Re: spinlocks() are severely broken in 2.2.X and 2.4.X for modules

From: David S. Miller (davem@redhat.com)
Date: Sat Jul 01 2000 - 05:52:42 EDT

  • Next message: Andi Kleen: "Re: spinlocks() are severely broken in 2.2.X and 2.4.X for modules"

       Date: Sat, 1 Jul 2000 10:09:56 +0200 (CEST)
       From: =?ISO-8859-1?Q?G=E9rard_Roudier?= <groudier@club-internet.fr>

       On Fri, 30 Jun 2000, Jeff V. Merkey wrote:

       [...]

    > void NWLockLRU(LRU_HANDLE *lru_handle)
    > {
    > spinlock_irqsave(&lru_handle->LRU_spinlock, lru_handle->LRU_flags);

       Does not seem to me safe to store irq flags in shared data structures.
       SMP .vs. UP does not seem to matter here.

    In fact this is also a problem in that it encourages people to believe
    that doing irqsave/irqrestore for the same "flags" in different
    functions is ok. It is not and will crash on some architectures.

    For example, this is OK:

    {
            unsigned long flags;

            spin_lock_irqsave(&lock, flags);
            ...
            spin_unlock_irqrestore(&lock, flags);
    }

    This is not OK:

    void bar(unsigned long flags)
    {
            spin_unlock_irqrestore(&lock, flags);
    }

    void foo(void)
    {
            spin_lock_irqsave(&lock, flags);
            ...
            bar();
    }

    It will crash on Sparc for example, it will probably cause problems
    on some other non-x86 ports as well.

    Later,
    David S. Miller
    davem@redhat.com

    -
    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 Jul 01 2000 - 06:05:31 EDT