From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Morse Subject: [PATCH v5 04/16] arm64: alternatives: use tpidr_el2 on VHE hosts Date: Wed, 6 Dec 2017 19:01:30 +0000 Message-ID: <20171206190142.9246-5-james.morse@arm.com> References: <20171206190142.9246-1-james.morse@arm.com> Return-path: In-Reply-To: <20171206190142.9246-1-james.morse-5wv7dgnIgG8@public.gmane.org> Sender: devicetree-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org Cc: kvmarm-FPEHb7Xf0XXUo1n7N8X6UoWGPAHP3yOg@public.gmane.org, devicetree-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Will Deacon , Catalin Marinas , Mark Rutland , Rob Herring , Marc Zyngier , Christoffer Dall , Lorenzo Pieralisi , Loc Ho , Heyi Guo List-Id: devicetree@vger.kernel.org Now that KVM uses tpidr_el2 in the same way as Linux's cpu_offset in tpidr_el1, merge the two. This saves KVM from save/restoring tpidr_el1 on VHE hosts, and allows future code to blindly access per-cpu variables without triggering world-switch. Signed-off-by: James Morse Reviewed-by: Christoffer Dall Reviewed-by: Catalin Marinas --- Changes since v3: * Moved 'alternatives_applied' test all into C, * Made enable method static and dragged up before first-use. Changes since v1: * cpu_copy_el2regs()'s 'have I been patched' test now always sets a register, just in case the compiler optimises out part of the logic. arch/arm64/include/asm/alternative.h | 2 ++ arch/arm64/include/asm/assembler.h | 8 ++++++++ arch/arm64/include/asm/percpu.h | 11 +++++++++-- arch/arm64/kernel/alternative.c | 9 +++++---- arch/arm64/kernel/cpufeature.c | 17 +++++++++++++++++ arch/arm64/mm/proc.S | 8 ++++++++ 6 files changed, 49 insertions(+), 6 deletions(-) diff --git a/arch/arm64/include/asm/alternative.h b/arch/arm64/include/asm/alternative.h index 4a85c6952a22..669028172fd6 100644 --- a/arch/arm64/include/asm/alternative.h +++ b/arch/arm64/include/asm/alternative.h @@ -12,6 +12,8 @@ #include #include +extern int alternatives_applied; + struct alt_instr { s32 orig_offset; /* offset to original instruction */ s32 alt_offset; /* offset to replacement instruction */ diff --git a/arch/arm64/include/asm/assembler.h b/arch/arm64/include/asm/assembler.h index c45bc94f15d0..c070c263021e 100644 --- a/arch/arm64/include/asm/assembler.h +++ b/arch/arm64/include/asm/assembler.h @@ -254,7 +254,11 @@ lr .req x30 // link register #else adr_l \dst, \sym #endif +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN mrs \tmp, tpidr_el1 +alternative_else + mrs \tmp, tpidr_el2 +alternative_endif add \dst, \dst, \tmp .endm @@ -265,7 +269,11 @@ lr .req x30 // link register */ .macro ldr_this_cpu dst, sym, tmp adr_l \dst, \sym +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN mrs \tmp, tpidr_el1 +alternative_else + mrs \tmp, tpidr_el2 +alternative_endif ldr \dst, [\dst, \tmp] .endm diff --git a/arch/arm64/include/asm/percpu.h b/arch/arm64/include/asm/percpu.h index 3bd498e4de4c..43393208229e 100644 --- a/arch/arm64/include/asm/percpu.h +++ b/arch/arm64/include/asm/percpu.h @@ -16,11 +16,15 @@ #ifndef __ASM_PERCPU_H #define __ASM_PERCPU_H +#include #include static inline void set_my_cpu_offset(unsigned long off) { - asm volatile("msr tpidr_el1, %0" :: "r" (off) : "memory"); + asm volatile(ALTERNATIVE("msr tpidr_el1, %0", + "msr tpidr_el2, %0", + ARM64_HAS_VIRT_HOST_EXTN) + :: "r" (off) : "memory"); } static