KVM_RISCV_ISA_EXT_ID {
/* ISA Extension registers are mapped as type 7 */
#define KVM_REG_RISCV_ISA_EXT (0x07 << KVM_REG_RISCV_TYPE_SHIFT)
+/* V extension registers are mapped as type 7 */
+#define KVM_REG_RISCV_VECTOR (0x07 << KVM_REG_RISCV_TYPE_SHIFT)
+#define KVM_REG_RISCV_VECTOR_CSR_REG(name) \
+ (offsetof(struct __riscv_v_state, name) / sizeof(unsigned long))
+#define KVM_REG_RISCV_VECTOR_REG(n) \
+ ((n) + sizeof(struct __riscv_v_state) / sizeof(unsigned long))
#endif
#endif /* __LINUX_KVM_RISCV_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index 80316ef7bb78..2540b9146072 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -278,6 +278,13 @@ void asm_offsets(void)
OFFSET(KVM_ARCH_FP_D_F31, kvm_cpu_context, fp.d.f[31]);
OFFSET(KVM_ARCH_FP_D_FCSR, kvm_cpu_context, fp.d.fcsr);
+ /* V extension */
+
+ OFFSET(KVM_ARCH_VECTOR_VSTART, kvm_cpu_context, vector.vstart);
+ OFFSET(KVM_ARCH_VECTOR_VL, kvm_cpu_context, vector.vl);
+ OFFSET(KVM_ARCH_VECTOR_VTYPE, kvm_cpu_context, vector.vtype);
+ OFFSET(KVM_ARCH_VECTOR_VCSR, kvm_cpu_context, vector.vcsr);
+ OFFSET(KVM_ARCH_VECTOR_DATAP, kvm_cpu_context, vector.datap);
/*
* THREAD_{F,X}* might be larger than a S-type offset can handle, but
* these are used in performance-sensitive assembly so we can't resort
diff --git a/arch/riscv/kvm/Makefile b/arch/riscv/kvm/Makefile
index 019df9208bdd..b26bc605a267 100644
--- a/arch/riscv/kvm/Makefile
+++ b/arch/riscv/kvm/Makefile
@@ -17,6 +17,7 @@ kvm-y += mmu.o
kvm-y += vcpu.o
kvm-y += vcpu_exit.o
kvm-y += vcpu_fp.o
+kvm-y += vcpu_vector.o
kvm-y += vcpu_insn.o
kvm-y += vcpu_switch.o
kvm-y += vcpu_sbi.o
diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
index d0f08d5b4282..76941937e745 100644
--- a/arch/riscv/kvm/vcpu.c
+++ b/arch/riscv/kvm/vcpu.c
@@ -19,6 +19,7 @@
#include <linux/kvm_host.h>
#include <asm/csr.h>
#include <asm/hwcap.h>
+#include <asm/switch_to.h>
const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
KVM_GENERIC_VCPU_STATS(),
@@ -51,6 +52,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
RISCV_ISA_EXT_h,
RISCV_ISA_EXT_i,
RISCV_ISA_EXT_m,
+ RISCV_ISA_EXT_v,
RISCV_ISA_EXT_SVPBMT,
RISCV_ISA_EXT_SSTC,
};
@@ -79,6 +81,7 @@ static bool kvm_riscv_vcpu_isa_enable_allowed(unsigned long ext)
return true;
}
+//CMS FIXME
static bool kvm_riscv_vcpu_isa_disable_allowed(unsigned long ext)
{
switch (ext) {
@@ -121,6 +124,8 @@ static void kvm_riscv_reset_vcpu(struct kvm_vcpu *vcpu)
kvm_riscv_vcpu_fp_reset(vcpu);
+ kvm_riscv_vcpu_vector_reset(vcpu);
+
kvm_riscv_vcpu_timer_reset(vcpu);
WRITE_ONCE(vcpu->arch.irqs_pending, 0);
@@ -171,6 +176,15 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
cntx->hstatus |= HSTATUS_SPVP;
cntx->hstatus |= HSTATUS_SPV;
+ if (has_vector()) {
+ cntx->vector.datap = kmalloc(riscv_vsize, GFP_KERNEL);
+ if (!cntx->vector.datap)
+ return -ENOMEM;
+ vcpu->arch.host_context.vector.datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ if (!vcpu->arch.host_context.vector.datap)
+ return -ENOMEM;
+ }
+
/* By default, make CY, TM, and IR counters accessible in VU mode */
reset_csr->scounteren = 0x7;
@@ -201,6 +215,9 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
/* Free unused pages pre-allocated for G-stage page table mappings */
kvm_mmu_free_memory_cache(&vcpu->arch.mmu_page_cache);
+
+ /* Free vector context space for host and guest kernel */
+ kvm_riscv_vcpu_free_vector_context(vcpu);
}
int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
@@ -539,6 +556,9 @@ static int kvm_riscv_vcpu_set_reg(struct kvm_vcpu *vcpu,
KVM_REG_RISCV_FP_D);
else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_ISA_EXT)
return kvm_riscv_vcpu_set_reg_isa_ext(vcpu, reg);
+ else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_VECTOR)
+ return kvm_riscv_vcpu_set_reg_vector(vcpu, reg,
+ KVM_REG_RISCV_VECTOR);
return -EINVAL;
}
@@ -562,6 +582,9 @@ static int kvm_riscv_vcpu_get_reg(struct kvm_vcpu *vcpu,
KVM_REG_RISCV_FP_D);
else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_ISA_EXT)
return kvm_riscv_vcpu_get_reg_isa_ext(vcpu, reg);
+ else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_VECTOR)
+ return kvm_riscv_vcpu_get_reg_vector(vcpu, reg,
+ KVM_REG_RISCV_VECTOR);
return -EINVAL;
}
@@ -818,6 +841,9 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
kvm_riscv_vcpu_host_fp_save(&vcpu->arch.host_context);
kvm_riscv_vcpu_guest_fp_restore(&vcpu->arch.guest_context,
vcpu->arch.isa);
+ kvm_riscv_vcpu_host_vector_save(&vcpu->arch.host_context);
+ kvm_riscv_vcpu_guest_vector_restore(&vcpu->arch.guest_context,
+ vcpu->arch.isa);
vcpu->cpu = cpu;
}
@@ -834,6 +860,12 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
kvm_riscv_vcpu_timer_save(vcpu);
+ kvm_riscv_vcpu_guest_vector_save(&vcpu->arch.guest_context,
+ vcpu->arch.isa);
+ kvm_riscv_vcpu_host_vector_restore(&vcpu->arch.host_context);
+
+ csr_write(CSR_HGATP, 0);
+
csr->vsstatus = csr_read(CSR_VSSTATUS);
csr->vsie = csr_read(CSR_VSIE);
csr->vstvec = csr_read(CSR_VSTVEC);
diff --git a/arch/riscv/kvm/vcpu_switch.S b/arch/riscv/kvm/vcpu_switch.S
index d74df8eb4d71..730dc9b8c644 100644
--- a/arch/riscv/kvm/vcpu_switch.S
+++ b/arch/riscv/kvm/vcpu_switch.S
@@ -406,3 +406,72 @@ __kvm_riscv_fp_d_restore:
csrw CSR_SSTATUS, t2
ret
#endif
+
+#ifdef CONFIG_VECTOR
+
+#define vstatep a0
+#define datap a1
+#define x_vstart t0
+#define x_vtype t1
+#define x_vl t2
+#define x_vcsr t3
+#define incr t4
+#define status t5
+
+ENTRY(__kvm_riscv_vector_save)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ li a2, KVM_ARCH_VECTOR_DATAP
+ add datap, a0, a2
+ ld datap, (datap)
+ csrr x_vstart, CSR_VSTART
+ csrr x_vtype, CSR_VTYPE
+ csrr x_vl, CSR_VL
+ csrr x_vcsr, CSR_VCSR
+ vsetvli incr, x0, e8, m8, ta, ma
+ vse8.v v0, (datap)
+ add datap, datap, incr
+ vse8.v v8, (datap)
+ add datap, datap, incr
+ vse8.v v16, (datap)
+ add datap, datap, incr
+ vse8.v v24, (datap)
+
+ REG_S x_vstart, KVM_ARCH_VECTOR_VSTART(vstatep)
+ REG_S x_vtype, KVM_ARCH_VECTOR_VTYPE(vstatep)
+ REG_S x_vl, KVM_ARCH_VECTOR_VL(vstatep)
+ REG_S x_vcsr, KVM_ARCH_VECTOR_VCSR(vstatep)
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__kvm_riscv_vector_save)
+
+ENTRY(__kvm_riscv_vector_restore)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ li a2, KVM_ARCH_VECTOR_DATAP
+ add datap, a0, a2
+ ld datap, (datap)
+ vsetvli incr, x0, e8, m8, ta, ma
+ vle8.v v0, (datap)
+ add datap, datap, incr
+ vle8.v v8, (datap)
+ add datap, datap, incr
+ vle8.v v16, (datap)
+ add datap, datap, incr
+ vle8.v v24, (datap)
+
+ REG_L x_vstart, KVM_ARCH_VECTOR_VSTART(vstatep)
+ REG_L x_vtype, KVM_ARCH_VECTOR_VTYPE(vstatep)
+ REG_L x_vl, KVM_ARCH_VECTOR_VL(vstatep)
+ REG_L x_vcsr, KVM_ARCH_VECTOR_VCSR(vstatep)
+ vsetvl x0, x_vl, x_vtype
+ csrw CSR_VSTART, x_vstart
+ csrw CSR_VCSR, x_vcsr
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__kvm_riscv_vector_restore)
+#endif
diff --git a/arch/riscv/kvm/vcpu_vector.c b/arch/riscv/kvm/vcpu_vector.c
new file mode 100644
index 000000000000..37bf4ffd47dd
--- /dev/null
+++ b/arch/riscv/kvm/vcpu_vector.c
@@ -0,0 +1,173 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 SiFive
+ *
+ * Authors:
+ * Atish Patra <atish.patra@wdc.com>
+ * Anup Patel <anup.patel@wdc.com>
+ * Vincent Chen <vincent.chen@sifive.com>
+ * Greentime Hu <greentime.hu@sifive.com>
+ */
+
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/kvm_host.h>
+#include <linux/uaccess.h>
+#include <asm/hwcap.h>
+
+#ifdef CONFIG_VECTOR
+extern unsigned long riscv_vsize;
+void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
+{
+ unsigned long isa = vcpu->arch.isa;
+ struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
+
+ cntx->sstatus &= ~SR_VS;
+ if (riscv_isa_extension_available(&isa, v))
+ cntx->sstatus |= SR_VS_INITIAL;
+ else
+ cntx->sstatus |= SR_VS_OFF;
+
+ memset(cntx->vector.datap, 0, riscv_vsize);
+}
+
+static void kvm_riscv_vcpu_vector_clean(struct kvm_cpu_context *cntx)
+{
+ cntx->sstatus &= ~SR_VS;
+ cntx->sstatus |= SR_VS_CLEAN;
+}
+
+void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+ if ((cntx->sstatus & SR_VS) == SR_VS_DIRTY) {
+ if (riscv_isa_extension_available(&isa, v))
+ __kvm_riscv_vector_save(cntx);
+ kvm_riscv_vcpu_vector_clean(cntx);
+ }
+}
+
+void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+ if ((cntx->sstatus & SR_VS) != SR_VS_OFF) {
+ if (riscv_isa_extension_available(&isa, v))
+ __kvm_riscv_vector_restore(cntx);
+ kvm_riscv_vcpu_vector_clean(cntx);
+ }
+}
+
+void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx)
+{
+ /* No need to check host sstatus as it can be modified outside */
+ __kvm_riscv_vector_save(cntx);
+}
+
+void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx)
+{
+ __kvm_riscv_vector_restore(cntx);
+}
+
+void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu)
+{
+ kfree(vcpu->arch.guest_reset_context.vector.datap);
+ kfree(vcpu->arch.host_context.vector.datap);
+}
+#else
+#define riscv_vsize (0)
+#endif
+
+static void *kvm_riscv_vcpu_vreg_addr(struct kvm_vcpu *vcpu,
+ unsigned long reg_num,
+ size_t reg_size)
+{
+ struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
+ void *reg_val;
+ size_t vlenb = riscv_vsize / 32;
+
+ if (reg_num < KVM_REG_RISCV_VECTOR_REG(0)) {
+ if (reg_size != sizeof(unsigned long))
+ return NULL;
+ switch (reg_num) {
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vstart):
+ reg_val = &cntx->vector.vstart;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vl):
+ reg_val = &cntx->vector.vl;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vtype):
+ reg_val = &cntx->vector.vtype;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vcsr):
+ reg_val = &cntx->vector.vcsr;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(datap):
+ default:
+ return NULL;
+ }
+ } else if (reg_num <= KVM_REG_RISCV_VECTOR_REG(31)) {
+ if (reg_size != vlenb)
+ return NULL;
+ reg_val = cntx->vector.datap
+ + (reg_num - KVM_REG_RISCV_VECTOR_REG(0)) * vlenb;
+ } else {
+ return NULL;
+ }
+
+ return reg_val;
+}
+
+int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype)
+{
+ unsigned long isa = vcpu->arch.isa;
+ unsigned long __user *uaddr =
+ (unsigned long __user *)(unsigned long)reg->addr;
+ unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
+ KVM_REG_SIZE_MASK |
+ rtype);
+ void *reg_val;
+ size_t reg_size = KVM_REG_SIZE(reg->id);
+
+ if ((rtype == KVM_REG_RISCV_VECTOR) &&
+ riscv_isa_extension_available(&isa, v)) {
+ reg_val = kvm_riscv_vcpu_vreg_addr(vcpu, reg_num, reg_size);
+ }
+
+ if (!reg_val)
+ return -EINVAL;
+
+ if (copy_to_user(uaddr, reg_val, reg_size))
+ return -EFAULT;
+
+ return 0;
+}
+
+int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype)
+{
+ unsigned long isa = vcpu->arch.isa;
+ unsigned long __user *uaddr =
+ (unsigned long __user *)(unsigned long)reg->addr;
+ unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
+ KVM_REG_SIZE_MASK |
+ rtype);
+ void *reg_val = NULL;
+ size_t reg_size = KVM_REG_SIZE(reg->id);
+
+ if ((rtype == KVM_REG_RISCV_VECTOR) &&
+ riscv_isa_extension_available(&isa, v)) {
+ reg_val = kvm_riscv_vcpu_vreg_addr(vcpu, reg_num, reg_size);
+ }
+
+ if (!reg_val)
+ return -EINVAL;
+
+ if (copy_from_user(reg_val, uaddr, reg_size))
+ return -EFAULT;
+
+ return 0;
+}
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 16/17] riscv: KVM: Add vector lazy save/restore support
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Vincent Chen, Greentime Hu, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Chris Stillson,
Arnaud Pouliquen, Paolo Bonzini, Alexandre Ghiti, Qinglin Pan,
Arnd Bergmann, Heiko Stuebner, Dao Lu, Jisheng Zhang, Sunil V L,
Nick Knight, Han-Kuan Chen, Li Zhengyu, Changbin Du,
Alexander Graf, Ard Biesheuvel, Tsukasa OI, Yury Norov,
Nicolas Saenz Julienne, Frederic Weisbecker, Mark Rutland,
Vitaly Wool, Myrtle Shah, Catalin Marinas, Mark Brown,
Will Deacon, WANG Xuerui, Alexey Dobriyan, Huacai Chen,
Christian Brauner, Evgenii Stepanov, Eugene Syromiatnikov,
Colin Cross, Peter Collingbourne, Andrew Morton,
Suren Baghdasaryan, Barret Rhoden, Davidlohr Bueso, linux-riscv,
linux-kernel, linux-mm, kvm, kvm-riscv
From: Vincent Chen <vincent.chen@sifive.com>
This patch adds vector context save/restore for guest VCPUs. To reduce the
impact on KVM performance, the implementation imitates the FP context
switch mechanism to lazily store and restore the vector context only when
the kernel enters/exits the in-kernel run loop and not during the KVM
world switch.
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
---
arch/riscv/include/asm/kvm_host.h | 2 +
arch/riscv/include/asm/kvm_vcpu_vector.h | 65 +++++++++
arch/riscv/include/uapi/asm/kvm.h | 7 +
arch/riscv/kernel/asm-offsets.c | 7 +
arch/riscv/kvm/Makefile | 1 +
arch/riscv/kvm/vcpu.c | 32 +++++
arch/riscv/kvm/vcpu_switch.S | 69 +++++++++
arch/riscv/kvm/vcpu_vector.c | 173 +++++++++++++++++++++++
8 files changed, 356 insertions(+)
create mode 100644 arch/riscv/include/asm/kvm_vcpu_vector.h
create mode 100644 arch/riscv/kvm/vcpu_vector.c
diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
index 60c517e4d576..665ddb4cec62 100644
--- a/arch/riscv/include/asm/kvm_host.h
+++ b/arch/riscv/include/asm/kvm_host.h
@@ -17,6 +17,7 @@
#include <asm/hwcap.h>
#include <asm/kvm_vcpu_fp.h>
#include <asm/kvm_vcpu_insn.h>
+#include <asm/kvm_vcpu_vector.h>
#include <asm/kvm_vcpu_timer.h>
#define KVM_MAX_VCPUS 1024
@@ -143,6 +144,7 @@ struct kvm_cpu_context {
unsigned long sstatus;
unsigned long hstatus;
union __riscv_fp_state fp;
+ struct __riscv_v_state vector;
};
struct kvm_vcpu_csr {
diff --git a/arch/riscv/include/asm/kvm_vcpu_vector.h b/arch/riscv/include/asm/kvm_vcpu_vector.h
new file mode 100644
index 000000000000..1dcc1b2e05bb
--- /dev/null
+++ b/arch/riscv/include/asm/kvm_vcpu_vector.h
@@ -0,0 +1,65 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 SiFive
+ *
+ * Authors:
+ * Atish Patra <atish.patra@wdc.com>
+ * Anup Patel <anup.patel@wdc.com>
+ * Vincent Chen <vincent.chen@sifive.com>
+ * Greentime Hu <greentime.hu@sifive.com>
+ */
+
+#ifndef __KVM_VCPU_RISCV_VECTOR_H
+#define __KVM_VCPU_RISCV_VECTOR_H
+
+#include <linux/types.h>
+
+struct kvm_cpu_context;
+
+#ifdef CONFIG_VECTOR
+void __kvm_riscv_vector_save(struct kvm_cpu_context *context);
+void __kvm_riscv_vector_restore(struct kvm_cpu_context *context);
+void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu);
+void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa);
+void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa);
+void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx);
+void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx);
+void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu);
+#else
+static inline void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
+{
+}
+
+static inline void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+}
+
+static inline void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+}
+
+static inline void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx)
+{
+}
+
+static inline void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx)
+{
+}
+
+static inline void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu)
+{
+}
+#endif
+
+int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype);
+int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype);
+#endif
diff --git a/arch/riscv/include/uapi/asm/kvm.h b/arch/riscv/include/uapi/asm/kvm.h
index 7351417afd62..f4ba57b235a3 100644
--- a/arch/riscv/include/uapi/asm/kvm.h
+++ b/arch/riscv/include/uapi/asm/kvm.h
@@ -96,6 +96,7 @@ enum KVM_RISCV_ISA_EXT_ID {
KVM_RISCV_ISA_EXT_H,
KVM_RISCV_ISA_EXT_I,
KVM_RISCV_ISA_EXT_M,
+ KVM_RISCV_ISA_EXT_V,
KVM_RISCV_ISA_EXT_SVPBMT,
KVM_RISCV_ISA_EXT_SSTC,
KVM_RISCV_ISA_EXT_MAX,
@@ -145,6 +146,12 @@ enum KVM_RISCV_ISA_EXT_ID {
/* ISA Extension registers are mapped as type 7 */
#define KVM_REG_RISCV_ISA_EXT (0x07 << KVM_REG_RISCV_TYPE_SHIFT)
+/* V extension registers are mapped as type 7 */
+#define KVM_REG_RISCV_VECTOR (0x07 << KVM_REG_RISCV_TYPE_SHIFT)
+#define KVM_REG_RISCV_VECTOR_CSR_REG(name) \
+ (offsetof(struct __riscv_v_state, name) / sizeof(unsigned long))
+#define KVM_REG_RISCV_VECTOR_REG(n) \
+ ((n) + sizeof(struct __riscv_v_state) / sizeof(unsigned long))
#endif
#endif /* __LINUX_KVM_RISCV_H */
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c
index 80316ef7bb78..2540b9146072 100644
--- a/arch/riscv/kernel/asm-offsets.c
+++ b/arch/riscv/kernel/asm-offsets.c
@@ -278,6 +278,13 @@ void asm_offsets(void)
OFFSET(KVM_ARCH_FP_D_F31, kvm_cpu_context, fp.d.f[31]);
OFFSET(KVM_ARCH_FP_D_FCSR, kvm_cpu_context, fp.d.fcsr);
+ /* V extension */
+
+ OFFSET(KVM_ARCH_VECTOR_VSTART, kvm_cpu_context, vector.vstart);
+ OFFSET(KVM_ARCH_VECTOR_VL, kvm_cpu_context, vector.vl);
+ OFFSET(KVM_ARCH_VECTOR_VTYPE, kvm_cpu_context, vector.vtype);
+ OFFSET(KVM_ARCH_VECTOR_VCSR, kvm_cpu_context, vector.vcsr);
+ OFFSET(KVM_ARCH_VECTOR_DATAP, kvm_cpu_context, vector.datap);
/*
* THREAD_{F,X}* might be larger than a S-type offset can handle, but
* these are used in performance-sensitive assembly so we can't resort
diff --git a/arch/riscv/kvm/Makefile b/arch/riscv/kvm/Makefile
index 019df9208bdd..b26bc605a267 100644
--- a/arch/riscv/kvm/Makefile
+++ b/arch/riscv/kvm/Makefile
@@ -17,6 +17,7 @@ kvm-y += mmu.o
kvm-y += vcpu.o
kvm-y += vcpu_exit.o
kvm-y += vcpu_fp.o
+kvm-y += vcpu_vector.o
kvm-y += vcpu_insn.o
kvm-y += vcpu_switch.o
kvm-y += vcpu_sbi.o
diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
index d0f08d5b4282..76941937e745 100644
--- a/arch/riscv/kvm/vcpu.c
+++ b/arch/riscv/kvm/vcpu.c
@@ -19,6 +19,7 @@
#include <linux/kvm_host.h>
#include <asm/csr.h>
#include <asm/hwcap.h>
+#include <asm/switch_to.h>
const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
KVM_GENERIC_VCPU_STATS(),
@@ -51,6 +52,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
RISCV_ISA_EXT_h,
RISCV_ISA_EXT_i,
RISCV_ISA_EXT_m,
+ RISCV_ISA_EXT_v,
RISCV_ISA_EXT_SVPBMT,
RISCV_ISA_EXT_SSTC,
};
@@ -79,6 +81,7 @@ static bool kvm_riscv_vcpu_isa_enable_allowed(unsigned long ext)
return true;
}
+//CMS FIXME
static bool kvm_riscv_vcpu_isa_disable_allowed(unsigned long ext)
{
switch (ext) {
@@ -121,6 +124,8 @@ static void kvm_riscv_reset_vcpu(struct kvm_vcpu *vcpu)
kvm_riscv_vcpu_fp_reset(vcpu);
+ kvm_riscv_vcpu_vector_reset(vcpu);
+
kvm_riscv_vcpu_timer_reset(vcpu);
WRITE_ONCE(vcpu->arch.irqs_pending, 0);
@@ -171,6 +176,15 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
cntx->hstatus |= HSTATUS_SPVP;
cntx->hstatus |= HSTATUS_SPV;
+ if (has_vector()) {
+ cntx->vector.datap = kmalloc(riscv_vsize, GFP_KERNEL);
+ if (!cntx->vector.datap)
+ return -ENOMEM;
+ vcpu->arch.host_context.vector.datap = kzalloc(riscv_vsize, GFP_KERNEL);
+ if (!vcpu->arch.host_context.vector.datap)
+ return -ENOMEM;
+ }
+
/* By default, make CY, TM, and IR counters accessible in VU mode */
reset_csr->scounteren = 0x7;
@@ -201,6 +215,9 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
/* Free unused pages pre-allocated for G-stage page table mappings */
kvm_mmu_free_memory_cache(&vcpu->arch.mmu_page_cache);
+
+ /* Free vector context space for host and guest kernel */
+ kvm_riscv_vcpu_free_vector_context(vcpu);
}
int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
@@ -539,6 +556,9 @@ static int kvm_riscv_vcpu_set_reg(struct kvm_vcpu *vcpu,
KVM_REG_RISCV_FP_D);
else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_ISA_EXT)
return kvm_riscv_vcpu_set_reg_isa_ext(vcpu, reg);
+ else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_VECTOR)
+ return kvm_riscv_vcpu_set_reg_vector(vcpu, reg,
+ KVM_REG_RISCV_VECTOR);
return -EINVAL;
}
@@ -562,6 +582,9 @@ static int kvm_riscv_vcpu_get_reg(struct kvm_vcpu *vcpu,
KVM_REG_RISCV_FP_D);
else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_ISA_EXT)
return kvm_riscv_vcpu_get_reg_isa_ext(vcpu, reg);
+ else if ((reg->id & KVM_REG_RISCV_TYPE_MASK) == KVM_REG_RISCV_VECTOR)
+ return kvm_riscv_vcpu_get_reg_vector(vcpu, reg,
+ KVM_REG_RISCV_VECTOR);
return -EINVAL;
}
@@ -818,6 +841,9 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
kvm_riscv_vcpu_host_fp_save(&vcpu->arch.host_context);
kvm_riscv_vcpu_guest_fp_restore(&vcpu->arch.guest_context,
vcpu->arch.isa);
+ kvm_riscv_vcpu_host_vector_save(&vcpu->arch.host_context);
+ kvm_riscv_vcpu_guest_vector_restore(&vcpu->arch.guest_context,
+ vcpu->arch.isa);
vcpu->cpu = cpu;
}
@@ -834,6 +860,12 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
kvm_riscv_vcpu_timer_save(vcpu);
+ kvm_riscv_vcpu_guest_vector_save(&vcpu->arch.guest_context,
+ vcpu->arch.isa);
+ kvm_riscv_vcpu_host_vector_restore(&vcpu->arch.host_context);
+
+ csr_write(CSR_HGATP, 0);
+
csr->vsstatus = csr_read(CSR_VSSTATUS);
csr->vsie = csr_read(CSR_VSIE);
csr->vstvec = csr_read(CSR_VSTVEC);
diff --git a/arch/riscv/kvm/vcpu_switch.S b/arch/riscv/kvm/vcpu_switch.S
index d74df8eb4d71..730dc9b8c644 100644
--- a/arch/riscv/kvm/vcpu_switch.S
+++ b/arch/riscv/kvm/vcpu_switch.S
@@ -406,3 +406,72 @@ __kvm_riscv_fp_d_restore:
csrw CSR_SSTATUS, t2
ret
#endif
+
+#ifdef CONFIG_VECTOR
+
+#define vstatep a0
+#define datap a1
+#define x_vstart t0
+#define x_vtype t1
+#define x_vl t2
+#define x_vcsr t3
+#define incr t4
+#define status t5
+
+ENTRY(__kvm_riscv_vector_save)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ li a2, KVM_ARCH_VECTOR_DATAP
+ add datap, a0, a2
+ ld datap, (datap)
+ csrr x_vstart, CSR_VSTART
+ csrr x_vtype, CSR_VTYPE
+ csrr x_vl, CSR_VL
+ csrr x_vcsr, CSR_VCSR
+ vsetvli incr, x0, e8, m8, ta, ma
+ vse8.v v0, (datap)
+ add datap, datap, incr
+ vse8.v v8, (datap)
+ add datap, datap, incr
+ vse8.v v16, (datap)
+ add datap, datap, incr
+ vse8.v v24, (datap)
+
+ REG_S x_vstart, KVM_ARCH_VECTOR_VSTART(vstatep)
+ REG_S x_vtype, KVM_ARCH_VECTOR_VTYPE(vstatep)
+ REG_S x_vl, KVM_ARCH_VECTOR_VL(vstatep)
+ REG_S x_vcsr, KVM_ARCH_VECTOR_VCSR(vstatep)
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__kvm_riscv_vector_save)
+
+ENTRY(__kvm_riscv_vector_restore)
+ li status, SR_VS
+ csrs CSR_STATUS, status
+
+ li a2, KVM_ARCH_VECTOR_DATAP
+ add datap, a0, a2
+ ld datap, (datap)
+ vsetvli incr, x0, e8, m8, ta, ma
+ vle8.v v0, (datap)
+ add datap, datap, incr
+ vle8.v v8, (datap)
+ add datap, datap, incr
+ vle8.v v16, (datap)
+ add datap, datap, incr
+ vle8.v v24, (datap)
+
+ REG_L x_vstart, KVM_ARCH_VECTOR_VSTART(vstatep)
+ REG_L x_vtype, KVM_ARCH_VECTOR_VTYPE(vstatep)
