Linux Sound subsystem development
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Vijendar Mukunda" <Vijendar.Mukunda@amd.com>
Cc: linux-sound@vger.kernel.org
Subject: Re: [PATCH v7 20/23] soundwire: amd: fix interrupt gate and work drain ordering in PM ops
Date: Wed, 18 Nov 2026 11:21:56 +0000	[thread overview]
Message-ID: <sashiko-outbox-220819@kernel.org> (raw)
In-Reply-To: <20261118104353.331556-21-Vijendar.Mukunda@amd.com>

Thank you for your contribution! Sashiko AI review found 3 potential
issues to consider:
- [High] soundwire: amd: runtime PM deadlock in amd_suspend_runtime()
  work drain
- [Medium] soundwire: amd: stale status bits and early IRQ enable in
  amd_suspend()
- [Medium] soundwire: amd: draining amd_sdw_work re-unmasks IRQs in
  POWER_OFF_MODE
--

commit 51e6c84ae75bcfe935e29acd598ea807a66790ac
Author: Vijendar Mukunda <Vijendar.Mukunda@amd.com>

soundwire: amd: fix interrupt gate and work drain ordering in PM ops

This commit reorders interrupt masking, workqueue draining, and wake
configuration across the CLK_STOP_MODE and POWER_OFF_MODE paths in
amd_suspend() and amd_suspend_runtime().

> diff --git a/drivers/soundwire/amd_manager.c b/drivers/soundwire/amd_manager.c
> index 57238c81997a2..220abf73cb25e 100644
> --- a/drivers/soundwire/amd_manager.c
> +++ b/drivers/soundwire/amd_manager.c
> @@ -1430,17 +1430,20 @@ static int __maybe_unused amd_suspend(struct device *dev)
>  	}
>  
>  	if (amd_manager->power_mode_mask & AMD_SDW_CLK_STOP_MODE) {
> +		amd_disable_sdw_interrupts(amd_manager);
> +		cancel_work_sync(&amd_manager->amd_sdw_irq_thread);
>  		cancel_work_sync(&amd_manager->amd_sdw_work);
> -		amd_sdw_wake_enable(amd_manager, false);
>  		ret = amd_sdw_host_wake_enable_by_rev(amd_manager, false);
>  		if (ret)
>  			return ret;
>  		ret = amd_sdw_clock_stop(amd_manager);
>  		if (ret)
>  			return ret;
> +		amd_enable_sdw_interrupts(amd_manager);

[Severity: Medium]
Can disabling and draining interrupts without clearing
ACP_SW_STATE_CHANGE_STATUS_0TO7 and ACP_SW_STATE_CHANGE_STATUS_8TO11, and
then calling amd_enable_sdw_interrupts() while the bus is clock-stopped,
cause stale pre-suspend status bits to be processed while the link is down?

When a SoundWire interrupt fires just before amd_disable_sdw_interrupts(),
the ACP hard IRQ handler clears ACP_EXTERNAL_INTR_STAT and queues
amd_sdw_irq_thread, leaving the status bits latched in
ACP_SW_STATE_CHANGE_STATUS_0TO7 and ACP_SW_STATE_CHANGE_STATUS_8TO11.
amd_disable_sdw_interrupts() sets interrupt_enabled to false without
clearing those status registers, so an in-flight or cancelled
amd_sdw_irq_thread() exits early without clearing them either:

drivers/soundwire/amd_manager.c:amd_sdw_irq_thread() {
	if (!READ_ONCE(amd_manager->interrupt_enabled))
		return;

	status_change_8to11 = readl(amd_manager->mmio + ACP_SW_STATE_CHANGE_STATUS_8TO11);
	status_change_0to7 = readl(amd_manager->mmio + ACP_SW_STATE_CHANGE_STATUS_0TO7);
	...
}

In both amd_suspend() and amd_suspend_runtime() under
AMD_SDW_CLK_STOP_MODE, amd_enable_sdw_interrupts() is called immediately
after amd_sdw_clock_stop() (and in AMD_SDW_POWER_OFF_MODE on ACP63,
amd_suspend_runtime() sets interrupt_enabled back to true).  When a wake
interrupt later runs amd_sdw_irq_thread() during runtime suspend,
amd_sdw_process_wake_event() only queues an asynchronous
pm_request_resume(amd_manager->dev) before amd_sdw_irq_thread() falls
through, reads the stale status bits, and runs
amd_sdw_read_and_process_ping_status() and schedule_work(&amd_sdw_work)
while the clock is still stopped, resulting in MIPI command timeouts.

