Re: [bpf-next v1] bpf: hash map, suppress lockdep warning

[Hou Tao <houtao1@huawei.com>]

Hi,

On 1/6/2023 10:13 AM, Xu Kuohai wrote:
> On 1/5/2023 7:27 PM, tong@infragraf.org wrote:
>> From: Tonghao Zhang <tong@infragraf.org>
>>
>> The lock may be taken in both NMI and non-NMI contexts.
>> There is a lockdep warning (inconsistent lock state), if
>> enable lockdep. For performance, this patch doesn't use trylock,
>> and disable lockdep temporarily.
>>
>> [   82.474075] ================================
>> [   82.474076] WARNING: inconsistent lock state
>> [   82.474090] 6.1.0+ #48 Tainted: G            E
>> [   82.474093] --------------------------------
>> [   82.474100] inconsistent {INITIAL USE} -> {IN-NMI} usage.
>> [   82.474101] kprobe-load/1740 [HC1[1]:SC0[0]:HE0:SE1] takes:
>> [   82.474105] ffff88860a5cf7b0 (&htab->lockdep_key){....}-{2:2}, at:
>> htab_lock_bucket+0x61/0x6c
>> [   82.474120] {INITIAL USE} state was registered at:
>> [   82.474122]   mark_usage+0x1d/0x11d
>> [   82.474130]   __lock_acquire+0x3c9/0x6ed
>> [   82.474131]   lock_acquire+0x23d/0x29a
>> [   82.474135]   _raw_spin_lock_irqsave+0x43/0x7f
>> [   82.474148]   htab_lock_bucket+0x61/0x6c
>> [   82.474151]   htab_map_update_elem+0x11e/0x220
>> [   82.474155]   bpf_map_update_value+0x267/0x28e
>> [   82.474160]   map_update_elem+0x13e/0x17d
>> [   82.474164]   __sys_bpf+0x2ae/0xb2e
>> [   82.474167]   __do_sys_bpf+0xd/0x15
>> [   82.474171]   do_syscall_64+0x6d/0x84
>> [   82.474174]   entry_SYSCALL_64_after_hwframe+0x63/0xcd
>> [   82.474178] irq event stamp: 1496498
>> [   82.474180] hardirqs last  enabled at (1496497): [<ffffffff817eb9d9>]
>> syscall_enter_from_user_mode+0x63/0x8d
>> [   82.474184] hardirqs last disabled at (1496498): [<ffffffff817ea6b6>]
>> exc_nmi+0x87/0x109
>> [   82.474187] softirqs last  enabled at (1446698): [<ffffffff81a00347>]
>> __do_softirq+0x347/0x387
>> [   82.474191] softirqs last disabled at (1446693): [<ffffffff810b9b06>]
>> __irq_exit_rcu+0x67/0xc6
>> [   82.474195]
>> [   82.474195] other info that might help us debug this:
>> [   82.474196]  Possible unsafe locking scenario:
>> [   82.474196]
>> [   82.474197]        CPU0
>> [   82.474198]        ----
>> [   82.474198]   lock(&htab->lockdep_key);
>> [   82.474200]   <Interrupt>
>> [   82.474200]     lock(&htab->lockdep_key);
>> [   82.474201]
>> [   82.474201]  *** DEADLOCK ***
>> [   82.474201]
>> [   82.474202] no locks held by kprobe-load/1740.
>> [   82.474203]
>> [   82.474203] stack backtrace:
>> [   82.474205] CPU: 14 PID: 1740 Comm: kprobe-load Tainted: G           
>> E      6.1.0+ #48
>> [   82.474208] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
>> rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
>> [   82.474213] Call Trace:
>> [   82.474218]  <NMI>
>> [   82.474224]  dump_stack_lvl+0x57/0x81
>> [   82.474228]  lock_acquire+0x1f4/0x29a
>> [   82.474233]  ? htab_lock_bucket+0x61/0x6c
>> [   82.474237]  ? rcu_read_lock_held_common+0xe/0x38
>> [   82.474245]  _raw_spin_lock_irqsave+0x43/0x7f
>> [   82.474249]  ? htab_lock_bucket+0x61/0x6c
>> [   82.474253]  htab_lock_bucket+0x61/0x6c
>> [   82.474257]  htab_map_update_elem+0x11e/0x220
>> [   82.474264]  bpf_prog_df326439468c24a9_bpf_prog1+0x41/0x45
>> [   82.474276]  bpf_trampoline_6442457183_0+0x43/0x1000
>> [   82.474283]  nmi_handle+0x5/0x254
>> [   82.474289]  default_do_nmi+0x3d/0xf6
>> [   82.474293]  exc_nmi+0xa1/0x109
>> [   82.474297]  end_repeat_nmi+0x16/0x67
>> [   82.474300] RIP: 0010:cpu_online+0xa/0x12
>> [   82.474308] Code: 08 00 00 00 39 c6 0f 43 c6 83 c0 07 83 e0 f8 c3 cc cc cc
>> cc 0f 1f 44 00 00 31 c0 c3 cc cc cc cc 89 ff 48 0f a3 3d 5f 52 75 01 <0f> 92
>> c0 c3 cc cc cc cc 55 48 89 e5 41 57 49 89 f7 41 56 49 896
>> [   82.474310] RSP: 0018:ffffc9000131bd38 EFLAGS: 00000283
>> [   82.474313] RAX: ffff88860b85fe78 RBX: 0000000000102cc0 RCX: 0000000000000008
>> [   82.474315] RDX: 0000000000000004 RSI: ffff88860b85fe78 RDI: 000000000000000e
>> [   82.474316] RBP: 00000000ffffffff R08: 0000000000102cc0 R09: 00000000ffffffff
>> [   82.474318] R10: 0000000000000001 R11: 0000000000000000 R12: ffff888100042200
