Skip to contentSkip to conversation

2.6.20-rc2-mm1 -- BUG: at arch/i386/mm/highmem.c:50 kmap_atomic()

DISCUSSION Miles Lane 2 participants 2 messages Dec 29, 2006

lore
Share & tools
lore

Miles Lane

author
#0 lore raw
Quoted 30 lines · Jiri Slaby
Quoted 29 lines · Miles Lane
Quoted 27 lines · Jiri Slaby
Quoted 15 lines · Miles Lane

Sorry, I am not finding who maintains highmem. Please forward.

WARNING (1) at arch/i386/mm/highmem.c:41 kmap_atomic()

[<c0103c25>] dump_trace+0x68/0x1d2
[<c0103da7>] show_trace_log_lvl+0x18/0x2c
[<c0104410>] show_trace+0xf/0x11
[<c010449b>] dump_stack+0x12/0x14
[<c01144d9>] kmap_atomic+0x6f/0x1ca
[<f930e25d>] ntfs_end_buffer_async_read+0x25d/0x2ca [ntfs]
[<c017c294>] end_bio_bh_io_sync+0x2c/0x37
[<c017dc29>] bio_endio+0x5a/0x62
[<c01c8412>] __end_that_request_first+0x145/0x3ab
[<c0237695>] ide_end_request+0x80/0xd8
[<c023e3f0>] ide_dma_intr+0x55/0x9a
[<c02388dc>] ide_intr+0x182/0x1f2
Quoted 8 lines
[<c0140775>] handle_IRQ_event+0x1a/0x3f
[<c0141baa>] handle_edge_irq+0xc6/0x11c
[<c0105416>] do_IRQ+0x57/0x71
[<c010366b>] common_interrupt+0x23/0x28
[<f8826ee4>] acpi_processor_idle+0x1cc/0x36c [processor]
[<c010132b>] cpu_idle+0x3e/0x6c
[<c03f06d9>] start_kernel+0x2fa/0x2fe
=======================

Reported yet, you might see it here:
http://lkml.org/lkml/2006/12/15/222

Quoted 7 lines
Quoted 3 lines

It is certainly very similar, and probably has the same root cause.
Though, the trace isn't an exact match. So, who should look into
this?

The trace needn't be the same. The problem was, that kmap_atomic didn't know KM_BIO_SRC_IRQ which is called (twice) from ntfs_end_buffer_async_read. It doesn't matter who called this ntfs function and why.

With 2.6.20-rc2-mm1 I am seeing:

BUG: at arch/i386/mm/highmem.c:50 kmap_atomic()
[<c0114368>] kmap_atomic+0x8e/0x18b
[<f938525d>] ntfs_end_buffer_async_read+0x25d/0x2ca [ntfs]
[<c017a205>] end_bio_bh_io_sync+0x0/0x37
[<c017a231>] end_bio_bh_io_sync+0x2c/0x37
[<c017bb95>] bio_endio+0x5a/0x62
[<c0237da2>] scsi_delete_timer+0xb/0x1b
[<c01c641c>] __end_that_request_first+0x145/0x3ab
[<c01cb657>] as_put_io_context+0x43/0x4e
[<c0238ec7>] scsi_end_request+0x1d/0xb6
[<c0239092>] scsi_io_completion+0xf5/0x2ba
[<c011ff63>] do_timer+0x4fc/0x70c
[<f90024b6>] sd_rw_intr+0x15d/0x186 [sd_mod]
[<c023970f>] scsi_softirq_done+0x20/0xce
[<c02358cd>] scsi_finish_command+0x3f/0x43
[<c01c7f4c>] blk_done_softirq+0x4a/0x55
[<c011c1d1>] __do_softirq+0x35/0x75
[<c011c233>] do_softirq+0x22/0x26
[<c011c3fa>] irq_exit+0x29/0x62
[<c0106232>] do_IRQ+0x67/0x81
[<c01046e3>] common_interrupt+0x23/0x28
[<f8823ee4>] acpi_processor_idle+0x1cc/0x36c [processor]
[<f8823d18>] acpi_processor_idle+0x0/0x36c [processor]
[<c0102389>] cpu_idle+0x44/0x77
[<c0380a93>] start_kernel+0x2f9/0x2fd
[<c038042b>] unknown_bootoption+0x0/0x202
#1 lore raw
Quoted 35 lines · Miles Lane
Quoted 34 lines · Jiri Slaby
Quoted 32 lines · Miles Lane
Quoted 25 lines · Jiri Slaby
Quoted 15 lines · Miles Lane

Sorry, I am not finding who maintains highmem. Please forward.

