psi_enqueue() calls add_timer() with pi->lock and rq->lock held which
in-turn could allcate with debugobjcets in the locking order,
pi->lock
rq->lock
base->lock
batched_entropy_u32.lock
while the random code could always call into the scheduler via
try_to_wake_up() in the locking order,
batched_entropy_u32.lock
pi->lock
Thus, it could generate a lockdep splat below right after boot. Ideally,
psi_enqueue() might be able to be called without either pi->lock or
rq->lock held, but it is tricky to do.
Since,
1) debugobjects is only used in a debug kernel.
2) the chance to trigger a real deadlock is relative low.
3) once the splat happened, it will disable lockdep to prevent it from
catching any more important issues later.
just silent the splat by temporarily lettting lockdep ignore lockes
inside debug_timer_activate() which sounds like a reasonable tradeoff
for debug kernels.
00: [ 321.354464] WARNING: possible circular locking dependency detected
00: [ 321.354479] 5.6.0-rc5-next-20200312+ #5 Not tainted
00: [ 321.354489] ------------------------------------------------------
00: [ 321.354501] swapper/0/0 is trying to acquire lock:
00: [ 321.354512] 00000000163eaa48 (&p->pi_lock){-.-.}, at: try_to_wake_up+0xce
00: /0xd88
00: [ 321.354543]
00: [ 321.354543] but task is already holding lock:
00: [ 321.354554] 000000004e23d1b8 (random_write_wait.lock){..-.}, at: __wake_up_common_lock+0xd2/0x140
__wake_up_common_lock at kernel/sched/wait.c:123 (discriminator 1)
00: [ 321.354579]
00: [ 321.354579] which lock already depends on the new lock.
00: [ 321.354579]
00: [ 321.354592]
00: [ 321.354592] the existing dependency chain (in reverse order) is:
00: [ 321.354604]
00: [ 321.354604] -> #4 (random_write_wait.lock){..-.}:
00: [ 321.354623] lock_acquire+0x212/0x460
00: [ 321.354640] _raw_spin_lock_irqsave+0xc4/0xe0
00: [ 321.354652] __wake_up_common_lock+0xd2/0x140
00: [ 321.354664] __wake_up+0x2a/0x38
00: [ 321.354680] crng_reseed+0x7da/0xb70
00: [ 321.354692] _extract_crng+0xf6/0x100
00: [ 321.354704] crng_reseed+0x7a4/0xb70
00: [ 321.354716] _extract_crng+0xf6/0x100
00: [ 321.354729] get_random_u32+0x100/0x138
00: [ 321.354746] new_slab+0x188/0x760
00: [ 321.354758] ___slab_alloc+0x5d2/0x928
00: [ 321.354771] __slab_alloc+0x52/0x88
00: [ 321.355194] kmem_cache_alloc+0x34a/0x558
00: [ 321.355213] skb_clone+0x8c/0x138
00: [ 321.355228] arp_rcv+0x76/0x2d8
00: [ 321.355244] __netif_receive_skb_core+0xd08/0x1588
00: [ 321.355258] __netif_receive_skb_list_core+0x28a/0x530
00: [ 321.355273] netif_receive_skb_list_internal+0x4ae/0x760
00: [ 321.355288] napi_complete_done+0x178/0x320
00: [ 321.355302] qeth_poll+0xffa/0x1130
00: [ 321.355315] net_rx_action+0x30a/0x9e0
00: [ 321.355330] __do_softirq+0x1da/0xa28
00: [ 321.355347] do_softirq_own_stack+0xe4/0x100
00: [ 321.355361] irq_exit+0x148/0x1c0
00: [ 321.355373] do_IRQ+0xb8/0x108
00: [ 321.355399] io_int_handler+0x134/0x2c0
00: [ 321.355412] enabled_wait+0x64/0x168
00: [ 321.355424] enabled_wait+0x42/0x168