In addition, amd_sdw_clock_stop() returns 0 on all failure paths while
leaving the SoundWire clock running, so if clock stop fails,
amd_enable_sdw_interrupts() re-enables state-change interrupts on an active
bus as the device enters D3.

>  	} else if (amd_manager->power_mode_mask & AMD_SDW_POWER_OFF_MODE) {
> +		amd_disable_sdw_interrupts(amd_manager);
> +		cancel_work_sync(&amd_manager->amd_sdw_irq_thread);
>  		cancel_work_sync(&amd_manager->amd_sdw_work);
> -		amd_sdw_wake_enable(amd_manager, false);
>  		ret = amd_sdw_host_wake_enable_by_rev(amd_manager, false);
>  		if (ret)
>  			return ret;
> @@ -1451,7 +1454,6 @@ static int __maybe_unused amd_suspend(struct device *dev)
>  		ret = amd_sdw_clock_stop(amd_manager);
>  		if (ret)
>  			return ret;
> -		amd_disable_sdw_interrupts(amd_manager);
>  		ret = amd_disable_sdw_manager(amd_manager);
>  		if (ret)
>  			return ret;

[Severity: Medium]
In AMD_SDW_POWER_OFF_MODE (both here and in amd_suspend_runtime()), can an
in-flight amd_sdw_update_slave_status_work() re-unmask
ACP_SW_STATE_CHANGE_STATUS_MASK_0TO7 and
ACP_SW_STATE_CHANGE_STATUS_MASK_8TO11 after amd_disable_sdw_interrupts()
zeroes them?

CPU0                                  CPU1
amd_suspend_runtime()                 amd_sdw_update_slave_status_work()
  amd_disable_sdw_interrupts()
    writel(0, ..._MASK_0TO7)
    writel(0, ..._MASK_8TO11)
  cancel_work_sync(&amd_sdw_work)
                                        sdw_handle_slave_status()
                                        if (status[0] == SDW_SLAVE_ATTACHED)
                                          writel(AMD_SDW_IRQ_MASK_0TO7, ...)
                                          writel(AMD_SDW_IRQ_MASK_8TO11, ...)
                                          <-- masks left enabled

Because the post-clock-stop amd_disable_sdw_interrupts() call was removed,
those mask registers stay set to AMD_SDW_IRQ_MASK_0TO7 and
AMD_SDW_IRQ_MASK_8TO11 after cancel_work_sync(&amd_manager->amd_sdw_work)
returns.  In amd_suspend_runtime(), amd_sdw_wake_enable(amd_manager, true)
then performs a read-modify-write on ACP_SW_STATE_CHANGE_STATUS_MASK_8TO11,
and on ACP63 with wake_en_mask set, interrupt_enabled and
ACP_EXTERNAL_INTR_CNTL are re-enabled with all state-change and PREQ masks
active while the manager is disabled.

> @@ -1491,18 +1493,24 @@ static int __maybe_unused amd_suspend_runtime(struct device *dev)
>  		return 0;
>  	}
>  	if (amd_manager->power_mode_mask & AMD_SDW_CLK_STOP_MODE) {
> -		amd_sdw_wake_enable(amd_manager, true);
> +		amd_disable_sdw_interrupts(amd_manager);
> +		cancel_work_sync(&amd_manager->amd_sdw_irq_thread);
> +		cancel_work_sync(&amd_manager->amd_sdw_work);

[Severity: High]
Can calling cancel_work_sync(&amd_manager->amd_sdw_work) inside
amd_suspend_runtime() deadlock with amd_sdw_update_slave_status_work()?

