mmc: set timeout for SDHCI host before sending busy cmds

Hi all,
From the previous discussion, I do not think we have got a clear conclusion about using maximum timeout value. At least we know from Jae hoon Chung using 0xE for every case is not a good. So I want to suggest only use 0xE for busy command. I personally preferred below implementation, which is similar with a RFC patch submitted by Jae hoon Chung, but only without adding a new quirk.

I think sdhci_calc_timeout should be left for data transfer since at least we can get a warning if 0xE is not enough for host to use. And if the host controller and the card have no bugs, then the calculated timeout should be safe. Left the old implementation unchanged is also compatible with all existed host controllers and cards.

But for busy command, we are not clear about how long is safe enough for waiting and there is also no function to do the calculation for them. So preferred just using 0xE. Below the patch and comment:

Set the timeout control register for SDHCI host when send some commands which need busy signal. Use the maximum timeout value 0xE will be safe.

Signed-off-by Chuanxiao Dong <[email protected]>

---
 drivers/mmc/host/sdhci.c |    9 ++++++++-
 1 files changed, 8 insertions(+), 1 deletions(-)

drivers/mmc/host/sdhci.c +8 -1

index 99c372e..8306323 100644
@@ -659,8 +659,15 @@ static void sdhci_prepare_data(struct sdhci_host *host, struct mmc_data *data)
 
 	WARN_ON(host->data);
 
-	if (data == NULL)
+	if (data == NULL) {
+		/*
+		 * set the timeout to be maximum value for commands those with
+		 * busy signal
+		 */
+		if (host->cmd->flags & MMC_RSP_BUSY)
+			sdhci_writeb(host, 0xE, SDHCI_TIMEOUT_CONTROL);
 		return;
+	}
 
 	/* Sanity checks */
 	BUG_ON(data->blksz * data->blocks > 524288);

Quoted 7 lines · Chuanxiao Dong

Hi all,
From the previous discussion, I do not think we have got a clear conclusion about using maximum timeout value. At least we know from Jae hoon Chung using 0xE for every case is not a good. So I want to suggest only use 0xE for busy command. I personally preferred below implementation, which is similar with a RFC patch submitted by Jae hoon Chung, but only without adding a new quirk.

thanks for remind.
Yes, i tested without quirks, i think that is not problem.
(Just sent RFC patch with quirks, because i want to ask how think about adding quirks or not).

Quoted 41 lines

I think sdhci_calc_timeout should be left for data transfer since at least we can get a warning if 0xE is not enough for host to use. And if the host controller and the card have no bugs, then the calculated timeout should be safe. Left the old implementation unchanged is also compatible with all existed host controllers and cards.

But for busy command, we are not clear about how long is safe enough for waiting and there is also no function to do the calculation for them. So preferred just using 0xE. Below the patch and comment:

Set the timeout control register for SDHCI host when send some commands which need busy signal. Use the maximum timeout value 0xE will be safe.

Signed-off-by Chuanxiao Dong <[email protected]>

---
drivers/mmc/host/sdhci.c | 9 ++++++++-

1 files changed, 8 insertions(+), 1 deletions(-)

drivers/mmc/host/sdhci.c +8 -1

index 99c372e..8306323 100644
@@ -659,8 +659,15 @@ static void sdhci_prepare_data(struct sdhci_host *host, struct mmc_data *data)
 
 	WARN_ON(host->data);
 
-	if (data == NULL)
+	if (data == NULL) {
+		/*
+		 * set the timeout to be maximum value for commands those with
+		 * busy signal
+		 */
+		if (host->cmd->flags & MMC_RSP_BUSY)
+			sdhci_writeb(host, 0xE, SDHCI_TIMEOUT_CONTROL);
 		return;
+	}
 
 	/* Sanity checks */
 	BUG_ON(data->blksz * data->blocks > 524288);

Quoted 11 lines · Jaehoon Chung

Quoted 7 lines · Chuanxiao Dong

Hi all,
From the previous discussion, I do not think we have got a clear conclusion about using maximum timeout value. At least we know from Jae hoon Chung using 0xE for every case is not a good. So I want to suggest only use 0xE for busy command. I personally preferred below implementation, which is similar with a RFC patch submitted by Jae hoon Chung, but only without adding a new quirk.

thanks for remind.
Yes, i tested without quirks, i think that is not problem.
(Just sent RFC patch with quirks, because i want to ask how think about adding quirks or not).

Sorry I am confused.

Setting 0x0E all the time does not solve the problem and has side effects ? What are the side effects ?

Using BUSY patch for 0x0e (below) works ?

Quoted 37 lines

Quoted 37 lines

I think sdhci_calc_timeout should be left for data transfer since at least we can get a warning if 0xE is not enough for host to use. And if the host controller and the card have no bugs, then the calculated timeout should be safe. Left the old implementation unchanged is also compatible with all existed host controllers and cards.

But for busy command, we are not clear about how long is safe enough for waiting and there is also no function to do the calculation for them. So preferred just using 0xE. Below the patch and comment:

Set the timeout control register for SDHCI host when send some commands which need busy signal. Use the maximum timeout value 0xE will be safe.

Signed-off-by Chuanxiao Dong <[email protected]>

---
drivers/mmc/host/sdhci.c | 9 ++++++++-

1 files changed, 8 insertions(+), 1 deletions(-)

drivers/mmc/host/sdhci.c +8 -1

index 99c372e..8306323 100644
@@ -659,8 +659,15 @@ static void sdhci_prepare_data(struct sdhci_host *host, struct mmc_data *data)
 
 	WARN_ON(host->data);
 
-	if (data == NULL)
+	if (data == NULL) {
+		/*
+		 * set the timeout to be maximum value for commands those with
+		 * busy signal
+		 */
+		if (host->cmd->flags & MMC_RSP_BUSY)
+			sdhci_writeb(host, 0xE, SDHCI_TIMEOUT_CONTROL);
 		return;
+	}
 

Quoted 2 lines

drivers/mmc/host/sdhci.c static void sdhci_prepare_data(struct sdhci_host *host, struct mmc_data *data) · line 673

Quoted 2 lines

drivers/mmc/host/sdhci.c +0 -0

 	/* Sanity checks */
 	BUG_ON(data->blksz * data->blocks > 524288);

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