In rev 1.72 a situation related to write/mmap was fixed which could result

in a user process gaining visibility into the 'old' contents of a filesystem
block.  There were two cases:  (1) when uiomove() fails (user process issues
illegal write), and (2) when uiomove() overlaps a mmap() of the same file at
the same offset (fault -> recursive buffer I/O reads contents of old block).

Unfortunately 1.72 also had the unintended effect of forcing the filesystem
to do a read-before-write in the case of a full-block-write (non append case),
e.g. 'dd if=/dev/zero of=test.dat bs=1m count=256 conv=notrunc'.  This
destroys performance.. not only is a read forced for every write, but
clustering breaks as well.

The solution is to clear the buffer manually in the full-block case rather
then asking BALLOC to do it (BALLOC issues the read-before-write).  In the
partial-block case we want BALLOC to do it because the read-before-write
is necessary.  This patch should greatly improve database and news-feed
server performance.

Found by: MKI <mki@mozone.net>
MFC after:	3 days

0 files changed, 0 insertions, 0 deletions