WARNING (1) at arch/i386/mm/highmem.c:41 kmap_atomic()

[<c0103c25>] dump_trace+0x68/0x1d2
[<c0103da7>] show_trace_log_lvl+0x18/0x2c
[<c0104410>] show_trace+0xf/0x11
[<c010449b>] dump_stack+0x12/0x14
[<c01144d9>] kmap_atomic+0x6f/0x1ca
[<f930e25d>] ntfs_end_buffer_async_read+0x25d/0x2ca [ntfs]
[<c017c294>] end_bio_bh_io_sync+0x2c/0x37
[<c017dc29>] bio_endio+0x5a/0x62
[<c01c8412>] __end_that_request_first+0x145/0x3ab
[<c0237695>] ide_end_request+0x80/0xd8
[<c023e3f0>] ide_dma_intr+0x55/0x9a
[<c02388dc>] ide_intr+0x182/0x1f2
Quoted 8 lines
[<c0140775>] handle_IRQ_event+0x1a/0x3f
[<c0141baa>] handle_edge_irq+0xc6/0x11c
[<c0105416>] do_IRQ+0x57/0x71
[<c010366b>] common_interrupt+0x23/0x28
[<f8826ee4>] acpi_processor_idle+0x1cc/0x36c [processor]
[<c010132b>] cpu_idle+0x3e/0x6c
[<c03f06d9>] start_kernel+0x2fa/0x2fe
=======================
Quoted 2 lines

Reported yet, you might see it here:
http://lkml.org/lkml/2006/12/15/222

It is certainly very similar, and probably has the same root cause.
Though, the trace isn't an exact match. So, who should look into
this?

Quoted 5 lines
Quoted 5 lines

The trace needn't be the same. The problem was, that kmap_atomic didn't know
KM_BIO_SRC_IRQ which is called (twice) from ntfs_end_buffer_async_read. It
doesn't matter who called this ntfs function and why.

Oh sorry, my fault -- it does matter now: hardirq/softirq path.

Quoted 20 lines

With 2.6.20-rc2-mm1 I am seeing:

BUG: at arch/i386/mm/highmem.c:50 kmap_atomic()
[<c0114368>] kmap_atomic+0x8e/0x18b
[<f938525d>] ntfs_end_buffer_async_read+0x25d/0x2ca [ntfs]
[<c017a205>] end_bio_bh_io_sync+0x0/0x37
[<c017a231>] end_bio_bh_io_sync+0x2c/0x37
[<c017bb95>] bio_endio+0x5a/0x62
[<c0237da2>] scsi_delete_timer+0xb/0x1b
[<c01c641c>] __end_that_request_first+0x145/0x3ab
[<c01cb657>] as_put_io_context+0x43/0x4e
[<c0238ec7>] scsi_end_request+0x1d/0xb6
[<c0239092>] scsi_io_completion+0xf5/0x2ba
[<c011ff63>] do_timer+0x4fc/0x70c
[<f90024b6>] sd_rw_intr+0x15d/0x186 [sd_mod]
[<c023970f>] scsi_softirq_done+0x20/0xce
[<c02358cd>] scsi_finish_command+0x3f/0x43
[<c01c7f4c>] blk_done_softirq+0x4a/0x55
[<c011c1d1>] __do_softirq+0x35/0x75
[<c011c233>] do_softirq+0x22/0x26
Quoted 8 lines
[<c011c3fa>] irq_exit+0x29/0x62
[<c0106232>] do_IRQ+0x67/0x81
[<c01046e3>] common_interrupt+0x23/0x28
[<f8823ee4>] acpi_processor_idle+0x1cc/0x36c [processor]
[<f8823d18>] acpi_processor_idle+0x0/0x36c [processor]
[<c0102389>] cpu_idle+0x44/0x77
[<c0380a93>] start_kernel+0x2f9/0x2fd
[<c038042b>] unknown_bootoption+0x0/0x202

Yes, me too. But
type != KM_BIO_SRC_IRQ && type != KM_BIO_DST_IRQ) {
line appeared in in_irq() part in this kernel. This is OK for some cases -- when it's called through hardirq path. softirq remains unresolved.

regards,

Signature · 4 lines
http://www.fi.muni.cz/~xslaby/            Jiri Slaby
faculty of informatics, masaryk university, brno, cz
e-mail: jirislaby gmail com, gpg pubkey fingerprint:
B674 9967 0407 CE62 ACC8  22A0 32CC 55C3 39D4 7A7E