Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client
Pull Ceph fixes from Sage Weil:
"Just two small items from Ilya:
The first patch fixes the RBD readahead to grab full objects. The
second fixes the write ops to prevent undue promotion when a cache
tier is configured on the server side"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
rbd: use writefull op for object size writes
rbd: set max_sectors explicitly
+17
-6
+8
-2
drivers/block/rbd.c
+8
-2
drivers/block/rbd.c
···
1863
1863
rbd_osd_read_callback(obj_request);
1864
1864
break;
1865
1865
case CEPH_OSD_OP_SETALLOCHINT:
1866
-
rbd_assert(osd_req->r_ops[1].op == CEPH_OSD_OP_WRITE);
1866
+
rbd_assert(osd_req->r_ops[1].op == CEPH_OSD_OP_WRITE ||
1867
+
osd_req->r_ops[1].op == CEPH_OSD_OP_WRITEFULL);
1867
1868
/* fall through */
1868
1869
case CEPH_OSD_OP_WRITE:
1870
+
case CEPH_OSD_OP_WRITEFULL:
1869
1871
rbd_osd_write_callback(obj_request);
1870
1872
break;
1871
1873
case CEPH_OSD_OP_STAT:
···
2403
2401
opcode = CEPH_OSD_OP_ZERO;
2404
2402
}
2405
2403
} else if (op_type == OBJ_OP_WRITE) {
2406
-
opcode = CEPH_OSD_OP_WRITE;
2404
+
if (!offset && length == object_size)
2405
+
opcode = CEPH_OSD_OP_WRITEFULL;
2406
+
else
2407
+
opcode = CEPH_OSD_OP_WRITE;
2407
2408
osd_req_op_alloc_hint_init(osd_request, num_ops,
2408
2409
object_size, object_size);
2409
2410
num_ops++;
···
3765
3760
/* set io sizes to object size */
3766
3761
segment_size = rbd_obj_bytes(&rbd_dev->header);
3767
3762
blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE);
3763
+
q->limits.max_sectors = queue_max_hw_sectors(q);
3768
3764
blk_queue_max_segments(q, segment_size / SECTOR_SIZE);
3769
3765
blk_queue_max_segment_size(q, segment_size);
3770
3766
blk_queue_io_min(q, segment_size);
+9
-4
net/ceph/osd_client.c
+9
-4
net/ceph/osd_client.c
···
285
285
switch (op->op) {
286
286
case CEPH_OSD_OP_READ:
287
287
case CEPH_OSD_OP_WRITE:
288
+
case CEPH_OSD_OP_WRITEFULL:
288
289
ceph_osd_data_release(&op->extent.osd_data);
289
290
break;
290
291
case CEPH_OSD_OP_CALL:
···
486
485
size_t payload_len = 0;
487
486
488
487
BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
489
-
opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE);
488
+
opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
489
+
opcode != CEPH_OSD_OP_TRUNCATE);
490
490
491
491
op->extent.offset = offset;
492
492
op->extent.length = length;
493
493
op->extent.truncate_size = truncate_size;
494
494
op->extent.truncate_seq = truncate_seq;
495
-
if (opcode == CEPH_OSD_OP_WRITE)
495
+
if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
496
496
payload_len += length;
497
497
498
498
op->payload_len = payload_len;
···
672
670
break;
673
671
case CEPH_OSD_OP_READ:
674
672
case CEPH_OSD_OP_WRITE:
673
+
case CEPH_OSD_OP_WRITEFULL:
675
674
case CEPH_OSD_OP_ZERO:
676
675
case CEPH_OSD_OP_TRUNCATE:
677
-
if (src->op == CEPH_OSD_OP_WRITE)
676
+
if (src->op == CEPH_OSD_OP_WRITE ||
677
+
src->op == CEPH_OSD_OP_WRITEFULL)
678
678
request_data_len = src->extent.length;
679
679
dst->extent.offset = cpu_to_le64(src->extent.offset);
680
680
dst->extent.length = cpu_to_le64(src->extent.length);
···
685
681
dst->extent.truncate_seq =
686
682
cpu_to_le32(src->extent.truncate_seq);
687
683
osd_data = &src->extent.osd_data;
688
-
if (src->op == CEPH_OSD_OP_WRITE)
684
+
if (src->op == CEPH_OSD_OP_WRITE ||
685
+
src->op == CEPH_OSD_OP_WRITEFULL)
689
686
ceph_osdc_msg_data_add(req->r_request, osd_data);
690
687
else
691
688
ceph_osdc_msg_data_add(req->r_reply, osd_data);