Whamcloud - gitweb
git://git.whamcloud.com / fs / lustre-release.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 3129537) | patch
b=21983 Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
authorDmitry Zogin <dmitry.zogin@sun.com>
Tue, 9 Mar 2010 14:48:23 +0000 (09:48 -0500)
committerJohann Lombardi <johann@sun.com>
Wed, 10 Mar 2010 09:40:08 +0000 (10:40 +0100)
commitb303e9e897f048414f87c5289858570399113aec
tree52fd0871c4255956f66db38986f688f5b5590f8ftree | snapshot
parent3129537c7f5da328abdc1cbd9cf5419ff8e2530ccommit | diff
b=21983 Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh

During an ll_readahead under ll_readpage, we have seen the the
OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
a page that is already locked by the readahead code.

Using CFS_ALLOC_IO instead of CFS_ALLOC_STD will prevent
ldlm_pools_shrink from actually freeing slab, so the call path
that blocks indefinitely can never happen.

 i=adilger
 i=dmitry.zogin
 i=johann
lustre/llite/rw.c diff | blob | history
Lustre primary development and releases