Hi,

since I upgraded from Ubuntu Natty last autumn, I am experiencing system crashes whenever I write large amounts of data to disk. The error messages mention a generic protection fault in kswapd0, for example:

Code:
[  386.305170] general protection fault: 0000 [#1] PREEMPT SMP 
[  386.306634] CPU 0 
[  386.306634] Modules linked in: ipv6 it87 hwmon_vid coretemp usb_storage uas joydev snd_hda_codec_hdmi radeon iTCO_wdt r8169 snd_hda_codec_realtek iTCO_vendor_support i2c_i801 ppdev drm_kms_helper ttm psmouse pcspkr drm serio_raw parport_pc i2c_algo_bit parport intel_agp evdev i2c_core intel_gtt snd_hda_intel snd_hda_codec snd_hwdep snd_pcm snd_page_alloc snd_timer snd processor button soundcore mii ahci libahci fuse hid_cherry usbhid hid ext4 crc16 jbd2 mbcache sd_mod pata_acpi ata_piix uhci_hcd libata ehci_hcd scsi_mod usbcore usb_common
[  386.306634] 
[  386.306634] Pid: 17, comm: kswapd0 Not tainted 3.2.5-1-ARCH #1 Gigabyte Technology Co., Ltd. G41M-ES2L/G41M-ES2L
[  386.306634] RIP: 0010:[<ffffffff811187e8>]  [<ffffffff811187e8>] page_evictable+0x28/0xa0
[  386.306634] RSP: 0018:ffff88007b785a30  EFLAGS: 00010282
[  386.306634] RAX: fffe88007cc253c8 RBX: ffff88007b785da0 RCX: 0000000000016b28
[  386.306634] RDX: 010000000000081d RSI: 0000000000000000 RDI: ffffea0000985bc0
[  386.306634] RBP: ffff88007b785a40 R08: 000000000000000a R09: ffff88007b785fd8
[  386.306634] R10: 0000000000000001 R11: 0000000000000000 R12: ffff88007b785bc8
[  386.306634] R13: ffffea0000985be0 R14: ffffea0000985bc0 R15: 0000000000000001
[  386.306634] FS:  0000000000000000(0000) GS:ffff88007fc00000(0000) knlGS:0000000000000000
[  386.306634] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[  386.306634] CR2: 0000000001817830 CR3: 000000004ca24000 CR4: 00000000000006f0
[  386.306634] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  386.306634] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  386.306634] Process kswapd0 (pid: 17, threadinfo ffff88007b784000, task ffff88007bf97200)
[  386.306634] Stack:
[  386.306634]  ffffea0000950620 ffffea0000950600 ffff88007b785b60 ffffffff81118a6a
[  386.306634]  ffff88007fedb700 0000000000000001 ffff88007cc253c8 ffff88007bf97200
[  386.306634]  ffff88007b785fd8 ffff88007fedb700 ffff88007bf97200 000000000000000e
[  386.306634] Call Trace:
[  386.306634]  [<ffffffff81118a6a>] shrink_page_list+0x12a/0x980
[  386.306634]  [<ffffffff811168e2>] ? update_isolated_counts+0x162/0x190
[  386.306634]  [<ffffffff811196f2>] shrink_inactive_list+0x152/0x4e0
[  386.306634]  [<ffffffff8105cfd4>] ? dequeue_entity+0x104/0x290
[  386.306634]  [<ffffffff8111a25c>] shrink_zone+0x47c/0x5c0
[  386.306634]  [<ffffffff8111b5b5>] balance_pgdat+0x4e5/0x6d0
[  386.306634]  [<ffffffff8111b910>] kswapd+0x170/0x440
[  386.306634]  [<ffffffff81088d70>] ? abort_exclusive_wait+0xb0/0xb0
[  386.306634]  [<ffffffff8111b7a0>] ? balance_pgdat+0x6d0/0x6d0
[  386.306634]  [<ffffffff8108842c>] kthread+0x8c/0xa0
[  386.306634]  [<ffffffff8142b8b4>] kernel_thread_helper+0x4/0x10
[  386.306634]  [<ffffffff810883a0>] ? kthread_worker_fn+0x190/0x190
[  386.306634]  [<ffffffff8142b8b0>] ? gs_change+0x13/0x13
[  386.306634] Code: 00 00 00 55 48 89 e5 48 83 ec 10 66 66 66 66 90 48 8b 17 48 8b 17 48 8b 47 08 f7 c2 00 00 01 00 75 6e a8 01 75 16 48 85 c0 74 11 <48> 8b 90 88 00 00 00 31 c0 f7 c2 00 00 00 08 75 2a 48 8b 17 31 
[  386.306634] RIP  [<ffffffff811187e8>] page_evictable+0x28/0xa0
[  386.306634]  RSP <ffff88007b785a30>
[  386.390899] ---[ end trace e05fbac4fc52248a ]---
The actual call traces vary - often the issue occurs in
Code:
mem_cgroup_del_lru
isolate_lru_pages
shrink_inctive_list
shrink_zone
This issue is perfectly reproducible for me - I just need to write zeros to disk using e.g. dd and after 5-20GB the system will crash.

This affects every kernel I've tried after the Ubuntu Natty version 2.6.38.8 (e.g. 3.1.0-7 in fc16 or 3.2.6-2 in arch). The system is running perfectly stable under Windows and memtest shows no memory errors so I assume it's not a hardware problem. It also seems, only writes are affected since dd-ing entire hard disks to /dev/null won't crash the system.

I'd really appreciate any hints that could help me narrow this down.

Thank you!
- Christoph