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, 264 insertions, 0 deletions

diff --git a/crypto/deflate.c b/crypto/deflate.c
new file mode 100644
index 000000000..710ebba7c
--- /dev/null
+++ b/crypto/deflate.c
@@ -0,0 +1,264 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Cryptographic API.
+ *
+ * Deflate algorithm (RFC 1951), implemented here primarily for use
+ * by IPCOMP (RFC 3173 & RFC 2394).
+ *
+ * Copyright (c) 2003 James Morris <jmorris@intercode.com.au>
+ * Copyright (c) 2023 Google, LLC. <ardb@kernel.org>
+ * Copyright (c) 2025 Herbert Xu <herbert@gondor.apana.org.au>
+ */
+#include <crypto/internal/acompress.h>
+#include <crypto/scatterwalk.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/overflow.h>
+#include <linux/percpu.h>
+#include <linux/scatterlist.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/zlib.h>
+
+#define DEFLATE_DEF_LEVEL		Z_DEFAULT_COMPRESSION
+#define DEFLATE_DEF_WINBITS		11
+#define DEFLATE_DEF_MEMLEVEL		MAX_MEM_LEVEL
+
+struct deflate_stream {
+	struct z_stream_s stream;
+	u8 workspace[];
+};
+
+static DEFINE_MUTEX(deflate_stream_lock);
+
+static void *deflate_alloc_stream(void)
+{
+	size_t size = max(zlib_inflate_workspacesize(),
+			  zlib_deflate_workspacesize(-DEFLATE_DEF_WINBITS,
+						     DEFLATE_DEF_MEMLEVEL));
+	struct deflate_stream *ctx;
+
+	ctx = kvmalloc_flex(*ctx, workspace, size);
+	if (!ctx)
+		return ERR_PTR(-ENOMEM);
+
+	ctx->stream.workspace = ctx->workspace;
+
+	return ctx;
+}
+
+static void deflate_free_stream(void *ctx)
+{
+	kvfree(ctx);
+}
+
+static struct crypto_acomp_streams deflate_streams = {
+	.alloc_ctx = deflate_alloc_stream,
+	.free_ctx = deflate_free_stream,
+};
+
+static int deflate_compress_one(struct acomp_req *req,
+				struct deflate_stream *ds)
+{
+	struct z_stream_s *stream = &ds->stream;
+	struct acomp_walk walk;
+	int ret;
+
+	ret = acomp_walk_virt(&walk, req, true);
+	if (ret)
+		return ret;
+
+	do {
+		unsigned int dcur;
+
+		dcur = acomp_walk_next_dst(&walk);
+		if (!dcur)
+			return -ENOSPC;
+
+		stream->avail_out = dcur;
+		stream->next_out = walk.dst.virt.addr;
+
+		do {
+			int flush = Z_FINISH;
+			unsigned int scur;
+
+			stream->avail_in = 0;
+			stream->next_in = NULL;
+
+			scur = acomp_walk_next_src(&walk);
+			if (scur) {
+				if (acomp_walk_more_src(&walk, scur))
+					flush = Z_NO_FLUSH;
+				stream->avail_in = scur;
+				stream->next_in = walk.src.virt.addr;
+			}
+
+			ret = zlib_deflate(stream, flush);
+
+			if (scur) {
+				scur -= stream->avail_in;
+				acomp_walk_done_src(&walk, scur);
+			}
+		} while (ret == Z_OK && stream->avail_out);
+
+		acomp_walk_done_dst(&walk, dcur);
+	} while (ret == Z_OK);
+
+	if (ret != Z_STREAM_END)
+		return -EINVAL;
+
+	req->dlen = stream->total_out;
+	return 0;
+}
+
+static int deflate_compress(struct acomp_req *req)
+{
+	struct crypto_acomp_stream *s;
+	struct deflate_stream *ds;
+	int err;
+
+	s = crypto_acomp_lock_stream_bh(&deflate_streams);
+	ds = s->ctx;
+
+	err = zlib_deflateInit2(&ds->stream, DEFLATE_DEF_LEVEL, Z_DEFLATED,
+				-DEFLATE_DEF_WINBITS, DEFLATE_DEF_MEMLEVEL,
+				Z_DEFAULT_STRATEGY);
+	if (err != Z_OK) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = deflate_compress_one(req, ds);
+
+out:
+	crypto_acomp_unlock_stream_bh(s);
+
+	return err;
+}
+
+static int deflate_decompress_one(struct acomp_req *req,
+				  struct deflate_stream *ds)
+{
+	struct z_stream_s *stream = &ds->stream;
+	bool out_of_space = false;
+	struct acomp_walk walk;
+	int ret;
+
+	ret = acomp_walk_virt(&walk, req, true);
+	if (ret)
+		return ret;
+
+	do {
+		unsigned int scur;
+
+		stream->avail_in = 0;
+		stream->next_in = NULL;
+
+		scur = acomp_walk_next_src(&walk);
+		if (scur) {
+			stream->avail_in = scur;
+			stream->next_in = walk.src.virt.addr;
+		}
+
+		do {
+			unsigned int dcur;
+			unsigned long avail_in;
+
+			dcur = acomp_walk_next_dst(&walk);
+
+			stream->avail_out = dcur;
+			stream->next_out = walk.dst.virt.addr;
+			avail_in = stream->avail_in;
+
+			ret = zlib_inflate(stream, Z_NO_FLUSH);
+
+			if (!dcur && avail_in == stream->avail_in) {
+				out_of_space = true;
+				break;
+			}
+
+			dcur -= stream->avail_out;
+			acomp_walk_done_dst(&walk, dcur);
+		} while (ret == Z_OK && stream->avail_in);
+
+		if (scur)
+			acomp_walk_done_src(&walk, scur);
+
+		if (out_of_space)
+			return -ENOSPC;
+	} while (ret == Z_OK);
+
+	if (ret != Z_STREAM_END)
+		return -EINVAL;
+
+	req->dlen = stream->total_out;
+	return 0;
+}
+
+static int deflate_decompress(struct acomp_req *req)
+{
+	struct crypto_acomp_stream *s;
+	struct deflate_stream *ds;
+	int err;
+
+	s = crypto_acomp_lock_stream_bh(&deflate_streams);
+	ds = s->ctx;
+
+	err = zlib_inflateInit2(&ds->stream, -DEFLATE_DEF_WINBITS);
+	if (err != Z_OK) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = deflate_decompress_one(req, ds);
+
+out:
+	crypto_acomp_unlock_stream_bh(s);
+
+	return err;
+}
+
+static int deflate_init(struct crypto_acomp *tfm)
+{
+	int ret;
+
+	mutex_lock(&deflate_stream_lock);
+	ret = crypto_acomp_alloc_streams(&deflate_streams);
+	mutex_unlock(&deflate_stream_lock);
+
+	return ret;
+}
+
+static struct acomp_alg acomp = {
+	.compress		= deflate_compress,
+	.decompress		= deflate_decompress,
+	.init			= deflate_init,
+	.base.cra_name		= "deflate",
+	.base.cra_driver_name	= "deflate-generic",
+	.base.cra_flags		= CRYPTO_ALG_REQ_VIRT,
+	.base.cra_module	= THIS_MODULE,
+};
+
+static int __init deflate_mod_init(void)
+{
+	return crypto_register_acomp(&acomp);
+}
+
+static void __exit deflate_mod_fini(void)
+{
+	crypto_unregister_acomp(&acomp);
+	crypto_acomp_free_streams(&deflate_streams);
+}
+
+module_init(deflate_mod_init);
+module_exit(deflate_mod_fini);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("Deflate Compression Algorithm for IPCOMP");
+MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
+MODULE_AUTHOR("Ard Biesheuvel <ardb@kernel.org>");
+MODULE_AUTHOR("Herbert Xu <herbert@gondor.apana.org.au>");
+MODULE_ALIAS_CRYPTO("deflate");
+MODULE_ALIAS_CRYPTO("deflate-generic");