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.
Signed-off-by: Can Guo <[email protected]>
---
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;
}
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
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, 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()?
>
> Signed-off-by: Can Guo <[email protected]>
> ---
> 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;
> }
>
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 <[email protected]>
>> ---
>> 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;
>> }
>>
On 6/10/21 8:01 PM, Can Guo wrote:
> 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.
Why are link transition failures handled in the error handler instead of
in the context where these errors are detected (ufshcd_resume())? Is it
even possible to recover from a link transition failure or does this
perhaps indicate a broken UFS controller?
>> 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()?
+1
> 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.
Why is host_sem used for that purpose instead of lock_system_sleep() and
unlock_system_sleep()?
Thanks,
Bart.
On 2021-06-12 04:58, Bart Van Assche wrote:
> On 6/10/21 8:01 PM, Can Guo wrote:
>> 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.
>
> Why are link transition failures handled in the error handler instead
> of
> in the context where these errors are detected (ufshcd_resume())? Is it
> even possible to recover from a link transition failure or does this
> perhaps indicate a broken UFS controller?
Basically, almost all UFS failures are caused by errors in underlaying
layers,
i.e., UIC errors, including link transition failures. And according to
UFSHCI
spec, SW should do a full reset to recover it, just like handle any
other
fatal UIC errors. All UIC errors are detected by HW and reported by IRQ
handler.
UFSHCI Spec Ver. 31
8.2.7 Hibernate Enter/Exit Error Handling
Hibernate Enter/Exit Error occurs when the UniPro link is broken. When
this condition occurs,
host software should reset the host controller by setting register HCE
to ‘0’, re-initialize the host
controller by setting register HCE to ‘1', and then start link startup
sequence as shown in Figure 16.
>
>>> 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()?
>
> +1
I've explained why I chose not to do this in my last reply to Adrian.
Please kindly check it.
>
>> 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.
>
> Why is host_sem used for that purpose instead of lock_system_sleep()
> and
> unlock_system_sleep()?
>
I was aware of it, but the situation is that host_sem is also used to
avoid concurrency among user access, error handling and shutdown, so
I think just use host_sem anyways to simply the lockings, otherwise
user access and error handling would have to take both
system_transition_mutex
and host_sem
Thanks,
Can Guo.
> Thanks,
>
> Bart.
Hi Bart,
On 2021-06-12 14:46, Can Guo wrote:
> On 2021-06-12 04:58, Bart Van Assche wrote:
>> On 6/10/21 8:01 PM, Can Guo wrote:
>>> 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.
>>
>> Why are link transition failures handled in the error handler instead
>> of
>> in the context where these errors are detected (ufshcd_resume())? Is
>> it
>> even possible to recover from a link transition failure or does this
>> perhaps indicate a broken UFS controller?
>
> Basically, almost all UFS failures are caused by errors in underlaying
> layers,
> i.e., UIC errors, including link transition failures. And according to
> UFSHCI
> spec, SW should do a full reset to recover it, just like handle any
> other
> fatal UIC errors. All UIC errors are detected by HW and reported by IRQ
> handler.
>
> UFSHCI Spec Ver. 31
> 8.2.7 Hibernate Enter/Exit Error Handling
> Hibernate Enter/Exit Error occurs when the UniPro link is broken. When
> this condition occurs,
> host software should reset the host controller by setting register HCE
> to ‘0’, re-initialize the host
> controller by setting register HCE to ‘1', and then start link startup
> sequence as shown in Figure 16.
>
>>
>>>> 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()?
>>
>> +1
>
> I've explained why I chose not to do this in my last reply to Adrian.
> Please kindly check it.
>
>>
>>> 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.
>>
>> Why is host_sem used for that purpose instead of lock_system_sleep()
>> and
>> unlock_system_sleep()?
>>
>
> I was aware of it, but the situation is that host_sem is also used to
> avoid concurrency among user access, error handling and shutdown, so
> I think just use host_sem anyways to simply the lockings, otherwise
> user access and error handling would have to take both
> system_transition_mutex
> and host_sem
On second thought, I will take your suggestion to use
lock_system_sleep()
and unlock_system_sleep() in error handler and remove the host_sem used
in suspend/resume, which can make the code more readable by keeping the
changes within error handler itself. However, please note that host_sem
will still be used to avoid concurrency of user access, error handler
and
shutdown.
Thanks,
Can Guo.
>
> Thanks,
>
> Can Guo.
>
>> Thanks,
>>
>> Bart.