| Message ID | 20191023161203.28955-2-jbrunet@baylibre.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 901af18b6baade6a327e532427cbb233f4945f5d |
| Headers | show |
| Series | ASoC: hdmi-codec: fix locking issue | expand |
On Wed, Oct 23, 2019 at 06:12:02PM +0200, Jerome Brunet wrote: > This reverts commit eb1ecadb7f67dde94ef0efd3ddaed5cb6c9a65ed. > > This fixes the following warning reported by lockdep and a potential > issue with hibernation Please submit patches using subject lines reflecting the style for the subsystem, this makes it easier for people to identify relevant patches. Look at what existing commits in the area you're changing are doing and make sure your subject lines visually resemble what they're doing. There's no need to resubmit to fix this alone.
On Wed, Oct 23, 2019 at 05:37:16PM +0100, Mark Brown wrote: > On Wed, Oct 23, 2019 at 06:12:02PM +0200, Jerome Brunet wrote: > > This reverts commit eb1ecadb7f67dde94ef0efd3ddaed5cb6c9a65ed. > > > > This fixes the following warning reported by lockdep and a potential > > issue with hibernation > > Please submit patches using subject lines reflecting the style for the > subsystem, this makes it easier for people to identify relevant patches. > Look at what existing commits in the area you're changing are doing and > make sure your subject lines visually resemble what they're doing. > There's no need to resubmit to fix this alone. Hi Mark, If you look at the git log for reverted commits, the vast majority of them follow _this_ style. From 5.3 back to the start of current git history, there are 3665 commits with "Revert" in their subject line, 3050 of those start with "Revert" with no subsystem prefix. It seems that there are a small number of subsystems that want something different, ASoC included. That will be an ongoing problem, people won't remember which want it when the majority don't. Maybe the revert format should be standardised in some manner?
On Wed, Oct 23, 2019 at 07:26:18PM +0100, Russell King - ARM Linux admin wrote: > If you look at the git log for reverted commits, the vast majority > of them follow _this_ style. From 5.3 back to the start of current > git history, there are 3665 commits with "Revert" in their subject > line, 3050 of those start with "Revert" with no subsystem prefix. That's assuming that all reverts have Revert in their subject line of course! > It seems that there are a small number of subsystems that want > something different, ASoC included. That will be an ongoing problem, > people won't remember which want it when the majority don't. > Maybe the revert format should be standardised in some manner? My general thought on this is that reverts are commits like any other and that the documentation for how you're supposed to do commit messages also applies to them, might be worth a note though as I do see people not writing a commit log at all for them sometimes.
diff --git a/sound/soc/codecs/hdmi-codec.c b/sound/soc/codecs/hdmi-codec.c index b5fd8f08726e..f8b5b960e597 100644 --- a/sound/soc/codecs/hdmi-codec.c +++ b/sound/soc/codecs/hdmi-codec.c @@ -274,7 +274,7 @@ struct hdmi_codec_priv { uint8_t eld[MAX_ELD_BYTES]; struct snd_pcm_chmap *chmap_info; unsigned int chmap_idx; - struct mutex lock; + unsigned long busy; struct snd_soc_jack *jack; unsigned int jack_status; }; @@ -390,8 +390,8 @@ static int hdmi_codec_startup(struct snd_pcm_substream *substream, struct hdmi_codec_priv *hcp = snd_soc_dai_get_drvdata(dai); int ret = 0; - ret = mutex_trylock(&hcp->lock); - if (!ret) { + ret = test_and_set_bit(0, &hcp->busy); + if (ret) { dev_err(dai->dev, "Only one simultaneous stream supported!\n"); return -EINVAL; } @@ -419,7 +419,7 @@ static int hdmi_codec_startup(struct snd_pcm_substream *substream, err: /* Release the exclusive lock on error */ - mutex_unlock(&hcp->lock); + clear_bit(0, &hcp->busy); return ret; } @@ -431,7 +431,7 @@ static void hdmi_codec_shutdown(struct snd_pcm_substream *substream, hcp->chmap_idx = HDMI_CODEC_CHMAP_IDX_UNKNOWN; hcp->hcd.ops->audio_shutdown(dai->dev->parent, hcp->hcd.data); - mutex_unlock(&hcp->lock); + clear_bit(0, &hcp->busy); } static int hdmi_codec_hw_params(struct snd_pcm_substream *substream, @@ -811,8 +811,6 @@ static int hdmi_codec_probe(struct platform_device *pdev) return -ENOMEM; hcp->hcd = *hcd; - mutex_init(&hcp->lock); - daidrv = devm_kcalloc(dev, dai_count, sizeof(*daidrv), GFP_KERNEL); if (!daidrv) return -ENOMEM;
This reverts commit eb1ecadb7f67dde94ef0efd3ddaed5cb6c9a65ed. This fixes the following warning reported by lockdep and a potential issue with hibernation ==================================== WARNING: pulseaudio/1297 still has locks held! 5.3.0+ #1826 Not tainted ------------------------------------ 1 lock held by pulseaudio/1297: #0: ee815308 (&hcp->lock){....}, at: hdmi_codec_startup+0x20/0x130 stack backtrace: CPU: 0 PID: 1297 Comm: pulseaudio Not tainted 5.3.0+ #1826 Hardware name: Marvell Dove (Cubox) [<c0017b4c>] (unwind_backtrace) from [<c0014d10>] (show_stack+0x10/0x14) [<c0014d10>] (show_stack) from [<c00a2d18>] (futex_wait_queue_me+0x13c/0x19c) [<c00a2d18>] (futex_wait_queue_me) from [<c00a3630>] (futex_wait+0x184/0x24c) [<c00a3630>] (futex_wait) from [<c00a5e1c>] (do_futex+0x334/0x598) [<c00a5e1c>] (do_futex) from [<c00a62e8>] (sys_futex_time32+0x118/0x180) [<c00a62e8>] (sys_futex_time32) from [<c0009000>] (ret_fast_syscall+0x0/0x54) Exception stack(0xebd31fa8 to 0xebd31ff0) 1fa0: 00000000 ffffffff 000c8748 00000189 00000001 00000000 1fc0: 00000000 ffffffff 00000000 000000f0 00000000 00000000 00000000 00056200 1fe0: 000000f0 beac03a8 b6d6c835 b6d6f456 Fixes: eb1ecadb7f67 ("ASoC: hdmi-codec: re-introduce mutex locking") Reported-by: Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com> --- sound/soc/codecs/hdmi-codec.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)