summaryrefslogtreecommitdiffstats
path: root/ipc/shm.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-01-30 11:37:22 -0800
committerLinus Torvalds <torvalds@linux-foundation.org>2009-01-30 11:37:22 -0800
commit33bfad54b58cf05cfe6678c3ec9235d4bc8db4c2 (patch)
treed6be0ff97effa23289b2566acf8e56a49aa5657e /ipc/shm.c
parentc01a25e7cf6dcb0fa69c155706d5dd1e76e53796 (diff)
downloadkernel-crypto-33bfad54b58cf05cfe6678c3ec9235d4bc8db4c2.tar.gz
kernel-crypto-33bfad54b58cf05cfe6678c3ec9235d4bc8db4c2.tar.xz
kernel-crypto-33bfad54b58cf05cfe6678c3ec9235d4bc8db4c2.zip
Allow opportunistic merging of VM_CAN_NONLINEAR areas
Commit de33c8db5910cda599899dd431cc30d7c1018cbf ("Fix OOPS in mmap_region() when merging adjacent VM_LOCKED file segments") unified the vma merging of anonymous and file maps to just one place, which simplified the code and fixed a use-after-free bug that could cause an oops. But by doing the merge opportunistically before even having called ->mmap() on the file method, it now compares two different 'vm_flags' values: the pre-mmap() value of the new not-yet-formed vma, and previous mappings of the same file around it. And in doing so, it refused to merge the common file case, which adds a marker to say "I can be made non-linear". This fixes it by just adding a set of flags that don't have to match, because we know they are ok to merge. Currently it's only that single VM_CAN_NONLINEAR flag, but at least conceptually there could be others in the future. Reported-and-acked-by: Hugh Dickins <hugh@veritas.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Greg KH <gregkh@suse.de> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'ipc/shm.c')
0 files changed, 0 insertions, 0 deletions