Hello,
This patchset fixes two long standing issues and one relatively new
css ref leak.
a. elvpriv alloc failure, including ioc and icq failures, fails
request allocation. As those aren't mempool backed and may fail
indefinitely, this can lead to deadlock under memory pressure.
b. blkgs don't have proper indexing. With enough number of
request_queues and blk-throttle enabled, block layer can spend
considerable amount of cpu cycles walking the same list over and
over again.
c. __blkg_lookup_create() was leaking a css ref on failure path.
This patchset contains the following four patches.
0001-block-collapse-blk_alloc_request-into-get_request.patch
0002-block-fix-elvpriv-allocation-failure-handling.patch
0003-blkcg-fix-blkcg-css-ref-leak-in-__blkg_lookup_create.patch
0004-blkcg-use-radix-tree-to-index-blkgs-from-blkcg.patch
0001-0002 fix #a. 0003 fixes #c. 0004 fixes #b.
This patchset is on top of
block/for-3.5/core 5bc4afb1ec "blkcg: drop BLKCG_STAT_{PRIV|POL|OFF} macros"
+ [1] [PATCHSET] block: per-queue policy activation, take#2
+ [2] [PATCHSET] block: cosmetic updates to blkcg API
and available in the following git branch.
git://git.kernel.org/pub/scm/linux/kernel/git/tj/misc.git blk-misc-fixes
diffstat follows.
block/blk-cgroup.c | 71 ++++++++++++++++++++++++++++++++-----------
block/blk-cgroup.h | 6 +++
block/blk-core.c | 87 ++++++++++++++++++++++++++++-------------------------
3 files changed, 106 insertions(+), 58 deletions(-)
Thanks.
--
tejun
[1] https://lkml.org/lkml/2012/4/13/380
[2] http://www.spinics.net/lists/cgroups/msg01748.html
--
To unsubscribe from this list: send the line "unsubscribe cgroups" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
