On Wed, Jan 13, 2021 at 10:18:09PM -0800, Dan Williams wrote:

On Wed, Jan 13, 2021 at 5:50 PM HORIGUCHI NAOYA(堀口　直也)
[off-list ref] wrote:

quoted

On Wed, Jan 13, 2021 at 04:43:32PM -0800, Dan Williams wrote:

quoted

The conversion to move pfn_to_online_page() internal to
soft_offline_page() missed that the get_user_pages() reference taken by
the madvise() path needs to be dropped when pfn_to_online_page() fails.
Note the direct sysfs-path to soft_offline_page() does not perform a
get_user_pages() lookup.

When soft_offline_page() is handed a pfn_valid() &&
!pfn_to_online_page() pfn the kernel hangs at dax-device shutdown due to
a leaked reference.

Fixes: feec24a6139d ("mm, soft-offline: convert parameter to pfn")
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Cc: Michal Hocko <mhocko@kernel.org>
Reviewed-by: David Hildenbrand <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: <redacted>
Signed-off-by: Dan Williams <redacted>

I'm OK if we don't have any other better approach, but the proposed changes
make code a little messy, and I feel that get_user_pages() might be the
right place to fix. Is get_user_pages() expected to return struct page with
holding refcount for offline valid pages?  I thought that such pages are
only used by drivers for dax-devices, but that might be wrong. Can I ask for
a little more explanation from this perspective?

The motivation for ZONE_DEVICE is to allow get_user_pages() for "offline" pfns.

I missed this point, thank you.

soft_offline_page() wants to only operate on "online" pfns.

Right.

get_user_pages() on a dax-mapping returns an "offline" ZONE_DEVICE page.

OK.

When pfn_to_online_page() fails the get_user_pages() reference needs
to be dropped.

The background is clear to me, and I agree with this patch.

Reviewed-by: Naoya Horiguchi <redacted>

To be honest I dislike the entire flags based scheme for communicating
the fact that page reference obtained by madvise needs to be dropped
later. I'd rather pass a non-NULL 'struct page *' than redo
pfn_to_page() conversions in the leaf functions, but that's a much
larger change.

As Oscar mentions in another email, removing MF_COUNT_INCREASED was
recently discussed and rejected. I think that if pfn-page conversion
layer is somehow factored out from soft/hard offline handler, code would
be more readable/maintainable.

Thanks,
Naoya Horiguchi