sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted

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 <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com

1 files changed, 278 insertions, 0 deletions

diff --git a/net/ceph/striper.c b/net/ceph/striper.c
new file mode 100644
index 000000000..2b8288858
--- /dev/null
+++ b/net/ceph/striper.c
@@ -0,0 +1,278 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#include <linux/ceph/ceph_debug.h>
+
+#include <linux/math64.h>
+#include <linux/slab.h>
+
+#include <linux/ceph/striper.h>
+#include <linux/ceph/types.h>
+
+/*
+ * Map a file extent to a stripe unit within an object.
+ * Fill in objno, offset into object, and object extent length (i.e. the
+ * number of bytes mapped, less than or equal to @l->stripe_unit).
+ *
+ * Example for stripe_count = 3, stripes_per_object = 4:
+ *
+ * blockno   |  0  3  6  9 |  1  4  7 10 |  2  5  8 11 | 12 15 18 21 | 13 16 19
+ * stripeno  |  0  1  2  3 |  0  1  2  3 |  0  1  2  3 |  4  5  6  7 |  4  5  6
+ * stripepos |      0      |      1      |      2      |      0      |      1
+ * objno     |      0      |      1      |      2      |      3      |      4
+ * objsetno  |                    0                    |                    1
+ */
+void ceph_calc_file_object_mapping(struct ceph_file_layout *l,
+				   u64 off, u64 len,
+				   u64 *objno, u64 *objoff, u32 *xlen)
+{
+	u32 stripes_per_object = l->object_size / l->stripe_unit;
+	u64 blockno;	/* which su in the file (i.e. globally) */
+	u32 blockoff;	/* offset into su */
+	u64 stripeno;	/* which stripe */
+	u32 stripepos;	/* which su in the stripe,
+			   which object in the object set */
+	u64 objsetno;	/* which object set */
+	u32 objsetpos;	/* which stripe in the object set */
+
+	blockno = div_u64_rem(off, l->stripe_unit, &blockoff);
+	stripeno = div_u64_rem(blockno, l->stripe_count, &stripepos);
+	objsetno = div_u64_rem(stripeno, stripes_per_object, &objsetpos);
+
+	*objno = objsetno * l->stripe_count + stripepos;
+	*objoff = objsetpos * l->stripe_unit + blockoff;
+	*xlen = min_t(u64, len, l->stripe_unit - blockoff);
+}
+EXPORT_SYMBOL(ceph_calc_file_object_mapping);
+
+/*
+ * Return the last extent with given objno (@object_extents is sorted
+ * by objno).  If not found, return NULL and set @add_pos so that the
+ * new extent can be added with list_add(add_pos, new_ex).
+ */
+static struct ceph_object_extent *
+lookup_last(struct list_head *object_extents, u64 objno,
+	    struct list_head **add_pos)
+{
+	struct list_head *pos;
+
+	list_for_each_prev(pos, object_extents) {
+		struct ceph_object_extent *ex =
+		    list_entry(pos, typeof(*ex), oe_item);
+
+		if (ex->oe_objno == objno)
+			return ex;
+
+		if (ex->oe_objno < objno)
+			break;
+	}
+
+	*add_pos = pos;
+	return NULL;
+}
+
+static struct ceph_object_extent *
+lookup_containing(struct list_head *object_extents, u64 objno,
+		  u64 objoff, u32 xlen)
+{
+	struct ceph_object_extent *ex;
+
+	list_for_each_entry(ex, object_extents, oe_item) {
+		if (ex->oe_objno == objno &&
+		    ex->oe_off <= objoff &&
+		    ex->oe_off + ex->oe_len >= objoff + xlen) /* paranoia */
+			return ex;
+
+		if (ex->oe_objno > objno)
+			break;
+	}
+
+	return NULL;
+}
+
+/*
+ * Map a file extent to a sorted list of object extents.
+ *
+ * We want only one (or as few as possible) object extents per object.
+ * Adjacent object extents will be merged together, each returned object
+ * extent may reverse map to multiple different file extents.
+ *
+ * Call @alloc_fn for each new object extent and @action_fn for each
+ * mapped stripe unit, whether it was merged into an already allocated
+ * object extent or started a new object extent.
+ *
+ * Newly allocated object extents are added to @object_extents.
+ * To keep @object_extents sorted, successive calls to this function
+ * must map successive file extents (i.e. the list of file extents that
+ * are mapped using the same @object_extents must be sorted).
+ *
+ * The caller is responsible for @object_extents.
+ */
+int ceph_file_to_extents(struct ceph_file_layout *l, u64 off, u64 len,
+			 struct list_head *object_extents,
+			 struct ceph_object_extent *alloc_fn(void *arg),
+			 void *alloc_arg,
+			 ceph_object_extent_fn_t action_fn,
+			 void *action_arg)
+{
+	struct ceph_object_extent *last_ex, *ex;
+
+	while (len) {
+		struct list_head *add_pos = NULL;
+		u64 objno, objoff;
+		u32 xlen;
+
+		ceph_calc_file_object_mapping(l, off, len, &objno, &objoff,
+					      &xlen);
+
+		last_ex = lookup_last(object_extents, objno, &add_pos);
+		if (!last_ex || last_ex->oe_off + last_ex->oe_len != objoff) {
+			ex = alloc_fn(alloc_arg);
+			if (!ex)
+				return -ENOMEM;
+
+			ex->oe_objno = objno;
+			ex->oe_off = objoff;
+			ex->oe_len = xlen;
+			if (action_fn)
+				action_fn(ex, xlen, action_arg);
+
+			if (!last_ex)
+				list_add(&ex->oe_item, add_pos);
+			else
+				list_add(&ex->oe_item, &last_ex->oe_item);
+		} else {
+			last_ex->oe_len += xlen;
+			if (action_fn)
+				action_fn(last_ex, xlen, action_arg);
+		}
+
+		off += xlen;
+		len -= xlen;
+	}
+
+	for (last_ex = list_first_entry(object_extents, typeof(*ex), oe_item),
+	     ex = list_next_entry(last_ex, oe_item);
+	     &ex->oe_item != object_extents;
+	     last_ex = ex, ex = list_next_entry(ex, oe_item)) {
+		if (last_ex->oe_objno > ex->oe_objno ||
+		    (last_ex->oe_objno == ex->oe_objno &&
+		     last_ex->oe_off + last_ex->oe_len >= ex->oe_off)) {
+			WARN(1, "%s: object_extents list not sorted!\n",
+			     __func__);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(ceph_file_to_extents);
+
+/*
+ * A stripped down, non-allocating version of ceph_file_to_extents(),
+ * for when @object_extents is already populated.
+ */
+int ceph_iterate_extents(struct ceph_file_layout *l, u64 off, u64 len,
+			 struct list_head *object_extents,
+			 ceph_object_extent_fn_t action_fn,
+			 void *action_arg)
+{
+	while (len) {
+		struct ceph_object_extent *ex;
+		u64 objno, objoff;
+		u32 xlen;
+
+		ceph_calc_file_object_mapping(l, off, len, &objno, &objoff,
+					      &xlen);
+
+		ex = lookup_containing(object_extents, objno, objoff, xlen);
+		if (!ex) {
+			WARN(1, "%s: objno %llu %llu~%u not found!\n",
+			     __func__, objno, objoff, xlen);
+			return -EINVAL;
+		}
+
+		action_fn(ex, xlen, action_arg);
+
+		off += xlen;
+		len -= xlen;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(ceph_iterate_extents);
+
+/*
+ * Reverse map an object extent to a sorted list of file extents.
+ *
+ * On success, the caller is responsible for:
+ *
+ *     kfree(file_extents)
+ */
+int ceph_extent_to_file(struct ceph_file_layout *l,
+			u64 objno, u64 objoff, u64 objlen,
+			struct ceph_file_extent **file_extents,
+			u32 *num_file_extents)
+{
+	u32 stripes_per_object = l->object_size / l->stripe_unit;
+	u64 blockno;	/* which su */
+	u32 blockoff;	/* offset into su */
+	u64 stripeno;	/* which stripe */
+	u32 stripepos;	/* which su in the stripe,
+			   which object in the object set */
+	u64 objsetno;	/* which object set */
+	u32 i = 0;
+
+	if (!objlen) {
+		*file_extents = NULL;
+		*num_file_extents = 0;
+		return 0;
+	}
+
+	*num_file_extents = DIV_ROUND_UP_ULL(objoff + objlen, l->stripe_unit) -
+				     DIV_ROUND_DOWN_ULL(objoff, l->stripe_unit);
+	*file_extents = kmalloc_objs(**file_extents, *num_file_extents,
+				     GFP_NOIO);
+	if (!*file_extents)
+		return -ENOMEM;
+
+	div_u64_rem(objoff, l->stripe_unit, &blockoff);
+	while (objlen) {
+		u64 off, len;
+
+		objsetno = div_u64_rem(objno, l->stripe_count, &stripepos);
+		stripeno = div_u64(objoff, l->stripe_unit) +
+						objsetno * stripes_per_object;
+		blockno = stripeno * l->stripe_count + stripepos;
+		off = blockno * l->stripe_unit + blockoff;
+		len = min_t(u64, objlen, l->stripe_unit - blockoff);
+
+		(*file_extents)[i].fe_off = off;
+		(*file_extents)[i].fe_len = len;
+
+		blockoff = 0;
+		objoff += len;
+		objlen -= len;
+		i++;
+	}
+
+	BUG_ON(i != *num_file_extents);
+	return 0;
+}
+EXPORT_SYMBOL(ceph_extent_to_file);
+
+u64 ceph_get_num_objects(struct ceph_file_layout *l, u64 size)
+{
+	u64 period = (u64)l->stripe_count * l->object_size;
+	u64 num_periods = DIV64_U64_ROUND_UP(size, period);
+	u64 remainder_bytes;
+	u64 remainder_objs = 0;
+
+	div64_u64_rem(size, period, &remainder_bytes);
+	if (remainder_bytes > 0 &&
+	    remainder_bytes < (u64)l->stripe_count * l->stripe_unit)
+		remainder_objs = l->stripe_count -
+			    DIV_ROUND_UP_ULL(remainder_bytes, l->stripe_unit);
+
+	return num_periods * l->stripe_count - remainder_objs;
+}
+EXPORT_SYMBOL(ceph_get_num_objects);