diff mbox

[v2,1/2] mmc: dw_mmc: fix bug that cause 'Timeout sending command'

Message ID 1423466726-20833-2-git-send-email-addy.ke@rock-chips.com (mailing list archive)
State New, archived
Headers show

Commit Message

addy ke Feb. 9, 2015, 7:25 a.m. UTC
Because of some uncertain factors, such as worse card or worse hardware,
DAT[3:0](the data lines) may be pulled down by card, and mmc controller
will be in busy state. This should not happend when mmc controller
send command to update card clocks. If this happends, mci_send_cmd will
be failed and we will get 'Timeout sending command', and then system will
be blocked. To avoid this, we need reset mmc controller.

Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
---
 drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

Comments

Jaehoon Chung Feb. 9, 2015, 10:01 a.m. UTC | #1
Hi, Addy.

On 02/09/2015 04:25 PM, Addy Ke wrote:
> Because of some uncertain factors, such as worse card or worse hardware,
> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
> will be in busy state. This should not happend when mmc controller
> send command to update card clocks. If this happends, mci_send_cmd will
> be failed and we will get 'Timeout sending command', and then system will
> be blocked. To avoid this, we need reset mmc controller.
> 
> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
> ---
>  drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>  1 file changed, 28 insertions(+)
> 
> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
> index 4d2e3c2..b0b57e3 100644
> --- a/drivers/mmc/host/dw_mmc.c
> +++ b/drivers/mmc/host/dw_mmc.c
> @@ -100,6 +100,7 @@ struct idmac_desc {
>  };
>  #endif /* CONFIG_MMC_DW_IDMAC */
>  
> +static int dw_mci_card_busy(struct mmc_host *mmc);
>  static bool dw_mci_reset(struct dw_mci *host);
>  static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>  
> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>  		cmd, arg, cmd_status);
>  }
>  
> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
> +{
> +	struct dw_mci *host = slot->host;
> +	unsigned long timeout = jiffies + msecs_to_jiffies(500);
> +
> +	do {
> +		if (!dw_mci_card_busy(slot->mmc))
> +			return;
> +		cpu_relax();
> +	} while (time_before(jiffies, timeout));
> +
> +	dev_err(host->dev, "Data busy (status %#x)\n",
> +		mci_readl(slot->host, STATUS));
> +
> +	/*
> +	 * Data busy, this should not happend when mmc controller send command
> +	 * to update card clocks in non-volt-switch state. If it happends, we
> +	 * should reset controller to avoid getting "Timeout sending command".
> +	 */
> +	dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);

If reset is failed, then dw_mci_ctrl_reset should return "false".

	ret = dw_mci_ctrl_reset();

	WARN_ON(!ret || dw_mci_card_busy(slot->mmc));

Is it right?

In my experiment, if reset is failed or card is busy, eMMC can't work anymore..right?
I think this patch is reasonable to prevent blocking issue.

Best Regards,
Jaehoon Chung


> +
> +	/* Fail to reset controller or still data busy, WARN_ON! */
> +	WARN_ON(dw_mci_card_busy(slot->mmc));
> +}
> +
>  static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>  {
>  	struct dw_mci *host = slot->host;
> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>  	/* We must continue to set bit 28 in CMD until the change is complete */
>  	if (host->state == STATE_WAITING_CMD11_DONE)
>  		sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
> +	else
> +		dw_mci_wait_busy(slot);
>  
>  	if (!clock) {
>  		mci_writel(host, CLKENA, 0);
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Alim Akhtar Feb. 10, 2015, 3:22 p.m. UTC | #2
Hi Addy,

On Mon, Feb 9, 2015 at 12:55 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
> Because of some uncertain factors, such as worse card or worse hardware,
> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
> will be in busy state. This should not happend when mmc controller
> send command to update card clocks. If this happends, mci_send_cmd will
> be failed and we will get 'Timeout sending command', and then system will
> be blocked. To avoid this, we need reset mmc controller.
>
> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
> ---
>  drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>  1 file changed, 28 insertions(+)
>
> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
> index 4d2e3c2..b0b57e3 100644
> --- a/drivers/mmc/host/dw_mmc.c
> +++ b/drivers/mmc/host/dw_mmc.c
> @@ -100,6 +100,7 @@ struct idmac_desc {
>  };
>  #endif /* CONFIG_MMC_DW_IDMAC */
>
> +static int dw_mci_card_busy(struct mmc_host *mmc);
>  static bool dw_mci_reset(struct dw_mci *host);
>  static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>
> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>                 cmd, arg, cmd_status);
>  }
>
> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
> +{
> +       struct dw_mci *host = slot->host;
> +       unsigned long timeout = jiffies + msecs_to_jiffies(500);
> +
Why 500 msec?

