page allocation failures under network load

[jcfp]

Binary package hint: linux-source-2.6.20

When there's is significant network load, numerous "page allocation failures" appear in syslog. System does not crash, but short time of total unresponsiveness or even complete xorg hangs do happen. Reproduces quite easily when there's high network traffic (100mbit), such as when wget'ing some dvd iso from a fast mirror, or watching movies on divx stage6 website (with firefox + mplayerplugin). Tried with different network cards but that made no difference.

To the best of my knowledge, the problem started appearing after the latest security update to 2.6.20-16.32 (USN-518-1).
Kubuntu Feisty, Linux hostname 2.6.20-16-lowlatency #2 SMP PREEMPT Sun Sep 23 19:54:02 UTC 2007 i686 GNU/Linux

Typical error message looks like so (full dmesg etc will be attached):

[ 1158.143000] syslogd: page allocation failure. order:1, mode:0x20
[ 1158.143000] [<c015e300>] __alloc_pages+0x310/0x330
[ 1158.143000] [<c01752b1>] cache_alloc_refill+0x311/0x550
[ 1158.143000] [<c0175574>] __kmalloc+0x84/0x90
[ 1158.143000] [<c0282dc3>] __alloc_skb+0x53/0x120
[ 1158.143000] [<c02b7ec1>] tcp_collapse+0x1a1/0x380
[ 1158.143000] [<c02b81bc>] tcp_prune_queue+0x11c/0x310
[ 1158.143000] [<c02b8bbe>] tcp_data_queue+0x80e/0xc00
[ 1158.143000] [<c02ba64b>] tcp_rcv_established+0x2ab/0x9b0
[ 1158.143000] [<f8f86946>] __nf_conntrack_find+0x106/0x120 [nf_conntrack]
[ 1158.143000] [<f8f8a770>] tcp_packet+0x0/0xbc0 [nf_conntrack]
[ 1158.143000] [<c02c158e>] tcp_v4_do_rcv+0x2ce/0x550
[ 1158.143000] [<c012d045>] local_bh_enable_ip+0x25/0x60
[ 1158.143000] [<f8f0f6fb>] ipt_do_table+0x1fb/0x320 [ip_tables]
[ 1158.143000] [<c012cf97>] local_bh_enable+0x37/0xc0
[ 1158.143000] [<f8f8b831>] nf_ct_deliver_cached_events+0x81/0xf0 [nf_conntrack]
[ 1158.143000] [<c012cf97>] local_bh_enable+0x37/0xc0
[ 1158.143000] [<c02c3fd6>] tcp_v4_rcv+0x796/0x9a0
[ 1158.143000] [<c02a63a0>] ip_local_deliver+0x130/0x2e0
[ 1158.143000] [<c02a59b0>] ip_local_deliver_finish+0x0/0x250
[ 1158.143000] [<c02a5fb0>] ip_rcv+0x2b0/0x570
[ 1158.143000] [<c02a56c0>] ip_rcv_finish+0x0/0x2f0
[ 1158.143000] [<c012d045>] local_bh_enable_ip+0x25/0x60
[ 1158.143000] [<f8d35f46>] pppoe_rcv_core+0xd6/0x180 [pppoe]
[ 1158.143000] [<f8d35ff0>] pppoe_rcv+0x0/0xe0 [pppoe]
[ 1158.143000] [<c02f5111>] _spin_unlock+0x11/0x30
[ 1158.143000] [<c02a5d00>] ip_rcv+0x0/0x570
[ 1158.143000] [<c02875cd>] netif_receive_skb+0x22d/0x350
[ 1158.143000] [<c0282dc3>] __alloc_skb+0x53/0x120
[ 1158.143000] [<c02892b2>] process_backlog+0x92/0x120
[ 1158.143000] [<c02895a0>] net_rx_action+0xd0/0x2a0
[ 1158.143000] [<c012cacb>] __do_softirq+0x7b/0x100
[ 1158.143000] [<c012cbbd>] do_softirq+0x6d/0x70
[ 1158.143000] [<c012cd73>] irq_exit+0x43/0x50
[ 1158.143000] [<c0105e15>] do_IRQ+0x45/0x80
[ 1158.143000] [<c0135149>] sys_rt_sigprocmask+0xd9/0xf0
[ 1158.143000] [<c01043a3>] common_interrupt+0x23/0x30
[ 1158.143000] =======================
[ 1158.143000] Mem-info:
[ 1158.143000] DMA per-cpu:
[ 1158.143000] CPU 0: Hot: hi: 0, btch: 1 usd: 0 Cold: hi: 0, btch: 1 usd: 0
[ 1158.143000] Normal per-cpu:
[ 1158.143000] CPU 0: Hot: hi: 186, btch: 31 usd: 167 Cold: hi: 62, btch: 15 usd: 61
[ 1158.143000] HighMem per-cpu:
[ 1158.143000] CPU 0: Hot: hi: 42, btch: 7 usd: 4 Cold: hi: 14, btch: 3 usd: 12
[ 1158.143000] Active:98882 inactive:134613 dirty:22156 writeback:0 unstable:0 free:2336 slab:16435 mapped:26157 pagetables:553
[ 1158.143000] DMA free:3524kB min:68kB low:84kB high:100kB active:5844kB inactive:2132kB present:16256kB pages_scanned:0 all_unreclaimable? no
[ 1158.143000] lowmem_reserve[]: 0 873 1000
[ 1158.143000] Normal free:5652kB min:3744kB low:4680kB high:5616kB active:352656kB inactive:448708kB present:894080kB pages_scanned:0 all_unreclaimable? no
[ 1158.143000] lowmem_reserve[]: 0 0 1015
[ 1158.143000] HighMem free:168kB min:128kB low:264kB high:400kB active:37028kB inactive:87612kB present:129988kB pages_scanned:32 all_unreclaimable? no
[ 1158.143000] lowmem_reserve[]: 0 0 0
[ 1158.143000] DMA: 23*4kB 1*8kB 0*16kB 1*32kB 1*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 1*2048kB 0*4096kB = 3524kB
[ 1158.143000] Normal: 1237*4kB 0*8kB 0*16kB 0*32kB 1*64kB 1*128kB 0*256kB 1*512kB 0*1024kB 0*2048kB 0*4096kB = 5652kB
[ 1158.143000] HighMem: 4*4kB 3*8kB 2*16kB 1*32kB 1*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 168kB
[ 1158.143000] Swap cache: add 4781, delete 1, find 0/0, race 0+0
[ 1158.143000] Free swap = 2077316kB
[ 1158.143000] Total swap = 2096440kB
[ 1158.143000] Free swap: 2077316kB
[ 1158.143000] 262128 pages of RAM
[ 1158.143000] 32752 pages of HIGHMEM
[ 1158.143000] 4014 reserved pages
[ 1158.143000] 186288 pages shared
[ 1158.143000] 4780 pages swap cached
[ 1158.143000] 22156 pages dirty
[ 1158.143000] 0 pages writeback
[ 1158.143000] 26157 pages mapped
[ 1158.143000] 16435 pages slab
[ 1158.143000] 553 pages pagetables

[Launchpad Janitor (janitor)]

Beginning with the Hardy Heron 8.04 development cycle, all open Ubuntu kernel bugs need to be reported against the "linux" kernel package. We are automatically migrating this bug to the new "linux" package. However, development has already began for the upcoming Intrepid Ibex 8.10 release. It would be helpful if you could test the upcoming release and verify if this is still an issue - http://www.ubuntu.com/testing . If the issue still exists, please update this report by changing the Status of the "linux" task from "Incomplete" to "New". We appreciate your patience and understanding as we make this transition. Thanks!

[jcfp (jcfp)]

gone in 8.04