#include "drbd_req.h"
static int w_make_ov_request(struct drbd_work *w, int cancel);
-static int w_make_resync_request(struct drbd_work *w, int cancel);
-
/* endio handlers:
* drbd_md_io_complete (defined here)
- * drbd_endio_pri (defined here)
- * drbd_endio_sec (defined here)
+ * drbd_request_endio (defined here)
+ * drbd_peer_request_endio (defined here)
* bm_async_io_complete (defined in drbd_bitmap.c)
*
* For all these callbacks, note the following:
/* About the global_state_lock
Each state transition on an device holds a read lock. In case we have
- to evaluate the sync after dependencies, we grab a write lock, because
+ to evaluate the resync after dependencies, we grab a write lock, because
we need stable states on all devices for that. */
rwlock_t global_state_lock;
void drbd_md_io_complete(struct bio *bio, int error)
{
struct drbd_md_io *md_io;
+ struct drbd_conf *mdev;
md_io = (struct drbd_md_io *)bio->bi_private;
+ mdev = container_of(md_io, struct drbd_conf, md_io);
+
md_io->error = error;
- complete(&md_io->event);
+ /* We grabbed an extra reference in _drbd_md_sync_page_io() to be able
+ * to timeout on the lower level device, and eventually detach from it.
+ * If this io completion runs after that timeout expired, this
+ * drbd_md_put_buffer() may allow us to finally try and re-attach.
+ * During normal operation, this only puts that extra reference
+ * down to 1 again.
+ * Make sure we first drop the reference, and only then signal
+ * completion, or we may (in drbd_al_read_log()) cycle so fast into the
+ * next drbd_md_sync_page_io(), that we trigger the
+ * ASSERT(atomic_read(&mdev->md_io_in_use) == 1) there.
+ */
+ drbd_md_put_buffer(mdev);
+ md_io->done = 1;
+ wake_up(&mdev->misc_wait);
+ bio_put(bio);
+ put_ldev(mdev);
}
/* reads on behalf of the partner,
{
unsigned long flags = 0;
struct drbd_conf *mdev = peer_req->w.mdev;
- sector_t e_sector;
+ struct drbd_interval i;
int do_wake;
u64 block_id;
int do_al_complete_io;
* we may no longer access it,
* it may be freed/reused already!
* (as soon as we release the req_lock) */
- e_sector = peer_req->i.sector;
+ i = peer_req->i;
do_al_complete_io = peer_req->flags & EE_CALL_AL_COMPLETE_IO;
block_id = peer_req->block_id;
spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
if (block_id == ID_SYNCER)
- drbd_rs_complete_io(mdev, e_sector);
+ drbd_rs_complete_io(mdev, i.sector);
if (do_wake)
wake_up(&mdev->ee_wait);
if (do_al_complete_io)
- drbd_al_complete_io(mdev, e_sector);
+ drbd_al_complete_io(mdev, &i);
wake_asender(mdev->tconn);
put_ldev(mdev);
/* writes on behalf of the partner, or resync writes,
* "submitted" by the receiver.
*/
-void drbd_endio_sec(struct bio *bio, int error)
+void drbd_peer_request_endio(struct bio *bio, int error)
{
struct drbd_peer_request *peer_req = bio->bi_private;
struct drbd_conf *mdev = peer_req->w.mdev;
/* read, readA or write requests on R_PRIMARY coming from drbd_make_request
*/
-void drbd_endio_pri(struct bio *bio, int error)
+void drbd_request_endio(struct bio *bio, int error)
{
unsigned long flags;
struct drbd_request *req = bio->bi_private;
if (cancel || mdev->state.pdsk != D_UP_TO_DATE) {
_req_mod(req, READ_RETRY_REMOTE_CANCELED);
spin_unlock_irq(&mdev->tconn->req_lock);
- return 1;
+ return 0;
}
spin_unlock_irq(&mdev->tconn->req_lock);
sg_init_table(&sg, 1);
crypto_hash_init(&desc);
- __bio_for_each_segment(bvec, bio, i, 0) {
+ bio_for_each_segment(bvec, bio, i) {
sg_set_page(&sg, bvec->bv_page, bvec->bv_len, bvec->bv_offset);
crypto_hash_update(&desc, &sg, sg.length);
}
crypto_hash_final(&desc, digest);
}
+/* MAYBE merge common code with w_e_end_ov_req */
static int w_e_send_csum(struct drbd_work *w, int cancel)
{
struct drbd_peer_request *peer_req = container_of(w, struct drbd_peer_request, w);
struct drbd_conf *mdev = w->mdev;
int digest_size;
void *digest;
- int ok = 1;
+ int err = 0;
if (unlikely(cancel))
goto out;
- if (likely((peer_req->flags & EE_WAS_ERROR) != 0))
+ if (unlikely((peer_req->flags & EE_WAS_ERROR) != 0))
goto out;
- digest_size = crypto_hash_digestsize(mdev->csums_tfm);
+ digest_size = crypto_hash_digestsize(mdev->tconn->csums_tfm);
digest = kmalloc(digest_size, GFP_NOIO);
if (digest) {
sector_t sector = peer_req->i.sector;
unsigned int size = peer_req->i.size;
- drbd_csum_ee(mdev, mdev->csums_tfm, peer_req, digest);
- /* Free e and pages before send.
