lkml.org 
[lkml]   [2013]   [May]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    Date
    From
    SubjectRe: [PATCH 7/9] mm: vmscan: Block kswapd if it is encountering pages under writeback
    On Wed, May 15, 2013 at 02:39:02PM -0700, Andrew Morton wrote:
    > On Mon, 13 May 2013 09:12:38 +0100 Mel Gorman <mgorman@suse.de> wrote:
    >
    > > Historically, kswapd used to congestion_wait() at higher priorities if it
    > > was not making forward progress. This made no sense as the failure to make
    > > progress could be completely independent of IO. It was later replaced by
    > > wait_iff_congested() and removed entirely by commit 258401a6 (mm: don't
    > > wait on congested zones in balance_pgdat()) as it was duplicating logic
    > > in shrink_inactive_list().
    > >
    > > This is problematic. If kswapd encounters many pages under writeback and
    > > it continues to scan until it reaches the high watermark then it will
    > > quickly skip over the pages under writeback and reclaim clean young
    > > pages or push applications out to swap.
    > >
    > > The use of wait_iff_congested() is not suited to kswapd as it will only
    > > stall if the underlying BDI is really congested or a direct reclaimer was
    > > unable to write to the underlying BDI. kswapd bypasses the BDI congestion
    > > as it sets PF_SWAPWRITE but even if this was taken into account then it
    > > would cause direct reclaimers to stall on writeback which is not desirable.
    > >
    > > This patch sets a ZONE_WRITEBACK flag if direct reclaim or kswapd is
    > > encountering too many pages under writeback. If this flag is set and
    > > kswapd encounters a PageReclaim page under writeback then it'll assume
    > > that the LRU lists are being recycled too quickly before IO can complete
    > > and block waiting for some IO to complete.
    > >
    > >
    > > ...
    > >
    > > if (PageWriteback(page)) {
    > > - /*
    > > - * memcg doesn't have any dirty pages throttling so we
    > > - * could easily OOM just because too many pages are in
    > > - * writeback and there is nothing else to reclaim.
    > > - *
    > > - * Check __GFP_IO, certainly because a loop driver
    > > - * thread might enter reclaim, and deadlock if it waits
    > > - * on a page for which it is needed to do the write
    > > - * (loop masks off __GFP_IO|__GFP_FS for this reason);
    > > - * but more thought would probably show more reasons.
    > > - *
    > > - * Don't require __GFP_FS, since we're not going into
    > > - * the FS, just waiting on its writeback completion.
    > > - * Worryingly, ext4 gfs2 and xfs allocate pages with
    > > - * grab_cache_page_write_begin(,,AOP_FLAG_NOFS), so
    > > - * testing may_enter_fs here is liable to OOM on them.
    > > - */
    > > - if (global_reclaim(sc) ||
    > > + /* Case 1 above */
    > > + if (current_is_kswapd() &&
    > > + PageReclaim(page) &&
    > > + zone_is_reclaim_writeback(zone)) {
    > > + wait_on_page_writeback(page);
    >
    > wait_on_page_writeback() is problematic.
    >
    > - The page could be against data which is at the remote end of the
    > disk and the wait takes far too long.
    >
    > - The page could be against a really slow device, perhaps one which
    > has a (relatively!) large amount of dirty data pending.
    >

    These are both similar points, the page being waited upon could take an
    abnormal amount of time to be written due to either slow storage or a
    deep writeback queue. This is true.

    > - (What happens if the wait is against a page which is backed by a
    > device which is failing or was unplugged or is taking 60 seconds per
    > -EIO or whatever?)
    >
    > - (Can the wait be against an NFS/NBD/whatever page whose ethernet
    > cable got unplugged?)
    >

    Yes it can and if it happens, kswapd will halt for long periods of time
    deferring all reclaim to direct reclaim. The user-visible impact is that
    unplugged storage may result in more stalls due to direct reclaim.

    The situation gets worse if dirty_ratio amount of pages are backed by
    disconnected storage and the storage is unwilling/unable to discard the
    data. Eventually such a system will have every dirtying process halt in
    balance_dirty_pages. You're correct in pointing out that this patch makes
    the situation slightly worse by indirectly adding kswapd to the list of
    processes that gets stalled.