CPU0                                  CPU1
rpm_suspend(amd_manager->dev)         amd_sdw_update_slave_status_work()
  runtime_status = RPM_SUSPENDING
  amd_suspend_runtime()
    cancel_work_sync(&amd_sdw_work)
                                        sdw_handle_slave_status()
                                          sdw_handle_slave_alerts()
                                            pm_runtime_get_sync(&slave->dev)
                                              rpm_resume(&slave->dev, ...)
                                                rpm_resume(bus->dev, 0)
                                                  rpm_resume(amd_manager->dev, 0)
                                                    waits on runtime_status

 *** DEADLOCK ***

Each peripheral slave->dev is a child of bus->dev, which is a child of
amd_manager->dev, and both children are in RPM_SUSPENDED when
amd_manager->dev enters rpm_suspend().  If a peripheral alert schedules
amd_sdw_work before interrupts are disabled and
amd_sdw_update_slave_status_work() starts executing before
cancel_work_sync(&amd_manager->amd_sdw_work) cancels it,
pm_runtime_get_sync(&slave->dev) propagates a synchronous rpm_resume() up
to amd_manager->dev.  Because amd_manager->dev is in RPM_SUSPENDING,
rpm_resume() sleeps on amd_manager->dev->power.wait_queue waiting for
amd_suspend_runtime() to finish, while amd_suspend_runtime() is blocked in
cancel_work_sync(&amd_manager->amd_sdw_work) in both AMD_SDW_CLK_STOP_MODE
and AMD_SDW_POWER_OFF_MODE.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261118104353.331556-1-Vijendar.Mukunda@amd.com?part=20

  reply	other threads:[~2026-11-18 11:21 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18 10:42 [PATCH v7 00/23] soundwire: amd: SoundWire manager driver bug fixes Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 01/23] soundwire: amd: consolidate pad register macros into amd_init.h Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 02/23] soundwire: amd: program pad SCHMEN and OE control for ACP 7.0/7.1/7.2 Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 03/23] soundwire: amd: add pad ctrl helpers for D0/D3 PM transitions Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 04/23] soundwire: amd: track full pad teardown across suspend/resume Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 05/23] soundwire: amd: fix missing wake interrupt and group wake-arming calls in runtime suspend Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 06/23] soundwire: amd: fix IRQ thread dropping STATUS_0TO7 on simultaneous wake Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 07/23] soundwire: amd: remove STATUS_8TO11 clear from amd_sdw_process_wake_event() Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 08/23] soundwire: amd: fix missing wake interrupt for ACP6.3 in POWER_OFF suspend Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 09/23] soundwire: amd: fix SDW command timeout return value handling Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 10/23] soundwire: amd: cache ping slave status to avoid spurious disconnect on timeout Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 11/23] soundwire: amd: assert msg_lock is held in amd_sdw_read_ping_status() Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 12/23] soundwire: amd: fix work drain ordering in remove path Vijendar Mukunda
2026-11-18 11:11   ` sashiko-bot
2026-11-18 10:42 ` [PATCH v7 13/23] soundwire: amd: defer dev_set_drvdata() until after INIT_WORK() Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 14/23] soundwire: amd: pm_runtime guard in remove path Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 15/23] soundwire: amd: use interrupt_enabled flag to guard IRQ work scheduling Vijendar Mukunda
2026-11-18 11:16   ` sashiko-bot
2026-11-18 10:42 ` [PATCH v7 16/23] soundwire: amd: fix *sdw_ctx leak when sdw_amd_startup() fails Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 17/23] soundwire: amd: propagate amd_init_sdw_manager() error on resume Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 18/23] soundwire: amd: drop amd_deinit_sdw_manager() in POWER_OFF suspend Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 19/23] soundwire: amd: replace >= ACP70 with explicit switch/case in PM paths Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 20/23] soundwire: amd: fix interrupt gate and work drain ordering in PM ops Vijendar Mukunda
2026-11-18 11:21   ` sashiko-bot [this message]
2026-11-18 10:42 ` [PATCH v7 21/23] soundwire: amd: check for NULL manager drvdata in sdw_amd_startup() Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 22/23] soundwire: amd: fix shared-register RMW race in amd_sdw_set_device_state() Vijendar Mukunda
2026-11-18 10:42 ` [PATCH v7 23/23] soundwire: amd: fix shared-register RMW race in amd_sdw_host_wake_enable() Vijendar Mukunda

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-220819@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=Vijendar.Mukunda@amd.com \
    --cc=linux-sound@vger.kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox