| Message ID | Pine.NEB.4.64.1701241855380.2138@sdf.lonestar.org |
|---|---|
| State | New |
| Headers | show |
On 24.01.2017 19:56, Valentin Plotkin wrote: > > Hi everyone, > > I looked at the "qemu-system-ppc -nographic -cpu G2leGP3 -M ppce500" on > the BiteSizedTasks page. The segfault was caused by machine > initialization code which expected a certain MMU model, checked, so > unused SPR were read, returning zeros. bamboo and virtex machines are > affected as well, but it doesn't always cause segfault, usually running > into unmapped memory and failing somewhat more nicely. > > I added the checks. It would be possible to add support for other MMU > models, but I'm not sure if there is any point (would any guest OS > support mutually exclusive CPU and machine)? Hi, great to have a fix for this crash! I don't think it make much sense to add support for other MMU models here, so the simple checks should be good enough. However, your new code obviously does not follow the QEMU coding style. Could you please run your patch through scripts/checkpatch.pl and fix all issues that it reports? And when you resubmit, please make sure to copy the maintainers on CC: as well (scripts/get_maintainers.pl is your friend here). Thanks, Thomas
diff --git a/hw/ppc/e500.c b/hw/ppc/e500.c index cf8b122..cd352f6 100644 --- a/hw/ppc/e500.c +++ b/hw/ppc/e500.c @@ -631,6 +631,11 @@ static uint64_t mmubooke_initial_mapsize(CPUPPCState *env) static void mmubooke_create_initial_mapping(CPUPPCState *env) { + if(env->mmu_model!=POWERPC_MMU_BOOKE206){ + fprintf(stderr,"MMU model %i not supported by this machine.\n", env->mmu_model); + exit(-1); + } + ppcmas_tlb_t *tlb = booke206_get_tlbm(env, 1, 0, 0); hwaddr size; int ps; diff --git a/hw/ppc/ppc440_bamboo.c b/hw/ppc/ppc440_bamboo.c index 5c535b1..61e7028 100644 --- a/hw/ppc/ppc440_bamboo.c +++ b/hw/ppc/ppc440_bamboo.c @@ -124,6 +124,11 @@ static void mmubooke_create_initial_mapping(CPUPPCState *env, target_ulong va, hwaddr pa) { + if(env->mmu_model!=POWERPC_MMU_BOOKE){ + fprintf(stderr,"MMU model %i not supported by this machine.\n", env->mmu_model); + exit(-1); + } + ppcemb_tlb_t *tlb = &env->tlb.tlbe[0]; tlb->attr = 0; diff --git a/hw/ppc/virtex_ml507.c b/hw/ppc/virtex_ml507.c index b97d966..4041ff3 100644 --- a/hw/ppc/virtex_ml507.c +++ b/hw/ppc/virtex_ml507.c @@ -69,6 +69,11 @@ static void mmubooke_create_initial_mapping(CPUPPCState *env, target_ulong va, hwaddr pa) { + if(env->mmu_model!=POWERPC_MMU_BOOKE){ + fprintf(stderr,"MMU model %i not supported by this machine.\n", env->mmu_model); + exit(-1); + } + ppcemb_tlb_t *tlb = &env->tlb.tlbe[0];
Hi everyone, I looked at the "qemu-system-ppc -nographic -cpu G2leGP3 -M ppce500" on the BiteSizedTasks page. The segfault was caused by machine initialization code which expected a certain MMU model, checked, so unused SPR were read, returning zeros. bamboo and virtex machines are affected as well, but it doesn't always cause segfault, usually running into unmapped memory and failing somewhat more nicely. I added the checks. It would be possible to add support for other MMU models, but I'm not sure if there is any point (would any guest OS support mutually exclusive CPU and machine)? --- hw/ppc/e500.c | 5 +++++ hw/ppc/ppc440_bamboo.c | 5 +++++ hw/ppc/virtex_ml507.c | 5 +++++ 3 files changed, 15 insertions(+) tlb->attr = 0;