+ drbd_csum_ee(mdev, mdev->tconn->csums_tfm, peer_req, digest);
+ /* Free peer_req and pages before send.
* In case we block on congestion, we could otherwise run into
* some distributed deadlock, if the other side blocks on
* congestion as well, because our receiver blocks in
- * drbd_pp_alloc due to pp_in_use > max_buffers. */
- drbd_free_ee(mdev, peer_req);
+ * drbd_alloc_pages due to pp_in_use > max_buffers. */
+ drbd_free_peer_req(mdev, peer_req);
peer_req = NULL;
inc_rs_pending(mdev);
- ok = drbd_send_drequest_csum(mdev, sector, size,
- digest, digest_size,
- P_CSUM_RS_REQUEST);
+ err = drbd_send_drequest_csum(mdev, sector, size,
+ digest, digest_size,
+ P_CSUM_RS_REQUEST);
kfree(digest);
} else {
dev_err(DEV, "kmalloc() of digest failed.\n");
- ok = 0;
+ err = -ENOMEM;
}
out:
if (peer_req)
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
- if (unlikely(!ok))
+ if (unlikely(err))
dev_err(DEV, "drbd_send_drequest(..., csum) failed\n");
- return ok;
+ return err;
}
#define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN)
/* GFP_TRY, because if there is no memory available right now, this may
* be rescheduled for later. It is "only" background resync, after all. */
- peer_req = drbd_alloc_ee(mdev, ID_SYNCER /* unused */, sector, size, GFP_TRY);
+ peer_req = drbd_alloc_peer_req(mdev, ID_SYNCER /* unused */, sector,
+ size, GFP_TRY);
if (!peer_req)
goto defer;
spin_unlock_irq(&mdev->tconn->req_lock);
atomic_add(size >> 9, &mdev->rs_sect_ev);
- if (drbd_submit_ee(mdev, peer_req, READ, DRBD_FAULT_RS_RD) == 0)
+ if (drbd_submit_peer_request(mdev, peer_req, READ, DRBD_FAULT_RS_RD) == 0)
return 0;
/* If it failed because of ENOMEM, retry should help. If it failed
list_del(&peer_req->w.list);
spin_unlock_irq(&mdev->tconn->req_lock);
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
defer:
put_ldev(mdev);
return -EAGAIN;
break;
}
- return 1;
+ return 0;
}
void resync_timer_fn(unsigned long data)
fb->values[i] += value;
}
+struct fifo_buffer *fifo_alloc(int fifo_size)
+{
+ struct fifo_buffer *fb;
+
+ fb = kzalloc(sizeof(struct fifo_buffer) + sizeof(int) * fifo_size, GFP_KERNEL);
+ if (!fb)
+ return NULL;
+
+ fb->head_index = 0;
+ fb->size = fifo_size;
+ fb->total = 0;
+
+ return fb;
+}
+
static int drbd_rs_controller(struct drbd_conf *mdev)
{
+ struct disk_conf *dc;
unsigned int sect_in; /* Number of sectors that came in since the last turn */
unsigned int want; /* The number of sectors we want in the proxy */
int req_sect; /* Number of sectors to request in this turn */
int steps; /* Number of time steps to plan ahead */
int curr_corr;
int max_sect;
+ struct fifo_buffer *plan;
sect_in = atomic_xchg(&mdev->rs_sect_in, 0); /* Number of sectors that came in */
mdev->rs_in_flight -= sect_in;
- spin_lock(&mdev->peer_seq_lock); /* get an atomic view on mdev->rs_plan_s */
+ dc = rcu_dereference(mdev->ldev->disk_conf);
+ plan = rcu_dereference(mdev->rs_plan_s);
- steps = mdev->rs_plan_s.size; /* (mdev->sync_conf.c_plan_ahead * 10 * SLEEP_TIME) / HZ; */
+ steps = plan->size; /* (dc->c_plan_ahead * 10 * SLEEP_TIME) / HZ; */
if (mdev->rs_in_flight + sect_in == 0) { /* At start of resync */
- want = ((mdev->sync_conf.rate * 2 * SLEEP_TIME) / HZ) * steps;
+ want = ((dc->resync_rate * 2 * SLEEP_TIME) / HZ) * steps;
} else { /* normal path */
