New: swapper page allocation failure

Bugme-new Andrew Morton 5 participants 35 messages Apr 8, 2010 – Jun 2, 2010 1 message not in the archive

Subsystems (4)

Networking Memory management NFS client NFS server

lore

Share & tools

lore

Andrew Morton

author maintainer

2010-04-08 19:36 UTC #1 lore raw

In reply to a message not in the archive

(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).

Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:

https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output

Quoted 11 lines

I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.

hm, this is a regression.

: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]  <IRQ>  [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]  <EOI>  [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared

Either page reclaim got worse or kvm/virtio-net got more aggressive.

Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?

Thanks.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Avi Kivity

2010-04-08 19:40 UTC #2 lore raw

cc mst

Quoted 105 lines · Andrew Morton

(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).

Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:

https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output

Quoted 12 lines

I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.

hm, this is a regression.

: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared

Either page reclaim got worse or kvm/virtio-net got more aggressive.

Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?

Thanks.

Signature · 8 lines

I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Michael S. Tsirkin

2010-04-08 20:09 UTC #3 lore raw

Quoted 108 lines · Avi Kivity

cc mst

Quoted 105 lines · Andrew Morton

(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).

Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:

https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output

Quoted 12 lines

I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.

hm, this is a regression.

: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared

Either page reclaim got worse or kvm/virtio-net got more aggressive.

Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?

Thanks.

On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...

Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Signature · 7 lines

MST

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Robert Wimmer

2010-04-09 10:15 UTC #4 lore raw

I'm not really a git hero so here is what I've done:

cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57

Then I've checked

drivers/net/virtio_net.c
drivers/net/smc91x.c

if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel

genkernel --menuconfig --lvm --oldconfig all

which finally gave me a 2.6.31-rc5. I should mention that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926

Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.

Please let me know if the git commands above are right and/or if you need other kernel options enabled.

Thanks!
Robert

Quoted 114 lines · Michael S. Tsirkin

Quoted 112 lines · Avi Kivity
cc mst
Quoted 108 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 13 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.

Quoted 10 lines

On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Michael S. Tsirkin

2010-04-11 11:07 UTC #5 lore raw

Quoted 7 lines · Robert Wimmer

I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57

Looks right.

Quoted 12 lines

Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel

Quoted 3 lines

genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.

That's right.

Quoted 4 lines

I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926

Hmm, so we see the error here as well?

Quoted 5 lines

Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.

Looks right. You don't have to add -b flag if you don't want to.

Quoted 2 lines

Thanks!
Robert

Hmm, I do not see anything else that seems related. Could you please try to bisect?

git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c

should help assuming the change that triggers this is in virtio.

Quoted 120 lines · Michael S. Tsirkin

Quoted 119 lines · Avi Kivity
Quoted 113 lines
cc mst
Quoted 108 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 13 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...

Quoted 5 lines

Quoted 5 lines
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Robert Wimmer

2010-04-12 09:25 UTC #6 lore raw

server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers

Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.

Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.

This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420. (All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).

I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:

"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html

It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.

Thanks!
Robert

Quoted 53 lines · Michael S. Tsirkin

Quoted 8 lines · Robert Wimmer
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 4 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 5 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 6 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 3 lines
Thanks!
Robert

Quoted 141 lines

Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 132 lines · Michael S. Tsirkin
Quoted 131 lines · Avi Kivity
Quoted 118 lines
cc mst
Quoted 111 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 14 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Michael S. Tsirkin

2010-04-12 11:28 UTC #7 lore raw

Quoted 6 lines · Robert Wimmer

server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers

Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.

Type git bisect log to see your history.

8. Now git reset --hard ABCDEF~1 and try again.

If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.

Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html

Quoted 10 lines

Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.

Quoted 5 lines

This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.

Hmm. Some subtle race?

Quoted 16 lines

(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html

Quoted 8 lines

It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert

For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.

Quoted 189 lines · Michael S. Tsirkin

Quoted 188 lines · Robert Wimmer
Quoted 9 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 4 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 5 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 6 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 3 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 126 lines · Michael S. Tsirkin
Quoted 125 lines · Avi Kivity
Quoted 118 lines
cc mst
Quoted 111 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 14 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...

Quoted 6 lines

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Robert Wimmer

2010-04-12 13:50 UTC #8 lore raw

Sorry but I need some more git help. Here is what I've done. Started with a fresh clone of the kernel:

cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57

Since I already knew that this commit wasn't good I did

git bisect start
git bisect bad

compiled and started over. As expected the problem returns. So I've done another

git bisect bad

but I always get the same commit:

kabul:/usr/src/linux # git bisect log
git bisect start
# bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57

I've expected that after each "git bisect bad" I get the previous commit before the "bad" one. How can get the previous commit? The bisect documentation couldn't help me.

Thanks!
Robert

Quoted 76 lines · Michael S. Tsirkin

Quoted 7 lines · Robert Wimmer
server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.
Type git bisect log to see your history.
8. Now git reset --hard ABCDEF~1 and try again.
If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.
Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
Quoted 10 lines
Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.
Quoted 6 lines
This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.
Hmm. Some subtle race?
Quoted 16 lines
(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html
Quoted 9 lines
It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert

Quoted 222 lines

For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.
Quoted 214 lines · Michael S. Tsirkin
Quoted 213 lines · Robert Wimmer
Quoted 11 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 5 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 6 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 7 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 4 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 139 lines · Michael S. Tsirkin
Quoted 138 lines · Avi Kivity
Quoted 123 lines
cc mst
Quoted 114 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 15 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Michael S. Tsirkin

2010-04-12 13:57 UTC #9 lore raw

Quoted 9 lines · Robert Wimmer

Sorry but I need some more git help. Here is what I've done. Started with a fresh clone of the kernel:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57

Quoted 4 lines

Since I already knew that this commit wasn't good I did
git bisect start
git bisect bad

I think what you miss is marking the good commit.
bisect does a binary search but it needs to know
both good and bad commits to search in the range.

Optionally, you can use '-- drivers/virtio/ drivers/net/virtio_net.c' what this does is limit bisect to commits that touch files in question. This way you get much less tests to run
(about 4) but after you find a first problematic commit you must verify that a commit just before it does not have the issue.

If this turns out not to be the case, you'll have to fallback on full bisect, and we will now this is some other change in kernel that triggered the regression.

Quoted 23 lines

compiled and started over. As expected the problem returns. So I've done another
git bisect bad
but I always get the same commit:
kabul:/usr/src/linux # git bisect log
git bisect start
# bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
I've expected that after each "git bisect bad" I get the previous commit before the "bad" one. How can get the previous commit? The bisect documentation couldn't help me.

Quoted 304 lines

Thanks!
Robert
Quoted 298 lines · Michael S. Tsirkin
Quoted 7 lines · Robert Wimmer
server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.
Type git bisect log to see your history.
8. Now git reset --hard ABCDEF~1 and try again.
If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.
Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
Quoted 10 lines
Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.
Quoted 6 lines
This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.
Hmm. Some subtle race?
Quoted 16 lines
(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html
Quoted 9 lines
It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert
For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.
Quoted 214 lines · Michael S. Tsirkin
Quoted 213 lines · Robert Wimmer
Quoted 11 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 5 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 6 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 7 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 4 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 139 lines · Michael S. Tsirkin
Quoted 138 lines · Avi Kivity
Quoted 123 lines
cc mst
Quoted 114 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 15 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Robert Wimmer

2010-04-13 08:51 UTC #10 lore raw

I've tried to do my very best. In general I can
say: All 2.6.30 versions work, all 2.6.31 fail. 2.6.31-rc3 fails with "soft lockup" and is the only one which don't show any "swapper page allocation failure".
But the result is finally the same... 2.6.31-rc4
don't show "soft lockups" but "swapper page allocation failure". Here is the dmesg output for 2.6.31-rc3:
https://bugzilla.kernel.org/attachment.cgi?id=25986

So here is what I've done. Started with a fresh tree and my 2.6.30 .config:

rm -fr /usr/src/linux
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57

Here is the "git bisect log" output:

# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30' '--' 'drivers/virtio/' 'drivers/net/virtio_net.c'
# good: [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
git bisect good e3353853730eb99c56b7b0aed1667d51c0e3699a
# good: [9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb] Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6 git bisect good 9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb
# bad: [ff52c3fc7188855ede75d87b022271f0da309e5b] virtio: fix memory leak on device removal
git bisect bad ff52c3fc7188855ede75d87b022271f0da309e5b
# good: [31278e71471399beaff9280737e52b47db4dc345] net: group address list and its count
git bisect good 31278e71471399beaff9280737e52b47db4dc345
# bad: [4b892e6582e3a4fe01f623aea386907270d5bf83] virtio-pci: correctly unregister root device on error
git bisect bad 4b892e6582e3a4fe01f623aea386907270d5bf83

Hopefully this gives you some hints. The problem for me is that I don't know what commit I should consider good or bad. Should I consider the commit with the "soft lockup" as good because it don't show the allocation failure? Currently it is marked as bad (4b892e6582e3a4fe01f623aea386907270d5bf83). What should I do next?

Thanks!
Robert

Quoted 26 lines · Michael S. Tsirkin

Quoted 9 lines · Robert Wimmer
Sorry but I need some more git help. Here is what I've done. Started with a fresh clone of the kernel:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Quoted 5 lines
Since I already knew that this commit wasn't good I did
git bisect start
git bisect bad
I think what you miss is marking the good commit.
bisect does a binary search but it needs to know
both good and bad commits to search in the range.
Optionally, you can use '-- drivers/virtio/ drivers/net/virtio_net.c' what this does is limit bisect to commits that touch files in question. This way you get much less tests to run
(about 4) but after you find a first problematic commit you must verify that a commit just before it does not have the issue.

Quoted 360 lines

If this turns out not to be the case, you'll have to fallback on full bisect, and we will now this is some other change in kernel that triggered the regression.
Quoted 354 lines
compiled and started over. As expected the problem returns. So I've done another
git bisect bad
but I always get the same commit:
kabul:/usr/src/linux # git bisect log
git bisect start
# bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
I've expected that after each "git bisect bad" I get the previous commit before the "bad" one. How can get the previous commit? The bisect documentation couldn't help me.
Thanks!
Robert
Quoted 324 lines · Michael S. Tsirkin
Quoted 8 lines · Robert Wimmer
server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.
Type git bisect log to see your history.
8. Now git reset --hard ABCDEF~1 and try again.
If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.
Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
Quoted 10 lines
Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.
Quoted 7 lines
This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.
Hmm. Some subtle race?
Quoted 16 lines
(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html
Quoted 10 lines
It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert
For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.
Quoted 233 lines · Michael S. Tsirkin
Quoted 232 lines · Robert Wimmer
Quoted 13 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 6 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 7 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 8 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 5 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 146 lines · Michael S. Tsirkin
Quoted 145 lines · Avi Kivity
Quoted 4 lines
Quoted 124 lines
cc mst
Quoted 117 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 16 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Robert Wimmer

2010-04-19 12:55 UTC #11 lore raw

Is there a possibility to track this further down? I've problems on two other KVMs since a few weeks which I think that they're related to this. Host for this KVMs are kernel 2.6.32. Guests until today were also running 2.6.32. Inside the KVMs we're using GlusterFS, NFSv4 and Apache with PHP. From time to time the httpd-processes are "hanging". When this happens then we're seeing a lot of soft lockups. This hosts are running Xeon X5560 processors. Until today I suspected that this problems only happens on older Xeon's but this doesn't seems to be true. I've attached the output from /var/log/messages
(https://bugzilla.kernel.org/attachment.cgi?id=26048)
from one of the hosts with GlusterFS. I've now downgraded to kernel 2.6.30 in the guests. But since this problem also exists in 2.6.34-rc3 I suspect that we're never ever will be able to do a kernel update in the guests when they're using NFS :-(

But what I definitely can say is that all the problems only happens with guests running kernel >= 2.6.31 and with a remote file system (NFS, GlusterFS). Some days ago another KVM have had a network shutdown using kernel 2.6.32 in host and guest + NFSv4. But this only happend once until now and there isn't so much traffic running through the interfaces of that host.

