i386: improve and correct inline asm memory constraints
commitb862f3b099f3ea672c7438c0b282ce8201d39dfc
authorLinus Torvalds <torvalds@macmini.osdl.org>
Sat, 8 Jul 2006 22:24:18 +0000 (8 15:24 -0700)
committerLinus Torvalds <torvalds@macmini.osdl.org>
Sat, 8 Jul 2006 22:24:18 +0000 (8 15:24 -0700)
tree62f8cc2dc2b1c9abb6364b16f3b218a04d121f3e
parente2a3d40258fe20d205f8ed592e1e2c0d5529c2e1
i386: improve and correct inline asm memory constraints

Use "+m" rather than a combination of "=m" and "m" for improved clarity
and consistency.

This also fixes some inlines that incorrectly didn't tell the compiler
that they read the old value at all, potentially causing the compiler to
generate bogus code.  It appear that all of those potential bugs were
hidden by the use of extra "volatile" specifiers on the data structures
in question, though.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
include/asm-i386/atomic.h
include/asm-i386/futex.h
include/asm-i386/local.h
include/asm-i386/posix_types.h
include/asm-i386/rwlock.h
include/asm-i386/rwsem.h
include/asm-i386/semaphore.h
include/asm-i386/spinlock.h