On Thu, Apr 27, 2017 at 06:34:07PM +0300, Alexander Gerasiov wrote:
> Hello Julia,
>
> On Tue, 25 Apr 2017 16:09:29 -0500
> Julia Cartwright <julia@xxxxxx> wrote:
>
> > Looking at this some more, I imagine the callpath which is problematic
> > is:
> >
> > <writeback kworker>
> > ...
> > add_stripe_bio()
> > stripe_add_to_batch_list()
> > lock_two_stripes()
> > local_irq_disable()
> > spin_lock(&sh{1,2}->stripe_lock) --> ___might_sleep() -->
> > BOOM
> >
> > That is, it's the local_irq_disable() in lock_two_stripes() which is
> > the problem. That's my gut feel, anyway. Can you give this patch a
> > try and see if works? You'll probably still want the second patch in
> > your series.
>
> Ouch, you are absolutely right, it was not atomic context this time, so
> changing spinlocks from mutex to real ones is not necessary. My fault.
>
> Do you want me to resubmit the patch, or do it yourself?
If you can confirm that the proposed patch fixes the problem you are
seeing, I'll send a proper patch.
Thanks,
Julia
PS. Also, interestingly, this is not an altogether uncommon problem:
virtual report
@r exists@
position p1;
position p2;
@@
(
local_irq_disable@p1
|
local_irq_save@p1
)(...);
... when != local_irq_enable(...)
when != local_irq_restore(...)
(
spin_lock@p2
|
spin_lock_irq@p2
|
spin_lock_irqsave@p2
)(...);
@script:python depends on r && report@
p1 << r.p1;
p2 << r.p2;
@@
msg="ERROR: sleeping spinlock acquired, though interrupts disabled on line %s" % (p1[0].line)
coccilib.report.print_report(p2[0], msg)
Reports:
drivers/tty/serial/pch_uart.c:1672:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 1662
drivers/tty/serial/st-asc.c:812:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 806
drivers/tty/serial/meson_uart.c:497:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 491
drivers/tty/serial/bcm63xx_uart.c:713:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 706
drivers/tty/serial/ar933x_uart.c:555:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 548
drivers/tty/serial/stm32-usart.c:940:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 934
drivers/net/ethernet/tehuti/tehuti.c:1632:1-10: ERROR: sleeping spinlock acquired, though interrupts disabled on line 1631
drivers/tty/serial/pxa.c:659:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 653
drivers/tty/serial/lpc32xx_hs.c:153:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 147
arch/x86/kernel/reboot.c:98:1-10: ERROR: sleeping spinlock acquired, though interrupts disabled on line 86
net/core/drop_monitor.c:171:1-10: ERROR: sleeping spinlock acquired, though interrupts disabled on line 169
drivers/tty/serial/sh-sci.c:2800:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 2791
drivers/usb/gadget/udc/dummy_hcd.c:755:1-10: ERROR: sleeping spinlock acquired, though interrupts disabled on line 754
arch/alpha/kernel/srmcons.c:81:1-10: ERROR: sleeping spinlock acquired, though interrupts disabled on line 74
lib/dma-debug.c:946:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 943
arch/mn10300/kernel/mn10300-serial.c:1594:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 1587
arch/metag/kernel/smp.c:451:2-11: ERROR: sleeping spinlock acquired, though interrupts disabled on line 446
There isn't an entirely straightforward way to fix each of these; it will
require some auditing, if anyone is looking for more work to do :).
Attachment:
signature.asc
Description: PGP signature
