Re: [PATCH v3 5/9] scsi: ufs: Simplify error handling preparation
From: Can Guo
Date: Thu Jun 10 2021 - 23:01:47 EST
Hi Adrian,
On 2021-06-10 20:30, Adrian Hunter wrote:
On 10/06/21 7:43 am, Can Guo wrote:
Commit cb7e6f05fce67c965194ac04467e1ba7bc70b069 ("scsi: ufs: core:
Enable
power management for wlun") moves UFS operations out of
ufshcd_resume(), so
in error handling preparation, if ufshcd hba has failed to resume,
there is
no point to re-enable IRQ/clk/pwr.
I am not sure how cb7e6f05fce67c965194ac04467e1ba7bc70b069 made things
any
different,
Previously, without commit cb7e6f05fce67c965194ac04467e1ba7bc70b069,
ufshcd_resume()
may turn off pwr and clk due to UFS error, e.g., link transition failure
and SSU
error/abort (and these UFS error would invoke error handling). When
error handling
kicks start, it should re-enable the pwr and clk before proceeding. Now,
commit
cb7e6f05fce67c965194ac04467e1ba7bc70b069 makes ufshcd_resume() purely
control pwr and
clk, meaning if ufshcd_resume() fails, there is nothing we can do about
it - pwr or
clk enabling must have failed, and it is not because of UFS error. This
is why I am
removing the re-enabling pwr/clk in error handling prepare.
but what I really wonder is why we don't just do recovery
directly in __ufshcd_wl_suspend() and __ufshcd_wl_resume() and strip
all
the PM complexity out of ufshcd_err_handling()?
This is a good question and I've been strugled with this idea ever since
I
started to fix error handling.
Just so you know, there are runtime and system suspend/resume. And error
handling has the same nature of user access - it is unpredictable,
meaning it
can be invoked at any time (from IRQ handler), even when there is no
ongoing
cmd/data transactions (like auto hibern8 failure and UIC errors, such as
DME
error and some errors in data link layer) [1], unless you disable UFS
IRQ.
For runtime suspend/resume, it is fine, since we call
pm_runtime_get/put_sync() in
error handling - error handling won't run into parallel with runtime
suspend/resume.
For system suspend/resume, since error handling has the same nature like
user
access, so we are using host_sem to avoid concurrency of error handling
and
system suspend/resume.
Back to your question - can we just do recovery directly in
__ufshcd_wl_suspend()
and __ufshcd_wl_resume()? Yes, we can.
However, the reasons why I choose not to do it that way are (althrough
error
handler prepare has became much more simple after apply this change)
1. I want to keep all the complexity within error handler, and re-direct
all error
recovery needs to error handler. It can avoid calling
ufshcd_reset_and_restore()
and/or flush_work(&hba->eh_work) here and there. The entire UFS
suspend/resume is
already complex enough, I don't want to mess up with it.
2. We do explicit recovery only when we see certain errors, e.g., H8
enter func
returns an error during suspend, but as mentioned above [1], error
handling can
be invoked already from IRQ handler (due to all kinds of UIC errors
before H8 enter
func returns). So, we still need host_sem (in case of system
suspend/resume) to
avoid concurrency.
3. During system suspend/resume, error handling can be invoked (due to
non-fatal
errors) but still UFS cmds return no error at all. Similar like above,
we need
host_sem to avoid concurrency.
There are more reasons why I chose this way, but it is really this way
or others.
I am glad to see someone cares about error handling and can make it
better and
more robust, no matter what that way is. :)
Thanks,
Can Guo.
Signed-off-by: Can Guo <cang@xxxxxxxxxxxxxx>
---
drivers/scsi/ufs/ufshcd.c | 58
+++++++++++++++++++++++++----------------------
1 file changed, 31 insertions(+), 27 deletions(-)
diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 7dc0fda..0afad6b 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -2727,8 +2727,8 @@ static int ufshcd_queuecommand(struct Scsi_Host
*host, struct scsi_cmnd *cmd)
break;
case UFSHCD_STATE_EH_SCHEDULED_FATAL:
/*
- * pm_runtime_get_sync() is used at error handling preparation
- * stage. If a scsi cmd, e.g. the SSU cmd, is sent from hba's
+ * ufshcd_rpm_get_sync() is used at error handling preparation
+ * stage. If a scsi cmd, e.g., the SSU cmd, is sent from the
* PM ops, it can never be finished if we let SCSI layer keep
* retrying it, which gets err handler stuck forever. Neither
* can we let the scsi cmd pass through, because UFS is in bad
@@ -5915,29 +5915,26 @@ static void ufshcd_clk_scaling_suspend(struct
ufs_hba *hba, bool suspend)
}
}
-static void ufshcd_err_handling_prepare(struct ufs_hba *hba)
+static int ufshcd_err_handling_prepare(struct ufs_hba *hba)
{
+ /*
+ * Exclusively call pm_runtime_get_sync(hba->dev) once, in case
+ * following ufshcd_rpm_get_sync() fails.
