Age | Commit message (Collapse) | Author | Files | Lines |
|
It's not used globally and could be static.
Signed-off-by: Wanpeng Li <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Fengguang Wu <[email protected]>
Cc: Joonsoo Kim <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Yasuaki Ishimatsu <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Jiri Kosina <[email protected]>
Cc: Wanpeng Li <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
This patch is based on KOSAKI's work and I add a little more description,
please refer https://lkml.org/lkml/2012/6/14/74.
Currently, I found system can enter a state that there are lots of free
pages in a zone but only order-0 and order-1 pages which means the zone is
heavily fragmented, then high order allocation could make direct reclaim
path's long stall(ex, 60 seconds) especially in no swap and no compaciton
enviroment. This problem happened on v3.4, but it seems issue still lives
in current tree, the reason is do_try_to_free_pages enter live lock:
kswapd will go to sleep if the zones have been fully scanned and are still
not balanced. As kswapd thinks there's little point trying all over again
to avoid infinite loop. Instead it changes order from high-order to
0-order because kswapd think order-0 is the most important. Look at
73ce02e9 in detail. If watermarks are ok, kswapd will go back to sleep
and may leave zone->all_unreclaimable =3D 0. It assume high-order users
can still perform direct reclaim if they wish.
Direct reclaim continue to reclaim for a high order which is not a
COSTLY_ORDER without oom-killer until kswapd turn on
zone->all_unreclaimble= . This is because to avoid too early oom-kill.
So it means direct_reclaim depends on kswapd to break this loop.
In worst case, direct-reclaim may continue to page reclaim forever when
kswapd sleeps forever until someone like watchdog detect and finally kill
the process. As described in:
http://thread.gmane.org/gmane.linux.kernel.mm/103737
We can't turn on zone->all_unreclaimable from direct reclaim path because
direct reclaim path don't take any lock and this way is racy. Thus this
patch removes zone->all_unreclaimable field completely and recalculates
zone reclaimable state every time.
Note: we can't take the idea that direct-reclaim see zone->pages_scanned
directly and kswapd continue to use zone->all_unreclaimable. Because, it
is racy. commit 929bea7c71 (vmscan: all_unreclaimable() use
zone->all_unreclaimable as a name) describes the detail.
[[email protected]: uninline zone_reclaimable_pages() and zone_reclaimable()]
Cc: Aaditya Kumar <[email protected]>
Cc: Ying Han <[email protected]>
Cc: Nick Piggin <[email protected]>
Acked-by: Rik van Riel <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Bob Liu <[email protected]>
Cc: Neil Zhang <[email protected]>
Cc: Russell King - ARM Linux <[email protected]>
Reviewed-by: Michal Hocko <[email protected]>
Acked-by: Minchan Kim <[email protected]>
Acked-by: Johannes Weiner <[email protected]>
Signed-off-by: KOSAKI Motohiro <[email protected]>
Signed-off-by: Lisa Du <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Currently munlock_vma_pages_range() calls follow_page_mask() to obtain
each individual struct page. This entails repeated full page table
translations and page table lock taken for each page separately.
This patch avoids the costly follow_page_mask() where possible, by
iterating over ptes within single pmd under single page table lock. The
first pte is obtained by get_locked_pte() for non-THP page acquired by the
initial follow_page_mask(). The rest of the on-stack pagevec for munlock
is filled up using pte_walk as long as pte_present() and vm_normal_page()
are sufficient to obtain the struct page.
After this patch, a 14% speedup was measured for munlocking a 56GB large
memory area with THP disabled.
Signed-off-by: Vlastimil Babka <[email protected]>
Cc: Jörn Engel <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Michel Lespinasse <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Vlastimil Babka <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Pavel reported that in case if vma area get unmapped and then mapped (or
expanded) in-place, the soft dirty tracker won't be able to recognize this
situation since it works on pte level and ptes are get zapped on unmap,
loosing soft dirty bit of course.
So to resolve this situation we need to track actions on vma level, there
VM_SOFTDIRTY flag comes in. When new vma area created (or old expanded)
we set this bit, and keep it here until application calls for clearing
soft dirty bit.
Thus when user space application track memory changes now it can detect if
vma area is renewed.
Reported-by: Pavel Emelyanov <[email protected]>
Signed-off-by: Cyrill Gorcunov <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Matt Mackall <[email protected]>
Cc: Xiao Guangrong <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Stephen Rothwell <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Cc: Rob Landley <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Current early_pfn_to_nid() on arch that support memblock go over
memblock.memory one by one, so will take too many try near the end.
We can use existing memblock_search to find the node id for given pfn,
that could save some time on bigger system that have many entries
memblock.memory array.
Here are the timing differences for several machines. In each case with
the patch less time was spent in __early_pfn_to_nid().
3.11-rc5 with patch difference (%)
-------- ---------- --------------
UV1: 256 nodes 9TB: 411.66 402.47 -9.19 (2.23%)
UV2: 255 nodes 16TB: 1141.02 1138.12 -2.90 (0.25%)
UV2: 64 nodes 2TB: 128.15 126.53 -1.62 (1.26%)
UV2: 32 nodes 2TB: 121.87 121.07 -0.80 (0.66%)
Time in seconds.
Signed-off-by: Yinghai Lu <[email protected]>
Cc: Tejun Heo <[email protected]>
Acked-by: Russ Anderson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Currently hugepage migration works well only for pmd-based hugepages
(mainly due to lack of testing,) so we had better not enable migration of
other levels of hugepages until we are ready for it.
