Date: Thu, 16 Jul 2009 22:22:10 +0200 From: Hans Petter Selasky <hselasky@c2i.net> To: Patrick Lamaiziere <patfbsd@davenulle.org> Cc: freebsd-current@freebsd.org Subject: Re: ulpt problem (USB_ERR_IOERROR) Message-ID: <200907162222.11736.hselasky@c2i.net> In-Reply-To: <20090716215528.271fc177@baby-jane.lamaiziere.net> References: <20090703172600.1971111e@baby-jane.lamaiziere.net> <200907130846.51980.hselasky@c2i.net> <20090716215528.271fc177@baby-jane.lamaiziere.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi,
On Thursday 16 July 2009 21:55:28 Patrick Lamaiziere wrote:
> Le Mon, 13 Jul 2009 08:46:50 +0200,
>
> Hans Petter Selasky <hselasky@c2i.net> a =E9crit :
> > The printing works besides from the status failing?
>
> Yes.
>
>
> Do you mean here? :
Yes.
> [ULPT_INTR_DT_RD] =3D { .type =3D UE_CONTROL,
> ...
> .timeout =3D 20000, /* 20 second */
>
> If yes, I tried but this does not change anything. I also tried to
> change the ticks of the callout_reset in usb_watchdog().
>
> But If I remove the transfer start in usb_watchdog() (so I never read
> the status of the printer), it works well.
Could you try to add an "if (sc->sc_fflags =3D=3D 0)" in front, like this:
/*=20
* Only read status while the device is not opened, due to
* possible hardware or firmware bug in some printers.
*/
if (sc->sc_fflags =3D=3D 0)
usbd_transfer_start(sc->sc_xfer[ULPT_INTR_DT_RD]);
> With cups and without the deframentation of the datas to write it seems
> to work too (the same configuration as the top of this thread). Are you
> sure this it required by my printer?
No. I've removed that patch.
> Not related but I noticed that usb_write() in usb_dev.c returns -1 if I
> hit Ctrl-C when doing a "cat file > /dev/unlpt0". Because
> usb_fifo_wait(f) returns -1 (should be EINTR?).
If cv_wait_sig() returns -1, then they do not behave like in the manual pag=
e:
man cv_wait_sig
You have to do a "kill" to kill it. What I've seen is the code hangs in the=
=20
flushing loop in the close call:
/* check if flushed already */
while (f->flag_flushing &&
(!f->flag_iserror)) {
/* wait until all data has been written */
f->flag_sleeping =3D 1;
err =3D cv_wait_sig(&f->cv_io, f->priv_mtx);
if (err) {
DPRINTF("signal received\n");
break;
}
}
I have no idea why. Seems like the signals are blocked, so that CTRL+C is n=
ot=20
passed another time! Please investigate, if you have some time.
New patch:
1) cvsup to 8-current.
2) Replace serial/ulpt.c with the one from USB P4
http://perforce.freebsd.org/chv.cgi?CH=3D166176
=2D-HPS
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200907162222.11736.hselasky>