00: [ 321.355435] arch_cpu_idle+0xc2/0xc8
00: [ 321.355447] default_idle_call+0xdc/0xf8
00: [ 321.355463] do_idle+0x1fa/0x2a8
00: [ 321.355474] cpu_startup_entry+0x36/0x40
00: [ 321.355490] arch_call_rest_init+0xa8/0xb8
00: [ 321.355501]
00: [ 321.355501] -> #3 (batched_entropy_u32.lock){-.-.}:
00: [ 321.355523] lock_acquire+0x212/0x460
00: [ 321.355536] _raw_spin_lock_irqsave+0xc4/0xe0
00: [ 321.355551] get_random_u32+0x5a/0x138
00: [ 321.355564] new_slab+0x188/0x760
00: [ 321.355576] ___slab_alloc+0x5d2/0x928
00: [ 321.355589] __slab_alloc+0x52/0x88
00: [ 321.355801] kmem_cache_alloc+0x34a/0x558
00: [ 321.355819] fill_pool+0x29e/0x490
00: [ 321.355835] __debug_object_init+0xa0/0x828
00: [ 321.355848] debug_object_activate+0x200/0x368
00: [ 321.355864] add_timer+0x242/0x538
00: [ 321.355877] queue_delayed_work_on+0x13e/0x148
00: [ 321.355893] init_mm_internals+0x4c6/0x550
00: [ 321.355905] kernel_init_freeable+0x224/0x590
00: [ 321.355921] kernel_init+0x22/0x188
00: [ 321.355933] ret_from_fork+0x30/0x34
00: [ 321.355942]
00: [ 321.355942] -> #2 (&base->lock){-.-.}:
00: [ 321.355961] lock_acquire+0x212/0x460
00: [ 321.355973] _raw_spin_lock_irqsave+0xc4/0xe0
00: [ 321.355986] lock_timer_base+0xca/0x1a0
00: [ 321.355998] add_timer+0xe6/0x538
00: [ 321.356010] queue_delayed_work_on+0x13e/0x148
00: [ 321.356025] psi_task_change+0x510/0x8d0
00: [ 321.356037] activate_task+0x1f8/0x280
00: [ 321.356050] wake_up_new_task+0x410/0x7e0
00: [ 321.356064] _do_fork+0x24e/0xab0
00: [ 321.356076] kernel_thread+0xbc/0xd8
00: [ 321.356089] rest_init+0x36/0x338
00: [ 321.356100] arch_call_rest_init+0xa8/0xb8
00: [ 321.356110]
00: [ 321.356110] -> #1 (&rq->lock){-.-.}:
00: [ 321.356127] lock_acquire+0x212/0x460
00: [ 321.356140] _raw_spin_lock+0x4c/0x60
00: [ 321.356152] task_fork_fair+0x4a/0x270
00: [ 321.356165] sched_fork+0x262/0x3e0
00: [ 321.356326] copy_process+0x1160/0x2d20
00: [ 321.356339] _do_fork+0x134/0xab0
00: [ 321.356351] kernel_thread+0xbc/0xd8
00: [ 321.356364] rest_init+0x36/0x338
00: [ 321.356376] arch_call_rest_init+0xa8/0xb8
00: [ 321.356385]
00: [ 321.356385] -> #0 (&p->pi_lock){-.-.}:
00: [ 321.356403] check_noncircular+0x330/0x3d8
00: [ 321.356416] __lock_acquire+0x23d2/0x49f8
00: [ 321.356428] lock_acquire+0x212/0x460
00: [ 321.356441] _raw_spin_lock_irqsave+0xc4/0xe0
00: [ 321.356454] try_to_wake_up+0xce/0xd88
00: [ 321.356470] pollwake+0x118/0x160
00: [ 321.356482] __wake_up_common+0xf2/0x2e8
00: [ 321.356495] __wake_up_common_lock+0x100/0x140
00: [ 321.356507] __wake_up+0x2a/0x38
00: [ 321.356519] crng_reseed+0x7da/0xb70
00: [ 321.356531] _extract_crng+0xf6/0x100
00: [ 321.356543] crng_reseed+0x7a4/0xb70
00: [ 321.356555] _extract_crng+0xf6/0x100
00: [ 321.356568] get_random_u32+0x100/0x138
00: [ 321.356581] new_slab+0x188/0x760
00: [ 321.356593] ___slab_alloc+0x5d2/0x928
00: [ 321.356606] __slab_alloc+0x52/0x88
00: [ 321.356619] kmem_cache_alloc+0x34a/0x558
00: [ 321.356632] skb_clone+0x8c/0x138
