| Message ID | 20220124180637.160524-1-hy50.seo@samsung.com |
|---|---|
| State | New |
| Headers | show |
| Series | [v2] scsi: ufs: disable auto hibern8 while entering suspend | expand |
Hi, Please check this patch. If there is any other opinion, please give me comments Thanks > -----Original Message----- > From: 서호영 [mailto:hy50.seo@samsung.com] > Sent: Wednesday, January 26, 2022 2:35 PM > To: 'linux-scsi@vger.kernel.org'; 'linux-kernel@vger.kernel.org'; > 'alim.akhtar@samsung.com'; 'avri.altman@wdc.com'; 'jejb@linux.ibm.com'; > 'martin.petersen@oracle.com'; 'beanhuo@micron.com'; > 'asutoshd@codeaurora.org'; 'cang@codeaurora.org'; 'bvanassche@acm.org'; > 'bhoon95.kim@samsung.com'; 'kwmad.kim@samsung.com' > Cc: 'kernel test robot' > Subject: RE: [PATCH v2] scsi: ufs: disable auto hibern8 while entering > suspend > > Hi, > I think content is lacking in the UFSHCI spec. > In the process, AH8 is in operation, and it seems that sending the command > hibern8 by manual has a defeat. > I don't know what the all hci vendor's hardware design will be, but there > is a possibility that ah8 and manual hibern8 may overlap. > So if is operating in ah8, it is thought that it will be safer to disable > ah8 before sending hibern8 command. > > > -----Original Message----- > > From: Stanley Chu [mailto:stanley.chu@mediatek.com] > > Sent: Wednesday, January 26, 2022 10:22 AM > > To: SEO HOYOUNG; linux-scsi@vger.kernel.org; > > linux-kernel@vger.kernel.org; alim.akhtar@samsung.com; > > avri.altman@wdc.com; jejb@linux.ibm.com; martin.petersen@oracle.com; > > beanhuo@micron.com; asutoshd@codeaurora.org; cang@codeaurora.org; > > bvanassche@acm.org; bhoon95.kim@samsung.com; kwmad.kim@samsung.com > > Cc: kernel test robot; peter.wang@mediatek.com > > Subject: Re: [PATCH v2] scsi: ufs: disable auto hibern8 while entering > > suspend > > > > Hi Hoyoung, > > > > On Tue, 2022-01-25 at 03:06 +0900, SEO HOYOUNG wrote: > > > v1-> v2: fixed no previous prototype warning > > > Reported-by: kernel test robot <lkp@intel.com> > > > > > > All warnings (new ones prefixed by >>): > > > > > drivers/scsi/ufs/ufshcd.c:4207:6: warning: no previous prototype > > > > > > for 'ufshcd_auto_hibern8_disable' [-Wmissing-prototypes] > > > 4207 | void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) > > > > > > If using auto hibern8 mode, need to disable auto hibern8 while > > > entering suspend. > > > When using auto hibern8 mode, it does not seem right to send a uic > > > command > > > > The UFSHCI spec does not mention the above rule. > > Why would you need to disable AH8 before using UIC command to enter H8? > > > > > for entry into hibern8 in the next > > > line(ufshcd_lik_state_transition(..)) > > > It seem right to send after disable auto hibern8. > > > > > > In addition, if the auto hibern8 mode supported, it is enabled in > > > resume. > > > So it seems that it will be paired only when auto hibern8 is > > > disabled while entering suspend. > > > > > > Signed-off-by: SEO HOYOUNG <hy50.seo@samsung.com> > > > --- > > > drivers/scsi/ufs/ufshcd.c | 14 ++++++++++++++ > > > 1 file changed, 14 insertions(+) > > > > > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c > > > index 460d2b440d2e..a6edbbd8ca2c 100644 > > > --- a/drivers/scsi/ufs/ufshcd.c > > > +++ b/drivers/scsi/ufs/ufshcd.c > > > @@ -254,6 +254,7 @@ static void > > > ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set); > > > static inline void ufshcd_wb_toggle_flush(struct ufs_hba *hba, bool > > > enable); static void ufshcd_hba_vreg_set_lpm(struct ufs_hba *hba); > > > static void ufshcd_hba_vreg_set_hpm(struct ufs_hba *hba); > > > +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba); > > > > > > static inline void ufshcd_enable_irq(struct ufs_hba *hba) { @@ > > > -4204,6 +4205,18 @@ void ufshcd_auto_hibern8_update(struct ufs_hba > > > *hba, u32 ahit) } EXPORT_SYMBOL_GPL(ufshcd_auto_hibern8_update); > > > > > > +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) { > > > + unsigned long flags; > > > + > > > + if (!ufshcd_is_auto_hibern8_supported(hba)) > > > + return; > > > + > > > + spin_lock_irqsave(hba->host->host_lock, flags); > > > + ufshcd_writel(hba, 0, REG_AUTO_HIBERNATE_IDLE_TIMER); > > > + spin_unlock_irqrestore(hba->host->host_lock, flags); } > > > + > > > void ufshcd_auto_hibern8_enable(struct ufs_hba *hba) { > > > unsigned long flags; > > > @@ -8925,6 +8938,7 @@ static int __ufshcd_wl_suspend(struct ufs_hba > > > *hba, enum ufs_pm_op pm_op) > > > * with the link off, so do not check for bkops. > > > */ > > > check_for_bkops = !ufshcd_is_ufs_dev_deepsleep(hba); > > > + ufshcd_auto_hibern8_disable(hba); > > > ret = ufshcd_link_state_transition(hba, req_link_state, > > > check_for_bkops); > > > if (ret) > > > goto set_dev_active;
> Hi, > Please check this patch. > If there is any other opinion, please give me comments Thanks > > > -----Original Message----- > > From: 서호영 [mailto:hy50.seo@samsung.com] > > Sent: Wednesday, January 26, 2022 2:35 PM > > To: 'linux-scsi@vger.kernel.org'; 'linux-kernel@vger.kernel.org'; > > 'alim.akhtar@samsung.com'; 'avri.altman@wdc.com'; > > 'jejb@linux.ibm.com'; 'martin.petersen@oracle.com'; > > 'beanhuo@micron.com'; 'asutoshd@codeaurora.org'; > > 'cang@codeaurora.org'; 'bvanassche@acm.org'; > 'bhoon95.kim@samsung.com'; 'kwmad.kim@samsung.com' > > Cc: 'kernel test robot' > > Subject: RE: [PATCH v2] scsi: ufs: disable auto hibern8 while entering > > suspend > > > > Hi, > > I think content is lacking in the UFSHCI spec. > > In the process, AH8 is in operation, and it seems that sending the > > command > > hibern8 by manual has a defeat. > > I don't know what the all hci vendor's hardware design will be, but > > there is a possibility that ah8 and manual hibern8 may overlap. > > So if is operating in ah8, it is thought that it will be safer to > > disable > > ah8 before sending hibern8 command. Hi, Thank you for your patch. Maybe better to promote your idea in JEDEC first. Thanks, Avri > > > > > -----Original Message----- > > > From: Stanley Chu [mailto:stanley.chu@mediatek.com] > > > Sent: Wednesday, January 26, 2022 10:22 AM > > > To: SEO HOYOUNG; linux-scsi@vger.kernel.org; > > > linux-kernel@vger.kernel.org; alim.akhtar@samsung.com; > > > avri.altman@wdc.com; jejb@linux.ibm.com; > martin.petersen@oracle.com; > > > beanhuo@micron.com; asutoshd@codeaurora.org; > cang@codeaurora.org; > > > bvanassche@acm.org; bhoon95.kim@samsung.com; > kwmad.kim@samsung.com > > > Cc: kernel test robot; peter.wang@mediatek.com > > > Subject: Re: [PATCH v2] scsi: ufs: disable auto hibern8 while > > > entering suspend > > > > > > Hi Hoyoung, > > > > > > On Tue, 2022-01-25 at 03:06 +0900, SEO HOYOUNG wrote: > > > > v1-> v2: fixed no previous prototype warning > > > > Reported-by: kernel test robot <lkp@intel.com> > > > > > > > > All warnings (new ones prefixed by >>): > > > > > > drivers/scsi/ufs/ufshcd.c:4207:6: warning: no previous > > > > > > prototype > > > > > > > > for 'ufshcd_auto_hibern8_disable' [-Wmissing-prototypes] > > > > 4207 | void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) > > > > > > > > If using auto hibern8 mode, need to disable auto hibern8 while > > > > entering suspend. > > > > When using auto hibern8 mode, it does not seem right to send a uic > > > > command > > > > > > The UFSHCI spec does not mention the above rule. > > > Why would you need to disable AH8 before using UIC command to enter > H8? > > > > > > > for entry into hibern8 in the next > > > > line(ufshcd_lik_state_transition(..)) > > > > It seem right to send after disable auto hibern8. > > > > > > > > In addition, if the auto hibern8 mode supported, it is enabled in > > > > resume. > > > > So it seems that it will be paired only when auto hibern8 is > > > > disabled while entering suspend. > > > > > > > > Signed-off-by: SEO HOYOUNG <hy50.seo@samsung.com> > > > > --- > > > > drivers/scsi/ufs/ufshcd.c | 14 ++++++++++++++ > > > > 1 file changed, 14 insertions(+) > > > > > > > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c > > > > index 460d2b440d2e..a6edbbd8ca2c 100644 > > > > --- a/drivers/scsi/ufs/ufshcd.c > > > > +++ b/drivers/scsi/ufs/ufshcd.c > > > > @@ -254,6 +254,7 @@ static void > > > > ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set); > > > > static inline void ufshcd_wb_toggle_flush(struct ufs_hba *hba, > > > > bool enable); static void ufshcd_hba_vreg_set_lpm(struct ufs_hba > > > > *hba); static void ufshcd_hba_vreg_set_hpm(struct ufs_hba *hba); > > > > +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba); > > > > > > > > static inline void ufshcd_enable_irq(struct ufs_hba *hba) { @@ > > > > -4204,6 +4205,18 @@ void ufshcd_auto_hibern8_update(struct > ufs_hba > > > > *hba, u32 ahit) } > EXPORT_SYMBOL_GPL(ufshcd_auto_hibern8_update); > > > > > > > > +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) { > > > > +unsigned long flags; > > > > + > > > > + if (!ufshcd_is_auto_hibern8_supported(hba)) > > > > + return; > > > > + > > > > + spin_lock_irqsave(hba->host->host_lock, flags); > > > > + ufshcd_writel(hba, 0, REG_AUTO_HIBERNATE_IDLE_TIMER); > > > > + spin_unlock_irqrestore(hba->host->host_lock, flags); } > > > > + > > > > void ufshcd_auto_hibern8_enable(struct ufs_hba *hba) { > > > > unsigned long flags; > > > > @@ -8925,6 +8938,7 @@ static int __ufshcd_wl_suspend(struct > > > > ufs_hba *hba, enum ufs_pm_op pm_op) > > > > * with the link off, so do not check for bkops. > > > > */ > > > > check_for_bkops = !ufshcd_is_ufs_dev_deepsleep(hba); > > > > + ufshcd_auto_hibern8_disable(hba); > > > > ret = ufshcd_link_state_transition(hba, req_link_state, > > > > check_for_bkops); > > > > if (ret) > > > > goto set_dev_active; > >
