When a cpuset isolated partition is created / updated or destroyed, the
interrupt threads are affine blindly to all the non-isolated CPUs. And this
happens without taking into account the interrupt threads initial affinity
that becomes ignored.
For example in a system with 8 CPUs, if an interrupt and its kthread are
initially affine to CPU 5, creating an isolated partition with only CPU 2
inside will eventually end up affining the interrupt kthread to all CPUs
but CPU 2 (that is CPUs 0,1,3-7), losing the kthread preference for CPU 5.
Besides the blind re-affinity, this doesn't take care of the actual low
level interrupt which isn't migrated. As of today the only way to isolate
non managed interrupts, along with their kthreads, is to overwrite their
affinity separately, for example through /proc/irq/
To avoid doing that manually, future development should focus on updating
the interrupt's affinity whenever cpuset isolated partitions are updated.
In the meantime, cpuset shouldn't fiddle with interrupt threads directly.
To prevent from that, set the PF_NO_SETAFFINITY flag to them.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251118143052.68778-2-frederic@kernel.org
---
kernel/irq/manage.c | 23 +++++++++++++++--------
1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index c1ce30c9c3ab..98b9b8b4de27 100644
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -1408,16 +1408,23 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary)
* references an already freed task_struct.
*/
new->thread = get_task_struct(t);
+
/*
- * Tell the thread to set its affinity. This is
- * important for shared interrupt handlers as we do
- * not invoke setup_affinity() for the secondary
- * handlers as everything is already set up. Even for
- * interrupts marked with IRQF_NO_BALANCE this is
- * correct as we want the thread to move to the cpu(s)
- * on which the requesting code placed the interrupt.
+ * The affinity may not yet be available, but it will be once
+ * the IRQ will be enabled. Delay and defer the actual setting
+ * to the thread itself once it is ready to run. In the meantime,
+ * prevent it from ever being reaffined directly by cpuset or
+ * housekeeping. The proper way to do it is to reaffine the whole
+ * vector.
*/
- set_bit(IRQTF_AFFINITY, &new->thread_flags);
+ kthread_bind_mask(t, cpu_possible_mask);
+
+ /*
+ * Ensure the thread adjusts the affinity once it reaches the
+ * thread function.
+ */
+ new->thread_flags = BIT(IRQTF_AFFINITY);
+
return 0;
}
--
2.51.1Hi,
On Fri, Nov 21, 2025 at 03:34:59PM +0100, Frederic Weisbecker wrote:
> When a cpuset isolated partition is created / updated or destroyed, the
> interrupt threads are affine blindly to all the non-isolated CPUs. And this
> happens without taking into account the interrupt threads initial affinity
> that becomes ignored.
>
> For example in a system with 8 CPUs, if an interrupt and its kthread are
> initially affine to CPU 5, creating an isolated partition with only CPU 2
> inside will eventually end up affining the interrupt kthread to all CPUs
> but CPU 2 (that is CPUs 0,1,3-7), losing the kthread preference for CPU 5.
>
> Besides the blind re-affinity, this doesn't take care of the actual low
> level interrupt which isn't migrated. As of today the only way to isolate
> non managed interrupts, along with their kthreads, is to overwrite their
> affinity separately, for example through /proc/irq/
>
> To avoid doing that manually, future development should focus on updating
> the interrupt's affinity whenever cpuset isolated partitions are updated.
>
> In the meantime, cpuset shouldn't fiddle with interrupt threads directly.
> To prevent from that, set the PF_NO_SETAFFINITY flag to them.
>
> Suggested-by: Thomas Gleixner <tglx@linutronix.de>
> Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> Link: https://patch.msgid.link/20251118143052.68778-2-frederic@kernel.org
This patch triggers crashes due to lockups on systems with CONFIG_PREEMPT_RT
enabled.
Example and bisect log below. Reverting this patch fixes the problem.
Guenter
---
Linux version 6.19.0-rc1 (groeck@mars) (x86_64-linux-gcc (GCC) 14.3.0, GNU ld (GNU Binutils) 2.44) #1 SMP PREEMPT_RT Mon Dec 15 11:02:44 PST 2025
[ 66.503958]
[ 66.503982] ============================================
[ 66.503990] WARNING: possible recursive locking detected
[ 66.504114] 6.19.0-rc1 #1 Tainted: G N
[ 66.504126] --------------------------------------------
[ 66.504135] irq/22-eth0/3560 is trying to acquire lock:
[ 66.504195] ffff9eca0442ce40 (&ei_local->page_lock){+.+.}-{3:3}, at: ei_start_xmit+0x83/0x3b0
[ 66.504386]
[ 66.504386] but task is already holding lock:
[ 66.504393] ffff9eca0442ce40 (&ei_local->page_lock){+.+.}-{3:3}, at: __ei_interrupt.isra.0+0x2c/0x310
[ 66.504454]
[ 66.504454] other info that might help us debug this:
[ 66.504487] Possible unsafe locking scenario:
[ 66.504487]
[ 66.504494] CPU0
[ 66.504500] ----
[ 66.504505] lock(&ei_local->page_lock);
[ 66.504524] lock(&ei_local->page_lock);
[ 66.504542]
[ 66.504542] *** DEADLOCK ***
[ 66.504542]
[ 66.504548] May be due to missing lock nesting notation
[ 66.504548]
[ 66.504577] 11 locks held by irq/22-eth0/3560:
[ 66.504599] #0: ffff9eca0442ce40 (&ei_local->page_lock){+.+.}-{3:3}, at: __ei_interrupt.isra.0+0x2c/0x310
[ 66.504636] #1: ffffffffb8c0f260 (rcu_read_lock){....}-{1:3}, at: rt_spin_lock+0xe3/0x1d0
[ 66.504699] #2: ffffffffb8c0f260 (rcu_read_lock){....}-{1:3}, at: __local_bh_disable_ip+0x141/0x220
