2009-09-26 03:07:19 +04:00
|
|
|
/*
|
|
|
|
drbd_req.c
|
|
|
|
|
|
|
|
This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
|
|
|
|
|
|
|
|
Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
|
|
|
|
Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
|
|
|
|
Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
|
|
|
|
|
|
|
|
drbd is free software; you can redistribute it and/or modify
|
|
|
|
it under the terms of the GNU General Public License as published by
|
|
|
|
the Free Software Foundation; either version 2, or (at your option)
|
|
|
|
any later version.
|
|
|
|
|
|
|
|
drbd is distributed in the hope that it will be useful,
|
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
GNU General Public License for more details.
|
|
|
|
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
|
|
along with drbd; see the file COPYING. If not, write to
|
|
|
|
the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
|
|
|
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/drbd.h>
|
|
|
|
#include "drbd_int.h"
|
|
|
|
#include "drbd_req.h"
|
|
|
|
|
|
|
|
|
2011-12-03 14:18:56 +04:00
|
|
|
static bool drbd_may_do_local_read(struct drbd_conf *mdev, sector_t sector, int size);
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/* Update disk stats at start of I/O request */
|
|
|
|
static void _drbd_start_io_acct(struct drbd_conf *mdev, struct drbd_request *req, struct bio *bio)
|
|
|
|
{
|
|
|
|
const int rw = bio_data_dir(bio);
|
|
|
|
int cpu;
|
|
|
|
cpu = part_stat_lock();
|
|
|
|
part_stat_inc(cpu, &mdev->vdisk->part0, ios[rw]);
|
|
|
|
part_stat_add(cpu, &mdev->vdisk->part0, sectors[rw], bio_sectors(bio));
|
2011-11-07 13:54:28 +04:00
|
|
|
(void) cpu; /* The macro invocations above want the cpu argument, I do not like
|
|
|
|
the compiler warning about cpu only assigned but never used... */
|
2009-11-18 17:52:51 +03:00
|
|
|
part_inc_in_flight(&mdev->vdisk->part0, rw);
|
2009-09-26 03:07:19 +04:00
|
|
|
part_stat_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update disk stats when completing request upwards */
|
|
|
|
static void _drbd_end_io_acct(struct drbd_conf *mdev, struct drbd_request *req)
|
|
|
|
{
|
|
|
|
int rw = bio_data_dir(req->master_bio);
|
|
|
|
unsigned long duration = jiffies - req->start_time;
|
|
|
|
int cpu;
|
|
|
|
cpu = part_stat_lock();
|
|
|
|
part_stat_add(cpu, &mdev->vdisk->part0, ticks[rw], duration);
|
|
|
|
part_round_stats(cpu, &mdev->vdisk->part0);
|
2009-11-18 17:52:51 +03:00
|
|
|
part_dec_in_flight(&mdev->vdisk->part0, rw);
|
2009-09-26 03:07:19 +04:00
|
|
|
part_stat_unlock();
|
|
|
|
}
|
|
|
|
|
2011-01-26 20:45:11 +03:00
|
|
|
static struct drbd_request *drbd_req_new(struct drbd_conf *mdev,
|
|
|
|
struct bio *bio_src)
|
|
|
|
{
|
|
|
|
struct drbd_request *req;
|
|
|
|
|
|
|
|
req = mempool_alloc(drbd_request_mempool, GFP_NOIO);
|
|
|
|
if (!req)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
drbd_req_make_private_bio(req, bio_src);
|
|
|
|
req->rq_state = bio_data_dir(bio_src) == WRITE ? RQ_WRITE : 0;
|
2011-02-08 17:08:49 +03:00
|
|
|
req->w.mdev = mdev;
|
2011-01-26 20:45:11 +03:00
|
|
|
req->master_bio = bio_src;
|
|
|
|
req->epoch = 0;
|
2011-01-28 12:31:04 +03:00
|
|
|
|
2011-01-26 20:45:11 +03:00
|
|
|
drbd_clear_interval(&req->i);
|
|
|
|
req->i.sector = bio_src->bi_sector;
|
|
|
|
req->i.size = bio_src->bi_size;
|
2011-01-27 16:42:51 +03:00
|
|
|
req->i.local = true;
|
2011-01-28 12:31:04 +03:00
|
|
|
req->i.waiting = false;
|
|
|
|
|
2011-01-26 20:45:11 +03:00
|
|
|
INIT_LIST_HEAD(&req->tl_requests);
|
|
|
|
INIT_LIST_HEAD(&req->w.list);
|
|
|
|
|
|
|
|
return req;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void drbd_req_free(struct drbd_request *req)
|
|
|
|
{
|
|
|
|
mempool_free(req, drbd_request_mempool);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* rw is bio_data_dir(), only READ or WRITE */
|
2009-09-26 03:07:19 +04:00
|
|
|
static void _req_is_done(struct drbd_conf *mdev, struct drbd_request *req, const int rw)
|
|
|
|
{
|
|
|
|
const unsigned long s = req->rq_state;
|
2010-05-27 17:07:43 +04:00
|
|
|
|
|
|
|
/* remove it from the transfer log.
|
|
|
|
* well, only if it had been there in the first
|
|
|
|
* place... if it had not (local only or conflicting
|
|
|
|
* and never sent), it should still be "empty" as
|
|
|
|
* initialized in drbd_req_new(), so we can list_del() it
|
|
|
|
* here unconditionally */
|
|
|
|
list_del(&req->tl_requests);
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/* if it was a write, we may have to set the corresponding
|
|
|
|
* bit(s) out-of-sync first. If it had a local part, we need to
|
|
|
|
* release the reference to the activity log. */
|
|
|
|
if (rw == WRITE) {
|
|
|
|
/* Set out-of-sync unless both OK flags are set
|
|
|
|
* (local only or remote failed).
|
|
|
|
* Other places where we set out-of-sync:
|
|
|
|
* READ with local io-error */
|
|
|
|
if (!(s & RQ_NET_OK) || !(s & RQ_LOCAL_OK))
|
2011-01-03 19:09:58 +03:00
|
|
|
drbd_set_out_of_sync(mdev, req->i.sector, req->i.size);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
if ((s & RQ_NET_OK) && (s & RQ_LOCAL_OK) && (s & RQ_NET_SIS))
|
2011-01-03 19:09:58 +03:00
|
|
|
drbd_set_in_sync(mdev, req->i.sector, req->i.size);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* one might be tempted to move the drbd_al_complete_io
|
2011-02-17 18:46:59 +03:00
|
|
|
* to the local io completion callback drbd_request_endio.
|
2009-09-26 03:07:19 +04:00
|
|
|
* but, if this was a mirror write, we may only
|
|
|
|
* drbd_al_complete_io after this is RQ_NET_DONE,
|
|
|
|
* otherwise the extent could be dropped from the al
|
|
|
|
* before it has actually been written on the peer.
|
|
|
|
* if we crash before our peer knows about the request,
|
|
|
|
* but after the extent has been dropped from the al,
|
|
|
|
* we would forget to resync the corresponding extent.
|
|
|
|
*/
|
|
|
|
if (s & RQ_LOCAL_MASK) {
|
|
|
|
if (get_ldev_if_state(mdev, D_FAILED)) {
|
2010-08-31 14:00:50 +04:00
|
|
|
if (s & RQ_IN_ACT_LOG)
|
2011-03-31 17:18:56 +04:00
|
|
|
drbd_al_complete_io(mdev, &req->i);
|
2009-09-26 03:07:19 +04:00
|
|
|
put_ldev(mdev);
|
|
|
|
} else if (__ratelimit(&drbd_ratelimit_state)) {
|
2011-03-31 17:18:56 +04:00
|
|
|
dev_warn(DEV, "Should have called drbd_al_complete_io(, %llu, %u), "
|
|
|
|
"but my Disk seems to have failed :(\n",
|
|
|
|
(unsigned long long) req->i.sector, req->i.size);
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-05-26 19:13:18 +04:00
|
|
|
drbd_req_free(req);
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
static void queue_barrier(struct drbd_conf *mdev)
|
|
|
|
{
|
|
|
|
struct drbd_tl_epoch *b;
|
2011-11-10 21:45:36 +04:00
|
|
|
struct drbd_tconn *tconn = mdev->tconn;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* We are within the req_lock. Once we queued the barrier for sending,
|
|
|
|
* we set the CREATE_BARRIER bit. It is cleared as soon as a new
|
|
|
|
* barrier/epoch object is added. This is the only place this bit is
|
|
|
|
* set. It indicates that the barrier for this epoch is already queued,
|
|
|
|
* and no new epoch has been created yet. */
|
2011-11-10 21:45:36 +04:00
|
|
|
if (test_bit(CREATE_BARRIER, &tconn->flags))
|
2009-09-26 03:07:19 +04:00
|
|
|
return;
|
|
|
|
|
2011-11-10 21:45:36 +04:00
|
|
|
b = tconn->newest_tle;
|
2009-09-26 03:07:19 +04:00
|
|
|
b->w.cb = w_send_barrier;
|
2011-02-08 17:08:49 +03:00
|
|
|
b->w.mdev = mdev;
|
2009-09-26 03:07:19 +04:00
|
|
|
/* inc_ap_pending done here, so we won't
|
|
|
|
* get imbalanced on connection loss.
