* Re: [Linux-usb-users] [2.6.10-rc2] ehci_hcd causes oops after some use of usb harddisk
[not found] <Pine.LNX.4.44L0.0411231557100.6467-100000@ida.rowland.org>
@ 2004-12-13 19:55 ` Jan De Luyck
0 siblings, 0 replies; only message in thread
From: Jan De Luyck @ 2004-12-13 19:55 UTC (permalink / raw)
To: Alan Stern, linux-kernel; +Cc: USB development list, USB users list
On Tuesday 23 November 2004 22:02, Alan Stern wrote:
> On Tue, 23 Nov 2004, Jan De Luyck wrote:
> > On Tuesday 23 November 2004 20:39, Alan Stern wrote:
> > > Try the patch below and see if it helps.
> > >
> > > Alan Stern
> > >
> > >
> > > ===== drivers/usb/storage/transport.c 1.152 vs edited =====
> > > --- 1.152/drivers/usb/storage/transport.c 2004-11-14 19:41:07 -05:00
> > > +++ edited/drivers/usb/storage/transport.c 2004-11-23 14:37:40 -05:00
> > > @@ -987,7 +987,7 @@
> > > /* Genesys Logic interface chips need a 100us delay between the
> > > * command phase and the data phase */
> > > if (us->pusb_dev->descriptor.idVendor == USB_VENDOR_ID_GENESYS)
> > > - udelay(100);
> > > + udelay(200);
> > >
> > > if (transfer_length) {
> > > unsigned int pipe = srb->sc_data_direction == DMA_FROM_DEVICE ?
> >
> > Indeed, this solved my problems.
> >
> > Thanks a lot.
>
> Just out of curiosity, could you try using several different values in
> that udelay() statement? Maybe 200 is larger than necessary. If
> something like 110 would work just as well, then it would help improve the
> I/O speed.
Alan,
Sorry for the late answer, I've been rather busy of late. Today I had the time
to recompile and test the drive.
I'm currently using the drive with a udelay of 110, and it works without any
problems. The transfer speed is indeed higher than with the 200msec delay ;)
Hope this helps, maybe this can now be integrated into the main kernel so
people no longer have a problem with it :)
If you need more information, feel free to ask.
Jan
--
I learned to play guitar just to get the girls, and anyone who says they
didn't is just lyin'!
-- Willie Nelson
^ permalink raw reply [flat|nested] only message in thread