Date: Mon, 26 May 2025 18:51:28 -0700 From: Rick Macklem <rick.macklem@gmail.com> To: Cy Schubert <Cy.Schubert@cschubert.com> Cc: freebsd-current@freebsd.org Subject: Re: mountd(8) Hangs Message-ID: <CAM5tNy48VzJz3zYE9w9VO%2Bs=DuCQrdhkNEe_byLJ0NCPthQiiA@mail.gmail.com> In-Reply-To: <20250526162557.97D6C318@slippy.cwsent.com> References: <20250526162557.97D6C318@slippy.cwsent.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, May 26, 2025 at 9:26=E2=80=AFAM Cy Schubert <Cy.Schubert@cschubert.= com> wrote: > > I've noticed over the past month or so mountd(8) hanging on an NFS server > here when the NFS server is running poudriere. This only occurs when > poudriere is building multiple little ports resulting in many mounts and > unmounts due to jails being constituted and torn down. After a while > mountd(8) ends up in a deadlocked state causing poudriere to seize. I cannot think why poudriere would care about mountd being hung? > The > workaround is to stop mountd before or immediately following initiation o= f > poudriere. Well, there is code in mountd.c that prevents it from reloading exports too often. RELOADDELAY is set to 0.25sec. However, this code has not changed in a long time, so unless clock_gettime() has somehow changed, I cannot see why it would have stopped working? The commits during the last month or so for mountd.c are pretty benign, but you could try reverting one or two of them. (One moves some of the code into libutil, but I cannot think why that might matter?) The only commit I did (on May 1) was 68daa78, which just adds one additiona= l check to the exports processing. You can easily revert this one to see if i= t makes any difference? Can you see where the mountd process is stuck with something like # ps axHl to see if it is hung in the kernel and, if so, what the wchan is? or try sending it a SIGHUP when it is hung to see if that brings it back to life? rick > > This used to be a problem without the -S flag but now suspend/resume of > nfsd threads doesn't mitigate the problem either. > > Something has changed but I cannot point to any individual commit or grou= p > of commits that caused this regression. > > > -- > Cheers, > Cy Schubert <Cy.Schubert@cschubert.com> > FreeBSD UNIX: <cy@FreeBSD.org> Web: https://FreeBSD.org > NTP: <cy@nwtime.org> Web: https://nwtime.org > > e^(i*pi)+1=3D0 > > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAM5tNy48VzJz3zYE9w9VO%2Bs=DuCQrdhkNEe_byLJ0NCPthQiiA>