    > - The termination of wait_on_page_writeback() simply doesn't tell us
    > what we want to know here: that there has been a useful amount of
    > writeback completion against the pages on the tail of this LRU.
    >

    Neither does wait_iff_congested() or congestion_wait() if it waits on the
    wrong queue, wakes up due to IO completing on an unrelated backing_dev or
    wakes up after the timeout with no IO completed. Even if the congestion
    functions wakeup due to IO being complete, there is no guarantee that
    the completed IO is for pages at the end of the LRU or even on the same
    node. As this was already marked PageReclaim and is under writeback there
    is a reasonable assumption it has been on the LRU for some time and that
    wait_on_page_writeback() is not necessarily the worst decision. This is
    what I was taking into account when choosing wait_on_page_writeback().

    However, the unplugged scenario is a good point that would be tricky
    to debug and of the choices available, congestion_wait() is better than
    wait_iff_congested() in this case. It is guaranteed to stall kswapd and we
    *know* at least one page is under writeback so it does not fall foul of the
    old situation where we stalled in congestion_wait() when no IO was in flight.

    Would you like to replace the patch with this version? It includes a
    comment explaining why wait_on_page_writeback() is not used.

    ---8<---
    mm: vmscan: Block kswapd if it is encountering pages under writeback

    Historically, kswapd used to congestion_wait() at higher priorities if it
    was not making forward progress. This made no sense as the failure to make
    progress could be completely independent of IO. It was later replaced by
    wait_iff_congested() and removed entirely by commit 258401a6 (mm: don't
    wait on congested zones in balance_pgdat()) as it was duplicating logic
    in shrink_inactive_list().

    This is problematic. If kswapd encounters many pages under writeback and
    it continues to scan until it reaches the high watermark then it will
    quickly skip over the pages under writeback and reclaim clean young
    pages or push applications out to swap.

    The use of wait_iff_congested() is not suited to kswapd as it will only
    stall if the underlying BDI is really congested or a direct reclaimer was
    unable to write to the underlying BDI. kswapd bypasses the BDI congestion
    as it sets PF_SWAPWRITE but even if this was taken into account then it
    would cause direct reclaimers to stall on writeback which is not desirable.

    This patch sets a ZONE_WRITEBACK flag if direct reclaim or kswapd is
    encountering too many pages under writeback. If this flag is set and
    kswapd encounters a PageReclaim page under writeback then it'll assume
    that the LRU lists are being recycled too quickly before IO can complete
    and block waiting for some IO to complete.

    Signed-off-by: Mel Gorman <mgorman@suse.de>
    Reviewed-by: Michal Hocko <mhocko@suse.cz>
    Acked-by: Rik van Riel <riel@redhat.com>
    ---
    include/linux/mmzone.h | 8 +++++
    mm/vmscan.c | 80 ++++++++++++++++++++++++++++++++++++--------------
    2 files changed, 66 insertions(+), 22 deletions(-)

    diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
    index 2aaf72f..fce64af 100644
    --- a/include/linux/mmzone.h
    +++ b/include/linux/mmzone.h
    @@ -499,6 +499,9 @@ typedef enum {
    * many dirty file pages at the tail
    * of the LRU.
    */
    + ZONE_WRITEBACK, /* reclaim scanning has recently found
    + * many pages under writeback
    + */
    } zone_flags_t;

    static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
    @@ -526,6 +529,11 @@ static inline int zone_is_reclaim_dirty(const struct zone *zone)
    return test_bit(ZONE_TAIL_LRU_DIRTY, &zone->flags);
    }

    +static inline int zone_is_reclaim_writeback(const struct zone *zone)
    +{
    + return test_bit(ZONE_WRITEBACK, &zone->flags);
    +}
    +
    static inline int zone_is_reclaim_locked(const struct zone *zone)
    {
    return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
    diff --git a/mm/vmscan.c b/mm/vmscan.c
    index d6c916d..45aee36 100644
    --- a/mm/vmscan.c
    +++ b/mm/vmscan.c
    @@ -724,25 +724,53 @@ static unsigned long shrink_page_list(struct list_head *page_list,
    may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
    (PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

