| Message ID | aa81586e-c71f-4b33-b45e-2e106e120034@gmail.com |
|---|---|
| State | New |
| Headers | show |
On Thu, Nov 10, 2016 at 12:42:37PM +1100, Balbir Singh wrote: > > > As per the ISA we need a cause for FU exceptions.Executing a tabort r9 > for example in libc, causes a EXCP_FU exception. We don't wire up the > IC (cause) when we post the exception. The cause is required > for the kernel to do the right thing. I caught this issue while testing > the latest kernel against Cedrics' latest pnv ipmi branch. > > Signed-off-by: Balbir singh <bsingharora@gmail.com> Applied to ppc-for-2.8 > --- > target-ppc/excp_helper.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/target-ppc/excp_helper.c b/target-ppc/excp_helper.c > index 808760b..cccea8d 100644 > --- a/target-ppc/excp_helper.c > +++ b/target-ppc/excp_helper.c > @@ -427,6 +427,7 @@ static inline void powerpc_excp(PowerPCCPU *cpu, int excp_model, int excp) > case POWERPC_EXCP_VPU: /* Vector unavailable exception */ > case POWERPC_EXCP_VSXU: /* VSX unavailable exception */ > case POWERPC_EXCP_FU: /* Facility unavailable exception */ > + env->spr[SPR_FSCR] |= ((target_ulong)env->error_code << 56); > break; > case POWERPC_EXCP_PIT: /* Programmable interval timer interrupt */ > LOG_EXCP("PIT exception\n");
On Thu, Nov 10, 2016 at 01:06:17PM +1100, David Gibson wrote: > On Thu, Nov 10, 2016 at 12:42:37PM +1100, Balbir Singh wrote: > > > > > > As per the ISA we need a cause for FU exceptions.Executing a tabort r9 > > for example in libc, causes a EXCP_FU exception. We don't wire up the > > IC (cause) when we post the exception. The cause is required > > for the kernel to do the right thing. I caught this issue while testing > > the latest kernel against Cedrics' latest pnv ipmi branch. > > > > Signed-off-by: Balbir singh <bsingharora@gmail.com> > > Applied to ppc-for-2.8 Or not.. This causes compile breakage for the ppc32 targets. > > > --- > > target-ppc/excp_helper.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/target-ppc/excp_helper.c b/target-ppc/excp_helper.c > > index 808760b..cccea8d 100644 > > --- a/target-ppc/excp_helper.c > > +++ b/target-ppc/excp_helper.c > > @@ -427,6 +427,7 @@ static inline void powerpc_excp(PowerPCCPU *cpu, int excp_model, int excp) > > case POWERPC_EXCP_VPU: /* Vector unavailable exception */ > > case POWERPC_EXCP_VSXU: /* VSX unavailable exception */ > > case POWERPC_EXCP_FU: /* Facility unavailable exception */ > > + env->spr[SPR_FSCR] |= ((target_ulong)env->error_code << 56); > > break; > > case POWERPC_EXCP_PIT: /* Programmable interval timer interrupt */ > > LOG_EXCP("PIT exception\n"); >
On 10/11/16 13:46, David Gibson wrote: > On Thu, Nov 10, 2016 at 01:06:17PM +1100, David Gibson wrote: >> On Thu, Nov 10, 2016 at 12:42:37PM +1100, Balbir Singh wrote: >>> >>> >>> As per the ISA we need a cause for FU exceptions.Executing a tabort r9 >>> for example in libc, causes a EXCP_FU exception. We don't wire up the >>> IC (cause) when we post the exception. The cause is required >>> for the kernel to do the right thing. I caught this issue while testing >>> the latest kernel against Cedrics' latest pnv ipmi branch. >>> >>> Signed-off-by: Balbir singh <bsingharora@gmail.com> >> >> Applied to ppc-for-2.8 > > Or not.. > > This causes compile breakage for the ppc32 targets. > Sorry, I'll double check, I did a minimal compile. I'll redo and repost Balbir Singh.
diff --git a/target-ppc/excp_helper.c b/target-ppc/excp_helper.c index 808760b..cccea8d 100644 --- a/target-ppc/excp_helper.c +++ b/target-ppc/excp_helper.c @@ -427,6 +427,7 @@ static inline void powerpc_excp(PowerPCCPU *cpu, int excp_model, int excp) case POWERPC_EXCP_VPU: /* Vector unavailable exception */ case POWERPC_EXCP_VSXU: /* VSX unavailable exception */ case POWERPC_EXCP_FU: /* Facility unavailable exception */ + env->spr[SPR_FSCR] |= ((target_ulong)env->error_code << 56); break; case POWERPC_EXCP_PIT: /* Programmable interval timer interrupt */ LOG_EXCP("PIT exception\n");
As per the ISA we need a cause for FU exceptions.Executing a tabort r9 for example in libc, causes a EXCP_FU exception. We don't wire up the IC (cause) when we post the exception. The cause is required for the kernel to do the right thing. I caught this issue while testing the latest kernel against Cedrics' latest pnv ipmi branch. Signed-off-by: Balbir singh <bsingharora@gmail.com> --- target-ppc/excp_helper.c | 1 + 1 file changed, 1 insertion(+)