+ REG_L x_vl, KVM_ARCH_VECTOR_VL(vstatep)
+ REG_L x_vcsr, KVM_ARCH_VECTOR_VCSR(vstatep)
+ vsetvl x0, x_vl, x_vtype
+ csrw CSR_VSTART, x_vstart
+ csrw CSR_VCSR, x_vcsr
+
+ csrc CSR_STATUS, status
+ ret
+ENDPROC(__kvm_riscv_vector_restore)
+#endif
diff --git a/arch/riscv/kvm/vcpu_vector.c b/arch/riscv/kvm/vcpu_vector.c
new file mode 100644
index 000000000000..37bf4ffd47dd
--- /dev/null
+++ b/arch/riscv/kvm/vcpu_vector.c
@@ -0,0 +1,173 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 SiFive
+ *
+ * Authors:
+ * Atish Patra <atish.patra@wdc.com>
+ * Anup Patel <anup.patel@wdc.com>
+ * Vincent Chen <vincent.chen@sifive.com>
+ * Greentime Hu <greentime.hu@sifive.com>
+ */
+
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/kvm_host.h>
+#include <linux/uaccess.h>
+#include <asm/hwcap.h>
+
+#ifdef CONFIG_VECTOR
+extern unsigned long riscv_vsize;
+void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
+{
+ unsigned long isa = vcpu->arch.isa;
+ struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
+
+ cntx->sstatus &= ~SR_VS;
+ if (riscv_isa_extension_available(&isa, v))
+ cntx->sstatus |= SR_VS_INITIAL;
+ else
+ cntx->sstatus |= SR_VS_OFF;
+
+ memset(cntx->vector.datap, 0, riscv_vsize);
+}
+
+static void kvm_riscv_vcpu_vector_clean(struct kvm_cpu_context *cntx)
+{
+ cntx->sstatus &= ~SR_VS;
+ cntx->sstatus |= SR_VS_CLEAN;
+}
+
+void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+ if ((cntx->sstatus & SR_VS) == SR_VS_DIRTY) {
+ if (riscv_isa_extension_available(&isa, v))
+ __kvm_riscv_vector_save(cntx);
+ kvm_riscv_vcpu_vector_clean(cntx);
+ }
+}
+
+void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
+ unsigned long isa)
+{
+ if ((cntx->sstatus & SR_VS) != SR_VS_OFF) {
+ if (riscv_isa_extension_available(&isa, v))
+ __kvm_riscv_vector_restore(cntx);
+ kvm_riscv_vcpu_vector_clean(cntx);
+ }
+}
+
+void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx)
+{
+ /* No need to check host sstatus as it can be modified outside */
+ __kvm_riscv_vector_save(cntx);
+}
+
+void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx)
+{
+ __kvm_riscv_vector_restore(cntx);
+}
+
+void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu)
+{
+ kfree(vcpu->arch.guest_reset_context.vector.datap);
+ kfree(vcpu->arch.host_context.vector.datap);
+}
+#else
+#define riscv_vsize (0)
+#endif
+
+static void *kvm_riscv_vcpu_vreg_addr(struct kvm_vcpu *vcpu,
+ unsigned long reg_num,
+ size_t reg_size)
+{
+ struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
+ void *reg_val;
+ size_t vlenb = riscv_vsize / 32;
+
+ if (reg_num < KVM_REG_RISCV_VECTOR_REG(0)) {
+ if (reg_size != sizeof(unsigned long))
+ return NULL;
+ switch (reg_num) {
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vstart):
+ reg_val = &cntx->vector.vstart;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vl):
+ reg_val = &cntx->vector.vl;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vtype):
+ reg_val = &cntx->vector.vtype;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(vcsr):
+ reg_val = &cntx->vector.vcsr;
+ break;
+ case KVM_REG_RISCV_VECTOR_CSR_REG(datap):
+ default:
+ return NULL;
+ }
+ } else if (reg_num <= KVM_REG_RISCV_VECTOR_REG(31)) {
+ if (reg_size != vlenb)
+ return NULL;
+ reg_val = cntx->vector.datap
+ + (reg_num - KVM_REG_RISCV_VECTOR_REG(0)) * vlenb;
+ } else {
+ return NULL;
+ }
+
+ return reg_val;
+}
+
+int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype)
+{
+ unsigned long isa = vcpu->arch.isa;
+ unsigned long __user *uaddr =
+ (unsigned long __user *)(unsigned long)reg->addr;
+ unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
+ KVM_REG_SIZE_MASK |
+ rtype);
+ void *reg_val;
+ size_t reg_size = KVM_REG_SIZE(reg->id);
+
+ if ((rtype == KVM_REG_RISCV_VECTOR) &&
+ riscv_isa_extension_available(&isa, v)) {
+ reg_val = kvm_riscv_vcpu_vreg_addr(vcpu, reg_num, reg_size);
+ }
+
+ if (!reg_val)
+ return -EINVAL;
+
+ if (copy_to_user(uaddr, reg_val, reg_size))
+ return -EFAULT;
+
+ return 0;
+}
+
+int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
+ const struct kvm_one_reg *reg,
+ unsigned long rtype)
+{
+ unsigned long isa = vcpu->arch.isa;
+ unsigned long __user *uaddr =
+ (unsigned long __user *)(unsigned long)reg->addr;
+ unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
+ KVM_REG_SIZE_MASK |
+ rtype);
+ void *reg_val = NULL;
+ size_t reg_size = KVM_REG_SIZE(reg->id);
+
+ if ((rtype == KVM_REG_RISCV_VECTOR) &&
+ riscv_isa_extension_available(&isa, v)) {
+ reg_val = kvm_riscv_vcpu_vreg_addr(vcpu, reg_num, reg_size);
+ }
+
+ if (!reg_val)
+ return -EINVAL;
+
+ if (copy_from_user(reg_val, uaddr, reg_size))
+ return -EFAULT;
+
+ return 0;
+}
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread
* [PATCH v12 17/17] riscv: prctl to enable vector commands
2022-09-21 21:43 ` Chris Stillson
(?)
@ 2022-09-21 21:43 ` Chris Stillson
-1 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Chris Stillson, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Greentime Hu, Arnd Bergmann, Vincent Chen, Heiko Stuebner,
Jisheng Zhang, Dao Lu, Sunil V L, Han-Kuan Chen, Changbin Du,
Li Zhengyu, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Paul E. McKenney, Frederic Weisbecker,
Chen Lu, Vitaly Wool, Myrtle Shah, Mathieu Desnoyers,
Catalin Marinas, Mark Brown, Will Deacon, Luis Machado,
Janosch Frank, Huacai Chen, Alexey Dobriyan, Christian Brauner,
Peter Collingbourne, Colin Cross, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
This code makes enabling the vector extension on a riscv manchine
optional by adding an option to prctl() to allow a process to enable,
disable or query its vector context state.
-added prctl to enable/disable/query current vector state
-added actual function in riscv specific code to change/query the process
state
- Fixed problem with initial set of patches
(missing some EXPORT_SYMBOL() macro calls)
- rebased to 6.0-rc1
---
arch/riscv/configs/defconfig | 6 ++++++
arch/riscv/include/asm/kvm_vcpu_vector.h | 8 ++++----
arch/riscv/include/asm/processor.h | 6 ++++++
arch/riscv/include/asm/switch_to.h | 11 +++++++++++
arch/riscv/kernel/cpufeature.c | 3 ++-
arch/riscv/kernel/process.c | 20 +++++++++++++++++++-
arch/riscv/kvm/vcpu_vector.c | 14 +++++++-------
include/uapi/linux/prctl.h | 6 ++++++
kernel/sys.c | 7 +++++++
9 files changed, 68 insertions(+), 13 deletions(-)
diff --git a/arch/riscv/configs/defconfig b/arch/riscv/configs/defconfig
index aed332a9d4ea..fce054286b1f 100644
--- a/arch/riscv/configs/defconfig
+++ b/arch/riscv/configs/defconfig
@@ -209,3 +209,9 @@ CONFIG_RCU_EQS_DEBUG=y
# CONFIG_FTRACE is not set
# CONFIG_RUNTIME_TESTING_MENU is not set
CONFIG_MEMTEST=y
+CONFIG_ARCH_RV64I=y
+CONFIG_64BIT=y
+CONFIG_VECTOR=y
+CONFIG_ARCH_RV64I=y
+CONFIG_64BIT=y
+CONFIG_VECTOR=y
diff --git a/arch/riscv/include/asm/kvm_vcpu_vector.h b/arch/riscv/include/asm/kvm_vcpu_vector.h
index 1dcc1b2e05bb..c7101ff943a0 100644
--- a/arch/riscv/include/asm/kvm_vcpu_vector.h
+++ b/arch/riscv/include/asm/kvm_vcpu_vector.h
@@ -22,9 +22,9 @@ void __kvm_riscv_vector_save(struct kvm_cpu_context *context);
void __kvm_riscv_vector_restore(struct kvm_cpu_context *context);
void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu);
void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa);
+ unsigned long *isa);
void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa);
+ unsigned long *isa);
void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx);
void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx);
void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu);
@@ -34,12 +34,12 @@ static inline void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
}
static inline void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
}
static inline void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
}
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index a09141ecf6aa..f2d0a91ce174 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -88,6 +88,12 @@ extern void riscv_fill_hwcap(void);
extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
extern unsigned long signal_minsigstksz __ro_after_init;
+
+#ifdef CONFIG_VECTOR
+extern int rvv_proc_enable(unsigned long x);
+#define RVV_PROC_ENABLE(x) rvv_proc_enable(x)
+#endif
+
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index 527951c033d4..d9747450311c 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -80,6 +80,17 @@ extern unsigned long riscv_vsize;
extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
+static inline bool vstate_query(struct pt_regs *regs)
+{
+ return (regs->status & SR_VS) != 0;
+}
+
+static inline void vstate_on(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~(SR_VS)) | SR_VS_INITIAL;
+}
+
static inline void __vstate_clean(struct pt_regs *regs)
{
regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 0487ab19b234..3be469cb9266 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -37,6 +37,8 @@ __ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
#include <asm/vector.h>
__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
unsigned long riscv_vsize __read_mostly;
+EXPORT_SYMBOL(cpu_hwcap_vector);
+EXPORT_SYMBOL(riscv_vsize);
#endif
/**
@@ -346,4 +348,3 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
}
}
#endif
-}
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index e88a37fc77ed..a5a76d1374ec 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -15,6 +15,7 @@
#include <linux/tick.h>
#include <linux/ptrace.h>
#include <linux/uaccess.h>
+#include <linux/prctl.h>
#include <asm/unistd.h>
#include <asm/processor.h>
@@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
if (WARN_ON(!vstate->datap))
return;
}
- regs->status |= SR_VS_INITIAL;
/*
* Restore the initial value to the vector register
@@ -230,3 +229,22 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.sp = (unsigned long)childregs; /* kernel sp */
return 0;
}
+
+#ifdef CONFIG_VECTOR
+int rvv_proc_enable(unsigned long x)
+{
+ switch (x) {
+ case PR_RVV_DISABLE:
+ vstate_off(current, task_pt_regs(current));
+ return 0;
+ case PR_RVV_ENABLE:
+ vstate_on(current, task_pt_regs(current));
+ return 0;
+ case PR_RVV_QUERY:
+ return vstate_query(task_pt_regs(current));
+ default:
+ return -(EINVAL);
+
+ }
+}
+#endif
diff --git a/arch/riscv/kvm/vcpu_vector.c b/arch/riscv/kvm/vcpu_vector.c
index 37bf4ffd47dd..9d1613da561a 100644
--- a/arch/riscv/kvm/vcpu_vector.c
+++ b/arch/riscv/kvm/vcpu_vector.c
@@ -20,7 +20,7 @@
extern unsigned long riscv_vsize;
void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
cntx->sstatus &= ~SR_VS;
@@ -39,20 +39,20 @@ static void kvm_riscv_vcpu_vector_clean(struct kvm_cpu_context *cntx)
}
void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
if ((cntx->sstatus & SR_VS) == SR_VS_DIRTY) {
- if (riscv_isa_extension_available(&isa, v))
+ if (riscv_isa_extension_available(isa, v))
__kvm_riscv_vector_save(cntx);
kvm_riscv_vcpu_vector_clean(cntx);
}
}
void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
if ((cntx->sstatus & SR_VS) != SR_VS_OFF) {
- if (riscv_isa_extension_available(&isa, v))
+ if (riscv_isa_extension_available(isa, v))
__kvm_riscv_vector_restore(cntx);
kvm_riscv_vcpu_vector_clean(cntx);
}
@@ -122,7 +122,7 @@ int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
const struct kvm_one_reg *reg,
unsigned long rtype)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
unsigned long __user *uaddr =
(unsigned long __user *)(unsigned long)reg->addr;
unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
@@ -149,7 +149,7 @@ int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
const struct kvm_one_reg *reg,
unsigned long rtype)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
unsigned long __user *uaddr =
(unsigned long __user *)(unsigned long)reg->addr;
unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
index a5e06dcbba13..8ea56e4c48f8 100644
--- a/include/uapi/linux/prctl.h
+++ b/include/uapi/linux/prctl.h
@@ -281,6 +281,12 @@ struct prctl_mm_map {
# define PR_SME_VL_LEN_MASK 0xffff
# define PR_SME_VL_INHERIT (1 << 17) /* inherit across exec */
+/* RISC-V V vector extension */
+#define PR_RVV_STATE 65
+# define PR_RVV_DISABLE 0
+# define PR_RVV_ENABLE 1
+# define PR_RVV_QUERY 2
+
#define PR_SET_VMA 0x53564d41
# define PR_SET_VMA_ANON_NAME 0
diff --git a/kernel/sys.c b/kernel/sys.c
index b911fa6d81ab..3049b1823273 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -138,6 +138,9 @@
#ifndef GET_TAGGED_ADDR_CTRL
# define GET_TAGGED_ADDR_CTRL() (-EINVAL)
#endif
+#ifndef RVV_PROC_ENABLE
+# define RVV_PROC_ENABLE(x) (-EINVAL)
+#endif
/*
* this is where the system-wide overflow UID and GID are defined, for
@@ -2620,6 +2623,10 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
error = sched_core_share_pid(arg2, arg3, arg4, arg5);
break;
#endif
+ case PR_RVV_STATE:
+ error = RVV_PROC_ENABLE(arg2);
+ break;
+
case PR_SET_VMA:
error = prctl_set_vma(arg2, arg3, arg4, arg5);
break;
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 17/17] riscv: prctl to enable vector commands
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Chris Stillson, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Greentime Hu, Arnd Bergmann, Vincent Chen, Heiko Stuebner,
Jisheng Zhang, Dao Lu, Sunil V L, Han-Kuan Chen, Changbin Du,
Li Zhengyu, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Paul E. McKenney, Frederic Weisbecker,
Chen Lu, Vitaly Wool, Myrtle Shah, Mathieu Desnoyers,
Catalin Marinas, Mark Brown, Will Deacon, Luis Machado,
Janosch Frank, Huacai Chen, Alexey Dobriyan, Christian Brauner,
Peter Collingbourne, Colin Cross, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
This code makes enabling the vector extension on a riscv manchine
optional by adding an option to prctl() to allow a process to enable,
disable or query its vector context state.
-added prctl to enable/disable/query current vector state
-added actual function in riscv specific code to change/query the process
state
- Fixed problem with initial set of patches
(missing some EXPORT_SYMBOL() macro calls)
- rebased to 6.0-rc1
---
arch/riscv/configs/defconfig | 6 ++++++
arch/riscv/include/asm/kvm_vcpu_vector.h | 8 ++++----
arch/riscv/include/asm/processor.h | 6 ++++++
arch/riscv/include/asm/switch_to.h | 11 +++++++++++
arch/riscv/kernel/cpufeature.c | 3 ++-
arch/riscv/kernel/process.c | 20 +++++++++++++++++++-
arch/riscv/kvm/vcpu_vector.c | 14 +++++++-------
include/uapi/linux/prctl.h | 6 ++++++
kernel/sys.c | 7 +++++++
9 files changed, 68 insertions(+), 13 deletions(-)
diff --git a/arch/riscv/configs/defconfig b/arch/riscv/configs/defconfig
index aed332a9d4ea..fce054286b1f 100644
--- a/arch/riscv/configs/defconfig
+++ b/arch/riscv/configs/defconfig
@@ -209,3 +209,9 @@ CONFIG_RCU_EQS_DEBUG=y
# CONFIG_FTRACE is not set
# CONFIG_RUNTIME_TESTING_MENU is not set
CONFIG_MEMTEST=y
+CONFIG_ARCH_RV64I=y
+CONFIG_64BIT=y
+CONFIG_VECTOR=y
+CONFIG_ARCH_RV64I=y
+CONFIG_64BIT=y
+CONFIG_VECTOR=y
diff --git a/arch/riscv/include/asm/kvm_vcpu_vector.h b/arch/riscv/include/asm/kvm_vcpu_vector.h
index 1dcc1b2e05bb..c7101ff943a0 100644
--- a/arch/riscv/include/asm/kvm_vcpu_vector.h
+++ b/arch/riscv/include/asm/kvm_vcpu_vector.h
@@ -22,9 +22,9 @@ void __kvm_riscv_vector_save(struct kvm_cpu_context *context);
void __kvm_riscv_vector_restore(struct kvm_cpu_context *context);
void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu);
void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa);
+ unsigned long *isa);
void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa);
+ unsigned long *isa);
void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx);
void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx);
void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu);
@@ -34,12 +34,12 @@ static inline void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
}
static inline void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
}
static inline void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
}
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index a09141ecf6aa..f2d0a91ce174 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -88,6 +88,12 @@ extern void riscv_fill_hwcap(void);
extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
extern unsigned long signal_minsigstksz __ro_after_init;
+
+#ifdef CONFIG_VECTOR
+extern int rvv_proc_enable(unsigned long x);
+#define RVV_PROC_ENABLE(x) rvv_proc_enable(x)
+#endif
+
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index 527951c033d4..d9747450311c 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -80,6 +80,17 @@ extern unsigned long riscv_vsize;
extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
+static inline bool vstate_query(struct pt_regs *regs)
+{
+ return (regs->status & SR_VS) != 0;
+}
+
+static inline void vstate_on(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~(SR_VS)) | SR_VS_INITIAL;
+}
+
static inline void __vstate_clean(struct pt_regs *regs)
{
regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 0487ab19b234..3be469cb9266 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -37,6 +37,8 @@ __ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
#include <asm/vector.h>
__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
unsigned long riscv_vsize __read_mostly;
+EXPORT_SYMBOL(cpu_hwcap_vector);
+EXPORT_SYMBOL(riscv_vsize);
#endif
/**
@@ -346,4 +348,3 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
}
}
#endif
-}
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index e88a37fc77ed..a5a76d1374ec 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -15,6 +15,7 @@
#include <linux/tick.h>
#include <linux/ptrace.h>
#include <linux/uaccess.h>
+#include <linux/prctl.h>
#include <asm/unistd.h>
#include <asm/processor.h>
@@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
if (WARN_ON(!vstate->datap))
return;
}
- regs->status |= SR_VS_INITIAL;
/*
* Restore the initial value to the vector register
@@ -230,3 +229,22 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.sp = (unsigned long)childregs; /* kernel sp */
return 0;
}
+
+#ifdef CONFIG_VECTOR
+int rvv_proc_enable(unsigned long x)
+{
+ switch (x) {
+ case PR_RVV_DISABLE:
+ vstate_off(current, task_pt_regs(current));
+ return 0;
+ case PR_RVV_ENABLE:
+ vstate_on(current, task_pt_regs(current));
+ return 0;
+ case PR_RVV_QUERY:
+ return vstate_query(task_pt_regs(current));
+ default:
+ return -(EINVAL);
+
+ }
+}
+#endif
diff --git a/arch/riscv/kvm/vcpu_vector.c b/arch/riscv/kvm/vcpu_vector.c
index 37bf4ffd47dd..9d1613da561a 100644
--- a/arch/riscv/kvm/vcpu_vector.c
+++ b/arch/riscv/kvm/vcpu_vector.c
@@ -20,7 +20,7 @@
extern unsigned long riscv_vsize;
void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
cntx->sstatus &= ~SR_VS;
@@ -39,20 +39,20 @@ static void kvm_riscv_vcpu_vector_clean(struct kvm_cpu_context *cntx)
}
void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
if ((cntx->sstatus & SR_VS) == SR_VS_DIRTY) {
- if (riscv_isa_extension_available(&isa, v))
+ if (riscv_isa_extension_available(isa, v))
__kvm_riscv_vector_save(cntx);
kvm_riscv_vcpu_vector_clean(cntx);
}
}
void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
if ((cntx->sstatus & SR_VS) != SR_VS_OFF) {
- if (riscv_isa_extension_available(&isa, v))
+ if (riscv_isa_extension_available(isa, v))
__kvm_riscv_vector_restore(cntx);
kvm_riscv_vcpu_vector_clean(cntx);
}
@@ -122,7 +122,7 @@ int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
const struct kvm_one_reg *reg,
unsigned long rtype)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
unsigned long __user *uaddr =
(unsigned long __user *)(unsigned long)reg->addr;
unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
@@ -149,7 +149,7 @@ int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
const struct kvm_one_reg *reg,
unsigned long rtype)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
unsigned long __user *uaddr =
(unsigned long __user *)(unsigned long)reg->addr;
unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
index a5e06dcbba13..8ea56e4c48f8 100644
--- a/include/uapi/linux/prctl.h
+++ b/include/uapi/linux/prctl.h
@@ -281,6 +281,12 @@ struct prctl_mm_map {
# define PR_SME_VL_LEN_MASK 0xffff
# define PR_SME_VL_INHERIT (1 << 17) /* inherit across exec */
+/* RISC-V V vector extension */
+#define PR_RVV_STATE 65
+# define PR_RVV_DISABLE 0
+# define PR_RVV_ENABLE 1
+# define PR_RVV_QUERY 2
+
#define PR_SET_VMA 0x53564d41
# define PR_SET_VMA_ANON_NAME 0
diff --git a/kernel/sys.c b/kernel/sys.c
index b911fa6d81ab..3049b1823273 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -138,6 +138,9 @@
#ifndef GET_TAGGED_ADDR_CTRL
# define GET_TAGGED_ADDR_CTRL() (-EINVAL)
#endif
+#ifndef RVV_PROC_ENABLE
+# define RVV_PROC_ENABLE(x) (-EINVAL)
+#endif
/*
* this is where the system-wide overflow UID and GID are defined, for
@@ -2620,6 +2623,10 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
error = sched_core_share_pid(arg2, arg3, arg4, arg5);
break;
#endif
+ case PR_RVV_STATE:
+ error = RVV_PROC_ENABLE(arg2);
+ break;
+
case PR_SET_VMA:
error = prctl_set_vma(arg2, arg3, arg4, arg5);
break;
--
2.25.1
^ permalink raw reply related [flat|nested] 147+ messages in thread* [PATCH v12 17/17] riscv: prctl to enable vector commands
@ 2022-09-21 21:43 ` Chris Stillson
0 siblings, 0 replies; 147+ messages in thread
From: Chris Stillson @ 2022-09-21 21:43 UTC (permalink / raw)
Cc: Chris Stillson, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Eric Biederman, Kees Cook, Anup Patel, Atish Patra,
Oleg Nesterov, Guo Ren, Heinrich Schuchardt, Mayuresh Chitale,
Conor Dooley, Paolo Bonzini, Qinglin Pan, Alexandre Ghiti,
Greentime Hu, Arnd Bergmann, Vincent Chen, Heiko Stuebner,
Jisheng Zhang, Dao Lu, Sunil V L, Han-Kuan Chen, Changbin Du,
Li Zhengyu, Alexander Graf, Ard Biesheuvel, Tsukasa OI,
Yury Norov, Mark Rutland, Paul E. McKenney, Frederic Weisbecker,
Chen Lu, Vitaly Wool, Myrtle Shah, Mathieu Desnoyers,
Catalin Marinas, Mark Brown, Will Deacon, Luis Machado,
Janosch Frank, Huacai Chen, Alexey Dobriyan, Christian Brauner,
Peter Collingbourne, Colin Cross, Eugene Syromiatnikov,
Andrew Morton, Barret Rhoden, Suren Baghdasaryan,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv
This code makes enabling the vector extension on a riscv manchine
optional by adding an option to prctl() to allow a process to enable,
disable or query its vector context state.
-added prctl to enable/disable/query current vector state
-added actual function in riscv specific code to change/query the process
state
- Fixed problem with initial set of patches
(missing some EXPORT_SYMBOL() macro calls)
- rebased to 6.0-rc1
---
arch/riscv/configs/defconfig | 6 ++++++
arch/riscv/include/asm/kvm_vcpu_vector.h | 8 ++++----
arch/riscv/include/asm/processor.h | 6 ++++++
arch/riscv/include/asm/switch_to.h | 11 +++++++++++
arch/riscv/kernel/cpufeature.c | 3 ++-
arch/riscv/kernel/process.c | 20 +++++++++++++++++++-
arch/riscv/kvm/vcpu_vector.c | 14 +++++++-------
include/uapi/linux/prctl.h | 6 ++++++
kernel/sys.c | 7 +++++++
9 files changed, 68 insertions(+), 13 deletions(-)
diff --git a/arch/riscv/configs/defconfig b/arch/riscv/configs/defconfig
index aed332a9d4ea..fce054286b1f 100644
--- a/arch/riscv/configs/defconfig
+++ b/arch/riscv/configs/defconfig
@@ -209,3 +209,9 @@ CONFIG_RCU_EQS_DEBUG=y
# CONFIG_FTRACE is not set
# CONFIG_RUNTIME_TESTING_MENU is not set
CONFIG_MEMTEST=y
+CONFIG_ARCH_RV64I=y
+CONFIG_64BIT=y
+CONFIG_VECTOR=y
+CONFIG_ARCH_RV64I=y
+CONFIG_64BIT=y
+CONFIG_VECTOR=y
diff --git a/arch/riscv/include/asm/kvm_vcpu_vector.h b/arch/riscv/include/asm/kvm_vcpu_vector.h
index 1dcc1b2e05bb..c7101ff943a0 100644
--- a/arch/riscv/include/asm/kvm_vcpu_vector.h
+++ b/arch/riscv/include/asm/kvm_vcpu_vector.h
@@ -22,9 +22,9 @@ void __kvm_riscv_vector_save(struct kvm_cpu_context *context);
void __kvm_riscv_vector_restore(struct kvm_cpu_context *context);
void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu);
void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa);
+ unsigned long *isa);
void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa);
+ unsigned long *isa);
void kvm_riscv_vcpu_host_vector_save(struct kvm_cpu_context *cntx);
void kvm_riscv_vcpu_host_vector_restore(struct kvm_cpu_context *cntx);
void kvm_riscv_vcpu_free_vector_context(struct kvm_vcpu *vcpu);
@@ -34,12 +34,12 @@ static inline void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
}
static inline void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
}
static inline void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
}
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index a09141ecf6aa..f2d0a91ce174 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -88,6 +88,12 @@ extern void riscv_fill_hwcap(void);
extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
extern unsigned long signal_minsigstksz __ro_after_init;
+
+#ifdef CONFIG_VECTOR
+extern int rvv_proc_enable(unsigned long x);
+#define RVV_PROC_ENABLE(x) rvv_proc_enable(x)
+#endif
+
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/include/asm/switch_to.h b/arch/riscv/include/asm/switch_to.h
index 527951c033d4..d9747450311c 100644
--- a/arch/riscv/include/asm/switch_to.h
+++ b/arch/riscv/include/asm/switch_to.h
@@ -80,6 +80,17 @@ extern unsigned long riscv_vsize;
extern void __vstate_save(struct __riscv_v_state *save_to, void *datap);
extern void __vstate_restore(struct __riscv_v_state *restore_from, void *datap);
+static inline bool vstate_query(struct pt_regs *regs)
+{
+ return (regs->status & SR_VS) != 0;
+}
+
+static inline void vstate_on(struct task_struct *task,
+ struct pt_regs *regs)
+{
+ regs->status = (regs->status & ~(SR_VS)) | SR_VS_INITIAL;
+}
+
static inline void __vstate_clean(struct pt_regs *regs)
{
regs->status = (regs->status & ~(SR_VS)) | SR_VS_CLEAN;
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index 0487ab19b234..3be469cb9266 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -37,6 +37,8 @@ __ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_fpu);
#include <asm/vector.h>
__ro_after_init DEFINE_STATIC_KEY_FALSE(cpu_hwcap_vector);
unsigned long riscv_vsize __read_mostly;
+EXPORT_SYMBOL(cpu_hwcap_vector);
+EXPORT_SYMBOL(riscv_vsize);
#endif
/**
@@ -346,4 +348,3 @@ void __init_or_module riscv_cpufeature_patch_func(struct alt_entry *begin,
}
}
#endif
-}
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index e88a37fc77ed..a5a76d1374ec 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -15,6 +15,7 @@
#include <linux/tick.h>
#include <linux/ptrace.h>
#include <linux/uaccess.h>
+#include <linux/prctl.h>
#include <asm/unistd.h>
#include <asm/processor.h>
@@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
if (WARN_ON(!vstate->datap))
return;
}
- regs->status |= SR_VS_INITIAL;
/*
* Restore the initial value to the vector register
@@ -230,3 +229,22 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
p->thread.sp = (unsigned long)childregs; /* kernel sp */
return 0;
}
+
+#ifdef CONFIG_VECTOR
+int rvv_proc_enable(unsigned long x)
+{
+ switch (x) {
+ case PR_RVV_DISABLE:
+ vstate_off(current, task_pt_regs(current));
+ return 0;
+ case PR_RVV_ENABLE:
+ vstate_on(current, task_pt_regs(current));
+ return 0;
+ case PR_RVV_QUERY:
+ return vstate_query(task_pt_regs(current));
+ default:
+ return -(EINVAL);
+
+ }
+}
+#endif
diff --git a/arch/riscv/kvm/vcpu_vector.c b/arch/riscv/kvm/vcpu_vector.c
index 37bf4ffd47dd..9d1613da561a 100644
--- a/arch/riscv/kvm/vcpu_vector.c
+++ b/arch/riscv/kvm/vcpu_vector.c
@@ -20,7 +20,7 @@
extern unsigned long riscv_vsize;
void kvm_riscv_vcpu_vector_reset(struct kvm_vcpu *vcpu)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
struct kvm_cpu_context *cntx = &vcpu->arch.guest_context;
cntx->sstatus &= ~SR_VS;
@@ -39,20 +39,20 @@ static void kvm_riscv_vcpu_vector_clean(struct kvm_cpu_context *cntx)
}
void kvm_riscv_vcpu_guest_vector_save(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
if ((cntx->sstatus & SR_VS) == SR_VS_DIRTY) {
- if (riscv_isa_extension_available(&isa, v))
+ if (riscv_isa_extension_available(isa, v))
__kvm_riscv_vector_save(cntx);
kvm_riscv_vcpu_vector_clean(cntx);
}
}
void kvm_riscv_vcpu_guest_vector_restore(struct kvm_cpu_context *cntx,
- unsigned long isa)
+ unsigned long *isa)
{
if ((cntx->sstatus & SR_VS) != SR_VS_OFF) {
- if (riscv_isa_extension_available(&isa, v))
+ if (riscv_isa_extension_available(isa, v))
__kvm_riscv_vector_restore(cntx);
kvm_riscv_vcpu_vector_clean(cntx);
}
@@ -122,7 +122,7 @@ int kvm_riscv_vcpu_get_reg_vector(struct kvm_vcpu *vcpu,
const struct kvm_one_reg *reg,
unsigned long rtype)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
unsigned long __user *uaddr =
(unsigned long __user *)(unsigned long)reg->addr;
unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
@@ -149,7 +149,7 @@ int kvm_riscv_vcpu_set_reg_vector(struct kvm_vcpu *vcpu,
const struct kvm_one_reg *reg,
unsigned long rtype)
{
- unsigned long isa = vcpu->arch.isa;
+ unsigned long isa = *vcpu->arch.isa;
unsigned long __user *uaddr =
(unsigned long __user *)(unsigned long)reg->addr;
unsigned long reg_num = reg->id & ~(KVM_REG_ARCH_MASK |
diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
index a5e06dcbba13..8ea56e4c48f8 100644
--- a/include/uapi/linux/prctl.h
+++ b/include/uapi/linux/prctl.h
@@ -281,6 +281,12 @@ struct prctl_mm_map {
# define PR_SME_VL_LEN_MASK 0xffff
# define PR_SME_VL_INHERIT (1 << 17) /* inherit across exec */
+/* RISC-V V vector extension */
+#define PR_RVV_STATE 65
+# define PR_RVV_DISABLE 0
+# define PR_RVV_ENABLE 1
+# define PR_RVV_QUERY 2
+
#define PR_SET_VMA 0x53564d41
# define PR_SET_VMA_ANON_NAME 0
diff --git a/kernel/sys.c b/kernel/sys.c
index b911fa6d81ab..3049b1823273 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -138,6 +138,9 @@
#ifndef GET_TAGGED_ADDR_CTRL
# define GET_TAGGED_ADDR_CTRL() (-EINVAL)
#endif
+#ifndef RVV_PROC_ENABLE
+# define RVV_PROC_ENABLE(x) (-EINVAL)
+#endif
/*
* this is where the system-wide overflow UID and GID are defined, for
@@ -2620,6 +2623,10 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
error = sched_core_share_pid(arg2, arg3, arg4, arg5);
break;
#endif
+ case PR_RVV_STATE:
+ error = RVV_PROC_ENABLE(arg2);
+ break;
+
case PR_SET_VMA:
error = prctl_set_vma(arg2, arg3, arg4, arg5);
break;
--
2.25.1
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply related [flat|nested] 147+ messages in thread* RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-09-21 21:43 ` Chris Stillson
@ 2022-12-09 5:16 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-09 5:16 UTC (permalink / raw)
To: Chris Stillson
Cc: Paul Walmsley, Palmer Dabbelt, Anup Patel, Atish Patra, Guo Ren,
Conor Dooley, Greentime Hu, Vincent Chen, Andy Chiu,
Andrew Waterman, Darius Rad, Arnd Bergmann, lkml, linux-riscv,
Björn Töpel, Florian Weimer, GNU C Library,
Conor Dooley, Christoph Müllner, Aaron Durbin, linux
Hi Darius, Andrew, Palmer
On 9/21/22 14:43, Chris Stillson wrote:
> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
>
> @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
> if (WARN_ON(!vstate->datap))
> return;
> }
> - regs->status |= SR_VS_INITIAL;
>
Perhaps not obvious from the patch, but this is a major user experience
change: As in V unit would be turned off for a new task and we will rely
on a userspace prctl (also introduced in this patch) to enable V.
I know some of you had different opinion on this in the past [1], so
this is to make sure everyone's on same page.
And if we agree this is the way to go, how exactly will this be done in
userspace.
glibc dynamic loader will invoke the prctl() ? How will it decide
whether to do this (or not) - will it be unconditional or will it use
the hwcap - does latter plumbing exist already ? If so is it AT_HWCAP /
HWCAP2.
Also for static linked executables, where will the prctl be called from ?
[1] https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
^ permalink raw reply [flat|nested] 147+ messages in thread
* RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 5:16 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-09 5:16 UTC (permalink / raw)
To: Chris Stillson
Cc: Paul Walmsley, Palmer Dabbelt, Anup Patel, Atish Patra, Guo Ren,
Conor Dooley, Greentime Hu, Vincent Chen, Andy Chiu,
Andrew Waterman, Darius Rad, Arnd Bergmann, lkml, linux-riscv,
Björn Töpel, Florian Weimer, GNU C Library,
Conor Dooley, Christoph Müllner, Aaron Durbin, linux
Hi Darius, Andrew, Palmer
On 9/21/22 14:43, Chris Stillson wrote:
> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
>
> @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
> if (WARN_ON(!vstate->datap))
> return;
> }
> - regs->status |= SR_VS_INITIAL;
>
Perhaps not obvious from the patch, but this is a major user experience
change: As in V unit would be turned off for a new task and we will rely
on a userspace prctl (also introduced in this patch) to enable V.
I know some of you had different opinion on this in the past [1], so
this is to make sure everyone's on same page.
And if we agree this is the way to go, how exactly will this be done in
userspace.
glibc dynamic loader will invoke the prctl() ? How will it decide
whether to do this (or not) - will it be unconditional or will it use
the hwcap - does latter plumbing exist already ? If so is it AT_HWCAP /
HWCAP2.
Also for static linked executables, where will the prctl be called from ?
[1] https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 5:16 ` Vineet Gupta
@ 2022-12-09 6:27 ` Palmer Dabbelt
-1 siblings, 0 replies; 147+ messages in thread
From: Palmer Dabbelt @ 2022-12-09 6:27 UTC (permalink / raw)
To: Vineet Gupta
Cc: stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, Andrew Waterman,
Darius Rad, arnd, linux-kernel, linux-riscv, bjorn, fweimer,
libc-alpha, Conor Dooley, christoph.muellner, Aaron Durbin,
linux
On Thu, 08 Dec 2022 21:16:06 PST (-0800), Vineet Gupta wrote:
> Hi Darius, Andrew, Palmer
>
> On 9/21/22 14:43, Chris Stillson wrote:
>> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
>>
>> @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
>> if (WARN_ON(!vstate->datap))
>> return;
>> }
>> - regs->status |= SR_VS_INITIAL;
>>
>
> Perhaps not obvious from the patch, but this is a major user experience
> change: As in V unit would be turned off for a new task and we will rely
> on a userspace prctl (also introduced in this patch) to enable V.
IMO that's the only viable option: enabling V adds more user-visible
state, which is a uABI break. I haven't really had time to poke through
all the versions here, but I'd have the call look something like
prctl(RISCV_ENABLE_V, min_vlenb, max_vlenb, flags);
where
* min_vlenb is the smallest VLENB that userspace can support. There's
alreday an LLVM argument for this, I haven't dug into the generated
code but I assume it'll blow up on smaller VLENB systems somehow.
* max_vlenb is the largest VLENB that userspace can support.
* flags is just a placeholder for now, with 0 meaning "V as defined by
1.0 for all threads in this proces". That should give us an out if
something more complicated happens in the future.
That way VLA code can call `prctl(RISCV_ENABLE_V, 128, 8192, 0)` as it
supports any V 1.0 implementation, while code with other constraints can
avoid having V turned on in an unsupported configuration.
I think we can start out with no flags, but there's a few I could see
being useful already:
* Cross process/thread enabling. I think a reasonable default is
"enable V for all current and future threads in this process", but one
could imagine flags for "just this thread" vs "all current threads", a
default for new threads, and a default for child processes. I don't
think it matters so much what we pick as a default, just that it's
written down.
* Setting the VLENB bounds vs updating them. I'm thinking for shared
libraries, where they'd only want to enable V in the shared library if
it's already in a supported configuration. I'm not sure what the
right rules are here, but again it's best to write that down.
* Some way to disable V. Maybe we just say `prctl(RISCV_ENABLE_V, 0, 0,
...)` disables V, or maybe it's a flag? Again, it should just be
written down.
* What exactly we're enabling -- is it the V extension, or just the V
registers?
There's a bunch of subtly here, though, so I think we'd at least want
glibc and gdb support posted before committing to any uABI. It's
probably also worth looking at what the Arm folks did for SVE: I gave it
a quick glance and it seems like there's a lot of similarities with what
I'm suggesting here, but again a lot of this is pretty subtle stuff so
it's hard to tell just at a glance.
> I know some of you had different opinion on this in the past [1], so
> this is to make sure everyone's on same page.
> And if we agree this is the way to go, how exactly will this be done in
> userspace.
>
> glibc dynamic loader will invoke the prctl() ? How will it decide
> whether to do this (or not) - will it be unconditional or will it use
> the hwcap - does latter plumbing exist already ? If so is it AT_HWCAP /
> HWCAP2.
That part I haven't sorted out yet, and I don't think it's sufficient to
just say "userspace should enable what it can support" because of how
pervasive V instructions are going to be.
I don't think we need HWCAP, as userspace will need to call the prctl()
anyway to turn on V and thus can just use the success/failure of that to
sort things out.
Maybe it's sufficient to rely on some sort of sticky prctl() (or sysctl
type thing, the differences there would be pretty subtle) and just not
worry about it, but having some way of encoding this in the ELF seems
nice. That said, we've had a bunch of trouble sorting out the ISA
encoding in ELFs so maybe it's just not worth bothering?
> Also for static linked executables, where will the prctl be called from ?
I guess that's pretty far in the weeds, but we could at least hook CRT
to insert the relevant code. We'd really need to sort out how we're
going to encode the V support in binaries, though.
> [1] https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 6:27 ` Palmer Dabbelt
0 siblings, 0 replies; 147+ messages in thread
From: Palmer Dabbelt @ 2022-12-09 6:27 UTC (permalink / raw)
To: Vineet Gupta
Cc: stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, Andrew Waterman,
Darius Rad, arnd, linux-kernel, linux-riscv, bjorn, fweimer,
libc-alpha, Conor Dooley, christoph.muellner, Aaron Durbin,
linux
On Thu, 08 Dec 2022 21:16:06 PST (-0800), Vineet Gupta wrote:
> Hi Darius, Andrew, Palmer
>
> On 9/21/22 14:43, Chris Stillson wrote:
>> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
>>
>> @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
>> if (WARN_ON(!vstate->datap))
>> return;
>> }
>> - regs->status |= SR_VS_INITIAL;
>>
>
> Perhaps not obvious from the patch, but this is a major user experience
> change: As in V unit would be turned off for a new task and we will rely
> on a userspace prctl (also introduced in this patch) to enable V.
IMO that's the only viable option: enabling V adds more user-visible
state, which is a uABI break. I haven't really had time to poke through
all the versions here, but I'd have the call look something like
prctl(RISCV_ENABLE_V, min_vlenb, max_vlenb, flags);
where
* min_vlenb is the smallest VLENB that userspace can support. There's
alreday an LLVM argument for this, I haven't dug into the generated
code but I assume it'll blow up on smaller VLENB systems somehow.
* max_vlenb is the largest VLENB that userspace can support.
* flags is just a placeholder for now, with 0 meaning "V as defined by
1.0 for all threads in this proces". That should give us an out if
something more complicated happens in the future.
That way VLA code can call `prctl(RISCV_ENABLE_V, 128, 8192, 0)` as it
supports any V 1.0 implementation, while code with other constraints can
avoid having V turned on in an unsupported configuration.
I think we can start out with no flags, but there's a few I could see
being useful already:
* Cross process/thread enabling. I think a reasonable default is
"enable V for all current and future threads in this process", but one
could imagine flags for "just this thread" vs "all current threads", a
default for new threads, and a default for child processes. I don't
think it matters so much what we pick as a default, just that it's
written down.
* Setting the VLENB bounds vs updating them. I'm thinking for shared
libraries, where they'd only want to enable V in the shared library if
it's already in a supported configuration. I'm not sure what the
right rules are here, but again it's best to write that down.
* Some way to disable V. Maybe we just say `prctl(RISCV_ENABLE_V, 0, 0,
...)` disables V, or maybe it's a flag? Again, it should just be
written down.
* What exactly we're enabling -- is it the V extension, or just the V
registers?
There's a bunch of subtly here, though, so I think we'd at least want
glibc and gdb support posted before committing to any uABI. It's
probably also worth looking at what the Arm folks did for SVE: I gave it
a quick glance and it seems like there's a lot of similarities with what
I'm suggesting here, but again a lot of this is pretty subtle stuff so
it's hard to tell just at a glance.
> I know some of you had different opinion on this in the past [1], so
> this is to make sure everyone's on same page.
> And if we agree this is the way to go, how exactly will this be done in
> userspace.
>
> glibc dynamic loader will invoke the prctl() ? How will it decide
> whether to do this (or not) - will it be unconditional or will it use
> the hwcap - does latter plumbing exist already ? If so is it AT_HWCAP /
> HWCAP2.
That part I haven't sorted out yet, and I don't think it's sufficient to
just say "userspace should enable what it can support" because of how
pervasive V instructions are going to be.
I don't think we need HWCAP, as userspace will need to call the prctl()
anyway to turn on V and thus can just use the success/failure of that to
sort things out.
Maybe it's sufficient to rely on some sort of sticky prctl() (or sysctl
type thing, the differences there would be pretty subtle) and just not
worry about it, but having some way of encoding this in the ELF seems
nice. That said, we've had a bunch of trouble sorting out the ISA
encoding in ELFs so maybe it's just not worth bothering?
> Also for static linked executables, where will the prctl be called from ?
I guess that's pretty far in the weeds, but we could at least hook CRT
to insert the relevant code. We'd really need to sort out how we're
going to encode the V support in binaries, though.
> [1] https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 6:27 ` Palmer Dabbelt
@ 2022-12-09 7:42 ` Andrew Waterman
-1 siblings, 0 replies; 147+ messages in thread
From: Andrew Waterman @ 2022-12-09 7:42 UTC (permalink / raw)
To: Palmer Dabbelt
Cc: Vineet Gupta, stillson, Paul Walmsley, anup, atishp, guoren,
Conor Dooley, greentime.hu, vincent.chen, andy.chiu, Darius Rad,
arnd, linux-kernel, linux-riscv, bjorn, fweimer, libc-alpha,
christoph.muellner, Aaron Durbin, linux
Requiring application programmers (i.e. those who write main()) to
make a prctl() call is obviously completely unacceptable, because
application programmers don't know whether the V extension is being
used. Auto-vectorization and libc-function implementations will use
the V extension without any application-programmer knowledge or
intervention. And obviously we don't want to preclude that.
This suggests that ld.so, early-stage libc, or possibly both will need
to make this prctl() call, perhaps by parsing the ELF headers of the
binary and each library to determine if the V extension is used.
Personally, I'm agnostic to whether we put this onus on the kernel or
on user-space--I just want to make sure we're all on the same page
that it needs to be hidden behind libc/ld.so/etc. The onus can't be
on the application programmer.
On Thu, Dec 8, 2022 at 8:27 PM Palmer Dabbelt <palmer@dabbelt.com> wrote:
>
> On Thu, 08 Dec 2022 21:16:06 PST (-0800), Vineet Gupta wrote:
> > Hi Darius, Andrew, Palmer
> >
> > On 9/21/22 14:43, Chris Stillson wrote:
> >> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
> >>
> >> @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
> >> if (WARN_ON(!vstate->datap))
> >> return;
> >> }
> >> - regs->status |= SR_VS_INITIAL;
> >>
> >
> > Perhaps not obvious from the patch, but this is a major user experience
> > change: As in V unit would be turned off for a new task and we will rely
> > on a userspace prctl (also introduced in this patch) to enable V.
>
> IMO that's the only viable option: enabling V adds more user-visible
> state, which is a uABI break. I haven't really had time to poke through
> all the versions here, but I'd have the call look something like
>
> prctl(RISCV_ENABLE_V, min_vlenb, max_vlenb, flags);
>
> where
>
> * min_vlenb is the smallest VLENB that userspace can support. There's
> alreday an LLVM argument for this, I haven't dug into the generated
> code but I assume it'll blow up on smaller VLENB systems somehow.
> * max_vlenb is the largest VLENB that userspace can support.
> * flags is just a placeholder for now, with 0 meaning "V as defined by
> 1.0 for all threads in this proces". That should give us an out if
> something more complicated happens in the future.
>
> That way VLA code can call `prctl(RISCV_ENABLE_V, 128, 8192, 0)` as it
> supports any V 1.0 implementation, while code with other constraints can
> avoid having V turned on in an unsupported configuration.
VLA code needs to read the vlenb CSR; it can't assume 8192 (or any
other small number) is a safe upper bound.
>
> I think we can start out with no flags, but there's a few I could see
> being useful already:
>
> * Cross process/thread enabling. I think a reasonable default is
> "enable V for all current and future threads in this process", but one
> could imagine flags for "just this thread" vs "all current threads", a
> default for new threads, and a default for child processes. I don't
> think it matters so much what we pick as a default, just that it's
> written down.
> * Setting the VLENB bounds vs updating them. I'm thinking for shared
> libraries, where they'd only want to enable V in the shared library if
> it's already in a supported configuration. I'm not sure what the
> right rules are here, but again it's best to write that down.
> * Some way to disable V. Maybe we just say `prctl(RISCV_ENABLE_V, 0, 0,
> ...)` disables V, or maybe it's a flag? Again, it should just be
> written down.
> * What exactly we're enabling -- is it the V extension, or just the V
> registers?
>
> There's a bunch of subtly here, though, so I think we'd at least want
> glibc and gdb support posted before committing to any uABI. It's
> probably also worth looking at what the Arm folks did for SVE: I gave it
> a quick glance and it seems like there's a lot of similarities with what
> I'm suggesting here, but again a lot of this is pretty subtle stuff so
> it's hard to tell just at a glance.
>
> > I know some of you had different opinion on this in the past [1], so
> > this is to make sure everyone's on same page.
> > And if we agree this is the way to go, how exactly will this be done in
> > userspace.
> >
> > glibc dynamic loader will invoke the prctl() ? How will it decide
> > whether to do this (or not) - will it be unconditional or will it use
> > the hwcap - does latter plumbing exist already ? If so is it AT_HWCAP /
> > HWCAP2.
>
> That part I haven't sorted out yet, and I don't think it's sufficient to
> just say "userspace should enable what it can support" because of how
> pervasive V instructions are going to be.
>
> I don't think we need HWCAP, as userspace will need to call the prctl()
> anyway to turn on V and thus can just use the success/failure of that to
> sort things out.
>
> Maybe it's sufficient to rely on some sort of sticky prctl() (or sysctl
> type thing, the differences there would be pretty subtle) and just not
> worry about it, but having some way of encoding this in the ELF seems
> nice. That said, we've had a bunch of trouble sorting out the ISA
> encoding in ELFs so maybe it's just not worth bothering?
>
> > Also for static linked executables, where will the prctl be called from ?
>
> I guess that's pretty far in the weeds, but we could at least hook CRT
> to insert the relevant code. We'd really need to sort out how we're
> going to encode the V support in binaries, though.
>
> > [1] https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 7:42 ` Andrew Waterman
0 siblings, 0 replies; 147+ messages in thread
From: Andrew Waterman @ 2022-12-09 7:42 UTC (permalink / raw)
To: Palmer Dabbelt
Cc: Vineet Gupta, stillson, Paul Walmsley, anup, atishp, guoren,
Conor Dooley, greentime.hu, vincent.chen, andy.chiu, Darius Rad,
arnd, linux-kernel, linux-riscv, bjorn, fweimer, libc-alpha,
christoph.muellner, Aaron Durbin, linux
Requiring application programmers (i.e. those who write main()) to
make a prctl() call is obviously completely unacceptable, because
application programmers don't know whether the V extension is being
used. Auto-vectorization and libc-function implementations will use
the V extension without any application-programmer knowledge or
intervention. And obviously we don't want to preclude that.
This suggests that ld.so, early-stage libc, or possibly both will need
to make this prctl() call, perhaps by parsing the ELF headers of the
binary and each library to determine if the V extension is used.
Personally, I'm agnostic to whether we put this onus on the kernel or
on user-space--I just want to make sure we're all on the same page
that it needs to be hidden behind libc/ld.so/etc. The onus can't be
on the application programmer.
