[1235] in linux-scsi channel archive
Re: NCR 53c825 (53c8xx vs. 53c7,8xx)
daemon@ATHENA.MIT.EDU (Dirk Foersterling)
Wed Jan 8 01:28:48 1997
Date: Wed, 8 Jan 1997 05:37:05 +0100
From: dirk@informatik.uni-frankfurt.de (Dirk Foersterling)
To: groudier@club-internet.fr (Gerard Roudier)
Cc: linux-scsi@vger.rutgers.edu (Linux SCSI list)
In-Reply-To: <Pine.LNX.3.91.970107235558.288A-100000@localhost>; from Gerard Roudier on Jan 8, 1997 00:08:47 +0000
On Jan 7 1997, Gerard Roudier wrote:
>
> >
> > scsi_cmd 0x08 (write len=6, read len=39424)
>
[snip]
> opinion, the command above cannot succeed since the max buffer size
> used by the sg driver is currently 32768.
[snap]
> Edit linux/include/scsi/sg.h and increase the value of SG_BIG_BUF,
> for example:
>
> - #define SG_BIG_BUF 32768
> + #define SG_BIG_BUF (32768+8192)
>
Hmm... this is another thing that was missing. I have the following in sg.h:
#define SG_BIG_BUF 131072
since I use the program. It is stated in the README file. But more on
this file later in this text... <8^]
On Jan 8 1997, Gerard Roudier wrote:
>
> Hi Dirk!
>
> On Sat, 4 Jan 1997, Dirk Foersterling wrote:
>
> I've checked all the commands and the following seems to me questionnable:
>
> > scsi_cmd 0x11 (write len=6, read len=5)
>
> This command normally is a SPACE command for sequential devices and
> should not return any data, but your application seems to expect up to
> 5 data bytes.
This is also a proprietary command "CCD_DISTANCE". (NOW I know this :( )
>
> However, this command seems to not cause problems.
>
> Since the driver is not able to process commands without knowledge of
> data direction, I will reduce the number of commands for which direction
> is checked to only very common opcodes. (sure 0x11 will be removed).
>
> Thanks for your help for debugging.
This is due to my dumbness. Call me a jerk. I have to RTFM again and again.
That's the problem, if you ignore something that doesn't apply to you and
then you change hardware...
I could have shortened the debugging (or never started this, what would
have helped nobody) by re-reading the mscan README file:
- the ncr53c800 driver that I'm using somehow has problems
with the CCD_LD (CCD line-distance) command of the Mustek
scanner; this problem manifests itself by an entry in
/var/log/messages saying something like:
ncr53c810-0-<target 6, lun 0>: extraneous data discarded.
If you see this problem, too you might want to turn on
-DLINE_DISTANCE_WORKAROUND in the Makefile and take a
look at the relevant code in msci.c. The workaround probably
works for 600 dpi scanners only, but for other scanners, the
values should be analogous.
>
> PS: I received your reply today (Tuesday). Seems vger.rutgers.edu have
> had problems since Saturday.
I didn't care since I receive my mail only every 24 hours.
Anyway, thanks for your help. Next time I try to read all READMEs more
carfull to do a shorter report...
-dirk
--
D i r k F "o r s t e r l i n g
dirk@informatik.uni-frankfurt.de http://www.uni-frankfurt.de/~dirk
----------
"Das ist aber gerade e hoch 3 z. Das sieht doch jeder!" - R.K.
