At Mon, 19 Nov 2012 11:56:32 +0100, Julian Wollrath wrote:
I have an Thinkpad X121e with the following soundcard: 00:01.1 Audio device [0403]: Advanced Micro Devices [AMD] nee ATI Wrestler HDMI Audio [Radeon HD 6250/6310] [1002:1314]
Of course I did not mean the hdmi device but this one, sorry: 00:14.2 Audio device [0403]: Advanced Micro Devices [AMD] nee ATI SBx00 Azalia (Intel HDA) [1002:4383] (rev 40)
So, for this particular issue (at least the recursive suspend call), I think the easiest way is a patch like below. This is no ideal solution but would be good enough as a bandaid at this late stage for 3.7. Julian, could you check whether this works?
Sadly the patch does not work, I still get the spurious responses
Note that spurious responses won't be fixed by this patch at all. It's a different problem from the hung. The patch tried only avoiding the recursive suspend call.
and a trace, which looked minimaly different this time: [ 240.434881] INFO: task kworker/u:14:3617 blocked for more than 120 seconds. [ 240.434894] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 240.434901] kworker/u:14 D ffff88011ed128c0 0 3617 2 0x00000000 [ 240.434918] ffff880106272cd0 0000000000000046 0000000000000086 ffff88011a6e7950 [ 240.434931] ffff8801178f7fd8 ffff8801178f7fd8 ffff8801178f7fd8 ffff880106272cd0 [ 240.434944] 0000000000000003 ffff8801178f7c20 ffff8801178f7c18 ffff880106272cd0 [ 240.434957] Call Trace: [ 240.434978] [<ffffffff81403342>] ? schedule_timeout+0x1e2/0x2a0 [ 240.434992] [<ffffffff8106c32d>] ? select_task_rq_fair+0x53d/0x5d0 [ 240.435004] [<ffffffff814049c2>] ? wait_for_common+0xd2/0x180 [ 240.435017] [<ffffffff810693f0>] ? try_to_wake_up+0x270/0x270 [ 240.435030] [<ffffffff81052911>] ? flush_work+0xe1/0x160 [ 240.435041] [<ffffffff81051c10>] ? gcwq_release_assoc_and_unlock+0x40/0x40 [ 240.435052] [<ffffffff8105369d>] ? __cancel_work_timer+0x5d/0xa0 [ 240.435093] [<ffffffffa0208f5c>] ? hda_call_codec_suspend+0x14c/0x160 [snd_hda_codec] [ 240.435114] [<ffffffffa0208fb0>] ? snd_hda_suspend+0x40/0x60 [snd_hda_codec] [ 240.435131] [<ffffffffa01e06cb>] ? azx_bus_reset+0xab/0xc0 [snd_hda_intel] [ 240.435152] [<ffffffffa0207e55>] ? codec_exec_verb+0x185/0x1c0 [snd_hda_codec] [ 240.435173] [<ffffffffa0207ee8>] ? snd_hda_codec_read+0x58/0x90 [snd_hda_codec] [ 240.435193] [<ffffffffa0208b68>] ? snd_hda_shutup_pins+0x58/0x70 [snd_hda_codec] [ 240.435210] [<ffffffffa033e369>] ? conexant_suspend+0x9/0x10 [snd_hda_codec_conexant] [ 240.435229] [<ffffffffa0208e2e>] ? hda_call_codec_suspend+0x1e/0x160 [snd_hda_codec] [ 240.435249] [<ffffffffa0209048>] ? hda_power_work+0x78/0xc0 [snd_hda_codec] [ 240.435260] [<ffffffff81052f06>] ? process_one_work+0x126/0x490 [ 240.435280] [<ffffffffa0208fd0>] ? snd_hda_suspend+0x60/0x60 [snd_hda_codec] [ 240.435291] [<ffffffff81054c3d>] ? worker_thread+0x15d/0x450 [ 240.435302] [<ffffffff81054ae0>] ? flush_delayed_work+0x40/0x40 [ 240.435313] [<ffffffff810599a3>] ? kthread+0xb3/0xc0 [ 240.435323] [<ffffffff810598f0>] ? kthread_create_on_node+0x110/0x110 [ 240.435335] [<ffffffff814061ac>] ? ret_from_fork+0x7c/0xb0 [ 240.435345] [<ffffffff810598f0>] ? kthread_create_on_node+0x110/0x110
OK, then this code path is irrelevant with the real suspend/resume, but it's in a normal power-saving path.
