| Message ID | 20200618140312.155157-1-jagan@amarulasolutions.com |
|---|---|
| State | Accepted |
| Commit | f12341a9529540113f01989149bbbeb68662a829 |
| Headers | show |
| Series | [v4] mmc: sdhci: Fix HISPD bit handling | expand |
Jaehoon, > Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling Are you fine with this v4? Thanks, Peng. > > SDHCI HISPD bits need to be configured based on desired mmc timings mode > and some HISPD quirks. > > So, handle the HISPD bit based on the mmc computed selected mode(timing > parameter) rather than fixed mmc card clock frequency. > > Linux handle the HISPD similar like this in below commit but no > SDHCI_QUIRK_BROKEN_HISPD_MODE, > > commit <501639bf2173> ("mmc: sdhci: fix SDHCI_QUIRK_NO_HISPD_BIT > handling") > > This eventually fixed the mmc write issue observed in > rk3399 sdhci controller. > > Bug log for refernece, > => gpt write mmc 0 $partitions > Writing GPT: mmc write failed > ** Can't write to device 0 ** > ** Can't write to device 0 ** > error! > > Cc: Kever Yang <kever.yang at rock-chips.com> > Cc: Peng Fan <peng.fan at nxp.com> > Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # roc-rk3399-pc > Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> > --- > Changes for v4: > - update commit message > - simplify the logic. > > drivers/mmc/sdhci.c | 23 +++++++++++++++++------ > 1 file changed, 17 insertions(+), 6 deletions(-) > > diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index > 92cc8434af..6cb702111b 100644 > --- a/drivers/mmc/sdhci.c > +++ b/drivers/mmc/sdhci.c > @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) #endif > u32 ctrl; > struct sdhci_host *host = mmc->priv; > + bool no_hispd_bit = false; > > if (host->ops && host->ops->set_control_reg) > host->ops->set_control_reg(host); > @@ -594,14 +595,24 @@ static int sdhci_set_ios(struct mmc *mmc) > ctrl &= ~SDHCI_CTRL_4BITBUS; > } > > - if (mmc->clock > 26000000) > - ctrl |= SDHCI_CTRL_HISPD; > - else > - ctrl &= ~SDHCI_CTRL_HISPD; > - > if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || > (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) > - ctrl &= ~SDHCI_CTRL_HISPD; > + no_hispd_bit = true; > + > + if (!no_hispd_bit) { > + if (mmc->selected_mode == MMC_HS || > + mmc->selected_mode == SD_HS || > + mmc->selected_mode == MMC_DDR_52 || > + mmc->selected_mode == MMC_HS_200 || > + mmc->selected_mode == MMC_HS_400 || > + mmc->selected_mode == UHS_SDR25 || > + mmc->selected_mode == UHS_SDR50 || > + mmc->selected_mode == UHS_SDR104 || > + mmc->selected_mode == UHS_DDR50) > + ctrl |= SDHCI_CTRL_HISPD; > + else > + ctrl &= ~SDHCI_CTRL_HISPD; > + } > > sdhci_writeb(host, ctrl, SDHCI_HOST_CONTROL); > > -- > 2.25.1
Hi Peng, On 6/22/20 10:49 AM, Peng Fan wrote: > Jaehoon, > >> Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling > > Are you fine with this v4? Thanks for sharing it. i didn't see patch v4. > > Thanks, > Peng. > >> >> SDHCI HISPD bits need to be configured based on desired mmc timings mode >> and some HISPD quirks. >> >> So, handle the HISPD bit based on the mmc computed selected mode(timing >> parameter) rather than fixed mmc card clock frequency. >> >> Linux handle the HISPD similar like this in below commit but no >> SDHCI_QUIRK_BROKEN_HISPD_MODE, >> >> commit <501639bf2173> ("mmc: sdhci: fix SDHCI_QUIRK_NO_HISPD_BIT >> handling") >> >> This eventually fixed the mmc write issue observed in >> rk3399 sdhci controller. >> >> Bug log for refernece, >> => gpt write mmc 0 $partitions >> Writing GPT: mmc write failed >> ** Can't write to device 0 ** >> ** Can't write to device 0 ** >> error! >> >> Cc: Kever Yang <kever.yang at rock-chips.com> >> Cc: Peng Fan <peng.fan at nxp.com> >> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # roc-rk3399-pc >> Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> >> --- >> Changes for v4: >> - update commit message >> - simplify the logic. >> >> drivers/mmc/sdhci.c | 23 +++++++++++++++++------ >> 1 file changed, 17 insertions(+), 6 deletions(-) >> >> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index >> 92cc8434af..6cb702111b 100644 >> --- a/drivers/mmc/sdhci.c >> +++ b/drivers/mmc/sdhci.c >> @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) #endif >> u32 ctrl; >> struct sdhci_host *host = mmc->priv; >> + bool no_hispd_bit = false; >> >> if (host->ops && host->ops->set_control_reg) >> host->ops->set_control_reg(host); >> @@ -594,14 +595,24 @@ static int sdhci_set_ios(struct mmc *mmc) >> ctrl &= ~SDHCI_CTRL_4BITBUS; >> } >> >> - if (mmc->clock > 26000000) >> - ctrl |= SDHCI_CTRL_HISPD; >> - else >> - ctrl &= ~SDHCI_CTRL_HISPD; >> - >> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || >> (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) >> - ctrl &= ~SDHCI_CTRL_HISPD; >> + no_hispd_bit = true; No. ctrl variable is set to bit[2] of HOST_CONTROL register. But Some samsung SoCs are using its bit[2] as other purpose. So it needs to revert the value with "ctrl &= ~SDHCI_CTRL_HISPD". Because it's possible to set to 1. SDHCI_QUIRK_NO_HISPD_BIT means that it's used other purpose. Best Regards, Jaehoon Chung >> + >> + if (!no_hispd_bit) { >> + if (mmc->selected_mode == MMC_HS || >> + mmc->selected_mode == SD_HS || >> + mmc->selected_mode == MMC_DDR_52 || >> + mmc->selected_mode == MMC_HS_200 || >> + mmc->selected_mode == MMC_HS_400 || >> + mmc->selected_mode == UHS_SDR25 || >> + mmc->selected_mode == UHS_SDR50 || >> + mmc->selected_mode == UHS_SDR104 || >> + mmc->selected_mode == UHS_DDR50) >> + ctrl |= SDHCI_CTRL_HISPD; >> + else >> + ctrl &= ~SDHCI_CTRL_HISPD; >> + } >> >> sdhci_writeb(host, ctrl, SDHCI_HOST_CONTROL); >> >> -- >> 2.25.1 > >
On Mon, Jun 22, 2020 at 2:54 PM Jaehoon Chung <jh80.chung at samsung.com> wrote: > > Hi Peng, > > On 6/22/20 10:49 AM, Peng Fan wrote: > > Jaehoon, > > > >> Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling > > > > Are you fine with this v4? > > Thanks for sharing it. i didn't see patch v4. > > > > > Thanks, > > Peng. > > > >> > >> SDHCI HISPD bits need to be configured based on desired mmc timings mode > >> and some HISPD quirks. > >> > >> So, handle the HISPD bit based on the mmc computed selected mode(timing > >> parameter) rather than fixed mmc card clock frequency. > >> > >> Linux handle the HISPD similar like this in below commit but no > >> SDHCI_QUIRK_BROKEN_HISPD_MODE, > >> > >> commit <501639bf2173> ("mmc: sdhci: fix SDHCI_QUIRK_NO_HISPD_BIT > >> handling") > >> > >> This eventually fixed the mmc write issue observed in > >> rk3399 sdhci controller. > >> > >> Bug log for refernece, > >> => gpt write mmc 0 $partitions > >> Writing GPT: mmc write failed > >> ** Can't write to device 0 ** > >> ** Can't write to device 0 ** > >> error! > >> > >> Cc: Kever Yang <kever.yang at rock-chips.com> > >> Cc: Peng Fan <peng.fan at nxp.com> > >> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # roc-rk3399-pc > >> Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> > >> --- > >> Changes for v4: > >> - update commit message > >> - simplify the logic. > >> > >> drivers/mmc/sdhci.c | 23 +++++++++++++++++------ > >> 1 file changed, 17 insertions(+), 6 deletions(-) > >> > >> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index > >> 92cc8434af..6cb702111b 100644 > >> --- a/drivers/mmc/sdhci.c > >> +++ b/drivers/mmc/sdhci.c > >> @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) #endif > >> u32 ctrl; > >> struct sdhci_host *host = mmc->priv; > >> + bool no_hispd_bit = false; > >> > >> if (host->ops && host->ops->set_control_reg) > >> host->ops->set_control_reg(host); > >> @@ -594,14 +595,24 @@ static int sdhci_set_ios(struct mmc *mmc) > >> ctrl &= ~SDHCI_CTRL_4BITBUS; > >> } > >> > >> - if (mmc->clock > 26000000) > >> - ctrl |= SDHCI_CTRL_HISPD; > >> - else > >> - ctrl &= ~SDHCI_CTRL_HISPD; > >> - > >> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || > >> (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) > >> - ctrl &= ~SDHCI_CTRL_HISPD; > >> + no_hispd_bit = true; > > No. ctrl variable is set to bit[2] of HOST_CONTROL register. > But Some samsung SoCs are using its bit[2] as other purpose. > So it needs to revert the value with "ctrl &= ~SDHCI_CTRL_HISPD". > Because it's possible to set to 1. > > SDHCI_QUIRK_NO_HISPD_BIT means that it's used other purpose. So, what are you suggesting? could you please elaborate? Jagan.