+ */
+ pm_runtime_get_sync(hba->dev);
+ /* End of the world. */
+ if (pm_runtime_suspended(hba->dev)) {
+ pm_runtime_put(hba->dev);
+ return -EINVAL;
+ }
+
+ ufshcd_set_eh_in_progress(hba);
ufshcd_rpm_get_sync(hba);
- if (pm_runtime_status_suspended(&hba->sdev_ufs_device->sdev_gendev)
||
+ if (pm_runtime_suspended(&hba->sdev_ufs_device->sdev_gendev) ||
hba->is_wl_sys_suspended) {
- enum ufs_pm_op pm_op;
+ enum ufs_pm_op pm_op = hba->is_wl_sys_suspended ?
+ UFS_SYSTEM_PM : UFS_RUNTIME_PM;
- /*
- * Don't assume anything of resume, if
- * resume fails, irq and clocks can be OFF, and powers
- * can be OFF or in LPM.
- */
- ufshcd_setup_hba_vreg(hba, true);
- ufshcd_setup_vreg(hba, true);
- ufshcd_config_vreg_hpm(hba, hba->vreg_info.vccq);
- ufshcd_config_vreg_hpm(hba, hba->vreg_info.vccq2);
- ufshcd_hold(hba, false);
- if (!ufshcd_is_clkgating_allowed(hba)) {
- ufshcd_setup_clocks(hba, true);
- ufshcd_enable_irq(hba);
- }
- ufshcd_release(hba);
- pm_op = hba->is_wl_sys_suspended ? UFS_SYSTEM_PM : UFS_RUNTIME_PM;
ufshcd_vops_resume(hba, pm_op);
} else {
ufshcd_hold(hba, false);
@@ -5951,22 +5948,25 @@ static void ufshcd_err_handling_prepare(struct
ufs_hba *hba)
down_write(&hba->clk_scaling_lock);
up_write(&hba->clk_scaling_lock);
cancel_work_sync(&hba->eeh_work);
+ return 0;
}
static void ufshcd_err_handling_unprepare(struct ufs_hba *hba)
{
+ ufshcd_clear_eh_in_progress(hba);
ufshcd_scsi_unblock_requests(hba);
ufshcd_release(hba);
if (ufshcd_is_clkscaling_supported(hba))
ufshcd_clk_scaling_suspend(hba, false);
ufshcd_clear_ua_wluns(hba);
ufshcd_rpm_put(hba);
+ pm_runtime_put(hba->dev);
}
static inline bool ufshcd_err_handling_should_stop(struct ufs_hba
*hba)
{
return (!hba->is_powered || hba->shutting_down ||
- !hba->sdev_ufs_device ||
+ !hba->sdev_ufs_device || hba->is_sys_suspended ||
hba->ufshcd_state == UFSHCD_STATE_ERROR ||
(!(hba->saved_err || hba->saved_uic_err || hba->force_reset ||
ufshcd_is_link_broken(hba))));
@@ -6052,9 +6052,13 @@ static void ufshcd_err_handler(struct
work_struct *work)
up(&hba->host_sem);
return;
}
- ufshcd_set_eh_in_progress(hba);
spin_unlock_irqrestore(hba->host->host_lock, flags);
- ufshcd_err_handling_prepare(hba);
+ if (ufshcd_err_handling_prepare(hba)) {
+ dev_err(hba->dev, "%s: error handling preparation failed\n",
+ __func__);
+ up(&hba->host_sem);
+ return;
+ }
/* Complete requests that have door-bell cleared by h/w */
ufshcd_complete_requests(hba);
spin_lock_irqsave(hba->host->host_lock, flags);
@@ -6198,7 +6202,6 @@ static void ufshcd_err_handler(struct
work_struct *work)
dev_err_ratelimited(hba->dev, "%s: exit: saved_err 0x%x
saved_uic_err 0x%x",
__func__, hba->saved_err, hba->saved_uic_err);
}
- ufshcd_clear_eh_in_progress(hba);
spin_unlock_irqrestore(hba->host->host_lock, flags);
ufshcd_err_handling_unprepare(hba);
up(&hba->host_sem);
@@ -8999,6 +9002,9 @@ static int __ufshcd_wl_resume(struct ufs_hba
*hba, enum ufs_pm_op pm_op)
/* Enable Auto-Hibernate if configured */
ufshcd_auto_hibern8_enable(hba);
+
+ hba->clk_gating.is_suspended = false;
+ ufshcd_release(hba);
goto out;
set_old_link_state:
@@ -9008,8 +9014,6 @@ static int __ufshcd_wl_resume(struct ufs_hba
*hba, enum ufs_pm_op pm_op)
out:
if (ret)
ufshcd_update_evt_hist(hba, UFS_EVT_WL_RES_ERR, (u32)ret);
- hba->clk_gating.is_suspended = false;
- ufshcd_release(hba);
hba->wl_pm_op_in_progress = false;
return ret <= 0 ? ret : -EINVAL;
}