On Thu, Dec 8, 2022 at 8:27 PM Palmer Dabbelt <palmer@dabbelt.com> wrote:
>
> On Thu, 08 Dec 2022 21:16:06 PST (-0800), Vineet Gupta wrote:
> > Hi Darius, Andrew, Palmer
> >
> > On 9/21/22 14:43, Chris Stillson wrote:
> >> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
> >>
> >> @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs, unsigned long pc,
> >> if (WARN_ON(!vstate->datap))
> >> return;
> >> }
> >> - regs->status |= SR_VS_INITIAL;
> >>
> >
> > Perhaps not obvious from the patch, but this is a major user experience
> > change: As in V unit would be turned off for a new task and we will rely
> > on a userspace prctl (also introduced in this patch) to enable V.
>
> IMO that's the only viable option: enabling V adds more user-visible
> state, which is a uABI break. I haven't really had time to poke through
> all the versions here, but I'd have the call look something like
>
> prctl(RISCV_ENABLE_V, min_vlenb, max_vlenb, flags);
>
> where
>
> * min_vlenb is the smallest VLENB that userspace can support. There's
> alreday an LLVM argument for this, I haven't dug into the generated
> code but I assume it'll blow up on smaller VLENB systems somehow.
> * max_vlenb is the largest VLENB that userspace can support.
> * flags is just a placeholder for now, with 0 meaning "V as defined by
> 1.0 for all threads in this proces". That should give us an out if
> something more complicated happens in the future.
>
> That way VLA code can call `prctl(RISCV_ENABLE_V, 128, 8192, 0)` as it
> supports any V 1.0 implementation, while code with other constraints can
> avoid having V turned on in an unsupported configuration.
VLA code needs to read the vlenb CSR; it can't assume 8192 (or any
other small number) is a safe upper bound.
>
> I think we can start out with no flags, but there's a few I could see
> being useful already:
>
> * Cross process/thread enabling. I think a reasonable default is
> "enable V for all current and future threads in this process", but one
> could imagine flags for "just this thread" vs "all current threads", a
> default for new threads, and a default for child processes. I don't
> think it matters so much what we pick as a default, just that it's
> written down.
> * Setting the VLENB bounds vs updating them. I'm thinking for shared
> libraries, where they'd only want to enable V in the shared library if
> it's already in a supported configuration. I'm not sure what the
> right rules are here, but again it's best to write that down.
> * Some way to disable V. Maybe we just say `prctl(RISCV_ENABLE_V, 0, 0,
> ...)` disables V, or maybe it's a flag? Again, it should just be
> written down.
> * What exactly we're enabling -- is it the V extension, or just the V
> registers?
>
> There's a bunch of subtly here, though, so I think we'd at least want
> glibc and gdb support posted before committing to any uABI. It's
> probably also worth looking at what the Arm folks did for SVE: I gave it
> a quick glance and it seems like there's a lot of similarities with what
> I'm suggesting here, but again a lot of this is pretty subtle stuff so
> it's hard to tell just at a glance.
>
> > I know some of you had different opinion on this in the past [1], so
> > this is to make sure everyone's on same page.
> > And if we agree this is the way to go, how exactly will this be done in
> > userspace.
> >
> > glibc dynamic loader will invoke the prctl() ? How will it decide
> > whether to do this (or not) - will it be unconditional or will it use
> > the hwcap - does latter plumbing exist already ? If so is it AT_HWCAP /
> > HWCAP2.
>
> That part I haven't sorted out yet, and I don't think it's sufficient to
> just say "userspace should enable what it can support" because of how
> pervasive V instructions are going to be.
>
> I don't think we need HWCAP, as userspace will need to call the prctl()
> anyway to turn on V and thus can just use the success/failure of that to
> sort things out.
>
> Maybe it's sufficient to rely on some sort of sticky prctl() (or sysctl
> type thing, the differences there would be pretty subtle) and just not
> worry about it, but having some way of encoding this in the ELF seems
> nice. That said, we've had a bunch of trouble sorting out the ISA
> encoding in ELFs so maybe it's just not worth bothering?
>
> > Also for static linked executables, where will the prctl be called from ?
>
> I guess that's pretty far in the weeds, but we could at least hook CRT
> to insert the relevant code. We'd really need to sort out how we're
> going to encode the V support in binaries, though.
>
> > [1] https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 7:42 ` Andrew Waterman
@ 2022-12-09 10:02 ` Florian Weimer
-1 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-09 10:02 UTC (permalink / raw)
To: Andrew Waterman
Cc: Palmer Dabbelt, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, Darius Rad, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
* Andrew Waterman:
> This suggests that ld.so, early-stage libc, or possibly both will need
> to make this prctl() call, perhaps by parsing the ELF headers of the
> binary and each library to determine if the V extension is used.
If the string functions use the V extension, it will be enabled
unconditionally. So I don't see why it's okay for libc to trigger this
alleged UAPI change, when the kernel can't do it by default.
Thanks,
Florian
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 10:02 ` Florian Weimer
0 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-09 10:02 UTC (permalink / raw)
To: Andrew Waterman
Cc: Palmer Dabbelt, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, Darius Rad, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
* Andrew Waterman:
> This suggests that ld.so, early-stage libc, or possibly both will need
> to make this prctl() call, perhaps by parsing the ELF headers of the
> binary and each library to determine if the V extension is used.
If the string functions use the V extension, it will be enabled
unconditionally. So I don't see why it's okay for libc to trigger this
alleged UAPI change, when the kernel can't do it by default.
Thanks,
Florian
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 10:02 ` Florian Weimer
@ 2022-12-09 12:21 ` Darius Rad
-1 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-09 12:21 UTC (permalink / raw)
To: Florian Weimer
Cc: Andrew Waterman, Palmer Dabbelt, Vineet Gupta, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
> * Andrew Waterman:
>
> > This suggests that ld.so, early-stage libc, or possibly both will need
> > to make this prctl() call, perhaps by parsing the ELF headers of the
> > binary and each library to determine if the V extension is used.
>
> If the string functions use the V extension, it will be enabled
> unconditionally. So I don't see why it's okay for libc to trigger this
> alleged UAPI change, when the kernel can't do it by default.
>
Because the call to enable can fail and userspace needs to deal with that.
// darius
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 12:21 ` Darius Rad
0 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-09 12:21 UTC (permalink / raw)
To: Florian Weimer
Cc: Andrew Waterman, Palmer Dabbelt, Vineet Gupta, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
> * Andrew Waterman:
>
> > This suggests that ld.so, early-stage libc, or possibly both will need
> > to make this prctl() call, perhaps by parsing the ELF headers of the
> > binary and each library to determine if the V extension is used.
>
> If the string functions use the V extension, it will be enabled
> unconditionally. So I don't see why it's okay for libc to trigger this
> alleged UAPI change, when the kernel can't do it by default.
>
Because the call to enable can fail and userspace needs to deal with that.
// darius
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 12:21 ` Darius Rad
@ 2022-12-09 12:32 ` Florian Weimer
-1 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-09 12:32 UTC (permalink / raw)
To: Andrew Waterman
Cc: Palmer Dabbelt, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
* Darius Rad:
> On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>> * Andrew Waterman:
>>
>> > This suggests that ld.so, early-stage libc, or possibly both will need
>> > to make this prctl() call, perhaps by parsing the ELF headers of the
>> > binary and each library to determine if the V extension is used.
>>
>> If the string functions use the V extension, it will be enabled
>> unconditionally. So I don't see why it's okay for libc to trigger this
>> alleged UAPI change, when the kernel can't do it by default.
>>
>
> Because the call to enable can fail and userspace needs to deal with that.
Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
zero, or perhaps a special CPU register (although that is more unusual).
It's possible to do this differently, but every mid-level startup code
will have to replicate it (the libcs, other run-time environments like
Go, and so on).
Still it's much better than executing the instruction to see if it
traps, so I won't complain too much.
Thanks,
Florian
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 12:32 ` Florian Weimer
0 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-09 12:32 UTC (permalink / raw)
To: Andrew Waterman
Cc: Palmer Dabbelt, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
* Darius Rad:
> On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>> * Andrew Waterman:
>>
>> > This suggests that ld.so, early-stage libc, or possibly both will need
>> > to make this prctl() call, perhaps by parsing the ELF headers of the
>> > binary and each library to determine if the V extension is used.
>>
>> If the string functions use the V extension, it will be enabled
>> unconditionally. So I don't see why it's okay for libc to trigger this
>> alleged UAPI change, when the kernel can't do it by default.
>>
>
> Because the call to enable can fail and userspace needs to deal with that.
Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
zero, or perhaps a special CPU register (although that is more unusual).
It's possible to do this differently, but every mid-level startup code
will have to replicate it (the libcs, other run-time environments like
Go, and so on).
Still it's much better than executing the instruction to see if it
traps, so I won't complain too much.
Thanks,
Florian
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 12:32 ` Florian Weimer
@ 2022-12-09 12:42 ` Darius Rad
-1 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-09 12:42 UTC (permalink / raw)
To: Florian Weimer
Cc: Andrew Waterman, Palmer Dabbelt, Vineet Gupta, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via Libc-alpha wrote:
> * Darius Rad:
>
> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
> >> * Andrew Waterman:
> >>
> >> > This suggests that ld.so, early-stage libc, or possibly both will need
> >> > to make this prctl() call, perhaps by parsing the ELF headers of the
> >> > binary and each library to determine if the V extension is used.
> >>
> >> If the string functions use the V extension, it will be enabled
> >> unconditionally. So I don't see why it's okay for libc to trigger this
> >> alleged UAPI change, when the kernel can't do it by default.
> >>
> >
> > Because the call to enable can fail and userspace needs to deal with that.
>
> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
> zero, or perhaps a special CPU register (although that is more unusual).
That would indicate that the extension is not present, which is one of, but
not the only way it can fail.
The vector extension relies on dynamically allocated memory in the kernel,
which can fail.
It also provides the opportunity for the kernel to deny access to the
vector extension, perhaps due to administrative policy or other future
mechanism.
// darius
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 12:42 ` Darius Rad
0 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-09 12:42 UTC (permalink / raw)
To: Florian Weimer
Cc: Andrew Waterman, Palmer Dabbelt, Vineet Gupta, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via Libc-alpha wrote:
> * Darius Rad:
>
> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
> >> * Andrew Waterman:
> >>
> >> > This suggests that ld.so, early-stage libc, or possibly both will need
> >> > to make this prctl() call, perhaps by parsing the ELF headers of the
> >> > binary and each library to determine if the V extension is used.
> >>
> >> If the string functions use the V extension, it will be enabled
> >> unconditionally. So I don't see why it's okay for libc to trigger this
> >> alleged UAPI change, when the kernel can't do it by default.
> >>
> >
> > Because the call to enable can fail and userspace needs to deal with that.
>
> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
> zero, or perhaps a special CPU register (although that is more unusual).
That would indicate that the extension is not present, which is one of, but
not the only way it can fail.
The vector extension relies on dynamically allocated memory in the kernel,
which can fail.
It also provides the opportunity for the kernel to deny access to the
vector extension, perhaps due to administrative policy or other future
mechanism.
// darius
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 12:42 ` Darius Rad
@ 2022-12-09 13:04 ` Florian Weimer
-1 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-09 13:04 UTC (permalink / raw)
To: Andrew Waterman
Cc: Palmer Dabbelt, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
* Darius Rad:
> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via Libc-alpha wrote:
>> * Darius Rad:
>>
>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>> >> * Andrew Waterman:
>> >>
>> >> > This suggests that ld.so, early-stage libc, or possibly both will need
>> >> > to make this prctl() call, perhaps by parsing the ELF headers of the
>> >> > binary and each library to determine if the V extension is used.
>> >>
>> >> If the string functions use the V extension, it will be enabled
>> >> unconditionally. So I don't see why it's okay for libc to trigger this
>> >> alleged UAPI change, when the kernel can't do it by default.
>> >>
>> >
>> > Because the call to enable can fail and userspace needs to deal with that.
>>
>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
>> zero, or perhaps a special CPU register (although that is more unusual).
>
> That would indicate that the extension is not present, which is one of, but
> not the only way it can fail.
I think you should bring down the number of failure modes. HWCAP has
the advantage that it communicates kernel/hypervisor/firmware/CPU
support in a single bit, which simplifies the programming model and
avoids hard-to-detect bugs. It's not clear why it would be beneficial
to continue on ENOMEM failures here because the system must clearly be
in bad shape at this point, and launching a new process is very unlikely
to improve matters. So I think the simpler programming model is the way
to go here.
> The vector extension relies on dynamically allocated memory in the kernel,
> which can fail.
But this failure can be reported as part of execve and clone.
> It also provides the opportunity for the kernel to deny access to the
> vector extension, perhaps due to administrative policy or other future
> mechanism.
HWCAP can do this, too.
Thanks,
Florian
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 13:04 ` Florian Weimer
0 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-09 13:04 UTC (permalink / raw)
To: Andrew Waterman
Cc: Palmer Dabbelt, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
* Darius Rad:
> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via Libc-alpha wrote:
>> * Darius Rad:
>>
>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>> >> * Andrew Waterman:
>> >>
>> >> > This suggests that ld.so, early-stage libc, or possibly both will need
>> >> > to make this prctl() call, perhaps by parsing the ELF headers of the
>> >> > binary and each library to determine if the V extension is used.
>> >>
>> >> If the string functions use the V extension, it will be enabled
>> >> unconditionally. So I don't see why it's okay for libc to trigger this
>> >> alleged UAPI change, when the kernel can't do it by default.
>> >>
>> >
>> > Because the call to enable can fail and userspace needs to deal with that.
>>
>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
>> zero, or perhaps a special CPU register (although that is more unusual).
>
> That would indicate that the extension is not present, which is one of, but
> not the only way it can fail.
I think you should bring down the number of failure modes. HWCAP has
the advantage that it communicates kernel/hypervisor/firmware/CPU
support in a single bit, which simplifies the programming model and
avoids hard-to-detect bugs. It's not clear why it would be beneficial
to continue on ENOMEM failures here because the system must clearly be
in bad shape at this point, and launching a new process is very unlikely
to improve matters. So I think the simpler programming model is the way
to go here.
> The vector extension relies on dynamically allocated memory in the kernel,
> which can fail.
But this failure can be reported as part of execve and clone.
> It also provides the opportunity for the kernel to deny access to the
> vector extension, perhaps due to administrative policy or other future
> mechanism.
HWCAP can do this, too.
Thanks,
Florian
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 13:04 ` Florian Weimer
@ 2022-12-09 17:21 ` Palmer Dabbelt
-1 siblings, 0 replies; 147+ messages in thread
From: Palmer Dabbelt @ 2022-12-09 17:21 UTC (permalink / raw)
To: fweimer
Cc: Andrew Waterman, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
On Fri, 09 Dec 2022 05:04:23 PST (-0800), fweimer@redhat.com wrote:
> * Darius Rad:
>
>> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via Libc-alpha wrote:
>>> * Darius Rad:
>>>
>>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>>> >> * Andrew Waterman:
>>> >>
>>> >> > This suggests that ld.so, early-stage libc, or possibly both will need
>>> >> > to make this prctl() call, perhaps by parsing the ELF headers of the
>>> >> > binary and each library to determine if the V extension is used.
>>> >>
>>> >> If the string functions use the V extension, it will be enabled
>>> >> unconditionally. So I don't see why it's okay for libc to trigger this
>>> >> alleged UAPI change, when the kernel can't do it by default.
>>> >>
>>> >
>>> > Because the call to enable can fail and userspace needs to deal with that.
>>>
>>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
>>> zero, or perhaps a special CPU register (although that is more unusual).
>>
>> That would indicate that the extension is not present, which is one of, but
>> not the only way it can fail.
>
> I think you should bring down the number of failure modes. HWCAP has
> the advantage that it communicates kernel/hypervisor/firmware/CPU
> support in a single bit, which simplifies the programming model and
> avoids hard-to-detect bugs. It's not clear why it would be beneficial
> to continue on ENOMEM failures here because the system must clearly be
> in bad shape at this point, and launching a new process is very unlikely
> to improve matters. So I think the simpler programming model is the way
> to go here.
>
>> The vector extension relies on dynamically allocated memory in the kernel,
>> which can fail.
The issue I'm worried about is that V needs more space in the
ucontext-type structures. We have an extensibility scheme there so we
think it can be made to work, but IIUC we'll need glibc to be updated to
handle the extended contexts in order to avoid losing state when doing
ucontext-related operations like signal handling.
I don't see a way to handle that with just HWCAP, as we essentially need
some bi-directional communicaton between userspace and the kernel so
they can both decide to turn on V. I don't think we strictly need a
system call to do that, we kicked around the idea of encoding this in
the ELF, but there's a lot of flavors of vector in RISC-V and we've had
trouble trying to encode these in binaries before so it seems easier to
just use the syscall.
> But this failure can be reported as part of execve and clone.
>
>> It also provides the opportunity for the kernel to deny access to the
>> vector extension, perhaps due to administrative policy or other future
>> mechanism.
>
> HWCAP can do this, too.
>
> Thanks,
> Florian
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 17:21 ` Palmer Dabbelt
0 siblings, 0 replies; 147+ messages in thread
From: Palmer Dabbelt @ 2022-12-09 17:21 UTC (permalink / raw)
To: fweimer
Cc: Andrew Waterman, Vineet Gupta, stillson, Paul Walmsley, anup,
atishp, guoren, Conor Dooley, greentime.hu, vincent.chen,
andy.chiu, arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
On Fri, 09 Dec 2022 05:04:23 PST (-0800), fweimer@redhat.com wrote:
> * Darius Rad:
>
>> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via Libc-alpha wrote:
>>> * Darius Rad:
>>>
>>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>>> >> * Andrew Waterman:
>>> >>
>>> >> > This suggests that ld.so, early-stage libc, or possibly both will need
>>> >> > to make this prctl() call, perhaps by parsing the ELF headers of the
>>> >> > binary and each library to determine if the V extension is used.
>>> >>
>>> >> If the string functions use the V extension, it will be enabled
>>> >> unconditionally. So I don't see why it's okay for libc to trigger this
>>> >> alleged UAPI change, when the kernel can't do it by default.
>>> >>
>>> >
>>> > Because the call to enable can fail and userspace needs to deal with that.
>>>
>>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
>>> zero, or perhaps a special CPU register (although that is more unusual).
>>
>> That would indicate that the extension is not present, which is one of, but
>> not the only way it can fail.
>
> I think you should bring down the number of failure modes. HWCAP has
> the advantage that it communicates kernel/hypervisor/firmware/CPU
> support in a single bit, which simplifies the programming model and
> avoids hard-to-detect bugs. It's not clear why it would be beneficial
> to continue on ENOMEM failures here because the system must clearly be
> in bad shape at this point, and launching a new process is very unlikely
> to improve matters. So I think the simpler programming model is the way
> to go here.
>
>> The vector extension relies on dynamically allocated memory in the kernel,
>> which can fail.
The issue I'm worried about is that V needs more space in the
ucontext-type structures. We have an extensibility scheme there so we
think it can be made to work, but IIUC we'll need glibc to be updated to
handle the extended contexts in order to avoid losing state when doing
ucontext-related operations like signal handling.
I don't see a way to handle that with just HWCAP, as we essentially need
some bi-directional communicaton between userspace and the kernel so
they can both decide to turn on V. I don't think we strictly need a
system call to do that, we kicked around the idea of encoding this in
the ELF, but there's a lot of flavors of vector in RISC-V and we've had
trouble trying to encode these in binaries before so it seems easier to
just use the syscall.
> But this failure can be reported as part of execve and clone.
>
>> It also provides the opportunity for the kernel to deny access to the
>> vector extension, perhaps due to administrative policy or other future
>> mechanism.
>
> HWCAP can do this, too.
>
> Thanks,
> Florian
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 17:21 ` Palmer Dabbelt
@ 2022-12-09 19:42 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-09 19:42 UTC (permalink / raw)
To: Palmer Dabbelt, fweimer
Cc: Andrew Waterman, stillson, Paul Walmsley, anup, atishp, guoren,
Conor Dooley, greentime.hu, vincent.chen, andy.chiu, arnd,
linux-kernel, linux-riscv, bjorn, libc-alpha, christoph.muellner,
Aaron Durbin, linux
On 12/9/22 09:21, Palmer Dabbelt wrote:
> On Fri, 09 Dec 2022 05:04:23 PST (-0800), fweimer@redhat.com wrote:
>> * Darius Rad:
>>
>>> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via
>>> Libc-alpha wrote:
>>>> * Darius Rad:
>>>>
>>>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>>>> >> * Andrew Waterman:
>>>> >>
>>>> >> > This suggests that ld.so, early-stage libc, or possibly both
>>>> will need
>>>> >> > to make this prctl() call, perhaps by parsing the ELF headers
>>>> of the
>>>> >> > binary and each library to determine if the V extension is used.
>>>> >>
>>>> >> If the string functions use the V extension, it will be enabled
>>>> >> unconditionally. So I don't see why it's okay for libc to
>>>> trigger this
>>>> >> alleged UAPI change, when the kernel can't do it by default.
>>>> >>
>>>> >
>>>> > Because the call to enable can fail and userspace needs to deal
>>>> with that.
>>>>
>>>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
>>>> zero, or perhaps a special CPU register (although that is more
>>>> unusual).
>>>
>>> That would indicate that the extension is not present, which is one
>>> of, but
>>> not the only way it can fail.
>>
>> I think you should bring down the number of failure modes. HWCAP has
>> the advantage that it communicates kernel/hypervisor/firmware/CPU
>> support in a single bit, which simplifies the programming model and
>> avoids hard-to-detect bugs. It's not clear why it would be beneficial
>> to continue on ENOMEM failures here because the system must clearly be
>> in bad shape at this point, and launching a new process is very unlikely
>> to improve matters. So I think the simpler programming model is the way
>> to go here.
>>
>>> The vector extension relies on dynamically allocated memory in the
>>> kernel,
>>> which can fail.
>
> The issue I'm worried about is that V needs more space in the
> ucontext-type structures. We have an extensibility scheme there so we
> think it can be made to work, but IIUC we'll need glibc to be updated
> to handle the extended contexts in order to avoid losing state when
> doing ucontext-related operations like signal handling.
Sorry this is not relevant to this thread. I started a different thread
on ABI/sigcontext aspects, lets discuss it there.
>
> I don't see a way to handle that with just HWCAP, as we essentially
> need some bi-directional communicaton between userspace and the kernel
> so they can both decide to turn on V. I don't think we strictly need
> a system call to do that, we kicked around the idea of encoding this
> in the ELF, but there's a lot of flavors of vector in RISC-V and we've
> had trouble trying to encode these in binaries before so it seems
> easier to just use the syscall.
>
>> But this failure can be reported as part of execve and clone.
>>
>>> It also provides the opportunity for the kernel to deny access to the
>>> vector extension, perhaps due to administrative policy or other future
>>> mechanism.
>>
>> HWCAP can do this, too.
Having the prctl as general purpose knob to disable the V unit for
various reasons makes sense.
But keeping the V unit disabled by default and using prctl as a
gatekeeper to enable it feels unnecessary and tedious.
Here's my reasoning below (I'm collating comments from prior msgs as well).
1. Doesn't it add another userspace ABI which is already a headache for
this feature. And that needs to be built into not just libc but
potentially other runtimes too. Even after implemention there will be an
interim pain as the new prctl takes time to trickle down into tooling
and headers. Besides the new stuff will never be compatible with older
kernel but that is a minor point since older kernel not supporting V is
a deal breaker anyways.
2. People want the prctl gatekeeping for ability to gracefully handle
memory allocation failure for the extra V-state within kernel. But that
is only additional 4K (for typical 128 wide V regs) per task. If that is
failing, the system is not doing well anyways. Besides it is not an
issue at all since ENOMEM in clone/execve for the additional space
should handle the failure anyways. Only very sophisticated apps would
downgrade from executing V to Scalar code if the prctl failed. Instead
memory allocation is more likely to be an issue when copying V state on
a deep user stack across signal handling but there's nothing we can do
about it.
3. Another argument to prctl gatekeeping is ensuring user-space conforms
to vector length. But isn't the holy grail of RV V-extension VLA (Vector
Length Agnostic) programming. I expect most implements to follow that.
If there are super sophisticated (or dumb) apps that don't follow, they
will fail randomly. I think of Vector Length as any other ISA extensions
- its not that currently apps are required to prctl() for bitmanip
extension if they want to use it. Sure they could use AT_HWCAP (or
/proc/cpuinfo or any other portable way) to query the capability, same
can be done for V as well. Besides vlen is readable from user space so
the app can read it to make sure. My worry is we are providing
additional safety net to a small category of apps at the expense of
making it tiresome for everyone else.
HWCAP solves the kernel to user-space communication of capabilities. The
prctl is for user-space to kernel communication but I'm not convinced it
is really solving problems for the common case.
Thx,
-Vineet
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 19:42 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-09 19:42 UTC (permalink / raw)
To: Palmer Dabbelt, fweimer
Cc: Andrew Waterman, stillson, Paul Walmsley, anup, atishp, guoren,
Conor Dooley, greentime.hu, vincent.chen, andy.chiu, arnd,
linux-kernel, linux-riscv, bjorn, libc-alpha, christoph.muellner,
Aaron Durbin, linux
On 12/9/22 09:21, Palmer Dabbelt wrote:
> On Fri, 09 Dec 2022 05:04:23 PST (-0800), fweimer@redhat.com wrote:
>> * Darius Rad:
>>
>>> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via
>>> Libc-alpha wrote:
>>>> * Darius Rad:
>>>>
>>>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
>>>> >> * Andrew Waterman:
>>>> >>
>>>> >> > This suggests that ld.so, early-stage libc, or possibly both
>>>> will need
>>>> >> > to make this prctl() call, perhaps by parsing the ELF headers
>>>> of the
>>>> >> > binary and each library to determine if the V extension is used.
>>>> >>
>>>> >> If the string functions use the V extension, it will be enabled
>>>> >> unconditionally. So I don't see why it's okay for libc to
>>>> trigger this
>>>> >> alleged UAPI change, when the kernel can't do it by default.
>>>> >>
>>>> >
>>>> > Because the call to enable can fail and userspace needs to deal
>>>> with that.
>>>>
>>>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
>>>> zero, or perhaps a special CPU register (although that is more
>>>> unusual).
>>>
>>> That would indicate that the extension is not present, which is one
>>> of, but
>>> not the only way it can fail.
>>
>> I think you should bring down the number of failure modes. HWCAP has
>> the advantage that it communicates kernel/hypervisor/firmware/CPU
>> support in a single bit, which simplifies the programming model and
>> avoids hard-to-detect bugs. It's not clear why it would be beneficial
>> to continue on ENOMEM failures here because the system must clearly be
>> in bad shape at this point, and launching a new process is very unlikely
>> to improve matters. So I think the simpler programming model is the way
>> to go here.
>>
>>> The vector extension relies on dynamically allocated memory in the
>>> kernel,
>>> which can fail.
