]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
writeback: fix possible underflow in write bandwidth calculation
authorTejun Heo <tj@kernel.org>
Mon, 23 Mar 2015 04:18:48 +0000 (00:18 -0400)
committerJiri Slaby <jslaby@suse.cz>
Wed, 22 Apr 2015 06:58:35 +0000 (08:58 +0200)
commitc75b0a8bac60682488393cddbb46959d37c1816e
tree0b7ee78d47bc9fc91b55a90043adbec6da893f5d
parent0d4958ffdbf7a14e58ea822968a40c05b3c9de0f
writeback: fix possible underflow in write bandwidth calculation

commit c72efb658f7c8b27ca3d0efb5cfd5ded9fcac89e upstream.

From 1ebf33901ecc75d9496862dceb1ef0377980587c Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Mon, 23 Mar 2015 00:08:19 -0400

2f800fbd777b ("writeback: fix dirtied pages accounting on redirty")
introduced account_page_redirty() which reverts stat updates for a
redirtied page, making BDI_DIRTIED no longer monotonically increasing.

bdi_update_write_bandwidth() uses the delta in BDI_DIRTIED as the
basis for bandwidth calculation.  While unlikely, since the above
patch, the newer value may be lower than the recorded past value and
underflow the bandwidth calculation leading to a wild result.

Fix it by subtracing min of the old and new values when calculating
delta.  AFAIK, there hasn't been any report of it happening but the
resulting erratic behavior would be non-critical and temporary, so
it's possible that the issue is happening without being reported.  The
risk of the fix is very low, so tagged for -stable.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Fixes: 2f800fbd777b ("writeback: fix dirtied pages accounting on redirty")
Signed-off-by: Jens Axboe <axboe@fb.com>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
mm/page-writeback.c