Re: 2.6.21-rc7-mm1 BUG at kernel/sched-clock.c:175 init_sched_clock()

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 25 Apr 2007 17:42:41 +0900 Tejun Heo <htejun@gmail.com> wrote:

> Andrew Morton wrote:
> >> [    2.953862] RIP: 0010:[<ffffffff803ae98b>]  [<ffffffff803ae98b>] scsi_schedule_eh+0xa/0x57
> >> [    3.058672]  [<ffffffff803e1e01>] ata_port_schedule_eh+0x4c/0x50
> >> [    3.064725]  [<ffffffff803e1ea7>] ata_port_abort+0xa2/0xae
> >> [    3.070248]  [<ffffffff803e1ef9>] ata_port_freeze+0x46/0x57
> >> [    3.075853]  [<ffffffff803e4c39>] ahci_interrupt+0x300/0x47a
> >> [    3.081552]  [<ffffffff8020eb83>] handle_IRQ_event+0x27/0x57
> >> [    3.087253]  [<ffffffff8029a021>] handle_edge_irq+0xee/0x133
> >> [    3.092960]  [<ffffffff8025f4e1>] do_IRQ+0x6d/0xd5
> >> [    3.097793]  [<ffffffff80255071>] ret_from_intr+0x0/0xa
> >> [    3.103059]  [<ffffffff8024e88b>] mwait_idle+0x46/0x4b
> >> [    3.108231]  [<ffffffff802422f4>] cpu_idle+0x87/0xaa
> >> [    3.113227]  [<ffffffff8025c988>] rest_init+0x49/0x4b
> >> [    3.118322]  [<ffffffff805dca5d>] start_kernel+0x291/0x29c
> >> [    3.123837]  [<ffffffff805dc13a>] _sinittext+0x13a/0x141
> >>
> > 
> > So we took an AHCI interrupt when ata_port.scsi_host was still NULL.
> > 
> > It appears that ATA is presently requesting its IRQ before allocating all
> > the resources which are needed to handle an interrupt.  Does this
> > (resource-leaky) patch fix things?
> 
> No, that would break host probing.  The port is in frozen (controller
> initialized and IRQs masked off) state, so it's not allowed to take
> interrupt.

Sounds dodgy.  What happens if the IRQ line is shared with some other
device?  We'll enter ahci_interrupt().  We've alread set up ->port_map and
->n_ports so we're wholly dependent upon that read from HOST_IRQ_STAT
returning zeroes.

        /* sigh.  0xffffffff is a valid return from h/w */

if that happens we're dead, aren't we?


>  If interrupt triggers at this point, it's low level driver
> bug.  Berck, how reliably can you reproduce this problem?  Can you post
> the result of 'lspci -nn'?

Berck doesn't apepar to be sharing that IRQ, so it'll be something else.

However do note that if CONFIG_DEBUG_SHIRQ is enabled, request_irq() will
deliberately run the handler, to catch buggy drivers.