[PATCH] ASoC: samsung: Prevent clk_get_rate() calls in atomic context

[PATCH] ASoC: samsung: Prevent clk_get_rate() calls in atomic context

[Sylwester Nawrocki]

This patch moves clk_get_rate() call from trigger() to hw_params()
callback to avoid calling sleeping clk API from atomic context
and prevent deadlock as indicated below.

Before this change clk_get_rate() was being called with same
spinlock held as the one passed to the clk API when registering
clocks exposed by the I2S driver.

[   82.109780] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:908
[   82.117009] in_atomic(): 1, irqs_disabled(): 128, pid: 1554, name: speaker-test
[   82.124235] 3 locks held by speaker-test/1554:
[   82.128653]  #0: cc8c5328 (snd_pcm_link_rwlock){...-}, at: snd_pcm_stream_lock_irq+0x20/0x38
[   82.137058]  #1: ec9eda17 (&(&substream->self_group.lock)->rlock){..-.}, at: snd_pcm_ioctl+0x900/0x1268
[   82.146417]  #2: 6ac279bf (&(&pri_dai->spinlock)->rlock){..-.}, at: i2s_trigger+0x64/0x6d4
[   82.154650] irq event stamp: 8144
[   82.157949] hardirqs last  enabled at (8143): [<c0a0f574>] _raw_read_unlock_irq+0x24/0x5c
[   82.166089] hardirqs last disabled at (8144): [<c0a0f6a8>] _raw_read_lock_irq+0x18/0x58
[   82.174063] softirqs last  enabled at (8004): [<c01024e4>] __do_softirq+0x3a4/0x66c
[   82.181688] softirqs last disabled at (7997): [<c012d730>] irq_exit+0x140/0x168
[   82.188964] Preemption disabled at:
[   82.188967] [<00000000>]   (null)
[   82.195728] CPU: 6 PID: 1554 Comm: speaker-test Not tainted 5.0.0-rc5-00192-ga6e6caca8f03 #191
[   82.204302] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[   82.210376] [<c0111a54>] (unwind_backtrace) from [<c010d8f4>] (show_stack+0x10/0x14)
[   82.218084] [<c010d8f4>] (show_stack) from [<c09ef004>] (dump_stack+0x90/0xc8)
[   82.225278] [<c09ef004>] (dump_stack) from [<c0152980>] (___might_sleep+0x22c/0x2c8)
[   82.232990] [<c0152980>] (___might_sleep) from [<c0a0a2e4>] (__mutex_lock+0x28/0xa3c)
[   82.240788] [<c0a0a2e4>] (__mutex_lock) from [<c0a0ad80>] (mutex_lock_nested+0x1c/0x24)
[   82.248763] [<c0a0ad80>] (mutex_lock_nested) from [<c04923dc>] (clk_prepare_lock+0x78/0xec)
[   82.257079] [<c04923dc>] (clk_prepare_lock) from [<c049538c>] (clk_core_get_rate+0xc/0x5c)
[   82.265309] [<c049538c>] (clk_core_get_rate) from [<c0766b18>] (i2s_trigger+0x490/0x6d4)
[   82.273369] [<c0766b18>] (i2s_trigger) from [<c074fec4>] (soc_pcm_trigger+0x100/0x140)
[   82.281254] [<c074fec4>] (soc_pcm_trigger) from [<c07378a0>] (snd_pcm_do_start+0x2c/0x30)
[   82.289400] [<c07378a0>] (snd_pcm_do_start) from [<c07376cc>] (snd_pcm_action_single+0x38/0x78)
[   82.298065] [<c07376cc>] (snd_pcm_action_single) from [<c073a450>] (snd_pcm_ioctl+0x910/0x1268)
[   82.306734] [<c073a450>] (snd_pcm_ioctl) from [<c0292344>] (do_vfs_ioctl+0x90/0x9ec)
[   82.314443] [<c0292344>] (do_vfs_ioctl) from [<c0292cd4>] (ksys_ioctl+0x34/0x60)
[   82.321808] [<c0292cd4>] (ksys_ioctl) from [<c0101000>] (ret_fast_syscall+0x0/0x28)
[   82.329431] Exception stack(0xeb875fa8 to 0xeb875ff0)
[   82.334459] 5fa0:                   00033c18 b6e31000 00000004 00004142 00033d80 00033d80
[   82.342605] 5fc0: 00033c18 b6e31000 00008000 00000036 00008000 00000000 beea38a8 00008000
[   82.350748] 5fe0: b6e3142c beea384c b6da9a30 b6c9212c
[   82.355789]
[   82.357245] ======================================================
[   82.363397] WARNING: possible circular locking dependency detected
[   82.369551] 5.0.0-rc5-00192-ga6e6caca8f03 #191 Tainted: G        W
[   82.376395] ------------------------------------------------------
[   82.382548] speaker-test/1554 is trying to acquire lock:
[   82.387834] 6d2007f4 (prepare_lock){+.+.}, at: clk_prepare_lock+0x78/0xec
[   82.394593]
[   82.394593] but task is already holding lock:
[   82.400398] 6ac279bf (&(&pri_dai->spinlock)->rlock){..-.}, at: i2s_trigger+0x64/0x6d4
[   82.408197]
[   82.408197] which lock already depends on the new lock.
[   82.416343]
[   82.416343] the existing dependency chain (in reverse order) is:
[   82.423795]
[   82.423795] -> #1 (&(&pri_dai->spinlock)->rlock){..-.}:
[   82.430472]        clk_mux_set_parent+0x34/0xb8
[   82.434975]        clk_core_set_parent_nolock+0x1c4/0x52c
[   82.440347]        clk_set_parent+0x38/0x6c
[   82.444509]        of_clk_set_defaults+0xc8/0x308
[   82.449186]        of_clk_add_provider+0x84/0xd0
[   82.453779]        samsung_i2s_probe+0x408/0x5f8
[   82.458376]        platform_drv_probe+0x48/0x98
[   82.462879]        really_probe+0x224/0x3f4
[   82.467037]        driver_probe_device+0x70/0x1c4
[   82.471716]        bus_for_each_drv+0x44/0x8c
[   82.476049]        __device_attach+0xa0/0x138
[   82.480382]        bus_probe_device+0x88/0x90
[   82.484715]        deferred_probe_work_func+0x6c/0xbc
[   82.489741]        process_one_work+0x200/0x740
[   82.494246]        worker_thread+0x2c/0x4c8
[   82.498408]        kthread+0x128/0x164
[   82.502131]        ret_from_fork+0x14/0x20
[   82.506204]          (null)
[   82.508976]
[   82.508976] -> #0 (prepare_lock){+.+.}:
[   82.514264]        __mutex_lock+0x60/0xa3c
[   82.518336]        mutex_lock_nested+0x1c/0x24
[   82.522756]        clk_prepare_lock+0x78/0xec
[   82.527088]        clk_core_get_rate+0xc/0x5c
[   82.531421]        i2s_trigger+0x490/0x6d4
[   82.535494]        soc_pcm_trigger+0x100/0x140
[   82.539913]        snd_pcm_do_start+0x2c/0x30
[   82.544246]        snd_pcm_action_single+0x38/0x78
[   82.549012]        snd_pcm_ioctl+0x910/0x1268
[   82.553345]        do_vfs_ioctl+0x90/0x9ec
[   82.557417]        ksys_ioctl+0x34/0x60
[   82.561229]        ret_fast_syscall+0x0/0x28
[   82.565477]        0xbeea384c
[   82.568421]
[   82.568421] other info that might help us debug this:
[   82.568421]
[   82.576394]  Possible unsafe locking scenario:
[   82.576394]
[   82.582285]        CPU0                    CPU1
[   82.586792]        ----                    ----
[   82.591297]   lock(&(&pri_dai->spinlock)->rlock);
[   82.595977]                                lock(prepare_lock);
[   82.601782]                                lock(&(&pri_dai->spinlock)->rlock);
[   82.608975]   lock(prepare_lock);
[   82.612268]
[   82.612268]  *** DEADLOCK ***

