[PATCH 1/5] powerpc/64s: Move dcbt/dcbtst sequence into a macro

[PATCH 1/5] powerpc/64s: Move dcbt/dcbtst sequence into a macro

[Michael Ellerman]

There's an almost identical code sequence to specify load/store access
hints in __copy_tofrom_user_power7(), copypage_power7() and
memcpy_power7().

Move the sequence into a common macro, which is passed the registers to
use as they differ slightly.

There also needs to be a copy in the selftests, it could be shared in
future if the headers are cleaned up / refactored.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/include/asm/ppc_asm.h                   | 12 ++++++++++++
 arch/powerpc/lib/copypage_power7.S                   | 12 +-----------
 arch/powerpc/lib/copyuser_power7.S                   | 12 +-----------
 arch/powerpc/lib/memcpy_power7.S                     | 10 +---------
 .../selftests/powerpc/copyloops/asm/ppc_asm.h        | 12 ++++++++++++
 5 files changed, 27 insertions(+), 31 deletions(-)

diff --git a/arch/powerpc/include/asm/ppc_asm.h b/arch/powerpc/include/asm/ppc_asm.h
index 041ee2595520..78c7548eac1e 100644
--- a/arch/powerpc/include/asm/ppc_asm.h
+++ b/arch/powerpc/include/asm/ppc_asm.h
@@ -510,6 +510,18 @@ END_FTR_SECTION_NESTED(CPU_FTR_CELL_TB_BUG, CPU_FTR_CELL_TB_BUG, 96)
        lis     scratch,0x60000000@h;			\
        dcbt    0,scratch,0b01010
 
+#define DCBT_SETUP_STREAMS(from, from_parms, to, to_parms, scratch)	\
+	lis	scratch,0x8000;	/* GO=1 */				\
+	clrldi	scratch,scratch,32;					\
+	/* setup read stream 0 */					\
+	dcbt	0,from,0b01000;		/* addr from */			\
+	dcbt	0,from_parms,0b01010;	/* length and depth from */	\
+	/* setup write stream 1 */					\
+	dcbtst	0,to,0b01000;		/* addr to */			\
+	dcbtst	0,to_parms,0b01010;	/* length and depth to */	\
+	eieio;								\
+	dcbt	0,scratch,0b01010;	/* all streams GO */
+
 /*
  * toreal/fromreal/tophys/tovirt macros. 32-bit BookE makes them
  * keep the address intact to be compatible with code shared with
diff --git a/arch/powerpc/lib/copypage_power7.S b/arch/powerpc/lib/copypage_power7.S
index a783973f1215..07e7cec4d135 100644
--- a/arch/powerpc/lib/copypage_power7.S
+++ b/arch/powerpc/lib/copypage_power7.S
@@ -27,17 +27,7 @@ _GLOBAL(copypage_power7)
 #endif
 	ori	r10,r7,1	/* stream=1 */
 
-	lis	r8,0x8000	/* GO=1 */
-	clrldi	r8,r8,32
-
-	/* setup read stream 0  */
-	dcbt	0,r4,0b01000  	/* addr from */
-	dcbt	0,r7,0b01010   /* length and depth from */
-	/* setup write stream 1 */
-	dcbtst	0,r9,0b01000   /* addr to */
-	dcbtst	0,r10,0b01010  /* length and depth to */
-	eieio
-	dcbt	0,r8,0b01010	/* all streams GO */
+	DCBT_SETUP_STREAMS(r4, r7, r9, r10, r8)
 
 #ifdef CONFIG_ALTIVEC
 	mflr	r0
diff --git a/arch/powerpc/lib/copyuser_power7.S b/arch/powerpc/lib/copyuser_power7.S
index ac41053c3a5a..8474c682a178 100644
--- a/arch/powerpc/lib/copyuser_power7.S
+++ b/arch/powerpc/lib/copyuser_power7.S
@@ -298,17 +298,7 @@ err1;	stb	r0,0(r3)
 	or	r7,r7,r0
 	ori	r10,r7,1	/* stream=1 */
 
-	lis	r8,0x8000	/* GO=1 */
-	clrldi	r8,r8,32
-
-	/* setup read stream 0 */
-	dcbt	0,r6,0b01000   /* addr from */
-	dcbt	0,r7,0b01010   /* length and depth from */
-	/* setup write stream 1 */
-	dcbtst	0,r9,0b01000   /* addr to */
-	dcbtst	0,r10,0b01010  /* length and depth to */
-	eieio
-	dcbt	0,r8,0b01010	/* all streams GO */
+	DCBT_SETUP_STREAMS(r6, r7, r9, r10, r8)
 
 	beq	cr1,.Lunwind_stack_nonvmx_copy
 
diff --git a/arch/powerpc/lib/memcpy_power7.S b/arch/powerpc/lib/memcpy_power7.S
index 9398b2b746c4..b7c5e7fca8b9 100644
--- a/arch/powerpc/lib/memcpy_power7.S
+++ b/arch/powerpc/lib/memcpy_power7.S
@@ -244,15 +244,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
 	or	r7,r7,r0
 	ori	r10,r7,1	/* stream=1 */
 