All other guests with kernel 2.6.30 (about 80 guests on 18 hosts) with NFS and KVM 0.12.3 are really running perfectly.

Thanks!
Robert

Quoted 49 lines · Robert Wimmer

I've tried to do my very best. In general I can
say: All 2.6.30 versions work, all 2.6.31 fail. 2.6.31-rc3 fails with "soft lockup" and is the only one which don't show any "swapper page allocation failure".
But the result is finally the same... 2.6.31-rc4
don't show "soft lockups" but "swapper page allocation failure". Here is the dmesg output for 2.6.31-rc3:
https://bugzilla.kernel.org/attachment.cgi?id=25986
So here is what I've done. Started with a fresh tree and my 2.6.30 .config:
rm -fr /usr/src/linux
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Here is the "git bisect log" output:
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30' '--' 'drivers/virtio/' 'drivers/net/virtio_net.c'
# good: [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
git bisect good e3353853730eb99c56b7b0aed1667d51c0e3699a
# good: [9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb] Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6 git bisect good 9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb
# bad: [ff52c3fc7188855ede75d87b022271f0da309e5b] virtio: fix memory leak on device removal
git bisect bad ff52c3fc7188855ede75d87b022271f0da309e5b
# good: [31278e71471399beaff9280737e52b47db4dc345] net: group address list and its count
git bisect good 31278e71471399beaff9280737e52b47db4dc345
# bad: [4b892e6582e3a4fe01f623aea386907270d5bf83] virtio-pci: correctly unregister root device on error
git bisect bad 4b892e6582e3a4fe01f623aea386907270d5bf83
Hopefully this gives you some hints. The problem for me is that I don't know what commit I should consider good or bad. Should I consider the commit with the "soft lockup" as good because it don't show the allocation failure? Currently it is marked as bad (4b892e6582e3a4fe01f623aea386907270d5bf83). What should I do next?

Quoted 422 lines

Thanks!
Robert
Quoted 416 lines · Michael S. Tsirkin
Quoted 9 lines · Robert Wimmer
Sorry but I need some more git help. Here is what I've done. Started with a fresh clone of the kernel:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Quoted 6 lines
Since I already knew that this commit wasn't good I did
git bisect start
git bisect bad
I think what you miss is marking the good commit.
bisect does a binary search but it needs to know
both good and bad commits to search in the range.
Optionally, you can use '-- drivers/virtio/ drivers/net/virtio_net.c' what this does is limit bisect to commits that touch files in question. This way you get much less tests to run
(about 4) but after you find a first problematic commit you must verify that a commit just before it does not have the issue.
If this turns out not to be the case, you'll have to fallback on full bisect, and we will now this is some other change in kernel that triggered the regression.
Quoted 381 lines
compiled and started over. As expected the problem returns. So I've done another
git bisect bad
but I always get the same commit:
kabul:/usr/src/linux # git bisect log
git bisect start
# bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
I've expected that after each "git bisect bad" I get the previous commit before the "bad" one. How can get the previous commit? The bisect documentation couldn't help me.
Thanks!
Robert
Quoted 350 lines · Michael S. Tsirkin
Quoted 9 lines · Robert Wimmer
server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.
Type git bisect log to see your history.
8. Now git reset --hard ABCDEF~1 and try again.
If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.
Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
Quoted 10 lines
Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.
Quoted 8 lines
This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.
Hmm. Some subtle race?
Quoted 16 lines
(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html
Quoted 11 lines
It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert
For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.
Quoted 252 lines · Michael S. Tsirkin
Quoted 251 lines · Robert Wimmer
Quoted 4 lines
Quoted 11 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 7 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 8 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 9 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 6 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 153 lines · Michael S. Tsirkin
Quoted 152 lines · Avi Kivity
Quoted 5 lines
Quoted 128 lines
cc mst
Quoted 120 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 17 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Michael S. Tsirkin

2010-04-19 13:22 UTC #12 lore raw

So it seems the change that created the problem was not specific to virtio.

To track this further down, I think the thing to try would be to do a full bisect.

That is instead of git bisect start 'v2.6.31' 'v2.6.30' '--' 'drivers/virtio/' 'drivers/net/virtio_net.c'

git bisect start 'v2.6.31' 'v2.6.30'

and then test kernel versions as they are generated.

Quoted 31 lines · Robert Wimmer

Is there a possibility to track this further down? I've problems on two other KVMs since a few weeks which I think that they're related to this. Host for this KVMs are kernel 2.6.32. Guests until today were also running 2.6.32. Inside the KVMs we're using GlusterFS, NFSv4 and Apache with PHP. From time to time the httpd-processes are "hanging". When this happens then we're seeing a lot of soft lockups. This hosts are running Xeon X5560 processors. Until today I suspected that this problems only happens on older Xeon's but this doesn't seems to be true. I've attached the output from /var/log/messages
(https://bugzilla.kernel.org/attachment.cgi?id=26048)
from one of the hosts with GlusterFS. I've now downgraded to kernel 2.6.30 in the guests. But since this problem also exists in 2.6.34-rc3 I suspect that we're never ever will be able to do a kernel update in the guests when they're using NFS :-(
But what I definitely can say is that all the problems only happens with guests running kernel >= 2.6.31 and with a remote file system (NFS, GlusterFS). Some days ago another KVM have had a network shutdown using kernel 2.6.32 in host and guest + NFSv4. But this only happend once until now and there isn't so much traffic running through the interfaces of that host.
All other guests with kernel 2.6.30 (about 80 guests on 18 hosts) with NFS and KVM 0.12.3 are really running perfectly.

Quoted 477 lines

Thanks!
Robert
Quoted 471 lines · Robert Wimmer
I've tried to do my very best. In general I can
say: All 2.6.30 versions work, all 2.6.31 fail. 2.6.31-rc3 fails with "soft lockup" and is the only one which don't show any "swapper page allocation failure".
But the result is finally the same... 2.6.31-rc4
don't show "soft lockups" but "swapper page allocation failure". Here is the dmesg output for 2.6.31-rc3:
https://bugzilla.kernel.org/attachment.cgi?id=25986
So here is what I've done. Started with a fresh tree and my 2.6.30 .config:
rm -fr /usr/src/linux
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Here is the "git bisect log" output:
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30' '--' 'drivers/virtio/' 'drivers/net/virtio_net.c'
# good: [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
git bisect good e3353853730eb99c56b7b0aed1667d51c0e3699a
# good: [9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb] Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6 git bisect good 9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb
# bad: [ff52c3fc7188855ede75d87b022271f0da309e5b] virtio: fix memory leak on device removal
git bisect bad ff52c3fc7188855ede75d87b022271f0da309e5b
# good: [31278e71471399beaff9280737e52b47db4dc345] net: group address list and its count
git bisect good 31278e71471399beaff9280737e52b47db4dc345
# bad: [4b892e6582e3a4fe01f623aea386907270d5bf83] virtio-pci: correctly unregister root device on error
git bisect bad 4b892e6582e3a4fe01f623aea386907270d5bf83
Hopefully this gives you some hints. The problem for me is that I don't know what commit I should consider good or bad. Should I consider the commit with the "soft lockup" as good because it don't show the allocation failure? Currently it is marked as bad (4b892e6582e3a4fe01f623aea386907270d5bf83). What should I do next?
Thanks!
Robert
Quoted 416 lines · Michael S. Tsirkin
Quoted 9 lines · Robert Wimmer
Sorry but I need some more git help. Here is what I've done. Started with a fresh clone of the kernel:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Quoted 6 lines
Since I already knew that this commit wasn't good I did
git bisect start
git bisect bad
I think what you miss is marking the good commit.
bisect does a binary search but it needs to know
both good and bad commits to search in the range.
Optionally, you can use '-- drivers/virtio/ drivers/net/virtio_net.c' what this does is limit bisect to commits that touch files in question. This way you get much less tests to run
(about 4) but after you find a first problematic commit you must verify that a commit just before it does not have the issue.
If this turns out not to be the case, you'll have to fallback on full bisect, and we will now this is some other change in kernel that triggered the regression.
Quoted 381 lines
compiled and started over. As expected the problem returns. So I've done another
git bisect bad
but I always get the same commit:
kabul:/usr/src/linux # git bisect log
git bisect start
# bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
I've expected that after each "git bisect bad" I get the previous commit before the "bad" one. How can get the previous commit? The bisect documentation couldn't help me.
Thanks!
Robert
Quoted 350 lines · Michael S. Tsirkin
Quoted 9 lines · Robert Wimmer
server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.
Type git bisect log to see your history.
8. Now git reset --hard ABCDEF~1 and try again.
If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.
Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
Quoted 10 lines
Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel
in the host and guest for testing? Currently I'm still using 2.6.32 in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.
Quoted 8 lines
This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.
Hmm. Some subtle race?
Quoted 16 lines
(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio ring buffers available and the guest make more buffers available before qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html
Quoted 11 lines
It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines. The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert
For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.
Quoted 252 lines · Michael S. Tsirkin
Quoted 251 lines · Robert Wimmer
Quoted 4 lines
Quoted 11 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux cd linux
git checkout -b mykernel 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 7 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 8 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 9 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 6 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 153 lines · Michael S. Tsirkin
Quoted 152 lines · Avi Kivity
Quoted 5 lines
Quoted 128 lines
cc mst
Quoted 120 lines · Andrew Morton
(switched to email. Please respond via emailed reply-to-all, not via the bugzilla web interface).
Quoted 22 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No
Created an attachment (id=25903)
--> (https://bugzilla.kernel.org/attachment.cgi?id=25903)
dmesg output
Quoted 17 lines
I'm having problems with "swapper page allocation failure's" since upgrading from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur inside a kernel virtual maschine (KVM). Running Gentoo with kernel 2.6.32 as host which works fine. As long as kernel 2.6.30 is used as guest kernel the guest runs fine. But after upgrading to 2.6.32 and 2.6.33 I get "swapper page allocation failure's" (see attachment of dmesg output). The guest is only running a Apache webserver and serves files from a NFS share. It has 1 GB RAM and 2 virtual CPUs. I've tried different kernel configurations (e.g. a unmodified version from Sabayon Linux Distribution) but doesn't help. Load of the guest (and host) is very low. Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [  454.006706] users: page allocation failure. order:0, mode:0x20
: [  454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [  454.006714] Call Trace:
: [  454.006717]<IRQ>   [<ffffffff8109dff7>] __alloc_pages_nodemask+0x5c8/0x615
: [  454.006796]  [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [  454.006820]  [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [  454.006842]  [<ffffffff8167f2c7>] try_fill_recv+0x5e/0x20f
: [  454.006846]  [<ffffffff8167fe13>] virtnet_poll+0x52a/0x5c7
: [  454.006858]  [<ffffffff8104fe74>] ? run_timer_softirq+0x1dc/0x1f4
: [  454.006873]  [<ffffffff8176035d>] net_rx_action+0xad/0x1a5
: [  454.006882]  [<ffffffff8104b6cd>] __do_softirq+0x9c/0x127
: [  454.006897]  [<ffffffff81008ffc>] call_softirq+0x1c/0x30
: [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
: [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
: [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1
: [  454.006926]  [<ffffffff8183bc13>] ret_from_intr+0x0/0x11
: [  454.006928]<EOI>   [<ffffffff81026b25>] ? kvm_deferred_mmu_op+0x5e/0xe7
: [  454.006942]  [<ffffffff81026b19>] ? kvm_deferred_mmu_op+0x52/0xe7
: [  454.006946]  [<ffffffff81026c03>] kvm_mmu_write+0x2e/0x35
: [  454.006949]  [<ffffffff81026c7d>] kvm_set_pte_at+0x19/0x1b
: [  454.006953]  [<ffffffff810aba67>] __do_fault+0x3c4/0x492
: [  454.006957]  [<ffffffff810adcf4>] handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>] do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ? _raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544 isolated_anon:0
: [  454.007005]  active_file:4882 inactive_file:205902 isolated_file:0
: [  454.007006]  unevictable:0 dirty:11 writeback:0 unstable:0
: [  454.007007]  free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [  454.007008]  mapped:1895 shmem:113 pagetables:1370 bounce:0
: [  454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007021] lowmem_reserve[]: 0 994 994 994
: [  454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [  454.007036] lowmem_reserve[]: 0 0 0 0
: [  454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB 6*64kB 4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB
: [  454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the 2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC demands upon the page allocator?
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow ...
Robert, could you pls try 3161e453e496eb5643faad30fff5a5ab183da0fe and if that *does* have the problem, 0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

kernel

2010-04-21 09:23 UTC #13 lore raw

So after the compiler was running hot I've now the following result:

server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus' of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae

The last "git bisect good" prints out:

server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50

Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400

NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace

As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error.

This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.

Signed-off-by Trond Myklebust <[email protected]>

Signed-off-by Linus Torvalds <[email protected]>

:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs

Does this help you any further?

Thanks!
Robert

Quoted 10 lines · Michael S. Tsirkin

So it seems the change that created the problem was not specific to virtio.
To track this further down, I think the thing to try would be to do a full bisect.
That is instead of git bisect start 'v2.6.31' 'v2.6.30' '--' 'drivers/virtio/'
'drivers/net/virtio_net.c'

Quoted 85 lines

do
git bisect start 'v2.6.31' 'v2.6.30'
and then test kernel versions as they are generated.
Quoted 77 lines · Robert Wimmer
Is there a possibility to track this further down? I've problems on two other KVMs since a few weeks which I think that they're related to this. Host for this KVMs are kernel 2.6.32. Guests until today were also running 2.6.32. Inside the KVMs we're using GlusterFS, NFSv4 and Apache with PHP. From time to time the httpd-processes are "hanging". When this happens then we're seeing a lot of soft lockups. This hosts are running Xeon X5560 processors. Until today I suspected that this problems only happens on older Xeon's but this doesn't seems to be true. I've attached the output from /var/log/messages
(https://bugzilla.kernel.org/attachment.cgi?id=26048)
from one of the hosts with GlusterFS. I've now downgraded to kernel 2.6.30 in the guests. But since this problem also exists in 2.6.34-rc3 I suspect that we're never ever will be able to do a kernel update in the guests when they're using NFS :-(
But what I definitely can say is that all the problems only happens with guests running kernel >= 2.6.31 and with a remote file system (NFS, GlusterFS). Some days ago another KVM have had a network shutdown using kernel 2.6.32 in host and guest + NFSv4. But this only happend once until now and there isn't so much traffic running through the interfaces of that host.
All other guests with kernel 2.6.30 (about 80 guests on 18 hosts) with NFS and KVM 0.12.3 are really running perfectly.
Thanks!
Robert
Quoted 40 lines · Robert Wimmer
I've tried to do my very best. In general I can
say: All 2.6.30 versions work, all 2.6.31 fail. 2.6.31-rc3 fails with "soft lockup" and is the only one which don't show any "swapper page allocation failure".
But the result is finally the same... 2.6.31-rc4
don't show "soft lockups" but "swapper page allocation failure". Here is the dmesg output for 2.6.31-rc3:
https://bugzilla.kernel.org/attachment.cgi?id=25986
So here is what I've done. Started with a fresh tree and my 2.6.30 .config:
rm -fr /usr/src/linux
cd /usr/src
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
linux
cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Here is the "git bisect log" output:
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30' '--' 'drivers/virtio/' 'drivers/net/virtio_net.c'
# good: [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching for virtio drivers
git bisect good e3353853730eb99c56b7b0aed1667d51c0e3699a
# good: [9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb] Merge branch 'master'
of master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6 git bisect good 9cbc1cb8cd46ce1f7645b9de249b2ce8460129bb
# bad: [ff52c3fc7188855ede75d87b022271f0da309e5b] virtio: fix memory leak on device removal
git bisect bad ff52c3fc7188855ede75d87b022271f0da309e5b
# good: [31278e71471399beaff9280737e52b47db4dc345] net: group address list and its count
git bisect good 31278e71471399beaff9280737e52b47db4dc345
# bad: [4b892e6582e3a4fe01f623aea386907270d5bf83] virtio-pci:

correctly

Quoted 19 lines

Quoted 19 lines
Quoted 19 lines
unregister root device on error
git bisect bad 4b892e6582e3a4fe01f623aea386907270d5bf83
Hopefully this gives you some hints. The problem for me is that I don't know what commit I should consider good or bad. Should I consider the commit with the "soft lockup" as good because it don't show the allocation failure? Currently it is marked as bad (4b892e6582e3a4fe01f623aea386907270d5bf83). What should I do next?
Thanks!
Robert
Quoted 3 lines · Michael S. Tsirkin
On Mon, Apr 12, 2010 at 03:50:31PM +0200, Robert Wimmer wrote:

Quoted 6 lines

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Sorry but I need some more git help. Here is what I've done. Started with a fresh clone of the kernel:
cd /usr/src
git clone

git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

Quoted 10 lines

Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
linux
cd linux
git checkout 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
Since I already knew that this commit wasn't good I did
git bisect start
git bisect bad

Quoted 5 lines

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
I think what you miss is marking the good commit. bisect does a binary search but it needs to know both good and bad commits to search in the range.
Optionally, you can use '-- drivers/virtio/

drivers/net/virtio_net.c'

Quoted 4 lines

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
what this does is limit bisect to commits that touch files in question. This way you get much less tests to run (about 4) but after you find a first problematic commit you must verify that a commit just before it does not have the

issue.

Quoted 78 lines

Quoted 78 lines
Quoted 78 lines
Quoted 78 lines
If this turns out not to be the case, you'll have to fallback on full bisect, and we will now this is some other change in kernel that triggered the regression.
Quoted 70 lines
compiled and started over. As expected the problem returns. So I've done another
git bisect bad
but I always get the same commit:
kabul:/usr/src/linux # git bisect log
git bisect start
# bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57 # bad: [0b4f2928f14c4a9770b0866923fc81beb7f4aa57] smc91x: fix compilation on SMP
git bisect bad 0b4f2928f14c4a9770b0866923fc81beb7f4aa57
I've expected that after each "git bisect bad" I get the previous commit before the "bad" one. How can get the previous commit? The bisect documentation couldn't help me.
Thanks!
Robert
Quoted 39 lines · Michael S. Tsirkin
Quoted 10 lines · Robert Wimmer
server10:/usr/src/linux # git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
Bisecting: 12 revisions left to test after this (roughly 4 steps) [e3353853730eb99c56b7b0aed1667d51c0e3699a] virtio: enhance id_matching
for virtio drivers
Sorry I wasn't clear. the way to use bisect is as follows:
- first start as you did now.
1. now build kernel, install and test
2. if bug is there, type 'git bisect bad'
3. if bug is not there, type 'git bisect good'
4. The above will give you another kernel version to test if so go back to step 1
6. this will be repeated about 4 times (number of steps above)
7. in the end you will get the first revision which has the problem. Let's assume it is revision ABCDEF.
Type git bisect log to see your history.
8. Now git reset --hard ABCDEF~1 and try again.
If you see the problem with ABCDEF but not ABCDEF~1 then we will have a good guess at the culprit.
Some more tips here:
http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html

Quoted 6 lines

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Today I've upgraded to qemu-kvm-0.12.3-r1 (Gentoo package) but doesn't help. Still getting "page allocation failure" with 2.6.31-rc5.
Does it makes sense to use the same 2.6.31-rc5 kernel in the host and guest for testing? Currently I'm still using

2.6.32

Quoted 66 lines

Quoted 66 lines
Quoted 66 lines
Quoted 66 lines
Quoted 66 lines
Quoted 66 lines
Quoted 4 lines
in host and testing 2.6.31-rc5 in guest until "crashes". Then I start the guest with 2.6.30 again which works without trouble with 2.6.32 as host.
Quoted 8 lines
This is really strange. I have hosts with 2.6.32 running guests with 2.6.32 which works perfectly. These hosts and guests running on HP DL 380 G6 with Intel Xeon X5560. The guests which don't work with 2.6.32 (and 2.6.32 as host) running on HP DL 380 G5 with Intel Xeon L5420.
Hmm. Some subtle race?
Quoted 18 lines
(All guests) and (all hosts) have the same packages and the same versions installed and the same kernel configs (hosts and guests using different .config but the difference is very small e.g. CONFIG_PARAVIRT_SPINLOCKS=y, CONFIG_PARAVIRT_GUEST=y in guests but not in hosts .config).
I've had problems with qemu-kvm 0.12.2 with high network traffic which was solved by a patch submitted by Tom Lendacky:
"Fix a race condition where qemu finds that there are not enough virtio
ring buffers available and the guest make more buffers available before
qemu can enable notifications."
http://www.mail-archive.com/[email protected]/msg28667.html
Quoted 12 lines
It was a real lifesaver for the HP DL 380 G6 mentioned above but maybe this is now causing the problems with the G5 machines.
The symptoms are the same. I can still log into the guest via VNC but the network is down.
Thanks!
Robert
For now the only thing we seem to know for sure is that on specific hardware there's a regression between 2.6.30 and 2.6.31-rc5. Yes, it is possible that all it does is expose a qemu bug, but it's hard to say.
Let's find out what change
does that, this should give us a hint.
Quoted 9 lines · Michael S. Tsirkin
Quoted 8 lines · Robert Wimmer

