| Message ID | 20180411150829.3881900-1-arnd@arndb.de |
|---|---|
| State | New |
| Headers | show |
| Series | sparc: compat: allow including asm/compat.h for 32-bit | expand |
Arnd Bergmann <arnd@arndb.de> writes: > We have several files on sparc that include linux/compat.h and expect > asm/compat.h not to be included, otherwise we get a build failure. Should this say: "We have several files on sparc that include linux/compat.h and expect asm/compat.h not be included when !CONFIG_COMPAT." ? I don't have your tree and I see asm/compat.h included from linux/compat.h already so the description above seems wrong. Eric
On Wed, Apr 11, 2018 at 9:32 PM, Eric W. Biederman <ebiederm@xmission.com> wrote: > Arnd Bergmann <arnd@arndb.de> writes: > >> We have several files on sparc that include linux/compat.h and expect >> asm/compat.h not to be included, otherwise we get a build failure. > > Should this say: "We have several files on sparc that include > linux/compat.h and expect asm/compat.h not be included when > !CONFIG_COMPAT." ? Correct, I'll fix it up. I'll say '32-bit builds' instead, which is what I'm interested in. On sparc64, CONFIG_COMPAT is enabled unconditionally, so it's really the same thing. > I don't have your tree and I see asm/compat.h included from > linux/compat.h already so the description above seems wrong. I removed the tree from linux-next before the merge window because of this build regression, if you are interested in the other changes, please see git://git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git y2038-next Arnd
diff --git a/arch/sparc/include/asm/compat.h b/arch/sparc/include/asm/compat.h index 161aeacbc9ad..4eb51d2dae98 100644 --- a/arch/sparc/include/asm/compat.h +++ b/arch/sparc/include/asm/compat.h @@ -157,6 +157,7 @@ static inline compat_uptr_t ptr_to_compat(void __user *uptr) return (u32)(unsigned long)uptr; } +#ifdef CONFIG_COMPAT static inline void __user *arch_compat_alloc_user_space(long len) { struct pt_regs *regs = current_thread_info()->kregs; @@ -173,6 +174,7 @@ static inline void __user *arch_compat_alloc_user_space(long len) return (void __user *) usp; } +#endif struct compat_ipc64_perm { compat_key_t key; @@ -232,6 +234,7 @@ struct compat_shmid64_ds { unsigned int __unused2; }; +#ifdef CONFIG_COMPAT static inline int is_compat_task(void) { return test_thread_flag(TIF_32BIT); @@ -243,5 +246,6 @@ static inline bool in_compat_syscall(void) return pt_regs_trap_type(current_pt_regs()) == 0x110; } #define in_compat_syscall in_compat_syscall +#endif #endif /* _ASM_SPARC64_COMPAT_H */
We have several files on sparc that include linux/compat.h and expect asm/compat.h not to be included, otherwise we get a build failure. Since we need to include asm/compat.h for compat time_t handling on all 32-bit architectures now, this hides some portions of asm/compat.h in order to let the rest of the file get included. Signed-off-by: Arnd Bergmann <arnd@arndb.de> --- arch/sparc/include/asm/compat.h | 4 ++++ 1 file changed, 4 insertions(+) This is the fix for the y2038 syscall series that missed the 4.17 merge window. I've now spent two more days doing extra build testing on all 64-bit archtectures building with and without CONFIG_COMPAT, as well as without CONFIG_64BIT. I did not run into any additional build failures after this patch, so my plan is to pick that up into my y2038 tree after -rc1 and send Thomas a pull request a few days later for inclusion in linux-4.18. -- 2.9.0