ore: Support for short read/writes
Memory conditions and max_bio constraints might cause us to
not comply to the full length of the requested IO. Instead of
failing the complete IO we can issue a shorter read/write and
report how much was actually executed in the ios->length
member.
All users must check ios->length at IO_done or upon return of
ore_read/write and re-issue the reminder of the bytes. Because
other wise there is no error returned like before.
This is part of the effort to support the pnfs-obj layout driver.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
diff --git a/fs/exofs/ore.c b/fs/exofs/ore.c
index d0f8db5..8354fe0 100644
--- a/fs/exofs/ore.c
+++ b/fs/exofs/ore.c
@@ -377,8 +377,8 @@
unsigned pg = *cur_pg;
struct request_queue *q =
osd_request_queue(_ios_od(ios, per_dev->dev));
-
- per_dev->length += cur_len;
+ unsigned len = cur_len;
+ int ret;
if (per_dev->bio == NULL) {
unsigned pages_in_stripe = ios->layout->group_width *
@@ -390,7 +390,8 @@
if (unlikely(!per_dev->bio)) {
ORE_DBGMSG("Failed to allocate BIO size=%u\n",
bio_size);
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto out;
}
}
@@ -403,15 +404,24 @@
added_len = bio_add_pc_page(q, per_dev->bio, ios->pages[pg],
pglen, pgbase);
- if (unlikely(pglen != added_len))
- return -ENOMEM;
+ if (unlikely(pglen != added_len)) {
+ ret = -ENOMEM;
+ goto out;
+ }
pgbase = 0;
++pg;
}
BUG_ON(cur_len);
+ per_dev->length += len;
*cur_pg = pg;
- return 0;
+ ret = 0;
+out: /* we fail the complete unit on an error eg don't advance
+ * per_dev->length and cur_pg. This means that we might have a bigger
+ * bio than the CDB requested length (per_dev->length). That's fine
+ * only the oposite is fatal.
+ */
+ return ret;
}
static int _prepare_for_striping(struct ore_io_state *ios)
@@ -472,7 +482,13 @@
out:
ios->numdevs = devs_in_group;
ios->pages_consumed = cur_pg;
- return ret;
+ if (unlikely(ret)) {
+ if (length == ios->length)
+ return ret;
+ else
+ ios->length -= length;
+ }
+ return 0;
}
int ore_create(struct ore_io_state *ios)