Re: [PATCH v4 1/2] io_uring: avoid whole io_wq_work copy for requests completed inline

[Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>]

hi,

> hi jens,
> 
>> On 5/30/20 11:14 AM, Jens Axboe wrote:
>>> On 5/30/20 10:44 AM, Jens Axboe wrote:
>>>> On 5/30/20 8:39 AM, Xiaoguang Wang wrote:
>>>>> If requests can be submitted and completed inline, we don't need to
>>>>> initialize whole io_wq_work in io_init_req(), which is an expensive
>>>>> operation, add a new 'REQ_F_WORK_INITIALIZED' to control whether
>>>>> io_wq_work is initialized.
>>>>>
>>>>> I use /dev/nullb0 to evaluate performance improvement in my physical
>>>>> machine:
>>>>>    modprobe null_blk nr_devices=1 completion_nsec=0
>>>>>    sudo taskset -c 60 fio  -name=fiotest -filename=/dev/nullb0 -iodepth=128
>>>>>    -thread -rw=read -ioengine=io_uring -direct=1 -bs=4k -size=100G -numjobs=1
>>>>>    -time_based -runtime=120
>>>>>
>>>>> before this patch:
>>>>> Run status group 0 (all jobs):
>>>>>     READ: bw=724MiB/s (759MB/s), 724MiB/s-724MiB/s (759MB/s-759MB/s),
>>>>>     io=84.8GiB (91.1GB), run=120001-120001msec
>>>>>
>>>>> With this patch:
>>>>> Run status group 0 (all jobs):
>>>>>     READ: bw=761MiB/s (798MB/s), 761MiB/s-761MiB/s (798MB/s-798MB/s),
>>>>>     io=89.2GiB (95.8GB), run=120001-120001msec
>>>>>
>>>>> About 5% improvement.
>>>>
>>>> There's something funky going on here. I ran the liburing test
>>>> suite on this, and get a lot of left behind workers:
>>>>
>>>> Tests _maybe_ failed:  ring-leak open-close open-close file-update file-update accept-reuse accept-reuse poll-v-poll poll-v-poll fadvise fadvise madvise madvise short-read short-read openat2 
>>>> openat2 probe probe shared-wq shared-wq personality personality eventfd eventfd send_recv send_recv eventfd-ring eventfd-ring across-fork across-fork sq-poll-kthread sq-poll-kthread splice splice 
>>>> lfs-openat lfs-openat lfs-openat-write lfs-openat-write iopoll iopoll d4ae271dfaae-test d4ae271dfaae-test eventfd-disable eventfd-disable write-file write-file buf-rw buf-rw statx statx
>>>>
>>>> and also saw this:
>>>>
>>>> [  168.208940] ==================================================================
>>>> [  168.209311] BUG: KASAN: use-after-free in __lock_acquire+0x8bf/0x3000
>>>> [  168.209626] Read of size 8 at addr ffff88806801c0d8 by task io_wqe_worker-0/41761
>>>> [  168.209987]
>>>> [  168.210069] CPU: 0 PID: 41761 Comm: io_wqe_worker-0 Not tainted 5.7.0-rc7+ #6318
>>>> [  168.210424] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014
>>>> [  168.210857] Call Trace:
>>>> [  168.210991]  dump_stack+0x97/0xe0
>>>> [  168.211164]  print_address_description.constprop.0+0x1a/0x210
>>>> [  168.211446]  ? __lock_acquire+0x8bf/0x3000
>>>> [  168.211649]  __kasan_report.cold+0x20/0x39
>>>> [  168.211851]  ? __lock_acquire+0x8bf/0x3000
>>>> [  168.212051]  kasan_report+0x30/0x40
>>>> [  168.212226]  __lock_acquire+0x8bf/0x3000
>>>> [  168.212432]  ? ret_from_fork+0x24/0x30
>>>> [  168.212623]  ? stack_trace_save+0x81/0xa0
>>>> [  168.212821]  ? lockdep_hardirqs_on+0x270/0x270
>>>> [  168.213039]  ? save_stack+0x32/0x40
>>>> [  168.213212]  lock_acquire+0x122/0x570
>>>> [  168.213398]  ? __close_fd_get_file+0x40/0x150
>>>> [  168.213615]  ? lock_release+0x3f0/0x3f0
