Re: [PATCH] i2c-cadence: Do not let signals interrupt I2C transfers

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 03/11/2014 05:49 AM, Suneel Garapati wrote:
Hi Mike/Soren,



Met vriendelijke groet / kind regards,

Mike Looijmans

TOPIC Embedded Systems
Eindhovenseweg 32-C, NL-5683 KH Best
Postbus 440, NL-5680 AK Best
Telefoon: (+31) (0) 499 33 69 79
Telefax:  (+31) (0) 499 33 69 70
E-mail: mike.looijmans@xxxxxxxx
Website: www.topic.nl

Please consider the environment before printing this e-mail

-----Original Message-----
From: Sören Brinkmann [mailto:soren.brinkmann@xxxxxxxxxx]
Sent: Tuesday, March 11, 2014 03:30
To: Mike Looijmans; Michal Simek
Cc: git; wsa@xxxxxxxxxxxxx; linux-i2c@xxxxxxxxxxxxxxx; linux-
kernel@xxxxxxxxxxxxxxx
Subject: Re: [PATCH] i2c-cadence: Do not let signals interrupt I2C transfers

Hi Mike,

The cadence driver is not in mainline yet. I think for our vendor tree we can
pretty much take it this way.
Regarding getting this into mainline, I'll send another iteration of the change
set and include these changes.

On Mon, 2014-03-10 at 12:12PM +0100, Mike Looijmans wrote:
Pressing CTRL-C while communicating with an I2C device leads to
erratic behaviour. The cause is that the controller will interrupt the
I2C transfer in progress, and leave the client device in an undefined
state. Many drivers do not handle error return codes on I2C transfers.
The calling driver has no way of telling how much of the transfer has
actually completed, so it cannot reliably determine the device's state.

The best solution here is to not handle signals in the I2C bus driver
at all, but always complete a transaction before returning control.

See for a related patch and discussion on this topic:
http://lkml.org/lkml/2014/1/9/246

Can we get your Signed-off-by, please?

Sure, I forgot the -s, will add it in the next version.


---
  drivers/i2c/busses/i2c-cadence.c |   12 ++++--------
  1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/drivers/i2c/busses/i2c-cadence.c
b/drivers/i2c/busses/i2c-cadence.c
index 86713d6..32ce2ee 100644
--- a/drivers/i2c/busses/i2c-cadence.c
+++ b/drivers/i2c/busses/i2c-cadence.c
@@ -452,16 +452,12 @@ static int cdns_i2c_process_msg(struct cdns_i2c
*id, struct i2c_msg *msg,
                     cdns_i2c_msend(id);

             /* Wait for the signal of completion */
-           ret = wait_for_completion_interruptible_timeout(
-                                                   &id->xfer_done, HZ);
-           if (ret < 1) {
+           ret = wait_for_completion_timeout(&id->xfer_done, HZ);
+           if (ret == 0) {

To match the style used throughout the file this should just be
       if (!ret) {

Will do.

Instead of discarding Ctrl+C, Can we wait until the current msg transfer completes
[to avoid client undefined state or re-init the host to a known state]

I see da-vinci driver handling in a similar way.

My patch for the davinci should have completely ignored the signals too. I haven't come across a driver that actually calls the xfer function with more than one item anyway. If drivers check the result of the xfer at all, they usually don't have a clue how to handle partial transfers. Many drivers just ignore the return code completely and don't even report failure.

The gain is a stable I2C system. The loss here is something like <1ms more latency to interrupt a process in the case it was just transferring I2C data. From experience I can tell that the gain far outweighs that loss.

Mike.


--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux GPIO]     [Linux SPI]     [Linux Hardward Monitoring]     [LM Sensors]     [Linux USB Devel]     [Linux Media]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux