From: Huang Ying
In this patch, splitting huge page is delayed from almost the first step
of swapping out to after allocating the swap space for the
THP (Transparent Huge Page) and adding the THP into the swap cache.
This will reduce lock acquiring/releasing for the locks used for the
swap cache
From: Huang Ying
This patchset is to optimize the performance of Transparent Huge Page
(THP) swap.
Hi, Andrew, could you help me to check whether the overall design is
reasonable?
Hi, Hugh, Shaohua, Minchan and Rik, could you help me to review the
swap part of the patchset? Especially [01/10
Andrew Morton writes:
> On Thu, 1 Sep 2016 08:16:54 -0700 "Huang, Ying" wrote:
>
>> From: Huang Ying
>>
>> In this patch, the size of the swap cluster is changed to that of the
>> THP (Transparent Huge Page) on x86_64 architecture (512). This is for
&
Andrew Morton writes:
> On Thu, 01 Sep 2016 16:04:57 -0700 "Huang\, Ying"
> wrote:
>
>> >> }
>> >>
>> >> -#define SWAPFILE_CLUSTER 256
>> >> +#define SWAPFILE_CLUSTER 512
>> >> #define LATENCY_LIMIT
Hi, Jaegeuk,
"Huang, Ying" writes:
> Hi,
>
> I checked the comparison result below and found this is a regression for
> fsmark.files_per_sec, not fsmark.app_overhead.
>
> Best Regards,
> Huang, Ying
>
> kernel test robot writes:
>
>> FYI, we notic
Jaegeuk Kim writes:
> On Thu, Aug 04, 2016 at 10:44:20AM -0700, Huang, Ying wrote:
>> Jaegeuk Kim writes:
>>
>> > Hi Huang,
>> >
>> > On Thu, Aug 04, 2016 at 10:00:41AM -0700, Huang, Ying wrote:
>> >> Hi, Jaegeuk,
>> >>
>
016 at 02:33:08PM -0700, Huang, Ying wrote:
>> Hi, Minchan,
>>
>> Minchan Kim writes:
>> > Anyway, I hope [1/11] should be merged regardless of the patchset because
>> > I believe anyone doesn't feel comfortable with cluser_info functions. ;-)
>>
>
ramdisk on a Xeon E5 v3 machine,
the swap out throughput improved 40.4%, from ~0.97GB/s to ~1.36GB/s.
What's your plan for this patch? If it can be merged soon, that will be
great!
I found some issues in the original patch to work with swap cache. Below
is my fixes to make it work for swap cac
"Huang, Ying" writes:
> Hi, Dave,
>
> Dave Hansen writes:
>
>> On 08/09/2016 09:17 AM, Huang, Ying wrote:
>>> File pages uses a set of radix tags (DIRTY, TOWRITE, WRITEBACK) to
>>> accelerate finding the pages with the specific tag in the the rad
Hi, Jaegeuk,
On Thu, Aug 11, 2016 at 6:22 PM, Jaegeuk Kim wrote:
> On Thu, Aug 11, 2016 at 03:49:41PM -0700, Huang, Ying wrote:
>> Hi, Kim,
>>
>> "Huang, Ying" writes:
>> >>
>> >> [lkp] [f2fs] 3bdad3c7ee: aim7.jobs-per-min -25.3% regression
From: Huang Ying
This is a code clean up patch without functionality changes. The
swap_cluster_list data structure and its operations are introduced to
provide some better encapsulation for the free cluster and discard
cluster list operations. This avoid some code duplication, improved
the
Hi, Peter,
Do you have a git tree branch for this patchset? We want to test it in
0day performance test. That will make it a little easier.
Best Regards,
Huang, Ying
%
perf-profile.cycles-pp._raw_spin_lock_irqsave.test_clear_page_writeback.end_page_writeback.page_endio.pmem_rw_page
Cc: Hugh Dickins
Cc: Shaohua Li
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Mel Gorman
Cc: Tejun Heo
Cc: Wu Fengguang
Cc: Dave Hansen
Signed-off-by: "Huang, Ying"
--
From: Huang Ying
One page may store a set of entries of the
sis->swap_map (swap_info_struct->swap_map) in multiple swap clusters.