|
|
|
|
* dec_ap_pending will be done in got_BarrierAck
|
|
|
|
* or (on connection loss) in tl_clear. */
|
|
|
|
inc_ap_pending(mdev);
|
2011-11-10 21:45:36 +04:00
|
|
|
drbd_queue_work(&tconn->data.work, &b->w);
|
|
|
|
set_bit(CREATE_BARRIER, &tconn->flags);
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
static void _about_to_complete_local_write(struct drbd_conf *mdev,
|
|
|
|
struct drbd_request *req)
|
|
|
|
{
|
|
|
|
const unsigned long s = req->rq_state;
|
|
|
|
|
2010-12-05 16:11:14 +03:00
|
|
|
/* Before we can signal completion to the upper layers,
|
|
|
|
* we may need to close the current epoch.
|
|
|
|
* We can skip this, if this request has not even been sent, because we
|
|
|
|
* did not have a fully established connection yet/anymore, during
|
|
|
|
* bitmap exchange, or while we are C_AHEAD due to congestion policy.
|
|
|
|
*/
|
|
|
|
if (mdev->state.conn >= C_CONNECTED &&
|
|
|
|
(s & RQ_NET_SENT) != 0 &&
|
2011-01-19 16:16:30 +03:00
|
|
|
req->epoch == mdev->tconn->newest_tle->br_number)
|
2009-09-26 03:07:19 +04:00
|
|
|
queue_barrier(mdev);
|
|
|
|
}
|
|
|
|
|
|
|
|
void complete_master_bio(struct drbd_conf *mdev,
|
|
|
|
struct bio_and_error *m)
|
|
|
|
{
|
|
|
|
bio_endio(m->bio, m->error);
|
|
|
|
dec_ap_bio(mdev);
|
|
|
|
}
|
|
|
|
|
2011-01-28 12:31:04 +03:00
|
|
|
|
|
|
|
static void drbd_remove_request_interval(struct rb_root *root,
|
|
|
|
struct drbd_request *req)
|
|
|
|
{
|
2011-02-08 17:08:49 +03:00
|
|
|
struct drbd_conf *mdev = req->w.mdev;
|
2011-01-28 12:31:04 +03:00
|
|
|
struct drbd_interval *i = &req->i;
|
|
|
|
|
|
|
|
drbd_remove_interval(root, i);
|
|
|
|
|
|
|
|
/* Wake up any processes waiting for this request to complete. */
|
|
|
|
if (i->waiting)
|
|
|
|
wake_up(&mdev->misc_wait);
|
|
|
|
}
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/* Helper for __req_mod().
|
|
|
|
* Set m->bio to the master bio, if it is fit to be completed,
|
|
|
|
* or leave it alone (it is initialized to NULL in __req_mod),
|
|
|
|
* if it has already been completed, or cannot be completed yet.
|
|
|
|
* If m->bio is set, the error status to be returned is placed in m->error.
|
|
|
|
*/
|
|
|
|
void _req_may_be_done(struct drbd_request *req, struct bio_and_error *m)
|
|
|
|
{
|
|
|
|
const unsigned long s = req->rq_state;
|
2011-02-08 17:08:49 +03:00
|
|
|
struct drbd_conf *mdev = req->w.mdev;
|
2011-07-05 17:38:59 +04:00
|
|
|
int rw = req->rq_state & RQ_WRITE ? WRITE : READ;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* we must not complete the master bio, while it is
|
|
|
|
* still being processed by _drbd_send_zc_bio (drbd_send_dblock)
|
|
|
|
* not yet acknowledged by the peer
|
|
|
|
* not yet completed by the local io subsystem
|
|
|
|
* these flags may get cleared in any order by
|
|
|
|
* the worker,
|
|
|
|
* the receiver,
|
|
|
|
* the bio_endio completion callbacks.
|
|
|
|
*/
|
2011-07-05 17:38:59 +04:00
|
|
|
if (s & RQ_LOCAL_PENDING && !(s & RQ_LOCAL_ABORTED))
|
2011-02-22 04:15:32 +03:00
|
|
|
return;
|
|
|
|
if (req->i.waiting) {
|
|
|
|
/* Retry all conflicting peer requests. */
|
|
|
|
wake_up(&mdev->misc_wait);
|
|
|
|
}
|
2009-09-26 03:07:19 +04:00
|
|
|
if (s & RQ_NET_QUEUED)
|
|
|
|
return;
|
|
|
|
if (s & RQ_NET_PENDING)
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (req->master_bio) {
|
2011-01-25 17:37:43 +03:00
|
|
|
/* this is DATA_RECEIVED (remote read)
|
2009-09-26 03:07:19 +04:00
|
|
|
* or protocol C P_WRITE_ACK
|
|
|
|
* or protocol B P_RECV_ACK
|
2011-01-25 17:37:43 +03:00
|
|
|
* or protocol A "HANDED_OVER_TO_NETWORK" (SendAck)
|
2009-09-26 03:07:19 +04:00
|
|
|
* or canceled or failed,
|
|
|
|
* or killed from the transfer log due to connection loss.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* figure out whether to report success or failure.
|
|
|
|
*
|
|
|
|
* report success when at least one of the operations succeeded.
|
|
|
|
* or, to put the other way,
|
|
|
|
* only report failure, when both operations failed.
|
|
|
|
*
|
|
|
|
* what to do about the failures is handled elsewhere.
|
|
|
|
* what we need to do here is just: complete the master_bio.
|
|
|
|
*
|
|
|
|
* local completion error, if any, has been stored as ERR_PTR
|
2011-02-17 18:46:59 +03:00
|
|
|
* in private_bio within drbd_request_endio.
|
2009-09-26 03:07:19 +04:00
|
|
|
*/
|
|
|
|
int ok = (s & RQ_LOCAL_OK) || (s & RQ_NET_OK);
|
|
|
|
int error = PTR_ERR(req->private_bio);
|
|
|
|
|
|
|
|
/* remove the request from the conflict detection
|
|
|
|
* respective block_id verification hash */
|
2011-01-21 19:18:39 +03:00
|
|
|
if (!drbd_interval_empty(&req->i)) {
|
|
|
|
struct rb_root *root;
|
|
|
|
|
|
|
|
if (rw == WRITE)
|
|
|
|
root = &mdev->write_requests;
|
|
|
|
else
|
|
|
|
root = &mdev->read_requests;
|
2011-01-28 12:31:04 +03:00
|
|
|
drbd_remove_request_interval(root, req);
|
2011-02-22 04:15:32 +03:00
|
|
|
} else if (!(s & RQ_POSTPONED))
|
2010-10-21 19:21:19 +04:00
|
|
|
D_ASSERT((s & (RQ_NET_MASK & ~RQ_NET_DONE)) == 0);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* for writes we need to do some extra housekeeping */
|
|
|
|
if (rw == WRITE)
|
|
|
|
_about_to_complete_local_write(mdev, req);
|
|
|
|
|
|
|
|
/* Update disk stats */
|
|
|
|
_drbd_end_io_acct(mdev, req);
|
|
|
|
|
2011-02-22 04:15:32 +03:00
|
|
|
if (!(s & RQ_POSTPONED)) {
|
|
|
|
m->error = ok ? 0 : (error ?: -EIO);
|
|
|
|
m->bio = req->master_bio;
|
|
|
|
}
|
2009-09-26 03:07:19 +04:00
|
|
|
req->master_bio = NULL;
|
|
|
|
}
|
|
|
|
|
2011-07-05 17:38:59 +04:00
|
|
|
if (s & RQ_LOCAL_PENDING)
|
|
|
|
return;
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
if ((s & RQ_NET_MASK) == 0 || (s & RQ_NET_DONE)) {
|
|
|
|
/* this is disconnected (local only) operation,
|
2011-08-17 14:43:25 +04:00
|
|
|
* or protocol A, B, or C P_BARRIER_ACK,
|
2009-09-26 03:07:19 +04:00
|
|
|
* or killed from the transfer log due to connection loss. */
|
|
|
|
_req_is_done(mdev, req, rw);
|
|
|
|
}
|
|
|
|
/* else: network part and not DONE yet. that is
|
2011-08-17 14:43:25 +04:00
|
|
|
* protocol A, B, or C, barrier ack still pending... */
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
2010-06-23 19:18:51 +04:00
|
|
|
static void _req_may_be_done_not_susp(struct drbd_request *req, struct bio_and_error *m)
|
|
|
|
{
|
2011-02-08 17:08:49 +03:00
|
|
|
struct drbd_conf *mdev = req->w.mdev;
|
2010-06-23 19:18:51 +04:00
|
|
|
|
2011-03-28 18:48:11 +04:00
|
|
|
if (!drbd_suspended(mdev))
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done(req, m);
|
|
|
|
}
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/* obviously this could be coded as many single functions
|
|
|
|
* instead of one huge switch,
|
|
|
|
* or by putting the code directly in the respective locations
|
|
|
|
* (as it has been before).
|
|
|
|
*
|
|
|
|
* but having it this way
|
|
|
|
* enforces that it is all in this one place, where it is easier to audit,
|
|
|
|
* it makes it obvious that whatever "event" "happens" to a request should
|
|
|
|
* happen "atomically" within the req_lock,
|
|
|
|
* and it enforces that we have to think in a very structured manner
|
|
|
|
* about the "events" that may happen to a request during its life time ...
