Message ID | 1583322863-21790-1-git-send-email-vbadigan@codeaurora.org (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | [V2] mmc: sdhci-msm: Disable CQE during SDHC reset | expand |
On 4/03/20 1:54 pm, Veerabhadrarao Badiganti wrote: > When SDHC gets reset (E.g. in suspend path), CQE also gets reset > and goes to disable state. But s/w state still points it as CQE > is in enabled state. Since s/w and h/w states goes out of sync, > it results in s/w request timeout for subsequent CQE requests. > > To synchronize CQE s/w and h/w state during SDHC reset, > explicitly disable CQE after reset. Shouldn't you be calling cqhci_suspend() / cqhci_resume() in the suspend and resume paths? > > Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org> > --- > Changes since V1: > - Disable CQE only when SDHC undergoes s/w reset for all. > --- > drivers/mmc/host/sdhci-msm.c | 9 ++++++++- > 1 file changed, 8 insertions(+), 1 deletion(-) > > diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c > index 53b79ee..75929d3 100644 > --- a/drivers/mmc/host/sdhci-msm.c > +++ b/drivers/mmc/host/sdhci-msm.c > @@ -1823,6 +1823,13 @@ static void sdhci_msm_set_regulator_caps(struct sdhci_msm_host *msm_host) > pr_debug("%s: supported caps: 0x%08x\n", mmc_hostname(mmc), caps); > } > > +static void sdhci_msm_reset(struct sdhci_host *host, u8 mask) > +{ > + sdhci_reset(host, mask); > + if ((host->mmc->caps2 & MMC_CAP2_CQE) && (mask & SDHCI_RESET_ALL)) > + cqhci_suspend(host->mmc); > +} > + > static const struct sdhci_msm_variant_ops mci_var_ops = { > .msm_readl_relaxed = sdhci_msm_mci_variant_readl_relaxed, > .msm_writel_relaxed = sdhci_msm_mci_variant_writel_relaxed, > @@ -1861,7 +1868,7 @@ static void sdhci_msm_set_regulator_caps(struct sdhci_msm_host *msm_host) > MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match); > > static const struct sdhci_ops sdhci_msm_ops = { > - .reset = sdhci_reset, > + .reset = sdhci_msm_reset, > .set_clock = sdhci_msm_set_clock, > .get_min_clock = sdhci_msm_get_min_clock, > .get_max_clock = sdhci_msm_get_max_clock, >
Hi Adrian On 3/4/2020 5:58 PM, Adrian Hunter wrote: > On 4/03/20 1:54 pm, Veerabhadrarao Badiganti wrote: >> When SDHC gets reset (E.g. in suspend path), CQE also gets reset >> and goes to disable state. But s/w state still points it as CQE >> is in enabled state. Since s/w and h/w states goes out of sync, >> it results in s/w request timeout for subsequent CQE requests. >> >> To synchronize CQE s/w and h/w state during SDHC reset, >> explicitly disable CQE after reset. > Shouldn't you be calling cqhci_suspend() / cqhci_resume() in the suspend and > resume paths? This issue is seen during mmc runtime suspend. I can add it sdhci_msm_runtime_suspend but sdhci_msm runtime delay is aggressive, its 50ms. It may get invoked very frequently. So Im of the opinion that disabling CQE very often from platform runtime suspend is overkill. >> Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org> >> --- >> Changes since V1: >> - Disable CQE only when SDHC undergoes s/w reset for all. >> --- >> drivers/mmc/host/sdhci-msm.c | 9 ++++++++- >> 1 file changed, 8 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c >> index 53b79ee..75929d3 100644 >> --- a/drivers/mmc/host/sdhci-msm.c >> +++ b/drivers/mmc/host/sdhci-msm.c >> @@ -1823,6 +1823,13 @@ static void sdhci_msm_set_regulator_caps(struct sdhci_msm_host *msm_host) >> pr_debug("%s: supported caps: 0x%08x\n", mmc_hostname(mmc), caps); >> } >> >> +static void sdhci_msm_reset(struct sdhci_host *host, u8 mask) >> +{ >> + sdhci_reset(host, mask); >> + if ((host->mmc->caps2 & MMC_CAP2_CQE) && (mask & SDHCI_RESET_ALL)) >> + cqhci_suspend(host->mmc); >> +} >> + >> static const struct sdhci_msm_variant_ops mci_var_ops = { >> .msm_readl_relaxed = sdhci_msm_mci_variant_readl_relaxed, >> .msm_writel_relaxed = sdhci_msm_mci_variant_writel_relaxed, >> @@ -1861,7 +1868,7 @@ static void sdhci_msm_set_regulator_caps(struct sdhci_msm_host *msm_host) >> MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match); >> >> static const struct sdhci_ops sdhci_msm_ops = { >> - .reset = sdhci_reset, >> + .reset = sdhci_msm_reset, >> .set_clock = sdhci_msm_set_clock, >> .get_min_clock = sdhci_msm_get_min_clock, >> .get_max_clock = sdhci_msm_get_max_clock, >>
On 4/03/20 3:10 pm, Veerabhadrarao Badiganti wrote: > Hi Adrian > > On 3/4/2020 5:58 PM, Adrian Hunter wrote: >> On 4/03/20 1:54 pm, Veerabhadrarao Badiganti wrote: >>> When SDHC gets reset (E.g. in suspend path), CQE also gets reset >>> and goes to disable state. But s/w state still points it as CQE >>> is in enabled state. Since s/w and h/w states goes out of sync, >>> it results in s/w request timeout for subsequent CQE requests. >>> >>> To synchronize CQE s/w and h/w state during SDHC reset, >>> explicitly disable CQE after reset. >> Shouldn't you be calling cqhci_suspend() / cqhci_resume() in the suspend and >> resume paths? > > This issue is seen during mmc runtime suspend. I can add it > sdhci_msm_runtime_suspend > > but sdhci_msm runtime delay is aggressive, its 50ms. It may get invoked very > frequently. > > So Im of the opinion that disabling CQE very often from platform runtime > suspend is overkill. It doesn't look like sdhci-msm calls any sdhci.c pm ops, so how does SDHC get reset? > >>> Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org> >>> --- >>> Changes since V1: >>> - Disable CQE only when SDHC undergoes s/w reset for all. >>> --- >>> drivers/mmc/host/sdhci-msm.c | 9 ++++++++- >>> 1 file changed, 8 insertions(+), 1 deletion(-) >>> >>> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c >>> index 53b79ee..75929d3 100644 >>> --- a/drivers/mmc/host/sdhci-msm.c >>> +++ b/drivers/mmc/host/sdhci-msm.c >>> @@ -1823,6 +1823,13 @@ static void sdhci_msm_set_regulator_caps(struct >>> sdhci_msm_host *msm_host) >>> pr_debug("%s: supported caps: 0x%08x\n", mmc_hostname(mmc), caps); >>> } >>> +static void sdhci_msm_reset(struct sdhci_host *host, u8 mask) >>> +{ >>> + sdhci_reset(host, mask); >>> + if ((host->mmc->caps2 & MMC_CAP2_CQE) && (mask & SDHCI_RESET_ALL)) >>> + cqhci_suspend(host->mmc); >>> +} >>> + >>> static const struct sdhci_msm_variant_ops mci_var_ops = { >>> .msm_readl_relaxed = sdhci_msm_mci_variant_readl_relaxed, >>> .msm_writel_relaxed = sdhci_msm_mci_variant_writel_relaxed, >>> @@ -1861,7 +1868,7 @@ static void sdhci_msm_set_regulator_caps(struct >>> sdhci_msm_host *msm_host) >>> MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match); >>> static const struct sdhci_ops sdhci_msm_ops = { >>> - .reset = sdhci_reset, >>> + .reset = sdhci_msm_reset, >>> .set_clock = sdhci_msm_set_clock, >>> .get_min_clock = sdhci_msm_get_min_clock, >>> .get_max_clock = sdhci_msm_get_max_clock, >>>