00: [ 321.356645] arp_rcv+0x76/0x2d8
00: [ 321.356658] __netif_receive_skb_core+0xd08/0x1588
00: [ 321.356673] __netif_receive_skb_list_core+0x28a/0x530
00: [ 321.356688] netif_receive_skb_list_internal+0x4ae/0x760
00: [ 321.356840] napi_complete_done+0x178/0x320
00: [ 321.356855] qeth_poll+0xffa/0x1130
00: [ 321.356867] net_rx_action+0x30a/0x9e0
00: [ 321.356880] __do_softirq+0x1da/0xa28
00: [ 321.356895] do_softirq_own_stack+0xe4/0x100
00: [ 321.356908] irq_exit+0x148/0x1c0
00: [ 321.356920] do_IRQ+0xb8/0x108
00: [ 321.356932] io_int_handler+0x134/0x2c0
00: [ 321.356945] enabled_wait+0x64/0x168
00: [ 321.356980] enabled_wait+0x42/0x168
00: [ 321.357004] arch_cpu_idle+0xc2/0xc8
00: [ 321.357018] default_idle_call+0xdc/0xf8
00: [ 321.357032] do_idle+0x1fa/0x2a8
00: [ 321.357043] cpu_startup_entry+0x36/0x40
00: [ 321.357057] arch_call_rest_init+0xa8/0xb8
00: [ 321.357067]
00: [ 321.357067] other info that might help us debug this:
00: [ 321.357067]
00: [ 321.357084] Chain exists of:
00: [ 321.357084] &p->pi_lock --> batched_entropy_u32.lock --> random_write_wait.lock
00: [ 321.357084]
00: [ 321.357108] Possible unsafe locking scenario:
00: [ 321.357108]
00: [ 321.357119] CPU0 CPU1
00: [ 321.357129] ---- ----
00: [ 321.357138] lock(random_write_wait.lock);
00: [ 321.357149] lock(batched_entropy_u32.lock);
00: [ 321.357163] lock(random_write_wait.lock);
00: [ 321.357177] lock(&p->pi_lock);
00: [ 321.357337]
00: [ 321.357337] *** DEADLOCK ***
00: [ 321.357337]
00: [ 321.357351] 3 locks held by swapper/0/0:
00: [ 321.357360] #0: 000000004df6a1a0 (rcu_read_lock){....}, at: netif_receive_skb_list_internal+0x27c/0x760
00: [ 321.357387] #1: 000000006cfe3840 (batched_entropy_u32.lock){-.-.}, at: get_random_u32+0x5a/0x138
00: [ 321.357411] #2: 000000004e23d1b8 (random_write_wait.lock){..-.}, at: __wake_up_common_lock+0xd2/0x140
00: [ 321.357434]
00: [ 321.357434] stack backtrace:
00: [ 321.357451] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.6.0-rc5-next-20200312+ #5
00: [ 321.357465] Hardware name: IBM 2964 N96 400 (z/VM 6.4.0)
00: [ 321.357476] Call Trace:
00: [ 321.357491] [<000000004d12def0>] show_stack+0x118/0x1b8
00: [ 321.357506] [<000000004da308de>] dump_stack+0x12e/0x180
00: [ 321.357521] [<000000004d224218>] check_noncircular+0x330/0x3d8
00: [ 321.357535] [<000000004d22a82a>] __lock_acquire+0x23d2/0x49f8
00: [ 321.357550] [<000000004d2274aa>] lock_acquire+0x212/0x460
00: [ 321.357564] [<000000004da6bf6c>] _raw_spin_lock_irqsave+0xc4/0xe0
00: [ 321.357579] [<000000004d1bf9f6>] try_to_wake_up+0xce/0xd88
00: [ 321.357594] [<000000004d53a748>] pollwake+0x118/0x160
00: [ 321.357607] [<000000004d1ffb8a>] __wake_up_common+0xf2/0x2e8
00: [ 321.357622] [<000000004d1ffe80>] __wake_up_common_lock+0x100/0x140
00: [ 321.357654] [<000000004d1ffeea>] __wake_up+0x2a/0x38
00: [ 321.357669] [<000000004d77cef2>] crng_reseed+0x7da/0xb70
00: [ 321.357683] [<000000004d77d37e>] _extract_crng+0xf6/0x100
00: [ 321.357834] [<000000004d77cebc>] crng_reseed+0x7a4/0xb70
