[RFC] removal of legacy cdrom drivers (Re: [PATCH] mcdx.c insanity removal)

viro_at_parcelfarce.linux.theplanet.co.uk
Date: 05/03/04

  • Next message: David Lang: "Re: [hsflinux] [PATCH] Blacklist binary-only modules lying about their license" 
    Date:	Mon, 3 May 2004 02:16:29 +0100
To: Linus Torvalds <torvalds@osdl.org>

    Aiiee...

    You know, mcdx.c is like a roadkill - just can't stop looking at the thing.

    a) it doesn't initialize when non-modular. Since 2.5.0. Fixed, can send.
    b) it leaks on failure exits; since forever. Fixed, can send.
    c) it has several lovely bugs (e.g. foo = bar | SINGLE ? ... : ..., where
    SINGLE is defined as 1; from the context it's obvious that it should've
    been bar & SINGLE; there since 1.3.71). Fixed, can send.
    d) it has tons of crimes against decency - e.g.
            foo = bar > (foo += baz) ? bar : foo;
    where foo, bar and baz and fairly long expressions. Partially fixed, can send.
    Oh, and let's not forget the lovely variable names - e.g. "stuffp" through
    the entire thing. That's "pointer to all stuff we have about that device".

    e) it does *entire* *damn* *IO* without dropping queue lock. And yes, it
    *does* block in there. A lot. Always had.

    I can fix the last one, however at that point I'm really starting to wonder
    if we want to keep the FPOS in the tree. I don't have the hardware and while
    I'm reasonably sure that I can split the transition into provably correct
    small steps, I doubt that there's any point in doing that. Driver is obviously
    not used by anyone and hadn't been used for years.

    The same goes for the rest of drivers/cdrom - cdrom.c is used, all right
    (ide-cd, sr, pcd), but everything else is
            * abandoned by maintainers 5-6 years ago
            * broken
            * obviously not used by anybody
            * impossible to debug due to lack of hardware
            * fucking ugly

    How about removing all that stuff instead of keeping the known broken ***
    in the tree? If somebody wants it back, they can always pick the versions
    circa 2.6.0 from archives.

    If you are OK with that (and nobody on l-k stands up and claims that they want
    it alive and *claims* *that* *right* *fucking* *NOW*) I'll send you a patch
    putting these buggers out of their misery.
    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

  • Next message: David Lang: "Re: [hsflinux] [PATCH] Blacklist binary-only modules lying about their license"