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 --- io_uring/sync.c | 114 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 114 insertions(+) create mode 100644 io_uring/sync.c (limited to 'io_uring/sync.c') diff --git a/io_uring/sync.c b/io_uring/sync.c new file mode 100644 index 000000000..ab7fa1cd7 --- /dev/null +++ b/io_uring/sync.c @@ -0,0 +1,114 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#include "io_uring.h" +#include "sync.h" + +struct io_sync { + struct file *file; + loff_t len; + loff_t off; + int flags; + int mode; +}; + +int io_sfr_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) +{ + struct io_sync *sync = io_kiocb_to_cmd(req, struct io_sync); + + if (unlikely(sqe->addr || sqe->buf_index || sqe->splice_fd_in)) + return -EINVAL; + + sync->off = READ_ONCE(sqe->off); + sync->len = READ_ONCE(sqe->len); + sync->flags = READ_ONCE(sqe->sync_range_flags); + req->flags |= REQ_F_FORCE_ASYNC; + + return 0; +} + +int io_sync_file_range(struct io_kiocb *req, unsigned int issue_flags) +{ + struct io_sync *sync = io_kiocb_to_cmd(req, struct io_sync); + int ret; + + /* sync_file_range always requires a blocking context */ + WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK); + + ret = sync_file_range(req->file, sync->off, sync->len, sync->flags); + io_req_set_res(req, ret, 0); + return IOU_COMPLETE; +} + +int io_fsync_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) +{ + struct io_sync *sync = io_kiocb_to_cmd(req, struct io_sync); + + if (unlikely(sqe->addr || sqe->buf_index || sqe->splice_fd_in)) + return -EINVAL; + + sync->flags = READ_ONCE(sqe->fsync_flags); + if (unlikely(sync->flags & ~IORING_FSYNC_DATASYNC)) + return -EINVAL; + + sync->off = READ_ONCE(sqe->off); + if (sync->off < 0) + return -EINVAL; + sync->len = READ_ONCE(sqe->len); + req->flags |= REQ_F_FORCE_ASYNC; + return 0; +} + +int io_fsync(struct io_kiocb *req, unsigned int issue_flags) +{ + struct io_sync *sync = io_kiocb_to_cmd(req, struct io_sync); + loff_t end = sync->off + sync->len; + int ret; + + /* fsync always requires a blocking context */ + WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK); + + ret = vfs_fsync_range(req->file, sync->off, end > 0 ? end : LLONG_MAX, + sync->flags & IORING_FSYNC_DATASYNC); + io_req_set_res(req, ret, 0); + return IOU_COMPLETE; +} + +int io_fallocate_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) +{ + struct io_sync *sync = io_kiocb_to_cmd(req, struct io_sync); + + if (sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in) + return -EINVAL; + + sync->off = READ_ONCE(sqe->off); + sync->len = READ_ONCE(sqe->addr); + sync->mode = READ_ONCE(sqe->len); + req->flags |= REQ_F_FORCE_ASYNC; + return 0; +} + +int io_fallocate(struct io_kiocb *req, unsigned int issue_flags) +{ + struct io_sync *sync = io_kiocb_to_cmd(req, struct io_sync); + int ret; + + /* fallocate always requiring blocking context */ + WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK); + + ret = vfs_fallocate(req->file, sync->mode, sync->off, sync->len); + if (ret >= 0) + fsnotify_modify(req->file); + io_req_set_res(req, ret, 0); + return IOU_COMPLETE; +} -- cgit v1.3.1