Very Computer

Hi,

I've got a little problem with my Adaptec AHA1541 Controller. I just
installed SuSE Linux 7.2 with Kernel 2.4.4. Now when I try to mount a CD-ROM
from my SCSI-drive I first get "unable to determine CD format" or something
like that two times and on the third try my kernel panics with "DMA Pool
exhausted"... The same problem occured when I tried to create a swap-space
on a SCSI-disk. The old 2.2.16 kernel (SuSE 7.0) worked just fine. Does
anyone know what the problem might be??

Thanks,
--Michael

P.S.: Here's a copy from my /var/log/messages - in case that helps:

Jun 15 20:23:44 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:44 bilbo kernel:  I/O error: dev 0b:01, sector 2
Jun 15 20:23:44 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:44 bilbo kernel:  I/O error: dev 0b:01, sector 2
Jun 15 20:23:44 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:44 bilbo kernel:  I/O error: dev 0b:01, sector 0
Jun 15 20:23:44 bilbo kernel: FAT bread failed
Jun 15 20:23:45 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:45 bilbo kernel:  I/O error: dev 0b:01, sector 128
Jun 15 20:23:45 bilbo kernel: read_super_block: unable to read superblock on
dev 0b:01
Jun 15 20:23:45 bilbo kernel: read_old_super_block: try to find super block
in old location
Jun 15 20:23:45 bilbo kernel: read_old_super_block: can't find a reiserfs
filesystem on dev 0b:01.
Jun 15 20:23:45 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:45 bilbo kernel:  I/O error: dev 0b:01, sector 0
Jun 15 20:23:45 bilbo kernel: FAT bread failed
Jun 15 20:23:45 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:45 bilbo kernel:  I/O error: dev 0b:01, sector 0
Jun 15 20:23:45 bilbo kernel: FAT bread failed
Jun 15 20:23:45 bilbo kernel: sr: ran out of mem for scatter pad
Jun 15 20:23:45 bilbo kernel:  I/O error: dev 0b:01, sector 128
Jun 15 20:23:45 bilbo kernel: read_super_block: unable to read superblock on
dev 0b:01
Jun 15 20:23:45 bilbo kernel: read_old_super_block: try to find super block
in old location
Jun 15 20:23:45 bilbo kernel: Warning - running low on DMA memory
Jun 15 20:23:45 bilbo kernel: dma_free_sectors:4
Jun 15 20:23:45 bilbo kernel: use_sg:0^Ii:0
Jun 15 20:23:45 bilbo kernel: request_bufflen:0
Jun 15 20:23:45 bilbo kernel: Total 0 sectors consumed

This should go to either the linux admin
list, the comp.os.linux.system newsgroup,
or the linux kernel mailing list.

It's not likely anything wrong with your
hardware, and the 154x driver has
been mature for years. But something
new in 2.4.x could be screwing it
up, and there might even be a record
of it in the kernel list archives. I don't
know immediately how one would search
past weeks of the list, that would depend
either on having some search engine
available at the site hosting the kernel
mailing list archive or some external www archive that searches
the kernel list when the messages
are displayed on one of the sites that
make traffic on the kernel list available
on the WWW.

There are a few WWW archives of the kernel
list: a hypermail archive at U. of Indiana, The Mail Archive keeps one,
*.*.erau.*
(I just think of it as "erau", some uni
site in Germany), etc.

Geocrawler could perhaps find where
the kernel list archive is kept online, and
it may even search it and other mailing
lists that it knows about for mention of the
154x (or "1540", "1542b", etc). You could probably find Geocrawler via Google
or All the Web, All of the Time, or one of the other big search engines.

(Sorry to send you on this roundabout
URL hunt, but I'm sitting at my mother's
AOL box at the moment instead of my
linux machine, and don't have my bookmarks handy).

Google would probably also find The
Mail Archive home site for you. Google
is easy to remember:

<http://www.google.com/>

They have also taken over Deja News
old Usenet archive, but it's at a different
URL than the main Google www search
engine interface (should be a link from
either to the other somewhere around
their front page, though).

The linux-kernel list isn't at vger.rutgers.edu anymore, that
machine ate its disks some months
ago, and the list was moved temporarily to a Redhat server. I think it now has
an address of

a nice abstract host.domain that one
can move around to anywhere and
still get to it via DNS without having
to change the host address for the
mailing list.

I have a couple of 154x boards, a
1540 with the last 154x Adaptec bios and
a 1542b, same bios version. They work great (slow compared to pci cards, but
have had rock solid drivers for the last six years), but I haven't upgraded
past 2.2.18 (plus Andrea Archangeli's global VM patch that went into 2.2.19)
yet, so I haven't seen any possible 2.4.x generated
problems with the 154x scsi driver personally.

I do think the kernel guys, though not many of them still use 154x cards, are
likely to recognize the problem more
quickly than anyone reading the comp.os.linux.hardware group. At least someone
there will know who if
anyone now maintains the kernel
driver (Drew Eckhart wrote the 1.0.9 driver that worked so well, and he
upgraded it for 2.0.x, but he's been busy elsewhere for years).

The traffic level here is just too high for
kernel/scsi/driver experts to have the time to read it.

Even a low-end pci symbios 810 will
blow the doors off the 154x cards for
throughput, not to mention the newer
symbios 85C8xx cards, so thinking about budgeting $25-50 for an upgrade (used,
probably, although I think Joseph Gadoury may still have some in bubble packs:
scsi-2 or fast scsi-2 only, not ultrawide). Asus made some cheap ones that you
can sometimes find for sale
used, but you have to check whether
it is complete (they used to have bios-on-mb and just an scsi pinout interface
card in the pci slot for some motherboards-adapter combos).

Even if you don't need the disk throughput, these old 154x cards won't last
forever.

Good luck.

Regards,

Clayton Weaver

"Everyone is ignorant, just about different things."  Will Rogers