| Message ID | 20170302193300.21188-1-brogers@suse.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Am 02.03.2017 um 20:33 schrieb Bruce Rogers: > Currently qemu-s390x segfaults. Avoid cpu model code which doesn't > apply in linux user case. That was also my impression, but it does apply. However, for now it doesn't matter, as also tcg isn't properly wired up. In general, linux-user can support _at least_ the same features as tcg, in theory even more (all that only apply to kernel space). We had a proper fix on this list by Richard Henderson, wonder what happened to that? This file should contain no CONFIG_USER_ONLY apart from the QMP interface stuff. > --- > target/s390x/cpu_models.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c > index 2a894eec65..97e4e75f68 100644 > --- a/target/s390x/cpu_models.c > +++ b/target/s390x/cpu_models.c > @@ -590,7 +590,6 @@ CpuModelBaselineInfo *arch_query_cpu_model_baseline(CpuModelInfo *infoa, > cpu_info_from_model(baseline_info->model, &model, true); > return baseline_info; > } > -#endif > > static void check_consistency(const S390CPUModel *model) > { > @@ -660,7 +659,6 @@ static void check_compatibility(const S390CPUModel *max_model, > > static S390CPUModel *get_max_cpu_model(Error **errp) > { > -#ifndef CONFIG_USER_ONLY > static S390CPUModel max_model; > static bool cached; > > @@ -680,9 +678,9 @@ static S390CPUModel *get_max_cpu_model(Error **errp) > cached = true; > return &max_model; > } > -#endif > return NULL; > } > +#endif > > static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) > { > @@ -718,6 +716,7 @@ static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) > > void s390_realize_cpu_model(CPUState *cs, Error **errp) > { > +#ifndef CONFIG_USER_ONLY > S390CPUClass *xcc = S390_CPU_GET_CLASS(cs); > S390CPU *cpu = S390_CPU(cs); > const S390CPUModel *max_model; > @@ -751,6 +750,7 @@ void s390_realize_cpu_model(CPUState *cs, Error **errp) > } > > apply_cpu_model(cpu->model, errp); > +#endif > } > > static void get_feature(Object *obj, Visitor *v, const char *name, >
Am 02.03.2017 um 20:50 schrieb David Hildenbrand: > Am 02.03.2017 um 20:33 schrieb Bruce Rogers: >> Currently qemu-s390x segfaults. Avoid cpu model code which doesn't >> apply in linux user case. > > That was also my impression, but it does apply. However, for now it > doesn't matter, as also tcg isn't properly wired up. > > In general, linux-user can support _at least_ the same features as tcg, > in theory even more (all that only apply to kernel space). > > We had a proper fix on this list by Richard Henderson, wonder what > happened to that? Actually it was Stefan Weil :) cc-ing him. > > This file should contain no CONFIG_USER_ONLY apart from the QMP > interface stuff. > >> --- >> target/s390x/cpu_models.c | 6 +++--- >> 1 file changed, 3 insertions(+), 3 deletions(-) >> >> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c >> index 2a894eec65..97e4e75f68 100644 >> --- a/target/s390x/cpu_models.c >> +++ b/target/s390x/cpu_models.c >> @@ -590,7 +590,6 @@ CpuModelBaselineInfo *arch_query_cpu_model_baseline(CpuModelInfo *infoa, >> cpu_info_from_model(baseline_info->model, &model, true); >> return baseline_info; >> } >> -#endif >> >> static void check_consistency(const S390CPUModel *model) >> { >> @@ -660,7 +659,6 @@ static void check_compatibility(const S390CPUModel *max_model, >> >> static S390CPUModel *get_max_cpu_model(Error **errp) >> { >> -#ifndef CONFIG_USER_ONLY >> static S390CPUModel max_model; >> static bool cached; >> >> @@ -680,9 +678,9 @@ static S390CPUModel *get_max_cpu_model(Error **errp) >> cached = true; >> return &max_model; >> } >> -#endif >> return NULL; >> } >> +#endif >> >> static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) >> { >> @@ -718,6 +716,7 @@ static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) >> >> void s390_realize_cpu_model(CPUState *cs, Error **errp) >> { >> +#ifndef CONFIG_USER_ONLY >> S390CPUClass *xcc = S390_CPU_GET_CLASS(cs); >> S390CPU *cpu = S390_CPU(cs); >> const S390CPUModel *max_model; >> @@ -751,6 +750,7 @@ void s390_realize_cpu_model(CPUState *cs, Error **errp) >> } >> >> apply_cpu_model(cpu->model, errp); >> +#endif >> } >> >> static void get_feature(Object *obj, Visitor *v, const char *name, >> > >