Quoted 5 lines

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
I'm not really a git hero so here is what I've done:
cd /usr/src
git clone

git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

Quoted 3 lines

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
linux
cd linux
git checkout -b mykernel

0b4f2928f14c4a9770b0866923fc81beb7f4aa57

Quoted 97 lines

Quoted 97 lines
Quoted 97 lines
Quoted 97 lines
Quoted 97 lines
Quoted 97 lines
Quoted 97 lines
Quoted 97 lines
Quoted 4 lines
Looks right.
Quoted 12 lines
Then I've checked
drivers/net/virtio_net.c
drivers/net/smc91x.c
if the changes commited where not in there. Next I build my kernel as usual. I used my .config from 2.6.30 (which is working fine in a several guests / .config see here:
https://bugzilla.kernel.org/attachment.cgi?id=25925)
and build the kernel
Quoted 7 lines
genkernel --menuconfig --lvm --oldconfig all
which finally gave me a 2.6.31-rc5.
That's right.
Quoted 8 lines
I should mention
that 2.6.30 was using SLUB. So here is the output from the 2.6.31-rc5 kernel running about 20 min.:
https://bugzilla.kernel.org/attachment.cgi?id=25926
Hmm, so we see the error here as well?
Quoted 9 lines
Seems not very usefull to me. I'm currently compiling the same kernel with SLAB.
Please let me know if the git commands above are right and/or if you need other kernel options enabled.
Looks right. You don't have to add -b flag if you don't want to.
Quoted 6 lines
Thanks!
Robert
Hmm, I do not see anything else that seems related. Could you please try to bisect?
git bisect start v2.6.31 v2.6.30 -- drivers/virtio/ drivers/net/virtio_net.c
should help assuming the change that triggers this is in virtio.
Quoted 13 lines · Michael S. Tsirkin
Quoted 12 lines · Avi Kivity
Quoted 7 lines
cc mst

Quoted 7 lines

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines · Andrew Morton
Quoted 6 lines
(switched to email. Please respond via emailed

reply-to-all,

Quoted 30 lines

Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
Quoted 30 lines
not via the
bugzilla web interface).
Quoted 18 lines · On Wed, 7 Apr 2010 10:29:20 GMT [email protected] wrote:
https://bugzilla.kernel.org/show_bug.cgi?id=15709
Summary: swapper page allocation failure
Product: Memory Management
Version: 2.5
Kernel Version: 2.6.32 and 2.6.33
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: normal
Priority: P1
Component: Slab Allocator
AssignedTo: [email protected]
ReportedBy: [email protected]
Regression: No

Quoted 2 lines

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Created an attachment (id=25903)
-->

(https://bugzilla.kernel.org/attachment.cgi?id=25903)

Quoted 3 lines

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
dmesg output
I'm having problems with "swapper page allocation

failure's"

Quoted 6 lines

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
since upgrading
from kernel 2.6.30 to 2.6.32/2.6.33. The problems occur
inside a kernel virtual
maschine (KVM). Running Gentoo with kernel 2.6.32 as host
which works fine. As
long as kernel 2.6.30 is used as guest kernel the guest

runs

Quoted 6 lines

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
fine. But after
upgrading to 2.6.32 and 2.6.33 I get "swapper page
allocation failure's" (see
attachment of dmesg output). The guest is only running a
Apache webserver and
serves files from a NFS share. It has 1 GB RAM and 2

virtual

Quoted 4 lines

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
CPUs. I've tried
different kernel configurations (e.g. a unmodified version
from Sabayon Linux
Distribution) but doesn't help. Load of the guest (and

host)

Quoted 23 lines

Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 23 lines
Quoted 9 lines
is very low.
Network traffic is about 20-50 MBit/s.
hm, this is a regression.
: [ 454.006706] users: page allocation failure. order:0, mode:0x20
: [ 454.006712] Pid: 7992, comm: users Not tainted 2.6.34-rc3-git6 #2
: [ 454.006714] Call Trace:
: [ 454.006717]<IRQ> [<ffffffff8109dff7>]
__alloc_pages_nodemask+0x5c8/0x615
: [ 454.006796] [<ffffffff817860ce>] ? ip_local_deliver+0x65/0x6d
: [ 454.006820] [<ffffffff810c39c4>] alloc_pages_current+0x96/0x9f
: [ 454.006842] [<ffffffff8167f2c7>]

Quoted 8 lines

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
try_fill_recv+0x5e/0x20f
: [ 454.006846] [<ffffffff8167fe13>]
virtnet_poll+0x52a/0x5c7
: [ 454.006858] [<ffffffff8104fe74>] ?
run_timer_softirq+0x1dc/0x1f4
: [ 454.006873] [<ffffffff8176035d>]
net_rx_action+0xad/0x1a5
: [ 454.006882] [<ffffffff8104b6cd>]

__do_softirq+0x9c/0x127

Quoted 1 line

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
: [ 454.006897] [<ffffffff81008ffc>]

call_softirq+0x1c/0x30
>> >>>>>>>>>> : [  454.006901]  [<ffffffff8100af01>] do_softirq+0x41/0x7e
>> >>>>>>>>>> : [  454.006904]  [<ffffffff8104b3e3>] irq_exit+0x36/0x75
>> >>>>>>>>>> : [  454.006907]  [<ffffffff8100a5ee>] do_IRQ+0xaa/0xc1

Quoted 1 line

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
: [ 454.006926] [<ffffffff8183bc13>]

ret_from_intr+0x0/0x11

Quoted 5 lines

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
: [ 454.006928]<EOI> [<ffffffff81026b25>] ?
kvm_deferred_mmu_op+0x5e/0xe7
: [ 454.006942] [<ffffffff81026b19>] ?
kvm_deferred_mmu_op+0x52/0xe7
: [ 454.006946] [<ffffffff81026c03>]

kvm_mmu_write+0x2e/0x35

Quoted 3 lines

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
: [ 454.006949] [<ffffffff81026c7d>]
kvm_set_pte_at+0x19/0x1b
: [ 454.006953] [<ffffffff810aba67>]

__do_fault+0x3c4/0x492

Quoted 34 lines

Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
: [ 454.006957] [<ffffffff810adcf4>]
handle_mm_fault+0x478/0x9d8
: [  454.006966]  [<ffffffff810deb59>] ? path_put+0x2c/0x30
: [  454.006975]  [<ffffffff8102f162>]
do_page_fault+0x2f6/0x31a
: [  454.006979]  [<ffffffff8183b81e>] ?
_raw_spin_lock+0x9/0xd
: [  454.006982]  [<ffffffff8183bef5>] page_fault+0x25/0x30
: [  454.006985] Mem-Info:
: [  454.006987] Node 0 DMA per-cpu:
: [  454.006990] CPU    0: hi:    0, btch:   1 usd:   0
: [  454.006992] CPU    1: hi:    0, btch:   1 usd:   0
: [  454.006993] Node 0 DMA32 per-cpu:
: [  454.006996] CPU    0: hi:  186, btch:  31 usd: 185
: [  454.006998] CPU    1: hi:  186, btch:  31 usd: 112
: [  454.007003] active_anon:8308 inactive_anon:8544
isolated_anon:0
: [ 454.007005] active_file:4882 inactive_file:205902 isolated_file:0
: [ 454.007006] unevictable:0 dirty:11 writeback:0 unstable:0
: [ 454.007007] free:1385 slab_reclaimable:2445 slab_unreclaimable:4466
: [ 454.007008] mapped:1895 shmem:113 pagetables:1370 bounce:0
: [ 454.007010] Node 0 DMA free:4000kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:11844kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15768kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:64kB slab_unreclaimable:32kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
: [ 454.007021] lowmem_reserve[]: 0 994 994 994

Quoted 8 lines

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
: [ 454.007025] Node 0 DMA32 free:1540kB min:4000kB low:5000kB high:6000kB active_anon:33232kB inactive_anon:34176kB active_file:19528kB inactive_file:811764kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:1018068kB mlocked:0kB dirty:44kB writeback:0kB mapped:7580kB shmem:452kB slab_reclaimable:9716kB slab_unreclaimable:17832kB kernel_stack:1144kB pagetables:5480kB unstable:0kB

bounce:0kB

Quoted 3 lines

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no : [ 454.007036] lowmem_reserve[]: 0 0 0 0
: [ 454.007040] Node 0 DMA: 0*4kB 4*8kB 6*16kB 5*32kB

6*64kB

Quoted 36 lines

Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 36 lines
Quoted 23 lines
Quoted 23 lines
4*128kB 1*256kB 1*512kB 0*1024kB 1*2048kB 0*4096kB = 4000kB : [ 454.007050] Node 0 DMA32: 13*4kB 2*8kB 3*16kB 1*32kB 2*64kB 0*128kB 1*256kB 0*512kB 1*1024kB 0*2048kB 0*4096kB = 1556kB
: [  454.007059] 210914 total pagecache pages
: [  454.007061] 0 pages in swap cache
: [  454.007063] Swap cache stats: add 0, delete 0, find 0/0
: [  454.007065] Free swap  = 1959924kB
: [  454.007067] Total swap = 1959924kB
: [  454.014238] 262140 pages RAM
: [  454.014241] 7489 pages reserved
: [  454.014242] 21430 pages shared
: [  454.014244] 247174 pages non-shared
Either page reclaim got worse or kvm/virtio-net got more aggressive.
Avi, Rusty: can you think of any changes in the KVM/virtio area in the
2.6.30 -> 2.6.32 timeframe which may have increased the GFP_ATOMIC
demands upon the page allocator?
Quoted 7 lines
Quoted 7 lines
Thanks.
On the contrary, with commit 3161e453e496eb5643faad30fff5a5ab183da0fe
we should be using GFP_ATOMIC less.
But maybe there's a bug and it has the reverse effect somehow
...

Quoted 11 lines

Quoted 11 lines
Quoted 11 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Robert, could you pls try
3161e453e496eb5643faad30fff5a5ab183da0fe
and if that *does* have the problem,
0b4f2928f14c4a9770b0866923fc81beb7f4aa57?

Michael S. Tsirkin

2010-04-21 09:47 UTC #14 lore raw

Quoted 73 lines · kernel

So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus' of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>

Quoted 7 lines

:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?
Thanks!
Robert

Looks suspiciously like some error in testing.
Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Signature · 7 lines

MST

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

kernel

2010-04-22 09:31 UTC #15 lore raw

Maybe some comments to my former mail about what I've done:
I started with a fresh clone (deleted the old /usr/src/linux of course).

git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux

Then I started bisect

git bisect start 'v2.6.31' 'v2.6.30'

and build the first kernel and then marked kernels which "crashed" with "soft lockup" or "swapper page allocation failure" as bad and the other ones as good. Before I've compiled a new kernel I've always done a "make mrproper". I don't know if this is needed but thought it wouldn't hurt.