Fixes: 647d04f8e07a ("ASoC: samsung: i2s: Ensure the RCLK rate is properly determined")
Reported-by: Krzysztof Kozłowski <krzk@kernel.org>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
---
 sound/soc/samsung/i2s.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/sound/soc/samsung/i2s.c b/sound/soc/samsung/i2s.c
index d6c62aa13041..ce00fe2f6aae 100644
--- a/sound/soc/samsung/i2s.c
+++ b/sound/soc/samsung/i2s.c
@@ -700,6 +700,7 @@ static int i2s_hw_params(struct snd_pcm_substream *substream,
 {
 	struct i2s_dai *i2s = to_info(dai);
 	u32 mod, mask = 0, val = 0;
+	struct clk *rclksrc;
 	unsigned long flags;
 
 	WARN_ON(!pm_runtime_active(dai->dev));
@@ -782,6 +783,10 @@ static int i2s_hw_params(struct snd_pcm_substream *substream,
 
 	i2s->frmclk = params_rate(params);
 
+	rclksrc = i2s->clk_table[CLK_I2S_RCLK_SRC];
+	if (rclksrc && !IS_ERR(rclksrc))
+		i2s->rclk_srcrate = clk_get_rate(rclksrc);
+
 	return 0;
 }
 
@@ -886,11 +891,6 @@ static int config_setup(struct i2s_dai *i2s)
 		return 0;
 
 	if (!(i2s->quirks & QUIRK_NO_MUXPSR)) {
-		struct clk *rclksrc = i2s->clk_table[CLK_I2S_RCLK_SRC];
-
-		if (rclksrc && !IS_ERR(rclksrc))
-			i2s->rclk_srcrate = clk_get_rate(rclksrc);
-
 		psr = i2s->rclk_srcrate / i2s->frmclk / rfs;
 		writel(((psr - 1) << 8) | PSR_PSREN, i2s->addr + I2SPSR);
 		dev_dbg(&i2s->pdev->dev,
-- 
2.20.1

Re: [PATCH] ASoC: samsung: Prevent clk_get_rate() calls in atomic context

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 728 bytes --]

On Thu, Feb 07, 2019 at 03:20:41PM +0100, Sylwester Nawrocki wrote:

> [   82.109780] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:908
> [   82.117009] in_atomic(): 1, irqs_disabled(): 128, pid: 1554, name: speaker-test
> [   82.124235] 3 locks held by speaker-test/1554:
> [   82.128653]  #0: cc8c5328 (snd_pcm_link_rwlock){...-}, at: snd_pcm_stream_lock_irq+0x20/0x38

Please think hard before including complete backtraces in upstream
reports, they are very large and contain almost no useful information
relative to their size so often obscure the relevant content in your
message. If part of the backtrace is usefully illustrative then it's
usually better to pull out the relevant sections.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] ASoC: samsung: Prevent clk_get_rate() calls in atomic context

[Sylwester Nawrocki]

On 2/7/19 15:32, Mark Brown wrote:
> On Thu, Feb 07, 2019 at 03:20:41PM +0100, Sylwester Nawrocki wrote:
> 
>> [   82.109780] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:908
>> [   82.117009] in_atomic(): 1, irqs_disabled(): 128, pid: 1554, name: speaker-test
>> [   82.124235] 3 locks held by speaker-test/1554:
>> [   82.128653]  #0: cc8c5328 (snd_pcm_link_rwlock){...-}, at: snd_pcm_stream_lock_irq+0x20/0x38

> Please think hard before including complete backtraces in upstream
> reports, they are very large and contain almost no useful information
> relative to their size so often obscure the relevant content in your
> message. If part of the backtrace is usefully illustrative then it's
> usually better to pull out the relevant sections.

OK, let me trim the backtrace some more and resend.

Re: [PATCH] ASoC: samsung: Prevent clk_get_rate() calls in atomic context

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 537 bytes --]

On Thu, Feb 07, 2019 at 03:47:14PM +0100, Sylwester Nawrocki wrote:
> On 2/7/19 15:32, Mark Brown wrote:

> > Please think hard before including complete backtraces in upstream
> > reports, they are very large and contain almost no useful information
> > relative to their size so often obscure the relevant content in your
> > message. If part of the backtrace is usefully illustrative then it's
> > usually better to pull out the relevant sections.

> OK, let me trim the backtrace some more and resend.

It's fine, I already applied.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]