vmalloc, xfs_buf_get: failed to map pages

DISCUSSION Mario Bachmann 2 participants 5 messages Jun 19, 2010 – Jul 1, 2010

lore

Share & tools

lore

Mario Bachmann

author

2010-06-19 12:58 UTC #0 lore raw

Hello,

on a 32bit machine, I have big problems with Linux 2.6.34. With Linux 2.6.33.4 everything runs fine.

I do not know where to search for known bugs in xfs.

/var/log/messages shows a lot of lines:

Jun 19 11:43:15 amd2000 kernel: vmap allocation for size 4194304 failed: use vmalloc=<size> to increase size.
Jun 19 11:43:15 amd2000 kernel: xfs_buf_get: failed to map pages

What to to?

Mario

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Christoph Hellwig

2010-06-21 07:18 UTC #1 lore raw

Quoted 5 lines · Mario Bachmann

Hello,
on a 32bit machine, I have big problems with Linux 2.6.34. With Linux 2.6.33.4 everything runs fine.

Quoted 8 lines

I do not know where to search for known bugs in xfs.
/var/log/messages shows a lot of lines:
Jun 19 11:43:15 amd2000 kernel: vmap allocation for size 4194304 failed: use vmalloc=<size> to increase size.
Jun 19 11:43:15 amd2000 kernel: xfs_buf_get: failed to map pages
What to to?

Strange. There have been no fixes to the vmap allocator nor to the XFS using this. What other modules are you ising? Maybe they increased their amount of used vmalloc space. For now the only thing you can do is using the vmalloc= kernel paramter to increase the vmalloc space.

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Mario Bachmann

author

2010-06-21 11:32 UTC #2 lore raw

Am Mon, 21 Jun 2010 03:21:18 -0400
schrieb Christoph Hellwig <[email protected]>:

Quoted 13 lines · Mario Bachmann

Quoted 12 lines
Hello,
on a 32bit machine, I have big problems with Linux 2.6.34. With Linux 2.6.33.4 everything runs fine.
I do not know where to search for known bugs in xfs.
/var/log/messages shows a lot of lines:
Jun 19 11:43:15 amd2000 kernel: vmap allocation for size 4194304 failed: use vmalloc=<size> to increase size.
Jun 19 11:43:15 amd2000 kernel: xfs_buf_get: failed to map pages

Quoted 6 lines

Quoted 1 line
What to to?
Strange. There have been no fixes to the vmap allocator nor to the XFS using this. What other modules are you ising? Maybe they increased their amount of used vmalloc space. For now the only thing you can do is using the vmalloc= kernel paramter to increase the vmalloc space.

What is the command to get all used modules?

Here is the output of the old kernel (2.6.33.4). I used the same kernel-.config for the new kernel (2.6.34):

