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 --- net/ceph/pagevec.c | 101 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 101 insertions(+) create mode 100644 net/ceph/pagevec.c (limited to 'net/ceph/pagevec.c') diff --git a/net/ceph/pagevec.c b/net/ceph/pagevec.c new file mode 100644 index 000000000..a6aa5b3b7 --- /dev/null +++ b/net/ceph/pagevec.c @@ -0,0 +1,101 @@ +// SPDX-License-Identifier: GPL-2.0 +#include + +#include +#include +#include +#include +#include +#include + +#include + +void ceph_release_page_vector(struct page **pages, int num_pages) +{ + int i; + + for (i = 0; i < num_pages; i++) + __free_pages(pages[i], 0); + kfree(pages); +} +EXPORT_SYMBOL(ceph_release_page_vector); + +/* + * allocate a vector new pages + */ +struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags) +{ + struct page **pages; + int i; + + pages = kmalloc_objs(*pages, num_pages, flags); + if (!pages) + return ERR_PTR(-ENOMEM); + for (i = 0; i < num_pages; i++) { + pages[i] = __page_cache_alloc(flags); + if (pages[i] == NULL) { + ceph_release_page_vector(pages, i); + return ERR_PTR(-ENOMEM); + } + } + return pages; +} +EXPORT_SYMBOL(ceph_alloc_page_vector); + +void ceph_copy_from_page_vector(struct page **pages, + void *data, + loff_t off, size_t len) +{ + int i = 0; + size_t po = off & ~PAGE_MASK; + size_t left = len; + + while (left > 0) { + size_t l = min_t(size_t, PAGE_SIZE-po, left); + + memcpy(data, page_address(pages[i]) + po, l); + data += l; + left -= l; + po += l; + if (po == PAGE_SIZE) { + po = 0; + i++; + } + } +} +EXPORT_SYMBOL(ceph_copy_from_page_vector); + +/* + * Zero an extent within a page vector. Offset is relative to the + * start of the first page. + */ +void ceph_zero_page_vector_range(int off, int len, struct page **pages) +{ + int i = off >> PAGE_SHIFT; + + off &= ~PAGE_MASK; + + dout("zero_page_vector_page %u~%u\n", off, len); + + /* leading partial page? */ + if (off) { + int end = min((int)PAGE_SIZE, off + len); + dout("zeroing %d %p head from %d\n", i, pages[i], + (int)off); + zero_user_segment(pages[i], off, end); + len -= (end - off); + i++; + } + while (len >= PAGE_SIZE) { + dout("zeroing %d %p len=%d\n", i, pages[i], len); + zero_user_segment(pages[i], 0, PAGE_SIZE); + len -= PAGE_SIZE; + i++; + } + /* trailing partial page? */ + if (len) { + dout("zeroing %d %p tail to %d\n", i, pages[i], (int)len); + zero_user_segment(pages[i], 0, len); + } +} +EXPORT_SYMBOL(ceph_zero_page_vector_range); -- cgit v1.3.1