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 --- arch/alpha/lib/ev67-strlen.S | 51 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 51 insertions(+) create mode 100644 arch/alpha/lib/ev67-strlen.S (limited to 'arch/alpha/lib/ev67-strlen.S') diff --git a/arch/alpha/lib/ev67-strlen.S b/arch/alpha/lib/ev67-strlen.S new file mode 100644 index 000000000..3d9078807 --- /dev/null +++ b/arch/alpha/lib/ev67-strlen.S @@ -0,0 +1,51 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * arch/alpha/lib/ev67-strlen.S + * 21264 version by Rick Gorton + * + * Finds length of a 0-terminated string. Optimized for the + * Alpha architecture: + * + * - memory accessed as aligned quadwords only + * - uses bcmpge to compare 8 bytes in parallel + * + * Much of the information about 21264 scheduling/coding comes from: + * Compiler Writer's Guide for the Alpha 21264 + * abbreviated as 'CWG' in other comments here + * ftp.digital.com/pub/Digital/info/semiconductor/literature/dsc-library.html + * Scheduling notation: + * E - either cluster + * U - upper subcluster; U0 - subcluster U0; U1 - subcluster U1 + * L - lower subcluster; L0 - subcluster L0; L1 - subcluster L1 + */ +#include + .set noreorder + .set noat + + .globl strlen + .ent strlen + .align 4 +strlen: + ldq_u $1, 0($16) # L : load first quadword ($16 may be misaligned) + lda $2, -1($31) # E : + insqh $2, $16, $2 # U : + andnot $16, 7, $0 # E : + + or $2, $1, $1 # E : + cmpbge $31, $1, $2 # E : $2 <- bitmask: bit i == 1 <==> i-th byte == 0 + nop # E : + bne $2, $found # U : + +$loop: ldq $1, 8($0) # L : + addq $0, 8, $0 # E : addr += 8 + cmpbge $31, $1, $2 # E : + beq $2, $loop # U : + +$found: + cttz $2, $3 # U0 : + addq $0, $3, $0 # E : + subq $0, $16, $0 # E : + ret $31, ($26) # L0 : + + .end strlen + EXPORT_SYMBOL(strlen) -- cgit v1.3.1