On 3/4/2020 7:40 PM, Adrian Hunter wrote: > On 4/03/20 3:10 pm, Veerabhadrarao Badiganti wrote: >> Hi Adrian >> >> On 3/4/2020 5:58 PM, Adrian Hunter wrote: >>> On 4/03/20 1:54 pm, Veerabhadrarao Badiganti wrote: >>>> When SDHC gets reset (E.g. in suspend path), CQE also gets reset >>>> and goes to disable state. But s/w state still points it as CQE >>>> is in enabled state. Since s/w and h/w states goes out of sync, >>>> it results in s/w request timeout for subsequent CQE requests. >>>> >>>> To synchronize CQE s/w and h/w state during SDHC reset, >>>> explicitly disable CQE after reset. >>> Shouldn't you be calling cqhci_suspend() / cqhci_resume() in the suspend and >>> resume paths? >> This issue is seen during mmc runtime suspend. I can add it >> sdhci_msm_runtime_suspend >> >> but sdhci_msm runtime delay is aggressive, its 50ms. It may get invoked very >> frequently. >> >> So Im of the opinion that disabling CQE very often from platform runtime >> suspend is overkill. > It doesn't look like sdhci-msm calls any sdhci.c pm ops, so how does SDHC > get reset? With MMC_CAP_AGGRESSIVE_PM flag enabled, it getting called from mmc_runtime_suspend() Below is the call stack() sdhci_reset sdhci_do_reset sdhci_init sdhci_set_ios mmc_set_initial_state mmc_power_off _mmc_suspend mmc_runtime_suspend >>>> Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org> >>>> --- >>>> Changes since V1: >>>> - Disable CQE only when SDHC undergoes s/w reset for all. >>>> --- >>>> drivers/mmc/host/sdhci-msm.c | 9 ++++++++- >>>> 1 file changed, 8 insertions(+), 1 deletion(-) >>>> >>>> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c >>>> index 53b79ee..75929d3 100644 >>>> --- a/drivers/mmc/host/sdhci-msm.c >>>> +++ b/drivers/mmc/host/sdhci-msm.c >>>> @@ -1823,6 +1823,13 @@ static void sdhci_msm_set_regulator_caps(struct >>>> sdhci_msm_host *msm_host) >>>> pr_debug("%s: supported caps: 0x%08x\n", mmc_hostname(mmc), caps); >>>> } >>>> +static void sdhci_msm_reset(struct sdhci_host *host, u8 mask) >>>> +{ >>>> + sdhci_reset(host, mask); >>>> + if ((host->mmc->caps2 & MMC_CAP2_CQE) && (mask & SDHCI_RESET_ALL)) >>>> + cqhci_suspend(host->mmc); >>>> +} >>>> + >>>> static const struct sdhci_msm_variant_ops mci_var_ops = { >>>> .msm_readl_relaxed = sdhci_msm_mci_variant_readl_relaxed, >>>> .msm_writel_relaxed = sdhci_msm_mci_variant_writel_relaxed, >>>> @@ -1861,7 +1868,7 @@ static void sdhci_msm_set_regulator_caps(struct >>>> sdhci_msm_host *msm_host) >>>> MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match); >>>> static const struct sdhci_ops sdhci_msm_ops = { >>>> - .reset = sdhci_reset, >>>> + .reset = sdhci_msm_reset, >>>> .set_clock = sdhci_msm_set_clock, >>>> .get_min_clock = sdhci_msm_get_min_clock, >>>> .get_max_clock = sdhci_msm_get_max_clock, >>>>
On 4/03/20 6:50 pm, Veerabhadrarao Badiganti wrote: > > On 3/4/2020 7:40 PM, Adrian Hunter wrote: >> On 4/03/20 3:10 pm, Veerabhadrarao Badiganti wrote: >>> Hi Adrian >>> >>> On 3/4/2020 5:58 PM, Adrian Hunter wrote: >>>> On 4/03/20 1:54 pm, Veerabhadrarao Badiganti wrote: >>>>> When SDHC gets reset (E.g. in suspend path), CQE also gets reset >>>>> and goes to disable state. But s/w state still points it as CQE >>>>> is in enabled state. Since s/w and h/w states goes out of sync, >>>>> it results in s/w request timeout for subsequent CQE requests. >>>>> >>>>> To synchronize CQE s/w and h/w state during SDHC reset, >>>>> explicitly disable CQE after reset. >>>> Shouldn't you be calling cqhci_suspend() / cqhci_resume() in the suspend >>>> and >>>> resume paths? >>> This issue is seen during mmc runtime suspend. I can add it >>> sdhci_msm_runtime_suspend >>> >>> but sdhci_msm runtime delay is aggressive, its 50ms. It may get invoked very >>> frequently. >>> >>> So Im of the opinion that disabling CQE very often from platform runtime >>> suspend is overkill. >> It doesn't look like sdhci-msm calls any sdhci.c pm ops, so how does SDHC >> get reset? > > With MMC_CAP_AGGRESSIVE_PM flag enabled, it getting called from > mmc_runtime_suspend() > > Below is the call stack() > > sdhci_reset > sdhci_do_reset > sdhci_init > sdhci_set_ios > mmc_set_initial_state > mmc_power_off > _mmc_suspend > mmc_runtime_suspend > OK, cqhci_suspend does the right thing, but it is not an appropriate function for this. I suggest introducing cqhci_deactivate() as below. From: Adrian Hunter <adrian.hunter@intel.com> Date: Thu, 5 Mar 2020 10:42:09 +0200 Subject: [PATCH] mmc: cqhci: Add cqhci_deactivate() Host controllers can reset CQHCI either directly or as a consequence of host controller reset. Add cqhci_deactivate() which puts the CQHCI driver into a state that is consistent with that. Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> --- drivers/mmc/host/cqhci.c | 6 +++--- drivers/mmc/host/cqhci.h | 5 ++++- 2 files changed, 7 insertions(+), 4 deletions(-) diff --git a/drivers/mmc/host/cqhci.c b/drivers/mmc/host/cqhci.c index e2ea2c4b6b94..d8d024a1682b 100644 --- a/drivers/mmc/host/cqhci.c +++ b/drivers/mmc/host/cqhci.c @@ -298,16 +298,16 @@ static void __cqhci_disable(struct cqhci_host *cq_host) cq_host->activated = false; } -int cqhci_suspend(struct mmc_host *mmc) +int cqhci_deactivate(struct mmc_host *mmc) { struct cqhci_host *cq_host = mmc->cqe_private; - if (cq_host->enabled) + if (cq_host->enabled && cq_host->activated) __cqhci_disable(cq_host); return 0; } -EXPORT_SYMBOL(cqhci_suspend); +EXPORT_SYMBOL(cqhci_deactivate); int cqhci_resume(struct mmc_host *mmc) { diff --git a/drivers/mmc/host/cqhci.h b/drivers/mmc/host/cqhci.h index def76e9b5cac..8648846a0213 100644 --- a/drivers/mmc/host/cqhci.h +++ b/drivers/mmc/host/cqhci.h @@ -230,7 +230,10 @@ irqreturn_t cqhci_irq(struct mmc_host *mmc, u32 intmask, int cmd_error, int data_error); int cqhci_init(struct cqhci_host *cq_host, struct mmc_host *mmc, bool dma64); struct cqhci_host *cqhci_pltfm_init(struct platform_device *pdev); -int cqhci_suspend(struct mmc_host *mmc); +static inline int cqhci_suspend(struct mmc_host *mmc) +{ + return cqhci_deactivate(mmc); +} int cqhci_resume(struct mmc_host *mmc); #endif
On 3/5/2020 2:29 PM, Adrian Hunter wrote: > On 4/03/20 6:50 pm, Veerabhadrarao Badiganti wrote: >> On 3/4/2020 7:40 PM, Adrian Hunter wrote: >>> On 4/03/20 3:10 pm, Veerabhadrarao Badiganti wrote: >>>> Hi Adrian >>>> >>>> On 3/4/2020 5:58 PM, Adrian Hunter wrote: >>>>> On 4/03/20 1:54 pm, Veerabhadrarao Badiganti wrote: >>>>>> When SDHC gets reset (E.g. in suspend path), CQE also gets reset >>>>>> and goes to disable state. But s/w state still points it as CQE >>>>>> is in enabled state. Since s/w and h/w states goes out of sync, >>>>>> it results in s/w request timeout for subsequent CQE requests. >>>>>> >>>>>> To synchronize CQE s/w and h/w state during SDHC reset, >>>>>> explicitly disable CQE after reset. >>>>> Shouldn't you be calling cqhci_suspend() / cqhci_resume() in the suspend >>>>> and >>>>> resume paths? >>>> This issue is seen during mmc runtime suspend. I can add it >>>> sdhci_msm_runtime_suspend >>>> >>>> but sdhci_msm runtime delay is aggressive, its 50ms. It may get invoked very >>>> frequently. >>>> >>>> So Im of the opinion that disabling CQE very often from platform runtime >>>> suspend is overkill. >>> It doesn't look like sdhci-msm calls any sdhci.c pm ops, so how does SDHC >>> get reset? >> With MMC_CAP_AGGRESSIVE_PM flag enabled, it getting called from >> mmc_runtime_suspend() >> >> Below is the call stack() >> >> sdhci_reset >> sdhci_do_reset >> sdhci_init >> sdhci_set_ios >> mmc_set_initial_state >> mmc_power_off >> _mmc_suspend >> mmc_runtime_suspend >> > OK, cqhci_suspend does the right thing, but it is not an > appropriate function for this. I suggest