-	lis	r8,0x8000	/* GO=1 */
-	clrldi	r8,r8,32
-
-	dcbt	0,r6,0b01000
-	dcbt	0,r7,0b01010
-	dcbtst	0,r9,0b01000
-	dcbtst	0,r10,0b01010
-	eieio
-	dcbt	0,r8,0b01010	/* GO */
+	DCBT_SETUP_STREAMS(r6, r7, r9, r10, r8)
 
 	beq	cr1,.Lunwind_stack_nonvmx_copy
 
diff --git a/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h b/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
index a89f1fbf86ec..1d293ab77185 100644
--- a/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
+++ b/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
@@ -47,4 +47,16 @@
 /* Default to taking the first of any alternative feature sections */
 test_feature = 1
 
+#define DCBT_SETUP_STREAMS(from, from_parms, to, to_parms, scratch)	\
+	lis	scratch,0x8000;	/* GO=1 */				\
+	clrldi	scratch,scratch,32;					\
+	/* setup read stream 0 */					\
+	dcbt	0,from,0b01000;		/* addr from */			\
+	dcbt	0,from_parms,0b01010;	/* length and depth from */	\
+	/* setup write stream 1 */					\
+	dcbtst	0,to,0b01000;		/* addr to */			\
+	dcbtst	0,to_parms,0b01010;	/* length and depth to */	\
+	eieio;								\
+	dcbt	0,scratch,0b01010;	/* all streams GO */
+
 #endif /* __SELFTESTS_POWERPC_PPC_ASM_H */
-- 
2.43.2

[PATCH 2/5] powerpc/64s: Use .machine power4 around dcbt

[Michael Ellerman]

There are multiple decodings for the "dcbt" mnemonic, so the assembler
has to pick one.

That requires passing -many to the assembler, which is not recommended.

Without -many the clang 14 / binutils 2.38 build fails with:

  arch/powerpc/kernel/exceptions-64s.S:2976: Error: junk at end of line: `0b01010'
  clang: error: assembler command failed with exit code 1 (use -v to see invocation)

Fix it by adding .machine directives around the use of dcbt to specify
which encoding is desired.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/include/asm/ppc_asm.h | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/include/asm/ppc_asm.h b/arch/powerpc/include/asm/ppc_asm.h
index 78c7548eac1e..1d1018c1e482 100644
--- a/arch/powerpc/include/asm/ppc_asm.h
+++ b/arch/powerpc/include/asm/ppc_asm.h
@@ -508,11 +508,16 @@ END_FTR_SECTION_NESTED(CPU_FTR_CELL_TB_BUG, CPU_FTR_CELL_TB_BUG, 96)
  */
 #define DCBT_BOOK3S_STOP_ALL_STREAM_IDS(scratch)	\
        lis     scratch,0x60000000@h;			\
-       dcbt    0,scratch,0b01010
+       .machine push;					\
+       .machine power4;					\
+       dcbt    0,scratch,0b01010;			\
+       .machine pop;
 
 #define DCBT_SETUP_STREAMS(from, from_parms, to, to_parms, scratch)	\
 	lis	scratch,0x8000;	/* GO=1 */				\
 	clrldi	scratch,scratch,32;					\
+	.machine push;							\
+	.machine power4;						\
 	/* setup read stream 0 */					\
 	dcbt	0,from,0b01000;		/* addr from */			\
 	dcbt	0,from_parms,0b01010;	/* length and depth from */	\
@@ -520,7 +525,8 @@ END_FTR_SECTION_NESTED(CPU_FTR_CELL_TB_BUG, CPU_FTR_CELL_TB_BUG, 96)
 	dcbtst	0,to,0b01000;		/* addr to */			\
 	dcbtst	0,to_parms,0b01010;	/* length and depth to */	\
 	eieio;								\
-	dcbt	0,scratch,0b01010;	/* all streams GO */
+	dcbt	0,scratch,0b01010;	/* all streams GO */		\
+	.machine pop;
 
 /*
  * toreal/fromreal/tophys/tovirt macros. 32-bit BookE makes them
-- 
2.43.2

[PATCH 3/5] powerpc/fsl: Fix mfpmr build errors with newer binutils

[Michael Ellerman]

Binutils 2.38 complains about the use of mfpmr when building
ppc6xx_defconfig:

    CC      arch/powerpc/kernel/pmc.o
  {standard input}: Assembler messages:
  {standard input}:45: Error: unrecognized opcode: `mfpmr'
  {standard input}:56: Error: unrecognized opcode: `mtpmr'

This is because by default the kernel is built with -mcpu=powerpc, and
the mt/mfpmr instructions are not defined.

It can be avoided by enabling CONFIG_E300C3_CPU, but just adding that to
the defconfig will leave open the possibility of randconfig failures.

So add machine directives around the mt/mfpmr instructions to tell
binutils how to assemble them.

Cc: stable@vger.kernel.org
Reported-by: Jan-Benedict Glaw <jbglaw@lug-owl.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/include/asm/reg_fsl_emb.h | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/include/asm/reg_fsl_emb.h b/arch/powerpc/include/asm/reg_fsl_emb.h
index a21f529c43d9..8359c06d92d9 100644
--- a/arch/powerpc/include/asm/reg_fsl_emb.h
+++ b/arch/powerpc/include/asm/reg_fsl_emb.h
@@ -12,9 +12,16 @@
 #ifndef __ASSEMBLY__
 /* Performance Monitor Registers */
 #define mfpmr(rn)	({unsigned int rval; \
-			asm volatile("mfpmr %0," __stringify(rn) \
+			asm volatile(".machine push; " \
+				     ".machine e300; " \
+				     "mfpmr %0," __stringify(rn) ";" \
+				     ".machine pop; " \
 				     : "=r" (rval)); rval;})
-#define mtpmr(rn, v)	asm volatile("mtpmr " __stringify(rn) ",%0" : : "r" (v))
+#define mtpmr(rn, v)	asm volatile(".machine push; " \
+				     ".machine e300; " \
+				     "mtpmr " __stringify(rn) ",%0; " \
+				     ".machine pop; " \
+				     : : "r" (v))
 #endif /* __ASSEMBLY__ */
 
 /* Freescale Book E Performance Monitor APU Registers */
-- 
2.43.2

[PATCH 4/5] powerpc/fsl: Modernise mt/mfpmr

[Michael Ellerman]

With the addition of the machine directives, these are no longer simple
1-2 liner macros. So modernise them to be static inlines and use named
asm parameters.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/include/asm/reg_fsl_emb.h | 32 +++++++++++++++++---------
 1 file changed, 21 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/include/asm/reg_fsl_emb.h b/arch/powerpc/include/asm/reg_fsl_emb.h
index 8359c06d92d9..b0563c30b062 100644
--- a/arch/powerpc/include/asm/reg_fsl_emb.h
+++ b/arch/powerpc/include/asm/reg_fsl_emb.h
@@ -11,17 +11,27 @@
 
 #ifndef __ASSEMBLY__
 /* Performance Monitor Registers */
-#define mfpmr(rn)	({unsigned int rval; \
-			asm volatile(".machine push; " \
-				     ".machine e300; " \
-				     "mfpmr %0," __stringify(rn) ";" \
-				     ".machine pop; " \
-				     : "=r" (rval)); rval;})
-#define mtpmr(rn, v)	asm volatile(".machine push; " \
-				     ".machine e300; " \
-				     "mtpmr " __stringify(rn) ",%0; " \
-				     ".machine pop; " \
-				     : : "r" (v))
+static inline unsigned int mfpmr(unsigned int rn)
+{
+	unsigned int rval;
+
+	asm (".machine push; "
+	     ".machine e300; "
+	     "mfpmr %[rval], %[rn];"
+	     ".machine pop;"
+	     : [rval] "=r" (rval) : [rn] "i" (rn));
+
+	return rval;
+}
+
+static inline void mtpmr(unsigned int rn, unsigned int val)
+{
+	asm (".machine push; "
+	     ".machine e300; "
+	     "mtpmr %[rn], %[val];"
+	     ".machine pop;"
+	     : [val] "=r" (val) : [rn] "i" (rn));
+}
 #endif /* __ASSEMBLY__ */
 
 /* Freescale Book E Performance Monitor APU Registers */
-- 
2.43.2

[PATCH 5/5] powerpc: Remove cpu-as-y completely

[Michael Ellerman]

From: Christophe Leroy <christophe.leroy@csgroup.eu>

cpu-as-y is there to force assembler building options. But there is no
need for that. GCC is passed the necessary options and it automatically
pass the appropriate option to GAS.

GCC is given -maltivec when relevant, so no need for -Wa,-maltivec
either.

And -Wa,-many is wrong as it will hide innapropriate instructions.
Better to detect them and handle them on a case by case basis.

