[PATCH 0/2] powerpc/sstep: Fix compilation on 32-bit PowerPC

[PATCH 0/2] powerpc/sstep: Fix compilation on 32-bit PowerPC

[Link Mauve]

Two local variables were only set but never read on 32-bit PowerPC,
which causes an error during compilation.

This series first makes the 64-bit guards consistent by converting all
uses of CONFIG_PPC64 to use __powerpc64__ instead, and then guard the
two variables rc and suffix behind __powerpc64__ checks.

Link Mauve (2):
  powerpc/sstep: Consistently use one define to check arch bit width
  powerpc/sstep: Don’t define variables we won’t be using

 arch/powerpc/lib/sstep.c | 26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)


base-commit: 78bc8af4affb9a732504eb22eeac7d1e50883853
-- 
2.55.0

[PATCH 1/2] powerpc/sstep: Consistently use one define to check arch bit width

[Link Mauve]

Previously, both CONFIG_PPC64 and __powerpc64__ were being used to check
whether we were on 32-bit or on 64-bit PowerPC, sometimes in a single
function.

I picked __powerpc64__ since it was used six times as much as
CONFIG_PPC64 in this file, but could be convinced to use the other one
if there is any reason to prefer it.

Signed-off-by: Link Mauve <linkmauve@linkmauve.fr>
---
 arch/powerpc/lib/sstep.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/arch/powerpc/lib/sstep.c b/arch/powerpc/lib/sstep.c
index f0d6aa657c1a..597a0d841fe8 100644
--- a/arch/powerpc/lib/sstep.c
+++ b/arch/powerpc/lib/sstep.c
@@ -15,7 +15,7 @@
 #include <asm/cputable.h>
 #include <asm/disassemble.h>
 
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 /* Bits in SRR1 that are copied from MSR */
 #define MSR_MASK	0xffffffff87c0ffffUL
 #else
@@ -1256,7 +1256,7 @@ static nokprobe_inline void do_popcnt(const struct pt_regs *regs,
 	op->val = out;	/* popcntd */
 }
 
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 static nokprobe_inline void do_bpermd(const struct pt_regs *regs,
 				      struct instruction_op *op,
 				      unsigned long v1, unsigned long v2)
@@ -1273,7 +1273,7 @@ static nokprobe_inline void do_bpermd(const struct pt_regs *regs,
 	}
 	op->val = perm;
 }
-#endif /* CONFIG_PPC64 */
+#endif /* __powerpc64__ */
 /*
  * The size parameter adjusts the equivalent prty instruction.
  * prtyw = 32, prtyd = 64
@@ -1340,7 +1340,7 @@ static nokprobe_inline int trap_compare(long v1, long v2)
 int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
 		  ppc_inst_t instr)
 {
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 	unsigned int suffixopcode, prefixtype, prefix_r;
 #endif
 	unsigned int opcode, ra, rb, rc, rd, spr, u;
@@ -1739,7 +1739,7 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
 			op->reg = rd;
 			op->val = 0xffffffff & ~(MSR_ME | MSR_LE);
 			return 0;
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 		case 178:	/* mtmsrd */
 			if (user_mode(regs))
 				goto priv;
@@ -2054,7 +2054,7 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
 		case 186:	/* prtyd */
 			do_prty(regs, op, regs->gpr[rd], 64);
 			goto logical_done_nocc;
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 		case 252:	/* bpermd */
 			do_bpermd(regs, op, regs->gpr[rd], regs->gpr[rb]);
 			goto logical_done_nocc;
@@ -2082,7 +2082,7 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
 		case 476:	/* nand */
 			op->val = ~(regs->gpr[rd] & regs->gpr[rb]);
 			goto logical_done;
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 		case 506:	/* popcntd */
 			do_popcnt(regs, op, regs->gpr[rd], 64);
 			goto logical_done_nocc;
@@ -3247,7 +3247,7 @@ void emulate_update_regs(struct pt_regs *regs, struct instruction_op *op)
 		case BARRIER_EIEIO:
 			eieio();
 			break;
-#ifdef CONFIG_PPC64
+#ifdef __powerpc64__
 		case BARRIER_LWSYNC:
 			asm volatile("lwsync" : : : "memory");
 			break;
-- 
2.55.0

Re: [PATCH 1/2] powerpc/sstep: Consistently use one define to check arch bit width

[LEROY Christophe]

Le 29/07/2026 à 22:49, Link Mauve a écrit :
> Previously, both CONFIG_PPC64 and __powerpc64__ were being used to check
> whether we were on 32-bit or on 64-bit PowerPC, sometimes in a single
> function.
> 
> I picked __powerpc64__ since it was used six times as much as
> CONFIG_PPC64 in this file, but could be convinced to use the other one
> if there is any reason to prefer it.

Well, we tend to use the other as much as possible:

$ git grep -w -e "#ifdef CONFIG_PPC64" -e "defined(CONFIG_PPC64)" | wc -l
368
$ git grep -w -e "#ifdef __powerpc64__"  | wc -l
118

Allthough in most cases you can use one or the other, they don't mean 
the same.

CONFIG_PPC64 means you are building a 64 bit kernel.
__powerpc64__ means you are building 64 bit code.

In a 64 bit kernel you can have 32 bit code, for instance the vdso32.

If you look into include/asm/feature-fixups.h for instance you have :

#if defined(CONFIG_PPC64) && !defined(__powerpc64__)
/* 64 bits kernel, 32 bits code (ie. vdso32) */
#define FTR_ENTRY_LONG		.8byte
#define FTR_ENTRY_OFFSET	.long 0xffffffff; .long
#elif defined(CONFIG_PPC64)
#define FTR_ENTRY_LONG		.8byte
#define FTR_ENTRY_OFFSET	.8byte
#else
#define FTR_ENTRY_LONG		.long
#define FTR_ENTRY_OFFSET	.long
#endif

Unless you are a good reason to use __powerpc64__ it is often better to 
use CONFIG_PPC64

Christophe

> 
> Signed-off-by: Link Mauve <linkmauve@linkmauve.fr>
> ---
>   arch/powerpc/lib/sstep.c | 16 ++++++++--------
>   1 file changed, 8 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/powerpc/lib/sstep.c b/arch/powerpc/lib/sstep.c
> index f0d6aa657c1a..597a0d841fe8 100644
> --- a/arch/powerpc/lib/sstep.c
> +++ b/arch/powerpc/lib/sstep.c
> @@ -15,7 +15,7 @@
>   #include <asm/cputable.h>
>   #include <asm/disassemble.h>
>   
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   /* Bits in SRR1 that are copied from MSR */
>   #define MSR_MASK	0xffffffff87c0ffffUL
>   #else
> @@ -1256,7 +1256,7 @@ static nokprobe_inline void do_popcnt(const struct pt_regs *regs,
>   	op->val = out;	/* popcntd */
>   }
>   
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   static nokprobe_inline void do_bpermd(const struct pt_regs *regs,
>   				      struct instruction_op *op,
>   				      unsigned long v1, unsigned long v2)
> @@ -1273,7 +1273,7 @@ static nokprobe_inline void do_bpermd(const struct pt_regs *regs,
>   	}
>   	op->val = perm;
>   }
> -#endif /* CONFIG_PPC64 */
> +#endif /* __powerpc64__ */
>   /*
>    * The size parameter adjusts the equivalent prty instruction.
>    * prtyw = 32, prtyd = 64
> @@ -1340,7 +1340,7 @@ static nokprobe_inline int trap_compare(long v1, long v2)
>   int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
>   		  ppc_inst_t instr)
>   {
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   	unsigned int suffixopcode, prefixtype, prefix_r;
>   #endif
>   	unsigned int opcode, ra, rb, rc, rd, spr, u;
> @@ -1739,7 +1739,7 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
>   			op->reg = rd;
>   			op->val = 0xffffffff & ~(MSR_ME | MSR_LE);
>   			return 0;
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   		case 178:	/* mtmsrd */
>   			if (user_mode(regs))
>   				goto priv;
> @@ -2054,7 +2054,7 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
>   		case 186:	/* prtyd */
>   			do_prty(regs, op, regs->gpr[rd], 64);
>   			goto logical_done_nocc;
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   		case 252:	/* bpermd */
>   			do_bpermd(regs, op, regs->gpr[rd], regs->gpr[rb]);
>   			goto logical_done_nocc;
> @@ -2082,7 +2082,7 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
>   		case 476:	/* nand */
>   			op->val = ~(regs->gpr[rd] & regs->gpr[rb]);
>   			goto logical_done;
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   		case 506:	/* popcntd */
>   			do_popcnt(regs, op, regs->gpr[rd], 64);
>   			goto logical_done_nocc;
> @@ -3247,7 +3247,7 @@ void emulate_update_regs(struct pt_regs *regs, struct instruction_op *op)
>   		case BARRIER_EIEIO:
>   			eieio();
>   			break;
> -#ifdef CONFIG_PPC64
> +#ifdef __powerpc64__
>   		case BARRIER_LWSYNC:
>   			asm volatile("lwsync" : : : "memory");
>   			break;

