During test campaign, and especially after several unbind/bind sequences,
it has been seen that the SD-card on SDMMC1 thread could freeze.
The freeze always appear on a CMD23 following a CMD19.
Checking SDMMC internal registers shows that the tuning command (CMD19)
has failed.
The freeze is then due to the delay block involved in the tuning sequence.
To correct this, clear the delay block register DLYB_CR register after
the tuning commands.
Signed-off-by: Christophe Kerello <[email protected]>
Signed-off-by: Yann Gautier <[email protected]>
---
drivers/mmc/host/mmci_stm32_sdmmc.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
index fdaa11f92fe6..a75d3dd34d18 100644
--- a/drivers/mmc/host/mmci_stm32_sdmmc.c
+++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
@@ -441,6 +441,8 @@ static int sdmmc_dlyb_phase_tuning(struct mmci_host *host, u32 opcode)
return -EINVAL;
}
+ writel_relaxed(0, dlyb->base + DLYB_CR);
+
phase = end_of_len - max_len / 2;
sdmmc_dlyb_set_cfgr(dlyb, dlyb->unit, phase, false);
--
2.17.1
On 12/15/21 15:17, Yann Gautier wrote:
> During test campaign, and especially after several unbind/bind sequences,
> it has been seen that the SD-card on SDMMC1 thread could freeze.
> The freeze always appear on a CMD23 following a CMD19.
> Checking SDMMC internal registers shows that the tuning command (CMD19)
> has failed.
> The freeze is then due to the delay block involved in the tuning sequence.
> To correct this, clear the delay block register DLYB_CR register after
> the tuning commands.
>
> Signed-off-by: Christophe Kerello <[email protected]>
> Signed-off-by: Yann Gautier <[email protected]>
> ---
> drivers/mmc/host/mmci_stm32_sdmmc.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
> index fdaa11f92fe6..a75d3dd34d18 100644
> --- a/drivers/mmc/host/mmci_stm32_sdmmc.c
> +++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
> @@ -441,6 +441,8 @@ static int sdmmc_dlyb_phase_tuning(struct mmci_host *host, u32 opcode)
> return -EINVAL;
> }
>
> + writel_relaxed(0, dlyb->base + DLYB_CR);
> +
> phase = end_of_len - max_len / 2;
> sdmmc_dlyb_set_cfgr(dlyb, dlyb->unit, phase, false);
Shouldn't this have a Fixes: tag and be CC stable ?
This seems like a bugfix, no ?
On 12/16/21 10:13 AM, Marek Vasut wrote:
> On 12/15/21 15:17, Yann Gautier wrote:
>> During test campaign, and especially after several unbind/bind sequences,
>> it has been seen that the SD-card on SDMMC1 thread could freeze.
>> The freeze always appear on a CMD23 following a CMD19.
>> Checking SDMMC internal registers shows that the tuning command (CMD19)
>> has failed.
>> The freeze is then due to the delay block involved in the tuning
>> sequence.
>> To correct this, clear the delay block register DLYB_CR register after
>> the tuning commands.
>>
>> Signed-off-by: Christophe Kerello <[email protected]>
>> Signed-off-by: Yann Gautier <[email protected]>
>> ---
>> drivers/mmc/host/mmci_stm32_sdmmc.c | 2 ++
>> 1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c
>> b/drivers/mmc/host/mmci_stm32_sdmmc.c
>> index fdaa11f92fe6..a75d3dd34d18 100644
>> --- a/drivers/mmc/host/mmci_stm32_sdmmc.c
>> +++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
>> @@ -441,6 +441,8 @@ static int sdmmc_dlyb_phase_tuning(struct
>> mmci_host *host, u32 opcode)
>> return -EINVAL;
>> }
>> + writel_relaxed(0, dlyb->base + DLYB_CR);
>> +
>> phase = end_of_len - max_len / 2;
>> sdmmc_dlyb_set_cfgr(dlyb, dlyb->unit, phase, false);
>
> Shouldn't this have a Fixes: tag and be CC stable ?
> This seems like a bugfix, no ?
Hi Marek,
You're right, I should have added:
Fixes: 1103f807a3b9 ("mmc: mmci_sdmmc: Add execute tuning with delay block")
Ulf, could you manage that, or should I push a new version?
Best regards,
Yann
On Thu, 16 Dec 2021 at 10:49, Yann Gautier <[email protected]> wrote:
>
> On 12/16/21 10:13 AM, Marek Vasut wrote:
> > On 12/15/21 15:17, Yann Gautier wrote:
> >> During test campaign, and especially after several unbind/bind sequences,
> >> it has been seen that the SD-card on SDMMC1 thread could freeze.
> >> The freeze always appear on a CMD23 following a CMD19.
> >> Checking SDMMC internal registers shows that the tuning command (CMD19)
> >> has failed.
> >> The freeze is then due to the delay block involved in the tuning
> >> sequence.
> >> To correct this, clear the delay block register DLYB_CR register after
> >> the tuning commands.
> >>
> >> Signed-off-by: Christophe Kerello <[email protected]>
> >> Signed-off-by: Yann Gautier <[email protected]>
> >> ---
> >> drivers/mmc/host/mmci_stm32_sdmmc.c | 2 ++
> >> 1 file changed, 2 insertions(+)
> >>
> >> diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c
> >> b/drivers/mmc/host/mmci_stm32_sdmmc.c
> >> index fdaa11f92fe6..a75d3dd34d18 100644
> >> --- a/drivers/mmc/host/mmci_stm32_sdmmc.c
> >> +++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
> >> @@ -441,6 +441,8 @@ static int sdmmc_dlyb_phase_tuning(struct
> >> mmci_host *host, u32 opcode)
> >> return -EINVAL;
> >> }
> >> + writel_relaxed(0, dlyb->base + DLYB_CR);
> >> +
> >> phase = end_of_len - max_len / 2;
> >> sdmmc_dlyb_set_cfgr(dlyb, dlyb->unit, phase, false);
> >
> > Shouldn't this have a Fixes: tag and be CC stable ?
> > This seems like a bugfix, no ?
>
> Hi Marek,
>
> You're right, I should have added:
> Fixes: 1103f807a3b9 ("mmc: mmci_sdmmc: Add execute tuning with delay block")
>
> Ulf, could you manage that, or should I push a new version?
Patch applied for fixes and by adding a stable tag too, thanks!
Kind regards
Uffe