Re: [PATCH] xen/mm: avoid watchdog timeout in dump_numa() on large domains

["Roger Pau Monné" <roger.pau@citrix.com>]

On Tue, Jun 02, 2026 at 01:51:13PM +0200, Jan Beulich wrote:
> On 02.06.2026 10:49, Bernhard Kaindl wrote:
> > Using the 'u' debug key invokes dump_numa(), which walks each domain's
> > page list under page_alloc_lock to compute per-NUMA-node counts. On
> > domains with many pages, this O(pages) operation can hold the lock long
> > enough to trigger a watchdog timeout.
> 
> In addition to what Roger said: Is it really the lock holding that's a
> problem here? That is, there would be no problem if there was no lock
> involved in this O(pages) operation?
> 
> > Replace the page-list walk with node_tot_pages[], a per-node counter
> > maintained in struct domain. This reduces dump_numa()'s per-domain work
> > from O(pages) to O(nodes).
> 
> Alternative approch for consideration: Purge dump_numa()? This big a
> change for making a keyhandler work better is somewhat questionable an
> approach, imo. The keyhandler isn't there for use in production anyway,
> it's (primarily) a debugging aid. If the data is still needed (and may
> e.g. be useful on production systems), make a (preemptible) domctl or
> sysctl or alike instead?

At some point a similar change will be needed for per-node memory
claims.  While we might refuse just for dump_numa(), the intrusive
changes in memory_exchange() are a requirement for per-node claims
IIRC.

Thanks, Roger.