|
|
|
|
*/
|
2010-06-09 16:07:43 +04:00
|
|
|
int __req_mod(struct drbd_request *req, enum drbd_req_event what,
|
2009-09-26 03:07:19 +04:00
|
|
|
struct bio_and_error *m)
|
|
|
|
{
|
2011-02-08 17:08:49 +03:00
|
|
|
struct drbd_conf *mdev = req->w.mdev;
|
2011-04-19 19:10:19 +04:00
|
|
|
struct net_conf *nc;
|
2011-04-14 03:24:47 +04:00
|
|
|
int p, rv = 0;
|
2011-02-22 04:15:32 +03:00
|
|
|
|
|
|
|
if (m)
|
|
|
|
m->bio = NULL;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
switch (what) {
|
|
|
|
default:
|
|
|
|
dev_err(DEV, "LOGIC BUG in %s:%u\n", __FILE__ , __LINE__);
|
|
|
|
break;
|
|
|
|
|
|
|
|
/* does not happen...
|
|
|
|
* initialization done in drbd_req_new
|
2011-01-25 17:37:43 +03:00
|
|
|
case CREATED:
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
*/
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case TO_BE_SENT: /* via network */
|
2011-02-22 04:15:32 +03:00
|
|
|
/* reached via __drbd_make_request
|
2009-09-26 03:07:19 +04:00
|
|
|
* and from w_read_retry_remote */
|
|
|
|
D_ASSERT(!(req->rq_state & RQ_NET_MASK));
|
|
|
|
req->rq_state |= RQ_NET_PENDING;
|
2011-04-19 19:10:19 +04:00
|
|
|
rcu_read_lock();
|
|
|
|
nc = rcu_dereference(mdev->tconn->net_conf);
|
|
|
|
p = nc->wire_protocol;
|
|
|
|
rcu_read_unlock();
|
2011-04-14 03:24:47 +04:00
|
|
|
req->rq_state |=
|
|
|
|
p == DRBD_PROT_C ? RQ_EXP_WRITE_ACK :
|
|
|
|
p == DRBD_PROT_B ? RQ_EXP_RECEIVE_ACK : 0;
|
2009-09-26 03:07:19 +04:00
|
|
|
inc_ap_pending(mdev);
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case TO_BE_SUBMITTED: /* locally */
|
2011-02-22 04:15:32 +03:00
|
|
|
/* reached via __drbd_make_request */
|
2009-09-26 03:07:19 +04:00
|
|
|
D_ASSERT(!(req->rq_state & RQ_LOCAL_MASK));
|
|
|
|
req->rq_state |= RQ_LOCAL_PENDING;
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case COMPLETED_OK:
|
2011-07-05 17:38:59 +04:00
|
|
|
if (req->rq_state & RQ_WRITE)
|
2011-01-03 19:09:58 +03:00
|
|
|
mdev->writ_cnt += req->i.size >> 9;
|
2009-09-26 03:07:19 +04:00
|
|
|
else
|
2011-01-03 19:09:58 +03:00
|
|
|
mdev->read_cnt += req->i.size >> 9;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
req->rq_state |= (RQ_LOCAL_COMPLETED|RQ_LOCAL_OK);
|
|
|
|
req->rq_state &= ~RQ_LOCAL_PENDING;
|
|
|
|
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
put_ldev(mdev);
|
|
|
|
break;
|
|
|
|
|
2011-07-05 17:38:59 +04:00
|
|
|
case ABORT_DISK_IO:
|
|
|
|
req->rq_state |= RQ_LOCAL_ABORTED;
|
|
|
|
if (req->rq_state & RQ_WRITE)
|
|
|
|
_req_may_be_done_not_susp(req, m);
|
|
|
|
else
|
|
|
|
goto goto_queue_for_net_read;
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case WRITE_COMPLETED_WITH_ERROR:
|
2009-09-26 03:07:19 +04:00
|
|
|
req->rq_state |= RQ_LOCAL_COMPLETED;
|
|
|
|
req->rq_state &= ~RQ_LOCAL_PENDING;
|
|
|
|
|
2010-12-09 17:03:57 +03:00
|
|
|
__drbd_chk_io_error(mdev, false);
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
put_ldev(mdev);
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case READ_AHEAD_COMPLETED_WITH_ERROR:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* it is legal to fail READA */
|
|
|
|
req->rq_state |= RQ_LOCAL_COMPLETED;
|
|
|
|
req->rq_state &= ~RQ_LOCAL_PENDING;
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
put_ldev(mdev);
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case READ_COMPLETED_WITH_ERROR:
|
2011-01-03 19:09:58 +03:00
|
|
|
drbd_set_out_of_sync(mdev, req->i.sector, req->i.size);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
req->rq_state |= RQ_LOCAL_COMPLETED;
|
|
|
|
req->rq_state &= ~RQ_LOCAL_PENDING;
|
|
|
|
|
|
|
|
D_ASSERT(!(req->rq_state & RQ_NET_MASK));
|
|
|
|
|
2010-12-09 17:03:57 +03:00
|
|
|
__drbd_chk_io_error(mdev, false);
|
2009-09-26 03:07:19 +04:00
|
|
|
put_ldev(mdev);
|
|
|
|
|
2011-07-05 17:38:59 +04:00
|
|
|
goto_queue_for_net_read:
|
|
|
|
|
2010-05-27 11:45:45 +04:00
|
|
|
/* no point in retrying if there is no good remote data,
|
|
|
|
* or we have no connection. */
|
|
|
|
if (mdev->state.pdsk != D_UP_TO_DATE) {
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2010-05-27 11:45:45 +04:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
/* _req_mod(req,TO_BE_SENT); oops, recursion... */
|
2010-05-27 11:45:45 +04:00
|
|
|
req->rq_state |= RQ_NET_PENDING;
|
|
|
|
inc_ap_pending(mdev);
|
2011-01-25 17:37:43 +03:00
|
|
|
/* fall through: _req_mod(req,QUEUE_FOR_NET_READ); */
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case QUEUE_FOR_NET_READ:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* READ or READA, and
|
|
|
|
* no local disk,
|
|
|
|
* or target area marked as invalid,
|
|
|
|
* or just got an io-error. */
|
2011-02-22 04:15:32 +03:00
|
|
|
/* from __drbd_make_request
|
2009-09-26 03:07:19 +04:00
|
|
|
* or from bio_endio during read io-error recovery */
|
|
|
|
|
|
|
|
/* so we can verify the handle in the answer packet
|
|
|
|
* corresponding hlist_del is in _req_may_be_done() */
|
2011-07-16 01:52:44 +04:00
|
|
|
D_ASSERT(drbd_interval_empty(&req->i));
|
2011-01-21 19:18:39 +03:00
|
|
|
drbd_insert_interval(&mdev->read_requests, &req->i);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2009-11-03 04:22:06 +03:00
|
|
|
set_bit(UNPLUG_REMOTE, &mdev->flags);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
D_ASSERT(req->rq_state & RQ_NET_PENDING);
|
|
|
|
req->rq_state |= RQ_NET_QUEUED;
|
|
|
|
req->w.cb = (req->rq_state & RQ_LOCAL_MASK)
|
|
|
|
? w_read_retry_remote
|
|
|
|
: w_send_read_req;
|
2011-01-19 15:55:45 +03:00
|
|
|
drbd_queue_work(&mdev->tconn->data.work, &req->w);
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case QUEUE_FOR_NET_WRITE:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* assert something? */
|
2011-02-22 04:15:32 +03:00
|
|
|
/* from __drbd_make_request only */
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* corresponding hlist_del is in _req_may_be_done() */
|
2011-07-16 01:52:44 +04:00
|
|
|
D_ASSERT(drbd_interval_empty(&req->i));
|
2011-01-20 17:00:24 +03:00
|
|
|
drbd_insert_interval(&mdev->write_requests, &req->i);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* NOTE
|
|
|
|
* In case the req ended up on the transfer log before being
|
|
|
|
* queued on the worker, it could lead to this request being
|
|
|
|
* missed during cleanup after connection loss.
|
|
|
|
* So we have to do both operations here,
|
|
|
|
* within the same lock that protects the transfer log.
|
|
|
|
*
|
|
|
|
* _req_add_to_epoch(req); this has to be after the
|
|
|
|
* _maybe_start_new_epoch(req); which happened in
|
2011-02-22 04:15:32 +03:00
|
|
|
* __drbd_make_request, because we now may set the bit
|
2009-09-26 03:07:19 +04:00
|
|
|
* again ourselves to close the current epoch.