For me it was not clear that maybe I should have had stopped testing after the first commit that came up with a "swapper page allocation failure". It was only one commit which cased the allocation failure. All the other commits marked as bad came up with a soft lockup. But I thought it is important to find the earliest commit which crashes. So should I find out the commit with the allocation failure?

As you requested I've now done now a

git checkout c02d7adf8c5429727a98bad1d039bccad4c61c50

which ended with a soft lockup within 3 min. after starting the VM (see
https://bugzilla.kernel.org/attachment.cgi?id=26089&action=edit)
with this kernel.

Then I've done a

git checkout cf8d2c11cb77f129675478792122f50827e5b0ae

compiled and restarted the VM with this kernel version (BTW: Of course I've always used the same .config for all kernels I've build.). cf8d2c11cb77f129675478792122f50827e5b0ae is running fine.

Thanks!
Robert

Quoted 59 lines · Michael S. Tsirkin

Quoted 58 lines · kernel
So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus'
of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus'
of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card
driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400

Quoted 4 lines

Quoted 4 lines
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code

currently

Quoted 7 lines

Quoted 7 lines
has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in
nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the

namespace

Quoted 13 lines

Quoted 13 lines
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>
:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?

Quoted 7 lines

Quoted 2 lines
Thanks!
Robert
Looks suspiciously like some error in testing. Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae
is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Michael S. Tsirkin

2010-04-22 10:03 UTC #16 lore raw

Quoted 17 lines · kernel

Maybe some comments to my former mail about what I've done:
I started with a fresh clone (deleted the old /usr/src/linux of course).
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux
Then I started bisect
git bisect start 'v2.6.31' 'v2.6.30'
and build the first kernel and then marked kernels which "crashed" with "soft lockup" or "swapper page allocation failure" as bad and the other ones as good. Before I've compiled a new kernel I've always done a "make mrproper". I don't know if this is needed but thought it wouldn't hurt.

Quoted 7 lines

For me it was not clear that maybe I should have had stopped testing after the first commit that came up with a "swapper page allocation failure". It was only one commit which cased the allocation failure. All the other commits marked as bad came up with a soft lockup. But I thought it is important to find the earliest commit which crashes. So should I find out the commit with the allocation failure?

I think you did the right thing. We'll have to figure out soft lockup thing, then if page allocation failure turns out to be a different issue, look at it.

Quoted 8 lines

As you requested I've now done now a
git checkout c02d7adf8c5429727a98bad1d039bccad4c61c50
which ended with a soft lockup within 3 min. after starting the VM (see
https://bugzilla.kernel.org/attachment.cgi?id=26089&action=edit)
with this kernel.

I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?

Quoted 5 lines

Then I've done a
git checkout cf8d2c11cb77f129675478792122f50827e5b0ae

Quoted 7 lines

compiled and restarted the VM with this kernel version (BTW: Of course I've always used the same .config for all kernels I've build.). cf8d2c11cb77f129675478792122f50827e5b0ae is running fine.
Thanks!
Robert

Well, so the soft lockup issue seems NFS-related?
Trond, commit cf8d2c11cb77f129675478792122f50827e5b0ae seems to be causing problems on some old kernels (See bisect below). Any idea why?

Quoted 66 lines

Quoted 59 lines · Michael S. Tsirkin
Quoted 58 lines · kernel
So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus'
of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus'
of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card
driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400
Quoted 4 lines
Quoted 4 lines
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code
currently

Quoted 28 lines

Quoted 7 lines
Quoted 7 lines
has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in
nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the
namespace
Quoted 20 lines
Quoted 8 lines
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>
Quoted 7 lines
:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?
Thanks!
Robert
Looks suspiciously like some error in testing. Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae
is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Robert Wimmer

2010-04-23 05:26 UTC #17 lore raw

Quoted 2 lines

I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?

I'm building the kernels always with "genkernel" a Gentoo helper programm for kernel building. But I've looked into the log file of genkernel and there is nothing mentioned about striping the kernel. There will be a future release of genkernel which supports this but this is currently not the case. Since I haven't stripped the kernel I would answer no. Maybe a kernel option which should be enabled?

Thanks!
Robert

Quoted 54 lines · Michael S. Tsirkin

Quoted 17 lines · kernel
Maybe some comments to my former mail about what I've done:
I started with a fresh clone (deleted the old /usr/src/linux of course).
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux
Then I started bisect
git bisect start 'v2.6.31' 'v2.6.30'
and build the first kernel and then marked kernels which "crashed" with "soft lockup" or "swapper page allocation failure" as bad and the other ones as good. Before I've compiled a new kernel I've always done a "make mrproper". I don't know if this is needed but thought it wouldn't hurt.
Quoted 8 lines
For me it was not clear that maybe I should have had stopped testing after the first commit that came up with a "swapper page allocation failure". It was only one commit which cased the allocation failure. All the other commits marked as bad came up with a soft lockup. But I thought it is important to find the earliest commit which crashes. So should I find out the commit with the allocation failure?
I think you did the right thing. We'll have to figure out soft lockup thing, then if page allocation failure turns out to be a different issue, look at it.
Quoted 9 lines
As you requested I've now done now a
git checkout c02d7adf8c5429727a98bad1d039bccad4c61c50
which ended with a soft lockup within 3 min. after starting the VM (see
https://bugzilla.kernel.org/attachment.cgi?id=26089&action=edit)
with this kernel.
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
Quoted 9 lines
Then I've done a
git checkout cf8d2c11cb77f129675478792122f50827e5b0ae
compiled and restarted the VM with this kernel version (BTW: Of course I've always used the same .config for all kernels I've build.). cf8d2c11cb77f129675478792122f50827e5b0ae is running fine.

Quoted 110 lines

Quoted 3 lines
Thanks!
Robert
Well, so the soft lockup issue seems NFS-related?
Trond, commit cf8d2c11cb77f129675478792122f50827e5b0ae seems to be causing problems on some old kernels (See bisect below). Any idea why?
Quoted 101 lines
Quoted 60 lines · Michael S. Tsirkin
Quoted 58 lines · kernel
So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus'
of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus'
of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card
driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400
Quoted 5 lines
Quoted 5 lines
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code
currently
Quoted 8 lines
Quoted 8 lines
has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in
nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the
namespace
Quoted 21 lines
Quoted 8 lines
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>
Quoted 8 lines
:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?
Thanks!
Robert
Looks suspiciously like some error in testing. Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae
is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Michael S. Tsirkin

2010-04-25 09:18 UTC #18 lore raw

Quoted 11 lines · Robert Wimmer

Quoted 2 lines
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
I'm building the kernels always with "genkernel" a Gentoo helper programm for kernel building. But I've looked into the log file of genkernel and there is nothing mentioned about striping the kernel. There will be a future release of genkernel which supports this but this is currently not the case. Since I haven't stripped the kernel I would answer no. Maybe a kernel option which should be enabled?

Quoted 3 lines

Thanks!
Robert

Hmm. I have these
CONFIG_KALLSYMS=y
CONFIG_KALLSYMS_ALL=y
CONFIG_KALLSYMS_EXTRA_PASS=y
# CONFIG_STRIP_ASM_SYMS is not set

Quoted 159 lines · Michael S. Tsirkin

Quoted 156 lines · kernel
Quoted 18 lines
Maybe some comments to my former mail about what I've done:
I started with a fresh clone (deleted the old /usr/src/linux of course).
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux
Then I started bisect
git bisect start 'v2.6.31' 'v2.6.30'
and build the first kernel and then marked kernels which "crashed" with "soft lockup" or "swapper page allocation failure" as bad and the other ones as good. Before I've compiled a new kernel I've always done a "make mrproper". I don't know if this is needed but thought it wouldn't hurt.
Quoted 8 lines
For me it was not clear that maybe I should have had stopped testing after the first commit that came up with a "swapper page allocation failure". It was only one commit which cased the allocation failure. All the other commits marked as bad came up with a soft lockup. But I thought it is important to find the earliest commit which crashes. So should I find out the commit with the allocation failure?
I think you did the right thing. We'll have to figure out soft lockup thing, then if page allocation failure turns out to be a different issue, look at it.
Quoted 9 lines
As you requested I've now done now a
git checkout c02d7adf8c5429727a98bad1d039bccad4c61c50
which ended with a soft lockup within 3 min. after starting the VM (see
https://bugzilla.kernel.org/attachment.cgi?id=26089&action=edit)
with this kernel.
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
Quoted 4 lines
Then I've done a
git checkout cf8d2c11cb77f129675478792122f50827e5b0ae
Quoted 8 lines
compiled and restarted the VM with this kernel version (BTW: Of course I've always used the same .config for all kernels I've build.). cf8d2c11cb77f129675478792122f50827e5b0ae is running fine.
Thanks!
Robert
Well, so the soft lockup issue seems NFS-related?
Trond, commit cf8d2c11cb77f129675478792122f50827e5b0ae seems to be causing problems on some old kernels (See bisect below). Any idea why?
Quoted 93 lines
Quoted 60 lines · Michael S. Tsirkin
Quoted 58 lines · kernel
So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus'
of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus'
of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card
driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400
Quoted 5 lines
Quoted 5 lines
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code
currently
Quoted 8 lines
Quoted 8 lines
has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in
nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the
namespace
Quoted 13 lines
Quoted 13 lines
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>
:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?

Quoted 8 lines

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 3 lines
Thanks!
Robert
Looks suspiciously like some error in testing. Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae
is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Robert Wimmer

2010-04-25 20:41 UTC #19 lore raw

I've added CONFIG_KALLSYMS and CONFIG_KALLSYMS_ALL to my .config. I've uploaded the dmesg output. Maybe it helps a little bit:

https://bugzilla.kernel.org/attachment.cgi?id=26138

- Robert

Quoted 17 lines · Michael S. Tsirkin

Quoted 11 lines · Robert Wimmer
Quoted 3 lines
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
I'm building the kernels always with "genkernel" a Gentoo helper programm for kernel building. But I've looked into the log file of genkernel and there is nothing mentioned about striping the kernel. There will be a future release of genkernel which supports this but this is currently not the case. Since I haven't stripped the kernel I would answer no. Maybe a kernel option which should be enabled?
Quoted 4 lines
Thanks!
Robert

Quoted 190 lines

Hmm. I have these
CONFIG_KALLSYMS=y
CONFIG_KALLSYMS_ALL=y
CONFIG_KALLSYMS_EXTRA_PASS=y
# CONFIG_STRIP_ASM_SYMS is not set
Quoted 182 lines · Michael S. Tsirkin
Quoted 180 lines · kernel
Quoted 19 lines
Maybe some comments to my former mail about what I've done:
I started with a fresh clone (deleted the old /usr/src/linux of course).
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux
Then I started bisect
git bisect start 'v2.6.31' 'v2.6.30'
and build the first kernel and then marked kernels which "crashed" with "soft lockup" or "swapper page allocation failure" as bad and the other ones as good. Before I've compiled a new kernel I've always done a "make mrproper". I don't know if this is needed but thought it wouldn't hurt.
Quoted 9 lines
For me it was not clear that maybe I should have had stopped testing after the first commit that came up with a "swapper page allocation failure". It was only one commit which cased the allocation failure. All the other commits marked as bad came up with a soft lockup. But I thought it is important to find the earliest commit which crashes. So should I find out the commit with the allocation failure?
I think you did the right thing. We'll have to figure out soft lockup thing, then if page allocation failure turns out to be a different issue, look at it.
Quoted 10 lines
As you requested I've now done now a
git checkout c02d7adf8c5429727a98bad1d039bccad4c61c50
which ended with a soft lockup within 3 min. after starting the VM (see
https://bugzilla.kernel.org/attachment.cgi?id=26089&action=edit)
with this kernel.
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
Quoted 4 lines
Then I've done a
git checkout cf8d2c11cb77f129675478792122f50827e5b0ae
Quoted 9 lines
compiled and restarted the VM with this kernel version (BTW: Of course I've always used the same .config for all kernels I've build.). cf8d2c11cb77f129675478792122f50827e5b0ae is running fine.
Thanks!
Robert
Well, so the soft lockup issue seems NFS-related?
Trond, commit cf8d2c11cb77f129675478792122f50827e5b0ae seems to be causing problems on some old kernels (See bisect below). Any idea why?
Quoted 109 lines
Quoted 61 lines · Michael S. Tsirkin
Quoted 58 lines · kernel
So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus'
of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus'
of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card
driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400
Quoted 6 lines
Quoted 6 lines
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code
currently
Quoted 9 lines
Quoted 9 lines
has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in
nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the
namespace
Quoted 23 lines
Quoted 8 lines
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>
Quoted 9 lines
:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?
Thanks!
Robert
Looks suspiciously like some error in testing. Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae
is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Michael S. Tsirkin

2010-04-25 20:49 UTC #20 lore raw

So, it's an NFS-related regression, which is consistent with the bisect results. I guess someone who knows about NFS will have to look at it... BTW, you probably want to label the bug as regression.

Quoted 4 lines · Robert Wimmer

I've added CONFIG_KALLSYMS and CONFIG_KALLSYMS_ALL to my .config. I've uploaded the dmesg output. Maybe it helps a little bit:

Quoted 213 lines

https://bugzilla.kernel.org/attachment.cgi?id=26138
- Robert
Quoted 207 lines · Michael S. Tsirkin
Quoted 11 lines · Robert Wimmer
Quoted 3 lines
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
I'm building the kernels always with "genkernel" a Gentoo helper programm for kernel building. But I've looked into the log file of genkernel and there is nothing mentioned about striping the kernel. There will be a future release of genkernel which supports this but this is currently not the case. Since I haven't stripped the kernel I would answer no. Maybe a kernel option which should be enabled?
Quoted 4 lines
Thanks!
Robert
Hmm. I have these
CONFIG_KALLSYMS=y
CONFIG_KALLSYMS_ALL=y
CONFIG_KALLSYMS_EXTRA_PASS=y
# CONFIG_STRIP_ASM_SYMS is not set
Quoted 182 lines · Michael S. Tsirkin
Quoted 180 lines · kernel
Quoted 19 lines
Maybe some comments to my former mail about what I've done:
I started with a fresh clone (deleted the old /usr/src/linux of course).
git clone
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux
Then I started bisect
git bisect start 'v2.6.31' 'v2.6.30'
and build the first kernel and then marked kernels which "crashed" with "soft lockup" or "swapper page allocation failure" as bad and the other ones as good. Before I've compiled a new kernel I've always done a "make mrproper". I don't know if this is needed but thought it wouldn't hurt.
Quoted 9 lines
For me it was not clear that maybe I should have had stopped testing after the first commit that came up with a "swapper page allocation failure". It was only one commit which cased the allocation failure. All the other commits marked as bad came up with a soft lockup. But I thought it is important to find the earliest commit which crashes. So should I find out the commit with the allocation failure?
I think you did the right thing. We'll have to figure out soft lockup thing, then if page allocation failure turns out to be a different issue, look at it.
Quoted 10 lines
As you requested I've now done now a
git checkout c02d7adf8c5429727a98bad1d039bccad4c61c50
which ended with a soft lockup within 3 min. after starting the VM (see
https://bugzilla.kernel.org/attachment.cgi?id=26089&action=edit)
with this kernel.
I'm not sure why the lockup backtrace does not show function names - is the kernel stripped?
Quoted 4 lines
Then I've done a
git checkout cf8d2c11cb77f129675478792122f50827e5b0ae
Quoted 9 lines
compiled and restarted the VM with this kernel version (BTW: Of course I've always used the same .config for all kernels I've build.). cf8d2c11cb77f129675478792122f50827e5b0ae is running fine.
Thanks!
Robert
Well, so the soft lockup issue seems NFS-related?
Trond, commit cf8d2c11cb77f129675478792122f50827e5b0ae seems to be causing problems on some old kernels (See bisect below). Any idea why?
Quoted 109 lines
Quoted 61 lines · Michael S. Tsirkin
Quoted 58 lines · kernel
So after the compiler was running hot I've now the following result:
server10:/usr/src/linux # git bisect log
# bad: [74fca6a42863ffacaf7ba6f1936a9f228950f657] Linux 2.6.31 # good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30 git bisect start 'v2.6.31' 'v2.6.30'
# good: [925d74ae717c9a12d3618eb4b36b9fb632e2cef3] V4L/DVB (11736):
videobuf: modify return value of VIDIOC_REQBUFS ioctl
git bisect good 925d74ae717c9a12d3618eb4b36b9fb632e2cef3
# bad: [a380137900fca5c79e6daa9500bdb6ea5649188e] ixgbe: Fix device capabilities of 82599 single speed fiber NICs.
git bisect bad a380137900fca5c79e6daa9500bdb6ea5649188e
# good: [1dbb5765acc7a6fe4bc1957c001037cc9d02ae03] Staging: android:
lowmemorykiller: fix up remaining checkpatch warnings
git bisect good 1dbb5765acc7a6fe4bc1957c001037cc9d02ae03
# good: [df36b439c5fedefe013d4449cb6a50d15e2f4d70] Merge branch 'for-2.6.31' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
git bisect good df36b439c5fedefe013d4449cb6a50d15e2f4d70
# bad: [a800faec1b21d7133b5f0c8c6dac593b7c4e118d] Merge branch 'for-linus'
of git://www.jni.nu/cris
git bisect bad a800faec1b21d7133b5f0c8c6dac593b7c4e118d
# good: [ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2] Merge
git://git.infradead.org/mtd-2.6
git bisect good ac1b7c378ef26fba6694d5f118fe7fc16fee2fe2
# bad: [37c6dbe290c05023b47f52528e30ce51336b93eb] V4L/DVB (12091):
gspca_sonixj: Add light frequency control
git bisect bad 37c6dbe290c05023b47f52528e30ce51336b93eb
# bad: [687d680985b1438360a9ba470ece8b57cd205c3b] Merge
git://git.infradead.org/~dwmw2/iommu-2.6.31
git bisect bad 687d680985b1438360a9ba470ece8b57cd205c3b
# bad: [1053414068bad659479e6efa62a67403b8b1ec0a] Merge branch 'for-linus'
of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
git bisect bad 1053414068bad659479e6efa62a67403b8b1ec0a
# good: [b01b4babbf204443b5a846a7494546501614cefc] firewire: net: fix card
driver reloading
git bisect good b01b4babbf204443b5a846a7494546501614cefc
# bad: [c02d7adf8c5429727a98bad1d039bccad4c61c50] NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace git bisect bad c02d7adf8c5429727a98bad1d039bccad4c61c50
# good: [616511d039af402670de8500d0e24495113a9cab] VFS: Uninline the function put_mnt_ns()
git bisect good 616511d039af402670de8500d0e24495113a9cab
# good: [cf8d2c11cb77f129675478792122f50827e5b0ae] VFS: Add VFS helper functions for setting up private namespaces
git bisect good cf8d2c11cb77f129675478792122f50827e5b0ae
The last "git bisect good" prints out:
server10:/usr/src/linux # git bisect good c02d7adf8c5429727a98bad1d039bccad4c61c50 is the first bad commit commit c02d7adf8c5429727a98bad1d039bccad4c61c50
Author: Trond Myklebust <[email protected]>
Date:   Mon Jun 22 15:09:14 2009 -0400
Quoted 6 lines
Quoted 6 lines
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private
namespace
As noted in the previous patch, the NFSv4 client mount code
currently
Quoted 9 lines
Quoted 9 lines
has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in
nfs4_path_walk() will fail with an error.
This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the
namespace
Quoted 23 lines
Quoted 8 lines
on the
server, then uses the ordinary VFS and NFS path lookup code to walk
down the
mount path in that namespace.
Signed-off-by Trond Myklebust <[email protected]>
Signed-off-by Linus Torvalds <[email protected]>
Quoted 9 lines
:040000 040000 97a18818f26ab9a0987f157257eb6f399c3cc1cc
9ab6c712bb64f1349b5ac9f2020191abb5780ca0 M fs
Does this help you any further?
Thanks!
Robert
Looks suspiciously like some error in testing. Could you pls retest and verify again that cf8d2c11cb77f129675478792122f50827e5b0ae
is good and c02d7adf8c5429727a98bad1d039bccad4c61c50 is bad?

Trond Myklebust

2010-04-26 12:15 UTC #21 lore raw

Quoted 4 lines · Michael S. Tsirkin

So, it's an NFS-related regression, which is consistent with the bisect results. I guess someone who knows about NFS will have to look at it... BTW, you probably want to label the bug as regression.

Quoted 9 lines

Quoted 8 lines · Robert Wimmer
I've added CONFIG_KALLSYMS and CONFIG_KALLSYMS_ALL to my .config. I've uploaded the dmesg output. Maybe it helps a little bit:
https://bugzilla.kernel.org/attachment.cgi?id=26138
- Robert

That last trace is just saying that the NFSv4 reboot recovery code is crashing (which is hardly surprising if the memory management is hosed).

The initial bisection makes little sense to me: it is basically blaming a page allocation problem on a change to the NFSv4 mount code. The only way I can see that possibly happen is if you are hitting a stack overflow.
So 2 questions:

- Are you able to reproduce the bug when using NFSv3 instead?
- Have you tried running with stack tracing enabled?

Cheers
Trond

Robert Wimmer

2010-04-26 20:25 UTC #22 lore raw

Quoted 13 lines

Quoted 10 lines
Quoted 10 lines
I've added CONFIG_KALLSYMS and CONFIG_KALLSYMS_ALL to my .config. I've uploaded the dmesg output. Maybe it helps a little bit:
https://bugzilla.kernel.org/attachment.cgi?id=26138
- Robert
That last trace is just saying that the NFSv4 reboot recovery code is crashing (which is hardly surprising if the memory management is hosed).

Quoted 8 lines

The initial bisection makes little sense to me: it is basically blaming a page allocation problem on a change to the NFSv4 mount code. The only way I can see that possibly happen is if you are hitting a stack overflow.
So 2 questions:
- Are you able to reproduce the bug when using NFSv3 instead?

