diff mbox series

[v2,2/3] HID: logitech-hidpp: Retry commands when device is busy

Message ID 20230206221256.129198-2-hadess@hadess.net
State New
Headers
Series [v2,1/3] HID: logitech-hidpp: Add more debug statements |

Commit Message

Bastien Nocera Feb. 6, 2023, 10:12 p.m. UTC 
  Handle the busy error coming from the device or receiver. The
documentation says a busy error can be returned when:
"
Device (or receiver) cannot answer immediately to this request
for any reason i.e:
- already processing a request from the same or another SW
- pipe full
"

Signed-off-by: Bastien Nocera <hadess@hadess.net>
---

Same as v1

 drivers/hid/hid-logitech-hidpp.c | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Benjamin Tissoires Feb. 9, 2023, 2:50 p.m. UTC | #1 
On Feb 06 2023, Bastien Nocera wrote:
> Handle the busy error coming from the device or receiver. The
> documentation says a busy error can be returned when:
> "
> Device (or receiver) cannot answer immediately to this request
> for any reason i.e:
> - already processing a request from the same or another SW
> - pipe full
> "
> 
> Signed-off-by: Bastien Nocera <hadess@hadess.net>
> ---
> 
> Same as v1
> 
>  drivers/hid/hid-logitech-hidpp.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/hid/hid-logitech-hidpp.c b/drivers/hid/hid-logitech-hidpp.c
> index 1952d8d3b6b2..9e94026de437 100644
> --- a/drivers/hid/hid-logitech-hidpp.c
> +++ b/drivers/hid/hid-logitech-hidpp.c
> @@ -295,6 +295,7 @@ static int hidpp_send_message_sync(struct hidpp_device *hidpp,
>  	 */
>  	*response = *message;
>  
> +retry:
>  	ret = __hidpp_send_report(hidpp->hid_dev, message);
>  
>  	if (ret) {
> @@ -321,6 +322,10 @@ static int hidpp_send_message_sync(struct hidpp_device *hidpp,
>  			response->report_id == REPORT_ID_HIDPP_VERY_LONG) &&
>  			response->fap.feature_index == HIDPP20_ERROR) {
>  		ret = response->fap.params[1];
> +		if (ret == HIDPP20_ERROR_BUSY) {
> +			dbg_hid("%s:got busy hidpp 2.0 error %02X, retrying\n", __func__, ret);
> +			goto retry;

I must confess, I blocked a little bit there to decide whether or not
using goto here was OK.

But then I reliazed that there is no way to leave that function if the
device is buggy and constantly sends back ERROR_BUSY. So I am not very
found of the idea of having that got after all.

Would you mind respinning that patch with a bounded loop for the retries
instead of using a goto? I'd like the driver to give up after a few
retries if the device is not fair.

Cheers,
Benjamin

> +		}
>  		dbg_hid("%s:got hidpp 2.0 error %02X\n", __func__, ret);
>  		goto exit;
>  	}
> -- 
> 2.39.1
>
Bastien Nocera Feb. 9, 2023, 3:50 p.m. UTC | #2 
On Thu, 2023-02-09 at 15:50 +0100, Benjamin Tissoires wrote:
> On Feb 06 2023, Bastien Nocera wrote:
> > Handle the busy error coming from the device or receiver. The
> > documentation says a busy error can be returned when:
> > "
> > Device (or receiver) cannot answer immediately to this request
> > for any reason i.e:
> > - already processing a request from the same or another SW
> > - pipe full
> > "
> > 
> > Signed-off-by: Bastien Nocera <hadess@hadess.net>
> > ---
> > 
> > Same as v1
> > 
> >  drivers/hid/hid-logitech-hidpp.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/drivers/hid/hid-logitech-hidpp.c b/drivers/hid/hid-
> > logitech-hidpp.c
> > index 1952d8d3b6b2..9e94026de437 100644
> > --- a/drivers/hid/hid-logitech-hidpp.c
> > +++ b/drivers/hid/hid-logitech-hidpp.c
> > @@ -295,6 +295,7 @@ static int hidpp_send_message_sync(struct
> > hidpp_device *hidpp,
> >          */
> >         *response = *message;
> >  
> > +retry:
> >         ret = __hidpp_send_report(hidpp->hid_dev, message);
> >  
> >         if (ret) {
> > @@ -321,6 +322,10 @@ static int hidpp_send_message_sync(struct
> > hidpp_device *hidpp,
> >                         response->report_id ==
> > REPORT_ID_HIDPP_VERY_LONG) &&
> >                         response->fap.feature_index ==
> > HIDPP20_ERROR) {
> >                 ret = response->fap.params[1];
> > +               if (ret == HIDPP20_ERROR_BUSY) {
> > +                       dbg_hid("%s:got busy hidpp 2.0 error %02X,
> > retrying\n", __func__, ret);
> > +                       goto retry;
> 
> I must confess, I blocked a little bit there to decide whether or not
> using goto here was OK.
> 
> But then I reliazed that there is no way to leave that function if
> the
> device is buggy and constantly sends back ERROR_BUSY. So I am not
> very
> found of the idea of having that got after all.
> 
> Would you mind respinning that patch with a bounded loop for the
> retries
> instead of using a goto? I'd like the driver to give up after a few
> retries if the device is not fair.

Done in v3.
diff mbox series

Patch

diff --git a/drivers/hid/hid-logitech-hidpp.c b/drivers/hid/hid-logitech-hidpp.c
index 1952d8d3b6b2..9e94026de437 100644
--- a/drivers/hid/hid-logitech-hidpp.c
+++ b/drivers/hid/hid-logitech-hidpp.c
@@ -295,6 +295,7 @@  static int hidpp_send_message_sync(struct hidpp_device *hidpp,
 	 */
 	*response = *message;
 
+retry:
 	ret = __hidpp_send_report(hidpp->hid_dev, message);
 
 	if (ret) {
@@ -321,6 +322,10 @@  static int hidpp_send_message_sync(struct hidpp_device *hidpp,
 			response->report_id == REPORT_ID_HIDPP_VERY_LONG) &&
 			response->fap.feature_index == HIDPP20_ERROR) {
 		ret = response->fap.params[1];
+		if (ret == HIDPP20_ERROR_BUSY) {
+			dbg_hid("%s:got busy hidpp 2.0 error %02X, retrying\n", __func__, ret);
+			goto retry;
+		}
 		dbg_hid("%s:got hidpp 2.0 error %02X\n", __func__, ret);
 		goto exit;
 	}