Re: [RFC] Flat /lib/modules/`uname -r`

From: Keith Owens (kaos@ocs.com.au)
Date: Sat Apr 22 2000 - 00:34:00 EDT

  • Next message: Dunlap, Randy: "RE: [RFC] Flat /lib/modules/`uname -r`"

    On Sat, 22 Apr 2000 14:15:50 +1000,
    Andrew Morton <andrewm@uow.edu.au> wrote:
    >Yes, but that requires 'modprobe' and not a bare 'insmod', no?

    Yes, insmod is bare to the metal, it does exactly what you tell it to
    do. I disagree with fiddling around with symlinks or anything else,
    trying to fool insmod. Part of the problem is the pcmcia package going
    straight to insmod first instead of using modprobe. modules.conf
    contains local tailoring of modules and using insmod bypasses that,
    which is why pcmcia needs module options in its config files as well as
    modules.conf. IMHO this is a mistake when pcmcia is integrated into
    the kernel, pcmcia should be using modprobe.

    >Dave is concerned about breaking people's ability to switch between 2.2
    >and 2.3 kernels with the same /etc/*. Adding an 'alias 3c575_cb 3c59x'
    >to conf.modules would cause modprobe to fail if one booted back to 2.2?

    if `kernelversion` != '2.2'
      alias 3c575_cb 3c59x
    endif

    >But it's all muddy now because the 2.3 tree includes the pcmcia
    >support. Some drivers are common (but different) between 2.3 and the
    >pcmcia-cs package. I'd be interested in an explanation of what's going
    >on here.

    So would I ;). But flattening the kernel modules into a single
    directory and scanning that before modules/.../pcmcia would pick up the
    kernel version first, which should be what we want.

    >It's just a matter of being nice to users and installation programs.
    >Maybe we can actually parse the driver source to extract the info?
    >
    > {"3c905B-FX Cyclone 100baseFx",
    > PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
    >becomes
    > {/*SDEVICE*/"3c905B-FX Cyclone 100baseFx",
    > PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
    >
    >this tagging will allow us to associate "3c905B-FX Cyclone 100baseFx"
    >with 3c59x.c, at least.

    Parse the source, surely you jest. That table is static and is
    specific to 3c59x.c. PCI modules are moving towards the standard
    struct pci_device_id which modutils will extract, 3c59x.c should be
    converted to use pci_device_id. BTW I have no objection to
    pci_device_id being expanded to include a human readable text field but
    that is not my decision.

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.rutgers.edu
    Please read the FAQ at http://www.tux.org/lkml/



    This archive was generated by hypermail 2b29 : Sat Apr 22 2000 - 00:36:53 EDT