> +       do {
> +               if (!dw_mci_card_busy(slot->mmc))
> +                       return;
> +               cpu_relax();
> +       } while (time_before(jiffies, timeout));
> +
> +       dev_err(host->dev, "Data busy (status %#x)\n",
> +               mci_readl(slot->host, STATUS));
> +
> +       /*
> +        * Data busy, this should not happend when mmc controller send command
> +        * to update card clocks in non-volt-switch state. If it happends, we
> +        * should reset controller to avoid getting "Timeout sending command".
> +        */
> +       dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
> +
Why you need to reset all blocks? may be CTRL_RESET is good enough here.

> +       /* Fail to reset controller or still data busy, WARN_ON! */
> +       WARN_ON(dw_mci_card_busy(slot->mmc));
> +}
> +
>  static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>  {
>         struct dw_mci *host = slot->host;
> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>         /* We must continue to set bit 28 in CMD until the change is complete */
>         if (host->state == STATE_WAITING_CMD11_DONE)
>                 sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
> +       else
> +               dw_mci_wait_busy(slot);
>
hmm...I would suggest you to call dw_mci_wait_busy() from inside
mci_send_cmd(), seems like dw_mmc hangs while sending update clock cmd
in multiple cases.see [1]

[1]: http://permalink.gmane.org/gmane.linux.kernel.mmc/31140

>         if (!clock) {
>                 mci_writel(host, CLKENA, 0);
> --
> 1.8.3.2
>
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
addy ke Feb. 11, 2015, 2:57 a.m. UTC | #3
On 2015/02/10 23:22, Alim Akhtar wrote:
> Hi Addy,
>
> On Mon, Feb 9, 2015 at 12:55 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
>> Because of some uncertain factors, such as worse card or worse hardware,
>> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
>> will be in busy state. This should not happend when mmc controller
>> send command to update card clocks. If this happends, mci_send_cmd will
>> be failed and we will get 'Timeout sending command', and then system will
>> be blocked. To avoid this, we need reset mmc controller.
>>
>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>> ---
>>   drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>>   1 file changed, 28 insertions(+)
>>
>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
>> index 4d2e3c2..b0b57e3 100644
>> --- a/drivers/mmc/host/dw_mmc.c
>> +++ b/drivers/mmc/host/dw_mmc.c
>> @@ -100,6 +100,7 @@ struct idmac_desc {
>>   };
>>   #endif /* CONFIG_MMC_DW_IDMAC */
>>
>> +static int dw_mci_card_busy(struct mmc_host *mmc);
>>   static bool dw_mci_reset(struct dw_mci *host);
>>   static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>>
>> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>>                  cmd, arg, cmd_status);
>>   }
>>
>> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
>> +{
>> +       struct dw_mci *host = slot->host;
>> +       unsigned long timeout = jiffies + msecs_to_jiffies(500);
>> +
> Why 500 msec?
This timeout value is the same as  mci_send_cmd:
static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
{
         struct dw_mci *host = slot->host;
         unsigned long timeout = jiffies + msecs_to_jiffies(500);
         ....
}

I have not clear that which is suitable.
Do you have any suggestion on it?
>
>> +       do {
>> +               if (!dw_mci_card_busy(slot->mmc))
>> +                       return;
>> +               cpu_relax();
>> +       } while (time_before(jiffies, timeout));
>> +
>> +       dev_err(host->dev, "Data busy (status %#x)\n",
>> +               mci_readl(slot->host, STATUS));
>> +
>> +       /*
>> +        * Data busy, this should not happend when mmc controller send command
>> +        * to update card clocks in non-volt-switch state. If it happends, we
>> +        * should reset controller to avoid getting "Timeout sending command".
>> +        */
>> +       dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
>> +
> Why you need to reset all blocks? may be CTRL_RESET is good enough here.
I have tested on rk3288, if only reset ctroller, data busy bit will not 
be cleaned,and we will still get

"Timeout sending command".