[PATCH 2/2] powerpc/sstep: Don’t define variables we won’t be using

[Link Mauve]

It seems this function has only compiled without warnings on PPC64, not
on PowerPC, so before this commit it would fail with:
```
../arch/powerpc/lib/sstep.c: In function 'analyse_instr':
../arch/powerpc/lib/sstep.c:1350:28: error: variable 'suffix' set but not used [-Werror=unused-but-set-variable]
 1350 |         unsigned int word, suffix;
      |                            ^~~~~~
../arch/powerpc/lib/sstep.c:1346:38: error: variable 'rc' set but not used [-Werror=unused-but-set-variable]
 1346 |         unsigned int opcode, ra, rb, rc, rd, spr, u;
      |                                      ^~
cc1: all warnings being treated as errors
```

Signed-off-by: Link Mauve <linkmauve@linkmauve.fr>
---
 arch/powerpc/lib/sstep.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/lib/sstep.c b/arch/powerpc/lib/sstep.c
index 597a0d841fe8..c866e4337f94 100644
--- a/arch/powerpc/lib/sstep.c
+++ b/arch/powerpc/lib/sstep.c
@@ -1342,16 +1342,20 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
 {
 #ifdef __powerpc64__
 	unsigned int suffixopcode, prefixtype, prefix_r;
+	unsigned int rc;
+	unsigned int suffix;
 #endif
-	unsigned int opcode, ra, rb, rc, rd, spr, u;
+	unsigned int opcode, ra, rb, rd, spr, u;
 	unsigned long int imm;
 	unsigned long int val, val2;
 	unsigned int mb, me, sh;
-	unsigned int word, suffix;
+	unsigned int word;
 	long ival;
 
 	word = ppc_inst_val(instr);
+#ifdef __powerpc64__
 	suffix = ppc_inst_suffix(instr);
+#endif
 
 	op->type = COMPUTE;
 
@@ -1469,7 +1473,9 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
 	rd = (word >> 21) & 0x1f;
 	ra = (word >> 16) & 0x1f;
 	rb = (word >> 11) & 0x1f;
+#ifdef __powerpc64__
 	rc = (word >> 6) & 0x1f;
+#endif
 
 	switch (opcode) {
 #ifdef __powerpc64__
-- 
2.55.0

Re: [PATCH 2/2] powerpc/sstep: Don’t define variables we won’t be using

[Christophe Leroy (CS GROUP)]

Le 29/07/2026 à 22:49, Link Mauve a écrit :
> It seems this function has only compiled without warnings on PPC64, not
> on PowerPC, so before this commit it would fail with:
> ```
> ../arch/powerpc/lib/sstep.c: In function 'analyse_instr':
> ../arch/powerpc/lib/sstep.c:1350:28: error: variable 'suffix' set but not used [-Werror=unused-but-set-variable]
>   1350 |         unsigned int word, suffix;
>        |                            ^~~~~~
> ../arch/powerpc/lib/sstep.c:1346:38: error: variable 'rc' set but not used [-Werror=unused-but-set-variable]
>   1346 |         unsigned int opcode, ra, rb, rc, rd, spr, u;
>        |                                      ^~
> cc1: all warnings being treated as errors
> ```
> 

I'm surprised we still have this issue. I remember seeing it reported 
several times allthough I've not been able to spot when.

By the way on my side it happens only when I do a W=1 build.