Some users of hugepage migration (mbind, move_pages, and migrate_pages) do
page table walk and check pud/pmd_huge() there, so they are safe. But the
other users (softoffline and memory hotremove) don't do this, so without
this patch they can try to migrate unexpected types of hugepages.
To prevent this, we introduce hugepage_migration_support() as an
architecture dependent check of whether hugepage are implemented on a pmd
basis or not. And on some architecture multiple sizes of hugepages are
available, so hugepage_migration_support() also checks hugepage size.
Signed-off-by: Naoya Horiguchi <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Hillf Danton <[email protected]>
Cc: Wanpeng Li <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Until now we can't offline memory blocks which contain hugepages because a
hugepage is considered as an unmovable page. But now with this patch
series, a hugepage has become movable, so by using hugepage migration we
can offline such memory blocks.
What's different from other users of hugepage migration is that we need to
decompose all the hugepages inside the target memory block into free buddy
pages after hugepage migration, because otherwise free hugepages remaining
in the memory block intervene the memory offlining. For this reason we
introduce new functions dissolve_free_huge_page() and
dissolve_free_huge_pages().
Other than that, what this patch does is straightforwardly to add hugepage
migration code, that is, adding hugepage code to the functions which scan
over pfn and collect hugepages to be migrated, and adding a hugepage
allocation function to alloc_migrate_target().
As for larger hugepages (1GB for x86_64), it's not easy to do hotremove
over them because it's larger than memory block. So we now simply leave
it to fail as it is.
[[email protected]: remove duplicated include]
Signed-off-by: Naoya Horiguchi <[email protected]>
Acked-by: Andi Kleen <[email protected]>
Cc: Hillf Danton <[email protected]>
Cc: Wanpeng Li <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Signed-off-by: Wei Yongjun <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Enable hugepage migration from migrate_pages(2), move_pages(2), and
mbind(2).
Signed-off-by: Naoya Horiguchi <[email protected]>
Acked-by: Hillf Danton <[email protected]>
Acked-by: Andi Kleen <[email protected]>
Reviewed-by: Wanpeng Li <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Extend do_mbind() to handle vma with VM_HUGETLB set. We will be able to
migrate hugepage with mbind(2) after applying the enablement patch which
comes later in this series.
Signed-off-by: Naoya Horiguchi <[email protected]>
Acked-by: Andi Kleen <[email protected]>
Reviewed-by: Wanpeng Li <[email protected]>
Acked-by: Hillf Danton <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Currently migrate_huge_page() takes a pointer to a hugepage to be migrated
as an argument, instead of taking a pointer to the list of hugepages to be
migrated. This behavior was introduced in commit 189ebff28 ("hugetlb:
simplify migrate_huge_page()"), and was OK because until now hugepage
migration is enabled only for soft-offlining which migrates only one
hugepage in a single call.
But the situation will change in the later patches in this series which
enable other users of page migration to support hugepage migration. They
can kick migration for both of normal pages and hugepages in a single
call, so we need to go back to original implementation which uses linked
lists to collect the hugepages to be migrated.
With this patch, soft_offline_huge_page() switches to use migrate_pages(),
and migrate_huge_page() is not used any more. So let's remove it.
Signed-off-by: Naoya Horiguchi <[email protected]>
Acked-by: Andi Kleen <[email protected]>
Reviewed-by: Wanpeng Li <[email protected]>
Acked-by: Hillf Danton <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Currently hugepage migration is available only for soft offlining, but
it's also useful for some other users of page migration (clearly because
users of hugepage can enjoy the benefit of mempolicy and memory hotplug.)
So this patchset tries to extend such users to support hugepage migration.
The target of this patchset is to enable hugepage migration for NUMA
related system calls (migrate_pages(2), move_pages(2), and mbind(2)), and
memory hotplug.
This patchset does not add hugepage migration for memory compaction,
because users of memory compaction mainly expect to construct thp by
arranging raw pages, and there's little or no need to compact hugepages.
CMA, another user of page migration, can have benefit from hugepage
migration, but is not enabled to support it for now (just because of lack
of testing and expertise in CMA.)
Hugepage migration of non pmd-based hugepage (for example 1GB hugepage in
x86_64, or hugepages in architectures like ia64) is not enabled for now
(again, because of lack of testing.)
As for how these are achived, I extended the API (migrate_pages()) to
handle hugepage (with patch 1 and 2) and adjusted code of each caller to
check and collect movable hugepages (with patch 3-7). Remaining 2 patches
are kind of miscellaneous ones to avoid unexpected behavior. Patch 8 is
about making sure that we only migrate pmd-based hugepages. And patch 9
is about choosing appropriate zone for hugepage allocation.
My test is mainly functional one, simply kicking hugepage migration via
each entry point and confirm that migration is done correctly. Test code
is available here:
git://github.com/Naoya-Horiguchi/test_hugepage_migration_extension.git
And I always run libhugetlbfs test when changing hugetlbfs's code. With
this patchset, no regression was found in the test.
This patch (of 9):
Before enabling each user of page migration to support hugepage,
this patch enables the list of pages for migration to link not only
LRU pages, but also hugepages. As a result, putback_movable_pages()
and migrate_pages() can handle both of LRU pages and hugepages.