I've tried with NFSv3 now. With v4 the error normally occur within 5 minutes. The VM is now running for one hour and no soft lockup so far. So I would say it can't be reproduced with v3.

Quoted 2 lines

- Have you tried running with stack tracing enabled?

Can you explain this a little bit more please? CONFIG_STACKTRACE=y was already enabled. I've now enabled

CONFIG_USER_STACKTRACE_SUPPORT=y
CONFIG_NOP_TRACER=y
CONFIG_HAVE_FTRACE_NMI_ENTER=y
CONFIG_HAVE_FUNCTION_TRACER=y
CONFIG_HAVE_FUNCTION_GRAPH_TRACER=y
CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST=y
CONFIG_HAVE_DYNAMIC_FTRACE=y
CONFIG_HAVE_FTRACE_MCOUNT_RECORD=y
CONFIG_HAVE_FTRACE_SYSCALLS=y
CONFIG_FTRACE_NMI_ENTER=y
CONFIG_CONTEXT_SWITCH_TRACER=y
CONFIG_GENERIC_TRACER=y
CONFIG_FTRACE=y
CONFIG_FUNCTION_TRACER=y
CONFIG_FUNCTION_GRAPH_TRACER=y
CONFIG_FTRACE_SYSCALLS=y
CONFIG_STACK_TRACER=y
CONFIG_KMEMTRACE=y
CONFIG_DYNAMIC_FTRACE=y
CONFIG_FTRACE_MCOUNT_RECORD=y
CONFIG_HAVE_MMIOTRACE_SUPPORT=y

and run

echo 1 > /proc/sys/kernel/stack_tracer_enabled

But the output is mostly the same in dmesg/ var/log/messages. Can you please guide me how I can enable the stack tracing you need?

Thanks!
Robert

Trond Myklebust

2010-04-26 21:04 UTC #23 lore raw

Quoted 4 lines · Robert Wimmer

I've tried with NFSv3 now. With v4 the error normally occur within 5 minutes. The VM is now running for one hour and no soft lockup so far. So I would say it can't be reproduced with v3.

Thanks! That's useful info.

Quoted 28 lines

Quoted 2 lines
- Have you tried running with stack tracing enabled?
Can you explain this a little bit more please? CONFIG_STACKTRACE=y was already enabled. I've now enabled
CONFIG_USER_STACKTRACE_SUPPORT=y
CONFIG_NOP_TRACER=y
CONFIG_HAVE_FTRACE_NMI_ENTER=y
CONFIG_HAVE_FUNCTION_TRACER=y
CONFIG_HAVE_FUNCTION_GRAPH_TRACER=y
CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST=y
CONFIG_HAVE_DYNAMIC_FTRACE=y
CONFIG_HAVE_FTRACE_MCOUNT_RECORD=y
CONFIG_HAVE_FTRACE_SYSCALLS=y
CONFIG_FTRACE_NMI_ENTER=y
CONFIG_CONTEXT_SWITCH_TRACER=y
CONFIG_GENERIC_TRACER=y
CONFIG_FTRACE=y
CONFIG_FUNCTION_TRACER=y
CONFIG_FUNCTION_GRAPH_TRACER=y
CONFIG_FTRACE_SYSCALLS=y
CONFIG_STACK_TRACER=y
CONFIG_KMEMTRACE=y
CONFIG_DYNAMIC_FTRACE=y
CONFIG_FTRACE_MCOUNT_RECORD=y
CONFIG_HAVE_MMIOTRACE_SUPPORT=y

Quoted 7 lines

and run
echo 1 > /proc/sys/kernel/stack_tracer_enabled
But the output is mostly the same in dmesg/ var/log/messages. Can you please guide me how I can enable the stack tracing you need?

Sure. In addition to what you did above, please do

mount -t debugfs none /sys/kernel/debug

and then cat the contents of the pseudofile at

/sys/kernel/debug/tracing/stack_trace

Please do this more or less immediately after you've finished mounting the NFSv4 client.

Does your server have the 'crossmnt' or 'nohide' flags set, or does it use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.

Cheers
Trond

Robert Wimmer

2010-04-26 22:18 UTC #24 lore raw

Quoted 12 lines

Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished mounting the NFSv4 client.

I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:

After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155

Quoted 4 lines

Does your server have the 'crossmnt' or 'nohide' flags set, or does it use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.

The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash

Thanks!
Robert

Trond Myklebust

2010-04-26 23:29 UTC #25 lore raw

Quoted 26 lines · Robert Wimmer

Quoted 11 lines
Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished mounting the NFSv4 client.
I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:
After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155
Quoted 4 lines
Does your server have the 'crossmnt' or 'nohide' flags set, or does it use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.

Quoted 7 lines

The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash
Thanks!
Robert

That second trace is more than 5.5K deep, more than half of which is socket overhead :-(((.

The process stack does not appear to have overflowed, however that trace doesn't include any IRQ stack overhead.

OK... So what happens if we get rid of half of that trace by forcing asynchronous tasks such as this to run entirely in rpciod instead of first trying to run in the process context?

See the attachment...

SUNRPC: Reduce asynchronous RPC task stack usage

From: Trond Myklebust <[email protected]>

We should just farm out asynchronous RPC tasks immediately to rpciod...

Signed-off-by Trond Myklebust <[email protected]>

---

 net/sunrpc/sched.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

net/sunrpc/sched.c +6 -1

index c8979ce..22a097f 100644
@@ -720,7 +720,12 @@ void rpc_execute(struct rpc_task *task)
 {
 	rpc_set_active(task);
 	rpc_set_running(task);
-	__rpc_execute(task);
+	if (RPC_IS_ASYNC(task)) {
+		INIT_WORK(&task->u.tk_work, rpc_async_schedule);
+		queue_work(rpciod_workqueue, &task->u.tk_work);
+
+	} else
+		__rpc_execute(task);
 }
 
 static void rpc_async_schedule(struct work_struct *work)

Robert Wimmer

2010-04-27 22:56 UTC #26 lore raw

I've applied the patch against the kernel which I got from "git clone ...." resulted in a kernel 2.6.34-rc5.

The stack trace after mounting NFS is here:
https://bugzilla.kernel.org/attachment.cgi?id=26166
/var/log/messages after soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26167

I hope that there is any usefull information in there.

Thanks!
Robert

Quoted 41 lines · Trond Myklebust

Quoted 28 lines · Robert Wimmer
Quoted 12 lines
Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished mounting the NFSv4 client.
I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:
After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155
Quoted 5 lines
Does your server have the 'crossmnt' or 'nohide' flags set, or does it use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.
Quoted 8 lines
The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash
Thanks!
Robert
That second trace is more than 5.5K deep, more than half of which is socket overhead :-(((.

Quoted 9 lines

The process stack does not appear to have overflowed, however that trace doesn't include any IRQ stack overhead.
OK... So what happens if we get rid of half of that trace by forcing asynchronous tasks such as this to run entirely in rpciod instead of first trying to run in the process context?
See the attachment...

kernel

2010-05-03 06:11 UTC #27 lore raw

Anything we can do to investigate this further?

Thanks!
Robert

Quoted 8 lines · Robert Wimmer

I've applied the patch against the kernel which I got from "git clone ...." resulted in a kernel 2.6.34-rc5.
The stack trace after mounting NFS is here:
https://bugzilla.kernel.org/attachment.cgi?id=26166
/var/log/messages after soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26167

Quoted 17 lines

I hope that there is any usefull information in there.
Thanks!
Robert
Quoted 11 lines · Trond Myklebust
Quoted 9 lines · Robert Wimmer
Quoted 9 lines
Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished

mounting

Quoted 14 lines

Quoted 14 lines
Quoted 14 lines
Quoted 3 lines
the NFSv4 client.
I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:
After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155

Quoted 1 line

Quoted 1 line
Quoted 1 line
Quoted 1 line
Does your server have the 'crossmnt' or 'nohide' flags set, or does

Quoted 12 lines

Quoted 12 lines
Quoted 12 lines
Quoted 4 lines
use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.
The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash
Thanks!
Robert

Quoted 4 lines

Quoted 4 lines
That second trace is more than 5.5K deep, more than half of which is socket overhead :-(((.
The process stack does not appear to have overflowed, however that

trace

Quoted 8 lines

Quoted 8 lines
doesn't include any IRQ stack overhead.
OK... So what happens if we get rid of half of that trace by forcing asynchronous tasks such as this to run entirely in rpciod instead of first trying to run in the process context?
See the attachment...

Robert Wimmer

2010-05-06 21:19 UTC #28 lore raw

I don't know if someone is still interested in this but I think Trond isn't further interested because the last error was of cource a "page allocation failure" and not a "soft lookup" which Trond was trying to solve. But the patch was for 2.6.34 and the "soft lookup" comes up only with some 2.6.30 and maybe some 2.6.31 kernel versions. But the first error I reported was a "page allocation failure" which all kernels >= 2.6.32 produces with this configuration I use (NFSv4).

Michael suggested to first solve the "soft lookup" before further investigating the "page allocation failure". We know that the "soft lookup" only pop's up with NFSv4 and not v3. I really want to use v4 but since I'm not a kernel hacker someone must guide me what to try next.

I know that you're all have a lot of other work to do but if there're no ideas left what to do next it's maybe best to close the bug for now and I stay with kernel 2.6.30 for now or go back to NFS v3 if I upgrade to a newer kernel. Maybe the error will be fixed "by accident" in >= 2.6.35 ;-)

Thanks!
Robert

Quoted 73 lines · On 05/03/10 10:11, [email protected] wrote:

Anything we can do to investigate this further?
Thanks!
Robert
Quoted 8 lines · Robert Wimmer
I've applied the patch against the kernel which I got from "git clone ...." resulted in a kernel 2.6.34-rc5.
The stack trace after mounting NFS is here:
https://bugzilla.kernel.org/attachment.cgi?id=26166
/var/log/messages after soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26167
Quoted 20 lines
I hope that there is any usefull information in there.
Thanks!
Robert
Quoted 13 lines · Trond Myklebust
Quoted 10 lines · Robert Wimmer
Quoted 10 lines
Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished
mounting
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 4 lines
the NFSv4 client.
I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:
After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Does your server have the 'crossmnt' or 'nohide' flags set, or does
it
Quoted 14 lines
Quoted 14 lines
Quoted 14 lines
Quoted 5 lines
use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.
The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash
Thanks!
Robert

Quoted 16 lines

Quoted 5 lines
Quoted 5 lines
That second trace is more than 5.5K deep, more than half of which is socket overhead :-(((.
The process stack does not appear to have overflowed, however that
trace
Quoted 9 lines
Quoted 9 lines
doesn't include any IRQ stack overhead.
OK... So what happens if we get rid of half of that trace by forcing asynchronous tasks such as this to run entirely in rpciod instead of first trying to run in the process context?
See the attachment...

Trond Myklebust

2010-05-06 21:30 UTC #29 lore raw

Sorry. I've been caught up in work in the past few days.

I can certainly help with the soft lockup if you are able to supply either a dump that includes all threads stuck in the NFS, or a (binary) wireshark dump that shows the NFSv4 traffic between the client and server around the time of the hang.

