From: "Yury Norov (NVIDIA)" <yury.norov@gmail.com>
Optimizations for grp_spread_init_one():
1. Drop most of housekeeping code in grp_spread_init_one() by using
for_each_cpu_and_andnot_from().
2. Fix Shlemiel the Painter's algorithm by adding 'sibl = cpu' line. This
improves the outer loop complexity from quadratic to linear.
3. Don't clear the nmsk because it's ignored in the caller code anyways,
and switch to non-atomic set_cpu() for irqmsk as the mask is local and
implies no concurrency.
Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
---
lib/group_cpus.c | 25 ++++++-------------------
1 file changed, 6 insertions(+), 19 deletions(-)
diff --git a/lib/group_cpus.c b/lib/group_cpus.c
index 6aae1560b796..35aba99d8cd0 100644
--- a/lib/group_cpus.c
+++ b/lib/group_cpus.c
@@ -17,27 +17,14 @@ static void grp_spread_init_one(struct cpumask *irqmsk, struct cpumask *nmsk,
const struct cpumask *siblmsk;
int cpu, sibl;
- for ( ; cpus_per_grp > 0; ) {
- cpu = cpumask_first(nmsk);
-
- /* Should not happen, but I'm too lazy to think about it */
- if (cpu >= nr_cpu_ids)
- return;
-
- cpumask_clear_cpu(cpu, nmsk);
- cpumask_set_cpu(cpu, irqmsk);
- cpus_per_grp--;
-
+ for_each_cpu(cpu, nmsk) {
/* If the cpu has siblings, use them first */
siblmsk = topology_sibling_cpumask(cpu);
- for (sibl = -1; cpus_per_grp > 0; ) {
- sibl = cpumask_next(sibl, siblmsk);
- if (sibl >= nr_cpu_ids)
- break;
- if (!cpumask_test_and_clear_cpu(sibl, nmsk))
- continue;
- cpumask_set_cpu(sibl, irqmsk);
- cpus_per_grp--;
+ sibl = cpu;
+ for_each_cpu_and_andnot_from(sibl, nmsk, siblmsk, irqmsk) {
+ __cpumask_set_cpu(sibl, irqmsk);
+ if (--cpus_per_grp)
+ return;
}
}
}
--
2.43.0
Hello,
kernel test robot noticed "BUG:kernel_NULL_pointer_dereference,address" on:
commit: 5df2998b7baa3fd4cb66272d1ea5625573b4a63f ("[PATCH 3/3] group_cpus: optimize grp_spread_init_one()")
url: https://github.com/intel-lab-lkp/linux/commits/Yury-Norov/bitmap-cpumask-introduce-and_andnot-search-helper-and-iterator/20250911-051023
base: https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git 8ad25ebfa70e86860559b306bbc923c7db4fcac6
patch link: https://lore.kernel.org/all/20250910210850.404834-4-yury.norov@gmail.com/
patch subject: [PATCH 3/3] group_cpus: optimize grp_spread_init_one()
in testcase: stress-ng
version: stress-ng-x86_64-665b4465f-1_20250912
with following parameters:
nr_threads: 100%
testtime: 60s
test: cpu-online
cpufreq_governor: performance
config: x86_64-rhel-9.4
compiler: gcc-14
test machine: 104 threads 2 sockets (Skylake) with 192G memory
(please refer to attached dmesg/kmsg for entire log/backtrace)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <oliver.sang@intel.com>
| Closes: https://lore.kernel.org/oe-lkp/202509192101.5d4e0282-lkp@intel.com
The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20250919/202509192101.5d4e0282-lkp@intel.com [1]
we observed the issue happens randomly in our tests.
=========================================================================================
tbox_group/testcase/rootfs/kconfig/compiler/nr_threads/testtime/test/cpufreq_governor:
lkp-skl-fpga01/stress-ng/debian-13-x86_64-20250902.cgz/x86_64-rhel-9.4/gcc-14/100%/60s/cpu-online/performance
commit:
6ade57f62d272 ("group_cpus: don't call cpumask_weight() prematurely")
5df2998b7baa3 ("group_cpus: optimize grp_spread_init_one()")
6ade57f62d272d3e 5df2998b7baa3fd4cb66272d1ea
---------------- ---------------------------
fail:runs %reproduction fail:runs
| | |
:30 43% 13:30 dmesg.BUG:kernel_NULL_pointer_dereference,address
:30 43% 13:30 dmesg.Kernel_panic-not_syncing:Fatal_exception
:30 43% 13:30 dmesg.Oops
:30 43% 13:30 dmesg.RIP:blk_mq_all_tag_iter
and sorry our bot failed to upload correct dmesg in above link [1]
attached one FYI.
[ 43.544050][ T130] BUG: kernel NULL pointer dereference, address: 0000000000000004
[ 43.552056][ T130] #PF: supervisor read access in kernel mode
[ 43.558218][ T130] #PF: error_code(0x0000) - not-present page
[ 43.564340][ T130] PGD 0 P4D 0
[ 43.567866][ T130] Oops: Oops: 0000 [#1] SMP PTI
[ 43.572855][ T130] CPU: 19 UID: 0 PID: 130 Comm: cpuhp/19 Not tainted 6.17.0-rc1-00014-g5df2998b7baa #1 VOLUNTARY
[ 43.583575][ T130] Hardware name: Intel Corporation S2600BT/S2600BT, BIOS SE5C620.86B.1D.01.0147.121320181755 12/13/2018
[ 43.594817][ T130] RIP: 0010:blk_mq_all_tag_iter+0x1a/0x230
[ 43.600779][ T130] Code: 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 66 0f 1f 00 0f 1f 44 00 00 41 57 41 56 41 55 41 54 49 89 fc 55 53 48 83 ec 50 <8b> 4f 04 48 89 34 24 48 89 54 24 08 65 48 8b 05 8a b4 48 02 48 89
[ 43.620822][ T130] RSP: 0018:ffffc9000ce5fda0 EFLAGS: 00010286
[ 43.627048][ T130] RAX: 0000604fe9627758 RBX: ffffc9000ce5fe28 RCX: 0000000000000068
[ 43.635176][ T130] RDX: ffffc9000ce5fe28 RSI: ffffffff8199e310 RDI: 0000000000000000
[ 43.643294][ T130] RBP: ffff88810dd98600 R08: ffffffff833d72a0 R09: ffff8881003c8710
[ 43.651410][ T130] R10: ffffc9000ce5fdb0 R11: ffffc9000ce5fdb8 R12: 0000000000000000
[ 43.659516][ T130] R13: 0000000000000013 R14: ffff88810dd98760 R15: ffff8897e08dbde8
[ 43.667621][ T130] FS: 0000000000000000(0000) GS:ffff88985caa3000(0000) knlGS:0000000000000000
[ 43.676681][ T130] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 43.683393][ T130] CR2: 0000000000000004 CR3: 000000303de24004 CR4: 00000000007726f0
[ 43.691491][ T130] PKRU: 55555554
[ 43.695175][ T130] Call Trace:
[ 43.698593][ T130] <TASK>
[ 43.701657][ T130] ? __call_rcu_common+0xb0/0x2f0
[ 43.707850][ T130] ? xas_load+0xd/0xf0
[ 43.712054][ T130] ? xa_load+0x58/0xb0
[ 43.716259][ T130] blk_mq_hctx_notify_offline+0xd1/0x170
[ 43.722025][ T130] ? __pfx_blk_mq_hctx_notify_offline+0x10/0x10
[ 43.728390][ T130] cpuhp_invoke_callback+0x1c7/0x370
[ 43.733805][ T130] ? __pfx_smpboot_thread_fn+0x10/0x10
[ 43.739386][ T130] cpuhp_thread_fun+0x98/0x170
[ 43.744273][ T130] smpboot_thread_fn+0xc8/0x1b0
[ 43.749254][ T130] kthread+0xe5/0x1f0
[ 43.753363][ T130] ? __pfx_kthread+0x10/0x10
[ 43.758082][ T130] ret_from_fork+0x132/0x170
[ 43.762799][ T130] ? __pfx_kthread+0x10/0x10
[ 43.767513][ T130] ret_from_fork_asm+0x1a/0x30
[ 43.772398][ T130] </TASK>
[ 43.775532][ T130] Modules linked in: intel_rapl_msr intel_rapl_common intel_uncore_frequency intel_uncore_frequency_common btrfs blake2b_generic xor raid6_pq skx_edac skx_edac_common nfit libnvdimm sd_mod sg x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel irdma kvm irqbypass ice snd_pcm ghash_clmulni_intel ahci snd_timer rapl gnss snd ast ib_uverbs nvme intel_cstate libahci ipmi_ssif drm_client_lib soundcore binfmt_misc acpi_power_meter mei_me drm_shmem_helper i2c_i801 ib_core ioatdma pcspkr nvme_core libata ipmi_si intel_uncore acpi_ipmi drm_kms_helper mei i2c_smbus intel_pch_thermal lpc_ich dca wmi ipmi_devintf ipmi_msghandler acpi_pad joydev drm fuse nfnetlink
[ 43.836729][ T130] CR2: 0000000000000004
[ 43.841005][ T130] ---[ end trace 0000000000000000 ]---
[ 43.855607][ T130] pstore: backend (erst) writing error (-28)
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
© 2016 - 2025 Red Hat, Inc.