Re: [PATCH 3/9] xfs: don't use xlog_in_core_2_t in struct xlog_in_core

["Darrick J. Wong" <djwong@kernel.org>]

On Wed, Oct 15, 2025 at 06:34:22AM +0200, Christoph Hellwig wrote:
> On Tue, Oct 14, 2025 at 02:47:42PM -0700, Darrick J. Wong wrote:
> > On Mon, Oct 13, 2025 at 11:42:07AM +0900, Christoph Hellwig wrote:
> > > Most accessed to the on-disk log record header are for the original
> > > xlog_rec_header.  Make that the main structure, and case for the
> > > single remaining place using other union legs.
> > > 
> > > This prepares for removing xlog_in_core_2_t entirely.
> > 
> > Er... so xlog_rec_header is the header that gets written out at the
> > start of any log buffer?
> 
> Yes.
> 
> > And if a log record has more than
> > XLOG_CYCLE_DATA_SIZE basic blocks (BBs) in it, then it'll have some
> > quantity of "extended" headers in the form of a xlog_rec_ext_header
> > right after the xlog_rec_header, right?
> 
> They are not directly behіnd the current definition of the
> xlog_rec_header, but rather at each multiple of 512 bytes past the
> start of the xlog_rec_header.

<nod> And that only became obvious after I read through that patch that
removes xlog_in_core_2_t.

> > And both the regular and ext
> > headers both have a __be32 array containing the original first four
> > bytes of each BB, because each BB has a munged version of the LSN cycle
> > stamped into the first four bytes, right?
> 
> Yes.
> 
> > The previous patch refactored how the cycle_data transformation
> > happened, right?
> 
> Yes.
> 
> > So this patch just gets rid of the strange ic_header #define, and
> > updates the code to access ic_data->hic_header directly?  And now that
> > we have xlog_cycle_data to abstract the xlog_rec_header ->
> > xlog_in_core_2_t casting, this just works fine here.  Right?
> 
> Yes.

I'm satisfied then.
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

--D