[ 66.504722] #3: ffffffffb8c0f260 (rcu_read_lock){....}-{1:3}, at: process_backlog+0x25/0x2b0
[ 66.504746] #4: ffffffffb8c0f260 (rcu_read_lock){....}-{1:3}, at: __neigh_update+0x3e6/0xfd0
[ 66.504767] #5: ffffffffb8b18a40 (local_bh){.+.+}-{1:3}, at: __local_bh_disable_ip+0x22/0x220
[ 66.504809] #6: ffffffffb8c0f220 (rcu_read_lock_bh){....}-{1:3}, at: __dev_queue_xmit+0x6e/0x15a0
[ 66.504830] #7: ffff9eca064d4328 (dev->qdisc_tx_busylock ?: &qdisc_tx_busylock){+...}-{3:3}, at: __dev_queue_xmit+0xf7a/0x15a0
[ 66.504853] #8: ffffffffb8c0f260 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x52/0x110
[ 66.504873] #9: ffff9eca04473958 (_xmit_ETHER#2){+...}-{3:3}, at: sch_direct_xmit+0x118/0x2b0
[ 66.504909] #10: ffffffffb8c0f260 (rcu_read_lock){....}-{1:3}, at: rt_spin_lock+0xe3/0x1d0
[ 66.504939]
[ 66.504939] stack backtrace:
[ 66.505132] CPU: 1 UID: 0 PID: 3560 Comm: irq/22-eth0 Tainted: G N 6.19.0-rc1 #1 PREEMPT_{RT,(full)}
[ 66.505194] Tainted: [N]=TEST
[ 66.505210] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS edk2-stable202408-prebuilt.qemu.org 08/13/2024
[ 66.505323] Call Trace:
[ 66.505390] <TASK>
[ 66.505454] dump_stack_lvl+0x6f/0xb0
[ 66.505513] print_deadlock_bug.cold+0xbd/0xcc
[ 66.505531] __lock_acquire+0x1396/0x1d20
[ 66.505556] lock_acquire+0xce/0x2d0
[ 66.505570] ? ei_start_xmit+0x83/0x3b0
[ 66.505590] ? __lock_acquire+0x7a3/0x1d20
[ 66.505605] rt_spin_lock+0x3a/0x1d0
[ 66.505615] ? ei_start_xmit+0x83/0x3b0
[ 66.505632] ei_start_xmit+0x83/0x3b0
[ 66.505659] dev_hard_start_xmit+0x67/0x210
[ 66.505676] sch_direct_xmit+0x142/0x2b0
[ 66.505696] __dev_queue_xmit+0xfe6/0x15a0
[ 66.505711] ? lock_acquire+0xce/0x2d0
[ 66.505719] ? __neigh_update+0x310/0xfd0
[ 66.505729] ? find_held_lock+0x2b/0x80
[ 66.505745] ? mark_held_locks+0x40/0x70
[ 66.505756] ? eth_header+0x25/0xb0
[ 66.505772] ? __neigh_update+0x3e6/0xfd0
[ 66.505783] __neigh_update+0x310/0xfd0
[ 66.505808] arp_process+0x2e5/0xb50
[ 66.505832] ? process_backlog+0x25/0x2b0
[ 66.505846] ? rt_mutex_slowunlock+0x3ca/0x430
[ 66.505861] ? process_backlog+0x25/0x2b0
[ 66.505870] __netif_receive_skb_one_core+0x87/0x90
[ 66.505887] process_backlog+0x3c/0x2b0
[ 66.505904] __napi_poll.constprop.0+0x25/0x1b0
[ 66.505919] net_rx_action+0x308/0x410
[ 66.505965] handle_softirqs.isra.0+0xac/0x310
[ 66.505985] __local_bh_enable_ip+0xab/0x170
[ 66.505997] netif_rx+0x116/0x180
[ 66.506007] ei_receive+0x290/0x2f0
[ 66.506027] __ei_interrupt.isra.0+0x1d0/0x310
[ 66.506045] ? irq_thread+0xc1/0x2c0
[ 66.506055] irq_thread_fn+0x1e/0x60
[ 66.506068] irq_thread+0x1a3/0x2c0
[ 66.506078] ? __pfx_irq_thread_fn+0x10/0x10
[ 66.506090] ? __pfx_irq_thread_dtor+0x10/0x10
[ 66.506104] ? __pfx_irq_thread+0x10/0x10
[ 66.506115] kthread+0x108/0x230
[ 66.506129] ? __pfx_kthread+0x10/0x10
[ 66.506144] ret_from_fork+0x248/0x290
[ 66.506154] ? __pfx_kthread+0x10/0x10
[ 66.506166] ret_from_fork_asm+0x1a/0x30
[ 66.506213] </TASK>
Network interface test failed
TPM selftest failed
[ 67.609615] exFAT-fs (sda2): invalid boot record signature
[ 67.609656] exFAT-fs (sda2): failed to read boot sector
[ 67.609683] exFAT-fs (sda2): failed to recognize exfat type
[ 67.926789] ntfs3(sda2): Primary boot signature is not NTFS.
[ 67.927057] ntfs3(sda2): try to read out of volume at offset 0x3fffe00
[ 67.990642] NILFS (sda2): segctord starting. Construction interval = 5 seconds, CP frequency < 30 seconds
[ 68.378765] random: crng init done
[ 92.888600] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
[ 92.888748] rcu: Tasks blocked on level-0 rcu_node (CPUs 0-3): P3560/6:b..l
[ 92.889076] rcu: (detected by 1, t=26002 jiffies, g=18041, q=1569 ncpus=4)
[ 92.889337] task:irq/22-eth0 state:D stack:12048 pid:3560 tgid:3560 ppid:2 task_flags:0x4208040 flags:0x00080000
[ 92.889517] Call Trace:
[ 92.889604] <TASK>
[ 92.889630] __schedule+0x5d4/0x1110
[ 92.889691] schedule_rtlock+0x1e/0x40
[ 92.889707] rtlock_slowlock_locked+0x61e/0x1bf0
[ 92.889761] rt_spin_lock+0x9b/0x1d0
[ 92.889786] ei_start_xmit+0x83/0x3b0
[ 92.889827] dev_hard_start_xmit+0x67/0x210
[ 92.889853] sch_direct_xmit+0x142/0x2b0
[ 92.889880] __dev_queue_xmit+0xfe6/0x15a0
[ 92.889905] ? lock_acquire+0xce/0x2d0
[ 92.889918] ? __neigh_update+0x310/0xfd0
[ 92.889930] ? find_held_lock+0x2b/0x80
[ 92.889950] ? mark_held_locks+0x40/0x70
[ 92.889965] ? eth_header+0x25/0xb0
[ 92.889988] ? __neigh_update+0x3e6/0xfd0
[ 92.890004] __neigh_update+0x310/0xfd0
[ 92.890037] arp_process+0x2e5/0xb50
[ 92.890070] ? process_backlog+0x25/0x2b0
[ 92.890088] ? rt_mutex_slowunlock+0x3ca/0x430
[ 92.890108] ? process_backlog+0x25/0x2b0
[ 92.890119] __netif_receive_skb_one_core+0x87/0x90
[ 92.890138] process_backlog+0x3c/0x2b0
[ 92.890162] __napi_poll.constprop.0+0x25/0x1b0
[ 92.890188] net_rx_action+0x308/0x410
[ 92.890252] handle_softirqs.isra.0+0xac/0x310
[ 92.890280] __local_bh_enable_ip+0xab/0x170
[ 92.890295] netif_rx+0x116/0x180
[ 92.890310] ei_receive+0x290/0x2f0
[ 92.890343] __ei_interrupt.isra.0+0x1d0/0x310
[ 92.890365] ? irq_thread+0xc1/0x2c0
[ 92.890383] irq_thread_fn+0x1e/0x60
[ 92.890402] irq_thread+0x1a3/0x2c0
[ 92.890416] ? __pfx_irq_thread_fn+0x10/0x10
[ 92.890431] ? __pfx_irq_thread_dtor+0x10/0x10
[ 92.890449] ? __pfx_irq_thread+0x10/0x10
[ 92.890463] kthread+0x108/0x230
[ 92.890488] ? __pfx_kthread+0x10/0x10
[ 92.890511] ret_from_fork+0x248/0x290
[ 92.890527] ? __pfx_kthread+0x10/0x10
[ 92.890542] ret_from_fork_asm+0x1a/0x30
[ 92.890584] </TASK>
[ 123.205685] INFO: task pool_workqueue_:3 blocked for more than 30 seconds.