|
|
|
|
*
|
|
|
|
* Add req to the (now) current epoch (barrier). */
|
|
|
|
|
2009-11-03 04:22:06 +03:00
|
|
|
/* otherwise we may lose an unplug, which may cause some remote
|
|
|
|
* io-scheduler timeout to expire, increasing maximum latency,
|
|
|
|
* hurting performance. */
|
|
|
|
set_bit(UNPLUG_REMOTE, &mdev->flags);
|
|
|
|
|
2011-02-22 04:15:32 +03:00
|
|
|
/* see __drbd_make_request,
|
2009-09-26 03:07:19 +04:00
|
|
|
* just after it grabs the req_lock */
|
2011-11-10 21:45:36 +04:00
|
|
|
D_ASSERT(test_bit(CREATE_BARRIER, &mdev->tconn->flags) == 0);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-01-19 16:16:30 +03:00
|
|
|
req->epoch = mdev->tconn->newest_tle->br_number;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* increment size of current epoch */
|
2011-01-19 16:16:30 +03:00
|
|
|
mdev->tconn->newest_tle->n_writes++;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* queue work item to send data */
|
|
|
|
D_ASSERT(req->rq_state & RQ_NET_PENDING);
|
|
|
|
req->rq_state |= RQ_NET_QUEUED;
|
|
|
|
req->w.cb = w_send_dblock;
|
2011-01-19 15:55:45 +03:00
|
|
|
drbd_queue_work(&mdev->tconn->data.work, &req->w);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* close the epoch, in case it outgrew the limit */
|
2011-04-19 19:10:19 +04:00
|
|
|
rcu_read_lock();
|
|
|
|
nc = rcu_dereference(mdev->tconn->net_conf);
|
|
|
|
p = nc->max_epoch_size;
|
|
|
|
rcu_read_unlock();
|
|
|
|
if (mdev->tconn->newest_tle->n_writes >= p)
|
2009-09-26 03:07:19 +04:00
|
|
|
queue_barrier(mdev);
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case QUEUE_FOR_SEND_OOS:
|
2010-10-27 16:33:00 +04:00
|
|
|
req->rq_state |= RQ_NET_QUEUED;
|
2010-12-20 01:53:14 +03:00
|
|
|
req->w.cb = w_send_out_of_sync;
|
2011-01-19 15:55:45 +03:00
|
|
|
drbd_queue_work(&mdev->tconn->data.work, &req->w);
|
2010-10-27 16:33:00 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case OOS_HANDED_TO_NETWORK:
|
2010-10-27 16:33:00 +04:00
|
|
|
/* actually the same */
|
2011-01-25 17:37:43 +03:00
|
|
|
case SEND_CANCELED:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* treat it the same */
|
2011-01-25 17:37:43 +03:00
|
|
|
case SEND_FAILED:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* real cleanup will be done from tl_clear. just update flags
|
|
|
|
* so it is no longer marked as on the worker queue */
|
|
|
|
req->rq_state &= ~RQ_NET_QUEUED;
|
|
|
|
/* if we did it right, tl_clear should be scheduled only after
|
|
|
|
* this, so this should not be necessary! */
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case HANDED_OVER_TO_NETWORK:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* assert something? */
|
2010-10-26 18:02:27 +04:00
|
|
|
if (bio_data_dir(req->master_bio) == WRITE)
|
2011-01-03 19:09:58 +03:00
|
|
|
atomic_add(req->i.size >> 9, &mdev->ap_in_flight);
|
2010-10-26 18:02:27 +04:00
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
if (bio_data_dir(req->master_bio) == WRITE &&
|
2011-04-14 03:24:47 +04:00
|
|
|
!(req->rq_state & (RQ_EXP_RECEIVE_ACK | RQ_EXP_WRITE_ACK))) {
|
2009-09-26 03:07:19 +04:00
|
|
|
/* this is what is dangerous about protocol A:
|
|
|
|
* pretend it was successfully written on the peer. */
|
|
|
|
if (req->rq_state & RQ_NET_PENDING) {
|
|
|
|
dec_ap_pending(mdev);
|
|
|
|
req->rq_state &= ~RQ_NET_PENDING;
|
|
|
|
req->rq_state |= RQ_NET_OK;
|
|
|
|
} /* else: neg-ack was faster... */
|
|
|
|
/* it is still not yet RQ_NET_DONE until the
|
|
|
|
* corresponding epoch barrier got acked as well,
|
|
|
|
* so we know what to dirty on connection loss */
|
|
|
|
}
|
|
|
|
req->rq_state &= ~RQ_NET_QUEUED;
|
|
|
|
req->rq_state |= RQ_NET_SENT;
|
|
|
|
/* because _drbd_send_zc_bio could sleep, and may want to
|
2011-01-25 17:37:43 +03:00
|
|
|
* dereference the bio even after the "WRITE_ACKED_BY_PEER" and
|
|
|
|
* "COMPLETED_OK" events came in, once we return from
|
2009-09-26 03:07:19 +04:00
|
|
|
* _drbd_send_zc_bio (drbd_send_dblock), we have to check
|
|
|
|
* whether it is done already, and end it. */
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case READ_RETRY_REMOTE_CANCELED:
|
2010-05-27 11:45:45 +04:00
|
|
|
req->rq_state &= ~RQ_NET_QUEUED;
|
|
|
|
/* fall through, in case we raced with drbd_disconnect */
|
2011-01-25 17:37:43 +03:00
|
|
|
case CONNECTION_LOST_WHILE_PENDING:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* transfer log cleanup after connection loss */
|
|
|
|
/* assert something? */
|
|
|
|
if (req->rq_state & RQ_NET_PENDING)
|
|
|
|
dec_ap_pending(mdev);
|
2011-12-03 14:18:56 +04:00
|
|
|
|
|
|
|
p = !(req->rq_state & RQ_WRITE) && req->rq_state & RQ_NET_PENDING;
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
req->rq_state &= ~(RQ_NET_OK|RQ_NET_PENDING);
|
|
|
|
req->rq_state |= RQ_NET_DONE;
|
2010-10-26 18:02:27 +04:00
|
|
|
if (req->rq_state & RQ_NET_SENT && req->rq_state & RQ_WRITE)
|
2011-01-03 19:09:58 +03:00
|
|
|
atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
|
2010-10-26 18:02:27 +04:00
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/* if it is still queued, we may not complete it here.
|
|
|
|
* it will be canceled soon. */
|
2011-12-03 14:18:56 +04:00
|
|
|
if (!(req->rq_state & RQ_NET_QUEUED)) {
|
|
|
|
if (p)
|
|
|
|
goto goto_read_retry_local;
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done(req, m); /* Allowed while state.susp */
|
2011-12-03 14:18:56 +04:00
|
|
|
}
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case WRITE_ACKED_BY_PEER_AND_SIS:
|
2009-09-26 03:07:19 +04:00
|
|
|
req->rq_state |= RQ_NET_SIS;
|
2011-02-22 04:15:32 +03:00
|
|
|
case DISCARD_WRITE:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* for discarded conflicting writes of multiple primaries,
|
|
|
|
* there is no need to keep anything in the tl, potential
|
|
|
|
* node crashes are covered by the activity log. */
|
|
|
|
req->rq_state |= RQ_NET_DONE;
|
|
|
|
/* fall through */
|
2011-01-25 17:37:43 +03:00
|
|
|
case WRITE_ACKED_BY_PEER:
|
2011-04-14 03:24:47 +04:00
|
|
|
D_ASSERT(req->rq_state & RQ_EXP_WRITE_ACK);
|
2009-09-26 03:07:19 +04:00
|
|
|
/* protocol C; successfully written on peer.
|
|
|
|
* Nothing to do here.
|
|
|
|
* We want to keep the tl in place for all protocols, to cater
|
|
|
|
* for volatile write-back caches on lower level devices.
|
|
|
|
*
|
|
|
|
* A barrier request is expected to have forced all prior
|
|
|
|
* requests onto stable storage, so completion of a barrier
|
|
|
|
* request could set NET_DONE right here, and not wait for the
|
|
|
|
* P_BARRIER_ACK, but that is an unnecessary optimization. */
|
|
|
|
|
2011-04-14 03:24:47 +04:00
|
|
|
goto ack_common;
|
2009-09-26 03:07:19 +04:00
|
|
|
/* this makes it effectively the same as for: */
|
2011-01-25 17:37:43 +03:00
|
|
|
case RECV_ACKED_BY_PEER:
|
2011-04-14 03:24:47 +04:00
|
|
|
D_ASSERT(req->rq_state & RQ_EXP_RECEIVE_ACK);
|
2009-09-26 03:07:19 +04:00
|
|
|
/* protocol B; pretends to be successfully written on peer.
|
2011-01-25 17:37:43 +03:00
|
|
|
* see also notes above in HANDED_OVER_TO_NETWORK about
|
2009-09-26 03:07:19 +04:00
|
|
|
* protocol != C */
|
2011-04-14 03:24:47 +04:00
|
|
|
ack_common:
|
2009-09-26 03:07:19 +04:00
|
|
|
req->rq_state |= RQ_NET_OK;
|
|
|
|
D_ASSERT(req->rq_state & RQ_NET_PENDING);
|
|
|
|
dec_ap_pending(mdev);
|
2011-01-03 19:09:58 +03:00
|
|
|
atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
|
2009-09-26 03:07:19 +04:00
|
|
|
req->rq_state &= ~RQ_NET_PENDING;
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-02-22 04:15:32 +03:00
|
|
|
case POSTPONE_WRITE:
|
2011-04-14 03:24:47 +04:00
|
|
|
D_ASSERT(req->rq_state & RQ_EXP_WRITE_ACK);
|
|
|
|
/* If this node has already detected the write conflict, the
|
2011-02-22 04:15:32 +03:00
|
|
|
* worker will be waiting on misc_wait. Wake it up once this
|
|
|
|
* request has completed locally.
