Hi Jörn, On Wed, 3 Jul 2013 10:48:03 -0400 Jörn Engel <joern@xxxxxxxxx> wrote: > > On Wed, 3 July 2013 12:25:41 +0100, Steve Capper wrote: > > > > Today's linux-next appears to introduce some BUGs when running the libhugetlbfs > > testsuite (under both ARM64 and x86_64): > > > > [ 94.320661] BUG: Bad rss-counter state mm:ffff880119461040 idx:1 val:-512 > > [ 94.330346] BUG: Bad rss-counter state mm:ffff880119460680 idx:1 val:-2560 > > [ 94.341746] BUG: Bad rss-counter state mm:ffff880119460d00 idx:1 val:-512 > > [ 94.347518] BUG: Bad rss-counter state mm:ffff880119460d00 idx:1 val:-512 > > [ 94.415203] BUG: Bad rss-counter state mm:ffff8801194f9040 idx:1 val:-1024 > > > > I've bisected this to commit: > > commit 446313e3e4e7274c4d1e449d107fab6c041ec59c > > hugetlb: properly account rss > > > > The bad counter values I get are all negative. > > > > To reproduce this, just run through the latest libhugetlbfs test suite, these > > warnings appear pretty much immediately. (I set aside 200 huge pages first): > > http://libhugetlbfs.sourceforge.net/ > > > > Reverting this commit allows the test suite to run without problems. > > Thank you, I will have a look. For the interim, dropping this patch > from -mm seems the best approach. OK, I have dropped that from the akpm tree in linux-next today. -- Cheers, Stephen Rothwell sfr@xxxxxxxxxxxxxxxx
Attachment:
pgpMIa0TkiRzg.pgp
Description: PGP signature
