Hi Oliver,
On Thursday 03 July 2008, Oliver Neukum wrote:
> Am Donnerstag 03 Juli 2008 00:16:52 schrieb Laurent Pinchart:
> > On Wednesday 02 July 2008, Oliver Neukum wrote:
> > > Am Dienstag 01 Juli 2008 14:52:33 schrieb Laurent Pinchart:
> > > > Hi Oliver,
> > > >
> > > > On Tuesday 01 July 2008, Oliver Neukum wrote:
> > > > > int uvc_status_resume(struct uvc_device *dev)
> > > > > {
> > > > > if (dev->int_urb == NULL)
> > > > > return 0;
> > > > >
> > > > > return usb_submit_urb(dev->int_urb, GFP_KERNEL);
> > > > > }
> > > > >
> > > > > Hi,
> > > > >
> > > > > if you are waking up from a system suspend, your swap device may
> > > > > still be asleep. Better use GFP_NOIO.
> > > >
> > > > Thanks for the report.
> > > >
> > > > I assume the same is true for the video resume path where I allocate
> > > > and submit the video data URBs. The amount of memory the driver
> > > > allocates there is higher, is there anything I should pay attention
> > > > to to lower the memory pressure introduced by GFP_NOIO ?
> > >
> > > How much memory is that?
> >
> > For isochronous transfers (bulk transfers use smaller buffers), the
> > driver allocates five video buffers of at 240kB each.
>
> That is over 1MB in the worst case. Can't you keep the buffers allocated?
Very good point, thanks.
Best regards,
Laurent Pinchart
_______________________________________________
Linux-uvc-devel mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/linux-uvc-devel