>> [   82.474320] R13: 0000000000000004 R14: ffffffff81271dc2 R15: ffff88860b85fe78
>> [   82.474322]  ? kvmalloc_node+0x44/0xd2
>> [   82.474333]  ? cpu_online+0xa/0x12
>> [   82.474338]  ? cpu_online+0xa/0x12
>> [   82.474342]  </NMI>
>> [   82.474343]  <TASK>
>> [   82.474343]  trace_kmalloc+0x7c/0xe6
>> [   82.474347]  ? kvmalloc_node+0x44/0xd2
>> [   82.474350]  __kmalloc_node+0x9a/0xaf
>> [   82.474354]  kvmalloc_node+0x44/0xd2
>> [   82.474359]  kvmemdup_bpfptr+0x29/0x66
>> [   82.474363]  map_update_elem+0x119/0x17d
>> [   82.474370]  __sys_bpf+0x2ae/0xb2e
>> [   82.474380]  __do_sys_bpf+0xd/0x15
>> [   82.474384]  do_syscall_64+0x6d/0x84
>> [   82.474387]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
>> [   82.474391] RIP: 0033:0x7fe75d4f752d
>> [   82.474394] Code: 00 c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89
>> f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d
>> 01 f0 ff ff 73 01 c3 48 8b 0d 2b 79 2c 00 f7 d8 64 89 018
>> [   82.474396] RSP: 002b:00007ffe95d1cd78 EFLAGS: 00000246 ORIG_RAX:
>> 0000000000000141
>> [   82.474398] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007fe75d4f752d
>> [   82.474400] RDX: 0000000000000078 RSI: 00007ffe95d1cd80 RDI: 0000000000000002
>> [   82.474401] RBP: 00007ffe95d1ce30 R08: 0000000000000000 R09: 0000000000000004
>> [   82.474403] R10: 00007ffe95d1cd80 R11: 0000000000000246 R12: 00000000004007f0
>> [   82.474405] R13: 00007ffe95d1cf10 R14: 0000000000000000 R15: 0000000000000000
>> [   82.474412]  </TASK>
>>
>> Signed-off-by: Tonghao Zhang <tong@infragraf.org>
>> Cc: Alexei Starovoitov <ast@kernel.org>
>> Cc: Daniel Borkmann <daniel@iogearbox.net>
>> Cc: Andrii Nakryiko <andrii@kernel.org>
>> Cc: Martin KaFai Lau <martin.lau@linux.dev>
>> Cc: Song Liu <song@kernel.org>
>> Cc: Yonghong Song <yhs@fb.com>
>> Cc: John Fastabend <john.fastabend@gmail.com>
>> Cc: KP Singh <kpsingh@kernel.org>
>> Cc: Stanislav Fomichev <sdf@google.com>
>> Cc: Hao Luo <haoluo@google.com>
>> Cc: Jiri Olsa <jolsa@kernel.org>
>> Cc: Hou Tao <houtao1@huawei.com>
>> ---
>> previous discussion:
>> https://lore.kernel.org/all/20221121100521.56601-2-xiangxia.m.yue@gmail.com/
>> ---
>>   kernel/bpf/hashtab.c | 16 ++++++++++++++++
>>   1 file changed, 16 insertions(+)
>>
>> diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c
>> index 974f104f47a0..146433c9bd1a 100644
>> --- a/kernel/bpf/hashtab.c
>> +++ b/kernel/bpf/hashtab.c
>> @@ -161,6 +161,19 @@ static inline int htab_lock_bucket(const struct bpf_htab
>> *htab,
>>           return -EBUSY;
>>       }
>>   +    /*
>> +     * The lock may be taken in both NMI and non-NMI contexts.
>> +     * There is a lockdep warning (inconsistent lock state), if
>> +     * enable lockdep. The potential deadlock happens when the
>> +     * lock is contended from the same cpu. map_locked rejects
>> +     * concurrent access to the same bucket from the same CPU.
>> +     * When the lock is contended from a remote cpu, we would
>> +     * like the remote cpu to spin and wait, instead of giving
>> +     * up immediately. As this gives better throughput. So replacing
>> +     * the current raw_spin_lock_irqsave() with trylock sacrifices
>> +     * this performance gain. lockdep_off temporarily.
>> +     */
>> +    lockdep_off();
>
> seems it's better to define map_locked as a raw spin lock and use
> trylock(map_locked[hash])
> to check if this cpu has locked the bucket.
Redefining map_locked as per-cpu raw spinlock array seems a good idea, so we
don't need to call lock_acquire() and lock_release() manually. But lockdep_off()
is still needed here, else there will always be lockdep warning.
>
>>       raw_spin_lock_irqsave(&b->raw_lock, flags);
>>       *pflags = flags;
>>   @@ -172,7 +185,10 @@ static inline void htab_unlock_bucket(const struct
>> bpf_htab *htab,
>>                         unsigned long flags)
>>   {
>>       hash = hash & min_t(u32, HASHTAB_MAP_LOCK_MASK, htab->n_buckets -1);
>> +
>>       raw_spin_unlock_irqrestore(&b->raw_lock, flags);
>> +    lockdep_on();
>> +
>>       __this_cpu_dec(*(htab->map_locked[hash]));
>>       preempt_enable();
>>   }
>
> .