2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* linux/mm/swap_state.c
|
|
|
|
*
|
|
|
|
* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
|
|
|
|
* Swap reorganised 29.12.95, Stephen Tweedie
|
|
|
|
*
|
|
|
|
* Rewritten to use page cache, (C) 1998 Stephen Tweedie
|
|
|
|
*/
|
|
|
|
#include <linux/mm.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 15:04:11 +07:00
|
|
|
#include <linux/gfp.h>
|
2005-04-17 05:20:36 +07:00
|
|
|
#include <linux/kernel_stat.h>
|
|
|
|
#include <linux/swap.h>
|
2008-02-05 13:28:41 +07:00
|
|
|
#include <linux/swapops.h>
|
2005-04-17 05:20:36 +07:00
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/backing-dev.h>
|
swap: allow swap readahead to be merged
Swap readahead works fine, but the I/O to disk is almost always done in
page size requests, despite the fact that readahead submits
1<<page-cluster pages at a time.
On older kernels the old per device plugging behavior might have captured
this and merged the requests, but currently all comes down to much more
I/Os than required.
On a single device this might not be an issue, but as soon as a server
runs on shared san resources savin I/Os not only improves swapin
throughput but also provides a lower resource utilization.
With a load running KVM in a lot of memory overcommitment (the hot memory
is 1.5 times the host memory) swapping throughput improves significantly
and the lead feels more responsive as well as achieves more throughput.
In a test setup with 16 swap disks running blocktrace on one of those disks
shows the improved merging:
Prior:
Reads Queued: 560,888, 2,243MiB Writes Queued: 226,242, 904,968KiB
Read Dispatches: 544,701, 2,243MiB Write Dispatches: 159,318, 904,968KiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 544,716, 2,243MiB Writes Completed: 159,321, 904,980KiB
Read Merges: 16,187, 64,748KiB Write Merges: 61,744, 246,976KiB
IO unplugs: 149,614 Timer unplugs: 2,940
With the patch:
Reads Queued: 734,315, 2,937MiB Writes Queued: 300,188, 1,200MiB
Read Dispatches: 214,972, 2,937MiB Write Dispatches: 215,176, 1,200MiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 214,971, 2,937MiB Writes Completed: 215,177, 1,200MiB
Read Merges: 519,343, 2,077MiB Write Merges: 73,325, 293,300KiB
IO unplugs: 337,130 Timer unplugs: 11,184
I got ~10% to ~40% more throughput in my cases and at the same time much
lower cpu consumption when broken down per transferred kilobyte (the
majority of that due to saved interrupts and better cache handling). In a
shared SAN others might get an additional benefit as well, because this
now causes less protocol overhead.
Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-08-01 06:41:44 +07:00
|
|
|
#include <linux/blkdev.h>
|
2006-01-06 15:10:55 +07:00
|
|
|
#include <linux/pagevec.h>
|
2006-03-22 15:09:12 +07:00
|
|
|
#include <linux/migrate.h>
|
2009-01-08 09:08:00 +07:00
|
|
|
#include <linux/page_cgroup.h>
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#include <asm/pgtable.h>
|
|
|
|
|
|
|
|
/*
|
|
|
|
* swapper_space is a fiction, retained to simplify the path through
|
2011-03-10 14:52:07 +07:00
|
|
|
* vmscan's shrink_page_list.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2006-06-28 18:26:44 +07:00
|
|
|
static const struct address_space_operations swap_aops = {
|
2005-04-17 05:20:36 +07:00
|
|
|
.writepage = swap_writepage,
|
2012-08-01 06:44:55 +07:00
|
|
|
.set_page_dirty = swap_set_page_dirty,
|
2006-02-01 18:05:41 +07:00
|
|
|
.migratepage = migrate_page,
|
2005-04-17 05:20:36 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
static struct backing_dev_info swap_backing_dev_info = {
|
2009-06-12 19:45:52 +07:00
|
|
|
.name = "swap",
|
2008-10-19 10:26:32 +07:00
|
|
|
.capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
|
2005-04-17 05:20:36 +07:00
|
|
|
};
|
|
|
|
|
2013-02-23 07:34:37 +07:00
|
|
|
struct address_space swapper_spaces[MAX_SWAPFILES] = {
|
|
|
|
[0 ... MAX_SWAPFILES - 1] = {
|
|
|
|
.page_tree = RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
|
|
|
|
.a_ops = &swap_aops,
|
|
|
|
.backing_dev_info = &swap_backing_dev_info,
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
|
|
|
|
|
|
|
|
static struct {
|
|
|
|
unsigned long add_total;
|
|
|
|
unsigned long del_total;
|
|
|
|
unsigned long find_success;
|
|
|
|
unsigned long find_total;
|
|
|
|
} swap_cache_info;
|
|
|
|
|
2013-02-23 07:34:37 +07:00
|
|
|
unsigned long total_swapcache_pages(void)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
unsigned long ret = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_SWAPFILES; i++)
|
|
|
|
ret += swapper_spaces[i].nrpages;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
static atomic_t swapin_readahead_hits = ATOMIC_INIT(4);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
void show_swap_cache_info(void)
|
|
|
|
{
|
2013-02-23 07:34:37 +07:00
|
|
|
printk("%lu pages in swap cache\n", total_swapcache_pages());
|
2008-07-26 09:46:01 +07:00
|
|
|
printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
|
2005-04-17 05:20:36 +07:00
|
|
|
swap_cache_info.add_total, swap_cache_info.del_total,
|
2008-02-05 13:28:49 +07:00
|
|
|
swap_cache_info.find_success, swap_cache_info.find_total);
|
swap: add per-partition lock for swapfile
swap_lock is heavily contended when I test swap to 3 fast SSD (even
slightly slower than swap to 2 such SSD). The main contention comes
from swap_info_get(). This patch tries to fix the gap with adding a new
per-partition lock.
