Re: [Linux-v3.19-9526-ga135c717d5cd + vfs.git] blk_update_request: I/O error, dev loop0

[Jens Axboe <axboe@fb.com>]

On 02/22/2015 12:12 PM, Sedat Dilek wrote:
> On Sun, Feb 22, 2015 at 9:04 PM, Jens Axboe <axboe@fb.com> wrote:
>> On Feb 22, 2015, at 12:02 PM, Sedat Dilek <sedat.dilek@gmail.com> wrote:
>>>
>>>> On Sun, Feb 22, 2015 at 5:06 PM, Jens Axboe <axboe@fb.com> wrote:
>>>>> On 02/22/2015 06:09 AM, Sedat Dilek wrote:
>>>>>
>>>>> Hi,
>>>>>
>>>>> when testing Linux-next (upcoming v3.20) I fell over similiar messages
>>>>> when running fio here.
>>>>>
>>>>> This commit helped which is now upstream [1]...
>>>>>
>>>>> commit 9f9ee1f2b2f94f19437ae2def7c0d6636d7fe02e
>>>>> "block: Quiesce zeroout wrapper"
>>>>>
>>>>> --- dmesg_3.19.0-9526.2-iniza-small_before-fio-2.2.5.txt
>>>>> 2015-02-22 14:09:35.411824880 +0100
>>>>> +++ dmesg_3.19.0-9526.2-iniza-small_after-fio-2.2.5.txt 2015-02-22
>>>>> 14:10:25.327824500 +0100
>>>>> @@ -853,3 +853,25 @@
>>>>>   [  428.226000] Adding 36k swap on swapfile29.  Priority:-29 extents:1
>>>>> across:36k FS
>>>>>   [  469.339645] Process 7691(waitpid02) has RLIMIT_CORE set to 1
>>>>>   [  469.339650] Aborting core
>>>>> +[ 1877.189057] fio-testcase.sh (10238): drop_caches: 3
>>>>> +[ 1882.250174] blk_update_request: I/O error, dev loop0, sector 2883712
>>>>> +[ 1882.807993] blk_update_request: I/O error, dev loop0, sector 32460632
>>>>> +[ 1884.655458] blk_update_request: I/O error, dev loop0, sector 16869712
>>>>> +[ 1884.656896] blk_update_request: I/O error, dev loop0, sector 16854368
>>>>> +[ 1884.659316] blk_update_request: I/O error, dev loop0, sector 16855008
>>>>> +[ 1884.660834] blk_update_request: I/O error, dev loop0, sector 16869792
>>>>> +[ 1884.661780] blk_update_request: I/O error, dev loop0, sector 16854848
>>>>> +[ 1884.663282] blk_update_request: I/O error, dev loop0, sector 16855432
>>>>> +[ 1884.664775] blk_update_request: I/O error, dev loop0, sector 16859808
>>>>> +[ 1884.671472] blk_update_request: I/O error, dev loop0, sector 16866400
>>>>> +[ 1892.239383] blk_update_request: 20 callbacks suppressed
>>>>> +[ 1892.239389] blk_update_request: I/O error, dev loop0, sector 7872768
>>>>> +[ 1897.221576] blk_update_request: I/O error, dev loop0, sector 23331008
>>>>> +[ 1898.299271] blk_update_request: I/O error, dev loop0, sector 25690688
>>>>> +[ 1898.444161] blk_update_request: I/O error, dev loop0, sector 25726496
>>>>> +[ 1898.445946] blk_update_request: I/O error, dev loop0, sector 25690832
>>>>> +[ 1898.448185] blk_update_request: I/O error, dev loop0, sector 25692312
>>>>> +[ 1898.449321] blk_update_request: I/O error, dev loop0, sector 25704840
>>>>> +[ 1898.450314] blk_update_request: I/O error, dev loop0, sector 25710392
>>>>> +[ 1898.451351] blk_update_request: I/O error, dev loop0, sector 25726080
>>>>> +[ 1898.452348] blk_update_request: I/O error, dev loop0, sector 25737608
>>>>>
>>>>> FYI: This is with Linux-v3.19-9526-ga135c717d5cd plus vfs.git#for-linus.
>>>>
>>>>
>>>> This really has nothing to do with commit 9f9ee1f2b2f, other than that they
>>>> both have some block related messages. The above is normal IO errors on the
>>>> device, we expect to see messages related to that. The question is mostly
>>>> why you are seeing those. What are you running on the device?
>>>>
>>>> And is this a regression since 3.19?
>>>
>>> OK, I mixed it up - I was on a hurry travelling back home.
>>>
>>> This is (still) a Ubuntu/precise AMD64 WUBI system where I did the
>>> block-loopmq testing.
>>>
>>> It seems to be a regression since 3.19 (see my attached tarball).
>>
>> All I get is an attached file for a checksum, doesn't really help me in any way.
>>
>
> It's 2 files, I have unpacked it and attached all single files.

BTW, the other messages are the same (just the sha attached). And like 
Ming said, I don't see the warnings in the files you attached. So lets 
back it up a bit. A good bug report contains:

1) What was being run (kernel version)
2) What was run (fio command, in this case). This includes setup of how 
to reproduce.
3) Result
4) Expected result

Otherwise it ends up being way too obtuse, people have to guess at what 
you (the reporter) thinks is wrong, in the cases where this isn't 
immediately obvious.

So lets get back to basics on this one. What, in your opinion, has 
regressed in 4.0-rc1 compared to 3.19? And what is the exact test case?

-- 
Jens Axboe