[lkml]   [2002]   [Jul]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: [patch 2/13] remove pages from the LRU in __free_pages_ok()
    Linus Torvalds wrote:
    > On Sun, 28 Jul 2002, David S. Miller wrote:
    > >
    > > Also skb_release_data(), ___pskb_trim() and __pskb_pull_tail(). Can these
    > > ever perform the final release against a page which is on the LRU?
    > > In interrupt context?
    > >
    > > These page releases run from either user or softint context.
    > >
    > > They must never run from HW irqs, in fact there is a BUG()
    > > check there against this.
    > >From a page cache standpoint softirq's are 100% equivalent to hardware
    > irq's, so that doesn't much help here.

    So we cannot take pagemap_lru_lock from softirq context.

    hmm. ia32's do_IRQ() doesn't run do_sotfirq() any more, but the
    other architectures do. What's up with that?

    > > Any page that can be found in the page cache can end up here. So
    > > whatever that mean for "release against a page which is on the LRU"
    > > applies here.
    > Being in the page cache can be ok. What is _not_ ok is if this function
    > can ever be the last user to release such a page (ie the original page
    > count of the page had better be held on by something else - which usually
    > is the page-cacheness itself, since shrinking the page cache will only
    > happen for pages that are unused).

    shrink_cache() explicitly removes the page from the LRU (well, it
    won't even get that far if someone else has a ref).

    truncate_complete_page() _used_ to explicitly remove the page from
    the lru, but we took that out. And it was never reliable anyway,
    because some pages were left there (invalidatepage failed).

    Anyway. I have patches against 2.5.24, which work, which
    turn pagemap_lru_lock into an innermost, irq-safe lock. If
    we get that in place then page_cache_release() from IRQ context
    is fine.

    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:0.022 / U:19.540 seconds]
    ©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site