>
>> +       /* Fail to reset controller or still data busy, WARN_ON! */
>> +       WARN_ON(dw_mci_card_busy(slot->mmc));
>> +}
>> +
>>   static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>>   {
>>          struct dw_mci *host = slot->host;
>> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>>          /* We must continue to set bit 28 in CMD until the change is complete */
>>          if (host->state == STATE_WAITING_CMD11_DONE)
>>                  sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
>> +       else
>> +               dw_mci_wait_busy(slot);
>>
> hmm...I would suggest you to call dw_mci_wait_busy() from inside
> mci_send_cmd(), seems like dw_mmc hangs while sending update clock cmd
> in multiple cases.see [1]
>
> [1]: http://permalink.gmane.org/gmane.linux.kernel.mmc/31140
I think this patch is more reasonable.
So I will resend patches based on this patch.
thank you!
>
>>          if (!clock) {
>>                  mci_writel(host, CLKENA, 0);
>> --
>> 1.8.3.2
>>
>>
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
>


--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
addy ke Feb. 11, 2015, 3:07 a.m. UTC | #4
On 2015/02/09 18:01, Jaehoon Chung wrote:
> Hi, Addy.
>
> On 02/09/2015 04:25 PM, Addy Ke wrote:
>> Because of some uncertain factors, such as worse card or worse hardware,
>> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
>> will be in busy state. This should not happend when mmc controller
>> send command to update card clocks. If this happends, mci_send_cmd will
>> be failed and we will get 'Timeout sending command', and then system will
>> be blocked. To avoid this, we need reset mmc controller.
>>
>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>> ---
>>   drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>>   1 file changed, 28 insertions(+)
>>
>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
>> index 4d2e3c2..b0b57e3 100644
>> --- a/drivers/mmc/host/dw_mmc.c
>> +++ b/drivers/mmc/host/dw_mmc.c
>> @@ -100,6 +100,7 @@ struct idmac_desc {
>>   };
>>   #endif /* CONFIG_MMC_DW_IDMAC */
>>   
>> +static int dw_mci_card_busy(struct mmc_host *mmc);
>>   static bool dw_mci_reset(struct dw_mci *host);
>>   static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>>   
>> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>>   		cmd, arg, cmd_status);
>>   }
>>   
>> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
>> +{
>> +	struct dw_mci *host = slot->host;
>> +	unsigned long timeout = jiffies + msecs_to_jiffies(500);
>> +
>> +	do {
>> +		if (!dw_mci_card_busy(slot->mmc))
>> +			return;
>> +		cpu_relax();
>> +	} while (time_before(jiffies, timeout));
>> +
>> +	dev_err(host->dev, "Data busy (status %#x)\n",
>> +		mci_readl(slot->host, STATUS));
>> +
>> +	/*
>> +	 * Data busy, this should not happend when mmc controller send command
>> +	 * to update card clocks in non-volt-switch state. If it happends, we
>> +	 * should reset controller to avoid getting "Timeout sending command".
>> +	 */
>> +	dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
> If reset is failed, then dw_mci_ctrl_reset should return "false".
>
> 	ret = dw_mci_ctrl_reset();
>
> 	WARN_ON(!ret || dw_mci_card_busy(slot->mmc));
>
> Is it right?
you are right, and I will update it in my next version patch. thank you.
>
> In my experiment, if reset is failed or card is busy, eMMC can't work anymore..right?
> I think this patch is reasonable to prevent blocking issue.
>
> Best Regards,
> Jaehoon Chung
>
>
>> +
>> +	/* Fail to reset controller or still data busy, WARN_ON! */
>> +	WARN_ON(dw_mci_card_busy(slot->mmc));
>> +}
>> +
>>   static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>>   {
>>   	struct dw_mci *host = slot->host;
>> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>>   	/* We must continue to set bit 28 in CMD until the change is complete */
>>   	if (host->state == STATE_WAITING_CMD11_DONE)
>>   		sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
>> +	else
>> +		dw_mci_wait_busy(slot);
>>   
>>   	if (!clock) {
>>   		mci_writel(host, CLKENA, 0);
>>
>
>
>