If some of the entries has sis->swap_map[offset] > SWAP_MAP_MAX,
multiple pages will be used to store the set of entries of the
sis->swap_map. An
From: Huang Ying
Now, when the page table is walked in the implementation of
/proc//pagemap, pmd_soft_dirty() is used for both the PMD huge
page map and the PMD migration entries. That is wrong,
pmd_swp_soft_dirty() should be used for the PMD migration entries
instead because the different page
Michal Hocko writes:
> On Tue 17-10-17 16:13:20, Huang, Ying wrote:
>> From: Huang Ying
>>
>> One page may store a set of entries of the
>> sis->swap_map (swap_info_struct->swap_map) in multiple swap clusters.
>> If some of the entries has sis->sw
Zi Yan writes:
> Huang, Ying wrote:
>> "Kirill A. Shutemov" writes:
>>
>>> On Tue, Oct 17, 2017 at 04:18:18PM +0800, Huang, Ying wrote:
>>>> From: Huang Ying
>>>>
>>>> Now, when the page table is walked in the implementa
Ming Lei writes:
> On Thu, May 25, 2017 at 02:46:28PM +0800, Huang, Ying wrote:
>> From: Huang Ying
>>
>> In this patch, BIO_MAX_PAGES is changed from 256 to HPAGE_PMD_NR if
>> CONFIG_THP_SWAP is enabled and HPAGE_PMD_NR > 256. This is to support
>&
e)
> - put_page(page);
> + if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, vmf->orig_pte))
> goto out;
> - }
The page table holding PTE may be unmapped in pte_unmap_same(), so is it
safe for us to access page table after this in d
Minchan Kim writes:
> Hi Huang,
>
> On Wed, Nov 01, 2017 at 01:41:00PM +0800, Huang, Ying wrote:
>> Hi, Minchan,
>>
>> Minchan Kim writes:
>>
>> > When I see recent change of swap readahead, I am very unhappy
>> > about current code structu
Minchan Kim writes:
> This patch makes do_swap_page no need to be aware of two different
> swap readahead algorithm. Just unify cluster-based and vma-based
> readahead function call.
>
> Signed-off-by: Minchan Kim
> ---
> include/linux/swap.h | 17 -
> mm/memory.c | 11
u need this? You saved copying one page from memory to memory
(COW) now, at the cost of reading a page from disk to memory later?
Best Regards,
Huang, Ying
> Signed-off-by: zhouxianrong
> ---
> mm/swapfile.c |9 +++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
&g
From: Huang Ying
When a page fault occurs for a swap entry, the physical swap readahead
(not the VMA base swap readahead) may readahead several swap entries
after the fault swap entry. The readahead algorithm calculates some
of the swap entries to readahead via increasing the offset of the
Hi, Michal,
Michal Hocko writes:
> On Tue 24-10-17 10:47:00, Huang, Ying wrote:
>> From: Ying Huang
>>
>> __swp_swapcount() is used in __read_swap_cache_async(). Where the
>> invalid swap entry (offset > max) may be supplied during swap
>> readahead. Bu
Michal Hocko writes:
> On Tue 24-10-17 23:15:32, Huang, Ying wrote:
>> Hi, Michal,
>>
>> Michal Hocko writes:
>>
>> > On Tue 24-10-17 10:47:00, Huang, Ying wrote:
>> >> From: Ying Huang
>> >>
>> >> __swp_swapcount() i
Minchan Kim writes:
> On Tue, Oct 24, 2017 at 10:47:00AM +0800, Huang, Ying wrote:
>> From: Ying Huang
>>
>> __swp_swapcount() is used in __read_swap_cache_async(). Where the
>> invalid swap entry (offset > max) may be supplied during swap
>> readahead. Bu
From: Huang Ying
When a page fault occurs for a swap entry, the physical swap readahead
(not the VMA base swap readahead) may readahead several swap entries
after the fault swap entry. The readahead algorithm calculates some
of the swap entries to readahead via increasing the offset of the
From: Huang Ying
If THP migration is enabled, the following situation is possible,
- A THP is mapped at source address
- Migration is started to move the THP to another node
- Page fault occurs
- The PMD (migration entry) is copied to the destination address in mremap
That is, it is possible
On Fri, Nov 3, 2017 at 11:00 PM, Zi Yan wrote:
> On 3 Nov 2017, at 3:52, Huang, Ying wrote:
>
>> From: Huang Ying
>>
>> If THP migration is enabled, the following situation is possible,
>>
>> - A THP is mapped at source address
>> - Migration is started
From: Huang Ying
Now, when the page table is walked in the implementation of
/proc//pagemap, pmd_soft_dirty() is used for both the PMD huge
page map and the PMD migration entries. That is wrong,
pmd_swp_soft_dirty() should be used for the PMD migration entries
instead because the different page
From: Huang Ying
One page may store a set of entries of the
sis->swap_map (swap_info_struct->swap_map) in multiple swap clusters.