Hi Hoyoung >-----Original Message----- >From: Hoyoung SEO [mailto:hy50.seo@samsung.com] >Sent: Thursday, February 3, 2022 5:43 AM >To: linux-scsi@vger.kernel.org; linux-kernel@vger.kernel.org; >alim.akhtar@samsung.com; avri.altman@wdc.com; jejb@linux.ibm.com; >martin.petersen@oracle.com; beanhuo@micron.com; >asutoshd@codeaurora.org; cang@codeaurora.org; bvanassche@acm.org; >bhoon95.kim@samsung.com; kwmad.kim@samsung.com >Cc: 'kernel test robot' <lkp@intel.com> >Subject: RE: [PATCH v2] scsi: ufs: disable auto hibern8 while entering suspend > >Hi, >Please check this patch. >If there is any other opinion, please give me comments Thanks > >> -----Original Message----- >> From: 서호영 [mailto:hy50.seo@samsung.com] >> Sent: Wednesday, January 26, 2022 2:35 PM >> To: 'linux-scsi@vger.kernel.org'; 'linux-kernel@vger.kernel.org'; >> 'alim.akhtar@samsung.com'; 'avri.altman@wdc.com'; >> 'jejb@linux.ibm.com'; 'martin.petersen@oracle.com'; >> 'beanhuo@micron.com'; 'asutoshd@codeaurora.org'; >> 'cang@codeaurora.org'; 'bvanassche@acm.org'; >'bhoon95.kim@samsung.com'; 'kwmad.kim@samsung.com' >> Cc: 'kernel test robot' >> Subject: RE: [PATCH v2] scsi: ufs: disable auto hibern8 while entering >> suspend >> >> Hi, >> I think content is lacking in the UFSHCI spec. >> In the process, AH8 is in operation, and it seems that sending the >> command >> hibern8 by manual has a defeat. >> I don't know what the all hci vendor's hardware design will be, but >> there is a possibility that ah8 and manual hibern8 may overlap. >> So if is operating in ah8, it is thought that it will be safer to >> disable >> ah8 before sending hibern8 command. >> I am not sure, if this problem is generic and faced by all other UFS vendors. If not, how about having a vendor specific call back for your platform only? Just a thought. >> > -----Original Message----- >> > From: Stanley Chu [mailto:stanley.chu@mediatek.com] >> > Sent: Wednesday, January 26, 2022 10:22 AM >> > To: SEO HOYOUNG; linux-scsi@vger.kernel.org; >> > linux-kernel@vger.kernel.org; alim.akhtar@samsung.com; >> > avri.altman@wdc.com; jejb@linux.ibm.com; >martin.petersen@oracle.com; >> > beanhuo@micron.com; asutoshd@codeaurora.org; >cang@codeaurora.org; >> > bvanassche@acm.org; bhoon95.kim@samsung.com; >kwmad.kim@samsung.com >> > Cc: kernel test robot; peter.wang@mediatek.com >> > Subject: Re: [PATCH v2] scsi: ufs: disable auto hibern8 while >> > entering suspend >> > >> > Hi Hoyoung, >> > >> > On Tue, 2022-01-25 at 03:06 +0900, SEO HOYOUNG wrote: >> > > v1-> v2: fixed no previous prototype warning >> > > Reported-by: kernel test robot <lkp@intel.com> >> > > >> > > All warnings (new ones prefixed by >>): >> > > > > drivers/scsi/ufs/ufshcd.c:4207:6: warning: no previous >> > > > > prototype >> > > >> > > for 'ufshcd_auto_hibern8_disable' [-Wmissing-prototypes] >> > > 4207 | void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) >> > > >> > > If using auto hibern8 mode, need to disable auto hibern8 while >> > > entering suspend. >> > > When using auto hibern8 mode, it does not seem right to send a uic >> > > command >> > >> > The UFSHCI spec does not mention the above rule. >> > Why would you need to disable AH8 before using UIC command to enter >H8? >> > >> > > for entry into hibern8 in the next >> > > line(ufshcd_lik_state_transition(..)) >> > > It seem right to send after disable auto hibern8. >> > > >> > > In addition, if the auto hibern8 mode supported, it is enabled in >> > > resume. >> > > So it seems that it will be paired only when auto hibern8 is >> > > disabled while entering suspend. >> > > >> > > Signed-off-by: SEO HOYOUNG <hy50.seo@samsung.com> >> > > --- >> > > drivers/scsi/ufs/ufshcd.c | 14 ++++++++++++++ >> > > 1 file changed, 14 insertions(+) >> > > >> > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c >> > > index 460d2b440d2e..a6edbbd8ca2c 