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, 199 insertions, 0 deletions

diff --git a/scripts/container b/scripts/container
new file mode 100755
index 000000000..b05333d85
--- /dev/null
+++ b/scripts/container
@@ -0,0 +1,199 @@
+#!/usr/bin/env python3
+# SPDX-License-Identifier: GPL-2.0-only
+# Copyright (C) 2025 Guillaume Tucker
+
+"""Containerized builds"""
+
+import abc
+import argparse
+import logging
+import os
+import pathlib
+import shutil
+import subprocess
+import sys
+import uuid
+
+
+class ContainerRuntime(abc.ABC):
+    """Base class for a container runtime implementation"""
+
+    name = None  # Property defined in each implementation class
+
+    def __init__(self, args, logger):
+        self._uid = args.uid or os.getuid()
+        self._gid = args.gid or args.uid or os.getgid()
+        self._env_file = args.env_file
+        self._shell = args.shell
+        self._logger = logger
+
+    @classmethod
+    def is_present(cls):
+        """Determine whether the runtime is present on the system"""
+        return shutil.which(cls.name) is not None
+
+    @abc.abstractmethod
+    def _do_run(self, image, cmd, container_name):
+        """Runtime-specific handler to run a command in a container"""
+
+    @abc.abstractmethod
+    def _do_abort(self, container_name):
+        """Runtime-specific handler to abort a running container"""
+
+    def run(self, image, cmd):
+        """Run a command in a runtime container"""
+        container_name = str(uuid.uuid4())
+        self._logger.debug("container: %s", container_name)
+        try:
+            return self._do_run(image, cmd, container_name)
+        except KeyboardInterrupt:
+            self._logger.error("user aborted")
+            self._do_abort(container_name)
+            return 1
+
+
+class CommonRuntime(ContainerRuntime):
+    """Common logic for Docker and Podman"""
+
+    def _do_run(self, image, cmd, container_name):
+        cmdline = [self.name, 'run']
+        cmdline += self._get_opts(container_name)
+        cmdline.append(image)
+        cmdline += cmd
+        self._logger.debug('command: %s', ' '.join(cmdline))
+        return subprocess.call(cmdline)
+
+    def _get_opts(self, container_name):
+        opts = [
+            '--name', container_name,
+            '--rm',
+            '--volume', f'{pathlib.Path.cwd()}:/src',
+            '--workdir', '/src',
+        ]
+        if self._env_file:
+            opts += ['--env-file', self._env_file]
+        if self._shell:
+            opts += ['--interactive', '--tty']
+        return opts
+
+    def _do_abort(self, container_name):
+        subprocess.call([self.name, 'kill', container_name])
+
+
+class DockerRuntime(CommonRuntime):
+    """Run a command in a Docker container"""
+
+    name = 'docker'
+
+    def _get_opts(self, container_name):
+        return super()._get_opts(container_name) + [
+            '--user', f'{self._uid}:{self._gid}'
+        ]
+
+
+class PodmanRuntime(CommonRuntime):
+    """Run a command in a Podman container"""
+
+    name = 'podman'
+
+    def _get_opts(self, container_name):
+        return super()._get_opts(container_name) + [
+            '--userns', f'keep-id:uid={self._uid},gid={self._gid}',
+        ]
+
+
+class Runtimes:
+    """List of all supported runtimes"""
+
+    runtimes = [PodmanRuntime, DockerRuntime]
+
+    @classmethod
+    def get_names(cls):
+        """Get a list of all the runtime names"""
+        return list(runtime.name for runtime in cls.runtimes)
+
+    @classmethod
+    def get(cls, name):
+        """Get a single runtime class matching the given name"""
+        for runtime in cls.runtimes:
+            if runtime.name == name:
+                if not runtime.is_present():
+                    raise ValueError(f"runtime not found: {name}")
+                return runtime
+        raise ValueError(f"unknown runtime: {name}")
+
+    @classmethod
+    def find(cls):
+        """Find the first runtime present on the system"""
+        for runtime in cls.runtimes:
+            if runtime.is_present():
+                return runtime
+        raise ValueError("no runtime found")
+
+
+def _get_logger(verbose):
+    """Set up a logger with the appropriate level"""
+    logger = logging.getLogger('container')
+    handler = logging.StreamHandler()
+    handler.setFormatter(logging.Formatter(
+        fmt='[container {levelname}] {message}', style='{'
+    ))
+    logger.addHandler(handler)
+    logger.setLevel(logging.DEBUG if verbose is True else logging.INFO)
+    return logger
+
+
+def main(args):
+    """Main entry point for the container tool"""
+    logger = _get_logger(args.verbose)
+    try:
+        cls = Runtimes.get(args.runtime) if args.runtime else Runtimes.find()
+    except ValueError as ex:
+        logger.error(ex)
+        return 1
+    logger.debug("runtime: %s", cls.name)
+    logger.debug("image: %s", args.image)
+    return cls(args, logger).run(args.image, args.cmd)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        'container',
+        description="See the documentation for more details: "
+        "https://docs.kernel.org/dev-tools/container.html"
+    )
+    parser.add_argument(
+        '-e', '--env-file',
+        help="Path to an environment file to load in the container."
+    )
+    parser.add_argument(
+        '-g', '--gid',
+        help="Group ID to use inside the container."
+    )
+    parser.add_argument(
+        '-i', '--image', required=True,
+        help="Container image name."
+    )
+    parser.add_argument(
+        '-r', '--runtime', choices=Runtimes.get_names(),
+        help="Container runtime name.  If not specified, the first one found "
+        "on the system will be used i.e. Podman if present, otherwise Docker."
+    )
+    parser.add_argument(
+        '-s', '--shell', action='store_true',
+        help="Run the container in an interactive shell."
+    )
+    parser.add_argument(
+        '-u', '--uid',
+        help="User ID to use inside the container.  If the -g option is not "
+        "specified, the user ID will also be set as the group ID."
+    )
+    parser.add_argument(
+        '-v', '--verbose', action='store_true',
+        help="Enable verbose output."
+    )
+    parser.add_argument(
+        'cmd', nargs='+',
+        help="Command to run in the container"
+    )
+    sys.exit(main(parser.parse_args(sys.argv[1:])))