Global data like nr_swapfiles, total_swap_pages, least_priority and
swap_list are still protected by swap_lock.
nr_swap_pages is an atomic now, it can be changed without swap_lock. In
theory, it's possible get_swap_page() finds no swap pages but actually
there are free swap pages. But sounds not a big problem.
Accessing partition specific data (like scan_swap_map and so on) is only
protected by swap_info_struct.lock.
Changing swap_info_struct.flags need hold swap_lock and
swap_info_struct.lock, because scan_scan_map() will check it. read the
flags is ok with either the locks hold.
If both swap_lock and swap_info_struct.lock must be hold, we always hold
the former first to avoid deadlock.
swap_entry_free() can change swap_list. To delete that code, we add a
new highest_priority_index. Whenever get_swap_page() is called, we
check it. If it's valid, we use it.
It's a pity get_swap_page() still holds swap_lock(). But in practice,
swap_lock() isn't heavily contended in my test with this patch (or I can
say there are other much more heavier bottlenecks like TLB flush). And
BTW, looks get_swap_page() doesn't really need the lock. We never free
swap_info[] and we check SWAP_WRITEOK flag. The only risk without the
lock is we could swapout to some low priority swap, but we can quickly
recover after several rounds of swap, so sounds not a big deal to me.
But I'd prefer to fix this if it's a real problem.
"swap: make each swap partition have one address_space" improved the
swapout speed from 1.7G/s to 2G/s. This patch further improves the
speed to 2.3G/s, so around 15% improvement. It's a multi-process test,
so TLB flush isn't the biggest bottleneck before the patches.
[arnd@arndb.de: fix it for nommu]
[hughd@google.com: add missing unlock]
[minchan@kernel.org: get rid of lockdep whinge on sys_swapon]
Signed-off-by: Shaohua Li <shli@fusionio.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Seth Jennings <sjenning@linux.vnet.ibm.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Dan Magenheimer <dan.magenheimer@oracle.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-23 07:34:38 +07:00
|
|
|
printk("Free swap = %ldkB\n",
|
|
|
|
get_nr_swap_pages() << (PAGE_SHIFT - 10));
|
2005-04-17 05:20:36 +07:00
|
|
|
printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-09-22 07:02:50 +07:00
|
|
|
* __add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
|
2005-04-17 05:20:36 +07:00
|
|
|
* but sets SwapCache flag and private instead of mapping and index.
|
|
|
|
*/
|
2013-04-30 05:08:34 +07:00
|
|
|
int __add_to_swap_cache(struct page *page, swp_entry_t entry)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
int error;
|
2013-02-23 07:34:37 +07:00
|
|
|
struct address_space *address_space;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2014-01-24 06:52:54 +07:00
|
|
|
VM_BUG_ON_PAGE(!PageLocked(page), page);
|
|
|
|
VM_BUG_ON_PAGE(PageSwapCache(page), page);
|
|
|
|
VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
|
2009-01-07 05:39:25 +07:00
|
|
|
|
2009-09-22 07:02:50 +07:00
|
|
|
page_cache_get(page);
|
|
|
|
SetPageSwapCache(page);
|
|
|
|
set_page_private(page, entry.val);
|
|
|
|
|
2013-02-23 07:34:37 +07:00
|
|
|
address_space = swap_address_space(entry);
|
|
|
|
spin_lock_irq(&address_space->tree_lock);
|
|
|
|
error = radix_tree_insert(&address_space->page_tree,
|
|
|
|
entry.val, page);
|
2009-09-22 07:02:50 +07:00
|
|
|
if (likely(!error)) {
|
2013-02-23 07:34:37 +07:00
|
|
|
address_space->nrpages++;
|
2009-09-22 07:02:50 +07:00
|
|
|
__inc_zone_page_state(page, NR_FILE_PAGES);
|
|
|
|
INC_CACHE_INFO(add_total);
|
|
|
|
}
|
2013-02-23 07:34:37 +07:00
|
|
|
spin_unlock_irq(&address_space->tree_lock);
|
2009-09-22 07:02:50 +07:00
|
|
|
|
|
|
|
if (unlikely(error)) {
|
2009-09-22 07:02:52 +07:00
|
|
|
/*
|
|
|
|
* Only the context which have set SWAP_HAS_CACHE flag
|
|
|
|
* would call add_to_swap_cache().
|
|
|
|
* So add_to_swap_cache() doesn't returns -EEXIST.
|
|
|
|
*/
|
|
|
|
VM_BUG_ON(error == -EEXIST);
|
2009-09-22 07:02:50 +07:00
|
|
|
set_page_private(page, 0UL);
|
|
|
|
ClearPageSwapCache(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
2013-09-12 04:26:05 +07:00
|
|
|
error = radix_tree_maybe_preload(gfp_mask);
|
2008-02-07 15:14:05 +07:00
|
|
|
if (!error) {
|
2009-09-22 07:02:50 +07:00
|
|
|
error = __add_to_swap_cache(page, entry);
|
2005-04-17 05:20:36 +07:00
|
|
|
radix_tree_preload_end();
|
2008-02-07 15:14:13 +07:00
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This must be called only on pages that have
|
|
|
|
* been verified to be in the swap cache.
|
|
|
|
*/
|
|
|
|
void __delete_from_swap_cache(struct page *page)
|
|
|
|
{
|
2013-02-23 07:34:37 +07:00
|
|
|
swp_entry_t entry;
|
|
|
|
struct address_space *address_space;
|
|
|
|
|
2014-01-24 06:52:54 +07:00
|
|
|
VM_BUG_ON_PAGE(!PageLocked(page), page);
|
|
|
|
VM_BUG_ON_PAGE(!PageSwapCache(page), page);
|
|
|
|
VM_BUG_ON_PAGE(PageWriteback(page), page);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2013-02-23 07:34:37 +07:00
|
|
|
entry.val = page_private(page);
|
|
|
|
address_space = swap_address_space(entry);
|
|
|
|
radix_tree_delete(&address_space->page_tree, page_private(page));
|
[PATCH] mm: split page table lock
Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with
a many-threaded application which concurrently initializes different parts of
a large anonymous area.
This patch corrects that, by using a separate spinlock per page table page, to
guard the page table entries in that page, instead of using the mm's single
page_table_lock. (But even then, page_table_lock is still used to guard page
table allocation, and anon_vma allocation.)
In this implementation, the spinlock is tucked inside the struct page of the
page table page: with a BUILD_BUG_ON in case it overflows - which it would in
the case of 32-bit PA-RISC with spinlock debugging enabled.
Splitting the lock is not quite for free: another cacheline access. Ideally,
I suppose we would use split ptlock only for multi-threaded processes on
multi-cpu machines; but deciding that dynamically would have its own costs.
So for now enable it by config, at some number of cpus - since the Kconfig
language doesn't support inequalities, let preprocessor compare that with
NR_CPUS. But I don't think it's worth being user-configurable: for good
testing of both split and unsplit configs, split now at 4 cpus, and perhaps
change that to 8 later.
There is a benefit even for singly threaded processes: kswapd can be attacking
one part of the mm while another part is busy faulting.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 08:16:40 +07:00
|
|
|
set_page_private(page, 0);
|
2005-04-17 05:20:36 +07:00
|
|
|
ClearPageSwapCache(page);
|
2013-02-23 07:34:37 +07:00
|
|
|
address_space->nrpages--;
|
2006-06-30 15:55:35 +07:00
|
|
|
__dec_zone_page_state(page, NR_FILE_PAGES);
|
2005-04-17 05:20:36 +07:00
|
|
|
INC_CACHE_INFO(del_total);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* add_to_swap - allocate swap space for a page
|
|
|
|
* @page: page we want to move to swap
|
|
|
|
*
|
|
|
|
* Allocate swap space for the page and add the page to the
|
|
|
|
* swap cache. Caller needs to hold the page lock.
|
|
|
|
*/
|
2013-04-30 05:08:36 +07:00
|
|
|
int add_to_swap(struct page *page, struct list_head *list)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
|
|
|
swp_entry_t entry;
|
|
|
|
int err;
|
|
|
|
|
2014-01-24 06:52:54 +07:00
|
|
|
VM_BUG_ON_PAGE(!PageLocked(page), page);
|
|
|
|
VM_BUG_ON_PAGE(!PageUptodate(page), page);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2009-09-22 07:02:52 +07:00
|
|
|
entry = get_swap_page();
|
|
|
|
if (!entry.val)
|
|
|
|
return 0;
|
|
|
|
|
2011-01-14 06:46:47 +07:00
|
|
|
if (unlikely(PageTransHuge(page)))
|
2013-04-30 05:08:36 +07:00
|
|
|
if (unlikely(split_huge_page_to_list(page, list))) {
|
2011-01-14 06:46:47 +07:00
|
|
|
swapcache_free(entry, NULL);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-22 07:02:52 +07:00
|
|
|
/*
|
|
|
|
* Radix-tree node allocations from PF_MEMALLOC contexts could
|
|
|
|
* completely exhaust the page allocator. __GFP_NOMEMALLOC
|
|
|
|
* stops emergency reserves from being allocated.
|
|
|
|
*
|
|
|
|
* TODO: this could cause a theoretical memory reclaim
|
|
|
|
* deadlock in the swap out path.
|
|
|
|
*/
|
|
|
|
/*
|
|
|
|
* Add it to the swap cache and mark it dirty
|
|
|
|
*/
|
|
|
|
err = add_to_swap_cache(page, entry,
|
|
|
|
__GFP_HIGH|__GFP_NOMEMALLOC|__GFP_NOWARN);
|
|
|
|
|
|
|
|
if (!err) { /* Success */
|
|
|
|
SetPageDirty(page);
|
|
|
|
return 1;
|
|
|
|
} else { /* -ENOMEM radix-tree allocation failure */
|
2005-05-01 22:58:37 +07:00
|
|
|
/*
|
2009-09-22 07:02:52 +07:00
|
|
|
* add_to_swap_cache() doesn't return -EEXIST, so we can safely
|
|
|
|
* clear SWAP_HAS_CACHE flag.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2009-09-22 07:02:52 +07:00
|
|
|
swapcache_free(entry, NULL);
|
|
|
|
return 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This must be called only on pages that have
|
|
|
|
* been verified to be in the swap cache and locked.
|
|
|
|
* It will never put the page into the free list,
|
|
|
|
* the caller has a reference on the page.
|
|
|
|
*/
|
|
|
|
void delete_from_swap_cache(struct page *page)
|
|
|
|
{
|
|
|
|
swp_entry_t entry;
|
2013-02-23 07:34:37 +07:00
|
|
|
struct address_space *address_space;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
[PATCH] mm: split page table lock
Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with
a many-threaded application which concurrently initializes different parts of
a large anonymous area.
This patch corrects that, by using a separate spinlock per page table page, to
guard the page table entries in that page, instead of using the mm's single
page_table_lock. (But even then, page_table_lock is still used to guard page
table allocation, and anon_vma allocation.)
In this implementation, the spinlock is tucked inside the struct page of the
page table page: with a BUILD_BUG_ON in case it overflows - which it would in
the case of 32-bit PA-RISC with spinlock debugging enabled.
Splitting the lock is not quite for free: another cacheline access. Ideally,
I suppose we would use split ptlock only for multi-threaded processes on
multi-cpu machines; but deciding that dynamically would have its own costs.
So for now enable it by config, at some number of cpus - since the Kconfig
language doesn't support inequalities, let preprocessor compare that with
NR_CPUS. But I don't think it's worth being user-configurable: for good
testing of both split and unsplit configs, split now at 4 cpus, and perhaps
change that to 8 later.
There is a benefit even for singly threaded processes: kswapd can be attacking
one part of the mm while another part is busy faulting.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 08:16:40 +07:00
|
|
|
entry.val = page_private(page);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2013-02-23 07:34:37 +07:00
|
|
|
address_space = swap_address_space(entry);
|
|
|
|
spin_lock_irq(&address_space->tree_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
__delete_from_swap_cache(page);
|
2013-02-23 07:34:37 +07:00
|
|
|
spin_unlock_irq(&address_space->tree_lock);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2009-06-17 05:32:52 +07:00
|
|
|
swapcache_free(entry, page);
|
2005-04-17 05:20:36 +07:00
|
|
|
page_cache_release(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we are the only user, then try to free up the swap cache.
|
|
|
|
*
|
|
|
|
* Its ok to check for PageSwapCache without the page lock
|
2009-01-07 05:39:36 +07:00
|
|
|
* here because we are going to recheck again inside
|
|
|
|
* try_to_free_swap() _with_ the lock.
|
2005-04-17 05:20:36 +07:00
|
|
|
* - Marcelo
|
|
|
|
*/
|
|
|
|
static inline void free_swap_cache(struct page *page)
|
|
|
|
{
|
2009-01-07 05:39:36 +07:00
|
|
|
if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
|
|
|
|
try_to_free_swap(page);
|
2005-04-17 05:20:36 +07:00
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Perform a free_page(), also freeing any swap cache associated with
|
2005-10-30 08:16:41 +07:00
|
|
|
* this page if it is the last user of the page.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
void free_page_and_swap_cache(struct page *page)
|
|
|
|
{
|
|
|
|
free_swap_cache(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Passed an array of pages, drop them all from swapcache and then release
|
|
|
|
* them. They are removed from the LRU and freed if this is their last use.
|
|
|
|
*/
|
|
|
|
void free_pages_and_swap_cache(struct page **pages, int nr)
|
|
|
|
{
|
|
|
|
struct page **pagep = pages;
|
|
|
|
|
|
|
|
lru_add_drain();
|
|
|
|
while (nr) {
|
2006-01-06 15:10:55 +07:00
|
|
|
int todo = min(nr, PAGEVEC_SIZE);
|
2005-04-17 05:20:36 +07:00
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < todo; i++)
|
|
|
|
free_swap_cache(pagep[i]);
|
|
|
|
release_pages(pagep, todo, 0);
|
|
|
|
pagep += todo;
|
|
|
|
nr -= todo;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup a swap entry in the swap cache. A found page will be returned
|
|
|
|
* unlocked and with its refcount incremented - we rely on the kernel
|
|
|
|
* lock getting page table operations atomic even if we drop the page
|
|
|
|
* lock before returning.
|
|
|
|
*/
|
|
|
|
struct page * lookup_swap_cache(swp_entry_t entry)
|
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
|
2013-02-23 07:34:37 +07:00
|
|
|
page = find_get_page(swap_address_space(entry), entry.val);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
if (page) {
|
2005-04-17 05:20:36 +07:00
|
|
|
INC_CACHE_INFO(find_success);
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
if (TestClearPageReadahead(page))
|
|
|
|
atomic_inc(&swapin_readahead_hits);
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
INC_CACHE_INFO(find_total);
|
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Locate a page of swap in physical memory, reserving swap cache space
|
|
|
|
* and reading the disk if it is not already cached.
|
|
|
|
* A failure return means that either the page allocation failed or that
|
|
|
|
* the swap entry is no longer in use.
|
|
|
|
*/
|
swapin needs gfp_mask for loop on tmpfs
Building in a filesystem on a loop device on a tmpfs file can hang when
swapping, the loop thread caught in that infamous throttle_vm_writeout.
In theory this is a long standing problem, which I've either never seen in
practice, or long ago suppressed the recollection, after discounting my load
and my tmpfs size as unrealistically high. But now, with the new aops, it has
become easy to hang on one machine.
Loop used to grab_cache_page before the old prepare_write to tmpfs, which
seems to have been enough to free up some memory for any swapin needed; but
the new write_begin lets tmpfs find or allocate the page (much nicer, since
grab_cache_page missed tmpfs pages in swapcache).
When allocating a fresh page, tmpfs respects loop's mapping_gfp_mask, which
has __GFP_IO|__GFP_FS stripped off, and throttle_vm_writeout is designed to
break out when __GFP_IO or GFP_FS is unset; but when tmfps swaps in,
read_swap_cache_async allocates with GFP_HIGHUSER_MOVABLE regardless of the
mapping_gfp_mask - hence the hang.
So, pass gfp_mask down the line from shmem_getpage to shmem_swapin to
swapin_readahead to read_swap_cache_async to add_to_swap_cache.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-05 13:28:42 +07:00
|
|
|
struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
|
2005-04-17 05:20:36 +07:00
|
|
|
struct vm_area_struct *vma, unsigned long addr)
|
|
|
|
{
|
|
|
|
struct page *found_page, *new_page = NULL;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
do {
|
|
|
|
/*
|
|
|
|
* First check the swap cache. Since this is normally
|
|
|
|
* called after lookup_swap_cache() failed, re-calling
|
|
|
|
* that would confuse statistics.
|
|
|
|
*/
|
2013-02-23 07:34:37 +07:00
|
|
|
found_page = find_get_page(swap_address_space(entry),
|
|
|
|
entry.val);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (found_page)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get a new page to read into from swap.
|
|
|
|
*/
|
|
|
|
if (!new_page) {
|
swapin needs gfp_mask for loop on tmpfs
Building in a filesystem on a loop device on a tmpfs file can hang when
swapping, the loop thread caught in that infamous throttle_vm_writeout.
In theory this is a long standing problem, which I've either never seen in
practice, or long ago suppressed the recollection, after discounting my load
and my tmpfs size as unrealistically high. But now, with the new aops, it has
become easy to hang on one machine.
Loop used to grab_cache_page before the old prepare_write to tmpfs, which
seems to have been enough to free up some memory for any swapin needed; but
the new write_begin lets tmpfs find or allocate the page (much nicer, since
grab_cache_page missed tmpfs pages in swapcache).
When allocating a fresh page, tmpfs respects loop's mapping_gfp_mask, which
has __GFP_IO|__GFP_FS stripped off, and throttle_vm_writeout is designed to
break out when __GFP_IO or GFP_FS is unset; but when tmfps swaps in,
read_swap_cache_async allocates with GFP_HIGHUSER_MOVABLE regardless of the
mapping_gfp_mask - hence the hang.
So, pass gfp_mask down the line from shmem_getpage to shmem_swapin to
swapin_readahead to read_swap_cache_async to add_to_swap_cache.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-05 13:28:42 +07:00
|
|
|
new_page = alloc_page_vma(gfp_mask, vma, addr);
|
2005-04-17 05:20:36 +07:00
|
|
|
if (!new_page)
|
|
|
|
break; /* Out of memory */
|
|
|
|
}
|
|
|
|
|
2009-09-22 07:02:50 +07:00
|
|
|
/*
|
|
|
|
* call radix_tree_preload() while we can wait.
|
|
|
|
*/
|
2013-09-12 04:26:05 +07:00
|
|
|
err = radix_tree_maybe_preload(gfp_mask & GFP_KERNEL);
|
2009-09-22 07:02:50 +07:00
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
|
2008-02-05 13:28:49 +07:00
|
|
|
/*
|
|
|
|
* Swap entry may have been freed since our caller observed it.
|
|
|
|
*/
|
2009-06-17 05:32:53 +07:00
|
|
|
err = swapcache_prepare(entry);
|
2013-06-13 04:04:49 +07:00
|
|
|
if (err == -EEXIST) {
|
2009-09-22 07:02:50 +07:00
|
|
|
radix_tree_preload_end();
|
2013-06-13 04:04:49 +07:00
|
|
|
/*
|
|
|
|
* We might race against get_swap_page() and stumble
|
|
|
|
* across a SWAP_HAS_CACHE swap_map entry whose page
|
|
|
|
* has not been brought into the swapcache yet, while
|
|
|
|
* the other end is scheduled away waiting on discard
|
|
|
|
* I/O completion at scan_swap_map().
|
|
|
|
*
|
|
|
|
* In order to avoid turning this transitory state
|
|
|
|
* into a permanent loop around this -EEXIST case
|
|
|
|
* if !CONFIG_PREEMPT and the I/O completion happens
|
|
|
|
* to be waiting on the CPU waitqueue where we are now
|
|
|
|
* busy looping, we just conditionally invoke the
|
|
|
|
* scheduler here, if there are some more important
|
|
|
|
* tasks to run.
|
|
|
|
*/
|
|
|
|
cond_resched();
|
2009-06-17 05:32:53 +07:00
|
|
|
continue;
|
2009-09-22 07:02:50 +07:00
|
|
|
}
|
|
|
|
if (err) { /* swp entry is obsolete ? */
|
|
|
|
radix_tree_preload_end();
|
2008-02-05 13:28:49 +07:00
|
|
|
break;
|
2009-09-22 07:02:50 +07:00
|
|
|
}
|
2008-02-05 13:28:49 +07:00
|
|
|
|
2009-09-22 07:02:52 +07:00
|
|
|
/* May fail (-ENOMEM) if radix-tree node allocation failed. */
|
2008-10-19 10:26:57 +07:00
|
|
|
__set_page_locked(new_page);
|
2008-10-19 10:26:30 +07:00
|
|
|
SetPageSwapBacked(new_page);
|
2009-09-22 07:02:50 +07:00
|
|
|
err = __add_to_swap_cache(new_page, entry);
|
2008-08-02 17:01:03 +07:00
|
|
|
if (likely(!err)) {
|
2009-09-22 07:02:50 +07:00
|
|
|
radix_tree_preload_end();
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Initiate read into locked page and return.
|
|
|
|
*/
|
2008-10-19 10:26:36 +07:00
|
|
|
lru_cache_add_anon(new_page);
|
2009-06-17 05:33:02 +07:00
|
|
|
swap_readpage(new_page);
|
2005-04-17 05:20:36 +07:00
|
|
|
return new_page;
|
|
|
|
}
|
2009-09-22 07:02:50 +07:00
|
|
|
radix_tree_preload_end();
|
2008-10-19 10:26:30 +07:00
|
|
|
ClearPageSwapBacked(new_page);
|
2008-10-19 10:26:57 +07:00
|
|
|
__clear_page_locked(new_page);
|
2009-09-22 07:02:52 +07:00
|
|
|
/*
|
|
|
|
* add_to_swap_cache() doesn't return -EEXIST, so we can safely
|
|
|
|
* clear SWAP_HAS_CACHE flag.
|
|
|
|
*/
|
2009-06-17 05:32:52 +07:00
|
|
|
swapcache_free(entry, NULL);
|
2008-02-05 13:28:49 +07:00
|
|
|
} while (err != -ENOMEM);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
if (new_page)
|
|
|
|
page_cache_release(new_page);
|
|
|
|
return found_page;
|
|
|
|
}
|
2008-02-05 13:28:41 +07:00
|
|
|
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
static unsigned long swapin_nr_pages(unsigned long offset)
|
|
|
|
{
|
|
|
|
static unsigned long prev_offset;
|
|
|
|
unsigned int pages, max_pages, last_ra;
|
|
|
|
static atomic_t last_readahead_pages;
|
|
|
|
|
|
|
|
max_pages = 1 << ACCESS_ONCE(page_cluster);
|
|
|
|
if (max_pages <= 1)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This heuristic has been found to work well on both sequential and
|
|
|
|
* random loads, swapping to hard disk or to SSD: please don't ask
|
|
|
|
* what the "+ 2" means, it just happens to work well, that's all.
|
|
|
|
*/
|
|
|
|
pages = atomic_xchg(&swapin_readahead_hits, 0) + 2;
|
|
|
|
if (pages == 2) {
|
|
|
|
/*
|
|
|
|
* We can have no readahead hits to judge by: but must not get
|
|
|
|
* stuck here forever, so check for an adjacent offset instead
|
|
|
|
* (and don't even bother to check whether swap type is same).
|
|
|
|
*/
|
|
|
|
if (offset != prev_offset + 1 && offset != prev_offset - 1)
|
|
|
|
pages = 1;
|
|
|
|
prev_offset = offset;
|
|
|
|
} else {
|
|
|
|
unsigned int roundup = 4;
|
|
|
|
while (roundup < pages)
|
|
|
|
roundup <<= 1;
|
|
|
|
pages = roundup;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (pages > max_pages)
|
|
|
|
pages = max_pages;
|
|
|
|
|
|
|
|
/* Don't shrink readahead too fast */
|
|
|
|
last_ra = atomic_read(&last_readahead_pages) / 2;
|
|
|
|
if (pages < last_ra)
|
|
|
|
pages = last_ra;
|
|
|
|
atomic_set(&last_readahead_pages, pages);
|
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
2008-02-05 13:28:41 +07:00
|
|
|
/**
|
|
|
|
* swapin_readahead - swap in pages in hope we need them soon
|
|
|
|
* @entry: swap entry of this memory
|
2008-03-20 07:00:40 +07:00
|
|
|
* @gfp_mask: memory allocation flags
|
2008-02-05 13:28:41 +07:00
|
|
|
* @vma: user vma this address belongs to
|
|
|
|
* @addr: target address for mempolicy
|
|
|
|
*
|
|
|
|
* Returns the struct page for entry and addr, after queueing swapin.
|
|
|
|
*
|
|
|
|
* Primitive swap readahead code. We simply read an aligned block of
|
|
|
|
* (1 << page_cluster) entries in the swap area. This method is chosen
|
|
|
|
* because it doesn't cost us any seek time. We also make sure to queue
|
|
|
|
* the 'original' request together with the readahead ones...
|
|
|
|
*
|
|
|
|
* This has been extended to use the NUMA policies from the mm triggering
|
|
|
|
* the readahead.
|
|
|
|
*
|
|
|
|
* Caller must hold down_read on the vma->vm_mm if vma is not NULL.
|
|
|
|
*/
|
swapin needs gfp_mask for loop on tmpfs
Building in a filesystem on a loop device on a tmpfs file can hang when
swapping, the loop thread caught in that infamous throttle_vm_writeout.
In theory this is a long standing problem, which I've either never seen in
practice, or long ago suppressed the recollection, after discounting my load
and my tmpfs size as unrealistically high. But now, with the new aops, it has
become easy to hang on one machine.
Loop used to grab_cache_page before the old prepare_write to tmpfs, which
seems to have been enough to free up some memory for any swapin needed; but
the new write_begin lets tmpfs find or allocate the page (much nicer, since
grab_cache_page missed tmpfs pages in swapcache).
When allocating a fresh page, tmpfs respects loop's mapping_gfp_mask, which
has __GFP_IO|__GFP_FS stripped off, and throttle_vm_writeout is designed to
break out when __GFP_IO or GFP_FS is unset; but when tmfps swaps in,
read_swap_cache_async allocates with GFP_HIGHUSER_MOVABLE regardless of the
mapping_gfp_mask - hence the hang.
So, pass gfp_mask down the line from shmem_getpage to shmem_swapin to
swapin_readahead to read_swap_cache_async to add_to_swap_cache.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-05 13:28:42 +07:00
|
|
|
struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
|
2008-02-05 13:28:41 +07:00
|
|
|
struct vm_area_struct *vma, unsigned long addr)
|
|
|
|
{
|
|
|
|
struct page *page;
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
unsigned long entry_offset = swp_offset(entry);
|
|
|
|
unsigned long offset = entry_offset;
|
2012-03-22 06:33:50 +07:00
|
|
|
unsigned long start_offset, end_offset;
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
unsigned long mask;
|
swap: allow swap readahead to be merged
Swap readahead works fine, but the I/O to disk is almost always done in
page size requests, despite the fact that readahead submits
1<<page-cluster pages at a time.
On older kernels the old per device plugging behavior might have captured
this and merged the requests, but currently all comes down to much more
I/Os than required.
On a single device this might not be an issue, but as soon as a server
runs on shared san resources savin I/Os not only improves swapin
throughput but also provides a lower resource utilization.
With a load running KVM in a lot of memory overcommitment (the hot memory
is 1.5 times the host memory) swapping throughput improves significantly
and the lead feels more responsive as well as achieves more throughput.
In a test setup with 16 swap disks running blocktrace on one of those disks
shows the improved merging:
Prior:
Reads Queued: 560,888, 2,243MiB Writes Queued: 226,242, 904,968KiB
Read Dispatches: 544,701, 2,243MiB Write Dispatches: 159,318, 904,968KiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 544,716, 2,243MiB Writes Completed: 159,321, 904,980KiB
Read Merges: 16,187, 64,748KiB Write Merges: 61,744, 246,976KiB
IO unplugs: 149,614 Timer unplugs: 2,940
With the patch:
Reads Queued: 734,315, 2,937MiB Writes Queued: 300,188, 1,200MiB
Read Dispatches: 214,972, 2,937MiB Write Dispatches: 215,176, 1,200MiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 214,971, 2,937MiB Writes Completed: 215,177, 1,200MiB
Read Merges: 519,343, 2,077MiB Write Merges: 73,325, 293,300KiB
IO unplugs: 337,130 Timer unplugs: 11,184
I got ~10% to ~40% more throughput in my cases and at the same time much
lower cpu consumption when broken down per transferred kilobyte (the
majority of that due to saved interrupts and better cache handling). In a
shared SAN others might get an additional benefit as well, because this
now causes less protocol overhead.
Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-08-01 06:41:44 +07:00
|
|
|
struct blk_plug plug;
|
2008-02-05 13:28:41 +07:00
|
|
|
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
mask = swapin_nr_pages(offset) - 1;
|
|
|
|
if (!mask)
|
|
|
|
goto skip;
|
|
|
|
|
2012-03-22 06:33:50 +07:00
|
|
|
/* Read a page_cluster sized and aligned cluster around offset. */
|
|
|
|
start_offset = offset & ~mask;
|
|
|
|
end_offset = offset | mask;
|
|
|
|
if (!start_offset) /* First page is swap header. */
|
|
|
|
start_offset++;
|
|
|
|
|
swap: allow swap readahead to be merged
Swap readahead works fine, but the I/O to disk is almost always done in
page size requests, despite the fact that readahead submits
1<<page-cluster pages at a time.
On older kernels the old per device plugging behavior might have captured
this and merged the requests, but currently all comes down to much more
I/Os than required.
On a single device this might not be an issue, but as soon as a server
runs on shared san resources savin I/Os not only improves swapin
throughput but also provides a lower resource utilization.
With a load running KVM in a lot of memory overcommitment (the hot memory
is 1.5 times the host memory) swapping throughput improves significantly
and the lead feels more responsive as well as achieves more throughput.
In a test setup with 16 swap disks running blocktrace on one of those disks
shows the improved merging:
Prior:
Reads Queued: 560,888, 2,243MiB Writes Queued: 226,242, 904,968KiB
Read Dispatches: 544,701, 2,243MiB Write Dispatches: 159,318, 904,968KiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 544,716, 2,243MiB Writes Completed: 159,321, 904,980KiB
Read Merges: 16,187, 64,748KiB Write Merges: 61,744, 246,976KiB
IO unplugs: 149,614 Timer unplugs: 2,940
With the patch:
Reads Queued: 734,315, 2,937MiB Writes Queued: 300,188, 1,200MiB
Read Dispatches: 214,972, 2,937MiB Write Dispatches: 215,176, 1,200MiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 214,971, 2,937MiB Writes Completed: 215,177, 1,200MiB
Read Merges: 519,343, 2,077MiB Write Merges: 73,325, 293,300KiB
IO unplugs: 337,130 Timer unplugs: 11,184
I got ~10% to ~40% more throughput in my cases and at the same time much
lower cpu consumption when broken down per transferred kilobyte (the
majority of that due to saved interrupts and better cache handling). In a
shared SAN others might get an additional benefit as well, because this
now causes less protocol overhead.
Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-08-01 06:41:44 +07:00
|
|
|
blk_start_plug(&plug);
|
2012-03-22 06:33:50 +07:00
|
|
|
for (offset = start_offset; offset <= end_offset ; offset++) {
|
2008-02-05 13:28:41 +07:00
|
|
|
/* Ok, do the async read-ahead now */
|
|
|
|
page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
|
swapin needs gfp_mask for loop on tmpfs
Building in a filesystem on a loop device on a tmpfs file can hang when
swapping, the loop thread caught in that infamous throttle_vm_writeout.
In theory this is a long standing problem, which I've either never seen in
practice, or long ago suppressed the recollection, after discounting my load
and my tmpfs size as unrealistically high. But now, with the new aops, it has
become easy to hang on one machine.
Loop used to grab_cache_page before the old prepare_write to tmpfs, which
seems to have been enough to free up some memory for any swapin needed; but
the new write_begin lets tmpfs find or allocate the page (much nicer, since
grab_cache_page missed tmpfs pages in swapcache).
When allocating a fresh page, tmpfs respects loop's mapping_gfp_mask, which
has __GFP_IO|__GFP_FS stripped off, and throttle_vm_writeout is designed to
break out when __GFP_IO or GFP_FS is unset; but when tmfps swaps in,
read_swap_cache_async allocates with GFP_HIGHUSER_MOVABLE regardless of the
mapping_gfp_mask - hence the hang.
So, pass gfp_mask down the line from shmem_getpage to shmem_swapin to
swapin_readahead to read_swap_cache_async to add_to_swap_cache.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-05 13:28:42 +07:00
|
|
|
gfp_mask, vma, addr);
|
2008-02-05 13:28:41 +07:00
|
|
|
if (!page)
|
2012-03-22 06:33:50 +07:00
|
|
|
continue;
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
if (offset != entry_offset)
|
|
|
|
SetPageReadahead(page);
|
2008-02-05 13:28:41 +07:00
|
|
|
page_cache_release(page);
|
|
|
|
}
|
swap: allow swap readahead to be merged
Swap readahead works fine, but the I/O to disk is almost always done in
page size requests, despite the fact that readahead submits
1<<page-cluster pages at a time.
On older kernels the old per device plugging behavior might have captured
this and merged the requests, but currently all comes down to much more
I/Os than required.
On a single device this might not be an issue, but as soon as a server
runs on shared san resources savin I/Os not only improves swapin
throughput but also provides a lower resource utilization.
With a load running KVM in a lot of memory overcommitment (the hot memory
is 1.5 times the host memory) swapping throughput improves significantly
and the lead feels more responsive as well as achieves more throughput.
In a test setup with 16 swap disks running blocktrace on one of those disks
shows the improved merging:
Prior:
Reads Queued: 560,888, 2,243MiB Writes Queued: 226,242, 904,968KiB
Read Dispatches: 544,701, 2,243MiB Write Dispatches: 159,318, 904,968KiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 544,716, 2,243MiB Writes Completed: 159,321, 904,980KiB
Read Merges: 16,187, 64,748KiB Write Merges: 61,744, 246,976KiB
IO unplugs: 149,614 Timer unplugs: 2,940
With the patch:
Reads Queued: 734,315, 2,937MiB Writes Queued: 300,188, 1,200MiB
Read Dispatches: 214,972, 2,937MiB Write Dispatches: 215,176, 1,200MiB
Reads Requeued: 0 Writes Requeued: 0
Reads Completed: 214,971, 2,937MiB Writes Completed: 215,177, 1,200MiB
Read Merges: 519,343, 2,077MiB Write Merges: 73,325, 293,300KiB
IO unplugs: 337,130 Timer unplugs: 11,184
I got ~10% to ~40% more throughput in my cases and at the same time much
lower cpu consumption when broken down per transferred kilobyte (the
majority of that due to saved interrupts and better cache handling). In a
shared SAN others might get an additional benefit as well, because this
now causes less protocol overhead.
Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-08-01 06:41:44 +07:00
|
|
|
blk_finish_plug(&plug);
|
|
|
|
|
2008-02-05 13:28:41 +07:00
|
|
|
lru_add_drain(); /* Push any new pages onto the LRU now */
|
swap: add a simple detector for inappropriate swapin readahead
This is a patch to improve swap readahead algorithm. It's from Hugh and
I slightly changed it.
Hugh's original changelog:
swapin readahead does a blind readahead, whether or not the swapin is
sequential. This may be ok on harddisk, because large reads have
relatively small costs, and if the readahead pages are unneeded they can
be reclaimed easily - though, what if their allocation forced reclaim of
useful pages? But on SSD devices large reads are more expensive than
small ones: if the readahead pages are unneeded, reading them in caused
significant overhead.
This patch adds very simplistic random read detection. Stealing the
PageReadahead technique from Konstantin Khlebnikov's patch, avoiding the
vma/anon_vma sophistications of Shaohua Li's patch, swapin_nr_pages()
simply looks at readahead's current success rate, and narrows or widens
its readahead window accordingly. There is little science to its
heuristic: it's about as stupid as can be whilst remaining effective.
The table below shows elapsed times (in centiseconds) when running a
single repetitive swapping load across a 1000MB mapping in 900MB ram
with 1GB swap (the harddisk tests had taken painfully too long when I
used mem=500M, but SSD shows similar results for that).
Vanilla is the 3.6-rc7 kernel on which I started; Shaohua denotes his
Sep 3 patch in mmotm and linux-next; HughOld denotes my Oct 1 patch
which Shaohua showed to be defective; HughNew this Nov 14 patch, with
page_cluster as usual at default of 3 (8-page reads); HughPC4 this same
patch with page_cluster 4 (16-page reads); HughPC0 with page_cluster 0
(1-page reads: no readahead).
HDD for swapping to harddisk, SSD for swapping to VertexII SSD. Seq for
sequential access to the mapping, cycling five times around; Rand for
the same number of random touches. Anon for a MAP_PRIVATE anon mapping;
Shmem for a MAP_SHARED anon mapping, equivalent to tmpfs.
One weakness of Shaohua's vma/anon_vma approach was that it did not
optimize Shmem: seen below. Konstantin's approach was perhaps mistuned,
50% slower on Seq: did not compete and is not shown below.
HDD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 73921 76210 75611 76904 78191 121542
Seq Shmem 73601 73176 73855 72947 74543 118322
Rand Anon 895392 831243 871569 845197 846496 841680
Rand Shmem 1058375 1053486 827935 764955 764376 756489
SSD Vanilla Shaohua HughOld HughNew HughPC4 HughPC0
Seq Anon 24634 24198 24673 25107 21614 70018
Seq Shmem 24959 24932 25052 25703 22030 69678
Rand Anon 43014 26146 28075 25989 26935 25901
Rand Shmem 45349 45215 28249 24268 24138 24332
These tests are, of course, two extremes of a very simple case: under
heavier mixed loads I've not yet observed any consistent improvement or
degradation, and wider testing would be welcome.
Shaohua Li:
Test shows Vanilla is slightly better in sequential workload than Hugh's
patch. I observed with Hugh's patch sometimes the readahead size is
shrinked too fast (from 8 to 1 immediately) in sequential workload if
there is no hit. And in such case, continuing doing readahead is good
actually.
I don't prepare a sophisticated algorithm for the sequential workload
because so far we can't guarantee sequential accessed pages are swap out
sequentially. So I slightly change Hugh's heuristic - don't shrink
readahead size too fast.
Here is my test result (unit second, 3 runs average):
Vanilla Hugh New
Seq 356 370 360
Random 4525 2447 2444
Attached graph is the swapin/swapout throughput I collected with 'vmstat
2'. The first part is running a random workload (till around 1200 of
the x-axis) and the second part is running a sequential workload.
swapin and swapout throughput are almost identical in steady state in
both workloads. These are expected behavior. while in Vanilla, swapin
is much bigger than swapout especially in random workload (because wrong
readahead).
Original patches by: Shaohua Li and Konstantin Khlebnikov.
[fengguang.wu@intel.com: swapin_nr_pages() can be static]
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-02-07 03:04:21 +07:00
|
|
|
skip:
|
swapin needs gfp_mask for loop on tmpfs
Building in a filesystem on a loop device on a tmpfs file can hang when
swapping, the loop thread caught in that infamous throttle_vm_writeout.
In theory this is a long standing problem, which I've either never seen in
practice, or long ago suppressed the recollection, after discounting my load
and my tmpfs size as unrealistically high. But now, with the new aops, it has
become easy to hang on one machine.
Loop used to grab_cache_page before the old prepare_write to tmpfs, which
seems to have been enough to free up some memory for any swapin needed; but
the new write_begin lets tmpfs find or allocate the page (much nicer, since
grab_cache_page missed tmpfs pages in swapcache).
When allocating a fresh page, tmpfs respects loop's mapping_gfp_mask, which
has __GFP_IO|__GFP_FS stripped off, and throttle_vm_writeout is designed to
break out when __GFP_IO or GFP_FS is unset; but when tmfps swaps in,
read_swap_cache_async allocates with GFP_HIGHUSER_MOVABLE regardless of the
mapping_gfp_mask - hence the hang.
So, pass gfp_mask down the line from shmem_getpage to shmem_swapin to
swapin_readahead to read_swap_cache_async to add_to_swap_cache.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-05 13:28:42 +07:00
|
|
|
return read_swap_cache_async(entry, gfp_mask, vma, addr);
|
2008-02-05 13:28:41 +07:00
|
|
|
}
|