diff options
author | Gerald Schaefer <gerald.schaefer@linux.ibm.com> | 2024-11-21 18:45:20 +0100 |
---|---|---|
committer | Heiko Carstens <hca@linux.ibm.com> | 2024-11-27 12:55:20 +0100 |
commit | ae1b9fb2d556e95001399dcd4e228525aead5122 (patch) | |
tree | c40cf63c96cf1946ec404ba3ac6f38075a570ea8 /samples/fprobe | |
parent | f93d6d62e4697b0df1474f0b1314c3c158335f0a (diff) |
s390/mm: Rearrange region-third and segment table entry SW bits
Rearrange region-third and segment table entry SW bits, in order to
make room for future encoding of region/segment table swap entries.
Also adjust _SEGMENT_ENTRY_GMAP_UC and _SEGMENT_ENTRY_GMAP_IN bits in
gmap code. Those should only apply for gmap PMDs, and not really depend
on or conflict with host PMD bits, but for consistency also adjust them:
- _SEGMENT_ENTRY_GMAP_UC "dirty (migration)" was using the same bit as
_SEGMENT_ENTRY_SOFT_DIRTY in the host PMD -> make it use the new
SOFT_DIRTY bit 63 (0x0002)
- _SEGMENT_ENTRY_GMAP_IN "invalidation notify bit" was using 0x8000,
which was an unused bit in the host PMD, that is now used for
_SEGMENT_ENTRY_WRITE -> make it use bit 52 (0x0800) instead, which is
still unused in the host PMD
This is a prerequisite for hugetlbfs PTE_MARKER support on s390, which
is needed to fix a regression introduced with commit 8a13897fb0da
("mm: userfaultfd: support UFFDIO_POISON for hugetlbfs"). That commit
depends on the availability of swap entries for hugetlbfs, which were
not available for s390 so far.
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Diffstat (limited to 'samples/fprobe')
0 files changed, 0 insertions, 0 deletions