Re: BUG: scheduling while atomic in blk_mq codepath?

[Jens Axboe <axboe@fb.com>]

On 2014-06-19 08:35, Theodore Ts'o wrote:
> While trying to bisect some problems which were introduced sometime
> between 3.15 and 3.16-rc1 (specifically, (1) reads to a block device
> at offset 262144 * 4k are failing with a short read, and (2) block
> device reads are sometimes causing the entire kernel to hang), the
> following BUG got hit.
>
> [    0.000000] Linux version 3.15.0-rc8-06047-gaaeb255 (tytso@closure) (gcc version 4.8.3 (Debian 4.8.3-2) ) #1902 SMP Thu Jun 19 11:16:10 EDT 2014
>
> [....] Checking file systems...fsck from util-linux 2.20.1
> /dev/vdg was not cleanly unmounted, check forced.
> [    4.161703] BUG: scheduling while atomic: fsck.ext4/2072/0x0000000266.5%
> [    4.163673] no locks held by fsck.ext4/2072.
> [    4.164318] Modules linked in:
> [    4.164845] CPU: 0 PID: 2072 Comm: fsck.ext4 Not tainted 3.15.0-rc8-06047-gaaeb255 #1902
> [    4.166047] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
> [    4.166917]  00000000 00000000 f52c5ba0 c0832655 f5158610 f52c5bac c082f88a f6501e40
> [    4.168188]  f52c5c20 c08362ca c0eb3e40 c0eb3e40 374d3933 00000001 0396a8da 00000000
> [    4.169474]  f5158610 f51f1674 f4f46a00 f52c5be4 c015dd4b f4f46a00 f52c5bf0 c015dd5e
> [    4.170781] Call Trace:
> [    4.171159]  [<c0832655>] dump_stack+0x48/0x60
> [    4.171838]  [<c082f88a>] __schedule_bug+0x5c/0x6d
> [    4.172572]  [<c08362ca>] __schedule+0x61/0x65a
> [    4.173228]  [<c015dd4b>] ? kvm_clock_read+0x1f/0x29
> [    4.173977]  [<c015dd5e>] ? kvm_clock_get_cycles+0x9/0xc
> [    4.174771]  [<c01b4cb9>] ? timekeeping_get_ns.constprop.14+0x10/0x56
> [    4.175701]  [<c0836922>] schedule+0x5f/0x61
> [    4.176345]  [<c0836aa2>] io_schedule+0x50/0x67
> [    4.177060]  [<c0423b2d>] bt_get+0xaf/0xd1
> [    4.177677]  [<c0198282>] ? wake_up_atomic_t+0x1f/0x1f
> [    4.178444]  [<c0423bfd>] blk_mq_get_tag+0x26/0x82
> [    4.179158]  [<c0420f14>] __blk_mq_alloc_request+0x2a/0x169
> [    4.180022]  [<c04222b5>] blk_mq_map_request+0x137/0x1e3
> [    4.180825]  [<c0422f89>] blk_sq_make_request+0x82/0x145
> [    4.181630]  [<c041a687>] generic_make_request+0x82/0xb5
> [    4.182430]  [<c041a7aa>] submit_bio+0xf0/0x109
> [    4.183113]  [<c019e97c>] ? trace_hardirqs_on_caller+0x14e/0x169
> [    4.184019]  [<c025de72>] _submit_bh+0x1ad/0x1ca
> [    4.184661]  [<c025de9e>] submit_bh+0xf/0x11
> [    4.185267]  [<c025f5c9>] block_read_full_page+0x1e2/0x1f2
> [    4.186073]  [<c025f8cd>] ? I_BDEV+0xa/0xa
> [    4.186695]  [<c020ad30>] ? __lru_cache_add+0x24/0x46
> [    4.187452]  [<c020af13>] ? lru_cache_add+0xd/0xf
> [    4.188130]  [<c025fc04>] blkdev_readpage+0x14/0x16
> [    4.188832]  [<c0209adf>] __do_page_cache_readahead+0x1c0/0x1eb
> [    4.189704]  [<c0209cb9>] ondemand_readahead+0x1af/0x1b9
> [    4.190508]  [<c0209d22>] page_cache_async_readahead+0x5f/0x6a
> [    4.191424]  [<c0202370>] generic_file_aio_read+0x226/0x4f4
> [    4.192272]  [<c0260841>] blkdev_aio_read+0x90/0x9e
> [    4.193017]  [<c02385cd>] do_sync_read+0x52/0x79
> [    4.193731]  [<c023857b>] ? fdput_pos+0x25/0x25
> [    4.194412]  [<c0238d27>] vfs_read+0x72/0xd1
> [    4.195064]  [<c02391da>] SyS_read+0x49/0x7c
> [    4.195700]  [<c083a0c9>] syscall_call+0x7/0xb
> [    4.196385]  [<c0830000>] ? print_usage_bug+0xcd/0x18e
>
> Is any of these known problems?  This is blocking me from doing any
> kind of testing at the moment...  (these problems are showing up while
> running KVM using virtio devices).

I believe you already reported this issue a while back, and it should be 
fixed by commit cb96a42c in the kernel.

The other issue, not sure, not a lot of detail. It may be fixed by the 
pull request I sent out yesterday. You can try pulling in:

git://git.kernel.dk/linux-block.git for-linus

and see if that fixes it for you.

-- 
Jens Axboe