diff mbox

target-ppc: Fix SRR0 when taking unaligned exceptions

Message ID 20150702144406.40bce749@kryten
State New
Headers show

Commit Message

Anton Blanchard July 2, 2015, 4:44 a.m. UTC
We are setting SRR0 to the instruction before the one causing the
unaligned exception. A quick testcase:

. = 0x100
.globl _start
_start:
	/* Cause a 0x600 */
	li	3,0x1
	stwcx.	3,0,3
1:	b	1b

. = 0x600
1:	b	1b

Built into something we can load as a BIOS image:

gcc -mbig -c test.S
ld -EB -Ttext 0x0 -o test test.o
objcopy -O binary test test.bin

Run with:

qemu-system-ppc64 -nographic -bios test.bin

Shows an incorrect SRR0 (points at the li):

SRR0 0000000000000100

With the patch we get the correct SRR0:

SRR0 0000000000000104

Signed-off-by: Anton Blanchard <anton@samba.org>
---
 linux-user/main.c        | 2 +-
 target-ppc/excp_helper.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Comments

Alexander Graf Aug. 26, 2015, 9:44 a.m. UTC | #1
On 02.07.15 06:44, Anton Blanchard wrote:
> We are setting SRR0 to the instruction before the one causing the
> unaligned exception. A quick testcase:
> 
> . = 0x100
> .globl _start
> _start:
> 	/* Cause a 0x600 */
> 	li	3,0x1
> 	stwcx.	3,0,3
> 1:	b	1b
> 
> . = 0x600
> 1:	b	1b
> 
> Built into something we can load as a BIOS image:
> 
> gcc -mbig -c test.S
> ld -EB -Ttext 0x0 -o test test.o
> objcopy -O binary test test.bin
> 
> Run with:
> 
> qemu-system-ppc64 -nographic -bios test.bin
> 
> Shows an incorrect SRR0 (points at the li):
> 
> SRR0 0000000000000100
> 
> With the patch we get the correct SRR0:
> 
> SRR0 0000000000000104
> 
> Signed-off-by: Anton Blanchard <anton@samba.org>

Thanks, applied to ppc-next.


Alex
Benjamin Herrenschmidt Sept. 8, 2015, 10:21 p.m. UTC | #2
On Thu, 2015-07-02 at 14:44 +1000, Anton Blanchard wrote:
> We are setting SRR0 to the instruction before the one causing the
> unaligned exception. A quick testcase:
> 
 ../..

> p_helper.c b/target-ppc/excp_helper.c
> index b803475..4250106 100644
> --- a/target-ppc/excp_helper.c
> +++ b/target-ppc/excp_helper.c
> @@ -200,7 +200,7 @@ static inline void powerpc_excp(PowerPCCPU *cpu,
> int excp_model, int excp)
>          /* Get rS/rD and rA from faulting opcode */
>          env->spr[SPR_DSISR] |= (cpu_ldl_code(env, (env->nip - 4))
>                                  & 0x03FF0000) >> 16;.

You need to also fix the above to use env->nip instead of env->nip - 4
when generating DSISR

> -        goto store_current;
> +        goto store_next;
>      case POWERPC_EXCP_PROGRAM:   /* Program exception               
>          */
>          switch (env->error_code & ~0xF) {
>          case POWERPC_EXCP_FP:
diff mbox

Patch

diff --git a/linux-user/main.c b/linux-user/main.c
index c855bcc..9100130 100644
--- a/linux-user/main.c
+++ b/linux-user/main.c
@@ -1650,7 +1650,7 @@  void cpu_loop(CPUPPCState *env)
             info.si_signo = TARGET_SIGBUS;
             info.si_errno = 0;
             info.si_code = TARGET_BUS_ADRALN;
-            info._sifields._sigfault._addr = env->nip - 4;
+            info._sifields._sigfault._addr = env->nip;
             queue_signal(env, info.si_signo, &info);
             break;
         case POWERPC_EXCP_PROGRAM:  /* Program exception                     */
diff --git a/target-ppc/excp_helper.c b/target-ppc/excp_helper.c
index b803475..4250106 100644
--- a/target-ppc/excp_helper.c
+++ b/target-ppc/excp_helper.c
@@ -200,7 +200,7 @@  static inline void powerpc_excp(PowerPCCPU *cpu, int excp_model, int excp)
         /* Get rS/rD and rA from faulting opcode */
         env->spr[SPR_DSISR] |= (cpu_ldl_code(env, (env->nip - 4))
                                 & 0x03FF0000) >> 16;
-        goto store_current;
+        goto store_next;
     case POWERPC_EXCP_PROGRAM:   /* Program exception                        */
         switch (env->error_code & ~0xF) {
         case POWERPC_EXCP_FP: