From a9e3760b0838299649c0d57cca44daaf40ba3c33 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 14 Sep 2026 22:12:34 -1000 Subject: sched_ext: Maintain an online cid mask in the scheduler arena Schedulers on the default cid mapping treat [0, nr_online_cids) as the online set and restart on hotplug. Schedulers that install their own mapping with scx_bpf_cid_override() have no way to learn which cids are online: the count no longer identifies members and the CPU-form cpumask is unusable from cid programs. This is an obvious hole in the cid API. Add scx_bpf_online_cmask(), a kernel-maintained cmask in the scheduler's arena, allocated alongside the per-CPU scratch masks and populated after the cid mapping is finalized and before ops.init(), for child schedulers too. The pointer stays valid through ops.exit() with no reference to take. It is the arena offset as a void pointer, the same form struct_ops arena arguments arrive in. The verifier types the void return as a scalar for the program's arena cast. The mask follows the SCX hotplug notifications: seeded from cpu_active_mask and updated before ops.cid_online/offline() runs, so it lags cpu_online_mask only inside a hotplug transition. Updates walk the scheduler list under the lock that also serializes unlinking. Reads are live, not atomic snapshots. Root initialization excludes hotplug. v2: Reworded the getter kerneldoc (Andrea Righi). Signed-off-by: Tejun Heo Reviewed-by: Andrea Righi --- tools/testing/selftests/filelock/.gitignore | 1 + tools/testing/selftests/filelock/Makefile | 5 ++ tools/testing/selftests/filelock/ofdlocks.c | 120 ++++++++++++++++++++++++++++ 3 files changed, 126 insertions(+) create mode 100644 tools/testing/selftests/filelock/.gitignore create mode 100644 tools/testing/selftests/filelock/Makefile create mode 100644 tools/testing/selftests/filelock/ofdlocks.c (limited to 'tools/testing/selftests/filelock') diff --git a/tools/testing/selftests/filelock/.gitignore b/tools/testing/selftests/filelock/.gitignore new file mode 100644 index 000000000..825e899a1 --- /dev/null +++ b/tools/testing/selftests/filelock/.gitignore @@ -0,0 +1 @@ +ofdlocks diff --git a/tools/testing/selftests/filelock/Makefile b/tools/testing/selftests/filelock/Makefile new file mode 100644 index 000000000..478e82f8b --- /dev/null +++ b/tools/testing/selftests/filelock/Makefile @@ -0,0 +1,5 @@ +# SPDX-License-Identifier: GPL-2.0 + +TEST_GEN_PROGS := ofdlocks + +include ../lib.mk diff --git a/tools/testing/selftests/filelock/ofdlocks.c b/tools/testing/selftests/filelock/ofdlocks.c new file mode 100644 index 000000000..68bac28b2 --- /dev/null +++ b/tools/testing/selftests/filelock/ofdlocks.c @@ -0,0 +1,120 @@ +// SPDX-License-Identifier: GPL-2.0 + +#define _GNU_SOURCE +#include +#include +#include +#include +#include +#include "kselftest.h" + +static int lock_set(int fd, struct flock *fl) +{ + int ret; + + fl->l_pid = 0; // needed for OFD locks + fl->l_whence = SEEK_SET; + ret = fcntl(fd, F_OFD_SETLK, fl); + if (ret) + ksft_perror("fcntl()"); + return ret; +} + +static int lock_get(int fd, struct flock *fl) +{ + int ret; + + fl->l_pid = 0; // needed for OFD locks + fl->l_whence = SEEK_SET; + ret = fcntl(fd, F_OFD_GETLK, fl); + if (ret) + ksft_perror("fcntl()"); + return ret; +} + +int main(void) +{ + int rc; + struct flock fl, fl2; + int fd = open("/tmp/aa", O_RDWR | O_CREAT | O_EXCL, 0600); + int fd2 = open("/tmp/aa", O_RDONLY); + + ksft_print_header(); + ksft_set_plan(4); + + unlink("/tmp/aa"); + assert(fd != -1); + assert(fd2 != -1); + ksft_print_msg("opened fds %i %i\n", fd, fd2); + + /* Set some read lock */ + fl.l_type = F_RDLCK; + fl.l_start = 5; + fl.l_len = 3; + rc = lock_set(fd, &fl); + ksft_test_result(rc == 0, "set OFD read lock on first fd\n"); + if (rc != 0) + ksft_finished(); + + /* Make sure read locks do not conflict on different fds. */ + fl.l_type = F_RDLCK; + fl.l_start = 5; + fl.l_len = 1; + rc = lock_get(fd2, &fl); + if (rc != 0) + ksft_finished(); + if (fl.l_type != F_UNLCK) + ksft_exit_fail_msg("read locks conflicted\n"); + + /* Make sure read/write locks do conflict on different fds. */ + fl.l_type = F_WRLCK; + fl.l_start = 5; + fl.l_len = 1; + rc = lock_get(fd2, &fl); + if (rc != 0) + ksft_finished(); + ksft_test_result(fl.l_type != F_UNLCK, + "read and write locks conflicted\n"); + if (fl.l_type == F_UNLCK) + ksft_finished(); + + /* Get info about the lock on first fd. */ + fl.l_type = F_UNLCK; + fl.l_start = 5; + fl.l_len = 1; + rc = lock_get(fd, &fl); + if (rc != 0) + ksft_exit_fail_msg("F_OFD_GETLK with F_UNLCK not supported\n"); + ksft_test_result(fl.l_type != F_UNLCK, + "F_OFD_GETLK with F_UNLCK returned lock info\n"); + if (fl.l_type == F_UNLCK) + ksft_exit_fail(); + ksft_print_msg("F_UNLCK test returns: locked, type %i pid %i len %zi\n", + fl.l_type, fl.l_pid, fl.l_len); + + /* Try the same but by locking everything by len==0. */ + fl2.l_type = F_UNLCK; + fl2.l_start = 0; + fl2.l_len = 0; + rc = lock_get(fd, &fl2); + if (rc != 0) + ksft_exit_fail_msg + ("F_OFD_GETLK with F_UNLCK not supported\n"); + ksft_test_result(memcmp(&fl, &fl2, sizeof(fl)) == 0, + "F_UNLCK with len==0 returned the same\n"); + if (memcmp(&fl, &fl2, sizeof(fl))) { + ksft_exit_fail_msg + ("F_UNLCK test returns: locked, type %i pid %i len %zi\n", + fl.l_type, fl.l_pid, fl.l_len); + } + + /* Get info about the lock on second fd - no locks on it. */ + fl.l_type = F_UNLCK; + fl.l_start = 0; + fl.l_len = 0; + lock_get(fd2, &fl); + ksft_test_result(fl.l_type == F_UNLCK, + "F_OFD_GETLK with F_UNLCK return lock info from another fd\n"); + + ksft_finished(); +} -- cgit v1.3.1