Same as previous patch, I'd prefer using CONFIG_PPC64 instead of 
__powerpc64__

> Signed-off-by: Link Mauve <linkmauve@linkmauve.fr>
> ---
>   arch/powerpc/lib/sstep.c | 10 ++++++++--
>   1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/powerpc/lib/sstep.c b/arch/powerpc/lib/sstep.c
> index 597a0d841fe8..c866e4337f94 100644
> --- a/arch/powerpc/lib/sstep.c
> +++ b/arch/powerpc/lib/sstep.c
> @@ -1342,16 +1342,20 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
>   {
>   #ifdef __powerpc64__
>   	unsigned int suffixopcode, prefixtype, prefix_r;
> +	unsigned int rc;
> +	unsigned int suffix;

A single line would be enough:

	unsigned int rc, suffix;

>   #endif
> -	unsigned int opcode, ra, rb, rc, rd, spr, u;
> +	unsigned int opcode, ra, rb, rd, spr, u;
>   	unsigned long int imm;
>   	unsigned long int val, val2;
>   	unsigned int mb, me, sh;
> -	unsigned int word, suffix;
> +	unsigned int word;
>   	long ival;
>   
>   	word = ppc_inst_val(instr);
> +#ifdef __powerpc64__
>   	suffix = ppc_inst_suffix(instr);
> +#endif
>   
>   	op->type = COMPUTE;
>   
> @@ -1469,7 +1473,9 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
>   	rd = (word >> 21) & 0x1f;
>   	ra = (word >> 16) & 0x1f;
>   	rb = (word >> 11) & 0x1f;
> +#ifdef __powerpc64__
>   	rc = (word >> 6) & 0x1f;
> +#endif
>   
>   	switch (opcode) {
>   #ifdef __powerpc64__

Re: [PATCH 2/2] powerpc/sstep: Don’t define variables we won’t be using

[Link Mauve]

On Thu, Jul 30, 2026 at 08:24:36AM +0200, Christophe Leroy (CS GROUP) wrote:
> 
> 
> Le 29/07/2026 à 22:49, Link Mauve a écrit :
> > It seems this function has only compiled without warnings on PPC64, not
> > on PowerPC, so before this commit it would fail with:
> > ```
> > ../arch/powerpc/lib/sstep.c: In function 'analyse_instr':
> > ../arch/powerpc/lib/sstep.c:1350:28: error: variable 'suffix' set but not used [-Werror=unused-but-set-variable]
> >   1350 |         unsigned int word, suffix;
> >        |                            ^~~~~~
> > ../arch/powerpc/lib/sstep.c:1346:38: error: variable 'rc' set but not used [-Werror=unused-but-set-variable]
> >   1346 |         unsigned int opcode, ra, rb, rc, rd, spr, u;
> >        |                                      ^~
> > cc1: all warnings being treated as errors
> > ```
> > 
> 
> I'm surprised we still have this issue. I remember seeing it reported
> several times allthough I've not been able to spot when.
> 
> By the way on my side it happens only when I do a W=1 build.

Oh indeed, I was using W=1 all along, thanks for spotting that!

> 
> Same as previous patch, I'd prefer using CONFIG_PPC64 instead of
> __powerpc64__

Will do in v2.

> 
> > Signed-off-by: Link Mauve <linkmauve@linkmauve.fr>
> > ---
> >   arch/powerpc/lib/sstep.c | 10 ++++++++--
> >   1 file changed, 8 insertions(+), 2 deletions(-)
> > 
> > diff --git a/arch/powerpc/lib/sstep.c b/arch/powerpc/lib/sstep.c
> > index 597a0d841fe8..c866e4337f94 100644
> > --- a/arch/powerpc/lib/sstep.c
> > +++ b/arch/powerpc/lib/sstep.c
> > @@ -1342,16 +1342,20 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
> >   {
> >   #ifdef __powerpc64__
> >   	unsigned int suffixopcode, prefixtype, prefix_r;
> > +	unsigned int rc;
> > +	unsigned int suffix;
> 
> A single line would be enough:

Will do in v2.