[ 123.205742] Tainted: G N 6.19.0-rc1 #1
[ 123.205763] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 123.205796] task:pool_workqueue_ state:D stack:14744 pid:3 tgid:3 ppid:2 task_flags:0x208040 flags:0x00080000
[ 123.205833] Call Trace:
[ 123.205841] <TASK>
[ 123.205861] __schedule+0x5d4/0x1110
[ 123.205911] schedule+0x35/0x130
[ 123.205926] schedule_timeout+0xd1/0x120
[ 123.205961] __wait_for_common+0xbf/0x1f0
[ 123.205975] ? __pfx_schedule_timeout+0x10/0x10
[ 123.206006] synchronize_rcu_normal+0xf8/0x2c0
[ 123.206074] ? _raw_spin_lock_irqsave+0x3f/0x60
[ 123.206088] ? lock_acquire+0x284/0x2d0
[ 123.206102] ? lock_acquire+0x284/0x2d0
[ 123.206111] ? lock_acquire+0x284/0x2d0
[ 123.206120] ? lock_release+0x204/0x2c0
[ 123.206138] synchronize_rcu_expedited+0x31b/0x560
[ 123.206173] pwq_release_workfn+0x1a6/0x270
[ 123.206197] ? kthread_worker_fn+0x47/0x340
[ 123.206207] ? __pfx_pwq_release_workfn+0x10/0x10
[ 123.206218] kthread_worker_fn+0xce/0x340
[ 123.206234] ? __pfx_kthread_worker_fn+0x10/0x10
[ 123.206255] kthread+0x108/0x230
[ 123.206274] ? __pfx_kthread+0x10/0x10
[ 123.206293] ret_from_fork+0x248/0x290
[ 123.206306] ? __pfx_kthread+0x10/0x10
[ 123.206321] ret_from_fork_asm+0x1a/0x30
[ 123.206363] </TASK>
[ 123.206393] INFO: task rcub/0:19 blocked for more than 30 seconds.
[ 123.206403] Tainted: G N 6.19.0-rc1 #1
[ 123.206409] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 123.206414] task:rcub/0 state:D stack:14376 pid:19 tgid:19 ppid:2 task_flags:0x208040 flags:0x00080000
[ 123.206438] Call Trace:
[ 123.206589] <TASK>
[ 123.206609] __schedule+0x5d4/0x1110
[ 123.206646] rt_mutex_schedule+0x2e/0x50
[ 123.206662] rt_mutex_slowlock_block.constprop.0+0x1b8/0x350
[ 123.206678] ? lock_acquire+0x284/0x2d0
[ 123.206697] __rt_mutex_slowlock_locked.constprop.0+0x113/0x320
[ 123.206738] rt_mutex_slowlock.constprop.0+0x48/0xb0
[ 123.206760] rt_mutex_lock_nested+0x85/0xa0
[ 123.206774] ? rcu_boost_kthread+0x41/0x530
[ 123.206786] rcu_boost_kthread+0x14b/0x530
[ 123.206802] ? __pfx_rcu_boost_kthread+0x10/0x10
[ 123.206817] kthread+0x108/0x230
[ 123.206834] ? __pfx_kthread+0x10/0x10
[ 123.206853] ret_from_fork+0x248/0x290
[ 123.206865] ? __pfx_kthread+0x10/0x10
[ 123.206880] ret_from_fork_asm+0x1a/0x30
[ 123.206921] </TASK>
[ 123.206990] INFO: task umount:3611 blocked for more than 30 seconds.
[ 123.207001] Tainted: G N 6.19.0-rc1 #1
[ 123.207007] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 123.207012] task:umount state:D stack:13096 pid:3611 tgid:3611 ppid:3590 task_flags:0x400100 flags:0x00080000
[ 123.207036] Call Trace:
[ 123.207041] <TASK>
[ 123.207055] __schedule+0x5d4/0x1110
[ 123.207088] schedule+0x35/0x130
[ 123.207102] schedule_timeout+0xd1/0x120
[ 123.207137] __wait_for_common+0xbf/0x1f0
[ 123.207150] ? __pfx_schedule_timeout+0x10/0x10
[ 123.207180] synchronize_rcu_normal+0xf8/0x2c0
[ 123.207220] ? check_bytes_and_report+0x56/0x100
[ 123.207234] ? lock_acquire+0x284/0x2d0
[ 123.207251] ? lock_acquire+0x284/0x2d0
[ 123.207266] ? lock_release+0x204/0x2c0
[ 123.207276] ? lock_release+0x204/0x2c0
[ 123.207297] synchronize_rcu_expedited+0x31b/0x560
[ 123.207313] ? rt_mutex_slowunlock+0x3ca/0x430
[ 123.207344] namespace_unlock+0x285/0x370
[ 123.207373] path_umount+0x186/0x580
[ 123.207401] __x64_sys_umount+0x78/0x90
[ 123.207420] do_syscall_64+0xbd/0x430
[ 123.207439] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 123.207638] RIP: 0033:0x7efdfcc7721d
[ 123.207876] RSP: 002b:00007ffc3c2d3d00 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6
[ 123.207916] RAX: ffffffffffffffda RBX: 00007efdfccf2dc0 RCX: 00007efdfcc7721d
[ 123.207932] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 00007efdfccf29e0
[ 123.207948] RBP: 00007efdfccf29f0 R08: 00007efdfccf29f0 R09: 0000000000000012
[ 123.207963] R10: 00007efdfccf29fc R11: 0000000000000246 R12: 00007efdfccf2dc0
[ 123.207977] R13: 00007efdfccf29e0 R14: 0000000000000000 R15: 00007ffc3c2d3e58
[ 123.208037] </TASK>
[ 123.208106] INFO: lockdep is turned off.
[ 153.413565] INFO: task pool_workqueue_:3 blocked for more than 60 seconds.
[ 153.413607] Tainted: G N 6.19.0-rc1 #1
[ 153.413615] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 153.413621] task:pool_workqueue_ state:D stack:14744 pid:3 tgid:3 ppid:2 task_flags:0x208040 flags:0x00080000
[ 153.413656] Call Trace:
[ 153.413665] <TASK>
[ 153.413685] __schedule+0x5d4/0x1110
[ 153.413735] schedule+0x35/0x130
[ 153.413749] schedule_timeout+0xd1/0x120
[ 153.413785] __wait_for_common+0xbf/0x1f0
[ 153.413798] ? __pfx_schedule_timeout+0x10/0x10
[ 153.413829] synchronize_rcu_normal+0xf8/0x2c0
[ 153.413878] ? _raw_spin_lock_irqsave+0x3f/0x60
[ 153.413891] ? lock_acquire+0x284/0x2d0
[ 153.413904] ? lock_acquire+0x284/0x2d0
[ 153.413914] ? lock_acquire+0x284/0x2d0
[ 153.413923] ? lock_release+0x204/0x2c0
[ 153.413940] synchronize_rcu_expedited+0x31b/0x560
[ 153.413976] pwq_release_workfn+0x1a6/0x270
[ 153.413999] ? kthread_worker_fn+0x47/0x340
[ 153.414010] ? __pfx_pwq_release_workfn+0x10/0x10
[ 153.414022] kthread_worker_fn+0xce/0x340
[ 153.414037] ? __pfx_kthread_worker_fn+0x10/0x10
[ 153.414050] kthread+0x108/0x230
[ 153.414067] ? __pfx_kthread+0x10/0x10
[ 153.414087] ret_from_fork+0x248/0x290
[ 153.414100] ? __pfx_kthread+0x10/0x10
[ 153.414115] ret_from_fork_asm+0x1a/0x30
[ 153.414157] </TASK>
[ 153.414171] INFO: task rcub/0:19 blocked for more than 60 seconds.
