[PATCH v2 22/29] maple_tree: Use maple copy node for mas_wr_rebalance() operation

[Liam R. Howlett]

* Mark Brown <broonie at kernel.org> [260128 05:53]:
> On Tue, Jan 27, 2026 at 11:05:54PM +0000, Mark Brown wrote:
> > On Wed, Jan 21, 2026 at 11:45:19AM -0500, Liam R. Howlett wrote:
> 
> > > Stop using the maple big node for rebalance operations by changing to
> > > more align with spanning store.  The rebalance operation needs its own
> > > data calculation in rebalance_data().
> 
> > > In the event of too much data, the rebalance tries to push the data
> > > using push_data_sib().  If there is insufficient data, the rebalance
> > > operation will rebalance against a sibling (found with rebalance_sib()).
> 
> > These appear to bisect to this commit, there were some issues with the
> > setup of the bisect which fortunately don't seem to have confused things
> > and some timeouts stopped the last couple of jobs completing but we've
> > got adjacent commits showing passes and fails and of the two candidates
> > the other is a refactoring that doesn't look at all plausible.  I'm
> > rerunning a clean bisect but expect it to confirm this result.
> 
> Confirmed:
> 
> # first bad commit: [6ad069085c621d39fad4d1835af7e563975dcf7e] maple_tree: use maple copy node for mas_wr_rebalance() operation


Thanks for bisecting and the test name (linkat02 on arm64) from ltp, I
really appreciate the details on this.

I'm working on a fix.  Please let me know if any other platforms or
tests showed errors and I'll try them here before it's re-added to
-next.

Cheers,
Liam