[sane-devel] avision: Out of memory

[Rene Rebe]

Hi.

On: Tue, 04 Feb 2003 21:52:59 +0100,
    abel deuring <adeuring at gmx.net> wrote:
> Henning Meier-Geinitz wrote:
> > Hi,
> > 
> > On Mon, Feb 03, 2003 at 11:29:04AM +0100, Patrick Begou wrote:
> > 
> >>$ sane-find-scanner
> >>found SCSI scanner "HP ScanJet 5370C 6.00" at /dev/scanner
> >>found SCSI scanner "HP ScanJet 5370C 6.00" at /dev/sg2
> >>found SCSI scanner "HP ScanJet 5370C 6.00" at /dev/sgc
> [...]
> >>I did this also. All seems to work fine  until get_calib_format wich
> >>returns Out of Memory (see attached file).
> >>I have 256 MBytes RAM and 128 Mbytes of (unused) swap but I dont't think
> >>the problem is here. Something that the driver could not understand with
> >>my software configuration ?
> > 
> > 
> >>[avision] get_calib_data: type 62, size 167076, line_size: 167076
> >>[avision] get_calib_data: read data failed (Out of memory)
> > 
> > 
> > The backend tries to get 167076 bytes of data at once. Tht doesn't
> > seem to work with the hp5300 driver (or your kernel). I think the
> > default buffer size is 128k. 
> 
> Right. As I understand it, the 5370C uses some sort of USB/SCSI 
> translation. sanei_scsi_req_enter2 (more precisely, the SG driver of the 
> Linux kernel) can return "out of memory" errors. If a Sane frontend is 
> run with SANE_DEBUG_SANEI_SCSI=255, a debug message like 
> "sanei_scsi.issue: SG_BIG_BUF inconsistency?" should be printed. (which 
> we should update, BTW. SG_BIG_BUF does not caused problems since quite 
> some time.)
> 
> Anyway, the backend really should not try to send more data than 
> suggested by sanei_scsi_max_request_size.

I have no choise, since there are firmware bugs, that do not allow to
read the calibration data in several tiny chunks for (some Avision)
USB devices ... - And this ever worked for me here ...

Btw. It would ne nice to CC the backend maintainer, since I currently
only have the time to read over sane-* one a week or so ...

> Abel

- René

--  
René Rebe - Europe/Germany/Berlin
e-mail:   rene.rebe at gmx.net, rene at rocklinux.org
web:      www.rocklinux.org, drocklinux.dyndns.org/rene/

Anyone sending unwanted advertising e-mail to this address will be
charged $25 for network traffic and computing time. By extracting my
address from this message or its header, you agree to these terms.