Re: [PATCH v4] blk-cgroup: synchoronize blkg creation against policy deactivation

["yukuai (C)" <yukuai3@huawei.com>]

On 2021/10/22 9:28, yukuai (C) wrote:
> Hi 2021/10/20 9:40, Yu Kuai wrote:
>> Out test report a null pointer dereference:
>>
>> [  168.534653] 
>> ==================================================================
>> [  168.535614] Disabling lock debugging due to kernel taint
>> [  168.536346] BUG: kernel NULL pointer dereference, address: 
>> 0000000000000008
>> [  168.537274] #PF: supervisor read access in kernel mode
>> [  168.537964] #PF: error_code(0x0000) - not-present page
>> [  168.538667] PGD 0 P4D 0
>> [  168.539025] Oops: 0000 [#1] PREEMPT SMP KASAN
>> [  168.539656] CPU: 13 PID: 759 Comm: bash Tainted: G    B             
>> 5.15.0-rc2-next-202100
>> [  168.540954] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), 
>> BIOS ?-20190727_0738364
>> [  168.542736] RIP: 0010:bfq_pd_init+0x88/0x1e0
>> [  168.543318] Code: 98 00 00 00 e8 c9 e4 5b ff 4c 8b 65 00 49 8d 7c 
>> 24 08 e8 bb e4 5b ff 4d0
>> [  168.545803] RSP: 0018:ffff88817095f9c0 EFLAGS: 00010002
>> [  168.546497] RAX: 0000000000000001 RBX: ffff888101a1c000 RCX: 
>> 0000000000000000
>> [  168.547438] RDX: 0000000000000003 RSI: 0000000000000002 RDI: 
>> ffff888106553428
>> [  168.548402] RBP: ffff888106553400 R08: ffffffff961bcaf4 R09: 
>> 0000000000000001
>> [  168.549365] R10: ffffffffa2e16c27 R11: fffffbfff45c2d84 R12: 
>> 0000000000000000
>> [  168.550291] R13: ffff888101a1c098 R14: ffff88810c7a08c8 R15: 
>> ffffffffa55541a0
>> [  168.551221] FS:  00007fac75227700(0000) GS:ffff88839ba80000(0000) 
>> knlGS:0000000000000000
>> [  168.552278] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>> [  168.553040] CR2: 0000000000000008 CR3: 0000000165ce7000 CR4: 
>> 00000000000006e0
>> [  168.554000] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 
>> 0000000000000000
>> [  168.554929] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 
>> 0000000000000400
>> [  168.555888] Call Trace:
>> [  168.556221]  <TASK>
>> [  168.556510]  blkg_create+0x1c0/0x8c0
>> [  168.556989]  blkg_conf_prep+0x574/0x650
>> [  168.557502]  ? stack_trace_save+0x99/0xd0
>> [  168.558033]  ? blkcg_conf_open_bdev+0x1b0/0x1b0
>> [  168.558629]  tg_set_conf.constprop.0+0xb9/0x280
>> [  168.559231]  ? kasan_set_track+0x29/0x40
>> [  168.559758]  ? kasan_set_free_info+0x30/0x60
>> [  168.560344]  ? tg_set_limit+0xae0/0xae0
>> [  168.560853]  ? do_sys_openat2+0x33b/0x640
>> [  168.561383]  ? do_sys_open+0xa2/0x100
>> [  168.561877]  ? __x64_sys_open+0x4e/0x60
>> [  168.562383]  ? __kasan_check_write+0x20/0x30
>> [  168.562951]  ? copyin+0x48/0x70
>> [  168.563390]  ? _copy_from_iter+0x234/0x9e0
>> [  168.563948]  tg_set_conf_u64+0x17/0x20
>> [  168.564467]  cgroup_file_write+0x1ad/0x380
>> [  168.565014]  ? cgroup_file_poll+0x80/0x80
>> [  168.565568]  ? __mutex_lock_slowpath+0x30/0x30
>> [  168.566165]  ? pgd_free+0x100/0x160
>> [  168.566649]  kernfs_fop_write_iter+0x21d/0x340