--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Ulf Hansson Feb. 11, 2015, 3:25 a.m. UTC | #5
On 11 February 2015 at 03:57, Addy <addy.ke@rock-chips.com> wrote:
>
> On 2015/02/10 23:22, Alim Akhtar wrote:
>>
>> Hi Addy,
>>
>> On Mon, Feb 9, 2015 at 12:55 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
>>>
>>> Because of some uncertain factors, such as worse card or worse hardware,
>>> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
>>> will be in busy state. This should not happend when mmc controller
>>> send command to update card clocks. If this happends, mci_send_cmd will
>>> be failed and we will get 'Timeout sending command', and then system will
>>> be blocked. To avoid this, we need reset mmc controller.
>>>
>>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>>> ---
>>>   drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>>>   1 file changed, 28 insertions(+)
>>>
>>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
>>> index 4d2e3c2..b0b57e3 100644
>>> --- a/drivers/mmc/host/dw_mmc.c
>>> +++ b/drivers/mmc/host/dw_mmc.c
>>> @@ -100,6 +100,7 @@ struct idmac_desc {
>>>   };
>>>   #endif /* CONFIG_MMC_DW_IDMAC */
>>>
>>> +static int dw_mci_card_busy(struct mmc_host *mmc);
>>>   static bool dw_mci_reset(struct dw_mci *host);
>>>   static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>>>
>>> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot,
>>> u32 cmd, u32 arg)
>>>                  cmd, arg, cmd_status);
>>>   }
>>>
>>> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
>>> +{
>>> +       struct dw_mci *host = slot->host;
>>> +       unsigned long timeout = jiffies + msecs_to_jiffies(500);
>>> +
>>
>> Why 500 msec?
>
> This timeout value is the same as  mci_send_cmd:
> static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
> {
>         struct dw_mci *host = slot->host;
>         unsigned long timeout = jiffies + msecs_to_jiffies(500);
>         ....
> }
>
> I have not clear that which is suitable.
> Do you have any suggestion on it?
>>
>>
>>> +       do {
>>> +               if (!dw_mci_card_busy(slot->mmc))
>>> +                       return;
>>> +               cpu_relax();
>>> +       } while (time_before(jiffies, timeout));
>>> +
>>> +       dev_err(host->dev, "Data busy (status %#x)\n",
>>> +               mci_readl(slot->host, STATUS));
>>> +
>>> +       /*
>>> +        * Data busy, this should not happend when mmc controller send
>>> command
>>> +        * to update card clocks in non-volt-switch state. If it
>>> happends, we
>>> +        * should reset controller to avoid getting "Timeout sending
>>> command".
>>> +        */
>>> +       dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
>>> +
>>
>> Why you need to reset all blocks? may be CTRL_RESET is good enough here.
>
> I have tested on rk3288, if only reset ctroller, data busy bit will not be
> cleaned,and we will still get
>
> "Timeout sending command".
>
>>
>>> +       /* Fail to reset controller or still data busy, WARN_ON! */
>>> +       WARN_ON(dw_mci_card_busy(slot->mmc));
>>> +}
>>> +
>>>   static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool
>>> force_clkinit)
>>>   {
>>>          struct dw_mci *host = slot->host;
>>> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot
>>> *slot, bool force_clkinit)
>>>          /* We must continue to set bit 28 in CMD until the change is
>>> complete */
>>>          if (host->state == STATE_WAITING_CMD11_DONE)
>>>                  sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
>>> +       else
>>> +               dw_mci_wait_busy(slot);
>>>
>> hmm...I would suggest you to call dw_mci_wait_busy() from inside
>> mci_send_cmd(), seems like dw_mmc hangs while sending update clock cmd
>> in multiple cases.see [1]
>>
>> [1]: http://permalink.gmane.org/gmane.linux.kernel.mmc/31140
>
> I think this patch is more reasonable.
> So I will resend patches based on this patch.
> thank you!

Addy, Jaehoon,

It would simply things if I could get pull requests for patches of
dw_mmc. It's currently a bit hard to keep track and follow the
progress all the time.
Anyone volunteering in helping out here? :-)

Kind regards
Uffe
--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jaehoon Chung Feb. 11, 2015, 3:46 a.m. UTC | #6
Dear, Ulf.

