kernel_optimize_test/mm
Shaohua Li ec8acf20af 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 17:50:17 -08:00
..
backing-dev.c bdi: allow block devices to say that they require stable page writes 2013-02-21 17:22:19 -08:00
balloon_compaction.c mm: introduce a common interface for balloon pages mobility 2012-12-11 17:22:26 -08:00
bootmem.c mm: Add alloc_bootmem_low_pages_nopanic() 2013-01-29 19:32:59 -08:00
bounce.c block: optionally snapshot page contents to provide stable pages during write 2013-02-21 17:22:20 -08:00
cleancache.c
compaction.c mm: remove MIGRATE_ISOLATE check in hotpath 2013-02-23 17:50:15 -08:00
debug-pagealloc.c
dmapool.c dmapool: make DMAPOOL_DEBUG detect corruption of free marker 2012-12-11 17:22:24 -08:00
fadvise.c
failslab.c
filemap_xip.c mm: move all mmu notifier invocations to be done outside the PT lock 2012-10-09 16:22:58 +09:00
filemap.c mm: only enforce stable page writes if the backing device requires it 2013-02-21 17:22:19 -08:00
fremap.c mm: introduce VM_POPULATE flag to better deal with racy userspace programs 2013-02-23 17:50:11 -08:00
frontswap.c
highmem.c Some nice cleanups, and even a patch my wife did as a "live" demo for 2012-12-20 08:37:05 -08:00
huge_memory.c mm: numa: cleanup flow of transhuge page migration 2013-02-23 17:50:17 -08:00
hugetlb_cgroup.c mm/hugetlb: create hugetlb cgroup file in hugetlb_init 2012-12-18 15:02:15 -08:00
hugetlb.c mm/hugetlb.c: convert to pr_foo() 2013-02-23 17:50:09 -08:00
hwpoison-inject.c
init-mm.c
internal.h mm: directly use __mlock_vma_pages_range() in find_extend_vma() 2013-02-23 17:50:11 -08:00
interval_tree.c mm: add CONFIG_DEBUG_VM_RB build option 2012-10-09 16:22:42 +09:00
Kconfig memory-hotplug: implement register_page_bootmem_info_section of sparse-vmemmap 2013-02-23 17:50:12 -08:00
Kconfig.debug
kmemcheck.c
kmemleak-test.c
kmemleak.c mm/kmemleak.c: remove obsolete simple_strtoul 2012-12-18 15:02:15 -08:00
ksm.c mm/ksm.c: use new hashtable implementation 2013-02-23 17:50:10 -08:00
maccess.c
madvise.c mm: make madvise(MADV_WILLNEED) support swap file prefetch 2013-02-23 17:50:10 -08:00
Makefile mm: introduce a common interface for balloon pages mobility 2012-12-11 17:22:26 -08:00
memblock.c mm/memblock.c: use CONFIG_HAVE_MEMBLOCK_NODE_MAP to protect movablecore_map in memblock_overlaps_region(). 2013-02-23 17:50:14 -08:00
memcontrol.c swap: make each swap partition have one address_space 2013-02-23 17:50:17 -08:00
memory_hotplug.c mm: increase totalram_pages when free pages allocated by bootmem allocator 2013-02-23 17:50:15 -08:00
memory-failure.c mm/memory-failure.c: fix wrong num_poisoned_pages in handling memory error on thp 2013-02-23 17:50:15 -08:00
memory.c mm: fold page->_last_nid into page->flags where possible 2013-02-23 17:50:17 -08:00
mempolicy.c mempolicy: fix is_valid_nodemask() 2013-02-23 17:50:13 -08:00
mempool.c
migrate.c mm: numa: cleanup flow of transhuge page migration 2013-02-23 17:50:17 -08:00
mincore.c swap: make each swap partition have one address_space 2013-02-23 17:50:17 -08:00
mlock.c mm: introduce VM_POPULATE flag to better deal with racy userspace programs 2013-02-23 17:50:11 -08:00
mm_init.c
mmap.c swap: add per-partition lock for swapfile 2013-02-23 17:50:17 -08:00
mmu_context.c
mmu_notifier.c mm/mmu_notifier: allocate mmu_notifier in advance 2012-10-25 14:37:53 -07:00
mmzone.c memcg: fix hotplugged memory zone oops 2012-11-16 14:33:04 -08:00
mprotect.c mm/mprotect.c: coding-style cleanups 2012-12-18 15:02:15 -08:00
mremap.c mm: use mm_populate() for mremap() of VM_LOCKED vmas 2013-02-23 17:50:11 -08:00
msync.c
nobootmem.c mm: Add alloc_bootmem_low_pages_nopanic() 2013-01-29 19:32:59 -08:00
nommu.c swap: add per-partition lock for swapfile 2013-02-23 17:50:17 -08:00
oom_kill.c memcg, oom: provide more precise dump info while memcg oom happening 2013-02-23 17:50:08 -08:00
page_alloc.c mm: teach mm by current context info to not do I/O during memory allocation 2013-02-23 17:50:16 -08:00
page_cgroup.c memcontrol: use N_MEMORY instead N_HIGH_MEMORY 2012-12-12 17:38:32 -08:00
page_io.c
page_isolation.c mm: fix zone_watermark_ok_safe() accounting of isolated pages 2013-01-04 16:11:46 -08:00
page-writeback.c block: optionally snapshot page contents to provide stable pages during write 2013-02-21 17:22:20 -08:00
pagewalk.c thp: change split_huge_page_pmd() interface 2012-12-12 17:38:31 -08:00
percpu-km.c
percpu-vm.c
percpu.c mm, percpu: Make sure percpu_alloc early parameter has an argument 2012-12-02 06:23:04 -08:00
pgtable-generic.c mm: Only flush the TLB when clearing an accessible pte 2012-12-11 14:28:34 +00:00
process_vm_access.c
quicklist.c
readahead.c
rmap.c s390/mm: implement software dirty bits 2013-02-14 15:55:23 +01:00
shmem.c mempolicy: remove arg from mpol_parse_str, mpol_to_str 2013-01-02 09:27:10 -08:00
slab_common.c slab: propagate tunable values 2012-12-18 15:02:14 -08:00
slab.c memcg: add comments clarifying aspects of cache attribute propagation 2012-12-18 15:02:15 -08:00
slab.h slab: propagate tunable values 2012-12-18 15:02:14 -08:00
slob.c sl[au]b: always get the cache from its page in kmem_cache_free() 2012-12-18 15:02:14 -08:00
slub.c slub: drop mutex before deleting sysfs entry 2012-12-18 15:02:15 -08:00
sparse-vmemmap.c
sparse.c memory-failure: use num_poisoned_pages instead of mce_bad_pages 2013-02-23 17:50:15 -08:00
swap_state.c swap: add per-partition lock for swapfile 2013-02-23 17:50:17 -08:00
swap.c swap: make each swap partition have one address_space 2013-02-23 17:50:17 -08:00
swapfile.c swap: add per-partition lock for swapfile 2013-02-23 17:50:17 -08:00
truncate.c mm: drop vmtruncate 2012-12-20 18:46:29 -05:00
util.c swap: make each swap partition have one address_space 2013-02-23 17:50:17 -08:00
vmalloc.c mm: use IS_ENABLED(CONFIG_NUMA) instead of NUMA_BUILD 2012-12-11 17:22:22 -08:00
vmscan.c swap: add per-partition lock for swapfile 2013-02-23 17:50:17 -08:00
vmstat.c mm: don't wait on congested zones in balance_pgdat() 2013-02-23 17:50:15 -08:00