[ 153.414180] Tainted: G N 6.19.0-rc1 #1
[ 153.414187] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 153.414191] task:rcub/0 state:D stack:14376 pid:19 tgid:19 ppid:2 task_flags:0x208040 flags:0x00080000
[ 153.414214] Call Trace:
[ 153.414219] <TASK>
[ 153.414232] __schedule+0x5d4/0x1110
[ 153.414265] rt_mutex_schedule+0x2e/0x50
[ 153.414282] rt_mutex_slowlock_block.constprop.0+0x1b8/0x350
[ 153.414297] ? lock_acquire+0x284/0x2d0
[ 153.414316] __rt_mutex_slowlock_locked.constprop.0+0x113/0x320
[ 153.414357] rt_mutex_slowlock.constprop.0+0x48/0xb0
[ 153.414379] rt_mutex_lock_nested+0x85/0xa0
[ 153.414393] ? rcu_boost_kthread+0x41/0x530
[ 153.414405] rcu_boost_kthread+0x14b/0x530
[ 153.414422] ? __pfx_rcu_boost_kthread+0x10/0x10
[ 153.414580] kthread+0x108/0x230
[ 153.414606] ? __pfx_kthread+0x10/0x10
[ 153.414627] ret_from_fork+0x248/0x290
[ 153.414640] ? __pfx_kthread+0x10/0x10
[ 153.414655] ret_from_fork_asm+0x1a/0x30
[ 153.414696] </TASK>
[ 153.414765] INFO: task umount:3611 blocked for more than 60 seconds.
[ 153.414776] Tainted: G N 6.19.0-rc1 #1
[ 153.414782] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 153.414787] task:umount state:D stack:13096 pid:3611 tgid:3611 ppid:3590 task_flags:0x400100 flags:0x00080000
[ 153.414811] Call Trace:
[ 153.414816] <TASK>
[ 153.414830] __schedule+0x5d4/0x1110
[ 153.414864] schedule+0x35/0x130
[ 153.414878] schedule_timeout+0xd1/0x120
[ 153.414912] __wait_for_common+0xbf/0x1f0
[ 153.414926] ? __pfx_schedule_timeout+0x10/0x10
[ 153.414956] synchronize_rcu_normal+0xf8/0x2c0
[ 153.414996] ? check_bytes_and_report+0x56/0x100
[ 153.415010] ? lock_acquire+0x284/0x2d0
[ 153.415022] ? lock_acquire+0x284/0x2d0
[ 153.415035] ? lock_release+0x204/0x2c0
[ 153.415046] ? lock_release+0x204/0x2c0
[ 153.415067] synchronize_rcu_expedited+0x31b/0x560
[ 153.415083] ? rt_mutex_slowunlock+0x3ca/0x430
[ 153.415114] namespace_unlock+0x285/0x370
[ 153.415141] path_umount+0x186/0x580
[ 153.415169] __x64_sys_umount+0x78/0x90
[ 153.415188] do_syscall_64+0xbd/0x430
[ 153.415207] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 153.415221] RIP: 0033:0x7efdfcc7721d
[ 153.415235] RSP: 002b:00007ffc3c2d3d00 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6
[ 153.415247] RAX: ffffffffffffffda RBX: 00007efdfccf2dc0 RCX: 00007efdfcc7721d
[ 153.415255] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 00007efdfccf29e0
[ 153.415262] RBP: 00007efdfccf29f0 R08: 00007efdfccf29f0 R09: 0000000000000012
[ 153.415269] R10: 00007efdfccf29fc R11: 0000000000000246 R12: 00007efdfccf2dc0
[ 153.415275] R13: 00007efdfccf29e0 R14: 0000000000000000 R15: 00007ffc3c2d3e58
[ 153.415313] </TASK>
[ 153.415324] INFO: lockdep is turned off.
[ 170.893470] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
[ 170.893501] rcu: Tasks blocked on level-0 rcu_node (CPUs 0-3): P3560/6:b..l
[ 170.893534] rcu: (detected by 0, t=104007 jiffies, g=18041, q=1569 ncpus=4)
[ 170.893549] task:irq/22-eth0 state:D stack:12048 pid:3560 tgid:3560 ppid:2 task_flags:0x4208040 flags:0x00080000
[ 170.893582] Call Trace:
[ 170.893592] <TASK>
[ 170.893612] __schedule+0x5d4/0x1110
[ 170.893661] schedule_rtlock+0x1e/0x40
[ 170.893676] rtlock_slowlock_locked+0x61e/0x1bf0
[ 170.893730] rt_spin_lock+0x9b/0x1d0
[ 170.893751] ei_start_xmit+0x83/0x3b0
[ 170.893789] dev_hard_start_xmit+0x67/0x210
[ 170.893814] sch_direct_xmit+0x142/0x2b0
[ 170.893839] __dev_queue_xmit+0xfe6/0x15a0
[ 170.893859] ? lock_acquire+0xce/0x2d0
[ 170.893872] ? __neigh_update+0x310/0xfd0
[ 170.893884] ? find_held_lock+0x2b/0x80
[ 170.893904] ? mark_held_locks+0x40/0x70
[ 170.893919] ? eth_header+0x25/0xb0
[ 170.893938] ? __neigh_update+0x3e6/0xfd0
[ 170.893952] __neigh_update+0x310/0xfd0
[ 170.893984] arp_process+0x2e5/0xb50
[ 170.894007] ? process_backlog+0x25/0x2b0
[ 170.894024] ? rt_mutex_slowunlock+0x3ca/0x430
[ 170.894044] ? process_backlog+0x25/0x2b0
[ 170.894055] __netif_receive_skb_one_core+0x87/0x90
[ 170.894074] process_backlog+0x3c/0x2b0
[ 170.894096] __napi_poll.constprop.0+0x25/0x1b0
[ 170.894115] net_rx_action+0x308/0x410
[ 170.894172] handle_softirqs.isra.0+0xac/0x310
[ 170.894199] __local_bh_enable_ip+0xab/0x170
[ 170.894214] netif_rx+0x116/0x180
[ 170.894228] ei_receive+0x290/0x2f0
[ 170.894253] __ei_interrupt.isra.0+0x1d0/0x310
[ 170.894275] ? irq_thread+0xc1/0x2c0
[ 170.894289] irq_thread_fn+0x1e/0x60
[ 170.894306] irq_thread+0x1a3/0x2c0
[ 170.894319] ? __pfx_irq_thread_fn+0x10/0x10
[ 170.894335] ? __pfx_irq_thread_dtor+0x10/0x10
[ 170.894353] ? __pfx_irq_thread+0x10/0x10
[ 170.894367] kthread+0x108/0x230
[ 170.894385] ? __pfx_kthread+0x10/0x10
[ 170.894405] ret_from_fork+0x248/0x290
[ 170.894418] ? __pfx_kthread+0x10/0x10
[ 170.894432] ret_from_fork_asm+0x1a/0x30
[ 170.894474] </TASK>
[ 183.621585] INFO: task pool_workqueue_:3 blocked for more than 90 seconds.
