Re: Current linux-next status of UBI/UBIFS? Re: new UBI co-maintainer

[Richard Weinberger <richard@nod.at>]

Am 29.01.2015 um 04:28 schrieb hujianyang:
> I wish it could be soon, because there are some other patches about
> mounting reliability improvement I want to send. Could I send them
> together? But I haven't finished this work, so I think one by one
> is the best way, and I can discuss the designing with you and Artem
> in time.

Please send them and describe the issues in details, provide a test case, etc...
It is also important to tell us *how* this bug can happen.

>>>    [PATCH] UBI: fix soft lockup in ubi_check_volume()
>>
>> I'm also fine with that.
>>
>>>    and ubidump v6
>>
>> For ubidump I have to read the whole thread again as I got lost in it. :)
>>
>>> The patch "UBI: fix soft lockup in ubi_check_volume()" solves a really
>>> important problem, I wish it could be merged into stable soon.
>>
>> Yes.
>>
>>> I've discussed with Richard about the recovery of an corrupted UBIFS
>>> image, for example, ECC error. And actually my colleagues and I had
>>> worked out some features to improve the reliability of UBIFS. We are
>>> happy to share our design and greatly aspire the help from community.
>>
>> Nice.
>>
>>> Also, I think we could start to add more functions to my ubidump to
>>> make it a useful tool.
>>
>> I have to look at ubidump in detail but it sounds good.
>> To debug fastmap issues I have also a tool to analyze UBI images.
>> Maybe we can merge. First I have to shape it up.
>>
> 
> I'm glad with it. I was using ubidump for debugging these days,
> but I'm not sure if this v6 is OK. I'd fixed some tiny issues
> after v5 so I don't know if there are still other issues left.
> Thanks for your reviewing.
> 
> I see ubifastmap is marked as "Experimental feature" now and
> you'd introduced multi-queue for ubiblock. I think some works,
> e.g. testing, are really needed before importing these feature
> into our products.

Fastmap is experimental for good reasons. :)
With my current patches nobody was able to kill it, so maybe I'll
remove the experimental tag in v3.25.

> By the way, I found some products in my company are using
> MTD_UBI_GLUEBI to implement a squashfs on top of UBI device.
> UBI_GLUEBI or UBI_BLOCK, which is better in your considering,
> and why?

UBI Block, it was designed for that case.
Just compare the number of layers. :)

Thanks,
//richard