Loading ...
Sorry, an error occurred while loading the content.

fixed CDROM... Re: no idescsi CDROM, no PCMCIA: can I read your 2.4.0 .config?

Expand Messages
  • jonh@cs.dartmouth.edu
    ... manager ... No, I ll look at that tonight. ... It lets cdrecord see the -R and -RW features of the drive. ... Yes -- I set all of the pcmcia options to
    Message 1 of 7 , Feb 1 4:57 PM
    • 0 Attachment
      > On Tue, Jan 30, 2001 at 07:02:47PM -0000, jonh@c... wrote:
      > > I'm still stumped on the PCMCIA/Xircom Tulip problem. The card
      manager
      > > still sees my card as an "Anonymous Memory" card.

      David Rodriguez <dibeas@i...> wrote:
      > Have you tried to see what /sbin/dump_cis says?
      No, I'll look at that tonight.

      > P.S. What advantages does scsi-emulation have over the ide driver?
      It lets cdrecord see the -R and -RW features of the drive.

      Zach Loafman <zml@z...> wrote:
      > Have you tried the standalone modules? I had trouble with
      > xircom_tulip_cb.o in the kernel (in 2.4.0), but I turned off all the
      > pcmcia modules in the kernel and compiled the pcmcia-cs stuff
      > against my 2.4.0 kernel and everything was beautiful.
      Yes -- I set all of the pcmcia options to 'm', built a completely
      fresh kernel (from mrproper -- I learned to stash away my .config the
      first time :v), and built and installed a fresh pcmcia-cs-3.1.24.

      If it's convenient, please post your 2.4.0 .config. (I'll try to post
      mine, but of course without a NIC on that box, it's tricky. :v) I'll
      also try David's suggestion.

      Thanks all,

      --Jon
    • jonh@cs.dartmouth.edu
      ... I tried /sbin/dump_cis. Under 2.2.17 (where the pcmcia card works fine), it describes the card (in slot 1) pretty reasonably. Under 2.4.0, it says both
      Message 2 of 7 , Feb 2 7:44 AM
      • 0 Attachment
        > > > I'm still stumped on the PCMCIA/Xircom Tulip problem. The card
        > manager
        > > > still sees my card as an "Anonymous Memory" card.
        > David Rodriguez <dibeas@i...> wrote:
        > > Have you tried to see what /sbin/dump_cis says?
        > No, I'll look at that tonight.
        I tried /sbin/dump_cis. Under 2.2.17 (where the pcmcia card works
        fine), it describes the card (in slot 1) pretty reasonably. Under
        2.4.0, it says both slots are empty. (Especially odd, since during
        boot, cs tries to load "memory_cs", the module for Anonymous Memory.)

        I suspect I just have to add a configuration to get the right PCMCIA
        module's attention, but I'm not sure where to start. Back to the
        HOWTO...

        In other news, I tried the bios upgrade from TuxTops (since I was
        still within my 30-day money-back guarantee period :v). It worked
        fine, but you should have seen the look on my face when the mandrake
        hardware reconfiguration screen came up and said, "The following
        hardware has been removed from your system: ATI Mobility Pro AGP/2X,
        Maestro 2ESS..." :v) Happily, the next screen reported that the same
        hardware had also been added, so apparently the new bios just moved
        them around. Whew!

        --Jon
      • jonh@cs.dartmouth.edu
        My friend and I just spent an exciting evening debugging the PCMCIA problem. The symptom: my Xircom card kept appearing as Anonymous memory when booting in
        Message 3 of 7 , Feb 2 11:17 PM
        • 0 Attachment
          My friend and I just spent an exciting evening debugging the PCMCIA
          problem. The symptom: my Xircom card kept appearing as "Anonymous
          memory" when booting in 2.4.0, but the 2.2.17-based config that came
          with Mandrake worked great. What's up with that?

          It came down to kernel config. My first try was CONFIG_PCMCIA=y. I
          read that maybe the in-kernel PCMCIA support in 2.4.0 was flaky, so I
          set it down to CONFIG_PCMCIA=m, and figured I'd be getting the modules
          from pcmcia-cs-3.1.24. The short answer was that with CONFIG_PCMCIA=m,
          the kernel was supplying (year-old) PCMCIA modules, and pcmcia-cs was
          happily ignoring the ones it came with. The solution is to set
          CONFIG_PCMCIA=n (and make oldconfig in the kernel directory to tidy up
          the rest of the config file), and then remake the pcmcia-cs stuff.

          Here are the longer details, to provide search fodder for the next
          poor slob to hit this problem.

          Here were the messages that I'd get that were clearly wrong:

          Feb 2 22:05:46 eustace cardmgr[1718]: socket 1: Anonymous Memory
          Feb 2 22:05:46 eustace cardmgr[1718]: executing: 'modprobe memory_cs'
          Feb 2 22:05:46 eustace cardmgr[1718]: + modprobe: Can't locate module
          memory_cs

          But the real trouble was this:

          Feb 2 22:05:42 eustace kernel: Intel PCIC probe:
          Feb 2 22:05:42 eustace kernel: Intel i82365sl DF ISA-to-PCMCIA at
          port 0x3e0 ofs 0x00, 2 sockets
          Feb 2 22:05:42 eustace kernel: host opts [0]: none
          Feb 2 22:05:42 eustace kernel: host opts [1]:
          none

          When booting on my old 2.2.17-mdk kernel, I'd see something more like
          this (this one is actually on my new 2.4.0 kernel):

          Feb 2 22:44:25 eustace kernel: Intel PCIC probe: PCI: Found IRQ 11
          for device 00:04.0
          Feb 2 22:44:25 eustace kernel: TI 1225 rev 01 PCI-to-CardBus at
          slot 00:04, mem 0x10000000
          Feb 2 22:44:25 eustace kernel: host opts [0]: [ring] [serial pci
          & irq] [pci irq 11] [lat 168/32] [bus 2/5]
          Feb 2 22:44:25 eustace kernel: host opts [1]: [ring] [serial pci
          & irq] [pci irq 11] [lat 168/32] [bus
          6/9]

          Notice that the correct modules more accurately recognize the host
          adaptor as a TI1225, not a generic form of i82365. After poking around
          in the pcmcia-cs-3.1.24/modules source files, I noticed that the TI
          1225 string should appear in the i82365.o module if the module is
          prepared to recognize the chip. Indeed, a 'strings' on
          /lib/modules/2.2.17-mdk/pcmcia/i82365.o showed a 'TI 1225' line, and
          the same strings on .../2.4.0/pcmcia/i82365.o showed the line missing.
          From there, the only remaining detective work was to realize that the
          2.4.0 module was coming from the in-kernel pcmcia sources, and to
          notice that the version info in the kernel's i82365.c was dated in
          1999.

          On to IR, or whatever else is still broken. :v)

          --Jon
        Your message has been successfully submitted and would be delivered to recipients shortly.