[bug] btrfs fi usage reports bogus 16EiB free space

[bug] btrfs fi usage reports bogus 16EiB free space

[Chris Murphy]

Filed a bug:
https://bugzilla.kernel.org/show_bug.cgi?id=110111

The gist is that with a full USB stick, 'btrfs fi usage' reports:

    Free (estimated):  16.00EiB (min: 16.00EiB)

Which is clearly wrong, but very exciting to imagine what such a thing
would look like!


-- 
Chris Murphy

Re: [bug] btrfs fi usage reports bogus 16EiB free space

[Duncan]

Chris Murphy posted on Mon, 28 Dec 2015 17:12:13 -0700 as excerpted:

> Filed a bug:
> https://bugzilla.kernel.org/show_bug.cgi?id=110111
> 
> The gist is that with a full USB stick, 'btrfs fi usage' reports:
> 
>     Free (estimated):  16.00EiB (min: 16.00EiB)
> 
> Which is clearly wrong, but very exciting to imagine what such a thing
> would look like!

[After double-checking the bug report to ensure it was as I suspected, as 
the critical line wasn't included above.]

ENOTIMPLEMENTED.  See that "WARNING: MIXED blockgroups not handled" line 
at the top? =:^)

FWIW I've been seeing this on my /boot and its backup since original fi 
usage implementation.

The patches implementing usage didn't cover the mixed-bg case, except 
with the warning, but were still taken because the report is useful in 
the default case.  IIRC it did something similar for raid56 mode, which 
wasn't complete at the time fi usage was introduced and thus wasn't 
implemented either.

Which is one reason (besides not being in older versions) that we still 
request btrfs fi df and btrfs fi show both, instead of btrfs fi usage.

It's still interesting to imagine a 16 EiB device (not even an array, a 
single device, imagine an array...), tho. Kinda like reading
Hitchhiker's Guide to the Galaxy for the first time, rather mind 
stretching!  =:^)

Of course back a few years ago TiB or even GiB would have seemed equally 
mind stretching...  I remember when the memory in my machine first 
surpassed the size of the entire hard drive I had in my first 486sx25, 4 
MiB RAM, 130 MB hard drive... and I counted myself lucky just to get a 
486 instead of a 386...   Yes, there's kids messing up my lawn too; I 
shout and shake my cane at 'em sometimes. =:^)

(But 16 EiB is STILL mind stretching!)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [bug] btrfs fi usage reports bogus 16EiB free space

[Chris Murphy]

On Mon, Dec 28, 2015 at 7:22 PM, Duncan <1i5t5.duncan@cox.net> wrote:
> Chris Murphy posted on Mon, 28 Dec 2015 17:12:13 -0700 as excerpted:
>
>> Filed a bug:
>> https://bugzilla.kernel.org/show_bug.cgi?id=110111
>>
>> The gist is that with a full USB stick, 'btrfs fi usage' reports:
>>
>>     Free (estimated):  16.00EiB (min: 16.00EiB)
>>
>> Which is clearly wrong, but very exciting to imagine what such a thing
>> would look like!
>
> [After double-checking the bug report to ensure it was as I suspected, as
> the critical line wasn't included above.]
>
> ENOTIMPLEMENTED.  See that "WARNING: MIXED blockgroups not handled" line
> at the top? =:^)

Unrelated. When the mkfs is new and nothing is on the fs, I get this:

    Free (estimated):           7.44GiB    (min: 7.44GiB)

I'm not certain at what point the fs starts to report 16.00EiB but
it's not immediate.




-- 
Chris Murphy

Re: [bug] btrfs fi usage reports bogus 16EiB free space

[Chris Murphy]

On Mon, Dec 28, 2015 at 7:29 PM, Chris Murphy <lists@colorremedies.com> wrote:
> On Mon, Dec 28, 2015 at 7:22 PM, Duncan <1i5t5.duncan@cox.net> wrote:
>> Chris Murphy posted on Mon, 28 Dec 2015 17:12:13 -0700 as excerpted:
>>
>>> Filed a bug:
>>> https://bugzilla.kernel.org/show_bug.cgi?id=110111
>>>
>>> The gist is that with a full USB stick, 'btrfs fi usage' reports:
>>>
>>>     Free (estimated):  16.00EiB (min: 16.00EiB)
>>>
>>> Which is clearly wrong, but very exciting to imagine what such a thing
>>> would look like!
>>
>> [After double-checking the bug report to ensure it was as I suspected, as
>> the critical line wasn't included above.]
>>
>> ENOTIMPLEMENTED.  See that "WARNING: MIXED blockgroups not handled" line
>> at the top? =:^)
>
> Unrelated. When the mkfs is new and nothing is on the fs, I get this:
>
>     Free (estimated):           7.44GiB    (min: 7.44GiB)
>
> I'm not certain at what point the fs starts to report 16.00EiB but
> it's not immediate.


Looks like "fi usage" Used amount overestimates used by just over 2x
compared to "fi df" reporting. Once the device is about 48% full, "fi
usage" says it's full, and that's when free space shoots off to 16EiB
probably because as you say mixedbg isn't supported, and what's
probably happening is free space has gone negative, hence the 16EiB
value.

I'll update the bug.



-- 
Chris Murphy

Re: [bug] btrfs fi usage reports bogus 16EiB free space

[Christoph Anton Mitterer]

[-- Attachment #1: Type: text/plain, Size: 576 bytes --]

Hey Chris.

On Mon, 2015-12-28 at 19:29 -0700, Chris Murphy wrote:
> Unrelated. When the mkfs is new and nothing is on the fs, I get this:
> 
>     Free (estimated):           7.44GiB    (min: 7.44GiB)
> 
> I'm not certain at what point the fs starts to report 16.00EiB but
> it's not immediate.
I had something similar a while ago:
http://thread.gmane.org/gmane.comp.file-systems.btrfs/49955

Filipe thought there would have been a patch for something, but I
couldn't check it anymore (or haven't found the time to do so).


Cheers,
Chris.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 5313 bytes --]