If some of the entries has sis->swap_map[offset] > SWAP_MAP_MAX,
multiple pages will be used to store the set of entries of the
sis->swap_map. An
Hi, Minchan,
Minchan Kim writes:
> Hi Huang,
>
> On Fri, Oct 27, 2017 at 01:53:27PM +0800, Huang, Ying wrote:
>> From: Huang Ying
>>
>> When a page fault occurs for a swap entry, the physical swap readahead
>> (not the VMA base swap readahead) may readahead
Minchan Kim writes:
> Hi Huang,
>
> On Tue, Oct 31, 2017 at 01:32:32PM +0800, Huang, Ying wrote:
>> Hi, Minchan,
>>
>> Minchan Kim writes:
>>
>> > Hi Huang,
>> >
>> > On Fri, Oct 27, 2017 at 01:53:27PM +0800, Huang, Ying wrote:
>&
12 19:27:20 /usr/bin/hackbench -g 8 --threads -l 6
> Running in threaded mode with 8 groups using 40 file descriptors each
> (== 320 tasks)
> Each sender will pass 6 messages of 100 bytes
> ...
> wait for background monitors: perf-profile uptime proc-vmstat proc-stat
> mem
155259 ± 0% +0.0% 155259 ± 0% -42.2% 89723 ± 0%
sched_debug.sysctl_sched.sysctl_sched_features
Best Regards,
Huang, Ying
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://v
in should be the first. So if we found one
cache with too small (old) cache->time_in, we can say there are no cache
with same contents and bigger (newer) cache->time_in, so that we can
make decision (break) earlier.
Best Regards,
Huang, Ying
--
To unsubscribe from this list: send the line
Borislav Petkov writes:
> On Mon, Oct 26, 2015 at 11:20:35AM +0800, Huang, Ying wrote:
>> In ghes_estatus_caches[], for caches with same contents, the cache with
>> biggest (newest) cache->time_in should be the first. So if we found one
>> cache with too small (old) ca
swap entry can be reclaimed
later eventually.
Signed-off-by: "Huang, Ying"
Cc: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
---
mm/swapfile.c | 57 +
1 file changed, 25
=y
>> CONFIG_STACKPROTECTOR=y
>> CONFIG_STACKPROTECTOR_STRONG=y
>> CONFIG_CC_HAS_SANE_STACKPROTECTOR=y
>
> so you're testing the "no overhead" case to the "worst overhead" case.
Do you have interest in some other comparison?
Best Regards,
Huang, Ying
From: Huang Ying
Previously, during swapout, all PMD page mapping will be split and
replaced with PTE swap mapping. And when clearing the SWAP_HAS_CACHE
flag for the huge swap cluster in swapcache_free_cluster(), the huge
swap cluster will be split. Now, during swapout, the PMD page mapping
From: Huang Ying
It's unreasonable to optimize swapping for THP without basic swapping
support. And this will cause build errors when THP_SWAP functions are
defined in swapfile.c and called elsewhere.