- want = mdev->sync_conf.c_fill_target ? mdev->sync_conf.c_fill_target :
- sect_in * mdev->sync_conf.c_delay_target * HZ / (SLEEP_TIME * 10);
+ want = dc->c_fill_target ? dc->c_fill_target :
+ sect_in * dc->c_delay_target * HZ / (SLEEP_TIME * 10);
}
- correction = want - mdev->rs_in_flight - mdev->rs_planed;
+ correction = want - mdev->rs_in_flight - plan->total;
/* Plan ahead */
cps = correction / steps;
- fifo_add_val(&mdev->rs_plan_s, cps);
- mdev->rs_planed += cps * steps;
+ fifo_add_val(plan, cps);
+ plan->total += cps * steps;
/* What we do in this step */
- curr_corr = fifo_push(&mdev->rs_plan_s, 0);
- spin_unlock(&mdev->peer_seq_lock);
- mdev->rs_planed -= curr_corr;
+ curr_corr = fifo_push(plan, 0);
+ plan->total -= curr_corr;
req_sect = sect_in + curr_corr;
if (req_sect < 0)
req_sect = 0;
- max_sect = (mdev->sync_conf.c_max_rate * 2 * SLEEP_TIME) / HZ;
+ max_sect = (dc->c_max_rate * 2 * SLEEP_TIME) / HZ;
if (req_sect > max_sect)
req_sect = max_sect;
static int drbd_rs_number_requests(struct drbd_conf *mdev)
{
int number;
- if (mdev->rs_plan_s.size) { /* mdev->sync_conf.c_plan_ahead */
+
+ rcu_read_lock();
+ if (rcu_dereference(mdev->rs_plan_s)->size) {
number = drbd_rs_controller(mdev) >> (BM_BLOCK_SHIFT - 9);
mdev->c_sync_rate = number * HZ * (BM_BLOCK_SIZE / 1024) / SLEEP_TIME;
} else {
- mdev->c_sync_rate = mdev->sync_conf.rate;
+ mdev->c_sync_rate = rcu_dereference(mdev->ldev->disk_conf)->resync_rate;
number = SLEEP_TIME * mdev->c_sync_rate / ((BM_BLOCK_SIZE / 1024) * HZ);
}
+ rcu_read_unlock();
/* ignore the amount of pending requests, the resync controller should
* throttle down to incoming reply rate soon enough anyways. */
return number;
}
-static int w_make_resync_request(struct drbd_work *w, int cancel)
+int w_make_resync_request(struct drbd_work *w, int cancel)
{
struct drbd_conf *mdev = w->mdev;
unsigned long bit;
int i = 0;
if (unlikely(cancel))
- return 1;
+ return 0;
if (mdev->rs_total == 0) {
/* empty resync? */
drbd_resync_finished(mdev);
- return 1;
+ return 0;
}
if (!get_ldev(mdev)) {
to continue resync with a broken disk makes no sense at
all */
dev_err(DEV, "Disk broke down during resync!\n");
- return 1;
+ return 0;
}
max_bio_size = queue_max_hw_sectors(mdev->rq_queue) << 9;
if (bit == DRBD_END_OF_BITMAP) {
mdev->bm_resync_fo = drbd_bm_bits(mdev);
put_ldev(mdev);
- return 1;
+ return 0;
}
sector = BM_BIT_TO_SECT(bit);
/* adjust very last sectors, in case we are oddly sized */
if (sector + (size>>9) > capacity)
size = (capacity-sector)<<9;
- if (mdev->tconn->agreed_pro_version >= 89 && mdev->csums_tfm) {
+ if (mdev->tconn->agreed_pro_version >= 89 && mdev->tconn->csums_tfm) {
switch (read_for_csum(mdev, sector, size)) {
case -EIO: /* Disk failure */
put_ldev(mdev);
- return 0;
+ return -EIO;
case -EAGAIN: /* allocation failed, or ldev busy */
drbd_rs_complete_io(mdev, sector);
mdev->bm_resync_fo = BM_SECT_TO_BIT(sector);
BUG();
}
} else {
+ int err;
+
inc_rs_pending(mdev);
- if (!drbd_send_drequest(mdev, P_RS_DATA_REQUEST,
- sector, size, ID_SYNCER)) {
+ err = drbd_send_drequest(mdev, P_RS_DATA_REQUEST,
+ sector, size, ID_SYNCER);
+ if (err) {
dev_err(DEV, "drbd_send_drequest() failed, aborting...\n");
dec_rs_pending(mdev);
put_ldev(mdev);
- return 0;
+ return err;
}
}
}
* until then resync "work" is "inactive" ...