>
> The issue I'm worried about is that V needs more space in the
> ucontext-type structures. We have an extensibility scheme there so we
> think it can be made to work, but IIUC we'll need glibc to be updated
> to handle the extended contexts in order to avoid losing state when
> doing ucontext-related operations like signal handling.
Sorry this is not relevant to this thread. I started a different thread
on ABI/sigcontext aspects, lets discuss it there.
>
> I don't see a way to handle that with just HWCAP, as we essentially
> need some bi-directional communicaton between userspace and the kernel
> so they can both decide to turn on V. I don't think we strictly need
> a system call to do that, we kicked around the idea of encoding this
> in the ELF, but there's a lot of flavors of vector in RISC-V and we've
> had trouble trying to encode these in binaries before so it seems
> easier to just use the syscall.
>
>> But this failure can be reported as part of execve and clone.
>>
>>> It also provides the opportunity for the kernel to deny access to the
>>> vector extension, perhaps due to administrative policy or other future
>>> mechanism.
>>
>> HWCAP can do this, too.
Having the prctl as general purpose knob to disable the V unit for
various reasons makes sense.
But keeping the V unit disabled by default and using prctl as a
gatekeeper to enable it feels unnecessary and tedious.
Here's my reasoning below (I'm collating comments from prior msgs as well).
1. Doesn't it add another userspace ABI which is already a headache for
this feature. And that needs to be built into not just libc but
potentially other runtimes too. Even after implemention there will be an
interim pain as the new prctl takes time to trickle down into tooling
and headers. Besides the new stuff will never be compatible with older
kernel but that is a minor point since older kernel not supporting V is
a deal breaker anyways.
2. People want the prctl gatekeeping for ability to gracefully handle
memory allocation failure for the extra V-state within kernel. But that
is only additional 4K (for typical 128 wide V regs) per task. If that is
failing, the system is not doing well anyways. Besides it is not an
issue at all since ENOMEM in clone/execve for the additional space
should handle the failure anyways. Only very sophisticated apps would
downgrade from executing V to Scalar code if the prctl failed. Instead
memory allocation is more likely to be an issue when copying V state on
a deep user stack across signal handling but there's nothing we can do
about it.
3. Another argument to prctl gatekeeping is ensuring user-space conforms
to vector length. But isn't the holy grail of RV V-extension VLA (Vector
Length Agnostic) programming. I expect most implements to follow that.
If there are super sophisticated (or dumb) apps that don't follow, they
will fail randomly. I think of Vector Length as any other ISA extensions
- its not that currently apps are required to prctl() for bitmanip
extension if they want to use it. Sure they could use AT_HWCAP (or
/proc/cpuinfo or any other portable way) to query the capability, same
can be done for V as well. Besides vlen is readable from user space so
the app can read it to make sure. My worry is we are providing
additional safety net to a small category of apps at the expense of
making it tiresome for everyone else.
HWCAP solves the kernel to user-space communication of capabilities. The
prctl is for user-space to kernel communication but I'm not convinced it
is really solving problems for the common case.
Thx,
-Vineet
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 19:42 ` Vineet Gupta
@ 2022-12-09 19:58 ` Andrew Waterman
-1 siblings, 0 replies; 147+ messages in thread
From: Andrew Waterman @ 2022-12-09 19:58 UTC (permalink / raw)
To: Vineet Gupta
Cc: Palmer Dabbelt, fweimer, stillson, Paul Walmsley, anup, atishp,
guoren, Conor Dooley, greentime.hu, vincent.chen, andy.chiu,
arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
On Fri, Dec 9, 2022 at 11:42 AM Vineet Gupta <vineetg@rivosinc.com> wrote:
>
>
> On 12/9/22 09:21, Palmer Dabbelt wrote:
> > On Fri, 09 Dec 2022 05:04:23 PST (-0800), fweimer@redhat.com wrote:
> >> * Darius Rad:
> >>
> >>> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via
> >>> Libc-alpha wrote:
> >>>> * Darius Rad:
> >>>>
> >>>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
> >>>> >> * Andrew Waterman:
> >>>> >>
> >>>> >> > This suggests that ld.so, early-stage libc, or possibly both
> >>>> will need
> >>>> >> > to make this prctl() call, perhaps by parsing the ELF headers
> >>>> of the
> >>>> >> > binary and each library to determine if the V extension is used.
> >>>> >>
> >>>> >> If the string functions use the V extension, it will be enabled
> >>>> >> unconditionally. So I don't see why it's okay for libc to
> >>>> trigger this
> >>>> >> alleged UAPI change, when the kernel can't do it by default.
> >>>> >>
> >>>> >
> >>>> > Because the call to enable can fail and userspace needs to deal
> >>>> with that.
> >>>>
> >>>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
> >>>> zero, or perhaps a special CPU register (although that is more
> >>>> unusual).
> >>>
> >>> That would indicate that the extension is not present, which is one
> >>> of, but
> >>> not the only way it can fail.
> >>
> >> I think you should bring down the number of failure modes. HWCAP has
> >> the advantage that it communicates kernel/hypervisor/firmware/CPU
> >> support in a single bit, which simplifies the programming model and
> >> avoids hard-to-detect bugs. It's not clear why it would be beneficial
> >> to continue on ENOMEM failures here because the system must clearly be
> >> in bad shape at this point, and launching a new process is very unlikely
> >> to improve matters. So I think the simpler programming model is the way
> >> to go here.
> >>
> >>> The vector extension relies on dynamically allocated memory in the
> >>> kernel,
> >>> which can fail.
> >
> > The issue I'm worried about is that V needs more space in the
> > ucontext-type structures. We have an extensibility scheme there so we
> > think it can be made to work, but IIUC we'll need glibc to be updated
> > to handle the extended contexts in order to avoid losing state when
> > doing ucontext-related operations like signal handling.
>
> Sorry this is not relevant to this thread. I started a different thread
> on ABI/sigcontext aspects, lets discuss it there.
>
> >
> > I don't see a way to handle that with just HWCAP, as we essentially
> > need some bi-directional communicaton between userspace and the kernel
> > so they can both decide to turn on V. I don't think we strictly need
> > a system call to do that, we kicked around the idea of encoding this
> > in the ELF, but there's a lot of flavors of vector in RISC-V and we've
> > had trouble trying to encode these in binaries before so it seems
> > easier to just use the syscall.
> >
> >> But this failure can be reported as part of execve and clone.
> >>
> >>> It also provides the opportunity for the kernel to deny access to the
> >>> vector extension, perhaps due to administrative policy or other future
> >>> mechanism.
> >>
> >> HWCAP can do this, too.
>
> Having the prctl as general purpose knob to disable the V unit for
> various reasons makes sense.
>
> But keeping the V unit disabled by default and using prctl as a
> gatekeeper to enable it feels unnecessary and tedious.
> Here's my reasoning below (I'm collating comments from prior msgs as well).
>
> 1. Doesn't it add another userspace ABI which is already a headache for
> this feature. And that needs to be built into not just libc but
> potentially other runtimes too. Even after implemention there will be an
> interim pain as the new prctl takes time to trickle down into tooling
> and headers. Besides the new stuff will never be compatible with older
> kernel but that is a minor point since older kernel not supporting V is
> a deal breaker anyways.
>
> 2. People want the prctl gatekeeping for ability to gracefully handle
> memory allocation failure for the extra V-state within kernel. But that
> is only additional 4K (for typical 128 wide V regs) per task. If that is
> failing, the system is not doing well anyways. Besides it is not an
> issue at all since ENOMEM in clone/execve for the additional space
> should handle the failure anyways. Only very sophisticated apps would
> downgrade from executing V to Scalar code if the prctl failed. Instead
> memory allocation is more likely to be an issue when copying V state on
> a deep user stack across signal handling but there's nothing we can do
> about it.
>
> 3. Another argument to prctl gatekeeping is ensuring user-space conforms
> to vector length. But isn't the holy grail of RV V-extension VLA (Vector
> Length Agnostic) programming.
Yes, a suitable ABI for the V extension should cater cleanly to the
VLA model, since that's a major selling point of this ISA. The
baseline assumption should be that programs will execute correctly
regardless of VLEN (subject to the constraint that the V extension
requires VLEN >= 128, of course).
It's of course valid to construct programs with VLEN-dependent
behavior (e.g. dynamic dispatch to routines optimized differently for
different VLEN), but it should be considered the program's
responsibility to get that right. I don't think the ABI needs to
furnish guard rails.
> I expect most implements to follow that.
> If there are super sophisticated (or dumb) apps that don't follow, they
> will fail randomly. I think of Vector Length as any other ISA extensions
> - its not that currently apps are required to prctl() for bitmanip
> extension if they want to use it. Sure they could use AT_HWCAP (or
> /proc/cpuinfo or any other portable way) to query the capability, same
> can be done for V as well. Besides vlen is readable from user space so
> the app can read it to make sure. My worry is we are providing
> additional safety net to a small category of apps at the expense of
> making it tiresome for everyone else.
>
> HWCAP solves the kernel to user-space communication of capabilities. The
> prctl is for user-space to kernel communication but I'm not convinced it
> is really solving problems for the common case.
>
> Thx,
> -Vineet
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 19:58 ` Andrew Waterman
0 siblings, 0 replies; 147+ messages in thread
From: Andrew Waterman @ 2022-12-09 19:58 UTC (permalink / raw)
To: Vineet Gupta
Cc: Palmer Dabbelt, fweimer, stillson, Paul Walmsley, anup, atishp,
guoren, Conor Dooley, greentime.hu, vincent.chen, andy.chiu,
arnd, linux-kernel, linux-riscv, bjorn, libc-alpha,
christoph.muellner, Aaron Durbin, linux
On Fri, Dec 9, 2022 at 11:42 AM Vineet Gupta <vineetg@rivosinc.com> wrote:
>
>
> On 12/9/22 09:21, Palmer Dabbelt wrote:
> > On Fri, 09 Dec 2022 05:04:23 PST (-0800), fweimer@redhat.com wrote:
> >> * Darius Rad:
> >>
> >>> On Fri, Dec 09, 2022 at 01:32:33PM +0100, Florian Weimer via
> >>> Libc-alpha wrote:
> >>>> * Darius Rad:
> >>>>
> >>>> > On Fri, Dec 09, 2022 at 11:02:57AM +0100, Florian Weimer wrote:
> >>>> >> * Andrew Waterman:
> >>>> >>
> >>>> >> > This suggests that ld.so, early-stage libc, or possibly both
> >>>> will need
> >>>> >> > to make this prctl() call, perhaps by parsing the ELF headers
> >>>> of the
> >>>> >> > binary and each library to determine if the V extension is used.
> >>>> >>
> >>>> >> If the string functions use the V extension, it will be enabled
> >>>> >> unconditionally. So I don't see why it's okay for libc to
> >>>> trigger this
> >>>> >> alleged UAPI change, when the kernel can't do it by default.
> >>>> >>
> >>>> >
> >>>> > Because the call to enable can fail and userspace needs to deal
> >>>> with that.
> >>>>
> >>>> Failure is usually indicated by an AT_HWCAP or AT_HWCAP2 bit remaining
> >>>> zero, or perhaps a special CPU register (although that is more
> >>>> unusual).
> >>>
> >>> That would indicate that the extension is not present, which is one
> >>> of, but
> >>> not the only way it can fail.
> >>
> >> I think you should bring down the number of failure modes. HWCAP has
> >> the advantage that it communicates kernel/hypervisor/firmware/CPU
> >> support in a single bit, which simplifies the programming model and
> >> avoids hard-to-detect bugs. It's not clear why it would be beneficial
> >> to continue on ENOMEM failures here because the system must clearly be
> >> in bad shape at this point, and launching a new process is very unlikely
> >> to improve matters. So I think the simpler programming model is the way
> >> to go here.
> >>
> >>> The vector extension relies on dynamically allocated memory in the
> >>> kernel,
> >>> which can fail.
> >
> > The issue I'm worried about is that V needs more space in the
> > ucontext-type structures. We have an extensibility scheme there so we
> > think it can be made to work, but IIUC we'll need glibc to be updated
> > to handle the extended contexts in order to avoid losing state when
> > doing ucontext-related operations like signal handling.
>
> Sorry this is not relevant to this thread. I started a different thread
> on ABI/sigcontext aspects, lets discuss it there.
>
> >
> > I don't see a way to handle that with just HWCAP, as we essentially
> > need some bi-directional communicaton between userspace and the kernel
> > so they can both decide to turn on V. I don't think we strictly need
> > a system call to do that, we kicked around the idea of encoding this
> > in the ELF, but there's a lot of flavors of vector in RISC-V and we've
> > had trouble trying to encode these in binaries before so it seems
> > easier to just use the syscall.
> >
> >> But this failure can be reported as part of execve and clone.
> >>
> >>> It also provides the opportunity for the kernel to deny access to the
> >>> vector extension, perhaps due to administrative policy or other future
> >>> mechanism.
> >>
> >> HWCAP can do this, too.
>
> Having the prctl as general purpose knob to disable the V unit for
> various reasons makes sense.
>
> But keeping the V unit disabled by default and using prctl as a
> gatekeeper to enable it feels unnecessary and tedious.
> Here's my reasoning below (I'm collating comments from prior msgs as well).
>
> 1. Doesn't it add another userspace ABI which is already a headache for
> this feature. And that needs to be built into not just libc but
> potentially other runtimes too. Even after implemention there will be an
> interim pain as the new prctl takes time to trickle down into tooling
> and headers. Besides the new stuff will never be compatible with older
> kernel but that is a minor point since older kernel not supporting V is
> a deal breaker anyways.
>
> 2. People want the prctl gatekeeping for ability to gracefully handle
> memory allocation failure for the extra V-state within kernel. But that
> is only additional 4K (for typical 128 wide V regs) per task. If that is
> failing, the system is not doing well anyways. Besides it is not an
> issue at all since ENOMEM in clone/execve for the additional space
> should handle the failure anyways. Only very sophisticated apps would
> downgrade from executing V to Scalar code if the prctl failed. Instead
> memory allocation is more likely to be an issue when copying V state on
> a deep user stack across signal handling but there's nothing we can do
> about it.
>
> 3. Another argument to prctl gatekeeping is ensuring user-space conforms
> to vector length. But isn't the holy grail of RV V-extension VLA (Vector
> Length Agnostic) programming.
Yes, a suitable ABI for the V extension should cater cleanly to the
VLA model, since that's a major selling point of this ISA. The
baseline assumption should be that programs will execute correctly
regardless of VLEN (subject to the constraint that the V extension
requires VLEN >= 128, of course).
It's of course valid to construct programs with VLEN-dependent
behavior (e.g. dynamic dispatch to routines optimized differently for
different VLEN), but it should be considered the program's
responsibility to get that right. I don't think the ABI needs to
furnish guard rails.
> I expect most implements to follow that.
> If there are super sophisticated (or dumb) apps that don't follow, they
> will fail randomly. I think of Vector Length as any other ISA extensions
> - its not that currently apps are required to prctl() for bitmanip
> extension if they want to use it. Sure they could use AT_HWCAP (or
> /proc/cpuinfo or any other portable way) to query the capability, same
> can be done for V as well. Besides vlen is readable from user space so
> the app can read it to make sure. My worry is we are providing
> additional safety net to a small category of apps at the expense of
> making it tiresome for everyone else.
>
> HWCAP solves the kernel to user-space communication of capabilities. The
> prctl is for user-space to kernel communication but I'm not convinced it
> is really solving problems for the common case.
>
> Thx,
> -Vineet
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 19:42 ` Vineet Gupta
@ 2022-12-13 16:43 ` Darius Rad
-1 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-13 16:43 UTC (permalink / raw)
To: Vineet Gupta
Cc: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>
> But keeping the V unit disabled by default and using prctl as a gatekeeper
> to enable it feels unnecessary and tedious.
> Here's my reasoning below (I'm collating comments from prior msgs as well).
Please reference the previous discussion [1] which has covered topics that
have not been discussed recently.
[1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
>
> 1. Doesn't it add another userspace ABI which is already a headache for this
> feature. And that needs to be built into not just libc but potentially other
> runtimes too. Even after implemention there will be an interim pain as the
> new prctl takes time to trickle down into tooling and headers. Besides the
> new stuff will never be compatible with older kernel but that is a minor
> point since older kernel not supporting V is a deal breaker anyways.
>
None of this is relevant because there is no existing user space ABI for
vector. It is being invented now. If this is done poorly, for example, by
missing this opportunity to add a mechanism for user space to request use
of the vector extension, it will be much more painful to add later.
> 2. People want the prctl gatekeeping for ability to gracefully handle memory
> allocation failure for the extra V-state within kernel. But that is only
> additional 4K (for typical 128 wide V regs) per task.
But vector state scales up to as much as 256k. Are you suggesting that
there is no possibility that future systems would support more than
VLEN=128?
> If that is failing,
> the system is not doing well anyways. Besides it is not an issue at all
> since ENOMEM in clone/execve for the additional space should handle the
> failure anyways. Only very sophisticated apps would downgrade from executing
> V to Scalar code if the prctl failed.
This seems unlikely. As vector support does not exist in any present
hardware, and the vector extension is only optional in the RISC-V profiles
that include it, I would think that it is almost certain that any
application that supports V would have a fallback path for when the V
extension is not available.
Another motivation for requiring that user space request use of the vector
extension is that the vector unit may be shared between multiple harts
and/or have power or performance implications in the system. By requiring
that user space request access, it allows the system to decline that
access, and user space can handle this gracefully.
If we add a mechanism for user space to request access to the vector
extension, and it turns out that it was unnecessary, the worst that has
happened is a slight inconvenience.
If we do not add such a mechanism, and later determine that it is
necessary, we have a much greater problem. There would be backward
compatibility issues with the ABI, and such a mechanism could probably not
be fully implemented at all due to the desire to support potential future
legacy vector code.
This is a similar problem on x86. According to some, it was handled poorly
with AVX-512 by missing this type of mechanism, and improved with AMX [2].
There is opportunity to learn from that experience and do things better on
RISC-V.
[2] https://lore.kernel.org/lkml/87k0ntazyn.ffs@nanos.tec.linutronix.de/
// darius
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-13 16:43 ` Darius Rad
0 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-13 16:43 UTC (permalink / raw)
To: Vineet Gupta
Cc: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>
> But keeping the V unit disabled by default and using prctl as a gatekeeper
> to enable it feels unnecessary and tedious.
> Here's my reasoning below (I'm collating comments from prior msgs as well).
Please reference the previous discussion [1] which has covered topics that
have not been discussed recently.
[1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
>
> 1. Doesn't it add another userspace ABI which is already a headache for this
> feature. And that needs to be built into not just libc but potentially other
> runtimes too. Even after implemention there will be an interim pain as the
> new prctl takes time to trickle down into tooling and headers. Besides the
> new stuff will never be compatible with older kernel but that is a minor
> point since older kernel not supporting V is a deal breaker anyways.
>
None of this is relevant because there is no existing user space ABI for
vector. It is being invented now. If this is done poorly, for example, by
missing this opportunity to add a mechanism for user space to request use
of the vector extension, it will be much more painful to add later.
> 2. People want the prctl gatekeeping for ability to gracefully handle memory
> allocation failure for the extra V-state within kernel. But that is only
> additional 4K (for typical 128 wide V regs) per task.
But vector state scales up to as much as 256k. Are you suggesting that
there is no possibility that future systems would support more than
VLEN=128?
> If that is failing,
> the system is not doing well anyways. Besides it is not an issue at all
> since ENOMEM in clone/execve for the additional space should handle the
> failure anyways. Only very sophisticated apps would downgrade from executing
> V to Scalar code if the prctl failed.
This seems unlikely. As vector support does not exist in any present
hardware, and the vector extension is only optional in the RISC-V profiles
that include it, I would think that it is almost certain that any
application that supports V would have a fallback path for when the V
extension is not available.
Another motivation for requiring that user space request use of the vector
extension is that the vector unit may be shared between multiple harts
and/or have power or performance implications in the system. By requiring
that user space request access, it allows the system to decline that
access, and user space can handle this gracefully.
If we add a mechanism for user space to request access to the vector
extension, and it turns out that it was unnecessary, the worst that has
happened is a slight inconvenience.
If we do not add such a mechanism, and later determine that it is
necessary, we have a much greater problem. There would be backward
compatibility issues with the ABI, and such a mechanism could probably not
be fully implemented at all due to the desire to support potential future
legacy vector code.
This is a similar problem on x86. According to some, it was handled poorly
with AVX-512 by missing this type of mechanism, and improved with AMX [2].
There is opportunity to learn from that experience and do things better on
RISC-V.
[2] https://lore.kernel.org/lkml/87k0ntazyn.ffs@nanos.tec.linutronix.de/
// darius
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-13 16:43 ` Darius Rad
@ 2022-12-14 20:07 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-14 20:07 UTC (permalink / raw)
To: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/13/22 08:43, Darius Rad wrote:
> On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>> But keeping the V unit disabled by default and using prctl as a gatekeeper
>> to enable it feels unnecessary and tedious.
>> Here's my reasoning below (I'm collating comments from prior msgs as well).
> Please reference the previous discussion [1] which has covered topics that
> have not been discussed recently.
>
> [1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
I sure read thru that thread, and many more :-) to get context.
The highlight is we should something because AVX/AMX do so (or failed to
do so).
But on the flip side ARM SVE is not disabling this by default.
Your other concerns seems to be potential power implications for leaving
it on and sharing of V unit across harts (see more on that below)
Maybe leaving it on all the time will be motivation for hw designers to
be more considerate of the idle power draw.
>
>> 2. People want the prctl gatekeeping for ability to gracefully handle memory
>> allocation failure for the extra V-state within kernel. But that is only
>> additional 4K (for typical 128 wide V regs) per task.
> But vector state scales up to as much as 256k. Are you suggesting that
> there is no possibility that future systems would support more than
> VLEN=128?
I mentioned "typical". And below also said that memory allocation
concerns are moot, since fork/execve failures due to failing to allocate
would take care of those anyways.
>> If that is failing,
>> the system is not doing well anyways. Besides it is not an issue at all
>> since ENOMEM in clone/execve for the additional space should handle the
>> failure anyways. Only very sophisticated apps would downgrade from executing
>> V to Scalar code if the prctl failed.
> This seems unlikely. As vector support does not exist in any present
> hardware, and the vector extension is only optional in the RISC-V profiles
> that include it, I would think that it is almost certain that any
> application that supports V would have a fallback path for when the V
> extension is not available.
For specialized cases sure we would expect fat binaries with IFUNC based
dispatches (glibc mem*/str* are obvious examples). But with newer
compilers autovec is increasing becoming default even at medium
optimization levels such as -O2. So V code littered all over is just a
matter of time for the profiles/variants which allow V. For less capable
systems w/o V this is all but moot discussion since kernel itself need
not be built with V enabled.
> Another motivation for requiring that user space request use of the vector
> extension is that the vector unit may be shared between multiple harts
> and/or have power or performance implications in the system. By requiring
> that user space request access, it allows the system to decline that
> access, and user space can handle this gracefully.
But in this specific example won't the prctl cause more pain. So 2
concurrent processes on 2 different harts with shared V unit. One sends
prctl to enable and other wants to disable, what would the kernel do.
Will it just be whoever ends up running later wins. Granted I'm not too
familiar with how such a cross-hart sharing would work in a Vector
instructions being part of ISA (vs. Vector accelerator with job
push/pull approach)
Honestly I'm sympathetic to your power concerns with keeping V enabled
all the time. But the mechanics of implementing this prctl makes me
wary. Assuming this is done from dynamic loader, it implies loader
itself needs to be built with V disabled. This also leaves bunch of perf
on table since loader does tons of of string and memory operations which
could potentially benefit from V enabled code, granted it is not deal
breaker.
> If we add a mechanism for user space to request access to the vector
> extension, and it turns out that it was unnecessary, the worst that has
> happened is a slight inconvenience.
>
> If we do not add such a mechanism, and later determine that it is
> necessary, we have a much greater problem. There would be backward
> compatibility issues with the ABI, and such a mechanism could probably not
> be fully implemented at all due to the desire to support potential future
> legacy vector code.
Very true, but this in itself is not sufficient of a reason to warrant
adding it now.
> This is a similar problem on x86. According to some, it was handled poorly
> with AVX-512 by missing this type of mechanism, and improved with AMX [2].
> There is opportunity to learn from that experience and do things better on
> RISC-V.
>
> [2] https://lore.kernel.org/lkml/87k0ntazyn.ffs@nanos.tec.linutronix.de/
Right, but then why did ARM SVE guys chose to not take this path.
-Vineet
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-14 20:07 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-14 20:07 UTC (permalink / raw)
To: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/13/22 08:43, Darius Rad wrote:
> On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>> But keeping the V unit disabled by default and using prctl as a gatekeeper
>> to enable it feels unnecessary and tedious.
>> Here's my reasoning below (I'm collating comments from prior msgs as well).
> Please reference the previous discussion [1] which has covered topics that
> have not been discussed recently.
>
> [1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
I sure read thru that thread, and many more :-) to get context.
The highlight is we should something because AVX/AMX do so (or failed to
do so).
But on the flip side ARM SVE is not disabling this by default.
Your other concerns seems to be potential power implications for leaving
it on and sharing of V unit across harts (see more on that below)
Maybe leaving it on all the time will be motivation for hw designers to
be more considerate of the idle power draw.
>
>> 2. People want the prctl gatekeeping for ability to gracefully handle memory
>> allocation failure for the extra V-state within kernel. But that is only
>> additional 4K (for typical 128 wide V regs) per task.
> But vector state scales up to as much as 256k. Are you suggesting that
> there is no possibility that future systems would support more than
> VLEN=128?
I mentioned "typical". And below also said that memory allocation
concerns are moot, since fork/execve failures due to failing to allocate
would take care of those anyways.
>> If that is failing,
>> the system is not doing well anyways. Besides it is not an issue at all
>> since ENOMEM in clone/execve for the additional space should handle the
>> failure anyways. Only very sophisticated apps would downgrade from executing
>> V to Scalar code if the prctl failed.
> This seems unlikely. As vector support does not exist in any present
> hardware, and the vector extension is only optional in the RISC-V profiles
> that include it, I would think that it is almost certain that any
> application that supports V would have a fallback path for when the V
> extension is not available.
For specialized cases sure we would expect fat binaries with IFUNC based
dispatches (glibc mem*/str* are obvious examples). But with newer
compilers autovec is increasing becoming default even at medium
optimization levels such as -O2. So V code littered all over is just a
matter of time for the profiles/variants which allow V. For less capable
systems w/o V this is all but moot discussion since kernel itself need
not be built with V enabled.
> Another motivation for requiring that user space request use of the vector
> extension is that the vector unit may be shared between multiple harts
> and/or have power or performance implications in the system. By requiring
> that user space request access, it allows the system to decline that
> access, and user space can handle this gracefully.
