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/netlink/specs/nlctrl.yaml | 208 ++++++++++++++++++++++++++++++++ 1 file changed, 208 insertions(+) create mode 100644 Documentation/netlink/specs/nlctrl.yaml (limited to 'Documentation/netlink/specs/nlctrl.yaml') diff --git a/Documentation/netlink/specs/nlctrl.yaml b/Documentation/netlink/specs/nlctrl.yaml new file mode 100644 index 000000000..8b4472a6a --- /dev/null +++ b/Documentation/netlink/specs/nlctrl.yaml @@ -0,0 +1,208 @@ +# SPDX-License-Identifier: ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) +--- +name: nlctrl +protocol: genetlink-legacy +uapi-header: linux/genetlink.h + +doc: | + genetlink meta-family that exposes information about all genetlink + families registered in the kernel (including itself). + +definitions: + - + name: op-flags + type: flags + enum-name: + entries: + - admin-perm + - cmd-cap-do + - cmd-cap-dump + - cmd-cap-haspol + - uns-admin-perm + - + name: attr-type + enum-name: netlink-attribute-type + type: enum + entries: + - invalid + - flag + - u8 + - u16 + - u32 + - u64 + - s8 + - s16 + - s32 + - s64 + - binary + - string + - nul-string + - nested + - nested-array + - bitfield32 + - sint + - uint + +attribute-sets: + - + name: ctrl-attrs + name-prefix: ctrl-attr- + attributes: + - + name: family-id + type: u16 + - + name: family-name + type: string + - + name: version + type: u32 + - + name: hdrsize + type: u32 + - + name: maxattr + type: u32 + - + name: ops + type: indexed-array + sub-type: nest + nested-attributes: op-attrs + - + name: mcast-groups + type: indexed-array + sub-type: nest + nested-attributes: mcast-group-attrs + - + name: policy + type: nest-type-value + type-value: [policy-id, attr-id] + nested-attributes: policy-attrs + - + name: op-policy + type: nest-type-value + type-value: [op-id] + nested-attributes: op-policy-attrs + - + name: op + type: u32 + - + name: mcast-group-attrs + name-prefix: ctrl-attr-mcast-grp- + enum-name: + attributes: + - + name: name + type: string + - + name: id + type: u32 + - + name: op-attrs + name-prefix: ctrl-attr-op- + enum-name: + attributes: + - + name: id + type: u32 + - + name: flags + type: u32 + enum: op-flags + enum-as-flags: true + - + name: policy-attrs + name-prefix: nl-policy-type-attr- + enum-name: + attributes: + - + name: type + type: u32 + enum: attr-type + - + name: min-value-s + type: s64 + - + name: max-value-s + type: s64 + - + name: min-value-u + type: u64 + - + name: max-value-u + type: u64 + - + name: min-length + type: u32 + - + name: max-length + type: u32 + - + name: policy-idx + type: u32 + - + name: policy-maxtype + type: u32 + - + name: bitfield32-mask + type: u32 + - + name: mask + type: u64 + - + name: pad + type: pad + - + name: op-policy-attrs + name-prefix: ctrl-attr-policy- + enum-name: + attributes: + - + name: do + type: u32 + - + name: dump + type: u32 + +operations: + enum-model: directional + name-prefix: ctrl-cmd- + list: + - + name: getfamily + doc: Get / dump genetlink families + attribute-set: ctrl-attrs + do: + request: + value: 3 + attributes: + - family-name + reply: &all-attrs + value: 1 + attributes: + - family-id + - family-name + - hdrsize + - maxattr + - mcast-groups + - ops + - version + dump: + reply: *all-attrs + - + name: getpolicy + doc: Get / dump genetlink policies + attribute-set: ctrl-attrs + dump: + request: + value: 10 + attributes: + - family-name + - family-id + - op + reply: + value: 10 + attributes: + - family-id + - op-policy + - policy -- cgit v1.3.1