introducing > cqhci_deactivate() as below. > > From: Adrian Hunter <adrian.hunter@intel.com> > Date: Thu, 5 Mar 2020 10:42:09 +0200 > Subject: [PATCH] mmc: cqhci: Add cqhci_deactivate() > > Host controllers can reset CQHCI either directly or as a consequence of > host controller reset. Add cqhci_deactivate() which puts the CQHCI > driver into a state that is consistent with that. > > Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> > --- > drivers/mmc/host/cqhci.c | 6 +++--- > drivers/mmc/host/cqhci.h | 5 ++++- > 2 files changed, 7 insertions(+), 4 deletions(-) > > diff --git a/drivers/mmc/host/cqhci.c b/drivers/mmc/host/cqhci.c > index e2ea2c4b6b94..d8d024a1682b 100644 > --- a/drivers/mmc/host/cqhci.c > +++ b/drivers/mmc/host/cqhci.c > @@ -298,16 +298,16 @@ static void __cqhci_disable(struct cqhci_host *cq_host) > cq_host->activated = false; > } > > -int cqhci_suspend(struct mmc_host *mmc) > +int cqhci_deactivate(struct mmc_host *mmc) > { > struct cqhci_host *cq_host = mmc->cqe_private; > > - if (cq_host->enabled) > + if (cq_host->enabled && cq_host->activated) > __cqhci_disable(cq_host); > > return 0; > } > -EXPORT_SYMBOL(cqhci_suspend); > +EXPORT_SYMBOL(cqhci_deactivate); > > int cqhci_resume(struct mmc_host *mmc) > { > diff --git a/drivers/mmc/host/cqhci.h b/drivers/mmc/host/cqhci.h > index def76e9b5cac..8648846a0213 100644 > --- a/drivers/mmc/host/cqhci.h > +++ b/drivers/mmc/host/cqhci.h > @@ -230,7 +230,10 @@ irqreturn_t cqhci_irq(struct mmc_host *mmc, u32 intmask, int cmd_error, > int data_error); > int cqhci_init(struct cqhci_host *cq_host, struct mmc_host *mmc, bool dma64); > struct cqhci_host *cqhci_pltfm_init(struct platform_device *pdev); > -int cqhci_suspend(struct mmc_host *mmc); > +static inline int cqhci_suspend(struct mmc_host *mmc) > +{ > + return cqhci_deactivate(mmc); > +} > int cqhci_resume(struct mmc_host *mmc); > > #endif Thanks Adrian for the suggestion. Will post this change and my updated fix. Thanks Veera
diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c index 53b79ee..75929d3 100644 --- a/drivers/mmc/host/sdhci-msm.c +++ b/drivers/mmc/host/sdhci-msm.c @@ -1823,6 +1823,13 @@ static void sdhci_msm_set_regulator_caps(struct sdhci_msm_host *msm_host) pr_debug("%s: supported caps: 0x%08x\n", mmc_hostname(mmc), caps); } +static void sdhci_msm_reset(struct sdhci_host *host, u8 mask) +{ + sdhci_reset(host, mask); + if ((host->mmc->caps2 & MMC_CAP2_CQE) && (mask & SDHCI_RESET_ALL)) + cqhci_suspend(host->mmc); +} + static const struct sdhci_msm_variant_ops mci_var_ops = { .msm_readl_relaxed = sdhci_msm_mci_variant_readl_relaxed, .msm_writel_relaxed = sdhci_msm_mci_variant_writel_relaxed, @@ -1861,7 +1868,7 @@ static void sdhci_msm_set_regulator_caps(struct sdhci_msm_host *msm_host) MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match); static const struct sdhci_ops sdhci_msm_ops = { - .reset = sdhci_reset, + .reset = sdhci_msm_reset, .set_clock = sdhci_msm_set_clock, .get_min_clock = sdhci_msm_get_min_clock, .get_max_clock = sdhci_msm_get_max_clock,
When SDHC gets reset (E.g. in suspend path), CQE also gets reset and goes to disable state. But s/w state still points it as CQE is in enabled state. Since s/w and h/w states goes out of sync, it results in s/w request timeout for subsequent CQE requests. To synchronize CQE s/w and h/w state during SDHC reset, explicitly disable CQE after reset. Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org> --- Changes since V1: - Disable CQE only when SDHC undergoes s/w reset for all. --- drivers/mmc/host/sdhci-msm.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)