Patchwork [U-Boot] mmc: omap: handle controller errors properly

login
register
mail settings
Submitter Grazvydas Ignotas
Date March 19, 2012, 10:12 p.m.
Message ID <1332195126-24719-1-git-send-email-notasas@gmail.com>
Download mbox | patch
Permalink /patch/147661/
State Accepted
Commit 25c719e2c3247ef0bab28ee562f978e2ce9211e3
Delegated to: Andy Fleming
Headers show

Comments

Grazvydas Ignotas - March 19, 2012, 10:12 p.m.
According to OMAP3 TRM, when the controller reports certain errors,
driver must perform a software reset. This is done by setting a bit
in SYSCTL and waiting it to clear:
- SRC on command timeout (CTO)
- SRD on data errors (DTO, DCRC and DEB)

This fixes a problem seen on OMAP3 pandora board with some cards
that won't work with a message printed multiple times:
  timedout waiting on cmd inhibit to clear

Code loosely based on Linux omap_hsmmc driver.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
---
 drivers/mmc/omap_hsmmc.c |   36 ++++++++++++++++++++++++++++++++++--
 1 files changed, 34 insertions(+), 2 deletions(-)
Tom Rini - March 19, 2012, 11:21 p.m.
On Tue, Mar 20, 2012 at 12:12:06AM +0200, Grazvydas Ignotas wrote:
> According to OMAP3 TRM, when the controller reports certain errors,
> driver must perform a software reset. This is done by setting a bit
> in SYSCTL and waiting it to clear:
> - SRC on command timeout (CTO)
> - SRD on data errors (DTO, DCRC and DEB)
> 
> This fixes a problem seen on OMAP3 pandora board with some cards
> that won't work with a message printed multiple times:
>   timedout waiting on cmd inhibit to clear
> 
> Code loosely based on Linux omap_hsmmc driver.
> 
> Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>

On a BeagleBoard xM (rev A), tested various class 2 and 4 uSD cards I
have (none of which have ever failed, however).  On a BeagleBoard rev C5
and am335x EVM and an omap4_panda (ES), tested with various class 2, 4
and 6 SD cards (or uSD in adapter).

Tested-by: Tom Rini <trini@ti.com>

And I've sent a private message to folks that have had failing cards
that previous changes have fixed to make sure this doesn't break them.
Pali Rohár - March 21, 2012, 7:11 p.m.
On Monday 19 March 2012 15:14:39 Tom Rini wrote:
> On 03/19/2012 03:12 PM, Grazvydas Ignotas wrote:
> > According to OMAP3 TRM, when the controller reports certain
> > errors, driver must perform a software reset. This is done by
> > setting a bit in SYSCTL and waiting it to clear:
> > - SRC on command timeout (CTO)
> > - SRD on data errors (DTO, DCRC and DEB)
> > 
> > This fixes a problem seen on OMAP3 pandora board with some cards
> > 
> > that won't work with a message printed multiple times:
> >    timedout waiting on cmd inhibit to clear
> > 
> > Code loosely based on Linux omap_hsmmc driver.
> > 
> > Signed-off-by: Grazvydas Ignotas<notasas@gmail.com>
> 
> Most excellent!  Pali, can you try this on N900 and see if that
> helps out the problem you were having as well?
> 
> Andy, I'll give this a spin through the boards I've got here but
> I'd like this to go, via your tree, into u-boot/master assuming
> all goes well.

Hi!

