Date: Mon, 23 Feb 2015 03:22:58 +0200 From: Konstantin Belousov <kostikbel@gmail.com> To: sbruno@freebsd.org Cc: freebsd-current@freebsd.org Subject: Re: panic on application core dump? Message-ID: <20150223012257.GO74514@kib.kiev.ua> In-Reply-To: <54EA5F89.1010102@ignoranthack.me> References: <54E8EA2A.7020904@ignoranthack.me> <20150221211712.GG74514@kib.kiev.ua> <54EA1325.6070009@ignoranthack.me> <20150222180425.GJ74514@kib.kiev.ua> <54EA241D.6020606@ignoranthack.me> <20150222185352.GL74514@kib.kiev.ua> <54EA25FE.60401@ignoranthack.me> <54EA5F89.1010102@ignoranthack.me>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Feb 22, 2015 at 03:00:25PM -0800, Sean Bruno wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA512 > > On 02/22/15 10:54, Sean Bruno wrote: > > On 02/22/15 10:53, Konstantin Belousov wrote: > >> On Sun, Feb 22, 2015 at 10:46:53AM -0800, Sean Bruno wrote: > >>> Hmm ... looks unrelated to signals (maybe). This looks like a > >>> common ZFS deadlock that is yet undiagnosed. I do not have a > >>> show alllocks command available in db> . I will show each > >>> lock information below: > >> Add witness. > > > >>> > >>> db> show lockedvnods Locked vnodes > >>> > >>> 0xfffff801141a6588: tag zfs, type VDIR usecount 19, writecount > >>> 0, refcount 20 mountedhere 0 flags (VV_ROOT|VI_ACTIVE) > >>> v_object 0xfffff80079be4500 ref 0 pages 0 cleanbuf 0 dirtybuf 0 > >>> lock type zfs: EXCL by thread 0xfffff801ca10c4a0 (pid 75907, > >>> sh, tid 101262) with exclusive waiters pending > >> Without backtraces of the acquisition, it is not useful. You > >> need DEBUG_VFS_LOCKS for this. > > > > > > > > Thank you. I will do so and restart my non-determinstic test and > > see what I can find. > > > > sean _______________________ > > Well, that was certainly enlightening. I was able to get a WITNESS > panic in imgact_binmisc.c in an hour or two. I need to *not* hold the > mtx protecting the list of activators over the bcopy in > imgact_binmisc_exec(). > > Jiles proposes that we switch to an sx lock here for simplicity of > change of the code. > Kernel page fault with the following non-sleepable locks held: > exclusive sleep mutex imgact_binmisc (imgact_binmisc) r = 0 > (0xffffffff82012418) locked @ > /usr/src/sys/modules/imgact_binmisc/../../kern/imgact_binmisc.c:596 So was it the reason for your troubles after the patch for handling core file' vnode refcount properly ? Or is there something more left ?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150223012257.GO74514>