Re: BUG: sleeping function called from invalid context at mm/slub.c:1717

Re: BUG: sleeping function called from invalid context at mm/slub.c:1717

[Minchan Kim]

Hi, Jiri.

On Tue, Sep 15, 2009 at 7:10 PM, Jiri Kosina <jkosina@suse.cz> wrote:
> On Tue, 15 Sep 2009, Jens Axboe wrote:
>
>> This is new with todays -git:
>>
>> BUG: sleeping function called from invalid context at mm/slub.c:1717
>> in_atomic(): 1, irqs_disabled(): 1, pid: 0, name: swapper
>> Pid: 0, comm: swapper Not tainted 2.6.31 #206
>> Call Trace:
>>  <IRQ>  [<ffffffff8103eb23>] __might_sleep+0xf3/0x110
>>  [<ffffffff810e4d83>] kmem_cache_alloc+0x123/0x170
>>  [<ffffffff813306c9>] hid_input_report+0x89/0x3a0
>>  [<ffffffffa00cd5f4>] hid_ctrl+0xa4/0x1f0 [usbhid]
>>  [<ffffffff8108a4c7>] ? handle_IRQ_event+0xa7/0x1e0
>>  [<ffffffffa004da1f>] usb_hcd_giveback_urb+0x3f/0xa0 [usbcore]
>>  [<ffffffffa0074ab4>] uhci_giveback_urb+0xb4/0x240 [uhci_hcd]
>>  [<ffffffffa00750e7>] uhci_scan_schedule+0x357/0xab0 [uhci_hcd]
>>  [<ffffffffa0077a01>] uhci_irq+0x91/0x190 [uhci_hcd]
>>  [<ffffffffa004d44e>] usb_hcd_irq+0x2e/0x70 [usbcore]
>>  [<ffffffff8108a4c7>] handle_IRQ_event+0xa7/0x1e0
>>  [<ffffffff8108c58c>] handle_fasteoi_irq+0x7c/0xf0
>>  [<ffffffff8100f176>] handle_irq+0x46/0xa0
>>  [<ffffffff8100e49a>] do_IRQ+0x6a/0xf0
>>  [<ffffffff8100c853>] ret_from_intr+0x0/0xa
>>
>> And I notice there's a HID merge from yesterday, Jiri CC'ed.
>
> Thanks for letting me know. The patch below should fix it.
>
>
>
> From: Jiri Kosina <jkosina@suse.cz>
> Subject: [PATCH] HID: fix non-atomic allocation in hid_input_report
>
> 'interrupt' variable can't be used to safely determine whether
> we are running in atomic context or not, as we might be called from
> during control transfer completion through hid_ctrl() in atomic
> context with interrupt == 0.

I am not a USB expert so It might be dump comment. :)

We have to change description of hid_input_report.

 * @interrupt: called from atomic?

I think it lost meaning.
I am worried that interrupt variable is propagated down
to sub functions. Is it right on sub functions?

One more thing, I am concerned about increasing
GFP_ATOMIC customers although we can avoid it.
Is it called rarely?
Could you find a alternative method to overcome this issue?


>
> Reported-by: Jens Axboe <jens.axboe@oracle.com>
> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
> ---
>  drivers/hid/hid-core.c |    3 +--
>  1 files changed, 1 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/hid/hid-core.c b/drivers/hid/hid-core.c
> index 342b7d3..ca9bb26 100644
> --- a/drivers/hid/hid-core.c
> +++ b/drivers/hid/hid-core.c
> @@ -1089,8 +1089,7 @@ int hid_input_report(struct hid_device *hid, int type, u8 *data, int size, int i
>                return -1;
>        }
>
> -       buf = kmalloc(sizeof(char) * HID_DEBUG_BUFSIZE,
> -                       interrupt ? GFP_ATOMIC : GFP_KERNEL);
> +       buf = kmalloc(sizeof(char) * HID_DEBUG_BUFSIZE, GFP_ATOMIC);
>
>        if (!buf) {
>                report = hid_get_report(report_enum, data);
> --
> 1.5.6
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>



-- 
Kind regards,
Minchan Kim

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

Re: BUG: sleeping function called from invalid context at mm/slub.c:1717

[Jiri Kosina]

On Wed, 16 Sep 2009, Minchan Kim wrote:

> We have to change description of hid_input_report.
> 
>  * @interrupt: called from atomic?
> I think it lost meaning.

Good point, I will change it, thanks.

> I am worried that interrupt variable is propagated down to sub 
> functions. Is it right on sub functions?

Yes. This variable is not used for chosing correct allocation flags 
anywhere else, it is just carrying the semantics what the HID core should 
do, what callbacks to call, etc. So it's correct.
But you are right that the name and kerneldocs is confusing, and I will 
fix that.

> One more thing, I am concerned about increasing GFP_ATOMIC customers 
> although we can avoid it. Is it called rarely? Could you find a 
> alternative method to overcome this issue?

This is just a temporary buffer for debugging output, it is freed almost 
immediately later in the function, and even if the allocation fails, 
nothing bad happens (just the debugging output is not delivered to the 
debugfs buffer).

Thanks,

-- 
Jiri Kosina
SUSE Labs, Novell Inc.

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