Hi Peter Kirchgessner,
Thank you for your reply. It helped motivate me to keep persisting. I tried all
your advice and unfortunately it did not help. You may be still right about
buffer overruns. My guess is that the scanner wouldn't pause for some reason
when the buffer filled up, leading to the I/O errors.
But the remarkable fact is that the scanner is now working. I tried different
kernel combinations until one worked.
And the even more mind-boggling fact is that a newly compiled SMP kernel works
as well without CPU lockups appearing in the message log!!!
The success must be the result of some combination of these factors:
(a) Experimental FIFO and DMA kernel support enabled for the parallel port
connected to the scanner.
(b) chmod 666 /dev/sg0 (had done this earlier as well).
(c) judicious use of make mrproper in kernel recompiling instead of just make
clean (kernel compiling bombed once with just make clean).
[(d) pixie dust.
(e) gamma radiation changing a few bits on my hard disk.]
BTW here's the new kernel messages:
kernel: SCSI subsystem driver Revision: 1.00
kernel: Winbond Super-IO detection, now testing ports 3F0,370,250,4E,2E ...
kernel: Winbond chip at EFER=0x3f0 key=0x87 devid=52 devrev=f4 oldid=ff
kernel: Winbond chip type 83977EF / SMSC 97w35x
kernel: Winbond LPT Config: cr_30=01 60,61=0378 70=07 74=03, f0=03
kernel: Winbond LPT Config: active=yes, io=0x0378 irq=7, dma=3
kernel: Winbond LPT Config: irqtype=pulsed low, high-Z, ECP fifo threshold=0
kernel: Winbond LPT Config: Port mode=ECP and EPP-1.9
kernel: SMSC Super-IO detection, now testing Ports 2F0, 370 ...
kernel: 0x378: FIFO is 16 bytes
kernel: 0x378: writeIntrThreshold is 16
kernel: 0x378: readIntrThreshold is 16
kernel: parport0: PC-style at 0x378 (0x778) [PCSPP,TRISTATE,COMPAT,EPP,ECP]
kernel: parport0: irq 7 detected
kernel: parport1: PC-style at 0x278 [PCSPP,TRISTATE,EPP]
kernel: ppSCSI 0.92 (0.92) installed
kernel: parport0: epst.1 already owner
kernel: parport0: epst.1 tried to release parport when not owner
kernel: epst.1: epst 0.92 (0.92), Shuttle EPST at 0x378 mode 2 (PS/2) dly 1
nice 0 sg 16
kernel: scsi0 : epst
kernel: Vendor: HP Model: C5190A Rev: 3740
kernel: Type: Processor ANSI SCSI revision: 02
kernel: Detected scsi generic sg0 at scsi0, channel 0, id 0, lun 0, type 3
Linux seems a quite a bit slower in transferring the data compared to a test I
did in Windows 2000. In Win2k I scanned a colour document at 150DPI in 12
seconds. It worked out the be a transfer rate or 600kB/s (not bad for a
parallel port).
I might be able to play around with the buffer size to improve the results.
Regards,
Adam
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
This archive was generated by hypermail 2b29 : Mon Feb 26 2001 - 00:41:55 PST