The setting of -Wa,-many was added by commit 960e30029863
("powerpc/Makefile: Fix PPC_BOOK3S_64 ASFLAGS") in order to fix an issue
with clang and the passed -Wa,-mpower4 option. But we have now removed
it expecting the compiler to automatically pass the proper options and
instructions based on -mcpu=power4.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/Makefile | 12 ------------
 1 file changed, 12 deletions(-)

diff --git a/arch/powerpc/Makefile b/arch/powerpc/Makefile
index 051247027da0..098133b34b68 100644
--- a/arch/powerpc/Makefile
+++ b/arch/powerpc/Makefile
@@ -219,18 +219,6 @@ KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
 # often slow when they are implemented at all
 KBUILD_CFLAGS		+= $(call cc-option,-mno-string)
 
-cpu-as-$(CONFIG_ALTIVEC)	+= $(call as-option,-Wa$(comma)-maltivec)
-
-# When using '-many -mpower4' gas will first try and find a matching power4
-# mnemonic and failing that it will allow any valid mnemonic that GAS knows
-# about. GCC will pass -many to GAS when assembling, clang does not.
-# LLVM IAS doesn't understand either flag: https://github.com/ClangBuiltLinux/linux/issues/675
-# but LLVM IAS only supports ISA >= 2.06 for Book3S 64 anyway...
-cpu-as-$(CONFIG_PPC_BOOK3S_64)	+= $(call as-option,-Wa$(comma)-mpower4) $(call as-option,-Wa$(comma)-many)
-
-KBUILD_AFLAGS += $(cpu-as-y)
-KBUILD_CFLAGS += $(cpu-as-y)
-
 KBUILD_AFLAGS += $(aflags-y)
 KBUILD_CFLAGS += $(cflags-y)
 
-- 
2.43.2

Re: [PATCH 2/5] powerpc/64s: Use .machine power4 around dcbt

[Segher Boessenkool]

On Thu, Feb 29, 2024 at 11:25:18PM +1100, Michael Ellerman wrote:
> There are multiple decodings for the "dcbt" mnemonic, so the assembler
> has to pick one.
> 
> That requires passing -many to the assembler, which is not recommended.
> 
> Without -many the clang 14 / binutils 2.38 build fails with:
> 
>   arch/powerpc/kernel/exceptions-64s.S:2976: Error: junk at end of line: `0b01010'
>   clang: error: assembler command failed with exit code 1 (use -v to see invocation)
> 
> Fix it by adding .machine directives around the use of dcbt to specify
> which encoding is desired.
> 
> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Looks good, thanks!

Acked-by: Segher Boessenkool <segher@kernel.crashing.org>


Segher

Re: [PATCH 4/5] powerpc/fsl: Modernise mt/mfpmr

[Segher Boessenkool]

On Thu, Feb 29, 2024 at 11:25:20PM +1100, Michael Ellerman wrote:
> With the addition of the machine directives, these are no longer simple
> 1-2 liner macros. So modernise them to be static inlines and use named
> asm parameters.
> 
> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

You got rid of the __stringify blight as well.  Great :-)

Acked-by: Segher Boessenkool <segher@kernel.crashing.org>


Segher

Re: [PATCH 5/5] powerpc: Remove cpu-as-y completely

[Segher Boessenkool]

On Thu, Feb 29, 2024 at 11:25:21PM +1100, Michael Ellerman wrote:
> From: Christophe Leroy <christophe.leroy@csgroup.eu>

Acked-by: Segher Boessenkool <segher@kernel.crashing.org>


Segher

Re: [PATCH 1/5] powerpc/64s: Move dcbt/dcbtst sequence into a macro

[Michael Ellerman]

On Thu, 29 Feb 2024 23:25:17 +1100, Michael Ellerman wrote:
> There's an almost identical code sequence to specify load/store access
> hints in __copy_tofrom_user_power7(), copypage_power7() and
> memcpy_power7().
> 
> Move the sequence into a common macro, which is passed the registers to
> use as they differ slightly.
> 
> [...]

Applied to powerpc/next.

[1/5] powerpc/64s: Move dcbt/dcbtst sequence into a macro
      https://git.kernel.org/powerpc/c/8488cdcb00fd5f238754005a43a3a7445860d344
[2/5] powerpc/64s: Use .machine power4 around dcbt
      https://git.kernel.org/powerpc/c/4e284e38ed586edeb8bdb2b0c544273a7f72021c
[3/5] powerpc/fsl: Fix mfpmr build errors with newer binutils
      https://git.kernel.org/powerpc/c/5f491356b7149564ab22323ccce79c8d595bfd0c
[4/5] powerpc/fsl: Modernise mt/mfpmr
      https://git.kernel.org/powerpc/c/f01dbd73ccf122486ad4b52e74f5505985dd6af4
[5/5] powerpc: Remove cpu-as-y completely
      https://git.kernel.org/powerpc/c/ca3d3aa14e7673f1b15e862b71998a4664d50ebe

cheers