Package: src:linux
Version: 3.2.35-2~bpo60+1
Severity: normal
Tags: upstream patch
*** Please type your report below this line ***
We are running a NFSv4 server on Debian Squeeze with 3.2.0-0.bpo.4-amd64
and recently started observing performance degradation in the form of
slower response time for all NFS call. Investigation found elevated
CPU usage at kernel level (%sys) which was tracked down to mutex_spin_on_owner
and nfsd4_release_lockowner.
Sample output of "perf top":
Events: 70K cycles
61.46% [kernel] [k] mutex_spin_on_owner
30.68% [nfsd] [k] nfsd4_release_lockowner
1.50% [kernel] [k] intel_idle
0.11% [kernel] [k] irq_entries_start
0.11% [sunrpc] [k] svc_recv
[...]
As the problem was first happening intermittently but for hours at a time,
network captures were also taken and compared. It was found that the problem
(slower NFS response time and elevated kernel CPU usage) was correlated with
an elevated rate of RELEASE_LOCKOWNER requests.
The source for nfsd4_release_lockowner looks like this:
4252 __be32
4253 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4254 struct nfsd4_compound_state *cstate,
4255 struct nfsd4_release_lockowner *rlockowner)
4256 {
....
4275 nfs4_lock_state();
4276
4277 status = nfserr_locks_held;
4278 /* XXX: we're doing a linear search through all the lockowners.
4279 * Yipes! For now we'll just hope clients aren't really using
4280 * release_lockowner much, but eventually we have to fix these
4281 * data structures. */
4282 INIT_LIST_HEAD(&matches);
4283 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4284 list_for_each_entry(sop,
&lock_ownerstr_hashtbl[i], so_strhash) {
4285 if (!same_owner_str(sop, owner, clid))
4286 continue;
4287 list_for_each_entry(stp, &sop->so_stateids,
4288 st_perstateowner) {
4289 lo = lockowner(sop);
4290 if (check_for_locks(stp->st_file, lo))
4291 goto out;
4292 list_add(&lo->lo_list, &matches);
4293 }
4294 }
4295 }
4296 /* Clients probably won't expect us to return with some
(but not all)
4297 * of the lockowner state released; so don't release any until all
4298 * have been checked. */
4299 status = nfs_ok;
4300 while (!list_empty(&matches)) {
4301 lo = list_entry(matches.next, struct nfs4_lockowner,
4302 lo_list);
4303 /* unhash_stateowner deletes so_perclient only
4304 * for openowners. */
4305 list_del(&lo->lo_list);
4306 release_lockowner(lo);
4307 }
4308 out:
4309 nfs4_unlock_state();
4310 return status;
4311 }
So the problem is even documented at that level. Looking through upstream
git it appears a fix was applied in
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=06f1f864d4ae5804e83785308d41f14a08e4b980
We have not verified this patch applies cleanly or actually resolves
the problem but it seems likely.
Can you please consider applying that patch?
I have looked through existing Debian bug reports and I could not find
anything relevant although this one has similar superficial symptoms:
http://bugs.debian.org/692957
For the record, Dropbox 1.4.0 seems to be triggering an excessive amount
of RELEASE_LOCKOWNER requests, thus causing this problem. A separate bug report
has been filled with Dropbox:
https://forums.dropbox.com/topic.php?id=96061&replies=1
-- Package-specific info:
** Version:
Linux version 3.2.0-0.bpo.4-amd64 (debian-ker...@lists.debian.org)
(gcc version 4.4.5 (Debian 4.4.5-8) ) #1 SMP Debian 3.2.35-2~bpo60+1
** Command line:
BOOT_IMAGE=/boot/vmlinuz-3.2.0-0.bpo.4-amd64
root=UUID=ca958596-33fb-4e2c-9d87-16a74a584ea2 ro console=tty0 quiet
** Not tainted
** Loaded modules:
uinput
nfsd
nfs
lockd
fscache
auth_rpcgss
nfs_acl
sunrpc
ipmi_devintf
ipmi_msghandler
loop
snd_hda_intel
snd_hda_codec
snd_hwdep
snd_pcm
i7core_edac
edac_core
ioatdma
i2c_i801
snd_timer
snd
psmouse
acpi_cpufreq
mperf
i2c_core
tpm_tis
serio_raw
coretemp
processor
button
evdev
pcspkr
dca
crc32c_intel
tpm
tpm_bios
soundcore
snd_page_alloc
thermal_sys
ext4
mbcache
jbd2
crc16
dm_mod
raid10
raid1
md_mod
sd_mod
crc_t10dif
usbhid
hid
uhci_hcd
ehci_hcd
usbcore
ahci
libahci
libata
e1000e
scsi_mod
usb_common
-- System Information:
Debian Release: 6.0.2
APT prefers stable-updates
APT policy: (500, 'stable-updates'), (500, 'stable')
Architecture: amd64 (x86_64)
Kernel: Linux 3.2.0-0.bpo.4-amd64 (SMP w/4 CPU cores)
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Versions of packages linux-image-3.2.0-0.bpo.4-amd64 depends on:
ii debconf [debconf-2.0] 1.5.36.1 Debian configuration management sy
ii initramfs-tools [linux-init 0.99~bpo60+1 tools for generating an initramfs
ii linux-base 3.4~bpo60+1 Linux image base package
ii module-init-tools 3.12-1 tools for managing Linux kernel mo
Versions of packages linux-image-3.2.0-0.bpo.4-amd64 recommends:
pn firmware-linux-free <none> (no description available)
Versions of packages linux-image-3.2.0-0.bpo.4-amd64 suggests:
pn debian-kernel-handbook <none> (no description available)
ii grub-pc 1.98+20100804-14 GRand Unified Bootloader, version
pn linux-doc-3.2 <none> (no description available)
Versions of packages linux-image-3.2.0-0.bpo.4-amd64 is related to:
pn firmware-atheros <none> (no description available)
pn firmware-bnx2 <none> (no description available)
pn firmware-bnx2x <none> (no description available)
pn firmware-brcm80211 <none> (no description available)
pn firmware-intelwimax <none> (no description available)
pn firmware-ipw2x00 <none> (no description available)
pn firmware-ivtv <none> (no description available)
pn firmware-iwlwifi <none> (no description available)
pn firmware-libertas <none> (no description available)
pn firmware-linux <none> (no description available)
pn firmware-linux-nonfree <none> (no description available)
pn firmware-myricom <none> (no description available)
pn firmware-netxen <none> (no description available)
pn firmware-qlogic <none> (no description available)
pn firmware-ralink <none> (no description available)
pn firmware-realtek <none> (no description available)
pn xen-hypervisor <none> (no description available)
-- debconf information:
linux-image-3.2.0-0.bpo.4-amd64/postinst/missing-firmware-3.2.0-0.bpo.4-amd64:
linux-image-3.2.0-0.bpo.4-amd64/prerm/removing-running-kernel-3.2.0-0.bpo.4-amd64:
true
linux-image-3.2.0-0.bpo.4-amd64/postinst/depmod-error-initrd-3.2.0-0.bpo.4-amd64:
false
linux-image-3.2.0-0.bpo.4-amd64/postinst/ignoring-ramdisk:
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
