re : 3.2-rc2 regression due to commit "USB: EHCI: fix HUB TT scheduling issue with iso transfer" 811c926c

re : 3.2-rc2 regression due to commit "USB: EHCI: fix HUB TT scheduling issue with iso transfer" 811c926c

[matthieu castet]

Hi,

could you try the following patch ?

It seems there is a flow in the error condition.


Matthieu


diff --git a/drivers/usb/host/ehci-sched.c b/drivers/usb/host/ehci-sched.c
index 56a3203..464b451 100644
--- a/drivers/usb/host/ehci-sched.c
+++ b/drivers/usb/host/ehci-sched.c
@@ -1475,6 +1475,7 @@ iso_stream_schedule (
 	 * jump until after the queue is primed.
 	 */
 	else {
+		int done = 0;
 		start = SCHEDULE_SLOP + (now & ~0x07);
 
 		/* NOTE:  assumes URB_ISO_ASAP, to limit complexity/bugs */
@@ -1491,19 +1492,23 @@ iso_stream_schedule (
 			/* check schedule: enough space? */
 			if (stream->highspeed) {
 				if (itd_slot_ok(ehci, mod, start,
-						stream->usecs, period))
+						stream->usecs, period)) {
+					done = 1;
 					break;
+				}
 			} else {
 				if ((start % 8) >= 6)
 					continue;
 				if (sitd_slot_ok(ehci, mod, stream,
-						start, sched, period))
+						start, sched, period)) {
+					done = 1;
 					break;
+				}
 			}
 		} while (start > next);
 
 		/* no room in the schedule */
-		if (start == next) {
+		if (!done) {
 			ehci_dbg(ehci, "iso resched full %p (now %d max %d)\n",
 				urb, now, now + mod);
 			status = -ENOSPC;

Re: re : 3.2-rc2 regression due to commit "USB: EHCI: fix HUB TT scheduling issue with iso transfer" 811c926c

[Sander Eikelenboom]

Hi Matthieu,

I can confirm this patch fixes my problem, after testing haven't seen any problems yet.

Thx !

--
Sander



Saturday, November 26, 2011, 11:13:04 PM, you wrote:

> Hi,

> could you try the following patch ?

> It seems there is a flow in the error condition.


> Matthieu


> diff --git a/drivers/usb/host/ehci-sched.c b/drivers/usb/host/ehci-sched.c
> index 56a3203..464b451 100644
> --- a/drivers/usb/host/ehci-sched.c
> +++ b/drivers/usb/host/ehci-sched.c
> @@ -1475,6 +1475,7 @@ iso_stream_schedule (
>          * jump until after the queue is primed.
>          */
>         else {
> +               int done = 0;
>                 start = SCHEDULE_SLOP + (now & ~0x07);
>  
>                 /* NOTE:  assumes URB_ISO_ASAP, to limit complexity/bugs */
> @@ -1491,19 +1492,23 @@ iso_stream_schedule (
>                         /* check schedule: enough space? */
>                         if (stream->highspeed) {
>                                 if (itd_slot_ok(ehci, mod, start,
> -                                               stream->usecs, period))
> +                                               stream->usecs, period)) {
> +                                       done = 1;
>                                         break;
> +                               }
>                         } else {
>                                 if ((start % 8) >= 6)
>                                         continue;
>                                 if (sitd_slot_ok(ehci, mod, stream,
> -                                               start, sched, period))
> +                                               start, sched, period)) {
> +                                       done = 1;
>                                         break;
> +                               }
>                         }
>                 } while (start > next);
>  
>                 /* no room in the schedule */
> -               if (start == next) {
> +               if (!done) {
>                         ehci_dbg(ehci, "iso resched full %p (now %d max %d)\n",
>                                 urb, now, now + mod);
>                         status = -ENOSPC;



-- 
Best regards,
 Sander                            mailto:linux@eikelenboom.it

Re: re : 3.2-rc2 regression due to commit "USB: EHCI: fix HUB TT scheduling issue with iso transfer" 811c926c

[Greg KH]

On Sun, Nov 27, 2011 at 09:08:13PM +0100, Sander Eikelenboom wrote:
> Hi Matthieu,
> 
> I can confirm this patch fixes my problem, after testing haven't seen any problems yet.

Wonderful, thanks for testing.

Matthieu, care to make this up into a "real" patch so that I can apply
it and get it into 3.2?

thanks,

greg k-h