| Message ID | 20211104140536.42573-20-imp@bsdimp.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | bsd-user: arm (32-bit) support | expand |
On 11/4/21 10:05 AM, Warner Losh wrote: > +typedef struct target_mcontext { > + uint32_t __gregs[32]; sys/arm/include/ucontext.h has #define _NGREG 17 typedef __greg_t __gregset_t[_NGREG]; With s/32/17/ here, Reviewed-by: Richard Henderson <richard.henderson@linaro.org> r~
On Thu, Nov 4, 2021 at 11:49 AM Richard Henderson < richard.henderson@linaro.org> wrote: > On 11/4/21 10:05 AM, Warner Losh wrote: > > +typedef struct target_mcontext { > > + uint32_t __gregs[32]; > > sys/arm/include/ucontext.h has > > #define _NGREG 17 > typedef __greg_t __gregset_t[_NGREG]; > > With s/32/17/ here, > Reviewed-by: Richard Henderson <richard.henderson@linaro.org> > Will do. And I'm going to add compile time asserts for the size of mcontext and ucontext in a future patch series. It won't test for layout, but will keep things like this from happening again by accident. Warner
On Thu, Nov 4, 2021, 11:58 AM Warner Losh <imp@bsdimp.com> wrote: > > > On Thu, Nov 4, 2021 at 11:49 AM Richard Henderson < > richard.henderson@linaro.org> wrote: > >> On 11/4/21 10:05 AM, Warner Losh wrote: >> > +typedef struct target_mcontext { >> > + uint32_t __gregs[32]; >> >> sys/arm/include/ucontext.h has >> >> #define _NGREG 17 >> typedef __greg_t __gregset_t[_NGREG]; >> > There is already a TARGET__NGREG so I'll use that. With s/32/17/ here, >> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> >> > > Will do. > > And I'm going to add compile time asserts for the size of mcontext and > ucontext in a future > patch series. It won't test for layout, but will keep things like this > from happening again by accident. > > Warner >
On Thu, Nov 4, 2021 at 11:58 AM Warner Losh <imp@bsdimp.com> wrote: > > > On Thu, Nov 4, 2021 at 11:49 AM Richard Henderson < > richard.henderson@linaro.org> wrote: > >> On 11/4/21 10:05 AM, Warner Losh wrote: >> > +typedef struct target_mcontext { >> > + uint32_t __gregs[32]; >> >> sys/arm/include/ucontext.h has >> >> #define _NGREG 17 >> typedef __greg_t __gregset_t[_NGREG]; >> >> With s/32/17/ here, >> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> >> > > Will do. > > And I'm going to add compile time asserts for the size of mcontext and > ucontext in a future > patch series. It won't test for layout, but will keep things like this > from happening again by accident. > So I added a G_STATIC_ASSERT and found I used a host's pointer instead of the target's pointer for the vfp stuff. Easy enough to fix... But it means using functions that aren't yet defined when this is included, which means I'll have to restructure things... shuffling the includes is likely out of the question, so I'll have to create an arch/signal.c to hold the now-inlined functions. All because I wanted to catch errors like this before I made them again in public... :) So thanks for inspiring me to do something that detected the error, but a good natured harumph at the work I have to do now... Warner
diff --git a/bsd-user/arm/target_arch_signal.h b/bsd-user/arm/target_arch_signal.h index 973183d99c..3aaced474b 100644 --- a/bsd-user/arm/target_arch_signal.h +++ b/bsd-user/arm/target_arch_signal.h @@ -54,4 +54,24 @@ #define TARGET_MINSIGSTKSZ (1024 * 4) /* min sig stack size */ #define TARGET_SIGSTKSZ (TARGET_MINSIGSTKSZ + 32768) /* recommended size */ +/* + * Floating point register state + */ +typedef struct target_mcontext_vfp { + abi_ullong mcv_reg[32]; + abi_ulong mcv_fpscr; +} target_mcontext_vfp_t; + +typedef struct target_mcontext { + uint32_t __gregs[32]; + + /* + * Originally, rest of this structure was named __fpu, 35 * 4 bytes + * long, never accessed from kernel. + */ + abi_long mc_vfp_size; + abi_ptr *mc_vfp_ptr; + abi_int mc_spare[33]; +} target_mcontext_t; + #endif /* !_TARGET_ARCH_SIGNAL_H_ */