Re: [PATCH] hid: hid-core: Fix a sleep-in-atomic-context bug in __hid_request()

[Jia-Ju Bai]

On 2018/9/5 16:29, Jiri Kosina wrote:
> On Sat, 1 Sep 2018, Jia-Ju Bai wrote:
>
> > The driver may sleep with holding a spinlock.
> >
> > The function call paths (from bottom to top) in Linux-4.16 are:
> >
> > [FUNC] hid_alloc_report_buf(GFP_KERNEL)
> > drivers/hid/hid-core.c, 1435:
> > 	hid_alloc_report_buf in __hid_request
> > ./include/linux/hid.h, 1023:
> > 	__hid_request in hid_hw_request
> > drivers/hid/hid-picolcd_core.c, 111:
> > 	hid_hw_request in picolcd_send_and_wait
> > drivers/hid/hid-picolcd_core.c, 100:
> > 	_raw_spin_lock_irqsave in picolcd_send_and_wait
> >
> > [FUNC] hid_alloc_report_buf(GFP_KERNEL)
> > drivers/hid/hid-core.c, 1435:
> > 	hid_alloc_report_buf in __hid_request
> > ./include/linux/hid.h, 1023:
> > 	__hid_request in hid_hw_request
> > drivers/hid/hid-picolcd_core.c, 245:
> > 	hid_hw_request in picolcd_reset
> > drivers/hid/hid-picolcd_core.c, 235:
> > 	_raw_spin_lock_irqsave in picolcd_reset
> >
> > [FUNC] hid_alloc_report_buf(GFP_KERNEL)
> > drivers/hid/hid-core.c, 1435:
> > 	hid_alloc_report_buf in __hid_request
> > ./include/linux/hid.h, 1023:
> > 	__hid_request in hid_hw_request
> > drivers/hid/hid-picolcd_fb.c, 215:
> > 	hid_hw_request in picolcd_fb_reset
> > drivers/hid/hid-picolcd_fb.c, 206:
> > 	_raw_spin_lock_irqsave in picolcd_fb_reset
> >
> > [FUNC] hid_alloc_report_buf(GFP_KERNEL)
> > drivers/hid/hid-core.c, 1435:
> > 	hid_alloc_report_buf in __hid_request
> > ./include/linux/hid.h, 1023:
> > 	__hid_request in hid_hw_request
> > drivers/hid/hid-lg4ff.c, 465:
> > 	hid_hw_request in lg4ff_play
> > drivers/hid/hid-lg4ff.c, 441:
> > 	_raw_spin_lock_irqsave in lg4ff_play
> >
> > To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC.
> >
> > This bug is found by my static analysis tool DSAC.
> Could you please rewrite the changelog so that it's human readable? The
> above is a bit hard to understand, I think something along the lines of
> "__hid_request() has to be allocating with GFP_ATOMIC because there are
> the following callchains leading to __hid_request() being an atomic
> context: ... a->b->c.._hid_request()" etc.

Okay, I will send a new patch.


Best wishes,
Jia-Ju Bai