|
|
|
|
*/
|
|
|
|
D_ASSERT(req->rq_state & RQ_NET_PENDING);
|
|
|
|
req->rq_state |= RQ_POSTPONED;
|
|
|
|
_req_may_be_done_not_susp(req, m);
|
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case NEG_ACKED:
|
2009-09-26 03:07:19 +04:00
|
|
|
/* assert something? */
|
2010-10-26 18:02:27 +04:00
|
|
|
if (req->rq_state & RQ_NET_PENDING) {
|
2009-09-26 03:07:19 +04:00
|
|
|
dec_ap_pending(mdev);
|
2011-01-03 19:09:58 +03:00
|
|
|
atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
|
2010-10-26 18:02:27 +04:00
|
|
|
}
|
2009-09-26 03:07:19 +04:00
|
|
|
req->rq_state &= ~(RQ_NET_OK|RQ_NET_PENDING);
|
|
|
|
|
|
|
|
req->rq_state |= RQ_NET_DONE;
|
2011-11-11 15:31:20 +04:00
|
|
|
|
2011-12-03 14:18:56 +04:00
|
|
|
if (!(req->rq_state & RQ_WRITE))
|
2011-11-11 15:31:20 +04:00
|
|
|
goto goto_read_retry_local;
|
|
|
|
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2011-01-25 17:37:43 +03:00
|
|
|
/* else: done by HANDED_OVER_TO_NETWORK */
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-11-11 15:31:20 +04:00
|
|
|
goto_read_retry_local:
|
2011-12-03 14:18:56 +04:00
|
|
|
if (!drbd_may_do_local_read(mdev, req->i.sector, req->i.size)) {
|
|
|
|
_req_may_be_done_not_susp(req, m);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
D_ASSERT(!(req->rq_state & RQ_LOCAL_PENDING));
|
2011-11-11 15:31:20 +04:00
|
|
|
req->rq_state |= RQ_LOCAL_PENDING;
|
2011-12-03 14:18:56 +04:00
|
|
|
|
|
|
|
get_ldev(mdev);
|
|
|
|
req->w.cb = w_restart_disk_io;
|
|
|
|
drbd_queue_work(&mdev->tconn->data.work, &req->w);
|
2011-11-11 15:31:20 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case FAIL_FROZEN_DISK_IO:
|
2010-05-31 12:14:17 +04:00
|
|
|
if (!(req->rq_state & RQ_LOCAL_COMPLETED))
|
|
|
|
break;
|
|
|
|
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done(req, m); /* Allowed while state.susp */
|
2010-05-31 12:14:17 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case RESTART_FROZEN_DISK_IO:
|
2010-05-31 12:14:17 +04:00
|
|
|
if (!(req->rq_state & RQ_LOCAL_COMPLETED))
|
|
|
|
break;
|
|
|
|
|
|
|
|
req->rq_state &= ~RQ_LOCAL_COMPLETED;
|
|
|
|
|
|
|
|
rv = MR_READ;
|
|
|
|
if (bio_data_dir(req->master_bio) == WRITE)
|
|
|
|
rv = MR_WRITE;
|
|
|
|
|
|
|
|
get_ldev(mdev);
|
|
|
|
req->w.cb = w_restart_disk_io;
|
2011-01-19 15:55:45 +03:00
|
|
|
drbd_queue_work(&mdev->tconn->data.work, &req->w);
|
2010-05-31 12:14:17 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case RESEND:
|
2010-05-12 19:08:26 +04:00
|
|
|
/* If RQ_NET_OK is already set, we got a P_WRITE_ACK or P_RECV_ACK
|
2010-06-18 15:56:57 +04:00
|
|
|
before the connection loss (B&C only); only P_BARRIER_ACK was missing.
|
2010-05-12 19:08:26 +04:00
|
|
|
Trowing them out of the TL here by pretending we got a BARRIER_ACK
|
2010-06-22 16:03:27 +04:00
|
|
|
We ensure that the peer was not rebooted */
|
2010-05-12 19:08:26 +04:00
|
|
|
if (!(req->rq_state & RQ_NET_OK)) {
|
|
|
|
if (req->w.cb) {
|
2011-01-19 15:55:45 +03:00
|
|
|
drbd_queue_work(&mdev->tconn->data.work, &req->w);
|
2010-05-12 19:08:26 +04:00
|
|
|
rv = req->rq_state & RQ_WRITE ? MR_WRITE : MR_READ;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
2011-01-25 17:37:43 +03:00
|
|
|
/* else, fall through to BARRIER_ACKED */
|
2010-05-12 19:08:26 +04:00
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case BARRIER_ACKED:
|
2010-05-27 17:07:43 +04:00
|
|
|
if (!(req->rq_state & RQ_WRITE))
|
|
|
|
break;
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
if (req->rq_state & RQ_NET_PENDING) {
|
2011-08-17 14:43:25 +04:00
|
|
|
/* barrier came in before all requests were acked.
|
2009-09-26 03:07:19 +04:00
|
|
|
* this is bad, because if the connection is lost now,
|
|
|
|
* we won't be able to clean them up... */
|
2011-01-25 17:37:43 +03:00
|
|
|
dev_err(DEV, "FIXME (BARRIER_ACKED but pending)\n");
|
2011-01-19 16:16:30 +03:00
|
|
|
list_move(&req->tl_requests, &mdev->tconn->out_of_sequence_requests);
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
2011-01-21 19:10:37 +03:00
|
|
|
if ((req->rq_state & RQ_NET_MASK) != 0) {
|
|
|
|
req->rq_state |= RQ_NET_DONE;
|
2011-04-14 03:24:47 +04:00
|
|
|
if (!(req->rq_state & (RQ_EXP_RECEIVE_ACK | RQ_EXP_WRITE_ACK)))
|
2011-01-19 15:12:45 +03:00
|
|
|
atomic_sub(req->i.size>>9, &mdev->ap_in_flight);
|
2011-01-21 19:10:37 +03:00
|
|
|
}
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done(req, m); /* Allowed while state.susp */
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
|
2011-01-25 17:37:43 +03:00
|
|
|
case DATA_RECEIVED:
|
2009-09-26 03:07:19 +04:00
|
|
|
D_ASSERT(req->rq_state & RQ_NET_PENDING);
|
|
|
|
dec_ap_pending(mdev);
|
|
|
|
req->rq_state &= ~RQ_NET_PENDING;
|
|
|
|
req->rq_state |= (RQ_NET_OK|RQ_NET_DONE);
|
2010-06-23 19:18:51 +04:00
|
|
|
_req_may_be_done_not_susp(req, m);
|
2009-09-26 03:07:19 +04:00
|
|
|
break;
|
|
|
|
};
|
2010-06-09 16:07:43 +04:00
|
|
|
|
|
|
|
return rv;
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* we may do a local read if:
|
|
|
|
* - we are consistent (of course),
|
|
|
|
* - or we are generally inconsistent,
|
|
|
|
* BUT we are still/already IN SYNC for this area.
|
|
|
|
* since size may be bigger than BM_BLOCK_SIZE,
|
|
|
|
* we may need to check several bits.
|
|
|
|
*/
|
2010-12-19 22:48:29 +03:00
|
|
|
static bool drbd_may_do_local_read(struct drbd_conf *mdev, sector_t sector, int size)
|
2009-09-26 03:07:19 +04:00
|
|
|
{
|
|
|
|
unsigned long sbnr, ebnr;
|
|
|
|
sector_t esector, nr_sectors;
|
|
|
|
|
|
|
|
if (mdev->state.disk == D_UP_TO_DATE)
|
2010-12-19 22:48:29 +03:00
|
|
|
return true;
|
2011-02-18 16:13:07 +03:00
|
|
|
if (mdev->state.disk != D_INCONSISTENT)
|
2010-12-19 22:48:29 +03:00
|
|
|
return false;
|
2009-09-26 03:07:19 +04:00
|
|
|
esector = sector + (size >> 9) - 1;
|
2011-02-21 14:34:58 +03:00
|
|
|
nr_sectors = drbd_get_capacity(mdev->this_bdev);
|
2009-09-26 03:07:19 +04:00
|
|
|
D_ASSERT(sector < nr_sectors);
|
|
|
|
D_ASSERT(esector < nr_sectors);
|
|
|
|
|
|
|
|
sbnr = BM_SECT_TO_BIT(sector);
|
|
|
|
ebnr = BM_SECT_TO_BIT(esector);
|
|
|
|
|
2010-12-19 22:48:29 +03:00
|
|
|
return drbd_bm_count_bits(mdev, sbnr, ebnr) == 0;
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
2011-11-17 13:12:31 +04:00
|
|
|
static bool remote_due_to_read_balancing(struct drbd_conf *mdev, sector_t sector)
|
2011-11-11 15:31:20 +04:00
|
|
|
{
|
|
|
|
enum drbd_read_balancing rbm;
|
|
|
|
struct backing_dev_info *bdi;
|
2011-11-17 13:12:31 +04:00
|
|
|
int stripe_shift;
|
2011-11-11 15:31:20 +04:00
|
|
|
|
|
|
|
if (mdev->state.pdsk < D_UP_TO_DATE)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
rbm = rcu_dereference(mdev->ldev->disk_conf)->read_balancing;
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
switch (rbm) {
|
|
|
|
case RB_CONGESTED_REMOTE:
|
|
|
|
bdi = &mdev->ldev->backing_bdev->bd_disk->queue->backing_dev_info;
|
|
|
|
return bdi_read_congested(bdi);
|
|
|
|
case RB_LEAST_PENDING:
|
|
|
|
return atomic_read(&mdev->local_cnt) >
|
|
|
|
atomic_read(&mdev->ap_pending_cnt) + atomic_read(&mdev->rs_pending_cnt);
|
2011-11-17 13:12:31 +04:00
|
|
|
case RB_32K_STRIPING: /* stripe_shift = 15 */
|
|
|
|
case RB_64K_STRIPING:
|
|
|
|
case RB_128K_STRIPING:
|
|
|
|
case RB_256K_STRIPING:
|
|
|
|
case RB_512K_STRIPING:
|
|
|
|
case RB_1M_STRIPING: /* stripe_shift = 20 */
|
|
|
|
stripe_shift = (rbm - RB_32K_STRIPING + 15);
|
|
|
|
return (sector >> (stripe_shift - 9)) & 1;
|
2011-11-11 15:31:20 +04:00
|
|
|
case RB_ROUND_ROBIN:
|
|
|
|
return test_and_change_bit(READ_BALANCE_RR, &mdev->flags);
|
|
|
|
case RB_PREFER_REMOTE:
|
|
|
|
return true;
|
|
|
|
case RB_PREFER_LOCAL:
|
|
|
|
default:
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-01-28 17:53:51 +03:00
|
|
|
/*
|
|
|
|
* complete_conflicting_writes - wait for any conflicting write requests
|
|
|
|
*
|
|
|
|
* The write_requests tree contains all active write requests which we
|
|
|
|
* currently know about. Wait for any requests to complete which conflict with
|
|
|
|
* the new one.
|
|
|
|
*/
|
|
|
|
static int complete_conflicting_writes(struct drbd_conf *mdev,
|
|
|
|
sector_t sector, int size)
|
|
|
|
{
|
|
|
|
for(;;) {
|
|
|
|
struct drbd_interval *i;
|
2011-02-22 04:15:32 +03:00
|
|
|
int err;
|
2011-01-28 17:53:51 +03:00
|
|
|
|
|
|
|
i = drbd_find_overlap(&mdev->write_requests, sector, size);
|
|
|
|
if (!i)
|
|
|
|
return 0;
|
2011-02-22 04:15:32 +03:00
|
|
|
err = drbd_wait_misc(mdev, i);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2011-01-28 17:53:51 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-02-22 04:15:32 +03:00
|
|
|
int __drbd_make_request(struct drbd_conf *mdev, struct bio *bio, unsigned long start_time)
|
2009-09-26 03:07:19 +04:00
|
|
|
{
|
|
|
|
const int rw = bio_rw(bio);
|
|
|
|
const int size = bio->bi_size;
|
|
|
|
const sector_t sector = bio->bi_sector;
|
|
|
|
struct drbd_tl_epoch *b = NULL;
|
|
|
|
struct drbd_request *req;
|
2011-04-19 19:10:19 +04:00
|
|
|
struct net_conf *nc;
|
2010-10-27 16:33:00 +04:00
|
|
|
int local, remote, send_oos = 0;
|
2011-01-28 17:53:51 +03:00
|
|
|
int err;
|
2010-05-10 18:42:23 +04:00
|
|
|
int ret = 0;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
/* allocate outside of all locks; */
|
|
|
|
req = drbd_req_new(mdev, bio);
|
|
|
|
if (!req) {
|
|
|
|
dec_ap_bio(mdev);
|
|
|
|
/* only pass the error to the upper layers.
|
|
|
|
* if user cannot handle io errors, that's not our business. */
|
|
|
|
dev_err(DEV, "could not kmalloc() req\n");
|
|
|
|
bio_endio(bio, -ENOMEM);
|
|
|
|
return 0;
|
|
|
|
}
|
2010-11-09 19:45:06 +03:00
|
|
|
req->start_time = start_time;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
|
|
|
local = get_ldev(mdev);
|
|
|
|
if (!local) {
|
|
|
|
bio_put(req->private_bio); /* or we get a bio leak */
|
|
|
|
req->private_bio = NULL;
|
|
|
|
}
|
|
|
|
if (rw == WRITE) {
|
|
|
|
remote = 1;
|
|
|
|
} else {
|
|
|
|
/* READ || READA */
|
|
|
|
if (local) {
|
2011-12-03 14:18:56 +04:00
|
|
|
if (!drbd_may_do_local_read(mdev, sector, size) ||
|
|
|
|
remote_due_to_read_balancing(mdev, sector)) {
|
2009-09-26 03:07:19 +04:00
|
|
|
/* we could kick the syncer to
|
|
|
|
* sync this extent asap, wait for
|
|
|
|
* it, then continue locally.
|
|
|
|
* Or just issue the request remotely.
|
|
|
|
*/
|
|
|
|
local = 0;
|
|
|
|
bio_put(req->private_bio);
|
|
|
|
req->private_bio = NULL;
|
|
|
|
put_ldev(mdev);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
remote = !local && mdev->state.pdsk >= D_UP_TO_DATE;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If we have a disk, but a READA request is mapped to remote,
|
|
|
|
* we are R_PRIMARY, D_INCONSISTENT, SyncTarget.
|
|
|
|
* Just fail that READA request right here.
|
|
|
|
*
|
|
|
|
* THINK: maybe fail all READA when not local?
|
|
|
|
* or make this configurable...
|
|
|
|
* if network is slow, READA won't do any good.
|
|
|
|
*/
|
|
|
|
if (rw == READA && mdev->state.disk >= D_INCONSISTENT && !local) {
|
|
|
|
err = -EWOULDBLOCK;
|
|
|
|
goto fail_and_free_req;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* For WRITES going to the local disk, grab a reference on the target
|
|
|
|
* extent. This waits for any resync activity in the corresponding
|
|
|
|
* resync extent to finish, and, if necessary, pulls in the target
|
|
|
|
* extent into the activity log, which involves further disk io because
|
|
|
|
* of transactional on-disk meta data updates. */
|
2010-08-31 14:00:50 +04:00
|
|
|
if (rw == WRITE && local && !test_bit(AL_SUSPENDED, &mdev->flags)) {
|
|
|
|
req->rq_state |= RQ_IN_ACT_LOG;
|
2011-03-31 17:18:56 +04:00
|
|
|
drbd_al_begin_io(mdev, &req->i);
|
2010-08-31 14:00:50 +04:00
|
|
|
}
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-01-17 22:27:30 +03:00
|
|
|
remote = remote && drbd_should_do_remote(mdev->state);
|
2010-12-20 01:53:14 +03:00
|
|
|
send_oos = rw == WRITE && drbd_should_send_out_of_sync(mdev->state);
|
2010-11-10 14:08:37 +03:00
|
|
|
D_ASSERT(!(remote && send_oos));
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-03-28 18:48:11 +04:00
|
|
|
if (!(local || remote) && !drbd_suspended(mdev)) {
|
2010-10-19 14:08:13 +04:00
|
|
|
if (__ratelimit(&drbd_ratelimit_state))
|
|
|
|
dev_err(DEV, "IO ERROR: neither local nor remote disk\n");
|
2011-01-28 17:53:51 +03:00
|
|
|
err = -EIO;
|
2009-09-26 03:07:19 +04:00
|
|
|
goto fail_free_complete;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* For WRITE request, we have to make sure that we have an
|
|
|
|
* unused_spare_tle, in case we need to start a new epoch.
|
|
|
|
* I try to be smart and avoid to pre-allocate always "just in case",
|
|
|
|
* but there is a race between testing the bit and pointer outside the
|
|
|
|
* spinlock, and grabbing the spinlock.
|
|
|
|
* if we lost that race, we retry. */
|
2010-10-27 16:33:00 +04:00
|
|
|
if (rw == WRITE && (remote || send_oos) &&
|
2011-01-19 16:16:30 +03:00
|
|
|
mdev->tconn->unused_spare_tle == NULL &&
|
2011-11-10 21:45:36 +04:00
|
|
|
test_bit(CREATE_BARRIER, &mdev->tconn->flags)) {
|
2009-09-26 03:07:19 +04:00
|
|
|
allocate_barrier:
|
|
|
|
b = kmalloc(sizeof(struct drbd_tl_epoch), GFP_NOIO);
|
|
|
|
if (!b) {
|
|
|
|
dev_err(DEV, "Failed to alloc barrier.\n");
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto fail_free_complete;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* GOOD, everything prepared, grab the spin_lock */
|
2011-01-19 16:16:30 +03:00
|
|
|
spin_lock_irq(&mdev->tconn->req_lock);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-01-28 17:53:51 +03:00
|
|
|
if (rw == WRITE) {
|
|
|
|
err = complete_conflicting_writes(mdev, sector, size);
|
|
|
|
if (err) {
|
2011-02-22 04:15:32 +03:00
|
|
|
if (err != -ERESTARTSYS)
|
|
|
|
_conn_request_state(mdev->tconn,
|
|
|
|
NS(conn, C_TIMEOUT),
|
|
|
|
CS_HARD);
|
2011-01-28 17:53:51 +03:00
|
|
|
spin_unlock_irq(&mdev->tconn->req_lock);
|
2011-02-22 04:15:32 +03:00
|
|
|
err = -EIO;
|
2011-01-28 17:53:51 +03:00
|
|
|
goto fail_free_complete;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-03-28 18:48:11 +04:00
|
|
|
if (drbd_suspended(mdev)) {
|
2010-05-10 18:42:23 +04:00
|
|
|
/* If we got suspended, use the retry mechanism of
|
|
|
|
generic_make_request() to restart processing of this
|
2010-12-13 19:48:19 +03:00
|
|
|
bio. In the next call to drbd_make_request
|
2010-05-10 18:42:23 +04:00
|
|
|
we sleep in inc_ap_bio() */
|
|
|
|
ret = 1;
|
2011-01-19 16:16:30 +03:00
|
|
|
spin_unlock_irq(&mdev->tconn->req_lock);
|
2010-05-10 18:42:23 +04:00
|
|
|
goto fail_free_complete;
|
|
|
|
}
|
|
|
|
|
2010-10-27 16:33:00 +04:00
|
|
|
if (remote || send_oos) {
|
2011-01-17 22:27:30 +03:00
|
|
|
remote = drbd_should_do_remote(mdev->state);
|
2010-12-20 01:53:14 +03:00
|
|
|
send_oos = rw == WRITE && drbd_should_send_out_of_sync(mdev->state);
|
2010-11-10 14:08:37 +03:00
|
|
|
D_ASSERT(!(remote && send_oos));
|
2010-10-27 16:33:00 +04:00
|
|
|
|
|
|
|
if (!(remote || send_oos))
|
2009-09-26 03:07:19 +04:00
|
|
|
dev_warn(DEV, "lost connection while grabbing the req_lock!\n");
|
|
|
|
if (!(local || remote)) {
|
|
|
|
dev_err(DEV, "IO ERROR: neither local nor remote disk\n");
|
2011-01-19 16:16:30 +03:00
|
|
|
spin_unlock_irq(&mdev->tconn->req_lock);
|
2011-01-28 17:53:51 +03:00
|
|
|
err = -EIO;
|
2009-09-26 03:07:19 +04:00
|
|
|
goto fail_free_complete;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-01-19 16:16:30 +03:00
|
|
|
if (b && mdev->tconn->unused_spare_tle == NULL) {
|
|
|
|
mdev->tconn->unused_spare_tle = b;
|
2009-09-26 03:07:19 +04:00
|
|
|
b = NULL;
|
|
|
|
}
|
2010-10-27 16:33:00 +04:00
|
|
|
if (rw == WRITE && (remote || send_oos) &&
|
2011-01-19 16:16:30 +03:00
|
|
|
mdev->tconn->unused_spare_tle == NULL &&
|
2011-11-10 21:45:36 +04:00
|
|
|
test_bit(CREATE_BARRIER, &mdev->tconn->flags)) {
|
2009-09-26 03:07:19 +04:00
|
|
|
/* someone closed the current epoch
|
|
|
|
* while we were grabbing the spinlock */
|
2011-01-19 16:16:30 +03:00
|
|
|
spin_unlock_irq(&mdev->tconn->req_lock);
|
2009-09-26 03:07:19 +04:00
|
|
|
goto allocate_barrier;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* Update disk stats */
|
|
|
|
_drbd_start_io_acct(mdev, req, bio);
|
|
|
|
|
|
|
|
/* _maybe_start_new_epoch(mdev);
|
|
|
|
* If we need to generate a write barrier packet, we have to add the
|
|
|
|
* new epoch (barrier) object, and queue the barrier packet for sending,
|
|
|
|
* and queue the req's data after it _within the same lock_, otherwise
|
|
|
|
* we have race conditions were the reorder domains could be mixed up.