00: [ 321.357849] [<000000004d77d37e>] _extract_crng+0xf6/0x100
00: [ 321.357865] [<000000004d77e1e0>] get_random_u32+0x100/0x138
00: [ 321.357880] [<000000004d4ad5a8>] new_slab+0x188/0x760
00: [ 321.357895] [<000000004d4b0cea>] ___slab_alloc+0x5d2/0x928
00: [ 321.357911] [<000000004d4b1092>] __slab_alloc+0x52/0x88
00: [ 321.357925] [<000000004d4b1412>] kmem_cache_alloc+0x34a/0x558
00: [ 321.357939] [<000000004d88b79c>] skb_clone+0x8c/0x138
00: [ 321.357953] [<000000004d9c7da6>] arp_rcv+0x76/0x2d8
00: [ 321.357968] [<000000004d8bd170>] __netif_receive_skb_core+0xd08/0x1588
00: [ 321.358018] [<000000004d8bec1a>] __netif_receive_skb_list_core+0x28a/0x5
00: [ 321.358045] [<000000004d8bf36e>] netif_receive_skb_list_internal+0x4ae/0
00: [ 321.358062] [<000000004d8c2050>] napi_complete_done+0x178/0x320
00: [ 321.358077] [<000000004d858a8a>] qeth_poll+0xffa/0x1130
00: [ 321.358093] [<000000004d8c2f8a>] net_rx_action+0x30a/0x9e0
00: [ 321.358108] [<000000004da6e272>] __do_softirq+0x1da/0xa28
00: [ 321.358124] [<000000004d124cc4>] do_softirq_own_stack+0xe4/0x100
00: [ 321.358164] [<000000004d16fa00>] irq_exit+0x148/0x1c0
00: [ 321.358179] [<000000004d124480>] do_IRQ+0xb8/0x108
00: [ 321.358194] [<000000004da6d4a4>] io_int_handler+0x134/0x2c0
00: [ 321.358208] [<000000004d115d84>] enabled_wait+0x64/0x168
00: [ 321.358247] ([<000000004d115d62>] enabled_wait+0x42/0x168)
00: [ 321.358263] [<000000004d116212>] arch_cpu_idle+0xc2/0xc8
00: [ 321.358277] [<000000004da6bc2c>] default_idle_call+0xdc/0xf8
00: [ 321.358292] [<000000004d1cb872>] do_idle+0x1fa/0x2a8
00: [ 321.358458] [<000000004d1cbd0e>] cpu_startup_entry+0x36/0x40
00: [ 321.358473] [<000000004e40c098>] arch_call_rest_init+0xa8/0xb8
00: [ 321.358484] INFO: lockdep is turned off.
Signed-off-by: Qian Cai <[email protected]>
---
kernel/time/timer.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/kernel/time/timer.c b/kernel/time/timer.c
index 4820823515e9..27bfb8376d71 100644
--- a/kernel/time/timer.c
+++ b/kernel/time/timer.c
@@ -1036,7 +1036,13 @@ __mod_timer(struct timer_list *timer, unsigned long expires, unsigned int option
}
}
+ /*
+ * It will allocate under rq->lock and trigger a lockdep slat with
+ * random code. Don't disable lockdep with debugobjects.
+ */
+ lockdep_off();
debug_timer_activate(timer);
+ lockdep_on();
timer->expires = expires;
/*
--
2.21.0 (Apple Git-122.2)
On Fri, Mar 13, 2020 at 11:42:21AM -0400, Qian Cai wrote:
> psi_enqueue() calls add_timer() with pi->lock and rq->lock held which
> in-turn could allcate with debugobjcets in the locking order,
>
> pi->lock
> rq->lock
> base->lock
> batched_entropy_u32.lock
>
> while the random code could always call into the scheduler via
> try_to_wake_up() in the locking order,
>
> batched_entropy_u32.lock
> pi->lock
>
> Thus, it could generate a lockdep splat below right after boot. Ideally,
> psi_enqueue() might be able to be called without either pi->lock or
> rq->lock held, but it is tricky to do.