But in this specific example won't the prctl cause more pain. So 2
concurrent processes on 2 different harts with shared V unit. One sends
prctl to enable and other wants to disable, what would the kernel do.
Will it just be whoever ends up running later wins. Granted I'm not too
familiar with how such a cross-hart sharing would work in a Vector
instructions being part of ISA (vs. Vector accelerator with job
push/pull approach)
Honestly I'm sympathetic to your power concerns with keeping V enabled
all the time. But the mechanics of implementing this prctl makes me
wary. Assuming this is done from dynamic loader, it implies loader
itself needs to be built with V disabled. This also leaves bunch of perf
on table since loader does tons of of string and memory operations which
could potentially benefit from V enabled code, granted it is not deal
breaker.
> If we add a mechanism for user space to request access to the vector
> extension, and it turns out that it was unnecessary, the worst that has
> happened is a slight inconvenience.
>
> If we do not add such a mechanism, and later determine that it is
> necessary, we have a much greater problem. There would be backward
> compatibility issues with the ABI, and such a mechanism could probably not
> be fully implemented at all due to the desire to support potential future
> legacy vector code.
Very true, but this in itself is not sufficient of a reason to warrant
adding it now.
> This is a similar problem on x86. According to some, it was handled poorly
> with AVX-512 by missing this type of mechanism, and improved with AMX [2].
> There is opportunity to learn from that experience and do things better on
> RISC-V.
>
> [2] https://lore.kernel.org/lkml/87k0ntazyn.ffs@nanos.tec.linutronix.de/
Right, but then why did ARM SVE guys chose to not take this path.
-Vineet
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-14 20:07 ` Vineet Gupta
@ 2022-12-14 23:13 ` Samuel Holland
-1 siblings, 0 replies; 147+ messages in thread
From: Samuel Holland @ 2022-12-14 23:13 UTC (permalink / raw)
To: Vineet Gupta, Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/14/22 14:07, Vineet Gupta wrote:
> On 12/13/22 08:43, Darius Rad wrote:
>> On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>>> If that is failing,
>>> the system is not doing well anyways. Besides it is not an issue at all
>>> since ENOMEM in clone/execve for the additional space should handle the
>>> failure anyways. Only very sophisticated apps would downgrade from
>>> executing
>>> V to Scalar code if the prctl failed.
>> This seems unlikely. As vector support does not exist in any present
>> hardware, and the vector extension is only optional in the RISC-V
>> profiles
>> that include it, I would think that it is almost certain that any
>> application that supports V would have a fallback path for when the V
>> extension is not available.
>
> For specialized cases sure we would expect fat binaries with IFUNC based
> dispatches (glibc mem*/str* are obvious examples). But with newer
> compilers autovec is increasing becoming default even at medium
> optimization levels such as -O2. So V code littered all over is just a
> matter of time for the profiles/variants which allow V. For less capable
Autovectorization is only possible where the profile *requires* V. And
no existing profile does that.
> systems w/o V this is all but moot discussion since kernel itself need
> not be built with V enabled.
Distro kernels will be built with V enabled, and will spend most of
their time running on systems without V.
>> Another motivation for requiring that user space request use of the
>> vector
>> extension is that the vector unit may be shared between multiple harts
>> and/or have power or performance implications in the system. By
>> requiring
>> that user space request access, it allows the system to decline that
>> access, and user space can handle this gracefully.
>
> But in this specific example won't the prctl cause more pain. So 2
> concurrent processes on 2 different harts with shared V unit. One sends
> prctl to enable and other wants to disable, what would the kernel do.
> Will it just be whoever ends up running later wins. Granted I'm not too
> familiar with how such a cross-hart sharing would work in a Vector
> instructions being part of ISA (vs. Vector accelerator with job
> push/pull approach)
The vector execution unit can be shared between harts, not the vector
state. Think SMT.
> Honestly I'm sympathetic to your power concerns with keeping V enabled
> all the time. But the mechanics of implementing this prctl makes me
> wary. Assuming this is done from dynamic loader, it implies loader
> itself needs to be built with V disabled. This also leaves bunch of perf
> on table since loader does tons of of string and memory operations which
> could potentially benefit from V enabled code, granted it is not deal
> breaker.
The dynamic loader can do dynamic dispatch based on the result of the
prctl() just as well as any other library.
And the distro's dynamic loader can't be compiled with autovectorization
enabled anyway, because again the profiles support processors without V.
>> If we add a mechanism for user space to request access to the vector
>> extension, and it turns out that it was unnecessary, the worst that has
>> happened is a slight inconvenience.
>>
>> If we do not add such a mechanism, and later determine that it is
>> necessary, we have a much greater problem. There would be backward
>> compatibility issues with the ABI, and such a mechanism could probably
>> not
>> be fully implemented at all due to the desire to support potential future
>> legacy vector code.
>
> Very true, but this in itself is not sufficient of a reason to warrant
> adding it now.
It is, because changing the sigcontext layout without an opt-in is
already an ABI break (it completely blows past MINSIGSTKSZ).
Regards,
Samuel
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-14 23:13 ` Samuel Holland
0 siblings, 0 replies; 147+ messages in thread
From: Samuel Holland @ 2022-12-14 23:13 UTC (permalink / raw)
To: Vineet Gupta, Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/14/22 14:07, Vineet Gupta wrote:
> On 12/13/22 08:43, Darius Rad wrote:
>> On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>>> If that is failing,
>>> the system is not doing well anyways. Besides it is not an issue at all
>>> since ENOMEM in clone/execve for the additional space should handle the
>>> failure anyways. Only very sophisticated apps would downgrade from
>>> executing
>>> V to Scalar code if the prctl failed.
>> This seems unlikely. As vector support does not exist in any present
>> hardware, and the vector extension is only optional in the RISC-V
>> profiles
>> that include it, I would think that it is almost certain that any
>> application that supports V would have a fallback path for when the V
>> extension is not available.
>
> For specialized cases sure we would expect fat binaries with IFUNC based
> dispatches (glibc mem*/str* are obvious examples). But with newer
> compilers autovec is increasing becoming default even at medium
> optimization levels such as -O2. So V code littered all over is just a
> matter of time for the profiles/variants which allow V. For less capable
Autovectorization is only possible where the profile *requires* V. And
no existing profile does that.
> systems w/o V this is all but moot discussion since kernel itself need
> not be built with V enabled.
Distro kernels will be built with V enabled, and will spend most of
their time running on systems without V.
>> Another motivation for requiring that user space request use of the
>> vector
>> extension is that the vector unit may be shared between multiple harts
>> and/or have power or performance implications in the system. By
>> requiring
>> that user space request access, it allows the system to decline that
>> access, and user space can handle this gracefully.
>
> But in this specific example won't the prctl cause more pain. So 2
> concurrent processes on 2 different harts with shared V unit. One sends
> prctl to enable and other wants to disable, what would the kernel do.
> Will it just be whoever ends up running later wins. Granted I'm not too
> familiar with how such a cross-hart sharing would work in a Vector
> instructions being part of ISA (vs. Vector accelerator with job
> push/pull approach)
The vector execution unit can be shared between harts, not the vector
state. Think SMT.
> Honestly I'm sympathetic to your power concerns with keeping V enabled
> all the time. But the mechanics of implementing this prctl makes me
> wary. Assuming this is done from dynamic loader, it implies loader
> itself needs to be built with V disabled. This also leaves bunch of perf
> on table since loader does tons of of string and memory operations which
> could potentially benefit from V enabled code, granted it is not deal
> breaker.
The dynamic loader can do dynamic dispatch based on the result of the
prctl() just as well as any other library.
And the distro's dynamic loader can't be compiled with autovectorization
enabled anyway, because again the profiles support processors without V.
>> If we add a mechanism for user space to request access to the vector
>> extension, and it turns out that it was unnecessary, the worst that has
>> happened is a slight inconvenience.
>>
>> If we do not add such a mechanism, and later determine that it is
>> necessary, we have a much greater problem. There would be backward
>> compatibility issues with the ABI, and such a mechanism could probably
>> not
>> be fully implemented at all due to the desire to support potential future
>> legacy vector code.
>
> Very true, but this in itself is not sufficient of a reason to warrant
> adding it now.
It is, because changing the sigcontext layout without an opt-in is
already an ABI break (it completely blows past MINSIGSTKSZ).
Regards,
Samuel
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-14 20:07 ` Vineet Gupta
@ 2022-12-15 2:09 ` Darius Rad
-1 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-15 2:09 UTC (permalink / raw)
To: Vineet Gupta
Cc: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Wed, Dec 14, 2022 at 12:07:03PM -0800, Vineet Gupta wrote:
> On 12/13/22 08:43, Darius Rad wrote:
> > On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
> > > But keeping the V unit disabled by default and using prctl as a gatekeeper
> > > to enable it feels unnecessary and tedious.
> > > Here's my reasoning below (I'm collating comments from prior msgs as well).
> > Please reference the previous discussion [1] which has covered topics that
> > have not been discussed recently.
> >
> > [1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
>
> I sure read thru that thread, and many more :-) to get context.
> The highlight is we should something because AVX/AMX do so (or failed to do
> so).
> But on the flip side ARM SVE is not disabling this by default.
> Your other concerns seems to be potential power implications for leaving it
> on and sharing of V unit across harts (see more on that below)
> Maybe leaving it on all the time will be motivation for hw designers to be
> more considerate of the idle power draw.
>
That is not quite the same take away I had from the AMX discussion. I
would also disagree that ARM SVE is not disabling things by default,
although the behavior is somewhat different.
The significant point that I see from that discussion, which also applies
to SVE, and also applies to RISC-V vector, is that the extension is
necessarily dependant upon a functional unit that is reasonably large with
respect to the size of the processor and has a significant amount of
additional architectural state. The argument there is that AMX/SVE/RVV is
a significant system resource and should be treated as such: by having the
kernel track usage of it and by having a process specifically request
access to it.
For any of AMX/SVE/RVV, use of the extension operates as follows:
1. A process requests access to the extension,
2. The kernel allocates memory for the additional state for this process,
3. The kernel enables the extension for the process, and finally
4. The process is able to use the instructions.
I don't recall the exact details, but my understanding is that AMX is going
to use an x86 specific mechanism and require and explict request from user
space to request access to AMX.
For SVE, it is in fact disabled by default in the kernel. When a thread
executes the first SVE instruction, it will cause an exception, the kernel
will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
instructions will proceed without exceptions. Although SVE is disabled by
default, it is enabled automatically. Since this is done automatically
during an exception handler, there is no opportunity for memory allocation
errors to be reported, as there are in the AMX case.
For RVV, I do not recall ever seeing Linux patches that automatically enable
vector. I have seen it enabled unconditionally, or with a manual enable
(i.e., prctl).
It is possible to write a program that does not ever use AMX, and when that
program is run, the process will not incur the power or memory overhead of
AMX. It is also possible to do that with SVE. This is simply not possible
if RISC-V will, by default for every process, enable and allocate state
memory for vector.
So my thought would be what is the motivation for being even less flexible
than SVE, if you feel that the AMX mechanism is too onerous?
> >
> > > 2. People want the prctl gatekeeping for ability to gracefully handle memory
> > > allocation failure for the extra V-state within kernel. But that is only
> > > additional 4K (for typical 128 wide V regs) per task.
> > But vector state scales up to as much as 256k. Are you suggesting that
> > there is no possibility that future systems would support more than
> > VLEN=128?
>
> I mentioned "typical". And below also said that memory allocation concerns
> are moot, since fork/execve failures due to failing to allocate would take
> care of those anyways.
>
But again, what if one were using such an admittedly atypical system? Why
should such a user be compelled to take a memory hit for every process,
even if they specifically go out of their way to avoid using vector
instructions?
> > > If that is failing,
> > > the system is not doing well anyways. Besides it is not an issue at all
> > > since ENOMEM in clone/execve for the additional space should handle the
> > > failure anyways. Only very sophisticated apps would downgrade from executing
> > > V to Scalar code if the prctl failed.
> > This seems unlikely. As vector support does not exist in any present
> > hardware, and the vector extension is only optional in the RISC-V profiles
> > that include it, I would think that it is almost certain that any
> > application that supports V would have a fallback path for when the V
> > extension is not available.
>
> For specialized cases sure we would expect fat binaries with IFUNC based
> dispatches (glibc mem*/str* are obvious examples). But with newer compilers
> autovec is increasing becoming default even at medium optimization levels
> such as -O2. So V code littered all over is just a matter of time for the
> profiles/variants which allow V. For less capable systems w/o V this is all
> but moot discussion since kernel itself need not be built with V enabled.
>
To me, that seems like a lot of speculation, and certainly not worth
relying on to limit functionality.
>
> > Another motivation for requiring that user space request use of the vector
> > extension is that the vector unit may be shared between multiple harts
> > and/or have power or performance implications in the system. By requiring
> > that user space request access, it allows the system to decline that
> > access, and user space can handle this gracefully.
>
> But in this specific example won't the prctl cause more pain. So 2
> concurrent processes on 2 different harts with shared V unit. One sends
> prctl to enable and other wants to disable, what would the kernel do. Will
> it just be whoever ends up running later wins. Granted I'm not too familiar
> with how such a cross-hart sharing would work in a Vector instructions being
> part of ISA (vs. Vector accelerator with job push/pull approach)
>
I think you are misunderstanding both the mechanism and the potential
implementation.
The mechanism is not a switch per hart, it is a switch per process (or
thread). It indicates to the kernel that this process (or thread) is using
the resource, and the kernel will allocate memory, enable instructions,
etc., as needed when that processes is scheduled on a hart.
The theoretical implementation is something like what was done with
floating point in the AMD Bulldozer architecture. All the sharing of
resources is done automatically in hardware, but obviously performance
measurements could demonstrate that each thread/hart does not have its own
execution unit. I would imagine that in such an implementation, the vector
unit would not be able to, for example, enter a low power mode unless both
harts that depend on it are not using it. I would also imagine that single
thread performance for vector would be optimized if only one thread were
using the vector unit at a time. If the vector unit is enabled
unconditionally by default for all processes, this is not possible.
// darius
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 2:09 ` Darius Rad
0 siblings, 0 replies; 147+ messages in thread
From: Darius Rad @ 2022-12-15 2:09 UTC (permalink / raw)
To: Vineet Gupta
Cc: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv, bjorn,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Wed, Dec 14, 2022 at 12:07:03PM -0800, Vineet Gupta wrote:
> On 12/13/22 08:43, Darius Rad wrote:
> > On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
> > > But keeping the V unit disabled by default and using prctl as a gatekeeper
> > > to enable it feels unnecessary and tedious.
> > > Here's my reasoning below (I'm collating comments from prior msgs as well).
> > Please reference the previous discussion [1] which has covered topics that
> > have not been discussed recently.
> >
> > [1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
>
> I sure read thru that thread, and many more :-) to get context.
> The highlight is we should something because AVX/AMX do so (or failed to do
> so).
> But on the flip side ARM SVE is not disabling this by default.
> Your other concerns seems to be potential power implications for leaving it
> on and sharing of V unit across harts (see more on that below)
> Maybe leaving it on all the time will be motivation for hw designers to be
> more considerate of the idle power draw.
>
That is not quite the same take away I had from the AMX discussion. I
would also disagree that ARM SVE is not disabling things by default,
although the behavior is somewhat different.
The significant point that I see from that discussion, which also applies
to SVE, and also applies to RISC-V vector, is that the extension is
necessarily dependant upon a functional unit that is reasonably large with
respect to the size of the processor and has a significant amount of
additional architectural state. The argument there is that AMX/SVE/RVV is
a significant system resource and should be treated as such: by having the
kernel track usage of it and by having a process specifically request
access to it.
For any of AMX/SVE/RVV, use of the extension operates as follows:
1. A process requests access to the extension,
2. The kernel allocates memory for the additional state for this process,
3. The kernel enables the extension for the process, and finally
4. The process is able to use the instructions.
I don't recall the exact details, but my understanding is that AMX is going
to use an x86 specific mechanism and require and explict request from user
space to request access to AMX.
For SVE, it is in fact disabled by default in the kernel. When a thread
executes the first SVE instruction, it will cause an exception, the kernel
will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
instructions will proceed without exceptions. Although SVE is disabled by
default, it is enabled automatically. Since this is done automatically
during an exception handler, there is no opportunity for memory allocation
errors to be reported, as there are in the AMX case.
For RVV, I do not recall ever seeing Linux patches that automatically enable
vector. I have seen it enabled unconditionally, or with a manual enable
(i.e., prctl).
It is possible to write a program that does not ever use AMX, and when that
program is run, the process will not incur the power or memory overhead of
AMX. It is also possible to do that with SVE. This is simply not possible
if RISC-V will, by default for every process, enable and allocate state
memory for vector.
So my thought would be what is the motivation for being even less flexible
than SVE, if you feel that the AMX mechanism is too onerous?
> >
> > > 2. People want the prctl gatekeeping for ability to gracefully handle memory
> > > allocation failure for the extra V-state within kernel. But that is only
> > > additional 4K (for typical 128 wide V regs) per task.
> > But vector state scales up to as much as 256k. Are you suggesting that
> > there is no possibility that future systems would support more than
> > VLEN=128?
>
> I mentioned "typical". And below also said that memory allocation concerns
> are moot, since fork/execve failures due to failing to allocate would take
> care of those anyways.
>
But again, what if one were using such an admittedly atypical system? Why
should such a user be compelled to take a memory hit for every process,
even if they specifically go out of their way to avoid using vector
instructions?
> > > If that is failing,
> > > the system is not doing well anyways. Besides it is not an issue at all
> > > since ENOMEM in clone/execve for the additional space should handle the
> > > failure anyways. Only very sophisticated apps would downgrade from executing
> > > V to Scalar code if the prctl failed.
> > This seems unlikely. As vector support does not exist in any present
> > hardware, and the vector extension is only optional in the RISC-V profiles
> > that include it, I would think that it is almost certain that any
> > application that supports V would have a fallback path for when the V
> > extension is not available.
>
> For specialized cases sure we would expect fat binaries with IFUNC based
> dispatches (glibc mem*/str* are obvious examples). But with newer compilers
> autovec is increasing becoming default even at medium optimization levels
> such as -O2. So V code littered all over is just a matter of time for the
> profiles/variants which allow V. For less capable systems w/o V this is all
> but moot discussion since kernel itself need not be built with V enabled.
>
To me, that seems like a lot of speculation, and certainly not worth
relying on to limit functionality.
>
> > Another motivation for requiring that user space request use of the vector
> > extension is that the vector unit may be shared between multiple harts
> > and/or have power or performance implications in the system. By requiring
> > that user space request access, it allows the system to decline that
> > access, and user space can handle this gracefully.
>
> But in this specific example won't the prctl cause more pain. So 2
> concurrent processes on 2 different harts with shared V unit. One sends
> prctl to enable and other wants to disable, what would the kernel do. Will
> it just be whoever ends up running later wins. Granted I'm not too familiar
> with how such a cross-hart sharing would work in a Vector instructions being
> part of ISA (vs. Vector accelerator with job push/pull approach)
>
I think you are misunderstanding both the mechanism and the potential
implementation.
The mechanism is not a switch per hart, it is a switch per process (or
thread). It indicates to the kernel that this process (or thread) is using
the resource, and the kernel will allocate memory, enable instructions,
etc., as needed when that processes is scheduled on a hart.
The theoretical implementation is something like what was done with
floating point in the AMD Bulldozer architecture. All the sharing of
resources is done automatically in hardware, but obviously performance
measurements could demonstrate that each thread/hart does not have its own
execution unit. I would imagine that in such an implementation, the vector
unit would not be able to, for example, enter a low power mode unless both
harts that depend on it are not using it. I would also imagine that single
thread performance for vector would be optimized if only one thread were
using the vector unit at a time. If the vector unit is enabled
unconditionally by default for all processes, this is not possible.
// darius
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 2:09 ` Darius Rad
@ 2022-12-15 11:48 ` Björn Töpel
-1 siblings, 0 replies; 147+ messages in thread
From: Björn Töpel @ 2022-12-15 11:48 UTC (permalink / raw)
To: Darius Rad, Vineet Gupta
Cc: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
Darius Rad <darius@bluespec.com> writes:
> On Wed, Dec 14, 2022 at 12:07:03PM -0800, Vineet Gupta wrote:
>> On 12/13/22 08:43, Darius Rad wrote:
>> > On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>> > > But keeping the V unit disabled by default and using prctl as a gatekeeper
>> > > to enable it feels unnecessary and tedious.
>> > > Here's my reasoning below (I'm collating comments from prior msgs as well).
>> > Please reference the previous discussion [1] which has covered topics that
>> > have not been discussed recently.
>> >
>> > [1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
>>
>> I sure read thru that thread, and many more :-) to get context.
>> The highlight is we should something because AVX/AMX do so (or failed to do
>> so).
>> But on the flip side ARM SVE is not disabling this by default.
>> Your other concerns seems to be potential power implications for leaving it
>> on and sharing of V unit across harts (see more on that below)
>> Maybe leaving it on all the time will be motivation for hw designers to be
>> more considerate of the idle power draw.
>>
>
> That is not quite the same take away I had from the AMX discussion. I
> would also disagree that ARM SVE is not disabling things by default,
> although the behavior is somewhat different.
>
> The significant point that I see from that discussion, which also applies
> to SVE, and also applies to RISC-V vector, is that the extension is
> necessarily dependant upon a functional unit that is reasonably large with
> respect to the size of the processor and has a significant amount of
> additional architectural state. The argument there is that AMX/SVE/RVV is
> a significant system resource and should be treated as such: by having the
> kernel track usage of it and by having a process specifically request
> access to it.
>
> For any of AMX/SVE/RVV, use of the extension operates as follows:
>
> 1. A process requests access to the extension,
>
> 2. The kernel allocates memory for the additional state for this process,
>
> 3. The kernel enables the extension for the process, and finally
>
> 4. The process is able to use the instructions.
>
> I don't recall the exact details, but my understanding is that AMX is going
> to use an x86 specific mechanism and require and explict request from user
> space to request access to AMX.
Yes, it uses arch_prctl, and on top of that a "lazy trigger" (AFAIK) as
SVE (first use enable via trap).
> For SVE, it is in fact disabled by default in the kernel. When a thread
> executes the first SVE instruction, it will cause an exception, the kernel
> will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
> instructions will proceed without exceptions. Although SVE is disabled by
> default, it is enabled automatically. Since this is done automatically
> during an exception handler, there is no opportunity for memory allocation
> errors to be reported, as there are in the AMX case.
Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
much all processes on an SVE capable system will enable SVE (lazily)? If
so, that's close to "enabled by default" (unless SVE is disabled system
wide).
> For RVV, I do not recall ever seeing Linux patches that automatically enable
> vector. I have seen it enabled unconditionally, or with a manual enable
> (i.e., prctl).
>
> It is possible to write a program that does not ever use AMX, and when that
> program is run, the process will not incur the power or memory overhead of
> AMX. It is also possible to do that with SVE. This is simply not possible
> if RISC-V will, by default for every process, enable and allocate state
> memory for vector.
>
> So my thought would be what is the motivation for being even less flexible
> than SVE, if you feel that the AMX mechanism is too onerous?
AMX is a bit different from SVE and V; SVE/V is/would be used by glibc
for memcpy and such, where I doubt that AMX would be used there. Then
again, there's AVX512 which many argue that "turned on by default" was a
mistake (ABI breakage/power consumption).
>> >
>> > > 2. People want the prctl gatekeeping for ability to gracefully handle memory
>> > > allocation failure for the extra V-state within kernel. But that is only
>> > > additional 4K (for typical 128 wide V regs) per task.
>> > But vector state scales up to as much as 256k. Are you suggesting that
>> > there is no possibility that future systems would support more than
>> > VLEN=128?
>>
>> I mentioned "typical". And below also said that memory allocation concerns
>> are moot, since fork/execve failures due to failing to allocate would take
>> care of those anyways.
>>
>
> But again, what if one were using such an admittedly atypical system? Why
> should such a user be compelled to take a memory hit for every process,
> even if they specifically go out of their way to avoid using vector
> instructions?
For the sake of discussion; Nobody is arguing against not having knobs
to turn V on/off per-process/per-system, right? The discussion is about
on/off, and broader what a "typical" RV system looks like. If most
systems that fold in the A profile has V, it might make sense not
requiring users to explicitly enable it, and vice-versa.
Using RVA23 as a ball-gazing aid, [1] states that it might mandate V. If
so, assuming that "most system will be designed for V usage" is not
crazy.
Now moving on! The thread is leaning towards "disabled by default" ("AMX
way"), let's try to move the discussion forward!
The Linux kernel itself would benefit from using V
(hashing/crypto). What kind of policy would determine if the kernel is
allowed to use V? Default off, with an explicit enable kernel knob
(cmdline/sysctl/sysfs/...)?
There will likely be V support in glibc (str*/mem*). For systems that
prefer having V "always-on", the UX of requiring all binaries to
explicitly call prctl() is not great (as Andrew pointed out in earlier
posts). A V knob based on some system policy in crt0? :-P
Björn
[1] https://lists.riscv.org/g/tech-profiles/message/48
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 11:48 ` Björn Töpel
0 siblings, 0 replies; 147+ messages in thread
From: Björn Töpel @ 2022-12-15 11:48 UTC (permalink / raw)
To: Darius Rad, Vineet Gupta
Cc: Palmer Dabbelt, fweimer, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
Darius Rad <darius@bluespec.com> writes:
> On Wed, Dec 14, 2022 at 12:07:03PM -0800, Vineet Gupta wrote:
>> On 12/13/22 08:43, Darius Rad wrote:
>> > On Fri, Dec 09, 2022 at 11:42:19AM -0800, Vineet Gupta wrote:
>> > > But keeping the V unit disabled by default and using prctl as a gatekeeper
>> > > to enable it feels unnecessary and tedious.
>> > > Here's my reasoning below (I'm collating comments from prior msgs as well).
>> > Please reference the previous discussion [1] which has covered topics that
>> > have not been discussed recently.
>> >
>> > [1] https://lists.infradead.org/pipermail/linux-riscv/2021-September/thread.html#8361
>>
>> I sure read thru that thread, and many more :-) to get context.
>> The highlight is we should something because AVX/AMX do so (or failed to do
>> so).
>> But on the flip side ARM SVE is not disabling this by default.
>> Your other concerns seems to be potential power implications for leaving it
>> on and sharing of V unit across harts (see more on that below)
>> Maybe leaving it on all the time will be motivation for hw designers to be
>> more considerate of the idle power draw.
>>
>
> That is not quite the same take away I had from the AMX discussion. I
> would also disagree that ARM SVE is not disabling things by default,
> although the behavior is somewhat different.
>
> The significant point that I see from that discussion, which also applies
> to SVE, and also applies to RISC-V vector, is that the extension is
> necessarily dependant upon a functional unit that is reasonably large with
> respect to the size of the processor and has a significant amount of
> additional architectural state. The argument there is that AMX/SVE/RVV is
> a significant system resource and should be treated as such: by having the
> kernel track usage of it and by having a process specifically request
> access to it.