    + /*
    + * If a page at the tail of the LRU is under writeback, there
    + * are three cases to consider.
    + *
    + * 1) If reclaim is encountering an excessive number of pages
    + * under writeback and this page is both under writeback and
    + * PageReclaim then it indicates that pages are being queued
    + * for IO but are being recycled through the LRU before the
    + * IO can complete. Waiting on the page itself risks an
    + * indefinite stall if it is impossible to writeback the
    + * page due to IO error or disconnected storage so instead
    + * block for HZ/10 or until some IO completes then clear the
    + * ZONE_WRITEBACK flag to recheck if the condition exists.
    + *
    + * 2) Global reclaim encounters a page, memcg encounters a
    + * page that is not marked for immediate reclaim or
    + * the caller does not have __GFP_IO. In this case mark
    + * the page for immediate reclaim and continue scanning.
    + *
    + * __GFP_IO is checked because a loop driver thread might
    + * enter reclaim, and deadlock if it waits on a page for
    + * which it is needed to do the write (loop masks off
    + * __GFP_IO|__GFP_FS for this reason); but more thought
    + * would probably show more reasons.
    + *
    + * Don't require __GFP_FS, since we're not going into the
    + * FS, just waiting on its writeback completion. Worryingly,
    + * ext4 gfs2 and xfs allocate pages with
    + * grab_cache_page_write_begin(,,AOP_FLAG_NOFS), so testing
    + * may_enter_fs here is liable to OOM on them.
    + *
    + * 3) memcg encounters a page that is not already marked
    + * PageReclaim. memcg does not have any dirty pages
    + * throttling so we could easily OOM just because too many
    + * pages are in writeback and there is nothing else to
    + * reclaim. Wait for the writeback to complete.
    + */
    if (PageWriteback(page)) {
    - /*
    - * memcg doesn't have any dirty pages throttling so we
    - * could easily OOM just because too many pages are in
    - * writeback and there is nothing else to reclaim.
    - *
    - * Check __GFP_IO, certainly because a loop driver
    - * thread might enter reclaim, and deadlock if it waits
    - * on a page for which it is needed to do the write
    - * (loop masks off __GFP_IO|__GFP_FS for this reason);
    - * but more thought would probably show more reasons.
    - *
    - * Don't require __GFP_FS, since we're not going into
    - * the FS, just waiting on its writeback completion.
    - * Worryingly, ext4 gfs2 and xfs allocate pages with
    - * grab_cache_page_write_begin(,,AOP_FLAG_NOFS), so
    - * testing may_enter_fs here is liable to OOM on them.
    - */
    - if (global_reclaim(sc) ||
    + /* Case 1 above */
    + if (current_is_kswapd() &&
    + PageReclaim(page) &&
    + zone_is_reclaim_writeback(zone)) {
    + congestion_wait(BLK_RW_ASYNC, HZ/10);
    + zone_clear_flag(zone, ZONE_WRITEBACK);
    +
    + /* Case 2 above */
    + } else if (global_reclaim(sc) ||
    !PageReclaim(page) || !(sc->gfp_mask & __GFP_IO)) {
    /*
    * This is slightly racy - end_page_writeback()
    @@ -757,9 +785,13 @@ static unsigned long shrink_page_list(struct list_head *page_list,
    */
    SetPageReclaim(page);
    nr_writeback++;
    +
    goto keep_locked;
    +
    + /* Case 3 above */
    + } else {
    + wait_on_page_writeback(page);
    }
    - wait_on_page_writeback(page);
    }

    if (!force_reclaim)
    @@ -1374,8 +1406,10 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
    * isolated page is PageWriteback
    */
    if (nr_writeback && nr_writeback >=
    - (nr_taken >> (DEF_PRIORITY - sc->priority)))
    + (nr_taken >> (DEF_PRIORITY - sc->priority))) {
    wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);
    + zone_set_flag(zone, ZONE_WRITEBACK);
    + }

    /*
    * Similarly, if many dirty pages are encountered that are not
    @@ -2669,8 +2703,8 @@ static bool prepare_kswapd_sleep(pg_data_t *pgdat, int order, long remaining,
    * the high watermark.
    *
    * Returns true if kswapd scanned at least the requested number of pages to
    - * reclaim. This is used to determine if the scanning priority needs to be
    - * raised.
    + * reclaim or if the lack of progress was due to pages under writeback.
    + * This is used to determine if the scanning priority needs to be raised.
    */
    static bool kswapd_shrink_zone(struct zone *zone,
    struct scan_control *sc,
    @@ -2697,6 +2731,8 @@ static bool kswapd_shrink_zone(struct zone *zone,
    if (nr_slab == 0 && !zone_reclaimable(zone))
    zone->all_unreclaimable = 1;

    + zone_clear_flag(zone, ZONE_WRITEBACK);
    +
    return sc->nr_scanned >= sc->nr_to_reclaim;
    }


    \
     
     \ /
      Last update: 2013-05-16 15:42    [W:3.163 / U:0.004 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site