On 02/11/2015 12:25 PM, Ulf Hansson wrote:
> On 11 February 2015 at 03:57, Addy <addy.ke@rock-chips.com> wrote:
>>
>> On 2015/02/10 23:22, Alim Akhtar wrote:
>>>
>>> Hi Addy,
>>>
>>> On Mon, Feb 9, 2015 at 12:55 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
>>>>
>>>> Because of some uncertain factors, such as worse card or worse hardware,
>>>> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
>>>> will be in busy state. This should not happend when mmc controller
>>>> send command to update card clocks. If this happends, mci_send_cmd will
>>>> be failed and we will get 'Timeout sending command', and then system will
>>>> be blocked. To avoid this, we need reset mmc controller.
>>>>
>>>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>>>> ---
>>>>   drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>>>>   1 file changed, 28 insertions(+)
>>>>
>>>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
>>>> index 4d2e3c2..b0b57e3 100644
>>>> --- a/drivers/mmc/host/dw_mmc.c
>>>> +++ b/drivers/mmc/host/dw_mmc.c
>>>> @@ -100,6 +100,7 @@ struct idmac_desc {
>>>>   };
>>>>   #endif /* CONFIG_MMC_DW_IDMAC */
>>>>
>>>> +static int dw_mci_card_busy(struct mmc_host *mmc);
>>>>   static bool dw_mci_reset(struct dw_mci *host);
>>>>   static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>>>>
>>>> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot,
>>>> u32 cmd, u32 arg)
>>>>                  cmd, arg, cmd_status);
>>>>   }
>>>>
>>>> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
>>>> +{
>>>> +       struct dw_mci *host = slot->host;
>>>> +       unsigned long timeout = jiffies + msecs_to_jiffies(500);
>>>> +
>>>
>>> Why 500 msec?
>>
>> This timeout value is the same as  mci_send_cmd:
>> static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>> {
>>         struct dw_mci *host = slot->host;
>>         unsigned long timeout = jiffies + msecs_to_jiffies(500);
>>         ....
>> }
>>
>> I have not clear that which is suitable.
>> Do you have any suggestion on it?
>>>
>>>
>>>> +       do {
>>>> +               if (!dw_mci_card_busy(slot->mmc))
>>>> +                       return;
>>>> +               cpu_relax();
>>>> +       } while (time_before(jiffies, timeout));
>>>> +
>>>> +       dev_err(host->dev, "Data busy (status %#x)\n",
>>>> +               mci_readl(slot->host, STATUS));
>>>> +
>>>> +       /*
>>>> +        * Data busy, this should not happend when mmc controller send
>>>> command
>>>> +        * to update card clocks in non-volt-switch state. If it
>>>> happends, we
>>>> +        * should reset controller to avoid getting "Timeout sending
>>>> command".
>>>> +        */
>>>> +       dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
>>>> +
>>>
>>> Why you need to reset all blocks? may be CTRL_RESET is good enough here.
>>
>> I have tested on rk3288, if only reset ctroller, data busy bit will not be
>> cleaned,and we will still get
>>
>> "Timeout sending command".
>>
>>>
>>>> +       /* Fail to reset controller or still data busy, WARN_ON! */
>>>> +       WARN_ON(dw_mci_card_busy(slot->mmc));
>>>> +}
>>>> +
>>>>   static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool
>>>> force_clkinit)
>>>>   {
>>>>          struct dw_mci *host = slot->host;
>>>> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot
>>>> *slot, bool force_clkinit)
>>>>          /* We must continue to set bit 28 in CMD until the change is
>>>> complete */
>>>>          if (host->state == STATE_WAITING_CMD11_DONE)
>>>>                  sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
>>>> +       else
>>>> +               dw_mci_wait_busy(slot);
>>>>
>>> hmm...I would suggest you to call dw_mci_wait_busy() from inside
>>> mci_send_cmd(), seems like dw_mmc hangs while sending update clock cmd
>>> in multiple cases.see [1]
>>>
>>> [1]: http://permalink.gmane.org/gmane.linux.kernel.mmc/31140
>>
>> I think this patch is more reasonable.
>> So I will resend patches based on this patch.
>> thank you!
> 
> Addy, Jaehoon,
> 
> It would simply things if I could get pull requests for patches of
> dw_mmc. It's currently a bit hard to keep track and follow the
> progress all the time.
> Anyone volunteering in helping out here? :-)

If you are ok, i will do it..