>
> For any of AMX/SVE/RVV, use of the extension operates as follows:
>
> 1. A process requests access to the extension,
>
> 2. The kernel allocates memory for the additional state for this process,
>
> 3. The kernel enables the extension for the process, and finally
>
> 4. The process is able to use the instructions.
>
> I don't recall the exact details, but my understanding is that AMX is going
> to use an x86 specific mechanism and require and explict request from user
> space to request access to AMX.
Yes, it uses arch_prctl, and on top of that a "lazy trigger" (AFAIK) as
SVE (first use enable via trap).
> For SVE, it is in fact disabled by default in the kernel. When a thread
> executes the first SVE instruction, it will cause an exception, the kernel
> will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
> instructions will proceed without exceptions. Although SVE is disabled by
> default, it is enabled automatically. Since this is done automatically
> during an exception handler, there is no opportunity for memory allocation
> errors to be reported, as there are in the AMX case.
Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
much all processes on an SVE capable system will enable SVE (lazily)? If
so, that's close to "enabled by default" (unless SVE is disabled system
wide).
> For RVV, I do not recall ever seeing Linux patches that automatically enable
> vector. I have seen it enabled unconditionally, or with a manual enable
> (i.e., prctl).
>
> It is possible to write a program that does not ever use AMX, and when that
> program is run, the process will not incur the power or memory overhead of
> AMX. It is also possible to do that with SVE. This is simply not possible
> if RISC-V will, by default for every process, enable and allocate state
> memory for vector.
>
> So my thought would be what is the motivation for being even less flexible
> than SVE, if you feel that the AMX mechanism is too onerous?
AMX is a bit different from SVE and V; SVE/V is/would be used by glibc
for memcpy and such, where I doubt that AMX would be used there. Then
again, there's AVX512 which many argue that "turned on by default" was a
mistake (ABI breakage/power consumption).
>> >
>> > > 2. People want the prctl gatekeeping for ability to gracefully handle memory
>> > > allocation failure for the extra V-state within kernel. But that is only
>> > > additional 4K (for typical 128 wide V regs) per task.
>> > But vector state scales up to as much as 256k. Are you suggesting that
>> > there is no possibility that future systems would support more than
>> > VLEN=128?
>>
>> I mentioned "typical". And below also said that memory allocation concerns
>> are moot, since fork/execve failures due to failing to allocate would take
>> care of those anyways.
>>
>
> But again, what if one were using such an admittedly atypical system? Why
> should such a user be compelled to take a memory hit for every process,
> even if they specifically go out of their way to avoid using vector
> instructions?
For the sake of discussion; Nobody is arguing against not having knobs
to turn V on/off per-process/per-system, right? The discussion is about
on/off, and broader what a "typical" RV system looks like. If most
systems that fold in the A profile has V, it might make sense not
requiring users to explicitly enable it, and vice-versa.
Using RVA23 as a ball-gazing aid, [1] states that it might mandate V. If
so, assuming that "most system will be designed for V usage" is not
crazy.
Now moving on! The thread is leaning towards "disabled by default" ("AMX
way"), let's try to move the discussion forward!
The Linux kernel itself would benefit from using V
(hashing/crypto). What kind of policy would determine if the kernel is
allowed to use V? Default off, with an explicit enable kernel knob
(cmdline/sysctl/sysfs/...)?
There will likely be V support in glibc (str*/mem*). For systems that
prefer having V "always-on", the UX of requiring all binaries to
explicitly call prctl() is not great (as Andrew pointed out in earlier
posts). A V knob based on some system policy in crt0? :-P
Björn
[1] https://lists.riscv.org/g/tech-profiles/message/48
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 11:48 ` Björn Töpel
@ 2022-12-15 12:28 ` Florian Weimer
-1 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-15 12:28 UTC (permalink / raw)
To: Björn Töpel
Cc: Darius Rad, Vineet Gupta, Palmer Dabbelt, Andrew Waterman,
stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, arnd, linux-kernel,
linux-riscv, libc-alpha, christoph.muellner, Aaron Durbin, linux
* Björn Töpel:
>> For SVE, it is in fact disabled by default in the kernel. When a thread
>> executes the first SVE instruction, it will cause an exception, the kernel
>> will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
>> instructions will proceed without exceptions. Although SVE is disabled by
>> default, it is enabled automatically. Since this is done automatically
>> during an exception handler, there is no opportunity for memory allocation
>> errors to be reported, as there are in the AMX case.
>
> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
> much all processes on an SVE capable system will enable SVE (lazily)? If
> so, that's close to "enabled by default" (unless SVE is disabled system
> wide).
Yes, see sysdeps/aarch64/multiarch/memcpy.c:
static inline __typeof (__redirect_memcpy) *
select_memcpy_ifunc (void)
{
INIT_ARCH ();
if (sve && HAVE_AARCH64_SVE_ASM)
{
if (IS_A64FX (midr))
return __memcpy_a64fx;
return __memcpy_sve;
}
if (IS_THUNDERX (midr))
return __memcpy_thunderx;
if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
return __memcpy_thunderx2;
if (IS_FALKOR (midr) || IS_PHECDA (midr))
return __memcpy_falkor;
return __memcpy_generic;
}
And the __memcpy_sve implementation actually uses SVE.
If there were a prctl to select the vector width and enable the vector
extension, we'd have to pick a width in glibc anyway. Likewise for any
other libc, the Go runtime, and so on. That's why I think the kernel is
in a better position to handle this.
> AMX is a bit different from SVE and V; SVE/V is/would be used by glibc
> for memcpy and such, where I doubt that AMX would be used there. Then
> again, there's AVX512 which many argue that "turned on by default" was a
> mistake (ABI breakage/power consumption).
I don't think AMX is useful for string operations or the math functions
currently implemented in glibc.
Not everything in AVX-512 has high power consumption on relevant CPUs.
Furthermore, the extra registers that don't need VZEROUPPER help us to
avoid transactions aborts in RTM mode. If we had to enable AVX-512
explicitly in every process, I'm not sure if we would be using it today.
The complicated choices around AVX-512 (and AVX2 for earlier CPUs)
aren't particularly unique. These functions have different trade-offs
(optimizing for single thread/single process usage vs global system
behavior) on other architectures, too.
> There will likely be V support in glibc (str*/mem*). For systems that
> prefer having V "always-on", the UX of requiring all binaries to
> explicitly call prctl() is not great (as Andrew pointed out in earlier
> posts). A V knob based on some system policy in crt0? :-P
It wouldn't be in crt0 (statically linked), it would be in the dynamic
loader. So not quite as bad if policy revisions are required. But
glibc is not the only provider of userspace startup code, so future
tuning of userspace policy will remain complicated.
Thanks,
Florian
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 12:28 ` Florian Weimer
0 siblings, 0 replies; 147+ messages in thread
From: Florian Weimer @ 2022-12-15 12:28 UTC (permalink / raw)
To: Björn Töpel
Cc: Darius Rad, Vineet Gupta, Palmer Dabbelt, Andrew Waterman,
stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, arnd, linux-kernel,
linux-riscv, libc-alpha, christoph.muellner, Aaron Durbin, linux
* Björn Töpel:
>> For SVE, it is in fact disabled by default in the kernel. When a thread
>> executes the first SVE instruction, it will cause an exception, the kernel
>> will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
>> instructions will proceed without exceptions. Although SVE is disabled by
>> default, it is enabled automatically. Since this is done automatically
>> during an exception handler, there is no opportunity for memory allocation
>> errors to be reported, as there are in the AMX case.
>
> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
> much all processes on an SVE capable system will enable SVE (lazily)? If
> so, that's close to "enabled by default" (unless SVE is disabled system
> wide).
Yes, see sysdeps/aarch64/multiarch/memcpy.c:
static inline __typeof (__redirect_memcpy) *
select_memcpy_ifunc (void)
{
INIT_ARCH ();
if (sve && HAVE_AARCH64_SVE_ASM)
{
if (IS_A64FX (midr))
return __memcpy_a64fx;
return __memcpy_sve;
}
if (IS_THUNDERX (midr))
return __memcpy_thunderx;
if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
return __memcpy_thunderx2;
if (IS_FALKOR (midr) || IS_PHECDA (midr))
return __memcpy_falkor;
return __memcpy_generic;
}
And the __memcpy_sve implementation actually uses SVE.
If there were a prctl to select the vector width and enable the vector
extension, we'd have to pick a width in glibc anyway. Likewise for any
other libc, the Go runtime, and so on. That's why I think the kernel is
in a better position to handle this.
> AMX is a bit different from SVE and V; SVE/V is/would be used by glibc
> for memcpy and such, where I doubt that AMX would be used there. Then
> again, there's AVX512 which many argue that "turned on by default" was a
> mistake (ABI breakage/power consumption).
I don't think AMX is useful for string operations or the math functions
currently implemented in glibc.
Not everything in AVX-512 has high power consumption on relevant CPUs.
Furthermore, the extra registers that don't need VZEROUPPER help us to
avoid transactions aborts in RTM mode. If we had to enable AVX-512
explicitly in every process, I'm not sure if we would be using it today.
The complicated choices around AVX-512 (and AVX2 for earlier CPUs)
aren't particularly unique. These functions have different trade-offs
(optimizing for single thread/single process usage vs global system
behavior) on other architectures, too.
> There will likely be V support in glibc (str*/mem*). For systems that
> prefer having V "always-on", the UX of requiring all binaries to
> explicitly call prctl() is not great (as Andrew pointed out in earlier
> posts). A V knob based on some system policy in crt0? :-P
It wouldn't be in crt0 (statically linked), it would be in the dynamic
loader. So not quite as bad if policy revisions are required. But
glibc is not the only provider of userspace startup code, so future
tuning of userspace policy will remain complicated.
Thanks,
Florian
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 12:28 ` Florian Weimer
@ 2022-12-15 15:33 ` Richard Henderson
-1 siblings, 0 replies; 147+ messages in thread
From: Richard Henderson @ 2022-12-15 15:33 UTC (permalink / raw)
To: Florian Weimer, Björn Töpel
Cc: Darius Rad, Vineet Gupta, Palmer Dabbelt, Andrew Waterman,
stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, arnd, linux-kernel,
linux-riscv, libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
> * Björn Töpel:
>
>>> For SVE, it is in fact disabled by default in the kernel. When a thread
>>> executes the first SVE instruction, it will cause an exception, the kernel
>>> will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
>>> instructions will proceed without exceptions. Although SVE is disabled by
>>> default, it is enabled automatically. Since this is done automatically
>>> during an exception handler, there is no opportunity for memory allocation
>>> errors to be reported, as there are in the AMX case.
>>
>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
>> much all processes on an SVE capable system will enable SVE (lazily)? If
>> so, that's close to "enabled by default" (unless SVE is disabled system
>> wide).
>
> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
>
> static inline __typeof (__redirect_memcpy) *
> select_memcpy_ifunc (void)
> {
> INIT_ARCH ();
>
> if (sve && HAVE_AARCH64_SVE_ASM)
> {
> if (IS_A64FX (midr))
> return __memcpy_a64fx;
> return __memcpy_sve;
> }
>
> if (IS_THUNDERX (midr))
> return __memcpy_thunderx;
>
> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
> return __memcpy_thunderx2;
>
> if (IS_FALKOR (midr) || IS_PHECDA (midr))
> return __memcpy_falkor;
>
> return __memcpy_generic;
> }
>
> And the __memcpy_sve implementation actually uses SVE.
>
> If there were a prctl to select the vector width and enable the vector
> extension, we'd have to pick a width in glibc anyway.
There *is* a prctl to adjust the SVE vector width, but glibc does not need to select
because SVE dynamically adjusts to the currently enabled width. The kernel selects a
default width that fits within the default signal frame size.
The other thing of note for SVE is that, with the default function ABI all of the SVE
state is call-clobbered, which allows the kernel to drop instead of save state across
system calls. (There is a separate vector function call ABI when SVE types are used.)
So while strcpy may enable SVE for the thread, the next syscall may disable it again.
r~
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 15:33 ` Richard Henderson
0 siblings, 0 replies; 147+ messages in thread
From: Richard Henderson @ 2022-12-15 15:33 UTC (permalink / raw)
To: Florian Weimer, Björn Töpel
Cc: Darius Rad, Vineet Gupta, Palmer Dabbelt, Andrew Waterman,
stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, arnd, linux-kernel,
linux-riscv, libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
> * Björn Töpel:
>
>>> For SVE, it is in fact disabled by default in the kernel. When a thread
>>> executes the first SVE instruction, it will cause an exception, the kernel
>>> will allocate memory for SVE state and enable TIF_SVE. Further use of SVE
>>> instructions will proceed without exceptions. Although SVE is disabled by
>>> default, it is enabled automatically. Since this is done automatically
>>> during an exception handler, there is no opportunity for memory allocation
>>> errors to be reported, as there are in the AMX case.
>>
>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
>> much all processes on an SVE capable system will enable SVE (lazily)? If
>> so, that's close to "enabled by default" (unless SVE is disabled system
>> wide).
>
> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
>
> static inline __typeof (__redirect_memcpy) *
> select_memcpy_ifunc (void)
> {
> INIT_ARCH ();
>
> if (sve && HAVE_AARCH64_SVE_ASM)
> {
> if (IS_A64FX (midr))
> return __memcpy_a64fx;
> return __memcpy_sve;
> }
>
> if (IS_THUNDERX (midr))
> return __memcpy_thunderx;
>
> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
> return __memcpy_thunderx2;
>
> if (IS_FALKOR (midr) || IS_PHECDA (midr))
> return __memcpy_falkor;
>
> return __memcpy_generic;
> }
>
> And the __memcpy_sve implementation actually uses SVE.
>
> If there were a prctl to select the vector width and enable the vector
> extension, we'd have to pick a width in glibc anyway.
There *is* a prctl to adjust the SVE vector width, but glibc does not need to select
because SVE dynamically adjusts to the currently enabled width. The kernel selects a
default width that fits within the default signal frame size.
The other thing of note for SVE is that, with the default function ABI all of the SVE
state is call-clobbered, which allows the kernel to drop instead of save state across
system calls. (There is a separate vector function call ABI when SVE types are used.)
So while strcpy may enable SVE for the thread, the next syscall may disable it again.
r~
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 15:33 ` Richard Henderson
@ 2022-12-15 18:57 ` Vineet Gupta
-1 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-15 18:57 UTC (permalink / raw)
To: Richard Henderson, Florian Weimer, Björn Töpel
Cc: Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/15/22 07:33, Richard Henderson wrote:
> On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
>> * Björn Töpel:
>>
>>>> For SVE, it is in fact disabled by default in the kernel. When a
>>>> thread
>>>> executes the first SVE instruction, it will cause an exception, the
>>>> kernel
>>>> will allocate memory for SVE state and enable TIF_SVE. Further use
>>>> of SVE
>>>> instructions will proceed without exceptions. Although SVE is
>>>> disabled by
>>>> default, it is enabled automatically. Since this is done
>>>> automatically
>>>> during an exception handler, there is no opportunity for memory
>>>> allocation
>>>> errors to be reported, as there are in the AMX case.
>>>
>>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
>>> much all processes on an SVE capable system will enable SVE
>>> (lazily)? If
>>> so, that's close to "enabled by default" (unless SVE is disabled system
>>> wide).
>>
>> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
>>
>> static inline __typeof (__redirect_memcpy) *
>> select_memcpy_ifunc (void)
>> {
>> INIT_ARCH ();
>> if (sve && HAVE_AARCH64_SVE_ASM)
>> {
>> if (IS_A64FX (midr))
>> return __memcpy_a64fx;
>> return __memcpy_sve;
>> }
>> if (IS_THUNDERX (midr))
>> return __memcpy_thunderx;
>> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
>> return __memcpy_thunderx2;
>> if (IS_FALKOR (midr) || IS_PHECDA (midr))
>> return __memcpy_falkor;
>> return __memcpy_generic;
>> }
>> And the __memcpy_sve implementation actually uses SVE.
>>
>> If there were a prctl to select the vector width and enable the vector
>> extension, we'd have to pick a width in glibc anyway.
>
> There *is* a prctl to adjust the SVE vector width, but glibc does not
> need to select because SVE dynamically adjusts to the currently
> enabled width. The kernel selects a default width that fits within
> the default signal frame size.
>
> The other thing of note for SVE is that, with the default function ABI
> all of the SVE state is call-clobbered, which allows the kernel to
> drop instead of save state across system calls. (There is a separate
> vector function call ABI when SVE types are used.)
For the RV psABI, it is similar - all V regs are
caller-saved/call-clobbered [1] and syscalls are not required to
preserve V regs [2]
However last I checked ARM documentation the ABI doc seemed to suggest
that some (parts) of the SVE regs are callee-saved [3]
>
> So while strcpy may enable SVE for the thread, the next syscall may
> disable it again.
Next syscall could trash them, but will it disable SVE ? Despite
syscall/function-call clobbers, using V in tight loops such as mem*/str*
still is a win.
[1]
https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-cc.adoc
[2]
https://github.com/riscv/riscv-v-spec/blob/master/calling-convention.adoc
[3]
https://github.com/ARM-software/abi-aa/blob/2982a9f3b512a5bfdc9e3fea5d3b298f9165c36b/aapcs64/aapcs64.rst#the-base-procedure-call-standard
Sec 6.1.3 ".... In other cases it need only preserve the low 64 bits of
z8-z15"
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 18:57 ` Vineet Gupta
0 siblings, 0 replies; 147+ messages in thread
From: Vineet Gupta @ 2022-12-15 18:57 UTC (permalink / raw)
To: Richard Henderson, Florian Weimer, Björn Töpel
Cc: Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/15/22 07:33, Richard Henderson wrote:
> On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
>> * Björn Töpel:
>>
>>>> For SVE, it is in fact disabled by default in the kernel. When a
>>>> thread
>>>> executes the first SVE instruction, it will cause an exception, the
>>>> kernel
>>>> will allocate memory for SVE state and enable TIF_SVE. Further use
>>>> of SVE
>>>> instructions will proceed without exceptions. Although SVE is
>>>> disabled by
>>>> default, it is enabled automatically. Since this is done
>>>> automatically
>>>> during an exception handler, there is no opportunity for memory
>>>> allocation
>>>> errors to be reported, as there are in the AMX case.
>>>
>>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
>>> much all processes on an SVE capable system will enable SVE
>>> (lazily)? If
>>> so, that's close to "enabled by default" (unless SVE is disabled system
>>> wide).
>>
>> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
>>
>> static inline __typeof (__redirect_memcpy) *
>> select_memcpy_ifunc (void)
>> {
>> INIT_ARCH ();
>> if (sve && HAVE_AARCH64_SVE_ASM)
>> {
>> if (IS_A64FX (midr))
>> return __memcpy_a64fx;
>> return __memcpy_sve;
>> }
>> if (IS_THUNDERX (midr))
>> return __memcpy_thunderx;
>> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
>> return __memcpy_thunderx2;
>> if (IS_FALKOR (midr) || IS_PHECDA (midr))
>> return __memcpy_falkor;
>> return __memcpy_generic;
>> }
>> And the __memcpy_sve implementation actually uses SVE.
>>
>> If there were a prctl to select the vector width and enable the vector
>> extension, we'd have to pick a width in glibc anyway.
>
> There *is* a prctl to adjust the SVE vector width, but glibc does not
> need to select because SVE dynamically adjusts to the currently
> enabled width. The kernel selects a default width that fits within
> the default signal frame size.
>
> The other thing of note for SVE is that, with the default function ABI
> all of the SVE state is call-clobbered, which allows the kernel to
> drop instead of save state across system calls. (There is a separate
> vector function call ABI when SVE types are used.)
For the RV psABI, it is similar - all V regs are
caller-saved/call-clobbered [1] and syscalls are not required to
preserve V regs [2]
However last I checked ARM documentation the ABI doc seemed to suggest
that some (parts) of the SVE regs are callee-saved [3]
>
> So while strcpy may enable SVE for the thread, the next syscall may
> disable it again.
Next syscall could trash them, but will it disable SVE ? Despite
syscall/function-call clobbers, using V in tight loops such as mem*/str*
still is a win.
[1]
https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-cc.adoc
[2]
https://github.com/riscv/riscv-v-spec/blob/master/calling-convention.adoc
[3]
https://github.com/ARM-software/abi-aa/blob/2982a9f3b512a5bfdc9e3fea5d3b298f9165c36b/aapcs64/aapcs64.rst#the-base-procedure-call-standard
Sec 6.1.3 ".... In other cases it need only preserve the low 64 bits of
z8-z15"
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 18:57 ` Vineet Gupta
@ 2022-12-15 18:59 ` Andrew Pinski
-1 siblings, 0 replies; 147+ messages in thread
From: Andrew Pinski @ 2022-12-15 18:59 UTC (permalink / raw)
To: Vineet Gupta
Cc: Richard Henderson, Florian Weimer, Björn Töpel,
Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Thu, Dec 15, 2022 at 10:57 AM Vineet Gupta <vineetg@rivosinc.com> wrote:
>
>
>
> On 12/15/22 07:33, Richard Henderson wrote:
> > On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
> >> * Björn Töpel:
> >>
> >>>> For SVE, it is in fact disabled by default in the kernel. When a
> >>>> thread
> >>>> executes the first SVE instruction, it will cause an exception, the
> >>>> kernel
> >>>> will allocate memory for SVE state and enable TIF_SVE. Further use
> >>>> of SVE
> >>>> instructions will proceed without exceptions. Although SVE is
> >>>> disabled by
> >>>> default, it is enabled automatically. Since this is done
> >>>> automatically
> >>>> during an exception handler, there is no opportunity for memory
> >>>> allocation
> >>>> errors to be reported, as there are in the AMX case.
> >>>
> >>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
> >>> much all processes on an SVE capable system will enable SVE
> >>> (lazily)? If
> >>> so, that's close to "enabled by default" (unless SVE is disabled system
> >>> wide).
> >>
> >> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
> >>
> >> static inline __typeof (__redirect_memcpy) *
> >> select_memcpy_ifunc (void)
> >> {
> >> INIT_ARCH ();
> >> if (sve && HAVE_AARCH64_SVE_ASM)
> >> {
> >> if (IS_A64FX (midr))
> >> return __memcpy_a64fx;
> >> return __memcpy_sve;
> >> }
> >> if (IS_THUNDERX (midr))
> >> return __memcpy_thunderx;
> >> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
> >> return __memcpy_thunderx2;
> >> if (IS_FALKOR (midr) || IS_PHECDA (midr))
> >> return __memcpy_falkor;
> >> return __memcpy_generic;
> >> }
> >> And the __memcpy_sve implementation actually uses SVE.
> >>
> >> If there were a prctl to select the vector width and enable the vector
> >> extension, we'd have to pick a width in glibc anyway.
> >
> > There *is* a prctl to adjust the SVE vector width, but glibc does not
> > need to select because SVE dynamically adjusts to the currently
> > enabled width. The kernel selects a default width that fits within
> > the default signal frame size.
> >
> > The other thing of note for SVE is that, with the default function ABI
> > all of the SVE state is call-clobbered, which allows the kernel to
> > drop instead of save state across system calls. (There is a separate
> > vector function call ABI when SVE types are used.)
>
> For the RV psABI, it is similar - all V regs are
> caller-saved/call-clobbered [1] and syscalls are not required to
> preserve V regs [2]
> However last I checked ARM documentation the ABI doc seemed to suggest
> that some (parts) of the SVE regs are callee-saved [3]
Yes the lower 64 bits which overlap with the floating point registers.
Thanks,
Andrew Pinski
>
> >
> > So while strcpy may enable SVE for the thread, the next syscall may
> > disable it again.
>
> Next syscall could trash them, but will it disable SVE ? Despite
> syscall/function-call clobbers, using V in tight loops such as mem*/str*
> still is a win.
>
>
> [1]
> https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-cc.adoc
> [2]
> https://github.com/riscv/riscv-v-spec/blob/master/calling-convention.adoc
> [3]
> https://github.com/ARM-software/abi-aa/blob/2982a9f3b512a5bfdc9e3fea5d3b298f9165c36b/aapcs64/aapcs64.rst#the-base-procedure-call-standard
> Sec 6.1.3 ".... In other cases it need only preserve the low 64 bits of
> z8-z15"
>
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 18:59 ` Andrew Pinski
0 siblings, 0 replies; 147+ messages in thread
From: Andrew Pinski @ 2022-12-15 18:59 UTC (permalink / raw)
To: Vineet Gupta
Cc: Richard Henderson, Florian Weimer, Björn Töpel,
Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Thu, Dec 15, 2022 at 10:57 AM Vineet Gupta <vineetg@rivosinc.com> wrote:
>
>
>
> On 12/15/22 07:33, Richard Henderson wrote:
> > On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
> >> * Björn Töpel:
> >>
> >>>> For SVE, it is in fact disabled by default in the kernel. When a
> >>>> thread
> >>>> executes the first SVE instruction, it will cause an exception, the
> >>>> kernel
> >>>> will allocate memory for SVE state and enable TIF_SVE. Further use
> >>>> of SVE
> >>>> instructions will proceed without exceptions. Although SVE is
> >>>> disabled by
> >>>> default, it is enabled automatically. Since this is done
> >>>> automatically
> >>>> during an exception handler, there is no opportunity for memory
> >>>> allocation
> >>>> errors to be reported, as there are in the AMX case.
> >>>
> >>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
> >>> much all processes on an SVE capable system will enable SVE
> >>> (lazily)? If
> >>> so, that's close to "enabled by default" (unless SVE is disabled system
> >>> wide).
> >>
> >> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
> >>
> >> static inline __typeof (__redirect_memcpy) *
> >> select_memcpy_ifunc (void)
> >> {
> >> INIT_ARCH ();
> >> if (sve && HAVE_AARCH64_SVE_ASM)
> >> {
> >> if (IS_A64FX (midr))
> >> return __memcpy_a64fx;
> >> return __memcpy_sve;
> >> }
> >> if (IS_THUNDERX (midr))
> >> return __memcpy_thunderx;
> >> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
> >> return __memcpy_thunderx2;
> >> if (IS_FALKOR (midr) || IS_PHECDA (midr))
> >> return __memcpy_falkor;
> >> return __memcpy_generic;
> >> }
> >> And the __memcpy_sve implementation actually uses SVE.
> >>
> >> If there were a prctl to select the vector width and enable the vector
> >> extension, we'd have to pick a width in glibc anyway.
> >
> > There *is* a prctl to adjust the SVE vector width, but glibc does not
> > need to select because SVE dynamically adjusts to the currently
> > enabled width. The kernel selects a default width that fits within
> > the default signal frame size.
> >
> > The other thing of note for SVE is that, with the default function ABI
> > all of the SVE state is call-clobbered, which allows the kernel to
> > drop instead of save state across system calls. (There is a separate
> > vector function call ABI when SVE types are used.)
