Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'for-5.7/dm-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper fixes from Mike Snitzer:
- Document DM integrity allow_discard feature that was added during 5.7
merge window.
- Fix potential for DM writecache data corruption during DM table
reloads.
- Fix DM verity's FEC support's hash block number calculation in
verity_fec_decode().
- Fix bio-based DM multipath crash due to use of stale copy of
MPATHF_QUEUE_IO flag state in __map_bio().
* tag 'for-5.7/dm-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
dm multipath: use updated MPATHF_QUEUE_IO on mapping for bio-based mpath
dm verity fec: fix hash block number in verity_fec_decode
dm writecache: fix data corruption when reloading the target
dm integrity: document allow_discard option
+50
-23
+8
-5
Documentation/admin-guide/device-mapper/dm-integrity.rst
+8
-5
Documentation/admin-guide/device-mapper/dm-integrity.rst
···
182
182
space-efficient. If this option is not present, large padding is
183
183
used - that is for compatibility with older kernels.
184
184
185
+
allow_discards
186
+
Allow block discard requests (a.k.a. TRIM) for the integrity device.
187
+
Discards are only allowed to devices using internal hash.
185
188
186
-
The journal mode (D/J), buffer_sectors, journal_watermark, commit_time can
187
-
be changed when reloading the target (load an inactive table and swap the
188
-
tables with suspend and resume). The other arguments should not be changed
189
-
when reloading the target because the layout of disk data depend on them
190
-
and the reloaded target would be non-functional.
189
+
The journal mode (D/J), buffer_sectors, journal_watermark, commit_time and
190
+
allow_discards can be changed when reloading the target (load an inactive
191
+
table and swap the tables with suspend and resume). The other arguments
192
+
should not be changed when reloading the target because the layout of disk
193
+
data depend on them and the reloaded target would be non-functional.
191
194
192
195
193
196
The layout of the formatted block device:
+4
-2
drivers/md/dm-mpath.c
+4
-2
drivers/md/dm-mpath.c
···
585
585
586
586
/* Do we need to select a new pgpath? */
587
587
pgpath = READ_ONCE(m->current_pgpath);
588
-
queue_io = test_bit(MPATHF_QUEUE_IO, &m->flags);
589
-
if (!pgpath || !queue_io)
588
+
if (!pgpath || !test_bit(MPATHF_QUEUE_IO, &m->flags))
590
589
pgpath = choose_pgpath(m, bio->bi_iter.bi_size);
590
+
591
+
/* MPATHF_QUEUE_IO might have been cleared by choose_pgpath. */
592
+
queue_io = test_bit(MPATHF_QUEUE_IO, &m->flags);
591
593
592
594
if ((pgpath && queue_io) ||
593
595
(!pgpath && test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags))) {
+1
-1
drivers/md/dm-verity-fec.c
+1
-1
drivers/md/dm-verity-fec.c
+37
-15
drivers/md/dm-writecache.c
+37
-15
drivers/md/dm-writecache.c
···
931
931
return 0;
932
932
}
933
933
934
+
static int writecache_read_metadata(struct dm_writecache *wc, sector_t n_sectors)
935
+
{
936
+
struct dm_io_region region;
937
+
struct dm_io_request req;
938
+
939
+
region.bdev = wc->ssd_dev->bdev;
940
+
region.sector = wc->start_sector;
941
+
region.count = n_sectors;
942
+
req.bi_op = REQ_OP_READ;
943
+
req.bi_op_flags = REQ_SYNC;
944
+
req.mem.type = DM_IO_VMA;
945
+
req.mem.ptr.vma = (char *)wc->memory_map;
946
+
req.client = wc->dm_io;
947
+
req.notify.fn = NULL;
948
+
949
+
return dm_io(&req, 1, ®ion, NULL);
950
+
}
951
+
934
952
static void writecache_resume(struct dm_target *ti)
935
953
{
936
954
struct dm_writecache *wc = ti->private;
···
959
941
960
942
wc_lock(wc);
961
943
962
-
if (WC_MODE_PMEM(wc))
944
+
if (WC_MODE_PMEM(wc)) {
963
945
persistent_memory_invalidate_cache(wc->memory_map, wc->memory_map_size);
946
+
} else {
947
+
r = writecache_read_metadata(wc, wc->metadata_sectors);
948
+
if (r) {
949
+
size_t sb_entries_offset;
950
+
writecache_error(wc, r, "unable to read metadata: %d", r);
951
+
sb_entries_offset = offsetof(struct wc_memory_superblock, entries);
952
+
memset((char *)wc->memory_map + sb_entries_offset, -1,
953
+
(wc->metadata_sectors << SECTOR_SHIFT) - sb_entries_offset);
954
+
}
955
+
}
964
956
965
957
wc->tree = RB_ROOT;
966
958
INIT_LIST_HEAD(&wc->lru);
···
2130
2102
ti->error = "Invalid block size";
2131
2103
goto bad;
2132
2104
}
2105
+
if (wc->block_size < bdev_logical_block_size(wc->dev->bdev) ||
2106
+
wc->block_size < bdev_logical_block_size(wc->ssd_dev->bdev)) {
2107
+
r = -EINVAL;
2108
+
ti->error = "Block size is smaller than device logical block size";
2109
+
goto bad;
2110
+
}
2133
2111
wc->block_size_bits = __ffs(wc->block_size);
2134
2112
2135
2113
wc->max_writeback_jobs = MAX_WRITEBACK_JOBS;
···
2234
2200
goto bad;
2235
2201
}
2236
2202
} else {
2237
-
struct dm_io_region region;
2238
-
struct dm_io_request req;
2239
2203
size_t n_blocks, n_metadata_blocks;
2240
2204
uint64_t n_bitmap_bits;
2241
2205
···
2290
2258
goto bad;
2291
2259
}
2292
2260
2293
-
region.bdev = wc->ssd_dev->bdev;
2294
-
region.sector = wc->start_sector;
2295
-
region.count = wc->metadata_sectors;
2296
-
req.bi_op = REQ_OP_READ;
2297
-
req.bi_op_flags = REQ_SYNC;
2298
-
req.mem.type = DM_IO_VMA;
2299
-
req.mem.ptr.vma = (char *)wc->memory_map;
2300
-
req.client = wc->dm_io;
2301
-
req.notify.fn = NULL;
2302
-
2303
-
r = dm_io(&req, 1, ®ion, NULL);
2261
+
r = writecache_read_metadata(wc, wc->block_size >> SECTOR_SHIFT);
2304
2262
if (r) {
2305
-
ti->error = "Unable to read metadata";
2263
+
ti->error = "Unable to read first block of metadata";
2306
2264
goto bad;
2307
2265
}
2308
2266
}