mm: don't allow ioremapping of ranges larger than vmalloc space
commit5dc331852848a38ca00a2817e5b98a1d0561b116
authorRobert Bragg <robert@sixbynine.org>
Tue, 5 Feb 2008 06:29:18 +0000 (4 22:29 -0800)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Tue, 5 Feb 2008 17:44:18 +0000 (5 09:44 -0800)
treec306874d0194ecf88c2073a042e6f932e710f67f
parenta7f75e25860ac0a7b70cf6e14c37618d2d2bb890
mm: don't allow ioremapping of ranges larger than vmalloc space

When running with a 16M IOREMAP_MAX_ORDER (on armv7) we found that the
vmlist search routine in __get_vm_area_node can mistakenly allow a driver
to ioremap a range larger than vmalloc space.

If at the time of the ioremap all existing vmlist areas sit below the
determined alignment then the search routine continues past all entries and
exits the for loop - straight into the found: label - without ever testing
for integer wrapping or that the requested size fits.

We were seeing a driver successfully ioremap 128M of flash even though
there was only 120M of vmalloc space.  From that point the system was left
with the remainder of the first 16M of space to vmalloc/ioremap within.

Signed-off-by: Robert Bragg <robert@sixbynine.org>
Acked-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/vmalloc.c