>>> On 3/2/2017 at 12:53 PM, <david@redhat.com> wrote: > Am 02.03.2017 um 20:50 schrieb David Hildenbrand: >> Am 02.03.2017 um 20:33 schrieb Bruce Rogers: >>> Currently qemu-s390x segfaults. Avoid cpu model code which doesn't >>> apply in linux user case. >> >> That was also my impression, but it does apply. However, for now it >> doesn't matter, as also tcg isn't properly wired up. >> >> In general, linux-user can support _at least_ the same features as tcg, >> in theory even more (all that only apply to kernel space). >> >> We had a proper fix on this list by Richard Henderson, wonder what >> happened to that? > > Actually it was Stefan Weil :) cc-ing him. I'm not all so familiar with this code, I only saw a failure and fixed it the most obvious way for me. I'm happy if there is a better fix out there. BTW: I resent the patch again with SOB. Brue > >> >> This file should contain no CONFIG_USER_ONLY apart from the QMP >> interface stuff. >> >>> --- >>> target/s390x/cpu_models.c | 6 +++--- >>> 1 file changed, 3 insertions(+), 3 deletions(-) >>> >>> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c >>> index 2a894eec65..97e4e75f68 100644 >>> --- a/target/s390x/cpu_models.c >>> +++ b/target/s390x/cpu_models.c >>> @@ -590,7 +590,6 @@ CpuModelBaselineInfo > *arch_query_cpu_model_baseline(CpuModelInfo *infoa, >>> cpu_info_from_model(baseline_info->model, &model, true); >>> return baseline_info; >>> } >>> -#endif >>> >>> static void check_consistency(const S390CPUModel *model) >>> { >>> @@ -660,7 +659,6 @@ static void check_compatibility(const S390CPUModel > *max_model, >>> >>> static S390CPUModel *get_max_cpu_model(Error **errp) >>> { >>> -#ifndef CONFIG_USER_ONLY >>> static S390CPUModel max_model; >>> static bool cached; >>> >>> @@ -680,9 +678,9 @@ static S390CPUModel *get_max_cpu_model(Error **errp) >>> cached = true; >>> return &max_model; >>> } >>> -#endif >>> return NULL; >>> } >>> +#endif >>> >>> static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) >>> { >>> @@ -718,6 +716,7 @@ static inline void apply_cpu_model(const S390CPUModel > *model, Error **errp) >>> >>> void s390_realize_cpu_model(CPUState *cs, Error **errp) >>> { >>> +#ifndef CONFIG_USER_ONLY >>> S390CPUClass *xcc = S390_CPU_GET_CLASS(cs); >>> S390CPU *cpu = S390_CPU(cs); >>> const S390CPUModel *max_model; >>> @@ -751,6 +750,7 @@ void s390_realize_cpu_model(CPUState *cs, Error **errp) >>> } >>> >>> apply_cpu_model(cpu->model, errp); >>> +#endif >>> } >>> >>> static void get_feature(Object *obj, Visitor *v, const char *name, >>> >> >> > > > -- > Thanks, > > David
Am 02.03.2017 um 20:59 schrieb Bruce Rogers: >>>> On 3/2/2017 at 12:53 PM, <david@redhat.com> wrote: >> Am 02.03.2017 um 20:50 schrieb David Hildenbrand: >>> Am 02.03.2017 um 20:33 schrieb Bruce Rogers: >>>> Currently qemu-s390x segfaults. Avoid cpu model code which doesn't >>>> apply in linux user case. >>> >>> That was also my impression, but it does apply. However, for now it >>> doesn't matter, as also tcg isn't properly wired up. >>> >>> In general, linux-user can support _at least_ the same features as tcg, >>> in theory even more (all that only apply to kernel space). >>> >>> We had a proper fix on this list by Richard Henderson, wonder what >>> happened to that? >> >> Actually it was Stefan Weil :) cc-ing him. > > I'm not all so familiar with this code, I only saw a failure and fixed it the > most obvious way for me. I'm happy if there is a better fix out there. Thanks for taking care, seems like the other patch somehow got stuck, so thanks for bringing that up! :) In my original series (when I thought linux-user wouldn't care about CPU models), your patch would have been the correct fix! > > BTW: I resent the patch again with SOB. > > Brue >
diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c index 2a894eec65..97e4e75f68 100644 --- a/target/s390x/cpu_models.c +++ b/target/s390x/cpu_models.c @@ -590,7 +590,6 @@ CpuModelBaselineInfo *arch_query_cpu_model_baseline(CpuModelInfo *infoa, cpu_info_from_model(baseline_info->model, &model, true); return baseline_info; } -#endif static void check_consistency(const S390CPUModel *model) { @@ -660,7 +659,6 @@ static void check_compatibility(const S390CPUModel *max_model, static S390CPUModel *get_max_cpu_model(Error **errp) { -#ifndef CONFIG_USER_ONLY static S390CPUModel max_model; static bool cached; @@ -680,9 +678,9 @@ static S390CPUModel *get_max_cpu_model(Error **errp) cached = true; return &max_model; } -#endif return NULL; } +#endif static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) { @@ -718,6 +716,7 @@ static inline void apply_cpu_model(const S390CPUModel *model, Error **errp) void s390_realize_cpu_model(CPUState *cs, Error **errp) { +#ifndef CONFIG_USER_ONLY S390CPUClass *xcc = S390_CPU_GET_CLASS(cs); S390CPU *cpu = S390_CPU(cs); const S390CPUModel *max_model; @@ -751,6 +750,7 @@ void s390_realize_cpu_model(CPUState *cs, Error **errp) } apply_cpu_model(cpu->model, errp); +#endif } static void get_feature(Object *obj, Visitor *v, const char *name,