Cheers
Trond

Quoted 121 lines · Robert Wimmer

I don't know if someone is still interested in this but I think Trond isn't further interested because the last error was of cource a "page allocation failure" and not a "soft lookup" which Trond was trying to solve. But the patch was for 2.6.34 and the "soft lookup" comes up only with some 2.6.30 and maybe some 2.6.31 kernel versions. But the first error I reported was a "page allocation failure" which all kernels >= 2.6.32 produces with this configuration I use (NFSv4).
Michael suggested to first solve the "soft lookup" before further investigating the "page allocation failure". We know that the "soft lookup" only pop's up with NFSv4 and not v3. I really want to use v4 but since I'm not a kernel hacker someone must guide me what to try next.
I know that you're all have a lot of other work to do but if there're no ideas left what to do next it's maybe best to close the bug for now and I stay with kernel 2.6.30 for now or go back to NFS v3 if I upgrade to a newer kernel. Maybe the error will be fixed "by accident" in >= 2.6.35 ;-)
Thanks!
Robert
Quoted 89 lines · On 05/03/10 10:11, [email protected] wrote:
Anything we can do to investigate this further?
Thanks!
Robert
Quoted 8 lines · Robert Wimmer
I've applied the patch against the kernel which I got from "git clone ...." resulted in a kernel 2.6.34-rc5.
The stack trace after mounting NFS is here:
https://bugzilla.kernel.org/attachment.cgi?id=26166
/var/log/messages after soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26167
Quoted 20 lines
I hope that there is any usefull information in there.
Thanks!
Robert
Quoted 13 lines · Trond Myklebust
Quoted 10 lines · Robert Wimmer
Quoted 10 lines
Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished
mounting
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 4 lines
the NFSv4 client.
I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:
After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Does your server have the 'crossmnt' or 'nohide' flags set, or does
it
Quoted 14 lines
Quoted 14 lines
Quoted 14 lines
Quoted 5 lines
use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.
The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash
Thanks!
Robert
Quoted 5 lines
Quoted 5 lines
That second trace is more than 5.5K deep, more than half of which is socket overhead :-(((.
The process stack does not appear to have overflowed, however that
trace
Quoted 9 lines
Quoted 9 lines
doesn't include any IRQ stack overhead.
OK... So what happens if we get rid of half of that trace by forcing asynchronous tasks such as this to run entirely in rpciod instead of first trying to run in the process context?
See the attachment...

Robert Wimmer

2010-05-13 21:09 UTC #30 lore raw

Finally I've had some time to do the next test.
Here is a wireshark dump (~750 MByte):
http://213.252.12.93/2.6.34-rc5.cap.gz

dmesg output after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26371

stack trace before page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26369

stack trace after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26370

I hope the wireshark dump is not to big to download. It was created with
tshark -f "tcp port 2049" -i eth0 -w 2.6.34-rc5.cap

Thanks!
Robert

Quoted 150 lines · Trond Myklebust

Sorry. I've been caught up in work in the past few days.
I can certainly help with the soft lockup if you are able to supply either a dump that includes all threads stuck in the NFS, or a (binary) wireshark dump that shows the NFSv4 traffic between the client and server around the time of the hang.
Cheers
Trond
Quoted 136 lines · Robert Wimmer
I don't know if someone is still interested in this but I think Trond isn't further interested because the last error was of cource a "page allocation failure" and not a "soft lookup" which Trond was trying to solve. But the patch was for 2.6.34 and the "soft lookup" comes up only with some 2.6.30 and maybe some 2.6.31 kernel versions. But the first error I reported was a "page allocation failure" which all kernels >= 2.6.32 produces with this configuration I use (NFSv4).
Michael suggested to first solve the "soft lookup" before further investigating the "page allocation failure". We know that the "soft lookup" only pop's up with NFSv4 and not v3. I really want to use v4 but since I'm not a kernel hacker someone must guide me what to try next.
I know that you're all have a lot of other work to do but if there're no ideas left what to do next it's maybe best to close the bug for now and I stay with kernel 2.6.30 for now or go back to NFS v3 if I upgrade to a newer kernel. Maybe the error will be fixed "by accident" in >= 2.6.35 ;-)
Thanks!
Robert
Quoted 103 lines · On 05/03/10 10:11, [email protected] wrote:
Anything we can do to investigate this further?
Thanks!
Robert
Quoted 8 lines · Robert Wimmer
I've applied the patch against the kernel which I got from "git clone ...." resulted in a kernel 2.6.34-rc5.
The stack trace after mounting NFS is here:
https://bugzilla.kernel.org/attachment.cgi?id=26166
/var/log/messages after soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26167
Quoted 23 lines
I hope that there is any usefull information in there.
Thanks!
Robert
Quoted 15 lines · Trond Myklebust
Quoted 11 lines · Robert Wimmer
Quoted 11 lines
Sure. In addition to what you did above, please do
mount -t debugfs none /sys/kernel/debug
and then cat the contents of the pseudofile at
/sys/kernel/debug/tracing/stack_trace
Please do this more or less immediately after you've finished
mounting
Quoted 18 lines
Quoted 18 lines
Quoted 18 lines
Quoted 5 lines
the NFSv4 client.
I've uploaded the stack trace. It was generated
directly after mounting. Here are the stacks:
After mounting:
https://bugzilla.kernel.org/attachment.cgi?id=26153
After the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26154
The dmesg output of the soft lockup:
https://bugzilla.kernel.org/attachment.cgi?id=26155
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Does your server have the 'crossmnt' or 'nohide' flags set, or does
it
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 6 lines
use the 'refer' export option anywhere? If so, then we might have to test further, since those may trigger the NFSv4 submount feature.
The server has the following settings:
rw,nohide,insecure,async,no_subtree_check,no_root_squash
Thanks!
Robert
Quoted 6 lines
Quoted 6 lines
That second trace is more than 5.5K deep, more than half of which is socket overhead :-(((.
The process stack does not appear to have overflowed, however that
trace
Quoted 10 lines
Quoted 10 lines
doesn't include any IRQ stack overhead.
OK... So what happens if we get rid of half of that trace by forcing asynchronous tasks such as this to run entirely in rpciod instead of first trying to run in the process context?
See the attachment...

Trond Myklebust

2010-05-13 21:13 UTC #31 lore raw

Quoted 13 lines · Robert Wimmer

Finally I've had some time to do the next test.
Here is a wireshark dump (~750 MByte):
http://213.252.12.93/2.6.34-rc5.cap.gz
dmesg output after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26371
stack trace before page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26369
stack trace after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26370

Quoted 6 lines

I hope the wireshark dump is not to big to download. It was created with
tshark -f "tcp port 2049" -i eth0 -w 2.6.34-rc5.cap
Thanks!
Robert

Hi Robert,

I tried the above wireshark dump URL, but it appears to point to an empty file.

Cheers
Trond

Robert Wimmer

2010-05-14 05:42 UTC #32 lore raw

Hi Trond,

I'm sorry. There was a Varnish in front of that webserver which doesn't like so big files ;-) Please try this url: http://213.252.12.34/2.6.34-rc5.cap.gz
It work's for me.

Thanks!
Robert

Quoted 22 lines · Trond Myklebust

Quoted 13 lines · Robert Wimmer
Finally I've had some time to do the next test.
Here is a wireshark dump (~750 MByte):
http://213.252.12.93/2.6.34-rc5.cap.gz
dmesg output after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26371
stack trace before page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26369
stack trace after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26370
Quoted 7 lines
I hope the wireshark dump is not to big to download. It was created with
tshark -f "tcp port 2049" -i eth0 -w 2.6.34-rc5.cap
Thanks!
Robert

Quoted 8 lines

Hi Robert,
I tried the above wireshark dump URL, but it appears to point to an empty file.
Cheers
Trond

kernel

2010-05-20 05:39 UTC #33 lore raw

In reply to Trond Myklebust

Hi Trond,

have you had some time to download the wireshark dump?

Thanks!
Robert

Quoted 21 lines · Trond Myklebust

Quoted 13 lines · Robert Wimmer
Finally I've had some time to do the next test.
Here is a wireshark dump (~750 MByte):
http://213.252.12.93/2.6.34-rc5.cap.gz
dmesg output after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26371
stack trace before page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26369
stack trace after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26370
Quoted 6 lines
I hope the wireshark dump is not to big to download. It was created with
tshark -f "tcp port 2049" -i eth0 -w 2.6.34-rc5.cap
Thanks!
Robert

Quoted 7 lines

Hi Robert,
I tried the above wireshark dump URL, but it appears to point to an empty file.
Cheers
Trond

Robert Wimmer

2010-05-25 20:01 UTC #34 lore raw

Hi Trond,

just a little reminder ;-)

Thanks!
Robert

Quoted 40 lines · On 05/20/10 09:39, [email protected] wrote:

Hi Trond,
have you had some time to download the wireshark dump?
Thanks!
Robert
Quoted 30 lines · Trond Myklebust
Quoted 13 lines · Robert Wimmer
Finally I've had some time to do the next test.
Here is a wireshark dump (~750 MByte):
http://213.252.12.93/2.6.34-rc5.cap.gz
dmesg output after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26371
stack trace before page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26369
stack trace after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26370
Quoted 7 lines
I hope the wireshark dump is not to big to download. It was created with
tshark -f "tcp port 2049" -i eth0 -w 2.6.34-rc5.cap
Thanks!
Robert
Hi Robert,
I tried the above wireshark dump URL, but it appears to point to an empty file.
Cheers
Trond

kernel

2010-06-02 09:57 UTC #35 lore raw

Hi Trond,

currently it seems that the problem was fixed by accident... ;-) Since 2.6.34 is now in Gentoo portage I thought I should give it a try. Using my 2.6.35-r5 .config the 2.6.34 release is now working for 4 hours (instead of 5-10 minutes before). Hmmm... Hopefully it will run for some more hours and days now. Since I've definitely changed nothing besides the kernel it must have been fixed (hopefully) in one of the 2.6.34-rc's.

If it's still running tomorrow I'll close the bug.

Greetings
Robert

Quoted 48 lines · Robert Wimmer

Hi Trond,
just a little reminder ;-)
Thanks!
Robert
Quoted 40 lines · On 05/20/10 09:39, [email protected] wrote:
Hi Trond,
have you had some time to download the wireshark dump?
Thanks!
Robert
Quoted 30 lines · Trond Myklebust
Quoted 13 lines · Robert Wimmer
Finally I've had some time to do the next test.
Here is a wireshark dump (~750 MByte):
http://213.252.12.93/2.6.34-rc5.cap.gz
dmesg output after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26371
stack trace before page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26369
stack trace after page allocation failure:
https://bugzilla.kernel.org/attachment.cgi?id=26370
Quoted 7 lines
I hope the wireshark dump is not to big to download. It was created with
tshark -f "tcp port 2049" -i eth0 -w 2.6.34-rc5.cap
Thanks!
Robert
Hi Robert,
I tried the above wireshark dump URL, but it appears to point to an empty file.
Cheers
Trond