Re: -CURRENT panics in NFS

2021-02-27 Thread Mateusz Guzik
Thanks. I adjusted the namecache. However, the nfs fix provided by Rick should go in regardless. On 2/27/21, Juraj Lutter wrote: > > >> On 27 Feb 2021, at 21:49, Mateusz Guzik wrote: >> >> Can you dump 'struct componentname *cnp'? This should do the trick: >> f 12 >> p cnp >> >> Most notably I w

Re: -CURRENT panics in NFS

2021-02-27 Thread Juraj Lutter
> On 27 Feb 2021, at 21:49, Mateusz Guzik wrote: > > Can you dump 'struct componentname *cnp'? This should do the trick: > f 12 > p cnp > > Most notably I want to know if the name to added is a literal dot. > Yes, it is a dot (the directory itself): cn_nameptr = 0xfe0011428018 ".", cn_

Re: -CURRENT panics in NFS

2021-02-27 Thread Mateusz Guzik
You should be able to just use kgdb on the old kernel and the crashdump you already collected, provided both are still around. Alternatively boot with this without the fix: diff --git a/sys/kern/vfs_cache.c b/sys/kern/vfs_cache.c index fef1e31d197b..c4d2990b155d 100644 --- a/sys/kern/vfs_cache.c +

Re: -CURRENT panics in NFS

2021-02-27 Thread Juraj Lutter
I am now running a patched kernel, without problems. I can boot to unpatched one and see the output of these ddb commands. otis — Juraj Lutter XMPP: juraj (at) lutter.sk GSM: +421907986576 > On 27 Feb 2021, at 21:49, Mateusz Guzik wrote: > > Can you dump 'struct componentname *cnp'? This shou

Re: -CURRENT panics in NFS

2021-02-27 Thread Mateusz Guzik
Can you dump 'struct componentname *cnp'? This should do the trick: f 12 p cnp Most notably I want to know if the name to added is a literal dot. That case is handled if necessary, but the assert was added to start making the interface stricter. If the name is a dot I'll be inclined to remove the

Re: -CURRENT panics in NFS

2021-02-27 Thread Juraj Lutter
Hi, thank you for the swift reaction. This patch fixed my problem. otis — Juraj Lutter XMPP: juraj (at) lutter.sk GSM: +421907986576 > On 27 Feb 2021, at 16:53, Rick Macklem wrote: > > I reproduced the problem and the attached trivial patch > seems to fix it. Please test the patch if you can.

Re: -CURRENT panics in NFS

2021-02-27 Thread Rick Macklem
h the cracks. Thanks for reporting it, rick From: owner-freebsd-curr...@freebsd.org on behalf of Juraj Lutter Sent: Saturday, February 27, 2021 9:31 AM To: freebsd-current Subject: Re: -CURRENT panics in NFS CAUTION: This email originated from outside of the

Re: -CURRENT panics in NFS

2021-02-27 Thread Juraj Lutter
And a kgdb backtrace: (kgdb) bt #0 __curthread () at /usr/src/sys/amd64/include/pcpu_aux.h:55 #1 doadump (textdump=textdump@entry=0) at /usr/src/sys/kern/kern_shutdown.c:399 #2 0x804c7b2a in db_dump (dummy=, dummy2=, dummy3=, dummy4=) at /usr/src/sys/ddb/db_command.c:575 #3 0x

Re: -CURRENT panics in NFS

2021-02-27 Thread Juraj Lutter
Reliably reproducible: VNASSERT failed: dvp != vp not true at /usr/src/sys/kern/vfs_cache.c:2269 (cache_enter_time) 0xf80079321e00: type VDIR usecount 4, writecount 0, refcount 3 seqc users 0 mountedhere 0 hold count flags () flags (VV_ROOT|VV_VMSIZEVNLOCK) v_object 0xf801