>
> Since,
>
> 1) debugobjects is only used in a debug kernel.
> 2) the chance to trigger a real deadlock is relative low.
> 3) once the splat happened, it will disable lockdep to prevent it from
> catching any more important issues later.
>
> just silent the splat by temporarily lettting lockdep ignore lockes
> inside debug_timer_activate() which sounds like a reasonable tradeoff
> for debug kernels.
> diff --git a/kernel/time/timer.c b/kernel/time/timer.c
> index 4820823515e9..27bfb8376d71 100644
> --- a/kernel/time/timer.c
> +++ b/kernel/time/timer.c
> @@ -1036,7 +1036,13 @@ __mod_timer(struct timer_list *timer, unsigned long expires, unsigned int option
> }
> }
>
> + /*
> + * It will allocate under rq->lock and trigger a lockdep slat with
> + * random code. Don't disable lockdep with debugobjects.
> + */
> + lockdep_off();
> debug_timer_activate(timer);
> + lockdep_on();
>
> timer->expires = expires;
> /*
You have to be f'ing kidding me. You've just earned yourself a lifetime
membership of 'the tinker crew'.
> 00: [ 321.355501] -> #3 (batched_entropy_u32.lock){-.-.}:
> 00: [ 321.355523] lock_acquire+0x212/0x460
> 00: [ 321.355536] _raw_spin_lock_irqsave+0xc4/0xe0
> 00: [ 321.355551] get_random_u32+0x5a/0x138
> 00: [ 321.355564] new_slab+0x188/0x760
> 00: [ 321.355576] ___slab_alloc+0x5d2/0x928
> 00: [ 321.355589] __slab_alloc+0x52/0x88
> 00: [ 321.355801] kmem_cache_alloc+0x34a/0x558
> 00: [ 321.355819] fill_pool+0x29e/0x490
> 00: [ 321.355835] __debug_object_init+0xa0/0x828
> 00: [ 321.355848] debug_object_activate+0x200/0x368
> 00: [ 321.355864] add_timer+0x242/0x538
> 00: [ 321.355877] queue_delayed_work_on+0x13e/0x148
> 00: [ 321.355893] init_mm_internals+0x4c6/0x550
> 00: [ 321.355905] kernel_init_freeable+0x224/0x590
> 00: [ 321.355921] kernel_init+0x22/0x188
> 00: [ 321.355933] ret_from_fork+0x30/0x34
Did you actually look at debug_object_activate() and read?
The only reason that is calling into __debug_object_init() is because it
hadn't been initialized yet when it got activated. That *immediately*
should've been a clue.
You can initialize this stuff early. For instance:
INIT_DELAYED_WORK()
__INIT_DELAYED_WORK()
__init_timer()
init_timer_key()
debug_init()
debug_timer_init()
debug_object_init()
__debug_object_init()
And we're right at where the above callchain goes wrong.
Now, it actually looks like kernel/sched/psi.c actually initializes all
delayed works it uses. This then leaves other random delayed works to
establish the base->lock <- entropy.lock relation.
This just means we need to find and kill all such delayed_work users
that fail to properly initialize their data structure.
I'm not going to do that just now, the kids need attention.
> On Mar 13, 2020, at 2:08 PM, Peter Zijlstra <[email protected]> wrote:
>
> On Fri, Mar 13, 2020 at 11:42:21AM -0400, Qian Cai wrote:
>> psi_enqueue() calls add_timer() with pi->lock and rq->lock held which
>> in-turn could allcate with debugobjcets in the locking order,
>>
>> pi->lock
>> rq->lock
>> base->lock
>> batched_entropy_u32.lock
>>
>> while the random code could always call into the scheduler via
>> try_to_wake_up() in the locking order,
>>
>> batched_entropy_u32.lock
>> pi->lock
>>
>> Thus, it could generate a lockdep splat below right after boot. Ideally,
>> psi_enqueue() might be able to be called without either pi->lock or
>> rq->lock held, but it is tricky to do.