[ 183.621630] Tainted: G N 6.19.0-rc1 #1
[ 183.621639] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 183.621645] task:pool_workqueue_ state:D stack:14744 pid:3 tgid:3 ppid:2 task_flags:0x208040 flags:0x00080000
[ 183.621688] Call Trace:
[ 183.621697] <TASK>
[ 183.621721] __schedule+0x5d4/0x1110
[ 183.621770] schedule+0x35/0x130
[ 183.621785] schedule_timeout+0xd1/0x120
[ 183.621820] __wait_for_common+0xbf/0x1f0
[ 183.621833] ? __pfx_schedule_timeout+0x10/0x10
[ 183.621864] synchronize_rcu_normal+0xf8/0x2c0
[ 183.621913] ? _raw_spin_lock_irqsave+0x3f/0x60
[ 183.621925] ? lock_acquire+0x284/0x2d0
[ 183.621939] ? lock_acquire+0x284/0x2d0
[ 183.621948] ? lock_acquire+0x284/0x2d0
[ 183.621957] ? lock_release+0x204/0x2c0
[ 183.621974] synchronize_rcu_expedited+0x31b/0x560
[ 183.622009] pwq_release_workfn+0x1a6/0x270
[ 183.622032] ? kthread_worker_fn+0x47/0x340
[ 183.622043] ? __pfx_pwq_release_workfn+0x10/0x10
[ 183.622054] kthread_worker_fn+0xce/0x340
[ 183.622070] ? __pfx_kthread_worker_fn+0x10/0x10
[ 183.622082] kthread+0x108/0x230
[ 183.622100] ? __pfx_kthread+0x10/0x10
[ 183.622119] ret_from_fork+0x248/0x290
[ 183.622132] ? __pfx_kthread+0x10/0x10
[ 183.622147] ret_from_fork_asm+0x1a/0x30
[ 183.622189] </TASK>
[ 183.622203] INFO: task rcub/0:19 blocked for more than 90 seconds.
[ 183.622212] Tainted: G N 6.19.0-rc1 #1
[ 183.622218] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 183.622223] task:rcub/0 state:D stack:14376 pid:19 tgid:19 ppid:2 task_flags:0x208040 flags:0x00080000
[ 183.622246] Call Trace:
[ 183.622250] <TASK>
[ 183.622264] __schedule+0x5d4/0x1110
[ 183.622296] rt_mutex_schedule+0x2e/0x50
[ 183.622312] rt_mutex_slowlock_block.constprop.0+0x1b8/0x350
[ 183.622327] ? lock_acquire+0x284/0x2d0
[ 183.622345] __rt_mutex_slowlock_locked.constprop.0+0x113/0x320
[ 183.622386] rt_mutex_slowlock.constprop.0+0x48/0xb0
[ 183.622408] rt_mutex_lock_nested+0x85/0xa0
[ 183.622422] ? rcu_boost_kthread+0x41/0x530
[ 183.622568] rcu_boost_kthread+0x14b/0x530
[ 183.622592] ? __pfx_rcu_boost_kthread+0x10/0x10
[ 183.622608] kthread+0x108/0x230
[ 183.622626] ? __pfx_kthread+0x10/0x10
[ 183.622645] ret_from_fork+0x248/0x290
[ 183.622662] ? __pfx_kthread+0x10/0x10
[ 183.622677] ret_from_fork_asm+0x1a/0x30
[ 183.622717] </TASK>
[ 183.622785] INFO: task umount:3611 blocked for more than 90 seconds.
[ 183.622796] Tainted: G N 6.19.0-rc1 #1
[ 183.622802] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 183.622807] task:umount state:D stack:13096 pid:3611 tgid:3611 ppid:3590 task_flags:0x400100 flags:0x00080000
[ 183.622830] Call Trace:
[ 183.622835] <TASK>
[ 183.622849] __schedule+0x5d4/0x1110
[ 183.622882] schedule+0x35/0x130
[ 183.622896] schedule_timeout+0xd1/0x120
[ 183.622930] __wait_for_common+0xbf/0x1f0
[ 183.622943] ? __pfx_schedule_timeout+0x10/0x10
[ 183.622973] synchronize_rcu_normal+0xf8/0x2c0
[ 183.623012] ? check_bytes_and_report+0x56/0x100
[ 183.623026] ? lock_acquire+0x284/0x2d0
[ 183.623037] ? lock_acquire+0x284/0x2d0
[ 183.623051] ? lock_release+0x204/0x2c0
[ 183.623061] ? lock_release+0x204/0x2c0
[ 183.623081] synchronize_rcu_expedited+0x31b/0x560
[ 183.623098] ? rt_mutex_slowunlock+0x3ca/0x430
[ 183.623128] namespace_unlock+0x285/0x370
[ 183.623155] path_umount+0x186/0x580
[ 183.623183] __x64_sys_umount+0x78/0x90
[ 183.623201] do_syscall_64+0xbd/0x430
[ 183.623221] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 183.623235] RIP: 0033:0x7efdfcc7721d
[ 183.623248] RSP: 002b:00007ffc3c2d3d00 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6
[ 183.623260] RAX: ffffffffffffffda RBX: 00007efdfccf2dc0 RCX: 00007efdfcc7721d
[ 183.623267] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 00007efdfccf29e0
[ 183.623273] RBP: 00007efdfccf29f0 R08: 00007efdfccf29f0 R09: 0000000000000012
[ 183.623280] R10: 00007efdfccf29fc R11: 0000000000000246 R12: 00007efdfccf2dc0
[ 183.623287] R13: 00007efdfccf29e0 R14: 0000000000000000 R15: 00007ffc3c2d3e58
[ 183.623324] </TASK>
[ 183.623336] INFO: lockdep is turned off.
[ 213.829537] INFO: task pool_workqueue_:3 blocked for more than 120 seconds.
[ 213.829581] Tainted: G N 6.19.0-rc1 #1
[ 213.829589] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 213.829595] task:pool_workqueue_ state:D stack:14744 pid:3 tgid:3 ppid:2 task_flags:0x208040 flags:0x00080000
[ 213.829631] Call Trace:
[ 213.829641] <TASK>
[ 213.829664] __schedule+0x5d4/0x1110
[ 213.829715] schedule+0x35/0x130
[ 213.829730] schedule_timeout+0xd1/0x120
[ 213.829765] __wait_for_common+0xbf/0x1f0
[ 213.829787] ? __pfx_schedule_timeout+0x10/0x10
[ 213.829819] synchronize_rcu_normal+0xf8/0x2c0
[ 213.829875] ? _raw_spin_lock_irqsave+0x3f/0x60
[ 213.829890] ? lock_acquire+0x284/0x2d0
[ 213.829903] ? lock_acquire+0x284/0x2d0
[ 213.829913] ? lock_acquire+0x284/0x2d0
[ 213.829922] ? lock_release+0x204/0x2c0
[ 213.829940] synchronize_rcu_expedited+0x31b/0x560
[ 213.829976] pwq_release_workfn+0x1a6/0x270
[ 213.830000] ? kthread_worker_fn+0x47/0x340
[ 213.830010] ? __pfx_pwq_release_workfn+0x10/0x10
[ 213.830023] kthread_worker_fn+0xce/0x340
[ 213.830038] ? __pfx_kthread_worker_fn+0x10/0x10
[ 213.830052] kthread+0x108/0x230
[ 213.830070] ? __pfx_kthread+0x10/0x10
[ 213.830089] ret_from_fork+0x248/0x290
[ 213.830104] ? __pfx_kthread+0x10/0x10
[ 213.830119] ret_from_fork_asm+0x1a/0x30
[ 213.830161] </TASK>
[ 213.830195] Future hung task reports are suppressed, see sysctl kernel.hung_task_warnings
[ 213.830291] INFO: lockdep is turned off.
[ 248.899451] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
[ 248.899481] rcu: Tasks blocked on level-0 rcu_node (CPUs 0-3): P3560/6:b..l
[ 248.899513] rcu: (detected by 2, t=182013 jiffies, g=18041, q=1569 ncpus=4)
[ 248.899528] task:irq/22-eth0 state:D stack:12048 pid:3560 tgid:3560 ppid:2 task_flags:0x4208040 flags:0x00080000
[ 248.899556] Call Trace:
[ 248.899565] <TASK>
[ 248.899588] __schedule+0x5d4/0x1110
[ 248.899638] schedule_rtlock+0x1e/0x40
[ 248.899653] rtlock_slowlock_locked+0x61e/0x1bf0
[ 248.899708] rt_spin_lock+0x9b/0x1d0
[ 248.899730] ei_start_xmit+0x83/0x3b0
[ 248.899769] dev_hard_start_xmit+0x67/0x210
[ 248.899794] sch_direct_xmit+0x142/0x2b0
[ 248.899820] __dev_queue_xmit+0xfe6/0x15a0
[ 248.899840] ? lock_acquire+0xce/0x2d0
[ 248.899854] ? __neigh_update+0x310/0xfd0
[ 248.899867] ? find_held_lock+0x2b/0x80
[ 248.899887] ? mark_held_locks+0x40/0x70
[ 248.899903] ? eth_header+0x25/0xb0
[ 248.899922] ? __neigh_update+0x3e6/0xfd0
[ 248.899937] __neigh_update+0x310/0xfd0
[ 248.899970] arp_process+0x2e5/0xb50
[ 248.899994] ? process_backlog+0x25/0x2b0
[ 248.900012] ? rt_mutex_slowunlock+0x3ca/0x430
[ 248.900032] ? process_backlog+0x25/0x2b0
[ 248.900043] __netif_receive_skb_one_core+0x87/0x90
[ 248.900062] process_backlog+0x3c/0x2b0
[ 248.900091] __napi_poll.constprop.0+0x25/0x1b0
[ 248.900110] net_rx_action+0x308/0x410
[ 248.900168] handle_softirqs.isra.0+0xac/0x310
[ 248.900197] __local_bh_enable_ip+0xab/0x170
[ 248.900213] netif_rx+0x116/0x180
[ 248.900227] ei_receive+0x290/0x2f0
[ 248.900253] __ei_interrupt.isra.0+0x1d0/0x310
[ 248.900276] ? irq_thread+0xc1/0x2c0
[ 248.900289] irq_thread_fn+0x1e/0x60
[ 248.900306] irq_thread+0x1a3/0x2c0
[ 248.900320] ? __pfx_irq_thread_fn+0x10/0x10
[ 248.900336] ? __pfx_irq_thread_dtor+0x10/0x10
[ 248.900355] ? __pfx_irq_thread+0x10/0x10
[ 248.900369] kthread+0x108/0x230
[ 248.900388] ? __pfx_kthread+0x10/0x10
[ 248.900407] ret_from_fork+0x248/0x290
[ 248.900420] ? __pfx_kthread+0x10/0x10
[ 248.900435] ret_from_fork_asm+0x1a/0x30
[ 248.900478] </TASK>
qemu-system-x86_64: terminating on signal 15 from pid 1344874 (/bin/bash)
---
bisect:
# bad: [559e608c46553c107dbba19dae0854af7b219400] Merge tag 'ntfs3_for_6.19' of https://github.com/Paragon-Software-Group/linux-ntfs3
# good: [7d0a66e4bb9081d75c82ec4957c50034cb0ea449] Linux 6.18
git bisect start '559e608c4655' 'v6.18'
# bad: [015e7b0b0e8e51f7321ec2aafc1d7fc0a8a5536f] Merge tag 'bpf-next-6.19' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
git bisect bad 015e7b0b0e8e51f7321ec2aafc1d7fc0a8a5536f
# bad: [2547f79b0b0cd969ae6f736890af4ebd9368cda5] Merge tag 's390-6.19-1' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
git bisect bad 2547f79b0b0cd969ae6f736890af4ebd9368cda5
# good: [63e6995005be8ceb8a1d56a18df1a1a40c28356d] Merge tag 'objtool-core-2025-12-01' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
git bisect good 63e6995005be8ceb8a1d56a18df1a1a40c28356d
# bad: [15b87bec89cb227b55b3689bf5de31b85cf88559] Merge tag 'irq-drivers-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
git bisect bad 15b87bec89cb227b55b3689bf5de31b85cf88559
# good: [4a26e7032d7d57c998598c08a034872d6f0d3945] Merge tag 'core-bugs-2025-12-01' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
git bisect good 4a26e7032d7d57c998598c08a034872d6f0d3945
# good: [2b09f480f0a1e68111ae36a7be9aa1c93e067255] Merge tag 'core-rseq-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
git bisect good 2b09f480f0a1e68111ae36a7be9aa1c93e067255
# bad: [3de5e46e50abc01a1cee7e12b657e083fc5ed638] genirq: Remove cpumask availability check on kthread affinity setting
git bisect bad 3de5e46e50abc01a1cee7e12b657e083fc5ed638
# good: [bdf4e2ac295fe77c94b570a1ad12c0882bc89b53] genirq: Allow per-cpu interrupt sharing for non-overlapping affinities
git bisect good bdf4e2ac295fe77c94b570a1ad12c0882bc89b53
# good: [ee2d50a9f524ae829d1a8ec296d7a0170e7b8ade] genirq: Kill irq_{g,s}et_percpu_devid_partition()
git bisect good ee2d50a9f524ae829d1a8ec296d7a0170e7b8ade
# good: [9ea2b810d51ae662cc5b5578f9395cb620a34a26] genirq/proc: Fix race in show_irq_affinity()
git bisect good 9ea2b810d51ae662cc5b5578f9395cb620a34a26
# good: [9d3faec60b1303fbec53d7a9b48a8c0fc5ae029b] genirq: Use raw_spinlock_irq() in irq_set_affinity_notifier()
git bisect good 9d3faec60b1303fbec53d7a9b48a8c0fc5ae029b
# bad: [801afdfbfcd90ff62a4b2469bbda1d958f7a5353] genirq: Fix interrupt threads affinity vs. cpuset isolated partitions
git bisect bad 801afdfbfcd90ff62a4b2469bbda1d958f7a5353
# good: [68775ca79af3b8d4c147598983ece012d7007bac] genirq: Prevent early spurious wake-ups of interrupt threads
git bisect good 68775ca79af3b8d4c147598983ece012d7007bac
# first bad commit: [801afdfbfcd90ff62a4b2469bbda1d958f7a5353] genirq: Fix interrupt threads affinity vs. cpuset isolated partitions
On Fri, 21 Nov 2025 15:34:59 +0100 Frederic Weisbecker <frederic@kernel.org> wrote:
Hi everyone,
I ran some recent commits through AI patch review and it flagged this one.
The comments below looks right to me, but I might be missing something:
> diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
> index c1ce30c9c3ab6..61da1c68ff82a 100644
> --- a/kernel/irq/manage.c
> +++ b/kernel/irq/manage.c
> @@ -1408,16 +1408,23 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary)
> * references an already freed task_struct.
> */
> new->thread = get_task_struct(t);
[ ... ]
> - set_bit(IRQTF_AFFINITY, &new->thread_flags);
> + kthread_bind_mask(t, cpu_possible_mask);
> +
> + /*
> + * Ensure the thread adjusts the affinity once it reaches the
> + * thread function.
> + */
> + new->thread_flags = BIT(IRQTF_AFFINITY);
^^^
Can this clobber IRQTF_FORCED_THREAD?
