On Thu, Mar 12, 2026 at 02:14:25PM -0700, Andrew Morton wrote: > On Thu, 12 Mar 2026 20:27:16 +0000 "Lorenzo Stoakes (Oracle)" > <[email protected]> wrote: > > > +int mmap_action_prepare(struct vm_area_desc *desc, > > + struct mmap_action *action) > > + > > { > > switch (action->type) { > > case MMAP_NOTHING: > > - break; > > + return 0; > > case MMAP_REMAP_PFN: > > - remap_pfn_range_prepare(desc, action->remap.start_pfn); > > - break; > > + return remap_pfn_range_prepare(desc, action); > > case MMAP_IO_REMAP_PFN: > > - io_remap_pfn_range_prepare(desc, action->remap.start_pfn, > > - action->remap.size); > > - break; > > + return io_remap_pfn_range_prepare(desc, action); > > } > > } > > EXPORT_SYMBOL(mmap_action_prepare); > > hm, was this the correct version? > > mm/util.c: In function 'mmap_action_prepare': > mm/util.c:1451:1: error: control reaches end of non-void function > [-Werror=return-type] > 1451 | }
Seems different compiler versions do different things :) In theory we should never hit that but memory corruption and err... rogue drivers? could cause it ofc :) Will fix on respin. > > --- a/mm/util.c~mm-various-small-mmap_prepare-cleanups-fix > +++ a/mm/util.c > @@ -1356,6 +1356,8 @@ int mmap_action_prepare(struct vm_area_d > return remap_pfn_range_prepare(desc, action); > case MMAP_IO_REMAP_PFN: > return io_remap_pfn_range_prepare(desc, action); > + default: > + BUG(); I'd probably prefer a WARN_ON_ONCE(1) return -EBLAH; will think about it on respin. > } > } > EXPORT_SYMBOL(mmap_action_prepare); > _ > Cheers, Lorenzo