Best Regards,
Jaehoon Chung

> 
> Kind regards
> Uffe
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Ulf Hansson Feb. 11, 2015, 3:52 a.m. UTC | #7
On 11 February 2015 at 04:46, Jaehoon Chung <jh80.chung@samsung.com> wrote:
> Dear, Ulf.
>
> On 02/11/2015 12:25 PM, Ulf Hansson wrote:
>> On 11 February 2015 at 03:57, Addy <addy.ke@rock-chips.com> wrote:
>>>
>>> On 2015/02/10 23:22, Alim Akhtar wrote:
>>>>
>>>> Hi Addy,
>>>>
>>>> On Mon, Feb 9, 2015 at 12:55 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
>>>>>
>>>>> Because of some uncertain factors, such as worse card or worse hardware,
>>>>> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
>>>>> will be in busy state. This should not happend when mmc controller
>>>>> send command to update card clocks. If this happends, mci_send_cmd will
>>>>> be failed and we will get 'Timeout sending command', and then system will
>>>>> be blocked. To avoid this, we need reset mmc controller.
>>>>>
>>>>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>>>>> ---
>>>>>   drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>>>>>   1 file changed, 28 insertions(+)
>>>>>
>>>>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
>>>>> index 4d2e3c2..b0b57e3 100644
>>>>> --- a/drivers/mmc/host/dw_mmc.c
>>>>> +++ b/drivers/mmc/host/dw_mmc.c
>>>>> @@ -100,6 +100,7 @@ struct idmac_desc {
>>>>>   };
>>>>>   #endif /* CONFIG_MMC_DW_IDMAC */
>>>>>
>>>>> +static int dw_mci_card_busy(struct mmc_host *mmc);
>>>>>   static bool dw_mci_reset(struct dw_mci *host);
>>>>>   static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>>>>>
>>>>> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot,
>>>>> u32 cmd, u32 arg)
>>>>>                  cmd, arg, cmd_status);
>>>>>   }
>>>>>
>>>>> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
>>>>> +{
>>>>> +       struct dw_mci *host = slot->host;
>>>>> +       unsigned long timeout = jiffies + msecs_to_jiffies(500);
>>>>> +
>>>>
>>>> Why 500 msec?
>>>
>>> This timeout value is the same as  mci_send_cmd:
>>> static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>>> {
>>>         struct dw_mci *host = slot->host;
>>>         unsigned long timeout = jiffies + msecs_to_jiffies(500);
>>>         ....
>>> }
>>>
>>> I have not clear that which is suitable.
>>> Do you have any suggestion on it?
>>>>
>>>>
>>>>> +       do {
>>>>> +               if (!dw_mci_card_busy(slot->mmc))
>>>>> +                       return;
>>>>> +               cpu_relax();
>>>>> +       } while (time_before(jiffies, timeout));
>>>>> +
>>>>> +       dev_err(host->dev, "Data busy (status %#x)\n",
>>>>> +               mci_readl(slot->host, STATUS));
>>>>> +
>>>>> +       /*
>>>>> +        * Data busy, this should not happend when mmc controller send
>>>>> command
>>>>> +        * to update card clocks in non-volt-switch state. If it
>>>>> happends, we
>>>>> +        * should reset controller to avoid getting "Timeout sending
>>>>> command".
>>>>> +        */
>>>>> +       dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
>>>>> +
>>>>
>>>> Why you need to reset all blocks? may be CTRL_RESET is good enough here.
>>>
>>> I have tested on rk3288, if only reset ctroller, data busy bit will not be
>>> cleaned,and we will still get
>>>
>>> "Timeout sending command".
>>>
>>>>
>>>>> +       /* Fail to reset controller or still data busy, WARN_ON! */
>>>>> +       WARN_ON(dw_mci_card_busy(slot->mmc));
>>>>> +}
>>>>> +
>>>>>   static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool
>>>>> force_clkinit)
>>>>>   {
>>>>>          struct dw_mci *host = slot->host;
>>>>> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot
>>>>> *slot, bool force_clkinit)
>>>>>          /* We must continue to set bit 28 in CMD until the change is
>>>>> complete */
>>>>>          if (host->state == STATE_WAITING_CMD11_DONE)
>>>>>                  sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
>>>>> +       else
>>>>> +               dw_mci_wait_busy(slot);
>>>>>
>>>> hmm...I would suggest you to call dw_mci_wait_busy() from inside
>>>> mci_send_cmd(), seems like dw_mmc hangs while sending update clock cmd
>>>> in multiple cases.see [1]
>>>>
>>>> [1]: http://permalink.gmane.org/gmane.linux.kernel.mmc/31140
>>>
>>> I think this patch is more reasonable.
>>> So I will resend patches based on this patch.
>>> thank you!
>>
>> Addy, Jaehoon,
>>
>> It would simply things if I could get pull requests for patches of
>> dw_mmc. It's currently a bit hard to keep track and follow the
>> progress all the time.
>> Anyone volunteering in helping out here? :-)
>
> If you are ok, i will do it..