>
> For the RV psABI, it is similar - all V regs are
> caller-saved/call-clobbered [1] and syscalls are not required to
> preserve V regs [2]
> However last I checked ARM documentation the ABI doc seemed to suggest
> that some (parts) of the SVE regs are callee-saved [3]
Yes the lower 64 bits which overlap with the floating point registers.
Thanks,
Andrew Pinski
>
> >
> > So while strcpy may enable SVE for the thread, the next syscall may
> > disable it again.
>
> Next syscall could trash them, but will it disable SVE ? Despite
> syscall/function-call clobbers, using V in tight loops such as mem*/str*
> still is a win.
>
>
> [1]
> https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-cc.adoc
> [2]
> https://github.com/riscv/riscv-v-spec/blob/master/calling-convention.adoc
> [3]
> https://github.com/ARM-software/abi-aa/blob/2982a9f3b512a5bfdc9e3fea5d3b298f9165c36b/aapcs64/aapcs64.rst#the-base-procedure-call-standard
> Sec 6.1.3 ".... In other cases it need only preserve the low 64 bits of
> z8-z15"
>
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 18:59 ` Andrew Pinski
@ 2022-12-15 19:01 ` Andrew Pinski
-1 siblings, 0 replies; 147+ messages in thread
From: Andrew Pinski @ 2022-12-15 19:01 UTC (permalink / raw)
To: Vineet Gupta
Cc: Richard Henderson, Florian Weimer, Björn Töpel,
Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Thu, Dec 15, 2022 at 10:59 AM Andrew Pinski <pinskia@gmail.com> wrote:
>
> On Thu, Dec 15, 2022 at 10:57 AM Vineet Gupta <vineetg@rivosinc.com> wrote:
> >
> >
> >
> > On 12/15/22 07:33, Richard Henderson wrote:
> > > On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
> > >> * Björn Töpel:
> > >>
> > >>>> For SVE, it is in fact disabled by default in the kernel. When a
> > >>>> thread
> > >>>> executes the first SVE instruction, it will cause an exception, the
> > >>>> kernel
> > >>>> will allocate memory for SVE state and enable TIF_SVE. Further use
> > >>>> of SVE
> > >>>> instructions will proceed without exceptions. Although SVE is
> > >>>> disabled by
> > >>>> default, it is enabled automatically. Since this is done
> > >>>> automatically
> > >>>> during an exception handler, there is no opportunity for memory
> > >>>> allocation
> > >>>> errors to be reported, as there are in the AMX case.
> > >>>
> > >>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
> > >>> much all processes on an SVE capable system will enable SVE
> > >>> (lazily)? If
> > >>> so, that's close to "enabled by default" (unless SVE is disabled system
> > >>> wide).
> > >>
> > >> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
> > >>
> > >> static inline __typeof (__redirect_memcpy) *
> > >> select_memcpy_ifunc (void)
> > >> {
> > >> INIT_ARCH ();
> > >> if (sve && HAVE_AARCH64_SVE_ASM)
> > >> {
> > >> if (IS_A64FX (midr))
> > >> return __memcpy_a64fx;
> > >> return __memcpy_sve;
> > >> }
> > >> if (IS_THUNDERX (midr))
> > >> return __memcpy_thunderx;
> > >> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
> > >> return __memcpy_thunderx2;
> > >> if (IS_FALKOR (midr) || IS_PHECDA (midr))
> > >> return __memcpy_falkor;
> > >> return __memcpy_generic;
> > >> }
> > >> And the __memcpy_sve implementation actually uses SVE.
> > >>
> > >> If there were a prctl to select the vector width and enable the vector
> > >> extension, we'd have to pick a width in glibc anyway.
> > >
> > > There *is* a prctl to adjust the SVE vector width, but glibc does not
> > > need to select because SVE dynamically adjusts to the currently
> > > enabled width. The kernel selects a default width that fits within
> > > the default signal frame size.
> > >
> > > The other thing of note for SVE is that, with the default function ABI
> > > all of the SVE state is call-clobbered, which allows the kernel to
> > > drop instead of save state across system calls. (There is a separate
> > > vector function call ABI when SVE types are used.)
> >
> > For the RV psABI, it is similar - all V regs are
> > caller-saved/call-clobbered [1] and syscalls are not required to
> > preserve V regs [2]
> > However last I checked ARM documentation the ABI doc seemed to suggest
> > that some (parts) of the SVE regs are callee-saved [3]
>
> Yes the lower 64 bits which overlap with the floating point registers.
I should expand on that, only the specific callee registers have to
save the lower 64bits because they overlap with the floating point
registers.
Thanks,
Andrew
>
> Thanks,
> Andrew Pinski
>
>
> >
> > >
> > > So while strcpy may enable SVE for the thread, the next syscall may
> > > disable it again.
> >
> > Next syscall could trash them, but will it disable SVE ? Despite
> > syscall/function-call clobbers, using V in tight loops such as mem*/str*
> > still is a win.
> >
> >
> > [1]
> > https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-cc.adoc
> > [2]
> > https://github.com/riscv/riscv-v-spec/blob/master/calling-convention.adoc
> > [3]
> > https://github.com/ARM-software/abi-aa/blob/2982a9f3b512a5bfdc9e3fea5d3b298f9165c36b/aapcs64/aapcs64.rst#the-base-procedure-call-standard
> > Sec 6.1.3 ".... In other cases it need only preserve the low 64 bits of
> > z8-z15"
> >
^ permalink raw reply [flat|nested] 147+ messages in thread* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 19:01 ` Andrew Pinski
0 siblings, 0 replies; 147+ messages in thread
From: Andrew Pinski @ 2022-12-15 19:01 UTC (permalink / raw)
To: Vineet Gupta
Cc: Richard Henderson, Florian Weimer, Björn Töpel,
Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On Thu, Dec 15, 2022 at 10:59 AM Andrew Pinski <pinskia@gmail.com> wrote:
>
> On Thu, Dec 15, 2022 at 10:57 AM Vineet Gupta <vineetg@rivosinc.com> wrote:
> >
> >
> >
> > On 12/15/22 07:33, Richard Henderson wrote:
> > > On 12/15/22 04:28, Florian Weimer via Libc-alpha wrote:
> > >> * Björn Töpel:
> > >>
> > >>>> For SVE, it is in fact disabled by default in the kernel. When a
> > >>>> thread
> > >>>> executes the first SVE instruction, it will cause an exception, the
> > >>>> kernel
> > >>>> will allocate memory for SVE state and enable TIF_SVE. Further use
> > >>>> of SVE
> > >>>> instructions will proceed without exceptions. Although SVE is
> > >>>> disabled by
> > >>>> default, it is enabled automatically. Since this is done
> > >>>> automatically
> > >>>> during an exception handler, there is no opportunity for memory
> > >>>> allocation
> > >>>> errors to be reported, as there are in the AMX case.
> > >>>
> > >>> Glibc has an SVE optimized memcpy, right? Doesn't that mean that pretty
> > >>> much all processes on an SVE capable system will enable SVE
> > >>> (lazily)? If
> > >>> so, that's close to "enabled by default" (unless SVE is disabled system
> > >>> wide).
> > >>
> > >> Yes, see sysdeps/aarch64/multiarch/memcpy.c:
> > >>
> > >> static inline __typeof (__redirect_memcpy) *
> > >> select_memcpy_ifunc (void)
> > >> {
> > >> INIT_ARCH ();
> > >> if (sve && HAVE_AARCH64_SVE_ASM)
> > >> {
> > >> if (IS_A64FX (midr))
> > >> return __memcpy_a64fx;
> > >> return __memcpy_sve;
> > >> }
> > >> if (IS_THUNDERX (midr))
> > >> return __memcpy_thunderx;
> > >> if (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr))
> > >> return __memcpy_thunderx2;
> > >> if (IS_FALKOR (midr) || IS_PHECDA (midr))
> > >> return __memcpy_falkor;
> > >> return __memcpy_generic;
> > >> }
> > >> And the __memcpy_sve implementation actually uses SVE.
> > >>
> > >> If there were a prctl to select the vector width and enable the vector
> > >> extension, we'd have to pick a width in glibc anyway.
> > >
> > > There *is* a prctl to adjust the SVE vector width, but glibc does not
> > > need to select because SVE dynamically adjusts to the currently
> > > enabled width. The kernel selects a default width that fits within
> > > the default signal frame size.
> > >
> > > The other thing of note for SVE is that, with the default function ABI
> > > all of the SVE state is call-clobbered, which allows the kernel to
> > > drop instead of save state across system calls. (There is a separate
> > > vector function call ABI when SVE types are used.)
> >
> > For the RV psABI, it is similar - all V regs are
> > caller-saved/call-clobbered [1] and syscalls are not required to
> > preserve V regs [2]
> > However last I checked ARM documentation the ABI doc seemed to suggest
> > that some (parts) of the SVE regs are callee-saved [3]
>
> Yes the lower 64 bits which overlap with the floating point registers.
I should expand on that, only the specific callee registers have to
save the lower 64bits because they overlap with the floating point
registers.
Thanks,
Andrew
>
> Thanks,
> Andrew Pinski
>
>
> >
> > >
> > > So while strcpy may enable SVE for the thread, the next syscall may
> > > disable it again.
> >
> > Next syscall could trash them, but will it disable SVE ? Despite
> > syscall/function-call clobbers, using V in tight loops such as mem*/str*
> > still is a win.
> >
> >
> > [1]
> > https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-cc.adoc
> > [2]
> > https://github.com/riscv/riscv-v-spec/blob/master/calling-convention.adoc
> > [3]
> > https://github.com/ARM-software/abi-aa/blob/2982a9f3b512a5bfdc9e3fea5d3b298f9165c36b/aapcs64/aapcs64.rst#the-base-procedure-call-standard
> > Sec 6.1.3 ".... In other cases it need only preserve the low 64 bits of
> > z8-z15"
> >
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-15 18:57 ` Vineet Gupta
@ 2022-12-15 19:56 ` Richard Henderson
-1 siblings, 0 replies; 147+ messages in thread
From: Richard Henderson @ 2022-12-15 19:56 UTC (permalink / raw)
To: Vineet Gupta, Florian Weimer, Björn Töpel
Cc: Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/15/22 10:57, Vineet Gupta wrote:
>> The other thing of note for SVE is that, with the default function ABI all of the SVE
>> state is call-clobbered, which allows the kernel to drop instead of save state across
>> system calls. (There is a separate vector function call ABI when SVE types are used.)
>
> For the RV psABI, it is similar - all V regs are caller-saved/call-clobbered [1] and
> syscalls are not required to preserve V regs [2]
> However last I checked ARM documentation the ABI doc seemed to suggest that some (parts)
> of the SVE regs are callee-saved [3]
As Pinski mentioned, just some low bits that overlap with scalar fp state; the high bits
and the predicate registers gets zeroed when re-enabling.
>> So while strcpy may enable SVE for the thread, the next syscall may disable it again.
>
> Next syscall could trash them, but will it disable SVE ?
Yes. See fp_user_discard() in arch/arm64/kernel/syscall.c.
r~
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-15 19:56 ` Richard Henderson
0 siblings, 0 replies; 147+ messages in thread
From: Richard Henderson @ 2022-12-15 19:56 UTC (permalink / raw)
To: Vineet Gupta, Florian Weimer, Björn Töpel
Cc: Darius Rad, Palmer Dabbelt, Andrew Waterman, stillson,
Paul Walmsley, anup, atishp, guoren, Conor Dooley, greentime.hu,
vincent.chen, andy.chiu, arnd, linux-kernel, linux-riscv,
libc-alpha, christoph.muellner, Aaron Durbin, linux
On 12/15/22 10:57, Vineet Gupta wrote:
>> The other thing of note for SVE is that, with the default function ABI all of the SVE
>> state is call-clobbered, which allows the kernel to drop instead of save state across
>> system calls. (There is a separate vector function call ABI when SVE types are used.)
>
> For the RV psABI, it is similar - all V regs are caller-saved/call-clobbered [1] and
> syscalls are not required to preserve V regs [2]
> However last I checked ARM documentation the ABI doc seemed to suggest that some (parts)
> of the SVE regs are callee-saved [3]
As Pinski mentioned, just some low bits that overlap with scalar fp state; the high bits
and the predicate registers gets zeroed when re-enabling.
>> So while strcpy may enable SVE for the thread, the next syscall may disable it again.
>
> Next syscall could trash them, but will it disable SVE ?
Yes. See fp_user_discard() in arch/arm64/kernel/syscall.c.
r~
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
2022-12-09 6:27 ` Palmer Dabbelt
@ 2022-12-09 13:58 ` Icenowy Zheng
-1 siblings, 0 replies; 147+ messages in thread
From: Icenowy Zheng @ 2022-12-09 13:58 UTC (permalink / raw)
To: Palmer Dabbelt, Vineet Gupta
Cc: stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, Andrew Waterman,
Darius Rad, arnd, linux-kernel, linux-riscv, bjorn, fweimer,
libc-alpha, christoph.muellner, Aaron Durbin, linux
在 2022-12-08星期四的 22:27 -0800，Palmer Dabbelt写道：
> On Thu, 08 Dec 2022 21:16:06 PST (-0800), Vineet Gupta wrote:
> > Hi Darius, Andrew, Palmer
> >
> > On 9/21/22 14:43, Chris Stillson wrote:
> > > diff --git a/arch/riscv/kernel/process.c
> > > b/arch/riscv/kernel/process.c
> > >
> > > @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs,
> > > unsigned long pc,
> > > if (WARN_ON(!vstate->datap))
> > > return;
> > > }
> > > - regs->status |= SR_VS_INITIAL;
> > >
> >
> > Perhaps not obvious from the patch, but this is a major user
> > experience
> > change: As in V unit would be turned off for a new task and we will
> > rely
> > on a userspace prctl (also introduced in this patch) to enable V.
>
> IMO that's the only viable option: enabling V adds more user-visible
> state, which is a uABI break. I haven't really had time to poke
> through
> all the versions here, but I'd have the call look something like
>
> prctl(RISCV_ENABLE_V, min_vlenb, max_vlenb, flags);
Should we make this extra switch more future-proof by not only limiting
it to V, but also other extensions that will introduce extra state,
e.g. P ?
>
> where
>
> * min_vlenb is the smallest VLENB that userspace can support.
> There's
> alreday an LLVM argument for this, I haven't dug into the generated
> code but I assume it'll blow up on smaller VLENB systems somehow.
> * max_vlenb is the largest VLENB that userspace can support.
> * flags is just a placeholder for now, with 0 meaning "V as defined
> by
> 1.0 for all threads in this proces". That should give us an out if
> something more complicated happens in the future.
>
> That way VLA code can call `prctl(RISCV_ENABLE_V, 128, 8192, 0)` as
> it
> supports any V 1.0 implementation, while code with other constraints
> can
> avoid having V turned on in an unsupported configuration.
>
> I think we can start out with no flags, but there's a few I could see
> being useful already:
>
> * Cross process/thread enabling. I think a reasonable default is
> "enable V for all current and future threads in this process", but
> one
> could imagine flags for "just this thread" vs "all current
> threads", a
> default for new threads, and a default for child processes. I
> don't
> think it matters so much what we pick as a default, just that it's
> written down.
> * Setting the VLENB bounds vs updating them. I'm thinking for shared
> libraries, where they'd only want to enable V in the shared library
> if
> it's already in a supported configuration. I'm not sure what the
> right rules are here, but again it's best to write that down.
> * Some way to disable V. Maybe we just say `prctl(RISCV_ENABLE_V, 0,
> 0,
> ...)` disables V, or maybe it's a flag? Again, it should just be
> written down.
> * What exactly we're enabling -- is it the V extension, or just the V
> registers?
>
> There's a bunch of subtly here, though, so I think we'd at least want
> glibc and gdb support posted before committing to any uABI. It's
> probably also worth looking at what the Arm folks did for SVE: I gave
> it
> a quick glance and it seems like there's a lot of similarities with
> what
> I'm suggesting here, but again a lot of this is pretty subtle stuff
> so
> it's hard to tell just at a glance.
>
> > I know some of you had different opinion on this in the past [1],
> > so
> > this is to make sure everyone's on same page.
> > And if we agree this is the way to go, how exactly will this be
> > done in
> > userspace.
> >
> > glibc dynamic loader will invoke the prctl() ? How will it decide
> > whether to do this (or not) - will it be unconditional or will it
> > use
> > the hwcap - does latter plumbing exist already ? If so is it
> > AT_HWCAP /
> > HWCAP2.
>
> That part I haven't sorted out yet, and I don't think it's sufficient
> to
> just say "userspace should enable what it can support" because of how
> pervasive V instructions are going to be.
>
> I don't think we need HWCAP, as userspace will need to call the
> prctl()
> anyway to turn on V and thus can just use the success/failure of that
> to
> sort things out.
>
> Maybe it's sufficient to rely on some sort of sticky prctl() (or
> sysctl
> type thing, the differences there would be pretty subtle) and just
> not
> worry about it, but having some way of encoding this in the ELF seems
> nice. That said, we've had a bunch of trouble sorting out the ISA
> encoding in ELFs so maybe it's just not worth bothering?
>
> > Also for static linked executables, where will the prctl be called
> > from ?
>
> I guess that's pretty far in the weeds, but we could at least hook
> CRT
> to insert the relevant code. We'd really need to sort out how we're
> going to encode the V support in binaries, though.
>
> > [1]
> > https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: RISCV Vector unit disabled by default for new task (was Re: [PATCH v12 17/17] riscv: prctl to enable vector commands)
@ 2022-12-09 13:58 ` Icenowy Zheng
0 siblings, 0 replies; 147+ messages in thread
From: Icenowy Zheng @ 2022-12-09 13:58 UTC (permalink / raw)
To: Palmer Dabbelt, Vineet Gupta
Cc: stillson, Paul Walmsley, anup, atishp, guoren, Conor Dooley,
greentime.hu, vincent.chen, andy.chiu, Andrew Waterman,
Darius Rad, arnd, linux-kernel, linux-riscv, bjorn, fweimer,
libc-alpha, christoph.muellner, Aaron Durbin, linux
在 2022-12-08星期四的 22:27 -0800，Palmer Dabbelt写道：
> On Thu, 08 Dec 2022 21:16:06 PST (-0800), Vineet Gupta wrote:
> > Hi Darius, Andrew, Palmer
> >
> > On 9/21/22 14:43, Chris Stillson wrote:
> > > diff --git a/arch/riscv/kernel/process.c
> > > b/arch/riscv/kernel/process.c
> > >
> > > @@ -134,7 +135,6 @@ void start_thread(struct pt_regs *regs,
> > > unsigned long pc,
> > > if (WARN_ON(!vstate->datap))
> > > return;
> > > }
> > > - regs->status |= SR_VS_INITIAL;
> > >
> >
> > Perhaps not obvious from the patch, but this is a major user
> > experience
> > change: As in V unit would be turned off for a new task and we will
> > rely
> > on a userspace prctl (also introduced in this patch) to enable V.
>
> IMO that's the only viable option: enabling V adds more user-visible
> state, which is a uABI break. I haven't really had time to poke
> through
> all the versions here, but I'd have the call look something like
>
> prctl(RISCV_ENABLE_V, min_vlenb, max_vlenb, flags);
Should we make this extra switch more future-proof by not only limiting
it to V, but also other extensions that will introduce extra state,
e.g. P ?
>
> where
>
> * min_vlenb is the smallest VLENB that userspace can support.
> There's
> alreday an LLVM argument for this, I haven't dug into the generated
> code but I assume it'll blow up on smaller VLENB systems somehow.
> * max_vlenb is the largest VLENB that userspace can support.
> * flags is just a placeholder for now, with 0 meaning "V as defined
> by
> 1.0 for all threads in this proces". That should give us an out if
> something more complicated happens in the future.
>
> That way VLA code can call `prctl(RISCV_ENABLE_V, 128, 8192, 0)` as
> it
> supports any V 1.0 implementation, while code with other constraints
> can
> avoid having V turned on in an unsupported configuration.
>
> I think we can start out with no flags, but there's a few I could see
> being useful already:
>
> * Cross process/thread enabling. I think a reasonable default is
> "enable V for all current and future threads in this process", but
> one
> could imagine flags for "just this thread" vs "all current
> threads", a
> default for new threads, and a default for child processes. I
> don't
> think it matters so much what we pick as a default, just that it's
> written down.
> * Setting the VLENB bounds vs updating them. I'm thinking for shared
> libraries, where they'd only want to enable V in the shared library
> if
> it's already in a supported configuration. I'm not sure what the
> right rules are here, but again it's best to write that down.
> * Some way to disable V. Maybe we just say `prctl(RISCV_ENABLE_V, 0,
> 0,
> ...)` disables V, or maybe it's a flag? Again, it should just be
> written down.
> * What exactly we're enabling -- is it the V extension, or just the V
> registers?
>
> There's a bunch of subtly here, though, so I think we'd at least want
> glibc and gdb support posted before committing to any uABI. It's
> probably also worth looking at what the Arm folks did for SVE: I gave
> it
> a quick glance and it seems like there's a lot of similarities with
> what
> I'm suggesting here, but again a lot of this is pretty subtle stuff
> so
> it's hard to tell just at a glance.
>
> > I know some of you had different opinion on this in the past [1],
> > so
> > this is to make sure everyone's on same page.
> > And if we agree this is the way to go, how exactly will this be
> > done in
> > userspace.
> >
> > glibc dynamic loader will invoke the prctl() ? How will it decide
> > whether to do this (or not) - will it be unconditional or will it
> > use
> > the hwcap - does latter plumbing exist already ? If so is it
> > AT_HWCAP /
> > HWCAP2.
>
> That part I haven't sorted out yet, and I don't think it's sufficient
> to
> just say "userspace should enable what it can support" because of how
> pervasive V instructions are going to be.
>
> I don't think we need HWCAP, as userspace will need to call the
> prctl()
> anyway to turn on V and thus can just use the success/failure of that
> to
> sort things out.
>
> Maybe it's sufficient to rely on some sort of sticky prctl() (or
> sysctl
> type thing, the differences there would be pretty subtle) and just
> not
> worry about it, but having some way of encoding this in the ELF seems
> nice. That said, we've had a bunch of trouble sorting out the ISA
> encoding in ELFs so maybe it's just not worth bothering?
>
> > Also for static linked executables, where will the prctl be called
> > from ?
>
> I guess that's pretty far in the weeds, but we could at least hook
> CRT
> to insert the relevant code. We'd really need to sort out how we're
> going to encode the V support in binaries, though.
>
> > [1]
> > https://sourceware.org/pipermail/libc-alpha/2021-November/132883.html
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 01/17] riscv: Rename __switch_to_aux -> fpu
2022-09-21 21:43 ` Chris Stillson
@ 2023-01-23 11:20 ` Heiko Stübner
-1 siblings, 0 replies; 147+ messages in thread
From: Heiko Stübner @ 2023-01-23 11:20 UTC (permalink / raw)
To: linux-riscv
Cc: Guo Ren, Guo Ren, Greentime Hu, Anup Patel, Palmer Dabbelt,
Paul Walmsley, Palmer Dabbelt, Albert Ou, Eric Biederman,
Kees Cook, Atish Patra, Oleg Nesterov, Guo Ren,
Heinrich Schuchardt, Conor Dooley, Arnaud Pouliquen,
Chris Stillson, Paolo Bonzini, Alexandre Ghiti, Arnd Bergmann,
Vincent Chen, Dao Lu, Jisheng Zhang, Sunil V L, Nick Knight,
Han-Kuan Chen, Changbin Du, Li Zhengyu, Alexander Graf,
Ard Biesheuvel, Tsukasa OI, Yury Norov, Nicolas Saenz Julienne,
Paul E. McKenney, Frederic Weisbecker, Mark Rutland, Vitaly Wool,
Myrtle Shah, Ruinland Tsai, Catalin Marinas, Mark Brown,
Will Deacon, Alexey Dobriyan, Huacai Chen, Janosch Frank,
Christian Brauner, Eugene Syromiatnikov, Peter Collingbourne,
Colin Cross, Andrew Morton, Suren Baghdasaryan, Barret Rhoden,
Davidlohr Bueso, linux-riscv, linux-kernel, linux-mm, kvm,
kvm-riscv, Chris Stillson
Am Mittwoch, 21. September 2022, 23:43:43 CET schrieb Chris Stillson:
> From: Guo Ren <ren_guo@c-sky.com>
>
> The name of __switch_to_aux is not clear and rename it with the
> determine function: __switch_to_fpu. Next we could add other regs'
> switch.
>
> Signed-off-by: Guo Ren <ren_guo@c-sky.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> Reviewed-by: Anup Patel <anup@brainfault.org>
> Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
^ permalink raw reply [flat|nested] 147+ messages in thread
* Re: [PATCH v12 01/17] riscv: Rename __switch_to_aux -> fpu
@ 2023-01-23 11:20 ` Heiko Stübner
0 siblings, 0 replies; 147+ messages in thread
From: Heiko Stübner @ 2023-01-23 11:20 UTC (permalink / raw)
To: linux-riscv
Cc: Mark Rutland, Guo Ren, kvm, Catalin Marinas, Mark Brown,
Paul Walmsley, Palmer Dabbelt, linux-kernel, linux-mm,
Conor Dooley, Guo Ren, Jisheng Zhang, Nicolas Saenz Julienne,
Will Deacon, Ard Biesheuvel, Vitaly Wool, Davidlohr Bueso,
Yury Norov, Nick Knight, Paul E. McKenney, Anup Patel,
Huacai Chen, Myrtle Shah, Eugene Syromiatnikov, linux-riscv,
Tsukasa OI, Ruinland Tsai, Guo Ren, Greentime Hu,
Arnaud Pouliquen, Li Zhengyu, Heinrich Schuchardt, Barret Rhoden,
Albert Ou, Kees Cook, Arnd Bergmann, Frederic Weisbecker,
Suren Baghdasaryan, Dao Lu, Alexander Graf, Atish Patra,
Alexandre Ghiti, Peter Collingbourne, Janosch Frank,
Chris Stillson, Christian Brauner, Oleg Nesterov, Colin Cross,
Alexey Dobriyan, Han-Kuan Chen, Vincent Chen, Changbin Du,
Palmer Dabbelt, Eric Biederman, kvm-riscv, Paolo Bonzini,
Andrew Morton
Am Mittwoch, 21. September 2022, 23:43:43 CET schrieb Chris Stillson:
> From: Guo Ren <ren_guo@c-sky.com>
>
> The name of __switch_to_aux is not clear and rename it with the
> determine function: __switch_to_fpu. Next we could add other regs'
> switch.
>
> Signed-off-by: Guo Ren <ren_guo@c-sky.com>
> Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
> Signed-off-by: Greentime Hu <greentime.hu@sifive.com>
> Reviewed-by: Anup Patel <anup@brainfault.org>
> Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 147+ messages in thread