>>
>> Since,
>>
>> 1) debugobjects is only used in a debug kernel.
>> 2) the chance to trigger a real deadlock is relative low.
>> 3) once the splat happened, it will disable lockdep to prevent it from
>> catching any more important issues later.
>>
>> just silent the splat by temporarily lettting lockdep ignore lockes
>> inside debug_timer_activate() which sounds like a reasonable tradeoff
>> for debug kernels.
>
>
>> diff --git a/kernel/time/timer.c b/kernel/time/timer.c
>> index 4820823515e9..27bfb8376d71 100644
>> --- a/kernel/time/timer.c
>> +++ b/kernel/time/timer.c
>> @@ -1036,7 +1036,13 @@ __mod_timer(struct timer_list *timer, unsigned long expires, unsigned int option
>> }
>> }
>>
>> + /*
>> + * It will allocate under rq->lock and trigger a lockdep slat with
>> + * random code. Don't disable lockdep with debugobjects.
>> + */
>> + lockdep_off();
>> debug_timer_activate(timer);
>> + lockdep_on();
>>
>> timer->expires = expires;
>> /*
>
> You have to be f'ing kidding me. You've just earned yourself a lifetime
> membership of 'the tinker crew'.
>
>> 00: [ 321.355501] -> #3 (batched_entropy_u32.lock){-.-.}:
>> 00: [ 321.355523] lock_acquire+0x212/0x460
>> 00: [ 321.355536] _raw_spin_lock_irqsave+0xc4/0xe0
>> 00: [ 321.355551] get_random_u32+0x5a/0x138
>> 00: [ 321.355564] new_slab+0x188/0x760
>> 00: [ 321.355576] ___slab_alloc+0x5d2/0x928
>> 00: [ 321.355589] __slab_alloc+0x52/0x88
>> 00: [ 321.355801] kmem_cache_alloc+0x34a/0x558
>> 00: [ 321.355819] fill_pool+0x29e/0x490
>> 00: [ 321.355835] __debug_object_init+0xa0/0x828
>> 00: [ 321.355848] debug_object_activate+0x200/0x368
>> 00: [ 321.355864] add_timer+0x242/0x538
>> 00: [ 321.355877] queue_delayed_work_on+0x13e/0x148
>> 00: [ 321.355893] init_mm_internals+0x4c6/0x550
>> 00: [ 321.355905] kernel_init_freeable+0x224/0x590
>> 00: [ 321.355921] kernel_init+0x22/0x188
>> 00: [ 321.355933] ret_from_fork+0x30/0x34
>
> Did you actually look at debug_object_activate() and read?
>
> The only reason that is calling into __debug_object_init() is because it
> hadn't been initialized yet when it got activated. That *immediately*
> should've been a clue.
>
> You can initialize this stuff early. For instance:
>
> INIT_DELAYED_WORK()
> __INIT_DELAYED_WORK()
> __init_timer()
> init_timer_key()
> debug_init()
> debug_timer_init()
> debug_object_init()
> __debug_object_init()
>
> And we're right at where the above callchain goes wrong.
>
> Now, it actually looks like kernel/sched/psi.c actually initializes all
> delayed works it uses. This then leaves other random delayed works to
> establish the base->lock <- entropy.lock relation.
>
> This just means we need to find and kill all such delayed_work users
> that fail to properly initialize their data structure.
Well, in mm/vmstat.c (init_mm_internals) case for example, it was initialized
as a static,
static DECLARE_DEFERRABLE_WORK(shepherd, vmstat_shepherd);
which will not call __debug_object_init().
On Fri, Mar 13, 2020 at 03:32:52PM -0400, Qian Cai wrote:
> Well, in mm/vmstat.c (init_mm_internals) case for example, it was initialized
> as a static,
>
> static DECLARE_DEFERRABLE_WORK(shepherd, vmstat_shepherd);
>
> which will not call __debug_object_init().
Then fix that.
