swapper: page allocation failure

Bob Chiodini chiodr at kscems.ksc.nasa.gov
Wed Apr 28 16:56:56 UTC 2004


On Wed, 2004-04-28 at 12:23, Dave Jones wrote:
> On Sun, 2004-04-25 at 23:06, Tom Mitchell wrote:
> > What do I need to make a usefull bug report?
> >     2.6.5-1.332 #1 Mon Apr 19 10:13:26 EDT 2004 i686 athlon i386 GNU/Linux
> > ----
> > Apr 25 13:19:44 xtl2 kernel: inode_doinit_with_dentry:  context_to_sid(system_u:object_r:at_spool_t) returned 22 for dev=hda2 ino=1458196
> > Apr 25 13:42:16 xtl2 ntpd[1214]: synchronized to 192.168.0.51, stratum=2
> > Apr 25 13:44:04 xtl2 kernel: swapper: page allocation failure. order:1, mode:0x20
> > Apr 25 13:44:04 xtl2 kernel: Call Trace:
> > Apr 25 13:44:04 xtl2 kernel:  [<0213fbd3>] __alloc_pages+0x277/0x284
> > Apr 25 13:44:04 xtl2 kernel:  [<0213fbf8>] __get_free_pages+0x18/0x24
> > Apr 25 13:44:04 xtl2 kernel:  [<02143be5>] cache_grow+0x16b/0x3e0
> > Apr 25 13:44:04 xtl2 kernel:  [<0214420f>] cache_alloc_refill+0x3b5/0x3f1
> > Apr 25 13:44:04 xtl2 kernel:  [<02144878>] __kmalloc+0x8d/0x189
> > Apr 25 13:44:04 xtl2 kernel:  [<02284cb8>] alloc_skb+0x32/0xc3
> > Apr 25 13:44:04 xtl2 kernel:  [<22839cc2>] via_rhine_rx+0x404/0x486 [via_rhine]
> > Apr 25 13:44:04 xtl2 kernel:  [<22839502>] via_rhine_interrupt+0x6b/0x147 [via_rhine]
> > 
> > The machine did not fall over.  However it would not 'yum install' todays
> > latest until I turned enforcing off.  Yesterday and last night
> > I did install a number of new packages but did nothing to relabel
> > the system.
> 
> Should be fixed soon. I'm fairly certain this is the same issue as
> mentioned earlier. For some reason ext3 inodes fill up the entire ram,
> then the vm struggles trying to do an order 1 allocation (2 contiguous
> 4KB pages).
> This was pretty much the same failure mode I saw when updatedb blew up
> my box, except it was the e100 driver complaining for the same reason
> (no memory).
> 
> 	Dave

Dave,

Is there any way to verify this?  Vmstat perhaps.

This morning I caught my FC2T2 machine before is lost connectivity with
the outside world (ssh worked).  Local X and VT access were gone. 
Vmstat at the time indicated ~3300 free (K bytes?) on a 512MB machine. 
Last night before I left it was around 339000.  /proc/meminfo seems to
coincide, but not top.

Bob...





More information about the fedora-test-list mailing list