In __setup_irq(), irq_setup_forced_threading() is called before
setup_irq_thread(). When forced threading is enabled,
irq_setup_forced_threading() sets IRQTF_FORCED_THREAD via set_bit():
set_bit(IRQTF_FORCED_THREAD, &new->thread_flags);
Then setup_irq_thread() overwrites thread_flags with a direct assignment:
new->thread_flags = BIT(IRQTF_AFFINITY);
This clears IRQTF_FORCED_THREAD. Later in irq_thread():
if (force_irqthreads() && test_bit(IRQTF_FORCED_THREAD,
&action->thread_flags))
handler_fn = irq_forced_thread_fn;
else
handler_fn = irq_thread_fn;
Without IRQTF_FORCED_THREAD, the handler uses irq_thread_fn() instead of
irq_forced_thread_fn(). The forced-threaded handler then runs without the
local_bh_disable() and local_irq_disable() protection that non-threaded
interrupt handlers expect.
Should this be:
new->thread_flags |= BIT(IRQTF_AFFINITY);
or:
set_bit(IRQTF_AFFINITY, &new->thread_flags);
> +
> return 0;
> }
-chris
On Thu, Dec 11 2025 at 17:48, Chris Mason wrote: > On Fri, 21 Nov 2025 15:34:59 +0100 Frederic Weisbecker <frederic@kernel.org> wrote: > I ran some recent commits through AI patch review and it flagged this one. > The comments below looks right to me, but I might be missing something: > >> diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c >> index c1ce30c9c3ab6..61da1c68ff82a 100644 >> --- a/kernel/irq/manage.c >> +++ b/kernel/irq/manage.c >> @@ -1408,16 +1408,23 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary) >> * references an already freed task_struct. >> */ >> new->thread = get_task_struct(t); > > [ ... ] > >> - set_bit(IRQTF_AFFINITY, &new->thread_flags); >> + kthread_bind_mask(t, cpu_possible_mask); >> + >> + /* >> + * Ensure the thread adjusts the affinity once it reaches the >> + * thread function. >> + */ >> + new->thread_flags = BIT(IRQTF_AFFINITY); > ^^^ > > Can this clobber IRQTF_FORCED_THREAD? > > In __setup_irq(), irq_setup_forced_threading() is called before > setup_irq_thread(). When forced threading is enabled, > irq_setup_forced_threading() sets IRQTF_FORCED_THREAD via set_bit(): > > set_bit(IRQTF_FORCED_THREAD, &new->thread_flags); > > Then setup_irq_thread() overwrites thread_flags with a direct assignment: > > new->thread_flags = BIT(IRQTF_AFFINITY); > > This clears IRQTF_FORCED_THREAD. Later in irq_thread(): Yep. That's broken. Nice catch.
Chris reported that the recent affinity management changes result in
overwriting the already initialized thread flags.
Use set_bit() to set the affinity bit instead of assigning the bit value to
the flags.
Fixes: 801afdfbfcd9 ("genirq: Fix interrupt threads affinity vs. cpuset isolated partitions")
Reported-by: Chris Mason <clm@meta.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Closes: https://lore.kernel.org/all/20251212014848.3509622-1-clm@meta.com
---
kernel/irq/manage.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -1414,7 +1414,7 @@ setup_irq_thread(struct irqaction *new,
* Ensure the thread adjusts the affinity once it reaches the
* thread function.
*/
- new->thread_flags = BIT(IRQTF_AFFINITY);
+ set_bit(IRQTF_AFFINITY, &new->thread_flags);
return 0;
}Le Fri, Dec 12, 2025 at 01:01:04PM +0900, Thomas Gleixner a écrit :
> Chris reported that the recent affinity management changes result in
> overwriting the already initialized thread flags.
>
> Use set_bit() to set the affinity bit instead of assigning the bit value to
> the flags.
>
> Fixes: 801afdfbfcd9 ("genirq: Fix interrupt threads affinity vs. cpuset isolated partitions")
> Reported-by: Chris Mason <clm@meta.com>
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> Closes: https://lore.kernel.org/all/20251212014848.3509622-1-clm@meta.com
Whoops!
Acked-by: Frederic Weisbecker <frederic@kernel.org>
--
Frederic Weisbecker
SUSE Labs
The following commit has been merged into the irq/urgent branch of tip:
Commit-ID: fbbd7ce627af733ded7971b2495b0d099a0a80da
Gitweb: https://git.kernel.org/tip/fbbd7ce627af733ded7971b2495b0d099a0a80da
Author: Thomas Gleixner <tglx@linutronix.de>
AuthorDate: Fri, 12 Dec 2025 13:01:04 +09:00
Committer: Thomas Gleixner <tglx@linutronix.de>
CommitterDate: Sat, 13 Dec 2025 10:29:33 +09:00
genirq: Don't overwrite interrupt thread flags on setup
Chris reported that the recent affinity management changes result in
overwriting the already initialized thread flags.
Use set_bit() to set the affinity bit instead of assigning the bit value to
the flags.
Fixes: 801afdfbfcd9 ("genirq: Fix interrupt threads affinity vs. cpuset isolated partitions")
Reported-by: Chris Mason <clm@meta.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://patch.msgid.link/87ecp0e4cf.ffs@tglx
Closes: https://lore.kernel.org/all/20251212014848.3509622-1-clm@meta.com
---
kernel/irq/manage.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index 8b1b4c8..349ae79 100644
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -1414,7 +1414,7 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary)
* Ensure the thread adjusts the affinity once it reaches the
* thread function.
*/
- new->thread_flags = BIT(IRQTF_AFFINITY);
+ set_bit(IRQTF_AFFINITY, &new->thread_flags);
return 0;
}On 11/21/25 9:34 AM, Frederic Weisbecker wrote: > When a cpuset isolated partition is created / updated or destroyed, the > interrupt threads are affine blindly to all the non-isolated CPUs. And this > happens without taking into account the interrupt threads initial affinity > that becomes ignored. > > For example in a system with 8 CPUs, if an interrupt and its kthread are > initially affine to CPU 5, creating an isolated partition with only CPU 2 > inside will eventually end up affining the interrupt kthread to all CPUs > but CPU 2 (that is CPUs 0,1,3-7), losing the kthread preference for CPU 5. > > Besides the blind re-affinity, this doesn't take care of the actual low > level interrupt which isn't migrated. As of today the only way to isolate > non managed interrupts, along with their kthreads, is to overwrite their > affinity separately, for example through /proc/irq/ > > To avoid doing that manually, future development should focus on updating > the interrupt's affinity whenever cpuset isolated partitions are updated. > > In the meantime, cpuset shouldn't fiddle with interrupt threads directly. > To prevent from that, set the PF_NO_SETAFFINITY flag to them. > > Suggested-by: Thomas Gleixner <tglx@linutronix.de> > Signed-off-by: Frederic Weisbecker <frederic@kernel.org> > Signed-off-by: Thomas Gleixner <tglx@linutronix.de> > Link: https://patch.msgid.link/20251118143052.68778-2-frederic@kernel.org > --- > kernel/irq/manage.c | 23 +++++++++++++++-------- > 1 file changed, 15 insertions(+), 8 deletions(-) > > diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c > index c1ce30c9c3ab..98b9b8b4de27 100644 > --- a/kernel/irq/manage.c > +++ b/kernel/irq/manage.c > @@ -1408,16 +1408,23 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary) > * references an already freed task_struct. > */ > new->thread = get_task_struct(t); > + > /* > - * Tell the thread to set its affinity. This is > - * important for shared interrupt handlers as we do > - * not invoke setup_affinity() for the secondary > - * handlers as everything is already set up. Even for > - * interrupts marked with IRQF_NO_BALANCE this is > - * correct as we want the thread to move to the cpu(s) > - * on which the requesting code placed the interrupt. > + * The affinity may not yet be available, but it will be once > + * the IRQ will be enabled. Delay and defer the actual setting > + * to the thread itself once it is ready to run. In the meantime, > + * prevent it from ever being reaffined directly by cpuset or > + * housekeeping. The proper way to do it is to reaffine the whole > + * vector. > */ > - set_bit(IRQTF_AFFINITY, &new->thread_flags); > + kthread_bind_mask(t, cpu_possible_mask); > + > + /* > + * Ensure the thread adjusts the affinity once it reaches the > + * thread function. > + */ > + new->thread_flags = BIT(IRQTF_AFFINITY); > + > return 0; > } > LGTM Acked-by: Waiman Long <longman@redhat.com>
The following commit has been merged into the irq/core branch of tip:
Commit-ID: 801afdfbfcd90ff62a4b2469bbda1d958f7a5353
Gitweb: https://git.kernel.org/tip/801afdfbfcd90ff62a4b2469bbda1d958f7a5353
Author: Frederic Weisbecker <frederic@kernel.org>
AuthorDate: Fri, 21 Nov 2025 15:34:59 +01:00
Committer: Ingo Molnar <mingo@kernel.org>
CommitterDate: Sat, 22 Nov 2025 09:26:18 +01:00
genirq: Fix interrupt threads affinity vs. cpuset isolated partitions
When a cpuset isolated partition is created / updated or destroyed, the
interrupt threads are affined blindly to all the non-isolated CPUs. This
happens without taking into account the interrupt threads initial affinity
that becomes ignored.
