OSDN Git Service

ASoC: SOF: Intel: hda-stream: fix a deadlock with bus->reg_lock
authorKeyon Jie <yang.jie@linux.intel.com>
Wed, 12 Jun 2019 17:23:41 +0000 (12:23 -0500)
committerMark Brown <broonie@kernel.org>
Mon, 17 Jun 2019 12:44:50 +0000 (13:44 +0100)
commit7fd572e7d317fa51049d623badb8b2874bfd0119
tree1bf4759fdf3345c0e519ea77cc14d8f8661ccbc7
parent6297a0dc4c14a62bea5a9137ceef280cb7a80665
ASoC: SOF: Intel: hda-stream: fix a deadlock with bus->reg_lock

We should use irq disabled mode when read/write hda registers from
thread context, as we need to hold the same bus->reg_lock in interrupt
context hda_dsp_stream_interrupt(), otherwise, when we are holding the
lock in hda_dsp_stream_hw_free() and the interrupt arrives, we will get
deadlock in the interrupt handler.

Error logs like this:

[    5.603606]        CPU0
[    5.603606]        ----
[    5.603607]   lock(&(&bus->reg_lock)->rlock);
[    5.603608]   <Interrupt>
[    5.603609]     lock(&(&bus->reg_lock)->rlock);
[    5.603610]
                *** DEADLOCK ***

[    5.603611] 2 locks held by pulseaudio/2329:
[    5.603612]  #0: 000000005fcf26c6 (&card->mutex/1){+.+.}, at: dpcm_fe_dai_hw_free+0x2b/0x110 [snd_soc_core]
[    5.603619]  #1: 00000000ef369faf (&rtd->pcm_mutex){+.+.}, at: soc_pcm_hw_free+0x2e/0x1c0 [snd_soc_core]

The fix is simple, let's switch to use spin_lock/unlock_irq().

Reported-by: Xun Zhang <xun2.zhang@intel.com>
Signed-off-by: Keyon Jie <yang.jie@linux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
sound/soc/sof/intel/hda-stream.c