>>>> [  168.213814]  ? __lock_acquire+0x87e/0x3000
>>>> [  168.214016]  _raw_spin_lock+0x2c/0x40
>>>> [  168.214196]  ? __close_fd_get_file+0x40/0x150
>>>> [  168.214408]  __close_fd_get_file+0x40/0x150
>>>> [  168.214618]  io_issue_sqe+0x57f/0x22f0
>>>> [  168.214803]  ? lockdep_hardirqs_on+0x270/0x270
>>>> [  168.215019]  ? mark_held_locks+0x24/0x90
>>>> [  168.215211]  ? quarantine_put+0x6f/0x190
>>>> [  168.215404]  ? io_assign_current_work+0x59/0x80
>>>> [  168.215623]  ? __ia32_sys_io_uring_setup+0x30/0x30
>>>> [  168.215855]  ? find_held_lock+0xcb/0x100
>>>> [  168.216054]  ? io_worker_handle_work+0x289/0x980
>>>> [  168.216280]  ? lock_downgrade+0x340/0x340
>>>> [  168.216476]  ? io_wq_submit_work+0x5d/0x140
>>>> [  168.216679]  ? _raw_spin_unlock_irq+0x24/0x40
>>>> [  168.216890]  io_wq_submit_work+0x5d/0x140
>>>> [  168.217087]  io_worker_handle_work+0x30a/0x980
>>>> [  168.217305]  ? io_wqe_dec_running.isra.0+0x70/0x70
>>>> [  168.217537]  ? do_raw_spin_lock+0x100/0x180
>>>> [  168.217742]  ? rwlock_bug.part.0+0x60/0x60
>>>> [  168.217943]  io_wqe_worker+0x5fd/0x780
>>>> [  168.218126]  ? lock_downgrade+0x340/0x340
>>>> [  168.218323]  ? io_worker_handle_work+0x980/0x980
>>>> [  168.218546]  ? lockdep_hardirqs_on+0x17d/0x270
>>>> [  168.218765]  ? __kthread_parkme+0xca/0xe0
>>>> [  168.218961]  ? io_worker_handle_work+0x980/0x980
>>>> [  168.219186]  kthread+0x1f0/0x220
>>>> [  168.219346]  ? kthread_create_worker_on_cpu+0xb0/0xb0
>>>> [  168.219590]  ret_from_fork+0x24/0x30
>>>> [  168.219768]
>>>> [  168.219846] Allocated by task 41758:
>>>> [  168.220021]  save_stack+0x1b/0x40
>>>> [  168.220185]  __kasan_kmalloc.constprop.0+0xc2/0xd0
>>>> [  168.220416]  kmem_cache_alloc+0xe0/0x290
>>>> [  168.220607]  dup_fd+0x4e/0x5a0
>>>> [  168.220758]  copy_process+0xe35/0x2bf0
>>>> [  168.220942]  _do_fork+0xd8/0x550
>>>> [  168.221102]  __do_sys_clone+0xb5/0xe0
>>>> [  168.221282]  do_syscall_64+0x5e/0xe0
>>>> [  168.221457]  entry_SYSCALL_64_after_hwframe+0x49/0xb3
>>>> [  168.221729]
>>>> [  168.221848] Freed by task 41759:
>>>> [  168.222088]  save_stack+0x1b/0x40
>>>> [  168.222336]  __kasan_slab_free+0x12f/0x180
>>>> [  168.222632]  slab_free_freelist_hook+0x4d/0x120
>>>> [  168.222959]  kmem_cache_free+0x90/0x2e0
>>>> [  168.223239]  do_exit+0x5d2/0x12e0
>>>> [  168.223482]  do_group_exit+0x6f/0x130
>>>> [  168.223754]  __x64_sys_exit_group+0x28/0x30
>>>> [  168.224061]  do_syscall_64+0x5e/0xe0
>>>> [  168.224326]  entry_SYSCALL_64_after_hwframe+0x49/0xb3
>>>> [  168.224686]
>>>>
>>>> which indicates that current->files is no longer valid.
>>>
>>> Narrowed it down to the test/open-close test, and in particular where
>>> it closes the ring itself:
>>>
>>> ret = test_close(&ring, ring.ring_fd, 1);
>>>
>>> This seems to be because you do io_req_init_async() after calling
>>> io_issue_sqe(), and the command handler may have set something
>>> else for ->func at that point. Hence we never call the right
>>> handler if the close needs to be deferred, as it needs to for
>>> the io_uring as it has ->flush() defined.
>>>
>>> Why isn't io_req_init_async() just doing:
>>>
>>> static inline void io_req_init_async(struct io_kiocb *req,
>>>                           void (*func)(struct io_wq_work **))
>>> {
