On 2024/4/23 5:56, Andrew Morton wrote:
On Fri, 19 Apr 2024 09:21:06 +0800 Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:
Hi Andrew,
On 2024/4/19 3:44, Andrew Morton wrote:
------------------------------------------------------
From: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
Subject: mm: swapfile: check usable swap device in __folio_throttle_swaprate()
Date: Thu, 18 Apr 2024 21:56:44 +0800
Skip blk_cgroup_congested() if there is no usable swap device since no
swapin/out will occur, Thereby avoid taking swap_lock. The difference is
shown below from perf date of CoW pagefault,
Thanks your for help to improving changelog, so kind to me.
perf report -g -i perf.data.swapoff | egrep "blk_cgroup_congested|__folio_throttle_swaprate"
I make a mistake, this should swapon
1.01% 0.16% page_fault2_pro [kernel.kallsyms] [k] __folio_throttle_swaprate
0.83% 0.80% page_fault2_pro [kernel.kallsyms] [k] blk_cgroup_congested
perf report -g -i perf.data.swapon | egrep "blk_cgroup_congested|__folio_throttle_swaprate"
and this is swapoff
We don't use swap in our scene, so avoid blk_cgroup_congested(), could
you help to correct it too, thanks
0.15% 0.15% page_fault2_pro [kernel.kallsyms] [k] __folio_throttle_swaprate
Please send me the new changelog text and I'll paste it in, thanks.
Here, thank you, Andrew
Skip blk_cgroup_congested() if there is no usable swap device since no
swapin/out will occur, Thereby avoid taking swap_lock. The difference is
shown below from perf date of CoW pagefault,
perf report -g -i perf.data.swapon | egrep
"blk_cgroup_congested|__folio_throttle_swaprate"
1.01% 0.16% page_fault2_pro [kernel.kallsyms] [k]
__folio_throttle_swaprate
0.83% 0.80% page_fault2_pro [kernel.kallsyms] [k]
blk_cgroup_congested
perf report -g -i perf.data.swapoff | egrep
"blk_cgroup_congested|__folio_throttle_swaprate"
0.15% 0.15% page_fault2_pro [kernel.kallsyms] [k]
__folio_throttle_swaprate
