Re: [PATCH v2 10/13] xfs: Unmap blocks according to forcealign

[John Garry]

On 09/07/2024 08:46, Christoph Hellwig wrote:

Hi Christoph,

> > Using isfa (instead of isforcealign) might be interpreted as something else
> The check should be used in one single place where we decided if
> we need to to the alignment based adjustments.  So IMHO just killing
> it and open coding it there seems way easier.   Yes, it is in a loop,
> but compared to all the work done is is really cheap.
>
> > > We've been long wanting to split the whole align / convert unwritten /
> > > etc code into a helper outside the main bumapi flow.  And when adding
> > > new logic to it this might indeed be a good time.
> > ok, I'll see if can come up with something
> I can take a look too. 

I was wondering what you plans are for any clean-up/refactoring here, as 
mentioned?

I was starting to look at the whole "if (forcealign) else if (big rt)" 
flow refactoring in this series to use xfs_inode_alloc_unitsize(); 
however, I figure that you have plans wider in scope, which affects this.

? There is some real mess in there like trying
> to account for cases where the transaction doesn't have a block
> reservation, which I think could have happen in truncate until
> Zhang Yi fixed it for the 6.11 merge window.

Cheers,
John