From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug alb_break_llc() decides whether to break LLC preference during active load balance. It does so by testing that every runnable fair task on the source rq prefers its LLC: env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable But the two counters cover different sets. nr_pref_llc_running is updated in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued, so it follows queued tasks. h_nr_runnable is updated in set_delayed()/ clear_delayed() and drops delay-dequeued tasks. So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks, alb_break_llc() returns false, and active balance is free to pull a task off its preferred LLC. Active balance only moves runnable tasks, and this is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB skips the per-task test in can_migrate_task(). The runnable set is the one we want. Fix it on the counter side. A task should be counted in nr_pref_llc_running exactly while it is both queued on its preferred LLC (pref_llc_queued) and runnable (!sched_delayed). Define that membership once in task_pref_llc_runnable(), and adjust the counter only through pref_llc_running_inc()/pref_llc_running_dec() from the four sites that change either input: account_llc_enqueue(), account_llc_dequeue(), set_delayed() and clear_delayed(). Gating every update on the same predicate keeps the delay, wake and dequeue paths from double-counting or underflowing; see the comments at those sites for the ordering. nr_llc_running and sd->llc_counts are not touched and stay on queued semantics. Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC") Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/ Reported-by: Zhan Xusheng Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- Documentation/ABI/testing/sysfs-fs-ext4 | 118 ++++++++++++++++++++++++++++++++ 1 file changed, 118 insertions(+) create mode 100644 Documentation/ABI/testing/sysfs-fs-ext4 (limited to 'Documentation/ABI/testing/sysfs-fs-ext4') diff --git a/Documentation/ABI/testing/sysfs-fs-ext4 b/Documentation/ABI/testing/sysfs-fs-ext4 new file mode 100644 index 000000000..2edd0a667 --- /dev/null +++ b/Documentation/ABI/testing/sysfs-fs-ext4 @@ -0,0 +1,118 @@ +What: /sys/fs/ext4//mb_stats +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + Controls whether the multiblock allocator should + collect statistics, which are shown during the unmount. + 1 means to collect statistics, 0 means not to collect + statistics + +What: /sys/fs/ext4//mb_group_prealloc +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + The multiblock allocator will round up allocation + requests to a multiple of this tuning parameter if the + stripe size is not set in the ext4 superblock + +What: /sys/fs/ext4//mb_max_to_scan +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + The maximum number of extents the multiblock allocator + will search to find the best extent + +What: /sys/fs/ext4//mb_min_to_scan +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + The minimum number of extents the multiblock allocator + will search to find the best extent + +What: /sys/fs/ext4//mb_order2_req +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + Tuning parameter which controls the minimum size for + requests (as a power of 2) where the buddy cache is + used + +What: /sys/fs/ext4//mb_stream_req +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + Files which have fewer blocks than this tunable + parameter will have their blocks allocated out of a + block group specific preallocation pool, so that small + files are packed closely together. Each large file + will have its blocks allocated out of its own unique + preallocation pool. + +What: /sys/fs/ext4//inode_readahead_blks +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + Tuning parameter which controls the maximum number of + inode table blocks that ext4's inode table readahead + algorithm will pre-read into the buffer cache + +What: /sys/fs/ext4//delayed_allocation_blocks +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + This file is read-only and shows the number of blocks + that are dirty in the page cache, but which do not + have their location in the filesystem allocated yet. + +What: /sys/fs/ext4//lifetime_write_kbytes +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + This file is read-only and shows the number of kilobytes + of data that have been written to this filesystem since it was + created. + +What: /sys/fs/ext4//session_write_kbytes +Date: March 2008 +Contact: "Theodore Ts'o" +Description: + This file is read-only and shows the number of + kilobytes of data that have been written to this + filesystem since it was mounted. + +What: /sys/fs/ext4//inode_goal +Date: June 2008 +Contact: "Theodore Ts'o" +Description: + Tuning parameter which (if non-zero) controls the goal + inode used by the inode allocator in preference to + all other allocation heuristics. This is intended for + debugging use only, and should be 0 on production + systems. + +What: /sys/fs/ext4//max_writeback_mb_bump +Date: September 2009 +Contact: "Theodore Ts'o" +Description: + The maximum number of megabytes the writeback code will + try to write out before move on to another inode. + +What: /sys/fs/ext4//extent_max_zeroout_kb +Date: August 2012 +Contact: "Theodore Ts'o" +Description: + The maximum number of kilobytes which will be zeroed + out in preference to creating a new uninitialized + extent when manipulating an inode's extent tree. Note + that using a larger value will increase the + variability of time necessary to complete a random + write operation (since a 4k random write might turn + into a much larger write due to the zeroout + operation). + +What: /sys/fs/ext4//journal_task +Date: February 2019 +Contact: "Theodore Ts'o" +Description: + This file is read-only and shows the pid of journal thread in + current pid-namespace or 0 if task is unreachable. -- cgit v1.3.1