|
|
|
|
*
|
|
|
|
* Even read requests may start a new epoch and queue the corresponding
|
|
|
|
* barrier packet. To get the write ordering right, we only have to
|
|
|
|
* make sure that, if this is a write request and it triggered a
|
|
|
|
* barrier packet, this request is queued within the same spinlock. */
|
2011-01-19 16:16:30 +03:00
|
|
|
if ((remote || send_oos) && mdev->tconn->unused_spare_tle &&
|
2011-11-10 21:45:36 +04:00
|
|
|
test_and_clear_bit(CREATE_BARRIER, &mdev->tconn->flags)) {
|
2011-02-21 16:29:27 +03:00
|
|
|
_tl_add_barrier(mdev->tconn, mdev->tconn->unused_spare_tle);
|
2011-01-19 16:16:30 +03:00
|
|
|
mdev->tconn->unused_spare_tle = NULL;
|
2009-09-26 03:07:19 +04:00
|
|
|
} else {
|
|
|
|
D_ASSERT(!(remote && rw == WRITE &&
|
2011-11-10 21:45:36 +04:00
|
|
|
test_bit(CREATE_BARRIER, &mdev->tconn->flags)));
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* NOTE
|
|
|
|
* Actually, 'local' may be wrong here already, since we may have failed
|
|
|
|
* to write to the meta data, and may become wrong anytime because of
|
|
|
|
* local io-error for some other request, which would lead to us
|
|
|
|
* "detaching" the local disk.
|
|
|
|
*
|
|
|
|
* 'remote' may become wrong any time because the network could fail.
|
|
|
|
*
|
|
|
|
* This is a harmless race condition, though, since it is handled
|
|
|
|
* correctly at the appropriate places; so it just defers the failure
|
|
|
|
* of the respective operation.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* mark them early for readability.
|
|
|
|
* this just sets some state flags. */
|
|
|
|
if (remote)
|
2011-01-25 17:37:43 +03:00
|
|
|
_req_mod(req, TO_BE_SENT);
|
2009-09-26 03:07:19 +04:00
|
|
|
if (local)
|
2011-01-25 17:37:43 +03:00
|
|
|
_req_mod(req, TO_BE_SUBMITTED);
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-01-19 16:16:30 +03:00
|
|
|
list_add_tail(&req->tl_requests, &mdev->tconn->newest_tle->requests);
|
2010-05-27 17:07:43 +04:00
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/* NOTE remote first: to get the concurrent write detection right,
|
|
|
|
* we must register the request before start of local IO. */
|
|
|
|
if (remote) {
|
|
|
|
/* either WRITE and C_CONNECTED,
|
|
|
|
* or READ, and no local disk,
|
|
|
|
* or READ, but not in sync.
|
|
|
|
*/
|
|
|
|
_req_mod(req, (rw == WRITE)
|
2011-01-25 17:37:43 +03:00
|
|
|
? QUEUE_FOR_NET_WRITE
|
|
|
|
: QUEUE_FOR_NET_READ);
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
2010-10-27 16:33:00 +04:00
|
|
|
if (send_oos && drbd_set_out_of_sync(mdev, sector, size))
|
2011-01-25 17:37:43 +03:00
|
|
|
_req_mod(req, QUEUE_FOR_SEND_OOS);
|
2010-10-27 14:21:30 +04:00
|
|
|
|
2011-04-19 19:10:19 +04:00
|
|
|
rcu_read_lock();
|
|
|
|
nc = rcu_dereference(mdev->tconn->net_conf);
|
2010-10-27 16:33:00 +04:00
|
|
|
if (remote &&
|
2011-04-19 19:10:19 +04:00
|
|
|
nc->on_congestion != OC_BLOCK && mdev->tconn->agreed_pro_version >= 96) {
|
2010-10-27 14:21:30 +04:00
|
|
|
int congested = 0;
|
|
|
|
|
2011-04-19 19:10:19 +04:00
|
|
|
if (nc->cong_fill &&
|
|
|
|
atomic_read(&mdev->ap_in_flight) >= nc->cong_fill) {
|
2010-10-27 14:21:30 +04:00
|
|
|
dev_info(DEV, "Congestion-fill threshold reached\n");
|
|
|
|
congested = 1;
|
|
|
|
}
|
|
|
|
|
2011-04-19 19:10:19 +04:00
|
|
|
if (mdev->act_log->used >= nc->cong_extents) {
|
2010-10-27 14:21:30 +04:00
|
|
|
dev_info(DEV, "Congestion-extents threshold reached\n");
|
|
|
|
congested = 1;
|
|
|
|
}
|
|
|
|
|
2011-01-14 21:20:34 +03:00
|
|
|
if (congested) {
|
2011-01-21 16:13:22 +03:00
|
|
|
queue_barrier(mdev); /* last barrier, after mirrored writes */
|
2010-10-27 16:33:00 +04:00
|
|
|
|
2011-04-19 19:10:19 +04:00
|
|
|
if (nc->on_congestion == OC_PULL_AHEAD)
|
2010-10-27 14:21:30 +04:00
|
|
|
_drbd_set_state(_NS(mdev, conn, C_AHEAD), 0, NULL);
|
2011-04-19 19:10:19 +04:00
|
|
|
else /*nc->on_congestion == OC_DISCONNECT */
|
2010-10-27 14:21:30 +04:00
|
|
|
_drbd_set_state(_NS(mdev, conn, C_DISCONNECTING), 0, NULL);
|
|
|
|
}
|
|
|
|
}
|
2011-04-19 19:10:19 +04:00
|
|
|
rcu_read_unlock();
|
2010-10-27 14:21:30 +04:00
|
|
|
|
2011-01-19 16:16:30 +03:00
|
|
|
spin_unlock_irq(&mdev->tconn->req_lock);
|
2009-09-26 03:07:19 +04:00
|
|
|
kfree(b); /* if someone else has beaten us to it... */
|
|
|
|
|
|
|
|
if (local) {
|
|
|
|
req->private_bio->bi_bdev = mdev->ldev->backing_bdev;
|
|
|
|
|
drbd: fix potential data divergence after multiple failures
If we get an IO-error during an activity log transaction,
if we failed to write the bitmap of the evicted extent,
we must not write the transaction itself.
If we failed to write the transaction,
we must not even submit the corresponding bio,
as its extent is not yet marked in the activity log.
Otherwise, if this was a disconneted Primary (degraded cluster), which
now lost its disk as well, and we later re-attach the same backend
storage, we possibly "forget" to resync some parts of the disk that
potentially have been changed.
