[lkml]   [2002]   [Jul]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: PCI config locking (WAS Re: [RFC/CFT] cmd640 irqlocking fixes)2
    Benjamin Herrenschmidt wrote:
    >>Martin this patch should do the job. It uses the correct pci_config_lock
    >>>and it also adds the 2.4 probe safety checks for deciding which pci
    >>>modes to use.
    >>Hrm... pci_config_lock is specific to arch/i386 it seems (and is even
    >>a static in 2.4.19rc3). That is no good as this isn't the only
    >>driver to do config access from interrupts, so at least PPC is
    >>broken in this regard.
    >>Wouldn't it make sense to generalize it and implement it on all archs ?
    >>(That is move extern declaration of it to linux/pci.h, definition to
    >>drivers/pci/pci.c, and so on...)
    >>I'd rather have a per-host lock, but on the other hand, the host bus
    >>mecanism is still quite arch-specific, thus making difficult to use
    >>a per-host lock in drivers, at least in 2.4
    > Ok, fixing my own crap...
    > So there seem to be a problem with your patch: pci_config_lock appears
    > to be an x86-only thing that lives deep inside arch/i386/xxx/pci-pc.c
    > (xxx beeing kernel or pci)
    > On the other hand, there is already such a lock provided by
    > drivers/pci/access.c (pci_lock). You should probably fix your patch
    > to use that one. (and eventually get rid of the pci_config_lock
    > in x86, provided I didn't miss something else). But does anybody
    > but x86 uses CMD640 ? :)

    I agree on the pci_lock item.
    And yes CMD640 chips where quite common on Sparcs about 4-6 years ago.
    I think some Alpha based systems used them too... but I'm not sure.
    Regarding the locking issue. I think the best place to
    put it would be just before call down to the corresonding low level
    functions in the generic IDE layer. We may "overlock" a bit here -
    But who cares? This is by no way a time critical operation.

    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to
    More majordomo info at
    Please read the FAQ at

     \ /
      Last update: 2005-03-22 13:27    [W:0.025 / U:3.668 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site