Message ID | 1621846046-22204-6-git-send-email-cang@codeaurora.org |
---|---|
State | New |
Headers | show |
Series | Complementary changes for error handling | expand |
Hi Bart, On 2021-05-25 00:56, Bart Van Assche wrote: > On 5/24/21 1:47 AM, Can Guo wrote: >> UFS error handling now is doing more than just re-probing, but also >> sending >> scsi cmds, e.g., for clearing UACs, and recovering runtime PM error, >> which >> may change runtime status of scsi devices. To protect system >> suspend/resume >> from being disturbed by error handling, move the host_sem from wl pm >> ops >> to ufshcd_suspend_prepare() and ufshcd_resume_complete(). > > Other SCSI LLDs can perform error handling while system suspend/resume > is in progress. Why can't the UFS driver do this? I don't know about other SCSI LLDs, but UFS error handling is basically doing a re-probe/re-initialization to UFS device. Having UFS error handling running in parallel with system suspend/resume, neither of them will end up well. I didn't design all this, it is just happening, I am trying to fix it and semaphore works well for me. I am really glad to see someone cares about error handling and fix it with better ideas (maybe using WQ_FREEZABLE) later. > > Additionally, please document what the purpose of host_sem is before > making any changes to how host_sem is used. The only documentation I > have found of host_sem is the following: "* @host_sem: semaphore used > to > serialize concurrent contexts". To me that text is less than useful > since semaphores are almost always used to serialize concurrent code. > Sure, host_sem is actually preventing cocurrency happens among any of contexts, such as sysfs access, shutdown, error handling, system suspend/resume and async probe, I will update its message in next version. Thanks, Can Guo. > Thanks, > > Bart.
diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c index 5e6e3ac..9a3bc04 100644 --- a/drivers/scsi/ufs/ufshcd.c +++ b/drivers/scsi/ufs/ufshcd.c @@ -9057,16 +9057,13 @@ static int ufshcd_wl_suspend(struct device *dev) ktime_t start = ktime_get(); hba = shost_priv(sdev->host); - down(&hba->host_sem); if (pm_runtime_suspended(dev)) goto out; ret = __ufshcd_wl_suspend(hba, UFS_SYSTEM_PM); - if (ret) { + if (ret) dev_err(&sdev->sdev_gendev, "%s failed: %d\n", __func__, ret); - up(&hba->host_sem); - } out: if (!ret) @@ -9099,7 +9096,6 @@ static int ufshcd_wl_resume(struct device *dev) hba->curr_dev_pwr_mode, hba->uic_link_state); if (!ret) hba->is_wl_sys_suspended = false; - up(&hba->host_sem); return ret; } #endif @@ -9666,6 +9662,7 @@ void ufshcd_resume_complete(struct device *dev) ufshcd_rpmb_rpm_put(hba); hba->rpmb_complete_put = false; } + up(&hba->host_sem); } EXPORT_SYMBOL_GPL(ufshcd_resume_complete); @@ -9692,6 +9689,7 @@ int ufshcd_suspend_prepare(struct device *dev) ufshcd_rpmb_rpm_get_sync(hba); hba->rpmb_complete_put = true; } + down(&hba->host_sem); return 0; } EXPORT_SYMBOL_GPL(ufshcd_suspend_prepare);
UFS error handling now is doing more than just re-probing, but also sending scsi cmds, e.g., for clearing UACs, and recovering runtime PM error, which may change runtime status of scsi devices. To protect system suspend/resume from being disturbed by error handling, move the host_sem from wl pm ops to ufshcd_suspend_prepare() and ufshcd_resume_complete(). Signed-off-by: Can Guo <cang@codeaurora.org> --- drivers/scsi/ufs/ufshcd.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-)