Synopsis: if shmem_writepage calls swap_writepage directly, most shmem
swap loads benefit, and a catastrophic interaction between SLUB and some
flash storage is avoided.
shmem_writepage() has always been peculiar in making no attempt to write:
it has just transferred a shmem page from file cache to swap cache, then
let that page make its way around the LRU again before being written and
freed.
The idea was that people use tmpfs because they want those pages to stay
in RAM; so although we give it an overflow to swap, we should resist
writing too soon, giving those pages a second chance before they can be
reclaimed.
That was always questionable, and I've toyed with this patch for years;
but never had a clear justification to depart from the original design.
It became more questionable in 2.6.28, when the split LRU patches classed
shmem and tmpfs pages as SwapBacked rather than as file_cache: that in
itself gives them more resistance to reclaim than normal file pages. I
prepared this patch for 2.6.29, but the merge window arrived before I'd
completed gathering statistics to justify sending it in.
Then while comparing SLQB against SLUB, running SLUB on a laptop I'd
habitually used with SLAB, I found SLUB to run my tmpfs kbuild swapping
tests five times slower than SLAB or SLQB - other machines slower too, but
nowhere near so bad. Simpler "cp -a" swapping tests showed the same.
slub_max_order=0 brings sanity to all, but heavy swapping is too far from
normal to justify such a tuning. The crucial factor on that laptop turns
out to be that I'm using an SD card for swap. What happens is this:
By default, SLUB uses order-2 pages for shmem_inode_cache (and many other
fs inodes), so creating tmpfs files under memory pressure brings lumpy
reclaim into play. One subpage of the order is chosen from the bottom of
the LRU as usual, then the other three picked out from their random
positions on the LRUs.
In a tmpfs load, many of these pages will be ones which already passed
through shmem_writepage, so already have swap allocated. And though their
offsets on swap were probably allocated sequentially, now that the pages
are picked off at random, their swap offsets are scattered.
But the flash storage on the SD card is very sensitive to having its
writes merged: once swap is written at scattered offsets, performance
falls apart. Rotating disk seeks increase too, but less disastrously.
So: stop giving shmem/tmpfs pages a second pass around the LRU, write them
out to swap as soon as their swap has been allocated.
It's surely possible to devise an artificial load which runs faster the
old way, one whose sizing is such that the tmpfs pages on their second
pass are the ones that are wanted again, and other pages not.
But I've not yet found such a load: on all machines, under the loads I've
tried, immediate swap_writepage speeds up shmem swapping: especially when
using the SLUB allocator (and more effectively than slub_max_order=0), but
also with the others; and it also reduces the variance between runs. How
much faster varies widely: a factor of five is rare, 5% is common.
One load which might have suffered: imagine a swapping shmem load in a
limited mem_cgroup on a machine with plenty of memory. Before 2.6.29 the
swapcache was not charged, and such a load would have run quickest with
the shmem swapcache never written to swap. But now swapcache is charged,
so even this load benefits from shmem_writepage directly to swap.
Apologies for the #ifndef CONFIG_SWAP swap_writepage() stub in swap.h:
it's silly because that will never get called; but refactoring shmem.c
sensibly according to CONFIG_SWAP will be a separate task.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
try_to_free_pages() is used for the direct reclaim of up to
SWAP_CLUSTER_MAX pages when watermarks are low. The caller to
alloc_pages_nodemask() can specify a nodemask of nodes that are allowed to
be used but this is not passed to try_to_free_pages(). This can lead to
unnecessary reclaim of pages that are unusable by the caller and int the
worst case lead to allocation failure as progress was not been make where
it is needed.
This patch passes the nodemask used for alloc_pages_nodemask() to
try_to_free_pages().
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When a shrinker has a negative number of objects to delete, the symbol
name of the shrinker should be printed, not shrink_slab. This also makes
the error message slightly more informative.
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Change the page_mkwrite prototype to take a struct vm_fault, and return
VM_FAULT_xxx flags. There should be no functional change.
This makes it possible to return much more detailed error information to
the VM (and also can provide more information eg. virtual_address to the
driver, which might be important in some special cases).
This is required for a subsequent fix. And will also make it easier to
merge page_mkwrite() with fault() in future.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <joel.becker@oracle.com>
Cc: Artem Bityutskiy <dedekind@infradead.org>
Cc: Felix Blyakher <felixb@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=9838
On i386, HZ=1000, jiffies_to_clock_t() converts time in a somewhat strange
way from the user's point of view:
# echo 500 >/proc/sys/vm/dirty_writeback_centisecs
# cat /proc/sys/vm/dirty_writeback_centisecs
499
So, we have 5000 jiffies converted to only 499 clock ticks and reported
back.
TICK_NSEC = 999848
ACTHZ = 256039
Keeping in-kernel variable in units passed from userspace will fix issue
of course, but this probably won't be right for every sysctl.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
CONFIG_DEBUG_PAGEALLOC is now supported by x86, powerpc, sparc64, and
s390. This patch implements it for the rest of the architectures by
filling the pages with poison byte patterns after free_pages() and
verifying the poison patterns before alloc_pages().
This generic one cannot detect invalid page accesses immediately but
invalid read access may cause invalid dereference by poisoned memory and
invalid write access can be detected after a long delay.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I notice there are many places doing copy_from_user() which follows
kmalloc():
dst = kmalloc(len, GFP_KERNEL);
if (!dst)
return -ENOMEM;
if (copy_from_user(dst, src, len)) {
kfree(dst);
return -EFAULT
}
memdup_user() is a wrapper of the above code. With this new function, we
don't have to write 'len' twice, which can lead to typos/mistakes. It
also produces smaller code and kernel text.
A quick grep shows 250+ places where memdup_user() *may* be used. I'll
prepare a patchset to do this conversion.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Americo Wang <xiyou.wangcong@gmail.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The pagevec_swap_free() at the end of shrink_active_list() was introduced
in 68a22394 "vmscan: free swap space on swap-in/activation" when
shrink_active_list() was still rotating referenced active pages.
In 7e9cd48 "vmscan: fix pagecache reclaim referenced bit check" this was
changed, the rotating removed but the pagevec_swap_free() after the
rotation loop was forgotten, applying now to the pagevec of the
deactivation loop instead.
Now swap space is freed for deactivated pages. And only for those that
happen to be on the pagevec after the deactivation loop.
Complete 7e9cd48 and remove the rest of the swap freeing.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In shrink_active_list() after the deactivation loop, we strip buffer heads
from the potentially remaining pages in the pagevec.
Currently, this drops the zone's lru lock for stripping, only to reacquire
it again afterwards to update statistics.
It is not necessary to strip the pages before updating the stats, so move
the whole thing out of the protected region and save the extra locking.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: MinChan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
During page allocation, there are two stages of direct reclaim that are
applied to each zone in the preferred list. The first stage using
zone_reclaim() reclaims unmapped file backed pages and slab pages if over
defined limits as these are cheaper to reclaim. The caller specifies the
order of the target allocation but the scan control is not being correctly
initialised.
The impact is that the correct number of pages are being reclaimed but
that lumpy reclaim is not being applied. This increases the chances of a
full direct reclaim via try_to_free_pages() is required.
This patch initialises the order field of the scan control as requested by
the caller.
[mel@csn.ul.ie: rewrote changelog]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
At first look, mark_page_accessed() in follow_page() seems a bit strange.
It seems pte_mkyoung() would be better consistent with other kernel code.
However, it is intentional. The commit log said:
------------------------------------------------
commit 9e45f61d69be9024a2e6bef3831fb04d90fac7a8
Author: akpm <akpm>
Date: Fri Aug 15 07:24:59 2003 +0000
[PATCH] Use mark_page_accessed() in follow_page()
Touching a page via follow_page() counts as a reference so we should be
either setting the referenced bit in the pte or running mark_page_accessed().