On 6/22/20 6:26 PM, Jagan Teki wrote: > On Mon, Jun 22, 2020 at 2:54 PM Jaehoon Chung <jh80.chung at samsung.com> wrote: >> >> Hi Peng, >> >> On 6/22/20 10:49 AM, Peng Fan wrote: >>> Jaehoon, >>> >>>> Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling >>> >>> Are you fine with this v4? >> >> Thanks for sharing it. i didn't see patch v4. >> >>> >>> Thanks, >>> Peng. >>> >>>> >>>> SDHCI HISPD bits need to be configured based on desired mmc timings mode >>>> and some HISPD quirks. >>>> >>>> So, handle the HISPD bit based on the mmc computed selected mode(timing >>>> parameter) rather than fixed mmc card clock frequency. >>>> >>>> Linux handle the HISPD similar like this in below commit but no >>>> SDHCI_QUIRK_BROKEN_HISPD_MODE, >>>> >>>> commit <501639bf2173> ("mmc: sdhci: fix SDHCI_QUIRK_NO_HISPD_BIT >>>> handling") >>>> >>>> This eventually fixed the mmc write issue observed in >>>> rk3399 sdhci controller. >>>> >>>> Bug log for refernece, >>>> => gpt write mmc 0 $partitions >>>> Writing GPT: mmc write failed >>>> ** Can't write to device 0 ** >>>> ** Can't write to device 0 ** >>>> error! >>>> >>>> Cc: Kever Yang <kever.yang at rock-chips.com> >>>> Cc: Peng Fan <peng.fan at nxp.com> >>>> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # roc-rk3399-pc >>>> Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> >>>> --- >>>> Changes for v4: >>>> - update commit message >>>> - simplify the logic. >>>> >>>> drivers/mmc/sdhci.c | 23 +++++++++++++++++------ >>>> 1 file changed, 17 insertions(+), 6 deletions(-) >>>> >>>> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index >>>> 92cc8434af..6cb702111b 100644 >>>> --- a/drivers/mmc/sdhci.c >>>> +++ b/drivers/mmc/sdhci.c >>>> @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) #endif >>>> u32 ctrl; >>>> struct sdhci_host *host = mmc->priv; >>>> + bool no_hispd_bit = false; >>>> >>>> if (host->ops && host->ops->set_control_reg) >>>> host->ops->set_control_reg(host); >>>> @@ -594,14 +595,24 @@ static int sdhci_set_ios(struct mmc *mmc) >>>> ctrl &= ~SDHCI_CTRL_4BITBUS; >>>> } >>>> >>>> - if (mmc->clock > 26000000) >>>> - ctrl |= SDHCI_CTRL_HISPD; >>>> - else >>>> - ctrl &= ~SDHCI_CTRL_HISPD; >>>> - >>>> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || >>>> (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) >>>> - ctrl &= ~SDHCI_CTRL_HISPD; >>>> + no_hispd_bit = true; >> >> No. ctrl variable is set to bit[2] of HOST_CONTROL register. >> But Some samsung SoCs are using its bit[2] as other purpose. >> So it needs to revert the value with "ctrl &= ~SDHCI_CTRL_HISPD". >> Because it's possible to set to 1. >> >> SDHCI_QUIRK_NO_HISPD_BIT means that it's used other purpose. > > So, what are you suggesting? could you please elaborate? if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || (host->quirks & SDHCI_QUIRK_NO_BROKEN_HISPD_MODE)) { ctrl &= ~SDHCI_CTRL_HISPD; no_hispd_bit = true; } Just adding "ctrl &= ~SDHCI_CTRL_HISPD;" into its condition. Then it's helpful to me. Best Regards, Jaehoon Chung > > Jagan. >