100644 >> > > --- a/drivers/scsi/ufs/ufshcd.c >> > > +++ b/drivers/scsi/ufs/ufshcd.c >> > > @@ -254,6 +254,7 @@ static void >> > > ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set); >> > > static inline void ufshcd_wb_toggle_flush(struct ufs_hba *hba, >> > > bool enable); static void ufshcd_hba_vreg_set_lpm(struct ufs_hba >> > > *hba); static void ufshcd_hba_vreg_set_hpm(struct ufs_hba *hba); >> > > +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba); >> > > >> > > static inline void ufshcd_enable_irq(struct ufs_hba *hba) { @@ >> > > -4204,6 +4205,18 @@ void ufshcd_auto_hibern8_update(struct ufs_hba >> > > *hba, u32 ahit) } >EXPORT_SYMBOL_GPL(ufshcd_auto_hibern8_update); >> > > >> > > +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) { >> > > + unsigned long flags; >> > > + >> > > + if (!ufshcd_is_auto_hibern8_supported(hba)) >> > > + return; >> > > + >> > > + spin_lock_irqsave(hba->host->host_lock, flags); >> > > + ufshcd_writel(hba, 0, REG_AUTO_HIBERNATE_IDLE_TIMER); >> > > + spin_unlock_irqrestore(hba->host->host_lock, flags); } >> > > + >> > > void ufshcd_auto_hibern8_enable(struct ufs_hba *hba) { >> > > unsigned long flags; >> > > @@ -8925,6 +8938,7 @@ static int __ufshcd_wl_suspend(struct >> > > ufs_hba *hba, enum ufs_pm_op pm_op) >> > > * with the link off, so do not check for bkops. >> > > */ >> > > check_for_bkops = !ufshcd_is_ufs_dev_deepsleep(hba); >> > > + ufshcd_auto_hibern8_disable(hba); >> > > ret = ufshcd_link_state_transition(hba, req_link_state, >> > > check_for_bkops); >> > > if (ret) >> > > goto set_dev_active; > >
On 2/2/22 22:50, Alim Akhtar wrote: > I am not sure, if this problem is generic and faced by all other UFS vendors. > If not, how about having a vendor specific call back for your platform only? > Just a thought. I agree with the above. Since the code change does not follow from the spec, it should be guarded with a check for a new quirk flag. Thanks, Bart.
diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c index 460d2b440d2e..a6edbbd8ca2c 100644 --- a/drivers/scsi/ufs/ufshcd.c +++ b/drivers/scsi/ufs/ufshcd.c @@ -254,6 +254,7 @@ static void ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set); static inline void ufshcd_wb_toggle_flush(struct ufs_hba *hba, bool enable); static void ufshcd_hba_vreg_set_lpm(struct ufs_hba *hba); static void ufshcd_hba_vreg_set_hpm(struct ufs_hba *hba); +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba); static inline void ufshcd_enable_irq(struct ufs_hba *hba) { @@ -4204,6 +4205,18 @@ void ufshcd_auto_hibern8_update(struct ufs_hba *hba, u32 ahit) } EXPORT_SYMBOL_GPL(ufshcd_auto_hibern8_update); +static void ufshcd_auto_hibern8_disable(struct ufs_hba *hba) +{ + unsigned long flags; + + if (!ufshcd_is_auto_hibern8_supported(hba)) + return; + + spin_lock_irqsave(hba->host->host_lock, flags); + ufshcd_writel(hba, 0, REG_AUTO_HIBERNATE_IDLE_TIMER); + spin_unlock_irqrestore(hba->host->host_lock, flags); +} + void ufshcd_auto_hibern8_enable(struct ufs_hba *hba) { unsigned long flags; @@ -8925,6 +8938,7 @@ static int __ufshcd_wl_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op) * with the link off, so do not check for bkops. */ check_for_bkops = !ufshcd_is_ufs_dev_deepsleep(hba); + ufshcd_auto_hibern8_disable(hba); ret = ufshcd_link_state_transition(hba, req_link_state, check_for_bkops); if (ret) goto set_dev_active;