Re: [PATCH 5/5] zfs extensible_dataset and large_blocks feature support

[Toomas Soome <tsoome@me.com>]

> On 04.05.2015, at 7:32, Andrei Borzenkov <arvidjaar@gmail.com> wrote:
> 
> В Sun, 03 May 2015 22:47:40 +0300
> Toomas Soome <tsoome@me.com> пишет:
> 
>> 
>>> On 03.05.2015, at 22:23, Andrei Borzenkov <arvidjaar@gmail.com> wrote:
>>> 
>>>>> 
>>>> 
>>>> they do implement read support extensible dataset… instead of fixed DMU types they dont specify type, making it possible to use fat zap objects from bonus area. 
>>>> 
>>>> reference commit:
>>>> https://github.com/illumos/illumos-gate/commit/2acef22db7808606888f8f92715629ff3ba555b9
>>>> 
>>> 
>>> OK; just to be sure - there are several remaining uses of
>>> DMU_OT_DSL_DATASET and DMU_OT_DSL_DIR; are these intentional?
>> 
>> 
>> they *should* be ok. at least browsing and loading files from datasets haven’t revealed other instances over the test time. but problem is, the cases of zapify (when “normal” structure is wrapped as zap object) I have seen, have been related to datasets using >128k recordsize, and its possible some cases may have been missed - the grub zfs code and real zfs code are not exactly easy to compare. at least so far I haven’t seen bugs related to this nor received any reports. the basic symptom can be either message about wrong type or “missing” directory entry in listing.
>> 
> 
> 
> Can https://savannah.gnu.org/bugs/?44448 be somehow related?

without grub debug=zfs its impossible to say - guessing there is not really helpful:) altho in general, feature check for label should generate error message if pool has enabled features not supported by version of grub used there.

basically, my first suggestion to cases like that is to set debug=zfs in grub and ask user to provide that debug output.

rgds,
toomas