[lkml]   [2006]   [Jun]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: [PATCH] tmpfs time granularity fix for [acm]time going backwards. Also VFS time granularity bug on creat(). (Repost, more content)
    On Sun, Jun 11, 2006 at 07:10:31PM +0100, Hugh Dickins wrote:
    > On Sun, 11 Jun 2006, Robin H. Johnson wrote:
    > > After some digging, I found that this was being caused by tmpfs not having a
    > > time granularity set, thus inheriting the default 1s granularity.
    > That's a great little discovery, and a very good report and analysis:
    > thank you. Seems tmpfs got missed when s_time_gran was added in 2.6.11,
    > and I (tmpfs maintainer) failed to notice that patch going past.
    Ah, ok, it was mentioned to me there was a maintainer for tmpfs, but I
    found no mention of you in the tmpfs source, or MAINTAINERS. Maybe
    submit a patch ;-).

    > Perhaps we could devise a debug WARN_ON somewhere to check consistent
    > granularity; but I don't have the ingenuity right now, and would need
    > an additional superblock field or flag to not spam the logs horribly.
    > Perhaps it's easier just to delete CURRENT_TIME, converting its users.
    Yes, I'd agree that replacing CURRENT_TIME in filesystems with
    current_fs_time should be worthwhile for all filesystems - That,
    combined with your patch below to ensure they all use s_time_gran,
    should ensure safety.

    A total removal of CURRENT_TIME wouldn't work, there are a few other
    users besides setting [acm]times - however as above, we should be able
    to kill it for all filesystems.

    However CURRENT_TIME_SEC looks safe to convert, all of it's users are

    > Setting that safety aside, the patch below (against 2.6.17-rc6) looks
    > to me like all that's currently needed in mainline - but ecryptfs and
    > reiser4 in the mm tree will also want fixing, and more discrepancies
    > are sure to trickle in later.
    I checked at well, and this does cover every filesystem I see in the

    > If anyone thinks tmpfs is the most important to fix (I would think
    > that, wouldn't I?), I can forward your fix to Linus ahead of the rest.
    > Or if people agree the patch below is good, I can sign it off and send;
    > or FS maintainers extract their own little parts.
    I'd appreciate it tmpfs either of the fixes actually making it to
    2.6.17, there are a reasonable number of Gentoo users that use tmpfs as
    temporary storage to compile stuff, and there's a long-standing argument
    that tmpfs wasn't safe for that, due to this bug ;-).

    Acked-By: Robin H. Johnson <>

    Robin Hugh Johnson
    E-Mail :
    GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85
    [unhandled content-type:application/pgp-signature]
     \ /
      Last update: 2006-06-12 07:12    [W:0.021 / U:13.740 seconds]
    ©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site