All: > Subject: Re: [PATCH v4] mmc: sdhci: Fix HISPD bit handling > > On 6/22/20 6:26 PM, Jagan Teki wrote: > > On Mon, Jun 22, 2020 at 2:54 PM Jaehoon Chung > <jh80.chung at samsung.com> wrote: > >> > >> Hi Peng, > >> > >> On 6/22/20 10:49 AM, Peng Fan wrote: > >>> Jaehoon, > >>> > >>>> Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling > >>> > >>> Are you fine with this v4? > >> > >> Thanks for sharing it. i didn't see patch v4. > >> > >>> > >>> Thanks, > >>> Peng. > >>> > >>>> > >>>> SDHCI HISPD bits need to be configured based on desired mmc timings > >>>> mode and some HISPD quirks. > >>>> > >>>> So, handle the HISPD bit based on the mmc computed selected > >>>> mode(timing > >>>> parameter) rather than fixed mmc card clock frequency. > >>>> > >>>> Linux handle the HISPD similar like this in below commit but no > >>>> SDHCI_QUIRK_BROKEN_HISPD_MODE, > >>>> > >>>> commit <501639bf2173> ("mmc: sdhci: fix > SDHCI_QUIRK_NO_HISPD_BIT > >>>> handling") > >>>> > >>>> This eventually fixed the mmc write issue observed in > >>>> rk3399 sdhci controller. > >>>> > >>>> Bug log for refernece, > >>>> => gpt write mmc 0 $partitions > >>>> Writing GPT: mmc write failed > >>>> ** Can't write to device 0 ** > >>>> ** Can't write to device 0 ** > >>>> error! > >>>> > >>>> Cc: Kever Yang <kever.yang at rock-chips.com> > >>>> Cc: Peng Fan <peng.fan at nxp.com> > >>>> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # > >>>> roc-rk3399-pc > >>>> Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> > >>>> --- > >>>> Changes for v4: > >>>> - update commit message > >>>> - simplify the logic. > >>>> > >>>> drivers/mmc/sdhci.c | 23 +++++++++++++++++------ > >>>> 1 file changed, 17 insertions(+), 6 deletions(-) > >>>> > >>>> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index > >>>> 92cc8434af..6cb702111b 100644 > >>>> --- a/drivers/mmc/sdhci.c > >>>> +++ b/drivers/mmc/sdhci.c > >>>> @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) > #endif > >>>> u32 ctrl; > >>>> struct sdhci_host *host = mmc->priv; > >>>> + bool no_hispd_bit = false; > >>>> > >>>> if (host->ops && host->ops->set_control_reg) > >>>> host->ops->set_control_reg(host); @@ -594,14 > +595,24 > >>>> @@ static int sdhci_set_ios(struct mmc *mmc) > >>>> ctrl &= ~SDHCI_CTRL_4BITBUS; > >>>> } > >>>> > >>>> - if (mmc->clock > 26000000) > >>>> - ctrl |= SDHCI_CTRL_HISPD; > >>>> - else > >>>> - ctrl &= ~SDHCI_CTRL_HISPD; > >>>> - > >>>> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || > >>>> (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) > >>>> - ctrl &= ~SDHCI_CTRL_HISPD; > >>>> + no_hispd_bit = true; > >> > >> No. ctrl variable is set to bit[2] of HOST_CONTROL register. > >> But Some samsung SoCs are using its bit[2] as other purpose. > >> So it needs to revert the value with "ctrl &= ~SDHCI_CTRL_HISPD". > >> Because it's possible to set to 1. > >> > >> SDHCI_QUIRK_NO_HISPD_BIT means that it's used other purpose. > > > > So, what are you suggesting? could you please elaborate? > > > if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || > (host->quirks & SDHCI_QUIRK_NO_BROKEN_HISPD_MODE)) { > ctrl &= ~SDHCI_CTRL_HISPD; > no_hispd_bit = true; > } > > Just adding "ctrl &= ~SDHCI_CTRL_HISPD;" into its condition. I pushed this patch with the upper code added to my CI with a update in commit log: https://travis-ci.org/github/MrVan/u-boot/builds/701486010 Please let me know if any objections. Thanks, Peng. > Then it's helpful to me. > > Best Regards, > Jaehoon Chung > > > > > Jagan. > >