amd2000 ~ # cat /proc/modules
sch_sfq 4645 1 - Live 0xf9537000
nfsd 219445 13 - Live 0xf9301000
ipv6 201317 34 - Live 0xf9283000
nf_conntrack_ftp 4780 0 - Live 0xf9238000
nf_conntrack_irc 2587 0 - Live 0xf922d000
ipt_REJECT 1785 2 - Live 0xf8782000
xt_tcpudp 1863 64 - Live 0xf85cc000
nf_conntrack_ipv4 7781 65 - Live 0xf85c0000
nf_defrag_ipv4 747 1 nf_conntrack_ipv4, Live 0xf85b4000
xt_state 919 65 - Live 0xf85ab000
ipt_LOG 4203 12 - Live 0xf8586000
xt_limit 1140 14 - Live 0xf857c000
iptable_filter 1034 1 - Live 0xf84d4000
ipt_addrtype 1449 0 - Live 0xf84ca000
xt_DSCP 1499 0 - Live 0xf84c1000
xt_dscp 1155 0 - Live 0xf84b8000
xt_string 1051 0 - Live 0xf84af000
xt_owner 791 6 - Live 0xf84a6000
xt_NFQUEUE 1565 0 - Live 0xf849d000
xt_multiport 2163 0 - Live 0xf8494000
xt_MARK 597 0 - Live 0xf848a000
xt_mark 597 0 - Live 0xf8481000
xt_iprange 1112 0 - Live 0xf8478000
xt_hashlimit 7490 0 - Live 0xf846e000
xt_conntrack 2183 0 - Live 0xf8462000
xt_CONNMARK 963 0 - Live 0xf8458000
xt_connmark 803 0 - Live 0xf803e000
nf_conntrack 38512 7 nf_conntrack_ftp,nf_conntrack_irc,nf_conntrack_ipv4,xt_state,xt_conntrack,xt_CONNMARK,xt_connmark, Live 0xf843f000
ip_tables 7645 1 iptable_filter, Live 0xf8034000
x_tables 9456 20 ipt_REJECT,xt_tcpudp,xt_state,ipt_LOG,xt_limit,ipt_addrtype,xt_DSCP,xt_dscp,xt_string,xt_owner,xt_NFQUEUE,xt_multiport,xt_MARK,xt_mark,xt_iprange,xt_hashlimit,xt_conntrack,xt_CONNMARK,xt_connmark,ip_tables, Live 0xf800d000 snd_ens1371 13650 0 - Live 0xfa21d000
gameport 6213 1 snd_ens1371, Live 0xfa20d000
snd_rawmidi 14248 1 snd_ens1371, Live 0xfa1fd000
snd_intel8x0 21243 0 - Live 0xfa1e8000
snd_seq_device 4105 1 snd_rawmidi, Live 0xfa1d5000
snd_ac97_codec 85509 2 snd_ens1371,snd_intel8x0, Live 0xfa190000
ac97_bus 702 1 snd_ac97_codec, Live 0xfa166000
snd_pcm 48336 3 snd_ens1371,snd_intel8x0,snd_ac97_codec, Live 0xfa0ef000
snd_timer 14100 1 snd_pcm, Live 0xfa0ca000
nvidia 9829303 0 - Live 0xf9636000 (P)
ppdev 5090 0 - Live 0xf8bdd000
parport_pc 23336 0 - Live 0xf8bbe000
snd 36462 7 snd_ens1371,snd_rawmidi,snd_intel8x0,snd_seq_device,snd_ac97_codec,snd_pcm,snd_timer, Live 0xf8b95000
parport 24435 2 ppdev,parport_pc, Live 0xf8b70000
snd_page_alloc 5345 2 snd_intel8x0,snd_pcm, Live 0xf8012000

What is a good value for vmalloc= ?
The machine has 1 GB of RAM.

Mario

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Mario Bachmann

author

2010-07-01 11:42 UTC #3 lore raw

In reply to Christoph Hellwig

Hi again,

what is a good value for vmalloc= ?
The machine has 1 GB of RAM.

Greetings
Mario

Am Mon, 21 Jun 2010 03:21:18 -0400
schrieb Christoph Hellwig <[email protected]>:

Quoted 13 lines · Mario Bachmann

Quoted 12 lines
Hello,
on a 32bit machine, I have big problems with Linux 2.6.34. With Linux 2.6.33.4 everything runs fine.
I do not know where to search for known bugs in xfs.
/var/log/messages shows a lot of lines:
Jun 19 11:43:15 amd2000 kernel: vmap allocation for size 4194304 failed: use vmalloc=<size> to increase size.
Jun 19 11:43:15 amd2000 kernel: xfs_buf_get: failed to map pages

Quoted 6 lines

Quoted 1 line
What to to?
Strange. There have been no fixes to the vmap allocator nor to the XFS using this. What other modules are you ising? Maybe they increased their amount of used vmalloc space. For now the only thing you can do is using the vmalloc= kernel paramter to increase the vmalloc space.

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Christoph Hellwig

2010-07-01 13:00 UTC #4 lore raw

Quoted 4 lines · Mario Bachmann

Hi again,
what is a good value for vmalloc= ?
The machine has 1 GB of RAM.

try vmalloc=256M

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs