hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I9ER2W

--------------------------------

When offlining cpus, it holds cpu_hotplug_lock and call cpuset_hotplug_workfn asynchronously, which holds and releases cpuset_mutex repeatly to update cpusets, and it will release cpu_hotplug_lock before cpuset_hotplug_workfn finish. It means that some interfaces like cpuset_write_resmask holding two locks may rebuild scheduler domains when some cpusets are not refreshed, which may lead to generate domains with offlining cpus and will panic.

As commit 406100f3da08 ("cpuset: fix race between hotplug work and later CPU offline") mentioned. This problem happen in cgroup v2:

This problem can also happen in cgroup v1 pressure test, which onlines and offlines cpus, and sets cpuset.cpus to rebuild domains with sched_load_balance off.

smpboot: CPU 1 is now offline BUG: unable to handle page fault for address: 0000342bad598668 PGD 0 P4D 0 Oops: 0000 [#1] PREEMPT SMP NOPTI CPU: 54 PID: 9966 Comm: sh Kdump: loaded Tainted: G S OE 6.8.0-rc7+ #5 Hardware name: Powerleader PR2715P3/T1DM-E2, BIOS OEM57.22 11/23/2021 RIP: 0010:sd_init+0x204/0x390 Code: 00 02 00 00 a8 80 0f 84 ad 00 00 00 41 c7 44 24 2c 6e 00 00 00 85 d2 74 2f 48 63 54 24 18 49 8b 45 20 48 8b 14 d5 60 0b 79 ab <48> 8b 04 02 49 89 84 24 10 01 00 00 f0 ff 00 49 8b 84 24 10 01 00 RSP: 0018:ffff9f90669afc18 EFLAGS: 00010206 RAX: 0000342c00e62668 RBX: 0000000000000000 RCX: 00000000ffffffa0 RDX: ffffffffac736000 RSI: 0000000000000060 RDI: ffff8b24b8143930 RBP: ffff8b24b8143920 R08: 0000000000000060 R09: ffff8b2635783670 R10: 0000000000000001 R11: 0000000000000001 R12: ffff8b24b8143800 R13: ffff8ae687511600 R14: ffffffffabc6b9b0 R15: ffff8b25c0040ea0 FS: 00007fd93c734740(0000) GS:ffff8b24bca00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000342bad598668 CR3: 00000001e0934005 CR4: 00000000007706f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: <TASK> ? __die+0x24/0x70 ? page_fault_oops+0x82/0x150 ? local_clock_noinstr+0xf/0xb0 ? exc_page_fault+0x69/0x1b0 ? asm_exc_page_fault+0x26/0x30 ? sd_init+0x204/0x390 ? sd_init+0x11c/0x390 build_sched_domains+0x171/0x640 partition_sched_domains_locked+0x2a1/0x3c0 rebuild_sched_domains_locked+0x14f/0x200 update_cpumask+0x27c/0x5f0 cpuset_write_resmask+0x423/0x530 kernfs_fop_write_iter+0x160/0x220 vfs_write+0x355/0x480 ksys_write+0x69/0xf0 do_syscall_64+0x66/0x180 entry_SYSCALL_64_after_hwframe+0x6e/0x76 RIP: 0033:0x7fd93c4fc907 Code: 0f 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24 RSP: 002b:00007fff153c7f08 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 0000000000000002 RCX: 00007fd93c4fc907 RDX: 0000000000000002 RSI: 000055d055679900 RDI: 0000000000000001 RBP: 000055d055679900 R08: 00007fd93c5b01e0 R09: 00007fd93c5b0260 R10: 00007fd93c5b0160 R11: 0000000000000246 R12: 0000000000000002 R13: 00007fd93c5f25a0 R14: 0000000000000002 R15: 00007fd93c5f27a0

It must guarantee that cpus in domains passing to partition_and_rebuild_sched_domains must be active. So the domains should be checked after generate_sched_domains.

Fixes: 0ccea8feb980 ("cpuset: Make generate_sched_domains() work with partition") Signed-off-by: Chen Ridong chenridong@huawei.com --- kernel/cgroup/cpuset.c | 7 +++++++ 1 file changed, 7 insertions(+)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c index 94c166d2cc6f..038efca71f28 100644 --- a/kernel/cgroup/cpuset.c +++ b/kernel/cgroup/cpuset.c @@ -1122,6 +1122,7 @@ static void rebuild_sched_domains_locked(void) cpumask_var_t *doms; struct cpuset *cs; int ndoms; + int i;

lockdep_assert_cpus_held(); lockdep_assert_held(&cpuset_mutex); @@ -1163,6 +1164,12 @@ static void rebuild_sched_domains_locked(void) /* Generate domain masks and attrs */ ndoms = generate_sched_domains(&doms, &attr);

+ /* guarantee no CPU offlining in doms */ + for (i = 0; i < ndoms; ++i) { + if (doms && !cpumask_subset(doms[i], cpu_active_mask)) + return; + } + /* Have scheduler rebuild the domains */ partition_and_rebuild_sched_domains(ndoms, doms, attr); }