I tried this patch and it 
really fixed problem on Nokia N900.
Pali Rohár - March 21, 2012, 7:57 p.m.
On Wednesday 21 March 2012 20:11:08 Pali Rohár wrote:
> On Monday 19 March 2012 15:14:39 Tom Rini wrote:
> > On 03/19/2012 03:12 PM, Grazvydas Ignotas wrote:
> > > According to OMAP3 TRM, when the controller reports certain
> > > errors, driver must perform a software reset. This is done by
> > > setting a bit in SYSCTL and waiting it to clear:
> > > - SRC on command timeout (CTO)
> > > - SRD on data errors (DTO, DCRC and DEB)
> > > 
> > > This fixes a problem seen on OMAP3 pandora board with some
> > > cards
> > > 
> > > that won't work with a message printed multiple times:
> > >    timedout waiting on cmd inhibit to clear
> > > 
> > > Code loosely based on Linux omap_hsmmc driver.
> > > 
> > > Signed-off-by: Grazvydas Ignotas<notasas@gmail.com>
> > 
> > Most excellent!  Pali, can you try this on N900 and see if that
> > helps out the problem you were having as well?
> > 
> > Andy, I'll give this a spin through the boards I've got here but
> > I'd like this to go, via your tree, into u-boot/master assuming
> > all goes well.
> 
> Hi!
> 
> I tried this patch and it
> really fixed problem on Nokia N900.

Tested-by: Pali Rohár <pali.rohar@gmail.com>

Patch

diff --git a/drivers/mmc/omap_hsmmc.c b/drivers/mmc/omap_hsmmc.c
index 4b81d0d..06cbebf 100644
--- a/drivers/mmc/omap_hsmmc.c
+++ b/drivers/mmc/omap_hsmmc.c
@@ -33,6 +33,10 @@ 
 #include <asm/arch/mmc_host_def.h>
 #include <asm/arch/sys_proto.h>
 
+/* common definitions for all OMAPs */
+#define SYSCTL_SRC	(1 << 25)
+#define SYSCTL_SRD	(1 << 26)
+
 /* If we fail after 1 second wait, something is really bad */
 #define MAX_RETRY_MS	1000
 
@@ -195,6 +199,27 @@  static int mmc_init_setup(struct mmc *mmc)
 	return 0;
 }
 
+/*
+ * MMC controller internal finite state machine reset
+ *
+ * Used to reset command or data internal state machines, using respectively
+ * SRC or SRD bit of SYSCTL register
+ */
+static void mmc_reset_controller_fsm(struct hsmmc *mmc_base, u32 bit)
+{
+	ulong start;
+
+	mmc_reg_out(&mmc_base->sysctl, bit, bit);
+
+	start = get_timer(0);
+	while ((readl(&mmc_base->sysctl) & bit) != 0) {
+		if (get_timer(0) - start > MAX_RETRY_MS) {
+			printf("%s: timedout waiting for sysctl %x to clear\n",
+				__func__, bit);
+			return;
+		}
+	}
+}
 
 static int mmc_send_cmd(struct mmc *mmc, struct mmc_cmd *cmd,
 			struct mmc_data *data)
@@ -284,9 +309,10 @@  static int mmc_send_cmd(struct mmc *mmc, struct mmc_cmd *cmd,
 		}
 	} while (!mmc_stat);
 
-	if ((mmc_stat & IE_CTO) != 0)
+	if ((mmc_stat & IE_CTO) != 0) {
+		mmc_reset_controller_fsm(mmc_base, SYSCTL_SRC);
 		return TIMEOUT;
-	else if ((mmc_stat & ERRI_MASK) != 0)
+	} else if ((mmc_stat & ERRI_MASK) != 0)
 		return -1;
 
 	if (mmc_stat & CC_MASK) {
@@ -337,6 +363,9 @@  static int mmc_read_data(struct hsmmc *mmc_base, char *buf, unsigned int size)
 			}
 		} while (mmc_stat == 0);
 
+		if ((mmc_stat & (IE_DTO | IE_DCRC | IE_DEB)) != 0)
+			mmc_reset_controller_fsm(mmc_base, SYSCTL_SRD);
+
 		if ((mmc_stat & ERRI_MASK) != 0)
 			return 1;
 
@@ -389,6 +418,9 @@  static int mmc_write_data(struct hsmmc *mmc_base, const char *buf,
 			}
 		} while (mmc_stat == 0);
 
+		if ((mmc_stat & (IE_DTO | IE_DCRC | IE_DEB)) != 0)
+			mmc_reset_controller_fsm(mmc_base, SYSCTL_SRD);
+
 		if ((mmc_stat & ERRI_MASK) != 0)
 			return 1;