Great! Much appreciated!

Don't hesitate to send several PR during a release cycle.

Kind regards
Uffe
--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Andrzej Hajda Feb. 11, 2015, 11:58 a.m. UTC | #8
Hi Alim,

On 02/11/2015 03:57 AM, Addy wrote:
> 
> On 2015/02/10 23:22, Alim Akhtar wrote:
>> Hi Addy,
>>
>> On Mon, Feb 9, 2015 at 12:55 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
>>> Because of some uncertain factors, such as worse card or worse hardware,
>>> DAT[3:0](the data lines) may be pulled down by card, and mmc controller
>>> will be in busy state. This should not happend when mmc controller
>>> send command to update card clocks. If this happends, mci_send_cmd will
>>> be failed and we will get 'Timeout sending command', and then system will
>>> be blocked. To avoid this, we need reset mmc controller.
>>>
>>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>>> ---
>>>   drivers/mmc/host/dw_mmc.c | 28 ++++++++++++++++++++++++++++
>>>   1 file changed, 28 insertions(+)
>>>
>>> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
>>> index 4d2e3c2..b0b57e3 100644
>>> --- a/drivers/mmc/host/dw_mmc.c
>>> +++ b/drivers/mmc/host/dw_mmc.c
>>> @@ -100,6 +100,7 @@ struct idmac_desc {
>>>   };
>>>   #endif /* CONFIG_MMC_DW_IDMAC */
>>>
>>> +static int dw_mci_card_busy(struct mmc_host *mmc);
>>>   static bool dw_mci_reset(struct dw_mci *host);
>>>   static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
>>>
>>> @@ -888,6 +889,31 @@ static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
>>>                  cmd, arg, cmd_status);
>>>   }
>>>
>>> +static void dw_mci_wait_busy(struct dw_mci_slot *slot)
>>> +{
>>> +       struct dw_mci *host = slot->host;
>>> +       unsigned long timeout = jiffies + msecs_to_jiffies(500);
>>> +
>> Why 500 msec?
> This timeout value is the same as  mci_send_cmd:
> static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
> {
>          struct dw_mci *host = slot->host;
>          unsigned long timeout = jiffies + msecs_to_jiffies(500);
>          ....
> }
> 
> I have not clear that which is suitable.
> Do you have any suggestion on it?
>>
>>> +       do {
>>> +               if (!dw_mci_card_busy(slot->mmc))
>>> +                       return;
>>> +               cpu_relax();
>>> +       } while (time_before(jiffies, timeout));
>>> +
>>> +       dev_err(host->dev, "Data busy (status %#x)\n",
>>> +               mci_readl(slot->host, STATUS));
>>> +
>>> +       /*
>>> +        * Data busy, this should not happend when mmc controller send command
>>> +        * to update card clocks in non-volt-switch state. If it happends, we
>>> +        * should reset controller to avoid getting "Timeout sending command".
>>> +        */
>>> +       dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
>>> +
>> Why you need to reset all blocks? may be CTRL_RESET is good enough here.
> I have tested on rk3288, if only reset ctroller, data busy bit will not 
> be cleaned,and we will still get
> 
> "Timeout sending command".
> 
>>
>>> +       /* Fail to reset controller or still data busy, WARN_ON! */
>>> +       WARN_ON(dw_mci_card_busy(slot->mmc));
>>> +}
>>> +
>>>   static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>>>   {
>>>          struct dw_mci *host = slot->host;
>>> @@ -899,6 +925,8 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
>>>          /* We must continue to set bit 28 in CMD until the change is complete */
>>>          if (host->state == STATE_WAITING_CMD11_DONE)
>>>                  sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
>>> +       else
>>> +               dw_mci_wait_busy(slot);
>>>
>> hmm...I would suggest you to call dw_mci_wait_busy() from inside
>> mci_send_cmd(), seems like dw_mmc hangs while sending update clock cmd
>> in multiple cases.see [1]
>>
>> [1]: http://permalink.gmane.org/gmane.linux.kernel.mmc/31140
> I think this patch is more reasonable.
> So I will resend patches based on this patch.
> thank you!