On the receiving side, when receiving from a peer with unhealthy disk,
checking for pdsk == D_DISKLESS is not enough, we need to set out of
sync and do AL transactions for everything pdsk < D_INCONSISTENT on the
receiving side.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
2010-10-19 01:04:07 +04:00
|
|
|
/* State may have changed since we grabbed our reference on the
|
|
|
|
* mdev->ldev member. Double check, and short-circuit to endio.
|
|
|
|
* In case the last activity log transaction failed to get on
|
|
|
|
* stable storage, and this is a WRITE, we may not even submit
|
|
|
|
* this bio. */
|
|
|
|
if (get_ldev(mdev)) {
|
2010-12-07 12:43:29 +03:00
|
|
|
if (drbd_insert_fault(mdev, rw == WRITE ? DRBD_FAULT_DT_WR
|
|
|
|
: rw == READ ? DRBD_FAULT_DT_RD
|
|
|
|
: DRBD_FAULT_DT_RA))
|
drbd: fix potential data divergence after multiple failures
If we get an IO-error during an activity log transaction,
if we failed to write the bitmap of the evicted extent,
we must not write the transaction itself.
If we failed to write the transaction,
we must not even submit the corresponding bio,
as its extent is not yet marked in the activity log.
Otherwise, if this was a disconneted Primary (degraded cluster), which
now lost its disk as well, and we later re-attach the same backend
storage, we possibly "forget" to resync some parts of the disk that
potentially have been changed.
On the receiving side, when receiving from a peer with unhealthy disk,
checking for pdsk == D_DISKLESS is not enough, we need to set out of
sync and do AL transactions for everything pdsk < D_INCONSISTENT on the
receiving side.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
2010-10-19 01:04:07 +04:00
|
|
|
bio_endio(req->private_bio, -EIO);
|
|
|
|
else
|
|
|
|
generic_make_request(req->private_bio);
|
|
|
|
put_ldev(mdev);
|
|
|
|
} else
|
2009-09-26 03:07:19 +04:00
|
|
|
bio_endio(req->private_bio, -EIO);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
fail_free_complete:
|
2011-05-16 17:31:45 +04:00
|
|
|
if (req->rq_state & RQ_IN_ACT_LOG)
|
2011-03-31 17:18:56 +04:00
|
|
|
drbd_al_complete_io(mdev, &req->i);
|
2009-09-26 03:07:19 +04:00
|
|
|
fail_and_free_req:
|
2011-12-03 14:18:56 +04:00
|
|
|
if (local) {
|
2009-09-26 03:07:19 +04:00
|
|
|
bio_put(req->private_bio);
|
|
|
|
req->private_bio = NULL;
|
|
|
|
put_ldev(mdev);
|
|
|
|
}
|
2010-05-10 18:42:23 +04:00
|
|
|
if (!ret)
|
|
|
|
bio_endio(bio, err);
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
drbd_req_free(req);
|
|
|
|
dec_ap_bio(mdev);
|
|
|
|
kfree(b);
|
|
|
|
|
2010-05-10 18:42:23 +04:00
|
|
|
return ret;
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
2010-12-13 19:48:19 +03:00
|
|
|
int drbd_make_request(struct request_queue *q, struct bio *bio)
|
2009-09-26 03:07:19 +04:00
|
|
|
{
|
|
|
|
struct drbd_conf *mdev = (struct drbd_conf *) q->queuedata;
|
2010-11-09 19:45:06 +03:00
|
|
|
unsigned long start_time;
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2010-11-09 19:45:06 +03:00
|
|
|
start_time = jiffies;
|
|
|
|
|
2009-09-26 03:07:19 +04:00
|
|
|
/*
|
|
|
|
* what we "blindly" assume:
|
|
|
|
*/
|
|
|
|
D_ASSERT(bio->bi_size > 0);
|
2011-02-21 14:41:39 +03:00
|
|
|
D_ASSERT(IS_ALIGNED(bio->bi_size, 512));
|
2009-09-26 03:07:19 +04:00
|
|
|
|
2011-03-31 18:36:43 +04:00
|
|
|
inc_ap_bio(mdev);
|
|
|
|
return __drbd_make_request(mdev, bio, start_time);
|
2009-09-26 03:07:19 +04:00
|
|
|
}
|
|
|
|
|
2011-03-31 18:36:43 +04:00
|
|
|
/* This is called by bio_add_page().
|
|
|
|
*
|
|
|
|
* q->max_hw_sectors and other global limits are already enforced there.
|
2009-09-26 03:07:19 +04:00
|
|
|
*
|
2011-03-31 18:36:43 +04:00
|
|
|
* We need to call down to our lower level device,
|
|
|
|
* in case it has special restrictions.
|
|
|
|
*
|
|
|
|
* We also may need to enforce configured max-bio-bvecs limits.
|
2009-09-26 03:07:19 +04:00
|
|
|
*
|
|
|
|
* As long as the BIO is empty we have to allow at least one bvec,
|
2011-03-31 18:36:43 +04:00
|
|
|
* regardless of size and offset, so no need to ask lower levels.
|
2009-09-26 03:07:19 +04:00
|
|
|
*/
|
|
|
|
int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec)
|
|
|
|
{
|
|
|
|
struct drbd_conf *mdev = (struct drbd_conf *) q->queuedata;
|
|
|
|
unsigned int bio_size = bvm->bi_size;
|
2011-03-31 18:36:43 +04:00
|
|
|
int limit = DRBD_MAX_BIO_SIZE;
|
|
|
|
int backing_limit;
|
|
|
|
|
|
|
|
if (bio_size && get_ldev(mdev)) {
|
2009-09-26 03:07:19 +04:00
|
|
|
struct request_queue * const b =
|
|
|
|
mdev->ldev->backing_bdev->bd_disk->queue;
|
2010-05-14 21:16:41 +04:00
|
|
|
if (b->merge_bvec_fn) {
|
2009-09-26 03:07:19 +04:00
|
|
|
backing_limit = b->merge_bvec_fn(b, bvm, bvec);
|
|
|
|
limit = min(limit, backing_limit);
|
|
|
|
}
|
|
|
|
put_ldev(mdev);
|
|
|
|
}
|
|
|
|
return limit;
|
|
|
|
}
|
2011-03-01 13:08:28 +03:00
|
|
|
|
|
|
|
void request_timer_fn(unsigned long data)
|
|
|
|
{
|
|
|
|
struct drbd_conf *mdev = (struct drbd_conf *) data;
|
2011-03-01 13:08:28 +03:00
|
|
|
struct drbd_tconn *tconn = mdev->tconn;
|
2011-03-01 13:08:28 +03:00
|
|
|
struct drbd_request *req; /* oldest request */
|
|
|
|
struct list_head *le;
|
2011-04-19 19:10:19 +04:00
|
|
|
struct net_conf *nc;
|
2011-07-15 15:53:06 +04:00
|
|
|
unsigned long ent = 0, dt = 0, et, nt; /* effective timeout = ko_count * timeout */
|
2011-04-19 19:10:19 +04:00
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
nc = rcu_dereference(tconn->net_conf);
|
2011-07-05 17:38:59 +04:00
|
|
|
ent = nc ? nc->timeout * HZ/10 * nc->ko_count : 0;
|
|
|
|
|
|
|
|
if (get_ldev(mdev)) {
|
|
|
|
dt = rcu_dereference(mdev->ldev->disk_conf)->disk_timeout * HZ / 10;
|
|
|
|
put_ldev(mdev);
|
|
|
|
}
|
2011-04-19 19:10:19 +04:00
|
|
|
rcu_read_unlock();
|
2011-03-01 13:08:28 +03:00
|
|
|
|
2011-07-05 17:38:59 +04:00
|
|
|
et = min_not_zero(dt, ent);
|
|
|
|
|
|
|
|
if (!et || (mdev->state.conn < C_WF_REPORT_PARAMS && mdev->state.disk <= D_FAILED))
|
2011-03-01 13:08:28 +03:00
|
|
|
return; /* Recurring timer stopped */
|
|
|
|
|
2011-03-01 13:08:28 +03:00
|
|
|
spin_lock_irq(&tconn->req_lock);
|
|
|
|
le = &tconn->oldest_tle->requests;
|
2011-03-01 13:08:28 +03:00
|
|
|
if (list_empty(le)) {
|
2011-03-01 13:08:28 +03:00
|
|
|
spin_unlock_irq(&tconn->req_lock);
|
2011-03-01 13:08:28 +03:00
|
|
|
mod_timer(&mdev->request_timer, jiffies + et);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
le = le->prev;
|
|
|
|
req = list_entry(le, struct drbd_request, tl_requests);
|
2011-07-05 17:38:59 +04:00
|
|
|
if (ent && req->rq_state & RQ_NET_PENDING) {
|
|
|
|
if (time_is_before_eq_jiffies(req->start_time + ent)) {
|
2011-03-01 13:08:28 +03:00
|
|
|
dev_warn(DEV, "Remote failed to finish a request within ko-count * timeout\n");
|
2011-07-05 17:38:59 +04:00
|
|
|
_drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE | CS_HARD, NULL);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (dt && req->rq_state & RQ_LOCAL_PENDING) {
|
|
|
|
if (time_is_before_eq_jiffies(req->start_time + dt)) {
|
|
|
|
dev_warn(DEV, "Local backing device failed to meet the disk-timeout\n");
|
|
|
|
__drbd_chk_io_error(mdev, 1);
|
2011-03-01 13:08:28 +03:00
|
|
|
}
|
|
|
|
}
|
2011-07-15 15:53:06 +04:00
|
|
|
nt = (time_is_before_eq_jiffies(req->start_time + et) ? jiffies : req->start_time) + et;
|
2011-03-01 13:08:28 +03:00
|
|
|
spin_unlock_irq(&tconn->req_lock);
|
2011-07-15 15:53:06 +04:00
|
|
|
mod_timer(&mdev->request_timer, nt);
|
2011-03-01 13:08:28 +03:00
|
|
|
}
|