diff mbox series

[v2] ufs: core: wlun suspend dev/link state error recovery

Message ID 20240329015036.15707-1-peter.wang@mediatek.com (mailing list archive)
State Accepted
Headers show
Series [v2] ufs: core: wlun suspend dev/link state error recovery | expand

Commit Message

Peter Wang (王信友) March 29, 2024, 1:50 a.m. UTC 
From: Peter Wang <peter.wang@mediatek.com>

When wl suspend error occurs, for example, BKOP or SSU timeout, the host
triggers an error handler and returns -EBUSY to break the wl suspend process.
However, it is possible for the runtime PM to enter wl suspend again before
the error handler has finished, and return -EINVAL because the device is
in an error state. To address this, ensure that the rumtime PM waits for the
error handler to finish, or trigger the error handler in such cases,
because returning -EINVAL can cause the I/O to hang.

Signed-off-by: Peter Wang <peter.wang@mediatek.com>
---
 drivers/ufs/core/ufshcd.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Comments

Bart Van Assche March 29, 2024, 6:26 p.m. UTC | #1 
On 3/28/24 6:50 PM, peter.wang@mediatek.com wrote:
> When wl suspend error occurs, for example, BKOP or SSU timeout, the host
> triggers an error handler and returns -EBUSY to break the wl suspend process.
> However, it is possible for the runtime PM to enter wl suspend again before
> the error handler has finished, and return -EINVAL because the device is
> in an error state. To address this, ensure that the rumtime PM waits for the
> error handler to finish, or trigger the error handler in such cases,
> because returning -EINVAL can cause the I/O to hang.

Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Martin K. Petersen April 2, 2024, 1:48 a.m. UTC | #2 
On Fri, 29 Mar 2024 09:50:36 +0800, peter.wang@mediatek.com wrote:

> When wl suspend error occurs, for example, BKOP or SSU timeout, the host
> triggers an error handler and returns -EBUSY to break the wl suspend process.
> However, it is possible for the runtime PM to enter wl suspend again before
> the error handler has finished, and return -EINVAL because the device is
> in an error state. To address this, ensure that the rumtime PM waits for the
> error handler to finish, or trigger the error handler in such cases,
> because returning -EINVAL can cause the I/O to hang.
> 
> [...]

Applied to 6.9/scsi-fixes, thanks!

[1/1] ufs: core: wlun suspend dev/link state error recovery
      https://git.kernel.org/mkp/scsi/c/6bc5e70b1c79
diff mbox series

Patch

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index e30fd125988d..292b06f361a2 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -9791,7 +9791,10 @@  static int __ufshcd_wl_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 
 	/* UFS device & link must be active before we enter in this function */
 	if (!ufshcd_is_ufs_dev_active(hba) || !ufshcd_is_link_active(hba)) {
-		ret = -EINVAL;
+		/*  Wait err handler finish or trigger err recovery */
+		if (!ufshcd_eh_in_progress(hba))
+			ufshcd_force_error_recovery(hba);
+		ret = -EBUSY;
 		goto enable_scaling;
 	}