On 6/24/20 10:36 AM, Peng Fan wrote: > All: > >> Subject: Re: [PATCH v4] mmc: sdhci: Fix HISPD bit handling >> >> On 6/22/20 6:26 PM, Jagan Teki wrote: >>> On Mon, Jun 22, 2020 at 2:54 PM Jaehoon Chung >> <jh80.chung at samsung.com> wrote: >>>> >>>> Hi Peng, >>>> >>>> On 6/22/20 10:49 AM, Peng Fan wrote: >>>>> Jaehoon, >>>>> >>>>>> Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling >>>>> >>>>> Are you fine with this v4? >>>> >>>> Thanks for sharing it. i didn't see patch v4. >>>> >>>>> >>>>> Thanks, >>>>> Peng. >>>>> >>>>>> >>>>>> SDHCI HISPD bits need to be configured based on desired mmc timings >>>>>> mode and some HISPD quirks. >>>>>> >>>>>> So, handle the HISPD bit based on the mmc computed selected >>>>>> mode(timing >>>>>> parameter) rather than fixed mmc card clock frequency. >>>>>> >>>>>> Linux handle the HISPD similar like this in below commit but no >>>>>> SDHCI_QUIRK_BROKEN_HISPD_MODE, >>>>>> >>>>>> commit <501639bf2173> ("mmc: sdhci: fix >> SDHCI_QUIRK_NO_HISPD_BIT >>>>>> handling") >>>>>> >>>>>> This eventually fixed the mmc write issue observed in >>>>>> rk3399 sdhci controller. >>>>>> >>>>>> Bug log for refernece, >>>>>> => gpt write mmc 0 $partitions >>>>>> Writing GPT: mmc write failed >>>>>> ** Can't write to device 0 ** >>>>>> ** Can't write to device 0 ** >>>>>> error! >>>>>> >>>>>> Cc: Kever Yang <kever.yang at rock-chips.com> >>>>>> Cc: Peng Fan <peng.fan at nxp.com> >>>>>> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # >>>>>> roc-rk3399-pc >>>>>> Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> >>>>>> --- >>>>>> Changes for v4: >>>>>> - update commit message >>>>>> - simplify the logic. >>>>>> >>>>>> drivers/mmc/sdhci.c | 23 +++++++++++++++++------ >>>>>> 1 file changed, 17 insertions(+), 6 deletions(-) >>>>>> >>>>>> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index >>>>>> 92cc8434af..6cb702111b 100644 >>>>>> --- a/drivers/mmc/sdhci.c >>>>>> +++ b/drivers/mmc/sdhci.c >>>>>> @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) >> #endif >>>>>> u32 ctrl; >>>>>> struct sdhci_host *host = mmc->priv; >>>>>> + bool no_hispd_bit = false; >>>>>> >>>>>> if (host->ops && host->ops->set_control_reg) >>>>>> host->ops->set_control_reg(host); @@ -594,14 >> +595,24 >>>>>> @@ static int sdhci_set_ios(struct mmc *mmc) >>>>>> ctrl &= ~SDHCI_CTRL_4BITBUS; >>>>>> } >>>>>> >>>>>> - if (mmc->clock > 26000000) >>>>>> - ctrl |= SDHCI_CTRL_HISPD; >>>>>> - else >>>>>> - ctrl &= ~SDHCI_CTRL_HISPD; >>>>>> - >>>>>> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || >>>>>> (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) >>>>>> - ctrl &= ~SDHCI_CTRL_HISPD; >>>>>> + no_hispd_bit = true; >>>> >>>> No. ctrl variable is set to bit[2] of HOST_CONTROL register. >>>> But Some samsung SoCs are using its bit[2] as other purpose. >>>> So it needs to revert the value with "ctrl &= ~SDHCI_CTRL_HISPD". >>>> Because it's possible to set to 1. >>>> >>>> SDHCI_QUIRK_NO_HISPD_BIT means that it's used other purpose. >>> >>> So, what are you suggesting? could you please elaborate? >> >> >> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || >> (host->quirks & SDHCI_QUIRK_NO_BROKEN_HISPD_MODE)) { >> ctrl &= ~SDHCI_CTRL_HISPD; >> no_hispd_bit = true; >> } >> >> Just adding "ctrl &= ~SDHCI_CTRL_HISPD;" into its condition. > > I pushed this patch with the upper code added to my CI > with a update in commit log: > https://protect2.fireeye.com/url?k=a0ad72d4-fd674763-a0acf99b-0cc47a3003e8-b7ff57d36725a01c&q=1&u=https%3A%2F%2Ftravis-ci.org%2Fgithub%2FMrVan%2Fu-boot%2Fbuilds%2F701486010 > > Please let me know if any objections. Thanks! > > Thanks, > Peng. > >> Then it's helpful to me. >> >> Best Regards, >> Jaehoon Chung >> >>> >>> Jagan. >>> >
On Wed, Jun 24, 2020 at 7:06 AM Peng Fan <peng.fan at nxp.com> wrote: > > All: > > > Subject: Re: [PATCH v4] mmc: sdhci: Fix HISPD bit handling > > > > On 6/22/20 6:26 PM, Jagan Teki wrote: > > > On Mon, Jun 22, 2020 at 2:54 PM Jaehoon Chung > > <jh80.chung at samsung.com> wrote: > > >> > > >> Hi Peng, > > >> > > >> On 6/22/20 10:49 AM, Peng Fan wrote: > > >>> Jaehoon, > > >>> > > >>>> Subject: [PATCH v4] mmc: sdhci: Fix HISPD bit handling > > >>> > > >>> Are you fine with this v4? > > >> > > >> Thanks for sharing it. i didn't see patch v4. > > >> > > >>> > > >>> Thanks, > > >>> Peng. > > >>> > > >>>> > > >>>> SDHCI HISPD bits need to be configured based on desired mmc timings > > >>>> mode and some HISPD quirks. > > >>>> > > >>>> So, handle the HISPD bit based on the mmc computed selected > > >>>> mode(timing > > >>>> parameter) rather than fixed mmc card clock frequency. > > >>>> > > >>>> Linux handle the HISPD similar like this in below commit but no > > >>>> SDHCI_QUIRK_BROKEN_HISPD_MODE, > > >>>> > > >>>> commit <501639bf2173> ("mmc: sdhci: fix > > SDHCI_QUIRK_NO_HISPD_BIT > > >>>> handling") > > >>>> > > >>>> This eventually fixed the mmc write issue observed in > > >>>> rk3399 sdhci controller. > > >>>> > > >>>> Bug log for refernece, > > >>>> => gpt write mmc 0 $partitions > > >>>> Writing GPT: mmc write failed > > >>>> ** Can't write to device 0 ** > > >>>> ** Can't write to device 0 ** > > >>>> error! > > >>>> > > >>>> Cc: Kever Yang <kever.yang at rock-chips.com> > > >>>> Cc: Peng Fan <peng.fan at nxp.com> > > >>>> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # > > >>>> roc-rk3399-pc > > >>>> Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> > > >>>> --- > > >>>> Changes for v4: > > >>>> - update commit message > > >>>> - simplify the logic. > > >>>> > > >>>> drivers/mmc/sdhci.c | 23 +++++++++++++++++------ > > >>>> 1 file changed, 17 insertions(+), 6 deletions(-) > > >>>> > > >>>> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index > > >>>> 92cc8434af..6cb702111b 100644 > > >>>> --- a/drivers/mmc/sdhci.c > > >>>> +++ b/drivers/mmc/sdhci.c > > >>>> @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) > > #endif > > >>>> u32 ctrl; > > >>>> struct sdhci_host *host = mmc->priv; > > >>>> + bool no_hispd_bit = false; > > >>>> > > >>>> if (host->ops && host->ops->set_control_reg) > > >>>> host->ops->set_control_reg(host); @@ -594,14 > > +595,24 > > >>>> @@ static int sdhci_set_ios(struct mmc *mmc) > > >>>> ctrl &= ~SDHCI_CTRL_4BITBUS; > > >>>> } > > >>>> > > >>>> - if (mmc->clock > 26000000) > > >>>> - ctrl |= SDHCI_CTRL_HISPD; > > >>>> - else > > >>>> - ctrl &= ~SDHCI_CTRL_HISPD; > > >>>> - > > >>>> if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || > > >>>> (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) > > >>>> - ctrl &= ~SDHCI_CTRL_HISPD; > > >>>> + no_hispd_bit = true; > > >> > > >> No. ctrl variable is set to bit[2] of HOST_CONTROL register. > > >> But Some samsung SoCs are using its bit[2] as other purpose. > > >> So it needs to revert the value with "ctrl &= ~SDHCI_CTRL_HISPD". > > >> Because it's possible to set to 1. > > >> > > >> SDHCI_QUIRK_NO_HISPD_BIT means that it's used other purpose. > > > > > > So, what are you suggesting? could you please elaborate? > > > > > > if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || > > (host->quirks & SDHCI_QUIRK_NO_BROKEN_HISPD_MODE)) { > > ctrl &= ~SDHCI_CTRL_HISPD; > > no_hispd_bit = true; > > } > > > > Just adding "ctrl &= ~SDHCI_CTRL_HISPD;" into its condition. > > I pushed this patch with the upper code added to my CI > with a update in commit log: > https://travis-ci.org/github/MrVan/u-boot/builds/701486010 > > Please let me know if any objections. Thanks for taking care of this. Jagan.
diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c index 92cc8434af..6cb702111b 100644 --- a/drivers/mmc/sdhci.c +++ b/drivers/mmc/sdhci.c @@ -567,6 +567,7 @@ static int sdhci_set_ios(struct mmc *mmc) #endif u32 ctrl; struct sdhci_host *host = mmc->priv; + bool no_hispd_bit = false; if (host->ops && host->ops->set_control_reg) host->ops->set_control_reg(host); @@ -594,14 +595,24 @@ static int sdhci_set_ios(struct mmc *mmc) ctrl &= ~SDHCI_CTRL_4BITBUS; } - if (mmc->clock > 26000000) - ctrl |= SDHCI_CTRL_HISPD; - else - ctrl &= ~SDHCI_CTRL_HISPD; - if ((host->quirks & SDHCI_QUIRK_NO_HISPD_BIT) || (host->quirks & SDHCI_QUIRK_BROKEN_HISPD_MODE)) - ctrl &= ~SDHCI_CTRL_HISPD; + no_hispd_bit = true; + + if (!no_hispd_bit) { + if (mmc->selected_mode == MMC_HS || + mmc->selected_mode == SD_HS || + mmc->selected_mode == MMC_DDR_52 || + mmc->selected_mode == MMC_HS_200 || + mmc->selected_mode == MMC_HS_400 || + mmc->selected_mode == UHS_SDR25 || + mmc->selected_mode == UHS_SDR50 || + mmc->selected_mode == UHS_SDR104 || + mmc->selected_mode == UHS_DDR50) + ctrl |= SDHCI_CTRL_HISPD; + else + ctrl &= ~SDHCI_CTRL_HISPD; + } sdhci_writeb(host, ctrl, SDHCI_HOST_CONTROL);
SDHCI HISPD bits need to be configured based on desired mmc timings mode and some HISPD quirks. So, handle the HISPD bit based on the mmc computed selected mode(timing parameter) rather than fixed mmc card clock frequency. Linux handle the HISPD similar like this in below commit but no SDHCI_QUIRK_BROKEN_HISPD_MODE, commit <501639bf2173> ("mmc: sdhci: fix SDHCI_QUIRK_NO_HISPD_BIT handling") This eventually fixed the mmc write issue observed in rk3399 sdhci controller. Bug log for refernece, => gpt write mmc 0 $partitions Writing GPT: mmc write failed ** Can't write to device 0 ** ** Can't write to device 0 ** error! Cc: Kever Yang <kever.yang at rock-chips.com> Cc: Peng Fan <peng.fan at nxp.com> Tested-by: Suniel Mahesh <sunil at amarulasolutions.com> # roc-rk3399-pc Signed-off-by: Jagan Teki <jagan at amarulasolutions.com> --- Changes for v4: - update commit message - simplify the logic. drivers/mmc/sdhci.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)