inline unsigned long __my_cpu_offset(void) @@ -31,7 +35,10 @@ static inline unsigned long __my_cpu_offset(void) * We want to allow caching the value, so avoid using volatile and * instead use a fake stack read to hazard against barrier(). */ - asm("mrs %0, tpidr_el1" : "=r" (off) : + asm(ALTERNATIVE("mrs %0, tpidr_el1", + "mrs %0, tpidr_el2", + ARM64_HAS_VIRT_HOST_EXTN) + : "=r" (off) : "Q" (*(const unsigned long *)current_stack_pointer)); return off; diff --git a/arch/arm64/kernel/alternative.c b/arch/arm64/kernel/alternative.c index 6dd0a3a3e5c9..414288a558c8 100644 --- a/arch/arm64/kernel/alternative.c +++ b/arch/arm64/kernel/alternative.c @@ -32,6 +32,8 @@ #define ALT_ORIG_PTR(a) __ALT_PTR(a, orig_offset) #define ALT_REPL_PTR(a) __ALT_PTR(a, alt_offset) +int alternatives_applied; + struct alt_region { struct alt_instr *begin; struct alt_instr *end; @@ -143,7 +145,6 @@ static void __apply_alternatives(void *alt_region, bool use_linear_alias) */ static int __apply_alternatives_multi_stop(void *unused) { - static int patched = 0; struct alt_region region = { .begin = (struct alt_instr *)__alt_instructions, .end = (struct alt_instr *)__alt_instructions_end, @@ -151,14 +152,14 @@ static int __apply_alternatives_multi_stop(void *unused) /* We always have a CPU 0 at this point (__init) */ if (smp_processor_id()) { - while (!READ_ONCE(patched)) + while (!READ_ONCE(alternatives_applied)) cpu_relax(); isb(); } else { - BUG_ON(patched); + BUG_ON(alternatives_applied); __apply_alternatives(®ion, true); /* Barriers provided by the cache flushing */ - WRITE_ONCE(patched, 1); + WRITE_ONCE(alternatives_applied, 1); } return 0; diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c index 98e6563015a4..cb9c4e5a72f8 100644 --- a/arch/arm64/kernel/cpufeature.c +++ b/arch/arm64/kernel/cpufeature.c @@ -879,6 +879,22 @@ static int __init parse_kpti(char *str) __setup("kpti=", parse_kpti); #endif /* CONFIG_UNMAP_KERNEL_AT_EL0 */ +static int cpu_copy_el2regs(void *__unused) +{ + /* + * Copy register values that aren't redirected by hardware. + * + * Before code patching, we only set tpidr_el1, all CPUs need to copy + * this value to tpidr_el2 before we patch the code. Once we've done + * that, freshly-onlined CPUs will set tpidr_el2, so we don't need to + * do anything here. + */ + if (!alternatives_applied) + write_sysreg(read_sysreg(tpidr_el1), tpidr_el2); + + return 0; +} + static const struct arm64_cpu_capabilities arm64_features[] = { { .desc = "GIC system register CPU interface", @@ -948,6 +964,7 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .capability = ARM64_HAS_VIRT_HOST_EXTN, .def_scope = SCOPE_SYSTEM, .matches = runs_at_el2, + .enable = cpu_copy_el2regs, }, { .desc = "32-bit EL0 Support", diff --git a/arch/arm64/mm/proc.S b/arch/arm64/mm/proc.S index 3146dc96f05b..2f5051411783 100644 --- a/arch/arm64/mm/proc.S +++ b/arch/arm64/mm/proc.S @@ -70,7 +70,11 @@ ENTRY(cpu_do_suspend) mrs x8, mdscr_el1 mrs x9, oslsr_el1 mrs x10, sctlr_el1 +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN mrs x11, tpidr_el1 +alternative_else + mrs x11, tpidr_el2 +alternative_endif mrs x12, sp_el0 stp x2, x3, [x0] stp x4, xzr, [x0, #16] @@ -116,7 +120,11 @@ ENTRY(cpu_do_resume) msr mdscr_el1, x10 msr sctlr_el1, x12 +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN msr tpidr_el1, x13 +alternative_else + msr tpidr_el2, x13 +alternative_endif msr sp_el0, x14 /* * Restore oslsr_el1 by writing oslar_el1 -- 2.15.0 -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html From mboxrd@z Thu Jan 1 00:00:00 1970 From: james.morse@arm.com (James Morse) Date: Wed, 6 Dec 2017 19:01:30 +0000 Subject: [PATCH v5 04/16] arm64: alternatives: use tpidr_el2 on VHE hosts In-Reply-To: <20171206190142.9246-1-james.morse@arm.com> References: <20171206190142.9246-1-james.morse@arm.com> Message-ID: <20171206190142.9246-5-james.morse@arm.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Now that KVM uses tpidr_el2 in the same way as Linux's cpu_offset in tpidr_el1, merge the two. This saves KVM from save/restoring tpidr_el1 on VHE hosts, and allows future code to blindly access per-cpu variables without triggering world-switch. Signed-off-by: James Morse Reviewed-by: Christoffer Dall Reviewed-by: Catalin Marinas --- Changes since v3: * Moved 'alternatives_applied' test all into C, * Made enable method static and dragged up before first-use. Changes since v1: * cpu_copy_el2regs()'s 'have I been patched' test now always sets a register, just in case the compiler optimises out part of the logic. arch/arm64/include/asm/alternative.h | 2 ++ arch/arm64/include/asm/assembler.h | 8 ++++++++ arch/arm64/include/asm/percpu.h | 11 +++++++++-- arch/arm64/kernel/alternative.c | 9 +++++---- arch/arm64/kernel/cpufeature.c | 17 +++++++++++++++++ arch/arm64/mm/proc.S | 8 ++++++++ 6 files changed, 49 insertions(+), 6 deletions(-) diff --git a/arch/arm64/include/asm/alternative.h b/arch/arm64/include/asm/alternative.h index 4a85c6952a22..669028172fd6 100644 --- a/arch/arm64/include/asm/alternative.h +++ b/arch/arm64/include/asm/alternative.h @@ -12,6 +12,8 @@ #include #include +extern int alternatives_applied; + struct alt_instr { s32 orig_offset; /* offset to original instruction */ s32 alt_offset; /* offset to replacement instruction */ diff --git a/arch/arm64/include/asm/assembler.h b/arch/arm64/include/asm/assembler.h index c45bc94f15d0..c070c263021e 100644 --- a/arch/arm64/include/asm/assembler.h +++ b/arch/arm64/include/asm/assembler.h @@ -254,7 +254,11 @@ lr .req x30 // link register #else adr_l \dst, \sym #endif +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN mrs \tmp, tpidr_el1 +alternative_else + mrs \tmp, tpidr_el2 +alternative_endif add \dst, \dst, \tmp .endm @@ -265,7 +269,11 @@ lr .req x30 // link register */ .macro ldr_this_cpu dst, sym, tmp adr_l \dst, \sym +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN mrs \tmp, tpidr_el1 +alternative_else + mrs \tmp, tpidr_el2 +alternative_endif ldr \dst, [\dst, \tmp] .endm diff --git a/arch/arm64/include/asm/percpu.h b/arch/arm64/include/asm/percpu.h index 3bd498e4de4c..43393208229e 100644 --- a/arch/arm64/include/asm/percpu.h +++ b/arch/arm64/include/asm/percpu.h @@ -16,11 +16,15 @@ #ifndef __ASM_PERCPU_H #define __ASM_PERCPU_H +#include #include static inline void set_my_cpu_offset(unsigned long off) { - asm volatile("msr tpidr_el1, %0" :: "r" (off) : "memory"); + asm volatile(ALTERNATIVE("msr tpidr_el1, %0", + "msr tpidr_el2, %0", + ARM64_HAS_VIRT_HOST_EXTN) + :: "r" (off) : "memory"); } static inline unsigned long __my_cpu_offset(void) @@ -31,7 +35,10 @@ static inline unsigned long __my_cpu_offset(void) * We