>>>          if (!(req->flags & REQ_F_WORK_INITIALIZED)) {
>>>                  req->work = (struct io_wq_work){ .func = func };
>>>                  req->flags |= REQ_F_WORK_INITIALIZED;
>>>          }
>>> }
>>>
>>> ?
>>
>> I guess that won't always work, if the request has been deferred and
>> we're now setting a new work func. So we really want the 'reset to
>> io_wq_submit_work' to only happen if the opcode hasn't already set
>> a private handler. Can you make that change?
>>
>> Also please fix up missing braces. The cases of:
>>
>> if (something) {
>>     line 1
>>     line 2
>> } else
>>     line 3
>>
>> should always includes braces, if one clause has it.
>>
>> A v5 with those two things would be ready to commit.
> First thanks for figuring out this bug.
> Indeed, I had run test cases in liburing/test multiple rounds before sending
> V4 patches, and all test cases pass, no kernel issues occurred, here is my kernel
> build config, seems that I had kasan enabled.
> [lege@localhost linux]$ cat .config | grep KASAN
> CONFIG_KASAN_SHADOW_OFFSET=0xdffffc0000000000
> CONFIG_HAVE_ARCH_KASAN=y
> CONFIG_HAVE_ARCH_KASAN_VMALLOC=y
> CONFIG_CC_HAS_KASAN_GENERIC=y
> CONFIG_KASAN=y
> CONFIG_KASAN_GENERIC=y
> CONFIG_KASAN_OUTLINE=y
> # CONFIG_KASAN_INLINE is not set
> CONFIG_KASAN_STACK=1
> # CONFIG_KASAN_VMALLOC is not set
> # CONFIG_TEST_KASAN is not set
> But I don't know why I did't reproduce the bug, sorry.
> 
> Are you fine below changes?
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index 12296ce3e8b9..2a3a02838f7b 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -907,9 +907,10 @@ static void io_file_put_work(struct work_struct *work);
>   static inline void io_req_init_async(struct io_kiocb *req,
>                          void (*func)(struct io_wq_work **))
>   {
> -       if (req->flags & REQ_F_WORK_INITIALIZED)
> -               req->work.func = func;
> -       else {
> +       if (req->flags & REQ_F_WORK_INITIALIZED) {
> +               if (!req->work.func)
> +                       req->work.func = func;
> +       } else {
>                  req->work = (struct io_wq_work){ .func = func };
>                  req->flags |= REQ_F_WORK_INITIALIZED;
>          }
> @@ -2920,6 +2921,8 @@ static int __io_splice_prep(struct io_kiocb *req,
>                  return ret;
>          req->flags |= REQ_F_NEED_CLEANUP;
> 
> +       /* Splice will be punted aync, so initialize io_wq_work firstly_*/
> +       io_req_init_async(req, io_wq_submit_work);
>          if (!S_ISREG(file_inode(sp->file_in)->i_mode))
>                  req->work.flags |= IO_WQ_WORK_UNBOUND;
> 
> @@ -3592,6 +3595,9 @@ static int io_statx(struct io_kiocb *req, bool force_nonblock)
> 
>   static int io_close_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>   {
> +        /* Close may be punted aync, so initialize io_wq_work firstly */
> +       io_req_init_async(req, io_wq_submit_work);
> +
For splice and close requests, these two about io_req_init_async() calls should be
io_req_init_async(req, NULL), because they change req->work.flags firstly.

Regards,
Xiaoguang Wang

>          /*
>           * If we queue this for async, it must not be cancellable. That would
>           * leave the 'file' in an undeterminate state.
> 
> Regards,
> Xiaoguang Wang
> 
> 
>>