For example in a system with 8 CPUs, if an interrupt and its kthread are
initially affine to CPU 5, creating an isolated partition with only CPU 2
inside will eventually end up affining the interrupt kthread to all CPUs
but CPU 2 (that is CPUs 0,1,3-7), losing the kthread preference for CPU 5.
Besides the blind re-affining, this doesn't take care of the actual low
level interrupt which isn't migrated. As of today the only way to isolate
non managed interrupts, along with their kthreads, is to overwrite their
affinity separately, for example through /proc/irq/
To avoid doing that manually, future development should focus on updating
the interrupt's affinity whenever cpuset isolated partitions are updated.
In the meantime, cpuset shouldn't fiddle with interrupt threads directly.
To prevent from that, set the PF_NO_SETAFFINITY flag to them.
This is done through kthread_bind_mask() by affining them initially to all
possible CPUs as at that point the interrupt is not started up which means
the affinity of the hard interrupt is not known. The thread will adjust
that once it reaches the handler, which is guaranteed to happen after the
initial affinity of the hard interrupt is established.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://patch.msgid.link/20251121143500.42111-3-frederic@kernel.org
---
kernel/irq/manage.c | 23 +++++++++++++++--------
1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index c1ce30c..61da1c6 100644
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -1408,16 +1408,23 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary)
* references an already freed task_struct.
*/
new->thread = get_task_struct(t);
+
/*
- * Tell the thread to set its affinity. This is
- * important for shared interrupt handlers as we do
- * not invoke setup_affinity() for the secondary
- * handlers as everything is already set up. Even for
- * interrupts marked with IRQF_NO_BALANCE this is
- * correct as we want the thread to move to the cpu(s)
- * on which the requesting code placed the interrupt.
+ * The affinity can not be established yet, but it will be once the
+ * interrupt is enabled. Delay and defer the actual setting to the
+ * thread itself once it is ready to run. In the meantime, prevent
+ * it from ever being re-affined directly by cpuset or
+ * housekeeping. The proper way to do it is to re-affine the whole
+ * vector.
*/
- set_bit(IRQTF_AFFINITY, &new->thread_flags);
+ kthread_bind_mask(t, cpu_possible_mask);
+
+ /*
+ * Ensure the thread adjusts the affinity once it reaches the
+ * thread function.
+ */
+ new->thread_flags = BIT(IRQTF_AFFINITY);
+
return 0;
}
The following commit has been merged into the irq/core branch of tip:
Commit-ID: 71b89ad36c06603c093f04e142972d67c9272f14
Gitweb: https://git.kernel.org/tip/71b89ad36c06603c093f04e142972d67c9272f14
Author: Frederic Weisbecker <frederic@kernel.org>
AuthorDate: Fri, 21 Nov 2025 15:34:59 +01:00
Committer: Thomas Gleixner <tglx@linutronix.de>
CommitterDate: Fri, 21 Nov 2025 20:50:30 +01:00
genirq: Fix interrupt threads affinity vs. cpuset isolated partitions
When a cpuset isolated partition is created / updated or destroyed, the
interrupt threads are affined blindly to all the non-isolated CPUs. This
happens without taking into account the interrupt threads initial affinity
that becomes ignored.
For example in a system with 8 CPUs, if an interrupt and its kthread are
initially affine to CPU 5, creating an isolated partition with only CPU 2
inside will eventually end up affining the interrupt kthread to all CPUs
but CPU 2 (that is CPUs 0,1,3-7), losing the kthread preference for CPU 5.
Besides the blind re-affining, this doesn't take care of the actual low
level interrupt which isn't migrated. As of today the only way to isolate
non managed interrupts, along with their kthreads, is to overwrite their
affinity separately, for example through /proc/irq/
To avoid doing that manually, future development should focus on updating
the interrupt's affinity whenever cpuset isolated partitions are updated.
In the meantime, cpuset shouldn't fiddle with interrupt threads directly.
To prevent from that, set the PF_NO_SETAFFINITY flag to them.
This is done through kthread_bind_mask() by affining them initially to all
possible CPUs as at that point the interrupt is not started up which means
the affinity of the hard interrupt is not known. The thread will adjust
that once it reaches the handler, which is guaranteed to happen after the
initial affinity of the hard interrupt is established.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251121143500.42111-3-frederic@kernel.org
---
kernel/irq/manage.c | 23 +++++++++++++++--------
1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index c1ce30c..61da1c6 100644
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -1408,16 +1408,23 @@ setup_irq_thread(struct irqaction *new, unsigned int irq, bool secondary)
* references an already freed task_struct.
*/
new->thread = get_task_struct(t);
+
/*
- * Tell the thread to set its affinity. This is
- * important for shared interrupt handlers as we do
- * not invoke setup_affinity() for the secondary
- * handlers as everything is already set up. Even for
- * interrupts marked with IRQF_NO_BALANCE this is
- * correct as we want the thread to move to the cpu(s)
- * on which the requesting code placed the interrupt.
+ * The affinity can not be established yet, but it will be once the
+ * interrupt is enabled. Delay and defer the actual setting to the
+ * thread itself once it is ready to run. In the meantime, prevent
+ * it from ever being re-affined directly by cpuset or
+ * housekeeping. The proper way to do it is to re-affine the whole
+ * vector.
*/
- set_bit(IRQTF_AFFINITY, &new->thread_flags);
+ kthread_bind_mask(t, cpu_possible_mask);
+
+ /*
+ * Ensure the thread adjusts the affinity once it reaches the
+ * thread function.
+ */
+ new->thread_flags = BIT(IRQTF_AFFINITY);
+
return 0;
}
© 2016 - 2026 Red Hat, Inc.