On 11/14/19 10:01 PM, Can Guo wrote:
On 2019-11-15 00:11, Bart Van Assche wrote:
On 11/13/19 8:03 AM, Bean Huo (beanhuo) wrote:
I think, you are asking for comment from Can. As for me, attached
patch is better.
Removing ufshcd_wait_for_doorbell_clr(), instead of reading doorbell
register, Now
using block layer blk_mq_{un,}freeze_queue(), looks good. I tested
your V5 patches,
didn't find problem yet.
Since my available platform doesn't support dynamic clk scaling, I
think, now seems only
Qcom UFS controllers support this feature. So we need Can Guo to
finally confirm it.
Do you agree with this patch series if patch 4 is replaced by the
patch attached to my previous e-mail? The entire (revised) series is
available at https://github.com/bvanassche/linux/tree/ufs-for-next.
After ufshcd_clock_scaling_prepare() returns(no error), all request
queues are frozen. If failure
happens(say power mode change command fails) after this point and error
handler kicks off,
we need to send dev commands(in ufshcd_probe_hba()) to bring UFS back to
functionality.
However, as the hba->cmd_queue is frozen, dev commands cannot be sent,
the error handler shall fail.
Hi Can,
My understanding of the current UFS driver code is that
ufshcd_clock_scaling_prepare() waits for ongoing commands to finish but
not for SCSI error handling to finish. Would you agree with changing
that behavior such that ufshcd_clock_scaling_prepare() returns an error
code if SCSI error handling is in progress? Do you agree that once that
change has been made that it is fine to invoke blk_freeze_queue_start()
for all three types of block layer request queues (SCSI commands, device
management commands and TMFs)? Do you agree that this would fix the
issue that it is possible today to submit TMFs from user space using
through the BSG queue while clock scaling is in progress?
Thanks,
Bart.
