[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20250304110318.159567-1-yizhou.tang@shopee.com>
Date: Tue, 4 Mar 2025 19:03:15 +0800
From: Tang Yizhou <yizhou.tang@...pee.com>
To: tj@...nel.org,
jack@...e.cz,
brauner@...nel.org,
willy@...radead.org,
akpm@...ux-foundation.org
Cc: rostedt@...dmis.org,
mhiramat@...nel.org,
ast@...nel.org,
linux-mm@...ck.org,
linux-fsdevel@...r.kernel.org,
linux-kernel@...r.kernel.org,
Tang Yizhou <yizhou.tang@...pee.com>
Subject: [PATCH v2 0/3] Fix calculations in trace_balance_dirty_pages() for cgwb
From: Tang Yizhou <yizhou.tang@...pee.com>
In my experiment, I found that the output of trace_balance_dirty_pages()
in the cgroup writeback scenario was strange because
trace_balance_dirty_pages() always uses global_wb_domain.dirty_limit for
related calculations instead of the dirty_limit of the corresponding
memcg's wb_domain.
The basic idea of the fix is to store the hard dirty limit value computed
in wb_position_ratio() into struct dirty_throttle_control and use it for
calculations in trace_balance_dirty_pages().
v2:
Adopt Tejun's suggestion and split the renaming code into Patch #2.
Pick up Tejun's Acked-by tag in Patch #3.
Tang Yizhou (3):
writeback: Let trace_balance_dirty_pages() take struct dtc as
parameter
writeback: Rename variables in trace_balance_dirty_pages()
writeback: Fix calculations in trace_balance_dirty_pages() for cgwb
include/linux/writeback.h | 24 +++++++++++++++++++++
include/trace/events/writeback.h | 33 ++++++++++++----------------
mm/page-writeback.c | 37 +++-----------------------------
3 files changed, 41 insertions(+), 53 deletions(-)
--
2.25.1
Powered by blists - more mailing lists