On Tue, Jun 09, 2026 at 08:11:42PM +0200, David Hildenbrand (Arm) wrote:
> On 6/9/26 17:35, Gregory Price wrote:
> > On Tue, Jun 09, 2026 at 12:26:07PM +0200, David Hildenbrand (Arm) wrote:
> >> On 6/5/26 23:19, Gregory Price wrote:
> >>
> >>>
> >>> +static int dax_kmem_parse_state(const char *buf)
> >>> +{
> >>> + if (sysfs_streq(buf, "unplugged"))
> >>> + return DAX_KMEM_UNPLUGGED;
> >>> + if (sysfs_streq(buf, "online"))
> >>> + return MMOP_ONLINE;
> >>> + if (sysfs_streq(buf, "online_kernel"))
> >>> + return MMOP_ONLINE_KERNEL;
> >>> + if (sysfs_streq(buf, "online_movable"))
> >>> + return MMOP_ONLINE_MOVABLE;
> >>> + return -EINVAL;
> >>
> >> Should we try making use of mhp_online_type_from_str()/online_type_to_str()
> >> [possibly a nicer exported function for the latter] to avoid duplicating
> >> this ...
> >>
> >
> > In patch 6 response i point out adding MMOP_UNPLUGGED
> >
> > If we add MMOP_UNPLUGGED as a state that is only use by callers of
> > memory hotplug to represent the current state - but not as a valid input
> > to memory_hotplug.c - then we can simply this as you request.
> >
> > Although we'll need to add a couple lines to memoryN/state parsing code
> > to disallow MMOP_UNPLUGGED as a valid input (otherwise you could
> > permanently unplug memory without the ability to get it back... unless
> > you want that? Seems useless to me.)
>
> I would just special-case your KMEM-special value, and then branch off to the
> core helpers?
>
> static int dax_kmem_parse_state(const char *buf)
> {
> if (sysfs_streq(buf, "unplugged"))
> return DAX_KMEM_UNPLUGGED;
> return mhp_online_type_from_str(...);
> }
>
> So there is no need to add a value to MM core?
>
Was going to propose this, but then I thought... well if CXL detactes
itself from dax and does this same pattern, maybe MMOP_UNPLUGGED makes
sense.
But i'll leave that for another day and do as you suggest.
Thanks!
~Gregory
