diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2018-04-05 20:33:38 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2018-04-05 20:33:38 -0700 |
commit | d66db9f6e427db122aeaad0f0cc94291ce6ddb82 (patch) | |
tree | 5b9ce3b73173d9d27b8342f7bbc5a4fbff00534f /include/uapi/asm-generic/siginfo.h | |
parent | 357aa6aefebe888c712152cb83c9e700f98eebd1 (diff) | |
parent | 4be33329d46f80e87afe7db61271d1370607260a (diff) |
Merge branch 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace
Pull siginfo updates from Eric Biederman:
"The work on cleaning up and getting the bugs out of siginfo generation
was largely stalled this round. The progress that was made was the
definition of FPE_FLTUNK. Which is usable to fix many of the cases
where siginfo generation is erroneously generating SI_USER by setting
si_code to 0, that has recently been tagged as FPE_FIXME.
You already have the change by way of the arm64 tree as that
definition was pulled into the arm64 tree to allow fixing the problem
there.
What remains is the second round of fixing for what I thought was a
trivial change to the struct siginfo when put the union in _sigfault
where it belongs. Do to historical reasons 32bit m68k only ensures
that pointers are 2 byte aligned. So I have added a m68k test case
made of BUILD_BUG_ONs to verify I have this fix correct and possibly
catch problems, and I have computed the number of bytes of padding
needed for the _addr_bnd and _addr_pkey cases and just use an array of
characters that size.
For pure paranoia I have written the code so if there is an
architecture out there that does not perform any alignment of
structures it should still work.
With the removal of all of the stale arechitectures this cycle future
work on cleaning up struct siginfo should be much easier. Almost all
of the conflicting si_code definitions have been removed with the
removal of (blackfin, tile, and frv). Plus some of the most difficult
to test cases have simply been removed from the tree.
Which means that with a little luck copy_siginfo_to_user can become a
light weight wrapper around copy_to_user in the next cycle"
* 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
m68k: Verify the offsets in struct siginfo never change.
signal: Correct the offset of si_pkey and si_lower in struct siginfo on m68k
Diffstat (limited to 'include/uapi/asm-generic/siginfo.h')
-rw-r--r-- | include/uapi/asm-generic/siginfo.h | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/include/uapi/asm-generic/siginfo.h b/include/uapi/asm-generic/siginfo.h index 6088bca89917..544208fd3db1 100644 --- a/include/uapi/asm-generic/siginfo.h +++ b/include/uapi/asm-generic/siginfo.h @@ -94,6 +94,9 @@ typedef struct siginfo { unsigned int _flags; /* see ia64 si_flags */ unsigned long _isr; /* isr */ #endif + +#define __ADDR_BND_PKEY_PAD (__alignof__(void *) < sizeof(short) ? \ + sizeof(short) : __alignof__(void *)) union { /* * used when si_code=BUS_MCEERR_AR or @@ -102,13 +105,13 @@ typedef struct siginfo { short _addr_lsb; /* LSB of the reported address */ /* used when si_code=SEGV_BNDERR */ struct { - void *_dummy_bnd; + char _dummy_bnd[__ADDR_BND_PKEY_PAD]; void __user *_lower; void __user *_upper; } _addr_bnd; /* used when si_code=SEGV_PKUERR */ struct { - void *_dummy_pkey; + char _dummy_pkey[__ADDR_BND_PKEY_PAD]; __u32 _pkey; } _addr_pkey; }; |