Altering the pte is tricky because we haven't implemented an atomic
pte_mkyoung(). And mark_page_accessed() is better anyway because it has more
aging state: it can move the page onto the active list.
BKrev: 3f3c8acbplT8FbwBVGtth7QmnqWkIw
------------------------------------------------
The atomic issue is still true nowadays. adding comment help to understand
code intention and it would be better.
[akpm@linux-foundation.org: clarify text]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
shrink_inactive_list() scans in sc->swap_cluster_max chunks until it hits
the scan limit it was passed.
shrink_inactive_list()
{
do {
isolate_pages(swap_cluster_max)
shrink_page_list()
} while (nr_scanned < max_scan);
}
This assumes that swap_cluster_max is not bigger than the scan limit
because the latter is checked only after at least one iteration.
In shrink_all_memory() sc->swap_cluster_max is initialized to the overall
reclaim goal in the beginning but not decreased while reclaim is making
progress which leads to subsequent calls to shrink_inactive_list()
reclaiming way too much in the one iteration that is done unconditionally.
Set sc->swap_cluster_max always to the proper goal before doing
shrink_all_zones()
shrink_list()
shrink_inactive_list().
While the current shrink_all_memory() happily reclaims more than actually
requested, this patch fixes it to never exceed the goal:
unpatched
wanted=10000 reclaimed=13356
wanted=10000 reclaimed=19711
wanted=10000 reclaimed=10289
wanted=10000 reclaimed=17306
wanted=10000 reclaimed=10700
wanted=10000 reclaimed=10004
wanted=10000 reclaimed=13301
wanted=10000 reclaimed=10976
wanted=10000 reclaimed=10605
wanted=10000 reclaimed=10088
wanted=10000 reclaimed=15000
patched
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=9599
wanted=10000 reclaimed=8476
wanted=10000 reclaimed=8326
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=9919
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=9624
wanted=10000 reclaimed=10000
wanted=10000 reclaimed=10000
wanted=8500 reclaimed=8092
wanted=316 reclaimed=316
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: MinChan Kim <minchan.kim@gmail.com>
Acked-by: Nigel Cunningham <ncunningham@crca.org.au>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit a79311c14e "vmscan: bail out of
direct reclaim after swap_cluster_max pages" moved the nr_reclaimed
counter into the scan control to accumulate the number of all reclaimed
pages in a reclaim invocation.
shrink_all_memory() can use the same mechanism. it increase code
consistency and redability.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: MinChan Kim <minchan.kim@gmail.com>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Impact: cleanup
In almost cases, for_each_zone() is used with populated_zone(). It's
because almost function doesn't need memoryless node information.
Therefore, for_each_populated_zone() can help to make code simplify.
This patch has no functional change.
[akpm@linux-foundation.org: small cleanup]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
sc.may_swap does not only influence reclaiming of anon pages but pages
mapped into pagetables in general, which also includes mapped file pages.
In shrink_page_list():
if (!sc->may_swap && page_mapped(page))
goto keep_locked;
For anon pages, this makes sense as they are always mapped and reclaiming
them always requires swapping.
But mapped file pages are skipped here as well and it has nothing to do
with swapping.
The real effect of the knob is whether mapped pages are unmapped and
reclaimed or not. Rename it to `may_unmap' to have its name match its
actual meaning more precisely.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: MinChan Kim <minchan.kim@gmail.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
vmap's dirty_list is unused. It's for optimizing flushing. but Nick
didn't write the code yet. so, we don't need it until time as it is
needed.
This patch removes vmap_block's dirty_list and codes related to it.
Signed-off-by: MinChan Kim <minchan.kim@gmail.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ingo Molnar
Hugh Dickins
KAMEZAWA Hiroyuki
David Rientjes
David Howells
Akinobu Mita
Nick Piggin
Alexey Dobriyan
Li Zefan
Roel Kluin
KOSAKI Motohiro
Johannes Weiner
Edward Shishkin
MinChan Kim
Cyrill Gorcunov