Hi Cameron, On Tue, Jan 06, 2004 at 03:00:20PM +0800, Cameron Patrick wrote: > I just attempted an installation on an alpha with a daily 'business > card' image from yesterday. After giving kernel boot parameters as > vorlon described, the initial boot was a success. The SCSI controller > and attached drives were detected. Unfortunately, some devfs weirdness > seemed to prevent the installer from seeing the CDROM drive: it showed > up in /dev/scsi/... but not in /dev/cdroms. Mounting the CD by hand > worked, but d-i still refused to recognise it. Since the driver for the > machine's network cards was on the CD, the installation got no further. > I was using a serial console, which mostly worked, although it seemed a bit > peculiar for it to ask for a keyboard layout, and the screen redraw was > significantly slower than it needed to be. > Some relevant output from capturing the serial console: > ~ # ls -l /dev/cdroms/ > lr-xr-xr-x 1 root root 34 Jan 1 1970 cdrom0 -> ?????????scsi/host0/bus0/target1/lun0/cd > ~ # ls -l /dev/scsi/host0/bus0/target1/lun0 > brw-rw-rw- 1 root root 11, 0 Jan 1 1970 cd > ~ # mount /dev/cdroms/cdrom0 /cdrom > mount: Mounting /dev/cdroms/cdrom0 on /cdrom failed: No such file or directory > ~ # mount /dev/scsi/host0/bus0/target1/lun0/cd /cdrom > mount: /dev/scsi/host0/bus0/target1/lun0/cd is write-protected, mounting read-only > mount: /dev/scsi/host0/bus0/target1/lun0/cd is write-protected, mounting read-only > mount: /dev/scsi/host0/bus0/target1/lun0/cd is write-protected, mounting read-only > It looks to me like the first three characters in the link in > /dev/cdroms should have been '../'. In case they get mangled by email, > a hex dump is: > 00000040 3e 20 c3 bf c3 bf c3 bf 73 63 73 69 2f 68 6f 73 |> ......scsi/hos| > Hmm... six bytes... utf8 weirdness happening somewhere? Thanks for giving it a try! I'll take a look at where this bad symlink is coming from, and hopefully have this fixed for tomorrow's build. (This problem didn't happen with my IDE CDROM.) > Also, I saw a lot of suspicious-looking "unaligned trap" kernel > messages. I've included a complete dmesg output at the end of this > message in case it helps anyone. > main-menu(252): unaligned trap at 0000020000044eb0: 0000000120014dbc 2d 31 > main-menu(252): unaligned trap at 0000020000044eb4: 0000000120014dc4 2d 9 > main-menu(252): unaligned trap at 000002000004183c: 0000000120015fec 2d 31 > main-menu(252): unaligned trap at 0000020000041840: 0000000120015fdc 2d 9 > main-menu(252): unaligned trap at 0000000120002838: 000000012001713c 29 9 > main-menu(252): unaligned trap at 0000000120002844: 000000012001c144 29 2 > main-menu(252): unaligned trap at 0000000120002880: 000000012001c13c 29 9 > main-menu(252): unaligned trap at 0000020000041fc8: 0000000120014dc4 29 3 > kbd-chooser(402): unaligned trap at 0000020000041fc8: 0000000120028be4 29 3 > kbd-chooser(402): unaligned trap at 0000020000041fc8: 0000000120028be4 29 3 > main-menu(252): unaligned trap at 0000000120002838: 00000001200199ec 29 9 > main-menu(252): unaligned trap at 0000000120002844: 000000012001d574 29 2 Hmm, I suppose the best way to find these might be to look at the build logs... If anyone has time to go digging, it should be easy to get line numbers for these buglets. (Note that unaligned traps are a bigger issue in 2.6, as these now generate signals to the userspace process on alpha.) Cheers, -- Steve Langasek postmodern programmer
Attachment:
pgpDzdP2wiLEJ.pgp
Description: PGP signature