Re: [PATCH 00/17] btrfs: initial ctree.h cleanups, simple stuff

[David Sterba <dsterba@suse.cz>]

On Thu, Sep 15, 2022 at 05:47:59PM +0800, Qu Wenruo wrote:
> 
> 
> On 2022/9/14 23:06, Josef Bacik wrote:
> > Hello,
> > 
> > This is the first part in what will probably be very many parts in my work to
> > cleanup ctree.h.  These are the smaller changes, mostly removing code that's not
> > used anymore, moving some stuff that's local to C files that don't need to be in
> > the header at all, and moving the rest of the on-disk definition stuff to
> > btrfs_tree.h.  There's a lot of patche here, but they're all relatively small,
> > the largest being the patch to move the on-disk definitions to btrfs_tree.h,
> > which is not very large compared to patches in the next several series.  This
> > has been built and tested, it's relatively low risk.  Thanks,
> 
> Looks good overall to me.
> 
> Just 3 small points of concern:
> 
> - About btrfs_tree.h usage.
> 
>    Really hope David can make it clear that, that UAPI header is for ALL
>    on-disk format definition.
> 
>    I'm not buying the old reason that the UAPI header is only for tree
>    search ioctl, and really want to move the whole super block definition
>    into UAPI header.

I have merged the patch but am expecting problems once people start
using the headers. I'll be addressed case by case.

>    (And I also think all upstream fses should have a concentrated UAPI
>     header too, just to make new comers easier to hack)
> 
> - Some tiny inline functions got unexported
> 
>    May not be a big thing though, just want to make sure we have no other
>    choice but make them uninlined.

This depends, trivial functions or where it's on a hot path it's better
to keep inline.

> 
> - Extra error tags in init_btrfs_fs()
                ^^^^
                labels

>    Just like open_ctree(), it's when but not whether to have wrong jump.
>    Thus a new series to make those cachep related exit function
>    conditional. (aka, we can call them unconditionally at error path)

You've sent the series that converts it to the array and we're going to
it that way. Please refresh the series as there were some minor changes
to the init/exit functions and resend. Thanks.