From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sedat Dilek Subject: Re: linux-next: Tree for January 15 (Call Trace in fs/dcache.c + autofs4) Date: Sat, 15 Jan 2011 12:23:57 +0100 Message-ID: References: <20110115075723.GJ19804@ZenIV.linux.org.uk> <20110115090702.GK19804@ZenIV.linux.org.uk> <20110115110848.GL19804@ZenIV.linux.org.uk> Reply-To: sedat.dilek@gmail.com Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Stephen Rothwell , linux-next@vger.kernel.org, LKML , linux-fsdevel To: Al Viro Return-path: In-Reply-To: Sender: linux-next-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Sat, Jan 15, 2011 at 12:16 PM, Sedat Dilek wrote: > On Sat, Jan 15, 2011 at 12:08 PM, Al Viro w= rote: >> On Sat, Jan 15, 2011 at 12:00:54PM +0100, Sedat Dilek wrote: >> >>> > Argh... ??In __do_follow_link() replace >>> > >>> > ?? ?? ?? ??if (link->mnt !=3D nd->path.mnt) >>> > with >>> > ?? ?? ?? ??if (link->mnt =3D=3D nd->path.mnt) >>> > >>> > Mismerge yesterday ;-/ ??I've pushed fix for that in for-next, wi= ll fold >>> > shortly. ??As for autofs4 breakage, I've a preliminary fix, testi= ng it >>> > now. >>> > >>> >>> Hey, cool and thanks. >>> >>> Not sure if you catched them all, I have noticed on my latest >>> ("buggy") linux-next kernel these Call Traces when doing an >>> update-grub. >> >> That might be vfsmount being dropped when it shouldn't or dentry lea= ked. >> And seeing that it's umount(8), I would suspect the latter... >> >> Anyway, with the latest from dhowells we probably should have >> d_set_d_op() mess on autofs4 under control (in #for-next). =C2=A0Whe= ther >> it's enough to actually fix the sucker is a separate question, of >> course - there might very well be more crap. >> >> I've instrumented mntput() et.al. here; hopefully that'll make catch= ing >> the remaining turds easier. =C2=A0As for dcache leaks... ouch. =C2=A0= Could you >> try to reproduce that one on the mainline kernel? =C2=A0At least tha= t'd >> isolate things a bit. >> > > Which version do you mean by "mainline"? 2.6.37-git12? > And with "[mismerge][foldme] breakage in __do_follow_link()" patch [1= ]? > I have to look by pulling-in(to) Linus-tree (linux-next) which patche= s > are not merged already. > > - Sedat - > > [1] http://git.kernel.org/?p=3Dlinux/kernel/git/viro/vfs-2.6.git;a=3D= commit;h=3D2f3740b3a80bc19498b3ec9b86a25d5941545436 > OK, I will try with these two patches and report later. - Sedat - P.S.: sd@tbox:~/src/linux-2.6/linux-2.6.37/debian/patches$ ls -l vfs-2.6-for-= next/ insgesamt 28 -rw-r--r-- 1 sd sd 737 15. Jan 12:21 0001-mismerge-foldme-breakage-in-__do_follow_link.patch -rw-r--r-- 1 sd sd 4446 15. Jan 12:21 0002-autofs4-Merge-the-remaining-dentry-ops-tables.patch -rw-r--r-- 1 sd sd 201 15. Jan 12:21 diffstat -rw-r--r-- 1 sd sd 232 15. Jan 12:21 Shortlog -rw-r--r-- 1 sd sd 4417 15. Jan 12:21 vfs-2.6-for-next.patch