Bug #512

layered nfs mounts - lockmgr panic

Added by vnc over 7 years ago. Updated over 6 years ago.

Status:ClosedStart date:
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:-
Target version:-

Description

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

Under Dragonfly-Preview (as of 3 days ago), mounting anything over
an
nfs mount crashes. I first stumbled upon this with 2 layered nfs
mounts, like so:

# mount_nfs -3 -s -i rasid:/safe/muz /muz
# ls /muz
main rip vnc
# ls /muz/main
^C
# mount_nfs -3 -s -i rasid:/safe/muz /muz
panic: lockmgr: locking against myself
Trace beginning at frame 0xcf0c7a38
panic(c054d6c6,c0616ac0,c057dcf4,cf0c7a68,0) at panic+0x99
panic(c057dcf4,c10c5bc0,10,133f,0) at panic+0x99
lockmgr(cf35c5cc,2,2,cf35c530,c13964c0) at lockmgr+0x365
vn_lock(cf35c530,2,cf35c530,c13967e4,cf0c7af4) at vn_lock+0x21
vget(cf35c530,2,1c,0,0) at vget+0x33
nfs_nget(c13964c0,cf3c4cd0,1c,cf0c7b0c,0) at nfs_nget+0xc3
nfs_root(c13964c0,cf0c7b9c,cf0c7ba4,c061b648,c061b648) at
nfs_root+0x38
checkdirs(cf35c530,c13967e4,0,c11c6c78,c11c6c78) at checkdirs+0x2f
sys_mount(cf0c7cf4,cf0c7d04,10,cf64eb900,0) at sys_mount+0x844
syscall2(2f,2f,2f,2,bfbff738) at syscall2+0x20e
Xint0x80_syscall() at Xint0x80_syscall0x2a
Debugger("panic")
Stopped at Debugger+0x44: movb $0,in_Debugger.0
ddb>

Isolating the problem further I found:
* mounting anything over a nfs partition triggers the same
crash,
* mounting nfs over a local partition does not crash,

Sorry for the possible typos, this is entirely reproductible, so I
can
test patches and play around in ddb if it helps to debug it.
However
the machine is not very fast so it _will_ take a while to get test
results if you cannot reproduce it.

I have only restarted following DragonFly development, so forgive
me if
I could have known that this was fixed in -HEAD.

I am sending this as a pre-bug report. I am now initiating the
build of
an -HEAD kernel to see if it was fixed. I should be able to report
on
this within an hour or so. [update: I will have to update the
toolchain before i can build the kernel. So expect to wait until
tomorrow for -HEAD results. Though I can interrupt the build and
test diffs - no problem)

- --Vincent
-----BEGIN PGP SIGNATURE-----
Note: This signature can be verified at https://www.hushtools.com/verify
Version: Hush 2.5

wpwEAQECAAYFAkWlkgUACgkQsBV9fcPttKApJAQAiC2Ss1TtNTSfi0kgjDWCX6t+XxXX
IjRdb26vQmVrV9LhDquBNJN7Axk0vur7xzPVRMpry53QvJpn96HZhV8uhzDXnNRGb284
hY4Y4JXLJClTw4AuTcd5qwjYCoCL7siiQF9Di30yPKxHlyLTuxj6gF0sy40lzv/xJY6S
QmcI7qM=
=eil4
-----END PGP SIGNATURE-----

History

#1 Updated by vnc over 7 years ago

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Well, it took a couple more days. But it still crashes with
yesterday's HEAD. The dump is the same (besides the functions
addresses obviously)

Sorry for not even attempting to fix it. Other than that, the
release is impressively stable. Although after installing the
Preview I removed the CD too early and got an endless stream of
error messages before I rebooted "the hard way". But that's a
minor problem and most likely it's fixed already.

Again, contact me if you cannot reproduce. Otherwise I suppose
testing on faster machines than what I've got here will be more
efficient for debugging :)
- --Vincent
-----BEGIN PGP SIGNATURE-----
Note: This signature can be verified at https://www.hushtools.com/verify
Version: Hush 2.5

wpwEAQECAAYFAkWqcAwACgkQsBV9fcPttKCXKgP+ICNeKjxTJzbMOcWA+QzOWH7CO/kR
APArjvSUcpatydjSruGp+N3i/eIG7y2KlXqLlmLtxJr6Y6g0hrnkKGVCgBTyZWKWO/Nt
I0pqC6BPd1C/nc6TjFET9llZebgeEs20Ap0KD6Upf51VGI9Ley5jVAYtHgI/K2C3SRL6
/7Ooa0s=
=8YAQ
-----END PGP SIGNATURE-----

#2 Updated by dillon about 7 years ago

:# mount_nfs -3 -s -i rasid:/safe/muz /muz
:# ls /muz
:main rip vnc
:# ls /muz/main

I believe I have fixed this problem in HEAD.

-Matt

#3 Updated by pavalos over 6 years ago

After 11 months, we're assuming this is fixed.

Also available in: Atom PDF