want to allow caching the value, so avoid using volatile and * instead use a fake stack read to hazard against barrier(). */ - asm("mrs %0, tpidr_el1" : "=r" (off) : + asm(ALTERNATIVE("mrs %0, tpidr_el1", + "mrs %0, tpidr_el2", + ARM64_HAS_VIRT_HOST_EXTN) + : "=r" (off) : "Q" (*(const unsigned long *)current_stack_pointer)); return off; diff --git a/arch/arm64/kernel/alternative.c b/arch/arm64/kernel/alternative.c index 6dd0a3a3e5c9..414288a558c8 100644 --- a/arch/arm64/kernel/alternative.c +++ b/arch/arm64/kernel/alternative.c @@ -32,6 +32,8 @@ #define ALT_ORIG_PTR(a) __ALT_PTR(a, orig_offset) #define ALT_REPL_PTR(a) __ALT_PTR(a, alt_offset) +int alternatives_applied; + struct alt_region { struct alt_instr *begin; struct alt_instr *end; @@ -143,7 +145,6 @@ static void __apply_alternatives(void *alt_region, bool use_linear_alias) */ static int __apply_alternatives_multi_stop(void *unused) { - static int patched = 0; struct alt_region region = { .begin = (struct alt_instr *)__alt_instructions, .end = (struct alt_instr *)__alt_instructions_end, @@ -151,14 +152,14 @@ static int __apply_alternatives_multi_stop(void *unused) /* We always have a CPU 0 at this point (__init) */ if (smp_processor_id()) { - while (!READ_ONCE(patched)) + while (!READ_ONCE(alternatives_applied)) cpu_relax(); isb(); } else { - BUG_ON(patched); + BUG_ON(alternatives_applied); __apply_alternatives(®ion, true); /* Barriers provided by the cache flushing */ - WRITE_ONCE(patched, 1); + WRITE_ONCE(alternatives_applied, 1); } return 0; diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c index 98e6563015a4..cb9c4e5a72f8 100644 --- a/arch/arm64/kernel/cpufeature.c +++ b/arch/arm64/kernel/cpufeature.c @@ -879,6 +879,22 @@ static int __init parse_kpti(char *str) __setup("kpti=", parse_kpti); #endif /* CONFIG_UNMAP_KERNEL_AT_EL0 */ +static int cpu_copy_el2regs(void *__unused) +{ + /* + * Copy register values that aren't redirected by hardware. + * + * Before code patching, we only set tpidr_el1, all CPUs need to copy + * this value to tpidr_el2 before we patch the code. Once we've done + * that, freshly-onlined CPUs will set tpidr_el2, so we don't need to + * do anything here. + */ + if (!alternatives_applied) + write_sysreg(read_sysreg(tpidr_el1), tpidr_el2); + + return 0; +} + static const struct arm64_cpu_capabilities arm64_features[] = { { .desc = "GIC system register CPU interface", @@ -948,6 +964,7 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .capability = ARM64_HAS_VIRT_HOST_EXTN, .def_scope = SCOPE_SYSTEM, .matches = runs_at_el2, + .enable = cpu_copy_el2regs, }, { .desc = "32-bit EL0 Support", diff --git a/arch/arm64/mm/proc.S b/arch/arm64/mm/proc.S index 3146dc96f05b..2f5051411783 100644 --- a/arch/arm64/mm/proc.S +++ b/arch/arm64/mm/proc.S @@ -70,7 +70,11 @@ ENTRY(cpu_do_suspend) mrs x8, mdscr_el1 mrs x9, oslsr_el1 mrs x10, sctlr_el1 +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN mrs x11, tpidr_el1 +alternative_else + mrs x11, tpidr_el2 +alternative_endif mrs x12, sp_el0 stp x2, x3, [x0] stp x4, xzr, [x0, #16] @@ -116,7 +120,11 @@ ENTRY(cpu_do_resume) msr mdscr_el1, x10 msr sctlr_el1, x12 +alternative_if_not ARM64_HAS_VIRT_HOST_EXTN msr tpidr_el1, x13 +alternative_else + msr tpidr_el2, x13 +alternative_endif msr sp_el0, x14 /* * Restore oslsr_el1 by writing oslar_el1 -- 2.15.0