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 --- tools/testing/vsock/util.h | 119 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 119 insertions(+) create mode 100644 tools/testing/vsock/util.h (limited to 'tools/testing/vsock/util.h') diff --git a/tools/testing/vsock/util.h b/tools/testing/vsock/util.h new file mode 100644 index 000000000..bf633cde8 --- /dev/null +++ b/tools/testing/vsock/util.h @@ -0,0 +1,119 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +#ifndef UTIL_H +#define UTIL_H + +#include +#include +#include +#include + +/* All known vsock transports, see callers of vsock_core_register() */ +#define KNOWN_TRANSPORTS(x) \ + x(LOOPBACK, "loopback") \ + x(VIRTIO, "virtio") \ + x(VHOST, "vhost") \ + x(VMCI, "vmci") \ + x(HYPERV, "hvs") + +enum transport { + TRANSPORT_COUNTER_BASE = __COUNTER__ + 1, + #define x(name, symbol) \ + TRANSPORT_##name = BIT(__COUNTER__ - TRANSPORT_COUNTER_BASE), + KNOWN_TRANSPORTS(x) + TRANSPORT_NUM = __COUNTER__ - TRANSPORT_COUNTER_BASE, + #undef x +}; + +static const char * const transport_ksyms[] = { + #define x(name, symbol) " " symbol "_transport", + KNOWN_TRANSPORTS(x) + #undef x +}; + +static_assert(ARRAY_SIZE(transport_ksyms) == TRANSPORT_NUM); +static_assert(BITS_PER_TYPE(int) >= TRANSPORT_NUM); + +#define TRANSPORTS_G2H (TRANSPORT_VIRTIO | TRANSPORT_VMCI | TRANSPORT_HYPERV) +#define TRANSPORTS_H2G (TRANSPORT_VHOST | TRANSPORT_VMCI) +#define TRANSPORTS_LOCAL (TRANSPORT_LOOPBACK) + +/* Tests can either run as the client or the server */ +enum test_mode { + TEST_MODE_UNSET, + TEST_MODE_CLIENT, + TEST_MODE_SERVER +}; + +#define DEFAULT_PEER_PORT 1234 + +/* Test runner options */ +struct test_opts { + enum test_mode mode; + unsigned int peer_cid; + unsigned int peer_port; +}; + +/* A test case definition. Test functions must print failures to stderr and + * terminate with exit(EXIT_FAILURE). + */ +struct test_case { + const char *name; /* human-readable name */ + + /* Called when test mode is TEST_MODE_CLIENT */ + void (*run_client)(const struct test_opts *opts); + + /* Called when test mode is TEST_MODE_SERVER */ + void (*run_server)(const struct test_opts *opts); + + bool skip; +}; + +void init_signals(void); +unsigned int parse_cid(const char *str); +unsigned int parse_port(const char *str); +int vsock_connect_fd(int fd, unsigned int cid, unsigned int port); +int vsock_connect(unsigned int cid, unsigned int port, int type); +int vsock_accept(unsigned int cid, unsigned int port, + struct sockaddr_vm *clientaddrp, int type); +int vsock_stream_connect(unsigned int cid, unsigned int port); +int vsock_bind_try(unsigned int cid, unsigned int port, int type); +int vsock_bind(unsigned int cid, unsigned int port, int type); +int vsock_bind_connect(unsigned int cid, unsigned int port, + unsigned int bind_port, int type); +int vsock_seqpacket_connect(unsigned int cid, unsigned int port); +int vsock_stream_accept(unsigned int cid, unsigned int port, + struct sockaddr_vm *clientaddrp); +int vsock_stream_listen(unsigned int cid, unsigned int port); +int vsock_seqpacket_accept(unsigned int cid, unsigned int port, + struct sockaddr_vm *clientaddrp); +void vsock_wait_remote_close(int fd); +bool vsock_ioctl_int(int fd, unsigned long op, int expected); +bool vsock_wait_sent(int fd); +void send_buf(int fd, const void *buf, size_t len, int flags, + ssize_t expected_ret); +void recv_buf(int fd, void *buf, size_t len, int flags, ssize_t expected_ret); +void send_byte(int fd, int expected_ret, int flags); +void recv_byte(int fd, int expected_ret, int flags); +void run_tests(const struct test_case *test_cases, + const struct test_opts *opts); +void list_tests(const struct test_case *test_cases); +void skip_test(struct test_case *test_cases, size_t test_cases_len, + const char *test_id_str); +void pick_test(struct test_case *test_cases, size_t test_cases_len, + const char *test_id_str); +unsigned long hash_djb2(const void *data, size_t len); +size_t iovec_bytes(const struct iovec *iov, size_t iovnum); +unsigned long iovec_hash_djb2(const struct iovec *iov, size_t iovnum); +struct iovec *alloc_test_iovec(const struct iovec *test_iovec, int iovnum); +void free_test_iovec(const struct iovec *test_iovec, + struct iovec *iovec, int iovnum); +void setsockopt_ull_check(int fd, int level, int optname, + unsigned long long val, char const *errmsg); +void setsockopt_int_check(int fd, int level, int optname, int val, + char const *errmsg); +void setsockopt_timeval_check(int fd, int level, int optname, + struct timeval val, char const *errmsg); +void enable_so_zerocopy_check(int fd); +void enable_so_linger(int fd, int timeout); +int get_transports(void); +#endif /* UTIL_H */ -- cgit v1.3.1