[lkml]   [2007]   [Oct]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: [PATCH resend] ramdisk: fix zeroed ramdisk pages on memory pressure
    On Monday 15 October 2007 19:05, Christian Borntraeger wrote:
    > Am Montag, 15. Oktober 2007 schrieb Nick Piggin:
    > > On Monday 15 October 2007 18:28, Christian Borntraeger wrote:
    > > > Andrew, this is a resend of a bugfix patch. Ramdisk seems a bit
    > > > unmaintained, so decided to sent the patch to you :-).
    > > > I have CCed Ted, who did work on the code in the 90s. I found no
    > > > current email address of Chad Page.
    > >
    > > This really needs to be fixed...
    > I obviously agree ;-)
    > We have seen this problem happen several times.
    > > I can't make up my mind between the approaches to fixing it.
    > >
    > > On one hand, I would actually prefer to really mark the buffers
    > > dirty (as in: Eric's fix for this problem[*]) than this patch,
    > > and this seems a bit like a bandaid...
    > I have never seen these patches, so I cannot comment on them.

    > > On the other hand, the wound being covered by the bandaid is
    > > actually the code in the buffer layer that does this latent
    > > "cleaning" of the page because it sadly doesn't really keep
    > > track of the pagecache state. But it *still* feels like we
    > > should be marking the rd page's buffers dirty which should
    > > avoid this problem anyway.
    > Yes, that would solve the problem as well. As long as we fix
    > the problem, I am happy. On the other hand, do you see any
    > obvious problem with this "bandaid"?

    I don't think so -- in fact, it could be the best candidate for
    a minimal fix for stable kernels (anyone disagree? if not, maybe
    you could also send this to the stable maintainers?).

    But I do want to have this fixed in a "nice" way. eg. I'd like
    it to mark the buffers dirty because that actually results in
    more reuse of generic kernel code, and also should make rd
    behave more naturally (I like using it to test filesystems
    because it can expose a lot more concurrency than something like
    loop on tmpfs). It should also be possible to actually have
    rd's buffer heads get reclaimed as well, preferably while
    exercising the common buffer paths and without writing much new

    All of that is secondary to fixing the data corruption problem
    of course! But the fact that those alternate patches do exist now
    means I want to just bring them into the discussion again before
    merging one or the other.
    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: 2007-10-15 11:31    [W:0.024 / U:3.856 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site