[PATCH AUTOSEL 6.16-6.6] soundwire: amd: serialize amd manager resume sequence during pm_prepare

[Sasha Levin <sashal@kernel.org>]

From: Vijendar Mukunda <Vijendar.Mukunda@amd.com>

[ Upstream commit 03837341790039d6f1cbf7a1ae7dfa2cb77ef0a4 ]

During pm_prepare callback, pm_request_resume() delays SoundWire manager D0
entry sequence. Synchronize runtime resume sequence for amd_manager
instance prior to invoking child devices resume sequence for both the amd
power modes(ClockStop Mode and Power off mode).
Change the power_mode_mask check and use pm_runtime_resume() in
amd_pm_prepare() callback.

Signed-off-by: Vijendar Mukunda <Vijendar.Mukunda@amd.com>
Link: https://lore.kernel.org/r/20250530054447.1645807-3-Vijendar.Mukunda@amd.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---

LLM Generated explanations, may be completely bogus:

**Backport Status: YES**

This commit should be backported to stable kernel trees for the
following reasons:

## Bug Fix Nature
The commit fixes a critical synchronization issue in the AMD SoundWire
driver's power management sequence. The original code used
`pm_request_resume()` which is asynchronous, leading to race conditions
where the manager device might not be fully resumed before child devices
start their resume sequence. This causes:
1. Missing device alerts on AMD platforms
2. Race conditions during system resume
3. Potential communication failures between the SoundWire manager and
   peripheral devices

## Code Analysis

1. **Critical Fix**: The change from `pm_request_resume()` to
   `pm_runtime_resume()` ensures synchronous resume operation,
   eliminating race conditions. This is similar to Intel's fix (commit
   6f4867fa5760) which addressed the same class of issue.

2. **Expanded Scope**: The condition change from `(power_mode_mask &
   AMD_SDW_CLK_STOP_MODE)` to just `(power_mode_mask)` ensures the fix
   applies to both power modes:
   - AMD_SDW_CLK_STOP_MODE (value 1): Clock stop mode
   - AMD_SDW_POWER_OFF_MODE (value 2): Power off mode

   This ensures proper synchronization for all power management
scenarios, not just clock stop mode.

3. **Small and Contained**: The fix is minimal (3 lines changed) and
   localized to the `amd_pm_prepare()` function, reducing regression
   risk.

4. **Clear Bug Symptoms**: The commit message explicitly states the
   observed bug: "device alerts are missing without pm_prepare on AMD
   platforms" - a user-visible problem affecting SoundWire
   functionality.

5. **No Architectural Changes**: This is purely a bug fix that corrects
   the synchronization mechanism without introducing new features or
   changing the driver architecture.

6. **Subsystem Pattern**: This follows an established pattern in the
   SoundWire subsystem where Intel had a similar issue and fix,
   indicating this is a known class of bugs that should be addressed in
   stable kernels.

The commit meets all stable kernel criteria: it fixes a real bug
affecting users, is minimal in scope, has low regression risk, and
doesn't introduce new functionality.

 drivers/soundwire/amd_manager.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/soundwire/amd_manager.c b/drivers/soundwire/amd_manager.c
index 7a671a786197..3b335d6eaa94 100644
--- a/drivers/soundwire/amd_manager.c
+++ b/drivers/soundwire/amd_manager.c
@@ -1178,10 +1178,10 @@ static int __maybe_unused amd_pm_prepare(struct device *dev)
 	 * device is not in runtime suspend state, observed that device alerts are missing
 	 * without pm_prepare on AMD platforms in clockstop mode0.
 	 */
-	if (amd_manager->power_mode_mask & AMD_SDW_CLK_STOP_MODE) {
-		ret = pm_request_resume(dev);
+	if (amd_manager->power_mode_mask) {
+		ret = pm_runtime_resume(dev);
 		if (ret < 0) {
-			dev_err(bus->dev, "pm_request_resume failed: %d\n", ret);
+			dev_err(bus->dev, "pm_runtime_resume failed: %d\n", ret);
 			return 0;
 		}
 	}
-- 
2.39.5