The comments are fixed too to reflect the latest progress.
Signed-off-by: "Huang,
From: Huang Ying
Hi, Andrew, could you help me to check whether the overall design is
reasonable?
Hi, Hugh, Shaohua, Minchan and Rik, could you help me to review the
swap part of the patchset? Especially [02/21], [03/21], [04/21],
[05/21], [06/21], [07/21], [08/21], [09/21], [10/21], [11/21
From: Huang Ying
When a PMD swap mapping is removed from a huge swap cluster, for
example, unmap a memory range mapped with PMD swap mapping, etc,
free_swap_and_cache() will be called to decrease the reference count
to the huge swap cluster. free_swap_and_cache() may also free or
split the huge
From: Huang Ying
Previously, the PMD swap operations are only enabled for
CONFIG_ARCH_ENABLE_THP_MIGRATION. Because they are only used by the
THP migration support. We will support PMD swap mapping to the huge
swap cluster and swapin the THP as a whole. That will be enabled via
From: Huang Ying
For a PMD swap mapping, zap_huge_pmd() will clear the PMD and call
free_swap_and_cache() to decrease the swap reference count and maybe
free or split the huge swap cluster and the THP in swap cache.
Signed-off-by: "Huang, Ying"
Cc: "Kirill A. Shutemov"
From: Huang Ying
During fork, the page table need to be copied from parent to child. A
PMD swap mapping need to be copied too and the swap reference count
need to be increased.
When the huge swap cluster has been split already, we need to split
the PMD swap mapping and fallback to PTE copying
From: Huang Ying
When madvise_free() found a PMD swap mapping, if only part of the huge
swap cluster is operated on, the PMD swap mapping will be split and
fallback to PTE swap mapping processing. Otherwise, if all huge swap
cluster is operated on, free_swap_and_cache() will be called to
From: Huang Ying
Swapin a THP as a whole isn't desirable at some situations. For
example, for random access pattern, swapin a THP as a whole will
inflate the reading greatly. So a sysfs interface:
/sys/kernel/mm/transparent_hugepage/swapin_enabled is added to
configure it. Three optio
From: Huang Ying
During mincore(), for PMD swap mapping, swap cache will be looked up.
If the resulting page isn't compound page, the PMD swap mapping will
be split and fallback to PTE swap mapping processing.
Signed-off-by: "Huang, Ying"
Cc: "Kirill A. Shutemov"
From: Huang Ying
This is the final step of the THP swapin support. When reclaiming a
anonymous THP, after allocating the huge swap cluster and add the THP
into swap cache, the PMD page mapping will be changed to the mapping
to the swap space. Previously, the PMD page mapping will be split
From: Huang Ying
Previously, to reclaim MADV_FREE THP, the THP will be split firstly,
then reclaim each sub-pages. This wastes cycles to split THP and
unmap and free each sub-pages, and split THP even if it has been
written since MADV_FREE. We have to do this because MADV_FREE THP
reclaiming
From: Huang Ying
Original code is only for PMD migration entry, it is revised to
support PMD swap mapping.
Signed-off-by: "Huang, Ying"
Cc: "Kirill A. Shutemov"
Cc: Andrea Arcangeli
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
From: Huang Ying
During swapoff, for a huge swap cluster, we need to allocate a THP,
read its contents into the THP and unuse the PMD and PTE swap mappings
to it. If failed to allocate a THP, the huge swap cluster will be
split.
During unuse, if it is found that the swap cluster mapped by a
From: Huang Ying
Previously the huge swap cluster will be split after the THP is
swapout. Now, to support to swapin the THP as a whole, the huge swap
cluster will not be split after the THP is reclaimed. So in memcg, we
need to move the swap account for PMD swap mappings in the process's
From: Huang Ying
During MADV_WILLNEED, for a PMD swap mapping, if THP swapin is enabled
for the VMA, the whole swap cluster will be swapin. Otherwise, the
huge swap cluster and the PMD swap mapping will be split and fallback
to PTE swap mapping.
Signed-off-by: "Huang, Ying"
Cc:
From: Huang Ying
When splitting a THP in swap cache or failing to allocate a THP when
swapin a huge swap cluster, the huge swap cluster will be split. In
addition to clear the huge flag of the swap cluster, the PMD swap
mapping count recorded in cluster_count() will be set to 0. But we
will
From: Huang Ying
2 new /proc/vmstat fields are added, "thp_swapin" and
"thp_swapin_fallback" to count swapin a THP from swap device as a
whole and fallback to normal page swapin.
Signed-off-by: "Huang, Ying"
Cc: "Kirill A. Shutemov"
Cc: Andrea Arcangel
From: Huang Ying
With this patch, when page fault handler find a PMD swap mapping, it
will swap in a THP as a whole. This avoids the overhead of
splitting/collapsing before/after the THP swapping. And improves the
swap performance greatly for reduced page fault count etc
From: Huang Ying
To swapin a THP as a whole, we need to read a huge swap cluster from
the swap device. This patch revised the __read_swap_cache_async() and
its callers and callees to support this. If __read_swap_cache_async()
find the swap cluster of the specified swap entry is huge, it will
From: Huang Ying
A huge PMD need to be split when zap a part of the PMD mapping etc.
If the PMD mapping is a swap mapping, we need to split it too. This
patch implemented the support for this. This is similar as splitting
the PMD page mapping, except we need to decrease the PMD swap mapping
From: Huang Ying
To support to swapin the THP as a whole, we need to create PMD swap
mapping during swapout, and maintain PMD swap mapping count. This
patch implements the support to increase the PMD swap mapping
count (for swapout, fork, etc.) and set SWAP_HAS_CACHE flag (for
swapin, etc
To improve the code readability. Some swap free related functions are
refactored.
This patchset is based on 8/23 HEAD of mmotm tree.
Best Regards,
Huang, Ying
the swap entry can be reclaimed
later eventually.
Signed-off-by: "Huang, Ying"
Cc: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
---
mm/swapfile.c | 57 +
1 file c
code.
Signed-off-by: "Huang, Ying"
Cc: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
---
mm/swapfile.c | 10 --
1 file changed, 4 insertions(+), 6 deletions(-)
diff --git a/mm/swapfile.c b/mm/swapfile.c
index 40
is moved to swap_free_cluster() to avoid
the downside.
Signed-off-by: "Huang, Ying"
Cc: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
---
mm/swapfile.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/mm/sw
; usually being 1 or 2 when idle, and only increasing to ~4-7 when stressed.
>> >
>> >
>> > PAGES_BETWEEN_RESCHED state AVG stddev
>> > 1 4 KiB idle36.086 1.920
>> > 16 64 KiB idle34.797 1.702
>> > 32 128 KiB idle35.104 1.752
>> > 64 256 KiB idle34.468 0.661
>> > 512 2048 KiBidle36.427 0.946
>> > 20488192 KiBidle34.988 2.406
>> > 262144 1048576 KiB idle36.792 0.193
>> > infin 512 GiB idle38.817 0.238 [causes softlockup]
>> > 1 4 KiB stress 55.562 0.661
>> > 16 64 KiB stress 57.509 0.248
>> > 32 128 KiB stress 69.265 3.913
>> > 64 256 KiB stress 70.217 4.534
>> > 512 2048 KiBstress 68.474 1.708
>> > 20488192 KiBstress 70.806 1.068
>> > 262144 1048576 KiB stress 55.217 1.184
>> > infin 512 GiB stress 55.062 0.291 [causes softlockup]
I think it may be good to separate the two optimization into 2 patches.
This makes it easier to evaluate the benefit of individual optimization.
Best Regards,
Huang, Ying
Dave Hansen writes:
> On 07/17/2018 08:25 PM, Huang, Ying wrote:
>>> Seriously, though, does it hurt us to add a comment or two to say
>>> something like:
>>>
>>> /*
>>> * Should not even be attempting cluster allocations when
>>&
is a public
function with a stub implementation for CONFIG_THP_SWAP=n in swap.h.
Signed-off-by: "Huang, Ying"
Suggested-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Wi
filename
base 242152028 340 2658367d7 mm/swapfile.o
head 241232004 340 264676763 mm/swapfile.o
Signed-off-by: "Huang, Ying"
Cc: Daniel Jordan
Cc: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
via merging huge/normal
code path/functions if possible.
One concern is that this may cause code size to dilate when
!CONFIG_TRANSPARENT_HUGEPAGE. The data shows that most refactoring
will only cause quite slight code size increase.
Best Regards,
Huang, Ying
nted out by Daniel, it is better to use "swap_count(map[i])"
here, because it works for "map[i] == 0" case too.
And this makes the implementation more consistent between normal and
huge swap entry.
Signed-off-by: "Huang, Ying"
Suggested-and-reviewed-by: Daniel Jordan
To improve the code readability.
Signed-off-by: "Huang, Ying"
Suggested-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Williams
---
mm/swapfile.c | 6 ++
1 file
=n.
Signed-off-by: "Huang, Ying"
Suggested-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Williams
---
mm/swapf
: "Huang, Ying"
Suggested-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Williams
---
mm/swapfile.c | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/mm/
The part of __swap_entry_free() with lock held is separated into a new
function __swap_entry_free_locked(). Because we want to reuse that
piece of code in some other places.
Just mechanical code refactoring, there is no any functional change in
this function.
Signed-off-by: "Huang,
. More difference comes from the ORC unwinder
segments: (1480 + 2220) - (1380 + 2070) = 250. If the frame pointer
unwinder is used, this costs nothing.
Signed-off-by: "Huang, Ying"
Reviewed-by: Daniel Jordan
Cc: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh D
E;
>> +else
>> +return false;
>
> Nitpick: no need for an else after a return:
>
> if (IS_ENABLED(CONFIG_THP_SWAP))
> return info->flags & CLUSTER_FLAG_HUGE;
> return false;
Sure. Will change this in next version.
Best Regards,
Huang, Ying
Christoph Hellwig writes:
> On Thu, Jul 19, 2018 at 04:48:35PM +0800, Huang Ying wrote:
>> +/*
>> + * Determine the locking method in use for this device. Return
>> + * swap_cluster_info if SSD-style cluster-based locking is in place.
>> + */
>> stati
via merging huge/normal
code path/functions if possible.
One concern is that this may cause code size to dilate when
!CONFIG_TRANSPARENT_HUGEPAGE. The data shows that most refactoring
will only cause quite slight code size increase.
Best Regards,
Huang, Ying
To improve the code readability.
Signed-off-by: "Huang, Ying"
Suggested-and-acked-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Williams
---
mm/swapfile.c | 9 +++
: "Huang, Ying"
Suggested-and-acked-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Williams
---
mm/swapfile.c | 7 +++
1 file changed, 3 insertions(+), 4 deletions(-)
diff
is a public
function with a stub implementation for CONFIG_THP_SWAP=n in swap.h.
Signed-off-by: "Huang, Ying"
Suggested-and-acked-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
C
. More difference comes from the ORC unwinder
segments: (1480 + 2220) - (1380 + 2070) = 250. If the frame pointer
unwinder is used, this costs nothing.
Signed-off-by: "Huang, Ying"
Reviewed-by: Daniel Jordan
Acked-by: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
=n.
Signed-off-by: "Huang, Ying"
Suggested-and-acked-by: Dave Hansen
Reviewed-by: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Dan Williams
---
mm/swapf
The part of __swap_entry_free() with lock held is separated into a new
function __swap_entry_free_locked(). Because we want to reuse that
piece of code in some other places.
Just mechanical code refactoring, there is no any functional change in
this function.
Signed-off-by: "Huang,
filename
base 242152028 340 2658367d7 mm/swapfile.o
head 241232004 340 264676763 mm/swapfile.o
Signed-off-by: "Huang, Ying"
Acked-by: Dave Hansen
Cc: Daniel Jordan
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
C
nted out by Daniel, it is better to use "swap_count(map[i])"
here, because it works for "map[i] == 0" case too.
And this makes the implementation more consistent between normal and
huge swap entry.
Signed-off-by: "Huang, Ying"
Suggested-and-reviewed-by: Daniel Jor
rom the llist.
Some other locks could be used to provide mutual exclusive between
- llist add, llist traverse
and
- llist delete
Is this your use case?
Best Regards,
Huang, Ying
Byungchul Park writes:
> On Tue, Jul 31, 2018 at 09:37:50AM +0800, Huang, Ying wrote:
>> Byungchul Park writes:
>>
>> > Hello folks,
>> >
>> > I'm careful in saying.. and curious about..
>> >
>> > In restrictive cases like only
e more information about where is your use case? Is it a
kernel driver? Then it is better to submit the patch together with its
user.
And I have the similar concern as Steven Rostedt. That is, if you are
the only user forever, it's not necessary to change the common code.
Best Regards,
Huang, Ying
From: Huang Ying
In this patch, locking related code is shared between huge/normal code
path in put_swap_page() to reduce code duplication. And `free_entries
== 0` case is merged into more general `free_entries !=
SWAPFILE_CLUSTER` case, because the new locking method makes it easy.
The added
via merging huge/normal
code path/functions if possible.
One concern is that this may cause code size to dilate when
!CONFIG_TRANSPARENT_HUGEPAGE. The data shows that most refactoring
will only cause quite slight code size increase.
Best Regards,
Huang, Ying
From: Huang Ying
In mm/swapfile.c, THP (Transparent Huge Page) swap specific code is
enclosed by #ifdef CONFIG_THP_SWAP/#endif to avoid code dilating when
THP isn't enabled. But #ifdef/#endif in .c file hurt the code
readability, so Dave suggested to use IS_ENABLED(CONFIG_THP_SWAP)
instea
From: Huang Ying
To improve the code readability.
Signed-off-by: "Huang, Ying"
Suggested-by: Dave Hansen
Cc: Michal Hocko
Cc: Johannes Weiner
Cc: Shaohua Li
Cc: Hugh Dickins
Cc: Minchan Kim
Cc: Rik van Riel
Cc: Daniel Jordan
Cc: Dan Williams
---
mm/swapfile.c | 6
From: Huang Ying
As suggested by Dave, we should unify the code path for normal and
huge swap support if possible to avoid duplicated code, bugs, etc. and
make it easier to review code.
In this patch, the normal/huge code path in swap_page_trans_huge_swapped()
is unified, the added and removed
From: Huang Ying
In this patch, the normal/huge code path in put_swap_page() and
several helper functions are unified to avoid duplicated code, bugs,
etc. and make it easier to review the code.
The removed lines are more than added lines. And the binary size is
kept exactly same when
From: Huang Ying
The part of __swap_entry_free() with lock held is separated into a new
function __swap_entry_free_locked(). Because we want to reuse that
piece of code in some other places.
Just mechanical code refactoring, there is no any functional change in
this function.
Signed-off-by
From: Huang Ying
CONFIG_THP_SWAP should depend on CONFIG_SWAP, because it's
unreasonable to optimize swapping for THP (Transparent Huge Page)
without basic swapping support.
In original code, when CONFIG_SWAP=n and CONFIG_THP_SWAP=y,
split_swap_cluster() will not be built because it
ster_info, and
si->swap_map[], this is described in comments of struct
swap_cluster_info. si->lock is used to protect other fields of si. If
two locks need to be held, hold si->lock first. This is for non-HDD.
For HDD, there are no cluster, so si->lock is used to protect
si->swap_map[].
Best Regards,
Huang, Ying
Daniel Jordan writes:
> On Fri, Jul 13, 2018 at 07:36:33AM +0800, Huang, Ying wrote:
>> diff --git a/mm/swapfile.c b/mm/swapfile.c
>> index 75c84aa763a3..160f78072667 100644
>> --- a/mm/swapfile.c
>> +++ b/mm/swapfile.c
>> @@ -270,7 +270,10 @@ static in
401 - 500 of 1854 matches
Mail list logo