I have tested your patches instead [1] above and they do not solve my issue:
Board: odroid-xu3/exynos5422/dw_mmc_250a.
MMC card: absent, broken-cd quirk
SD card: present

System hangs during boot after few minutes kernel spits:
[  242.188098] INFO: task kworker/u16:1:50 blocked for more than 120
seconds.
[  242.193524]       Not tainted
3.19.0-next-20150210-00002-gf96831b-dirty #3834
[  242.200622] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
[  242.208422] kworker/u16:1   D c04766ac     0    50      2 0x00000000
[  242.214756] Workqueue: kmmcd mmc_rescan
[  242.218553] [<c04766ac>] (__schedule) from [<c0476a58>]
(schedule+0x34/0x98)
[  242.225591] [<c0476a58>] (schedule) from [<c047a4dc>]
(schedule_timeout+0x110/0x164)
[  242.233302] [<c047a4dc>] (schedule_timeout) from [<c04774f0>]
(wait_for_common+0xb8/0x14c)
[  242.241539] [<c04774f0>] (wait_for_common) from [<c0362138>]
(mmc_wait_for_req+0x68/0x17c)
[  242.249861] [<c0362138>] (mmc_wait_for_req) from [<c03622cc>]
(mmc_wait_for_cmd+0x80/0xa0)
[  242.258002] [<c03622cc>] (mmc_wait_for_cmd) from [<c0367e50>]
(mmc_go_idle+0x78/0xf8)
[  242.265796] [<c0367e50>] (mmc_go_idle) from [<c0363e2c>]
(mmc_rescan+0x280/0x314)
[  242.273253] [<c0363e2c>] (mmc_rescan) from [<c0034764>]
(process_one_work+0x120/0x324)
[  242.281135] [<c0034764>] (process_one_work) from [<c00349cc>]
(worker_thread+0x30/0x42c)
[  242.289194] [<c00349cc>] (worker_thread) from [<c003926c>]
(kthread+0xd8/0xf4)
[  242.296389] [<c003926c>] (kthread) from [<c000e7c0>]
(ret_from_fork+0x14/0x34)

Just for record, Exynos4412/dw_mmc_240a with the same configuration
(no MMC card, broken-cd) works OK without patches.


Regards
Andrzej

>>
>>>          if (!clock) {
>>>                  mci_writel(host, CLKENA, 0);
>>> --
>>> 1.8.3.2
>>>
>>>
>>>
>>> _______________________________________________
>>> linux-arm-kernel mailing list
>>> linux-arm-kernel@lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>>
>>
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-doc" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
index 4d2e3c2..b0b57e3 100644
--- a/drivers/mmc/host/dw_mmc.c
+++ b/drivers/mmc/host/dw_mmc.c
@@ -100,6 +100,7 @@  struct idmac_desc {
 };
 #endif /* CONFIG_MMC_DW_IDMAC */
 
+static int dw_mci_card_busy(struct mmc_host *mmc);
 static bool dw_mci_reset(struct dw_mci *host);
 static bool dw_mci_ctrl_reset(struct dw_mci *host, u32 reset);
 
@@ -888,6 +889,31 @@  static void mci_send_cmd(struct dw_mci_slot *slot, u32 cmd, u32 arg)
 		cmd, arg, cmd_status);
 }
 
+static void dw_mci_wait_busy(struct dw_mci_slot *slot)
+{
+	struct dw_mci *host = slot->host;
+	unsigned long timeout = jiffies + msecs_to_jiffies(500);
+
+	do {
+		if (!dw_mci_card_busy(slot->mmc))
+			return;
+		cpu_relax();
+	} while (time_before(jiffies, timeout));
+
+	dev_err(host->dev, "Data busy (status %#x)\n",
+		mci_readl(slot->host, STATUS));
+
+	/*
+	 * Data busy, this should not happend when mmc controller send command
+	 * to update card clocks in non-volt-switch state. If it happends, we
+	 * should reset controller to avoid getting "Timeout sending command".
+	 */
+	dw_mci_ctrl_reset(host, SDMMC_CTRL_ALL_RESET_FLAGS);
+
+	/* Fail to reset controller or still data busy, WARN_ON! */
+	WARN_ON(dw_mci_card_busy(slot->mmc));
+}
+
 static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
 {
 	struct dw_mci *host = slot->host;
@@ -899,6 +925,8 @@  static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
 	/* We must continue to set bit 28 in CMD until the change is complete */
 	if (host->state == STATE_WAITING_CMD11_DONE)
 		sdmmc_cmd_bits |= SDMMC_CMD_VOLT_SWITCH;
+	else
+		dw_mci_wait_busy(slot);
 
 	if (!clock) {
 		mci_writel(host, CLKENA, 0);