That is, all of:
DECLARE_DEFERRABLE_WORK
KTHREAD_DELAYED_WORK_INIT
DEFINE_TIMER
are broken and need to go.
Unless of course, you can frob debugobjects such that we can provide the
required storage through is_static_object() and modify the above static
declarators to provide the storage.
Or, fix that random crud to do the wakeup outside of the lock.
Plenty options..
> On Mar 13, 2020, at 4:13 PM, Peter Zijlstra <[email protected]> wrote:
>
> On Fri, Mar 13, 2020 at 03:32:52PM -0400, Qian Cai wrote:
>> Well, in mm/vmstat.c (init_mm_internals) case for example, it was initialized
>> as a static,
>>
>> static DECLARE_DEFERRABLE_WORK(shepherd, vmstat_shepherd);
>>
>> which will not call __debug_object_init().
>
> Then fix that.
>
> That is, all of:
>
> DECLARE_DEFERRABLE_WORK
> KTHREAD_DELAYED_WORK_INIT
> DEFINE_TIMER
>
> are broken and need to go.
This looks invasive with so many callsites...
>
> Unless of course, you can frob debugobjects such that we can provide the
> required storage through is_static_object() and modify the above static
> declarators to provide the storage.
I’ll explore this which looks more promising.
>
> Or, fix that random crud to do the wakeup outside of the lock.
That is likely to be difficult until we can find a creative way to not “uglifying" the
random code by dropping locks in the middle etc just because of debugojects.
>
> Plenty options..
On Fri, Mar 13, 2020 at 05:00:41PM -0400, Qian Cai wrote:
> > On Mar 13, 2020, at 4:13 PM, Peter Zijlstra <[email protected]> wrote:
> >
> > Or, fix that random crud to do the wakeup outside of the lock.
>
> That is likely to be difficult until we can find a creative way to not “uglifying" the
> random code by dropping locks in the middle etc just because of debugojects.
https://git.kernel.org/pub/scm/linux/kernel/git/luto/linux.git/log/?h=random/fast
Doesn't look difficult at all.
> On Mar 13, 2020, at 5:26 PM, Peter Zijlstra <[email protected]> wrote:
>
> On Fri, Mar 13, 2020 at 05:00:41PM -0400, Qian Cai wrote:
>>> On Mar 13, 2020, at 4:13 PM, Peter Zijlstra <[email protected]> wrote:
>
>>>
>>> Or, fix that random crud to do the wakeup outside of the lock.
>>
>> That is likely to be difficult until we can find a creative way to not “uglifying" the
>> random code by dropping locks in the middle etc just because of debugojects.
>
> https://git.kernel.org/pub/scm/linux/kernel/git/luto/linux.git/log/?h=random/fast
>
> Doesn't look difficult at all.
Great. I cherry-picked the first two patches,
random: Consolidate entropy batches for u32 and u64
random: Remove batched entropy locking
which solved the issue here as well.
Andy, may I ask what your plan to post those?
> On Mar 13, 2020, at 10:53 PM, Qian Cai <[email protected]> wrote:
>
>
>
>> On Mar 13, 2020, at 5:26 PM, Peter Zijlstra <[email protected]> wrote:
>>
>> On Fri, Mar 13, 2020 at 05:00:41PM -0400, Qian Cai wrote:
>>>> On Mar 13, 2020, at 4:13 PM, Peter Zijlstra <[email protected]> wrote:
>>
>>>>
>>>> Or, fix that random crud to do the wakeup outside of the lock.
>>>
>>> That is likely to be difficult until we can find a creative way to not “uglifying" the
>>> random code by dropping locks in the middle etc just because of debugojects.
>>
>> https://git.kernel.org/pub/scm/linux/kernel/git/luto/linux.git/log/?h=random/fast
>>
>> Doesn't look difficult at all.
>
> Great. I cherry-picked the first two patches,
>
> random: Consolidate entropy batches for u32 and u64
> random: Remove batched entropy locking
>
> which solved the issue here as well.
>
> Andy, may I ask what your plan to post those?
Andy, do you mind me posting those 2 patches to Ted on your behalf if you have
no plan to do this in the near future?