On Wed, 15 Jul 2009, Stephan von Krawczynski wrote:Yes:
Jul 15 03:01:29 backup kernel: 1444268 total pagecache pages
Jul 15 03:01:29 backup kernel: 34 pages in swap cache
Jul 15 03:01:29 backup kernel: Swap cache stats: add 118, delete 84, find 0/2
Jul 15 03:01:29 backup kernel: Free swap = 2104080kB
Jul 15 03:01:29 backup kernel: Total swap = 2104488kB
I added Justin Piszcz to the cc since he was having the same problem as
described in http://bugzilla.kernel.org/show_bug.cgi?id=13648.
He was unable to get slabtop -o output when this was happening, though, so
maybe you could grab a snapshot of that when you get these failures? It
will help us figure out what cache the slab leak is in (assuming there is
one, >1G of slab on this machine is egregious).
Justin, were you using e1000e in your bug report?
If you have some additional time, it would also be helpful to get a
bisection of when the problem started occurring (it appears to be sometime
between 2.6.29 and 2.6.30).