>> [  168.567246]  ? cgroup_file_poll+0x80/0x80
>> [  168.567796]  new_sync_write+0x29f/0x3c0
>> [  168.568314]  ? new_sync_read+0x410/0x410
>> [  168.568840]  ? __handle_mm_fault+0x1c97/0x2d80
>> [  168.569425]  ? copy_page_range+0x2b10/0x2b10
>> [  168.570007]  ? _raw_read_lock_bh+0xa0/0xa0
>> [  168.570622]  vfs_write+0x46e/0x630
>> [  168.571091]  ksys_write+0xcd/0x1e0
>> [  168.571563]  ? __x64_sys_read+0x60/0x60
>> [  168.572081]  ? __kasan_check_write+0x20/0x30
>> [  168.572659]  ? do_user_addr_fault+0x446/0xff0
>> [  168.573264]  __x64_sys_write+0x46/0x60
>> [  168.573774]  do_syscall_64+0x35/0x80
>> [  168.574264]  entry_SYSCALL_64_after_hwframe+0x44/0xae
>> [  168.574960] RIP: 0033:0x7fac74915130
>> [  168.575456] Code: 73 01 c3 48 8b 0d 58 ed 2c 00 f7 d8 64 89 01 48 
>> 83 c8 ff c3 66 0f 1f 444
>> [  168.577969] RSP: 002b:00007ffc3080e288 EFLAGS: 00000246 ORIG_RAX: 
>> 0000000000000001
>> [  168.578986] RAX: ffffffffffffffda RBX: 0000000000000009 RCX: 
>> 00007fac74915130
>> [  168.579937] RDX: 0000000000000009 RSI: 000056007669f080 RDI: 
>> 0000000000000001
>> [  168.580884] RBP: 000056007669f080 R08: 000000000000000a R09: 
>> 00007fac75227700
>> [  168.581841] R10: 000056007655c8f0 R11: 0000000000000246 R12: 
>> 0000000000000009
>> [  168.582796] R13: 0000000000000001 R14: 00007fac74be55e0 R15: 
>> 00007fac74be08c0
>> [  168.583757]  </TASK>
>> [  168.584063] Modules linked in:
>> [  168.584494] CR2: 0000000000000008
>> [  168.584964] ---[ end trace 2475611ad0f77a1a ]---
>>
>> This is because blkg_alloc() is called from blkg_conf_prep() without
>> holding 'q->queue_lock', and elevator is exited before blkg_create():
>>
>> thread 1                            thread 2
>> blkg_conf_prep
>>   spin_lock_irq(&q->queue_lock);
>>   blkg_lookup_check -> return NULL
>>   spin_unlock_irq(&q->queue_lock);
>>
>>   blkg_alloc
>>    blkcg_policy_enabled -> true
>>    pd = ->pd_alloc_fn
>>    blkg->pd[i] = pd
>>                                     blk_mq_exit_sched
>>                                      bfq_exit_queue
>>                                       blkcg_deactivate_policy
>>                                        spin_lock_irq(&q->queue_lock);
>>                                        __clear_bit(pol->plid, 
>> q->blkcg_pols);
>>                                        spin_unlock_irq(&q->queue_lock);
>>                                      q->elevator = NULL;
>>    spin_lock_irq(&q->queue_lock);
>>     blkg_create
>>      if (blkg->pd[i])
>>       ->pd_init_fn -> q->elevator is NULL
>>    spin_unlock_irq(&q->queue_lock);
>>
>> Because blkcg_deactivate_policy() requires queue to be frozen, we can
>> grab q_usage_counter to synchoronize blkg_conf_prep() against
>> blkcg_deactivate_policy().
>>
>> Fixes: e21b7a0b9887 ("block, bfq: add full hierarchical scheduling and 
>> cgroups support")
>> Signed-off-by: Yu Kuai <yukuai3@huawei.com>
>> Acked-by: Tejun Heo <tj@kernel.org>
> 
> Hi, jens
> 
> Can you please apply this patch?
> 
> Thanks,
> Kuai

friendly ping ...