On Mon, 19 Oct 2009, Sarah Sharp wrote:
> @@ -1354,6 +1354,7 @@ iso_stream_schedule (
> period <<= 3;
>
> now = ehci_readl(ehci, &ehci->regs->frame_index) % mod;
> + next = now + ehci->i_thresh;
The same math can accomodate both full-speed and high-speed devices.
The difference can be confined to the definition of "next":
/* For HS devices, allow scheduling within the isochronous
* scheduling threshold. For FS/LS devices, don't.
* (Work around for ICH9 bug.)
*/
next = (stream->highspeed ? now : now + ehci->i_thresh);
I avoided using the unexplained abbreviation "IST" because it's likely
to confuse readers. It feels a little funny to talk about FS/LS
devices here, since LS devices can't do isochronous transfers, but I
guess it's okay.
> @@ -1365,16 +1366,29 @@ iso_stream_schedule (
> */
> if (likely (!list_empty (&stream->td_list))) {
> start = stream->next_uframe;
> - if (start < now)
> - start += mod;
>
> /* Fell behind (by up to twice the slop amount)? */
> - if (start >= max - 2 * SCHEDULE_SLOP)
> - start += period * DIV_ROUND_UP(
> - max - start, period) - mod;
> + /* For HS devices, allow scheduling within the IST.
> + * For FS/LS devices, don't. (Work around for ICH9 bug.)
> + */
> + if (stream->highspeed) {
> + if (start < now)
> + start += mod;
> +
> + if (start >= max - 2 * SCHEDULE_SLOP)
> + start += period * DIV_ROUND_UP(
> + max - start, period) - mod;
Then the code above isn't needed.
> + } else {
> + if (((start - next) & (mod - 1)) >=
> + mod - 2 * SCHEDULE_SLOP)
> + start += period * DIV_ROUND_UP(
> + (next - start) & (mod - 1),
> + period);
> + }
>
> /* Tried to schedule too far into the future? */
> - if (unlikely((start + sched->span) >= max)) {
> + if (unlikely(((start - now) & (mod - 1)) + sched->span
> + >= max - 2 * SCHEDULE_SLOP)) {
> status = -EFBIG;
> goto fail;
> }
The rest is okay.
Alan Stern
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