Signed-off-by: Naoya Horiguchi <[email protected]>
Acked-by: Andi Kleen <[email protected]>
Reviewed-by: Wanpeng Li <[email protected]>
Acked-by: Hillf Danton <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: "Aneesh Kumar K.V" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
In struct gen_pool_chunk, end_addr means the end address of memory chunk
(inclusive), but in the implementation it is treated as address + size of
memory chunk (exclusive), so it points to the address plus one instead of
correct ending address.
The ending address of memory chunk plus one will cause overflow on the
memory chunk including the last address of memory map, e.g. when starting
address is 0xFFF00000 and size is 0x100000 on 32bit machine, ending
address will be 0x100000000.
Use correct ending address like starting address + size - 1.
[[email protected]: add comment to struct gen_pool_chunk:end_addr]
Signed-off-by: Joonyoung Shim <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
The main idea behind this patchset is to reduce the vmstat update overhead
by avoiding interrupt enable/disable and the use of per cpu atomics.
This patch (of 3):
It is better to have a separate folding function because
refresh_cpu_vm_stats() also does other things like expire pages in the
page allocator caches.
If we have a separate function then refresh_cpu_vm_stats() is only called
from the local cpu which allows additional optimizations.
The folding function is only called when a cpu is being downed and
therefore no other processor will be accessing the counters. Also
simplifies synchronization.
[[email protected]: fix UP build]
Signed-off-by: Christoph Lameter <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
CC: Tejun Heo <[email protected]>
Cc: Joonsoo Kim <[email protected]>
Cc: Alexey Dobriyan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
PageSwapCache() is always false when !CONFIG_SWAP, so compiler
properly discard related code. Therefore, we don't need #ifdef explicitly.
Signed-off-by: Joonsoo Kim <[email protected]>
Acked-by: Johannes Weiner <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Rik van Riel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Each zone that holds userspace pages of one workload must be aged at a
speed proportional to the zone size. Otherwise, the time an individual
page gets to stay in memory depends on the zone it happened to be
allocated in. Asymmetry in the zone aging creates rather unpredictable
aging behavior and results in the wrong pages being reclaimed, activated
etc.
But exactly this happens right now because of the way the page allocator
and kswapd interact. The page allocator uses per-node lists of all zones
in the system, ordered by preference, when allocating a new page. When
the first iteration does not yield any results, kswapd is woken up and the
allocator retries. Due to the way kswapd reclaims zones below the high
watermark while a zone can be allocated from when it is above the low
watermark, the allocator may keep kswapd running while kswapd reclaim
ensures that the page allocator can keep allocating from the first zone in
the zonelist for extended periods of time. Meanwhile the other zones
rarely see new allocations and thus get aged much slower in comparison.
The result is that the occasional page placed in lower zones gets
relatively more time in memory, even gets promoted to the active list
after its peers have long been evicted. Meanwhile, the bulk of the
working set may be thrashing on the preferred zone even though there may
be significant amounts of memory available in the lower zones.
Even the most basic test -- repeatedly reading a file slightly bigger than
memory -- shows how broken the zone aging is. In this scenario, no single
page should be able stay in memory long enough to get referenced twice and
activated, but activation happens in spades:
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 0
nr_inactive_file 0
nr_active_file 8
nr_inactive_file 1582
nr_active_file 11994
$ cat data data data data >/dev/null
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 70
nr_inactive_file 258753
nr_active_file 443214
nr_inactive_file 149793
nr_active_file 12021
Fix this with a very simple round robin allocator. Each zone is allowed a
batch of allocations that is proportional to the zone's size, after which
it is treated as full. The batch counters are reset when all zones have
been tried and the allocator enters the slowpath and kicks off kswapd
reclaim. Allocation and reclaim is now fairly spread out to all
available/allowable zones:
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 0
nr_inactive_file 174
nr_active_file 4865
nr_inactive_file 53
nr_active_file 860
$ cat data data data data >/dev/null
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 0
nr_inactive_file 666622
nr_active_file 4988
nr_inactive_file 190969
nr_active_file 937
When zone_reclaim_mode is enabled, allocations will now spread out to all
zones on the local node, not just the first preferred zone (which on a 4G
node might be a tiny Normal zone).
Signed-off-by: Johannes Weiner <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Reviewed-by: Rik van Riel <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Cc: Paul Bolle <[email protected]>
Cc: Zlatko Calusic <[email protected]>
Tested-by: Kevin Hilman <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
tracepoint()
In the current code, the value of fallback_migratetype that is printed
using the mm_page_alloc_extfrag tracepoint, is the value of the
migratetype *after* it has been set to the preferred migratetype (if the
ownership was changed). Obviously that wouldn't have been the original
intent. (We already have a separate 'change_ownership' field to tell
whether the ownership of the pageblock was changed from the
fallback_migratetype to the preferred type.)
The intent of the fallback_migratetype field is to show the migratetype
from which we borrowed pages in order to satisfy the allocation request.
So fix the code to print that value correctly.
Signed-off-by: Srivatsa S. Bhat <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Cody P Schafer <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
swap cluster allocation is to get better request merge to improve
performance. But the cluster is shared globally, if multiple tasks are
doing swap, this will cause interleave disk access. While multiple tasks
swap is quite common, for example, each numa node has a kswapd thread
doing swap and multiple threads/processes doing direct page reclaim.
ioscheduler can't help too much here, because tasks don't send swapout IO
down to block layer in the meantime. Block layer does merge some IOs, but
a lot not, depending on how many tasks are doing swapout concurrently. In
practice, I've seen a lot of small size IO in swapout workloads.
We makes the cluster allocation per-cpu here. The interleave disk access
issue goes away. All tasks swapout to their own cluster, so swapout will
become sequential, which can be easily merged to big size IO. If one CPU
can't get its per-cpu cluster (for example, there is no free cluster
anymore in the swap), it will fallback to scan swap_map. The CPU can
still continue swap. We don't need recycle free swap entries of other
CPUs.
In my test (swap to a 2-disk raid0 partition), this improves around 10%
swapout throughput, and request size is increased significantly.
How does this impact swap readahead is uncertain though. On one side,
page reclaim always isolates and swaps several adjancent pages, this will
make page reclaim write the pages sequentially and benefit readahead. On
the other side, several CPU write pages interleave means the pages don't
live _sequentially_ but relatively _near_. In the per-cpu allocation
case, if adjancent pages are written by different cpus, they will live
relatively _far_. So how this impacts swap readahead depends on how many
pages page reclaim isolates and swaps one time. If the number is big,
this patch will benefit swap readahead. Of course, this is about
sequential access pattern. The patch has no impact for random access
pattern, because the new cluster allocation algorithm is just for SSD.
Alternative solution is organizing swap layout to be per-mm instead of
this per-cpu approach. In the per-mm layout, we allocate a disk range for
each mm, so pages of one mm live in swap disk adjacently. per-mm layout
has potential issues of lock contention if multiple reclaimers are swap
pages from one mm. For a sequential workload, per-mm layout is better to
implement swap readahead, because pages from the mm are adjacent in disk.
But per-cpu layout isn't very bad in this workload, as page reclaim always
isolates and swaps several pages one time, such pages will still live in
disk sequentially and readahead can utilize this. For a random workload,
per-mm layout isn't beneficial of request merge, because it's quite
possible pages from different mm are swapout in the meantime and IO can't
be merged in per-mm layout. while with per-cpu layout we can merge
requests from any mm. Considering random workload is more popular in
workloads with swap (and per-cpu approach isn't too bad for sequential
workload too), I'm choosing per-cpu layout.
[[email protected]: coding-style fixes]
Signed-off-by: Shaohua Li <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Kyungmin Park <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Rafael Aquini <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
swap can do cluster discard for SSD, which is good, but there are some
problems here:
1. swap do the discard just before page reclaim gets a swap entry and
writes the disk sectors. This is useless for high end SSD, because an
overwrite to a sector implies a discard to original sector too. A
discard + overwrite == overwrite.
2. the purpose of doing discard is to improve SSD firmware garbage
collection. Idealy we should send discard as early as possible, so
firmware can do something smart. Sending discard just after swap entry
is freed is considered early compared to sending discard before write.
Of course, if workload is already bound to gc speed, sending discard
earlier or later doesn't make
3. block discard is a sync API, which will delay scan_swap_map()
significantly.
4. Write and discard command can be executed parallel in PCIe SSD.
Making swap discard async can make execution more efficiently.
This patch makes swap discard async and moves discard to where swap entry
is freed. Discard and write have no dependence now, so above issues can
be avoided. Idealy we should do discard for any freed sectors, but some
SSD discard is very slow. This patch still does discard for a whole
cluster.
My test does a several round of 'mmap, write, unmap', which will trigger a
lot of swap discard. In a fusionio card, with this patch, the test
runtime is reduced to 18% of the time without it, so around 5.5x faster.
[[email protected]: coding-style fixes]
Signed-off-by: Shaohua Li <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Kyungmin Park <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Rafael Aquini <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
I'm using a fast SSD to do swap. scan_swap_map() sometimes uses up to
20~30% CPU time (when cluster is hard to find, the CPU time can be up to
80%), which becomes a bottleneck. scan_swap_map() scans a byte array to
search a 256 page cluster, which is very slow.
Here I introduced a simple algorithm to search cluster. Since we only
care about 256 pages cluster, we can just use a counter to track if a
cluster is free. Every 256 pages use one int to store the counter. If
the counter of a cluster is 0, the cluster is free. All free clusters
will be added to a list, so searching cluster is very efficient. With
this, scap_swap_map() overhead disappears.
This might help low end SD card swap too. Because if the cluster is
aligned, SD firmware can do flash erase more efficiently.
We only enable the algorithm for SSD. Hard disk swap isn't fast enough
and has downside with the algorithm which might introduce regression (see
below).
The patch slightly changes which cluster is choosen. It always adds free
cluster to list tail. This can help wear leveling for low end SSD too.
And if no cluster found, the scan_swap_map() will do search from the end
of last cluster. So if no cluster found, the scan_swap_map() will do
search from the end of last free cluster, which is random. For SSD, this
isn't a problem at all.
Another downside is the cluster must be aligned to 256 pages, which will
reduce the chance to find a cluster. I would expect this isn't a big
problem for SSD because of the non-seek penality. (And this is the reason
I only enable the algorithm for SSD).
Signed-off-by: Shaohua Li <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Kyungmin Park <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Rafael Aquini <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
The previous patch doing vmstats for TLB flushes ("mm: vmstats: tlb flush
counters") effectively missed UP since arch/x86/mm/tlb.c is only compiled
for SMP.
UP systems do not do remote TLB flushes, so compile those counters out on
UP.
arch/x86/kernel/cpu/mtrr/generic.c calls __flush_tlb() directly. This is
probably an optimization since both the mtrr code and __flush_tlb() write
cr4. It would probably be safe to make that a flush_tlb_all() (and then
get these statistics), but the mtrr code is ancient and I'm hesitant to
touch it other than to just stick in the counters.
[[email protected]: tweak comments]
Signed-off-by: Dave Hansen <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
I was investigating some TLB flush scaling issues and realized that we do
not have any good methods for figuring out how many TLB flushes we are
doing.
It would be nice to be able to do these in generic code, but the
arch-independent calls don't explicitly specify whether we actually need
to do remote flushes or not. In the end, we really need to know if we
actually _did_ global vs. local invalidations, so that leaves us with few
options other than to muck with the counters from arch-specific code.
Signed-off-by: Dave Hansen <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Simple cleanup. Every user of vma_set_policy() does the same work, this
looks a bit annoying imho. And the new trivial helper which does
mpol_dup() + vma_set_policy() to simplify the callers.
Signed-off-by: Oleg Nesterov <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: David Rientjes <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Read block device partition table from command line. The partition used
for fixed block device (eMMC) embedded device. It is no MBR, save
storage space. Bootloader can be easily accessed by absolute address of
data on the block device. Users can easily change the partition.
This code reference MTD partition, source "drivers/mtd/cmdlinepart.c"
About the partition verbose reference
"Documentation/block/cmdline-partition.txt"
[[email protected]: fix printk text]
[[email protected]: fix error return code in parse_parts()]
Signed-off-by: Cai Zhiyong <[email protected]>
Cc: Karel Zak <[email protected]>
Cc: "Wanglin (Albert)" <[email protected]>
Cc: Marius Groeger <[email protected]>
Cc: David Woodhouse <[email protected]>
Cc: Jens Axboe <[email protected]>
Cc: Brian Norris <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Signed-off-by: Wei Yongjun <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
same_thread_group/has_group_leader_pid
task_struct->pid/tgid should go away.
1. Change same_thread_group() to use task->signal for comparison.
2. Change has_group_leader_pid(task) to compare task_pid(task) with
signal->leader_pid.
Signed-off-by: Oleg Nesterov <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Sergey Dyasly <[email protected]>
Reviewed-by: "Eric W. Biederman" <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Revert commit c846ef7deba2 ("include/linux/smp.h:on_each_cpu(): switch
back to a macro"). It turns out that the problematic linux/irqflags.h
include was fixed within ia64 and mn10300.
Cc: Geert Uytterhoeven <[email protected]>
Cc: David Daney <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Pull networking fixes from David Miller:
1) Brown paper bag fix in HTB scheduler, class options set incorrectly
due to a typoe. Fix from Vimalkumar.
2) It's possible for the ipv6 FIB garbage collector to run before all
the necessary datastructure are setup during init, defer the
notifier registry to avoid this problem. Fix from Michal Kubecek.
3) New i40e ethernet driver from the Intel folks.
4) Add new qmi wwan device IDs, from Bjørn Mork.
5) Doorbell lock in bnx2x driver is not initialized properly in some
configurations, fix from Ariel Elior.
6) Revert an ipv6 packet option padding change that broke standardized
ipv6 implementation test suites. From Jiri Pirko.
7) Fix synchronization of ARP information in bonding layer, from
Nikolay Aleksandrov.
8) Fix missing error return resulting in illegal memory accesses in
openvswitch, from Daniel Borkmann.
9) SCTP doesn't signal poll events properly due to mistaken operator
precedence, fix also from Daniel Borkmann.
10) __netdev_pick_tx() passes wrong index to sk_tx_queue_set() which
essentially disables caching of TX queue in sockets :-/ Fix from
Eric Dumazet.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (29 commits)
net_sched: htb: fix a typo in htb_change_class()
net: qmi_wwan: add new Qualcomm devices
ipv6: don't call fib6_run_gc() until routing is ready
net: tilegx driver: avoid compiler warning
fib6_rules: fix indentation
irda: vlsi_ir: Remove casting the return value which is a void pointer
irda: donauboe: Remove casting the return value which is a void pointer
net: fix multiqueue selection
net: sctp: fix smatch warning in sctp_send_asconf_del_ip
net: sctp: fix bug in sctp_poll for SOCK_SELECT_ERR_QUEUE
net: fib: fib6_add: fix potential NULL pointer dereference
net: ovs: flow: fix potential illegal memory access in __parse_flow_nlattrs
bcm63xx_enet: remove deprecated IRQF_DISABLED
net: korina: remove deprecated IRQF_DISABLED
macvlan: Move skb_clone check closer to call
qlcnic: Fix warning reported by kbuild test robot.
bonding: fix bond_arp_rcv setting and arp validate desync state
bonding: fix store_arp_validate race with mode change
ipv6/exthdrs: accept tlv which includes only padding
bnx2x: avoid atomic allocations during initialization
...
|
|
When loading the ipv6 module, ndisc_init() is called before
ip6_route_init(). As the former registers a handler calling
fib6_run_gc(), this opens a window to run the garbage collector
before necessary data structures are initialized. If a network
device is initialized in this window, adding MAC address to it
triggers a NETDEV_CHANGEADDR event, leading to a crash in
fib6_clean_all().
Take the event handler registration out of ndisc_init() into a
separate function ndisc_late_init() and move it after
ip6_route_init().
Signed-off-by: Michal Kubecek <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs
Pull 9p updates from Eric Van Hensbergen:
"Minor 9p fixes and tweaks for 3.12 merge window
The first fixes namespace issues which causes a kernel NULL pointer
dereference, the second fixes uevent handling to work better with
udev, and the third switches some code to use srlcpy instead of
strncpy in order to be safer.
All changes have been baking in for-next for at least 2 weeks"
* tag 'for-linus-3.12-merge' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
fs/9p: avoid accessing utsname after namespace has been torn down
9p: send uevent after adding/removing mount_tag attribute
fs: 9p: use strlcpy instead of strncpy
|
|
* pm-cpufreq:
intel_pstate: Add Haswell CPU models
Revert "cpufreq: make sure frequency transitions are serialized"
cpufreq: Use signed type for 'ret' variable, to store negative error values
cpufreq: Remove temporary fix for race between CPU hotplug and sysfs-writes
cpufreq: Synchronize the cpufreq store_*() routines with CPU hotplug
cpufreq: Invoke __cpufreq_remove_dev_finish() after releasing cpu_hotplug.lock
cpufreq: Split __cpufreq_remove_dev() into two parts
cpufreq: Fix wrong time unit conversion
cpufreq: serialize calls to __cpufreq_governor()
cpufreq: don't allow governor limits to be changed when it is disabled
|
|
|
|
|
|
|
|
Pull battery/power supply driver updates from Anton Vorontsov:
"New drivers:
- APM X-Gene system reboot driver by Feng Kan and Loc Ho (APM).
- Qualcomm MSM reboot/poweroff driver by Abhimanyu Kapur (Codeaurora).
- Texas Instruments BQ24190 charger driver by Mark A. Greer (Animal
Creek Technologies).
- Texas Instruments TWL4030 MADC battery driver by Lukas Märdian and
Marek Belisko (Golden Delicious Computers). The driver is used on
Freerunner GTA04 phones.
Highlighted fixes and improvements:
- Suspend/wakeup logic improvements: power supply objects will block
system suspend until all power supply events are processed. Thanks
to Zoran Markovic (Linaro), Arve Hjonnevag and Todd Poynor (Google)"
* tag 'for-v3.12' of git://git.infradead.org/battery-2.6:
rx51_battery: Fix channel number when reading adc value
power: Add twl4030_madc battery driver.
bq24190_charger: Workaround SS definition problem on i386 builds
power_supply: Prevent suspend until power supply events are processed
vexpress-poweroff: Should depend on the required infrastructure
twl4030-charger: Fix compiler warning with regulator_enable()
rx51_battery: Replace hardcoded channels values.
bq24190_charger: Add support for TI BQ24190 Battery Charger
ab8500-charger: We print an unintended error message
max8925_power: Fix missing of_node_put
power_supply: Replace strict_strtol() with kstrtol()
power: Add APM X-Gene system reboot driver
power_supply: tosa_battery: Get rid of irq_to_gpio usage
power supply: collie_battery: Convert to use dev_pm_ops
power_supply: Make goldfish_battery depend on GOLDFISH || COMPILE_TEST
power: reset: Add msm restart support
MAINTAINERS: drivers/power: add entry for SmartReflex AVS drivers
|
|
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
Pull drm fixes from Dave Airlie:
"Daniel had some fixes queued up, that were delayed, the stolen memory
ones and vga arbiter ones are quite useful, along with his usual bunch
of stuff, nothing for HSW outputs yet.
The one nouveau fix is for a regression I caused with the poweroff stuff"
* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (30 commits)
drm/nouveau: fix oops on runtime suspend/resume
drm/i915: Delay disabling of VGA memory until vgacon->fbcon handoff is done
drm/i915: try not to lose backlight CBLV precision
drm/i915: Confine page flips to BCS on Valleyview
drm/i915: Skip stolen region initialisation if none is reserved
drm/i915: fix gpu hang vs. flip stall deadlocks
drm/i915: Hold an object reference whilst we shrink it
drm/i915: fix i9xx_crtc_clock_get for multiplied pixels
drm/i915: handle sdvo input pixel multiplier correctly again
drm/i915: fix hpd work vs. flush_work in the pageflip code deadlock
drm/i915: fix up the relocate_entry refactoring
drm/i915: Fix pipe config warnings when dealing with LVDS fixed mode
drm/i915: Don't call sg_free_table() if sg_alloc_table() fails
i915: Update VGA arbiter support for newer devices
vgaarb: Fix VGA decodes changes
vgaarb: Don't disable resources that are not owned
drm/i915: Pin pages whilst mapping the dma-buf
drm/i915: enable trickle feed on Haswell
x86: add early quirk for reserving Intel graphics stolen memory v5
drm/i915: split PCI IDs out into i915_drm.h v4
...
|
|
Pull nfsd updates from Bruce Fields:
"This was a very quiet cycle! Just a few bugfixes and some cleanup"
* 'nfsd-next' of git://linux-nfs.org/~bfields/linux:
rpc: let xdr layer allocate gssproxy receieve pages
rpc: fix huge kmalloc's in gss-proxy
rpc: comment on linux_cred encoding, treat all as unsigned
rpc: clean up decoding of gssproxy linux creds
svcrpc: remove unused rq_resused
nfsd4: nfsd4_create_clid_dir prints uninitialized data
nfsd4: fix leak of inode reference on delegation failure
Revert "nfsd: nfs4_file_get_access: need to be more careful with O_RDWR"
sunrpc: prepare NFS for 2038
nfsd4: fix setlease error return
nfsd: nfs4_file_get_access: need to be more careful with O_RDWR
|
|
This patch adds the Third Party Copy (3PC) bit to signal support
for EXTENDED_COPY within standard inquiry response data.
Also add emulate_3pc device attribute in configfs (enabled by default)
to allow the exposure of this bit to be disabled, if necessary.
Cc: Christoph Hellwig <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Cc: Martin Petersen <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Roland Dreier <[email protected]>
Cc: Zach Brown <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: Nicholas Bellinger <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
This patch adds support for EXTENDED_COPY emulation from SPC-3, that
enables full copy offload target support within both a single virtual
backend device, and across multiple virtual backend devices. It also
functions independent of target fabric, and supports copy offload
across multiple target fabric ports.
This implemenation supports both EXTENDED_COPY PUSH and PULL models
of operation, so the actual CDB may be received on either source or
desination logical unit.
For Target Descriptors, it currently supports the NAA IEEE Registered
Extended designator (type 0xe4), which allows the reference of target
ports to occur independent of fabric type using EVPD 0x83 WWNs.
For Segment Descriptors, it currently supports copy from block to
block (0x02) mode.
It also honors any present SCSI reservations of the destination target
port. Note that only Supports No List Identifier (SNLID=1) mode is
supported.
Also included is basic RECEIVE_COPY_RESULTS with service action type
OPERATING PARAMETERS (0x03) required for SNLID=1 operation.
v3 changes:
- Fix incorrect return type in target_do_receive_copy_results()
(Fengguang)
v2 changes:
- Use target_alloc_sgl() instead of transport_generic_get_mem()
- Convert debug output to use pr_debug()
- Convert target_xcopy_parse_target_descriptors() NAA IEEN WWN
dump to use 0x%16phN format specification
- Drop unnecessary xcopy_pt_cmd->xpt_passthrough_wsem, and
associated usage in xcopy_pt_write_pending() and
target_xcopy_issue_pt_cmd()
- Add check for unsupported EXTENDED_COPY(LID4) service action
bits in target_do_xcopy()
Cc: Christoph Hellwig <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Cc: Martin Petersen <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Roland Dreier <[email protected]>
Cc: Zach Brown <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: Nicholas Bellinger <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
EXTENDED_COPY needs to be able to search a global list of devices
based on NAA WWN device identifiers, so add a simple g_device_list
protected by g_device_mutex.
Cc: Christoph Hellwig <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Cc: Martin Petersen <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Roland Dreier <[email protected]>
Cc: Zach Brown <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: Nicholas Bellinger <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
Both target_alloc_sgl() and transport_generic_map_mem_to_cmd() are
required by EXTENDED_COPY logic when setting up internally dispatched
command descriptors, so go ahead and make both of these non static.
Cc: Christoph Hellwig <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Cc: Martin Petersen <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Roland Dreier <[email protected]>
Cc: Zach Brown <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: Nicholas Bellinger <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
target_core_fabric.h
Reversing the dma_data_direction for pci_map_sg() friends is useful
for other drivers, so move it from tcm_qla2xxx into inline code
within target_core_fabric.h.
Also drop internal usage of equivlient in tcm_qla2xxx fabric code.
Reported-by: Christoph Hellwig <[email protected]>
Cc: Roland Dreier <[email protected]>
Cc: Giridhar Malavali <[email protected]>
Cc: Chad Dupuis <[email protected]>
Cc: Nicholas Bellinger <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
This patch adds support for COMPARE_AND_WRITE emulation on a per block
basis. This logic is used as an atomic test and set primative currently
used by VMWare ESX VAAI for performing array side locking of individual
VMFS extent ownership.
This includes the COMPARE_AND_WRITE CDB parsing within sbc_parse_cdb(),
and does the majority of the work within the compare_and_write_callback()
to perform the verify instance user data comparision, and subsequent
write instance user data I/O submission upon a successfull comparision.
The synchronization is enforced by se_device->caw_sem, that is obtained
before the initial READ I/O submission in sbc_compare_and_write(). The
mutex is then released upon MISCOMPARE in compare_and_write_callback(),
or upon WRITE instance user-data completion in compare_and_write_post().
The implementation currently assumes a single logical block (NoLB=1).
v4 changes:
- Explicitly clear cmd->transport_complete_callback for two failure
cases in sbc_compare_and_write() in order to avoid double unlock
of ->caw_sem in compare_and_write_callback() (Dan Carpenter)
v3 changes:
- Convert se_device->caw_mutex to ->caw_sem
v2 changes:
- Set SCF_COMPARE_AND_WRITE and cmd->execute_cmd() to
sbc_compare_and_write() during setup in sbc_parse_cdb()
- Use sbc_compare_and_write() for initial READ submission with
DMA_FROM_DEVICE
- Reset cmd->execute_cmd() to sbc_execute_rw() for write instance
user-data in compare_and_write_callback()
- Drop SCF_BIDI command flag usage
- Set TRANSPORT_PROCESSING + transport_state flags before write
instance submission, and convert to __target_execute_cmd()
- Prevent sbc_get_size() from being being called twice to
generate incorrect size in sbc_parse_cdb()
- Enforce se_device->caw_mutex synchronization between initial
READ I/O submission, and final WRITE I/O completion.
Cc: Christoph Hellwig <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Cc: Martin Petersen <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: Nicholas Bellinger <[email protected]>
Signed-off-by: Nicholas Bellinger <[email protected]>
|
|
We currently use a compile-time constant to size the node array for the
list_lru structure. Due to this, we don't need to allocate any memory at
initialization time. But as a consequence, the structures that contain
embedded list_lru lists can become way too big (the superblock for
instance contains two of them).
This patch aims at ameliorating this situation by dynamically allocating
the node arrays with the firmware provided nr_node_ids.
Signed-off-by: Glauber Costa <[email protected]>
Cc: Dave Chinner <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
There are no more users of this API, so kill it dead, dead, dead and
quietly bury the corpse in a shallow, unmarked grave in a dark forest deep
in the hills...
[[email protected]: added flowers to the grave]
Signed-off-by: Dave Chinner <[email protected]>
Signed-off-by: Glauber Costa <[email protected]>
Reviewed-by: Greg Thelen <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
Now that the shrinker is passing a node in the scan control structure, we
can pass this to the the generic LRU list code to isolate reclaim to the
lists on matching nodes.
Signed-off-by: Dave Chinner <[email protected]>
Signed-off-by: Glauber Costa <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
The list_lru infrastructure already keeps per-node LRU lists in its
node-specific list_lru_node arrays and provide us with a per-node API, and
the shrinkers are properly equiped with node information. This means that
we can now focus our shrinking effort in a single node, but the work that
is deferred from one run to another is kept global at nr_in_batch. Work
can be deferred, for instance, during direct reclaim under a GFP_NOFS
allocation, where situation, all the filesystem shrinkers will be
prevented from running and accumulate in nr_in_batch the amount of work
they should have done, but could not.
This creates an impedance problem, where upon node pressure, work deferred
will accumulate and end up being flushed in other nodes. The problem we
describe is particularly harmful in big machines, where many nodes can
accumulate at the same time, all adding to the global counter nr_in_batch.
As we accumulate more and more, we start to ask for the caches to flush
even bigger numbers. The result is that the caches are depleted and do
not stabilize. To achieve stable steady state behavior, we need to tackle
it differently.
In this patch we keep the deferred count per-node, in the new array
nr_deferred[] (the name is also a bit more descriptive) and will never
accumulate that to other nodes.
Signed-off-by: Glauber Costa <[email protected]>
Cc: Dave Chinner <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
Pass the node of the current zone being reclaimed to shrink_slab(),
allowing the shrinker control nodemask to be set appropriately for node
aware shrinkers.
Signed-off-by: Dave Chinner <[email protected]>
Signed-off-by: Glauber Costa <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
The list_lru implementation has one function, list_lru_dispose_all, with
only one user (the dentry code). At first, such function appears to make
sense because we are really not interested in the result of isolating each
dentry separately - all of them are going away anyway. However, it's
implementation is buggy in the following way:
When we call list_lru_dispose_all in fs/dcache.c, we scan all dentries
marking them with DCACHE_SHRINK_LIST. However, this is done without the
nlru->lock taken. The imediate result of that is that someone else may
add or remove the dentry from the LRU at the same time. When list_lru_del
happens in that scenario we will see an element that is not yet marked
with DCACHE_SHRINK_LIST (even though it will be in the future) and
obviously remove it from an lru where the element no longer is. Since
list_lru_dispose_all will in effect count down nlru's nr_items and
list_lru_del will do the same, this will lead to an imbalance.
The solution for this would not be so simple: we can obviously just keep
the lru_lock taken, but then we have no guarantees that we will be able to
acquire the dentry lock (dentry->d_lock). To properly solve this, we need
a communication mechanism between the lru and dentry code, so they can
coordinate this with each other.
Such mechanism already exists in the form of the list_lru_walk_cb
callback. So it is possible to construct a dcache-side prune function
that does the right thing only by calling list_lru_walk in a loop until no
more dentries are available.
With only one user, plus the fact that a sane solution for the problem
would involve boucing between dcache and list_lru anyway, I see little
justification to keep the special case list_lru_dispose_all in tree.
Signed-off-by: Glauber Costa <[email protected]>
Cc: Michal Hocko <[email protected]>
Acked-by: Dave Chinner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
This patch adapts the list_lru API to accept an optional node argument, to
be used by NUMA aware shrinking functions. Code that does not care about
the NUMA placement of objects can still call into the very same functions
as before. They will simply iterate over all nodes.
Signed-off-by: Glauber Costa <[email protected]>
Cc: Dave Chinner <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|
|
Now that we have an LRU list API, we can start to enhance the
implementation. This splits the single LRU list into per-node lists and
locks to enhance scalability. Items are placed on lists according to the
node the memory belongs to. To make scanning the lists efficient, also
track whether the per-node lists have entries in them in a active
nodemask.
Note: We use a fixed-size array for the node LRU, this struct can be very
big if MAX_NUMNODES is big. If this becomes a problem this is fixable by
turning this into a pointer and dynamically allocating this to
nr_node_ids. This quantity is firwmare-provided, and still would provide
room for all nodes at the cost of a pointer lookup and an extra
allocation. Because that allocation will most likely come from a may very
well fail.
[[email protected]: fix warnings, added note about node lru]
Signed-off-by: Dave Chinner <[email protected]>
Signed-off-by: Glauber Costa <[email protected]>
Reviewed-by: Greg Thelen <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Cc: "Theodore Ts'o" <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Artem Bityutskiy <[email protected]>
Cc: Arve Hjønnevåg <[email protected]>
Cc: Carlos Maiolino <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Chuck Lever <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Gleb Natapov <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Stultz <[email protected]>
Cc: KAMEZAWA Hiroyuki <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Steven Whitehouse <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Al Viro <[email protected]>
|