[PATCH] i2c-mpc: Wait for STOP to hit the bus

mpc_i2c_stop() only initiates STOP but does not wait for it to
hit the I2C bus. This is a problem when using I2C devices which
uses fairly long clock stretching just before STOP if you also
have an i2c-mux which may switch to another bus before STOP has
been processed.

Signed-off-by: Joakim Tjernlund <redacted>
---
 drivers/i2c/busses/i2c-mpc.c |   18 +++++++++++++++++-
 1 files changed, 17 insertions(+), 1 deletions(-)

diff --git drivers/i2c/busses/i2c-mpc.c drivers/i2c/busses/i2c-mpc.c
index 3d31879..c08f287 100644
--- drivers/i2c/busses/i2c-mpc.c
+++ drivers/i2c/busses/i2c-mpc.c

@@ -574,7 +574,23 @@ static int mpc_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
 			    mpc_write(i2c, pmsg->addr, pmsg->buf, pmsg->len, i);
 		}
 	}
-	mpc_i2c_stop(i2c);
+	mpc_i2c_stop(i2c); /* Initiate STOP */
+	orig_jiffies = jiffies;
+	/* Wait until STOP is seen, allow up to 1 s */
+	while (readb(i2c->base + MPC_I2C_SR) & CSR_MBB) {
+		if (time_after(jiffies, orig_jiffies + HZ)) {
+			u8 status = readb(i2c->base + MPC_I2C_SR);
+
+			dev_dbg(i2c->dev, "timeout\n");
+			if ((status & (CSR_MCF | CSR_MBB | CSR_RXAK)) != 0) {
+				writeb(status & ~CSR_MAL,
+				       i2c->base + MPC_I2C_SR);
+				mpc_i2c_fixup(i2c);
+			}
+			return -EIO;
+		}
+		cond_resched();
+	}
 	return (ret < 0) ? ret : num;
 }
 

-- 
1.7.8.6

On Thu, Aug 30, 2012 at 5:40 AM, Joakim Tjernlund
[off-list ref] wrote:

-       mpc_i2c_stop(i2c);
+       mpc_i2c_stop(i2c); /* Initiate STOP */
+       orig_jiffies =3D jiffies;
+       /* Wait until STOP is seen, allow up to 1 s */
+       while (readb(i2c->base + MPC_I2C_SR) & CSR_MBB) {
+               if (time_after(jiffies, orig_jiffies + HZ)) {
+                       u8 status =3D readb(i2c->base + MPC_I2C_SR);
+
+                       dev_dbg(i2c->dev, "timeout\n");
+                       if ((status & (CSR_MCF | CSR_MBB | CSR_RXAK)) !=

=3D 0) {

+                               writeb(status & ~CSR_MAL,
+                                      i2c->base + MPC_I2C_SR);
+                               mpc_i2c_fixup(i2c);
+                       }
+                       return -EIO;
+               }
+               cond_resched();
+       }

Shouldn't the while-loop be inside mpc_i2c_stop() itself?

--=20
Timur Tabi
Linux kernel developer at Freescale=

Tabi Timur-B04825 [off-list ref] wrote on 2012/09/02 04:48:01:

On Thu, Aug 30, 2012 at 5:40 AM, Joakim Tjernlund
[off-list ref] wrote:

quoted

-       mpc_i2c_stop(i2c);
+       mpc_i2c_stop(i2c); /* Initiate STOP */
+       orig_jiffies = jiffies;
+       /* Wait until STOP is seen, allow up to 1 s */
+       while (readb(i2c->base + MPC_I2C_SR) & CSR_MBB) {
+               if (time_after(jiffies, orig_jiffies + HZ)) {
+                       u8 status = readb(i2c->base + MPC_I2C_SR);
+
+                       dev_dbg(i2c->dev, "timeout\n");
+                       if ((status & (CSR_MCF | CSR_MBB | CSR_RXAK)) != 0) {
+                               writeb(status & ~CSR_MAL,
+                                      i2c->base + MPC_I2C_SR);
+                               mpc_i2c_fixup(i2c);
+                       }
+                       return -EIO;
+               }
+               cond_resched();
+       }

Shouldn't the while-loop be inside mpc_i2c_stop() itself?

Possibly but I choosed to do it this way as there is a similar loop in the beginning of mpc_xfer().
I figured it has better visibility if it is in the same function.

 Jocke

On Thu, Aug 30, 2012 at 12:40:04PM +0200, Joakim Tjernlund wrote:

mpc_i2c_stop() only initiates STOP but does not wait for it to
hit the I2C bus. This is a problem when using I2C devices which
uses fairly long clock stretching just before STOP if you also
have an i2c-mux which may switch to another bus before STOP has
been processed.

Signed-off-by: Joakim Tjernlund <redacted>

Patch didn't apply, because it was a p0-patch. Checkpatch would have
warned you.

Fixed that and applied to -next, thanks!

-- 
Pengutronix e.K.                           | Wolfram Sang                |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

Wolfram Sang [off-list ref] wrote on 2012/09/14 16:02:34:

On Thu, Aug 30, 2012 at 12:40:04PM +0200, Joakim Tjernlund wrote:

quoted

mpc_i2c_stop() only initiates STOP but does not wait for it to
hit the I2C bus. This is a problem when using I2C devices which
uses fairly long clock stretching just before STOP if you also
have an i2c-mux which may switch to another bus before STOP has
been processed.

Signed-off-by: Joakim Tjernlund <redacted>

Patch didn't apply, because it was a p0-patch. Checkpatch would have
warned you.

Bugger, it is my git config setting:
[diff]
	noprefix = true

I always forget to remove that one before generating patches, I will have to remove it.

Fixed that and applied to -next, thanks!

Thanks