Could you try the patch below instead?
By the way, it seems like I get the traces every 120 seconds and not only once, forgot to mention that earlier.
The spurious call (or no proper reply for the verb) indicates that either the HD-audio controller or the HD-audio codec stalls. Is yours a machine with a discrete GPU? If so, and if you are using vga-switcheroo or anything manipulating the D-GPU, it's possibly the cause of the stall.
thanks,
Takashi
--- diff --git a/sound/pci/hda/hda_codec.c b/sound/pci/hda/hda_codec.c index a8e7b00..d9fd439 100644 --- a/sound/pci/hda/hda_codec.c +++ b/sound/pci/hda/hda_codec.c @@ -228,7 +228,7 @@ static int codec_exec_verb(struct hda_codec *codec, unsigned int cmd, } mutex_unlock(&bus->cmd_mutex); snd_hda_power_down(codec); - if (res && *res == -1 && bus->rirb_error) { + if (!codec->in_pm && res && *res == -1 && bus->rirb_error) { if (bus->response_reset) { snd_printd("hda_codec: resetting BUS due to " "fatal communication error\n"); @@ -238,7 +238,7 @@ static int codec_exec_verb(struct hda_codec *codec, unsigned int cmd, goto again; } /* clear reset-flag when the communication gets recovered */ - if (!err) + if (!err || codec->in_pm) bus->response_reset = 0; return err; } @@ -3655,6 +3655,8 @@ static unsigned int hda_call_codec_suspend(struct hda_codec *codec, bool in_wq) { unsigned int state;
+ codec->in_pm = 1; + if (codec->patch_ops.suspend) codec->patch_ops.suspend(codec); hda_cleanup_all_streams(codec); @@ -3669,6 +3671,7 @@ static unsigned int hda_call_codec_suspend(struct hda_codec *codec, bool in_wq) codec->power_transition = 0; codec->power_jiffies = jiffies; spin_unlock(&codec->power_lock); + codec->in_pm = 0; return state; }
@@ -3677,6 +3680,8 @@ static unsigned int hda_call_codec_suspend(struct hda_codec *codec, bool in_wq) */ static void hda_call_codec_resume(struct hda_codec *codec) { + codec->in_pm = 1; + /* set as if powered on for avoiding re-entering the resume * in the resume / power-save sequence */ @@ -3700,6 +3705,8 @@ static void hda_call_codec_resume(struct hda_codec *codec) snd_hda_jack_set_dirty_all(codec); snd_hda_jack_report_sync(codec); } + + codec->in_pm = 0; snd_hda_power_down(codec); /* flag down before returning */ } #endif /* CONFIG_PM */ diff --git a/sound/pci/hda/hda_codec.h b/sound/pci/hda/hda_codec.h index baad7bf..8665540 100644 --- a/sound/pci/hda/hda_codec.h +++ b/sound/pci/hda/hda_codec.h @@ -871,6 +871,7 @@ struct hda_codec { unsigned int power_on :1; /* current (global) power-state */ unsigned int d3_stop_clk:1; /* support D3 operation without BCLK */ unsigned int pm_down_notified:1; /* PM notified to controller */ + unsigned int in_pm:1; /* suspend/resume being performed */ int power_transition; /* power-state in transition */ int power_count; /* current (global) power refcount */ struct delayed_work power_work; /* delayed task for powerdown */