> 
> 	unsigned int rc, suffix;
> 
> >   #endif
> > -	unsigned int opcode, ra, rb, rc, rd, spr, u;
> > +	unsigned int opcode, ra, rb, rd, spr, u;
> >   	unsigned long int imm;
> >   	unsigned long int val, val2;
> >   	unsigned int mb, me, sh;
> > -	unsigned int word, suffix;
> > +	unsigned int word;
> >   	long ival;
> >   	word = ppc_inst_val(instr);
> > +#ifdef __powerpc64__
> >   	suffix = ppc_inst_suffix(instr);
> > +#endif
> >   	op->type = COMPUTE;
> > @@ -1469,7 +1473,9 @@ int analyse_instr(struct instruction_op *op, const struct pt_regs *regs,
> >   	rd = (word >> 21) & 0x1f;
> >   	ra = (word >> 16) & 0x1f;
> >   	rb = (word >> 11) & 0x1f;
> > +#ifdef __powerpc64__
> >   	rc = (word >> 6) & 0x1f;
> > +#endif
> >   	switch (opcode) {
> >   #ifdef __powerpc64__
> 

-- 
Link Mauve

Re: [PATCH 0/2] powerpc/sstep: Fix compilation on 32-bit PowerPC

[Christophe Leroy (CS GROUP)]

Le 29/07/2026 à 22:49, Link Mauve a écrit :
> Two local variables were only set but never read on 32-bit PowerPC,
> which causes an error during compilation.

Are those errors new ? This series and the one from yesterday on ppc64, 
I have the feeling that the code is not new. Has anything changed in the 
building flags or does this pops up with recent versions of GCC ? If it 
is the case it is worth saying it in the commit messages in order to 
clarify the reason for the fix.

> 
> This series first makes the 64-bit guards consistent by converting all
> uses of CONFIG_PPC64 to use __powerpc64__ instead, and then guard the
> two variables rc and suffix behind __powerpc64__ checks.
> 
> Link Mauve (2):
>    powerpc/sstep: Consistently use one define to check arch bit width
>    powerpc/sstep: Don’t define variables we won’t be using
> 
>   arch/powerpc/lib/sstep.c | 26 ++++++++++++++++----------
>   1 file changed, 16 insertions(+), 10 deletions(-)
> 
> 
> base-commit: 78bc8af4affb9a732504eb22eeac7d1e50883853

Re: [PATCH 0/2] powerpc/sstep: Fix compilation on 32-bit PowerPC

[Link Mauve]

On Thu, Jul 30, 2026 at 07:51:40AM +0200, Christophe Leroy (CS GROUP) wrote:
> 
> 
> Le 29/07/2026 à 22:49, Link Mauve a écrit :
> > Two local variables were only set but never read on 32-bit PowerPC,
> > which causes an error during compilation.
> 
> Are those errors new ? This series and the one from yesterday on ppc64, I
> have the feeling that the code is not new. Has anything changed in the
> building flags or does this pops up with recent versions of GCC ? If it is
> the case it is worth saying it in the commit messages in order to clarify
> the reason for the fix.

I think they started happening when I enabled Rust support, but I’ve
been doing that for quite a while and I don’t know if the two are
linked, and when I disable CONFIG_RUST it’s still erroring without this
series.

I’m pretty sure someone has built the kernel on 32-bit PowerPC since
650b55b707fdfa764e9f2b81314d3eb4216fb962 or
930d6288a26787d2e7f633705434171a506db9c5, both commits introducing the
two variables which are then never used.

I couldn’t follow the maze of Makefile enough to determine where
-Werror=unused-but-set-variable is set, if this is specific to recent
versions of gcc or not, if this is enabled by a different CONFIG_
option, if this could be disabled already.  But I believe fixing this
issue is good anyway, the compiler is right that without __powerpc64__
these two variables are never used.

> 
> > 
> > This series first makes the 64-bit guards consistent by converting all
> > uses of CONFIG_PPC64 to use __powerpc64__ instead, and then guard the
> > two variables rc and suffix behind __powerpc64__ checks.
> > 
> > Link Mauve (2):
> >    powerpc/sstep: Consistently use one define to check arch bit width
> >    powerpc/sstep: Don’t define variables we won’t be using
> > 
> >   arch/powerpc/lib/sstep.c | 26 ++++++++++++++++----------
> >   1 file changed, 16 insertions(+), 10 deletions(-)
> > 
> > 
> > base-commit: 78bc8af4affb9a732504eb22eeac7d1e50883853
> 

-- 
Link Mauve