*/
put_ldev(mdev);
- return 1;
+ return 0;
}
requeue:
mdev->rs_in_flight += (i << (BM_BLOCK_SHIFT - 9));
mod_timer(&mdev->resync_timer, jiffies + SLEEP_TIME);
put_ldev(mdev);
- return 1;
+ return 0;
}
static int w_make_ov_request(struct drbd_work *w, int cancel)
size = (capacity-sector)<<9;
inc_rs_pending(mdev);
- if (!drbd_send_ov_request(mdev, sector, size)) {
+ if (drbd_send_ov_request(mdev, sector, size)) {
dec_rs_pending(mdev);
return 0;
}
{
struct drbd_conf *mdev = w->mdev;
kfree(w);
- ov_oos_print(mdev);
+ ov_out_of_sync_print(mdev);
drbd_resync_finished(mdev);
- return 1;
+ return 0;
}
static int w_resync_finished(struct drbd_work *w, int cancel)
drbd_resync_finished(mdev);
- return 1;
+ return 0;
}
static void ping_peer(struct drbd_conf *mdev)
w = kmalloc(sizeof(struct drbd_work), GFP_ATOMIC);
if (w) {
w->cb = w_resync_finished;
+ w->mdev = mdev;
drbd_queue_work(&mdev->tconn->data.work, w);
return 1;
}
ping_peer(mdev);
spin_lock_irq(&mdev->tconn->req_lock);
- os = mdev->state;
+ os = drbd_read_state(mdev);
verify_done = (os.conn == C_VERIFY_S || os.conn == C_VERIFY_T);
if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T)
khelper_cmd = "after-resync-target";
- if (mdev->csums_tfm && mdev->rs_total) {
+ if (mdev->tconn->csums_tfm && mdev->rs_total) {
const unsigned long s = mdev->rs_same_csum;
const unsigned long t = mdev->rs_total;
const int ratio =
/* helper */
static void move_to_net_ee_or_free(struct drbd_conf *mdev, struct drbd_peer_request *peer_req)
{
- if (drbd_ee_has_active_page(peer_req)) {
+ if (drbd_peer_req_has_active_page(peer_req)) {
/* This might happen if sendpage() has not finished */
int i = (peer_req->i.size + PAGE_SIZE -1) >> PAGE_SHIFT;
atomic_add(i, &mdev->pp_in_use_by_net);
spin_unlock_irq(&mdev->tconn->req_lock);
wake_up(&drbd_pp_wait);
} else
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
}
/**
{
struct drbd_peer_request *peer_req = container_of(w, struct drbd_peer_request, w);
struct drbd_conf *mdev = w->mdev;
- int ok;
+ int err;
if (unlikely(cancel)) {
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
dec_unacked(mdev);
- return 1;
+ return 0;
}
if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
- ok = drbd_send_block(mdev, P_DATA_REPLY, peer_req);
+ err = drbd_send_block(mdev, P_DATA_REPLY, peer_req);
} else {
if (__ratelimit(&drbd_ratelimit_state))
dev_err(DEV, "Sending NegDReply. sector=%llus.\n",
(unsigned long long)peer_req->i.sector);
- ok = drbd_send_ack(mdev, P_NEG_DREPLY, peer_req);
+ err = drbd_send_ack(mdev, P_NEG_DREPLY, peer_req);
}
dec_unacked(mdev);
move_to_net_ee_or_free(mdev, peer_req);
- if (unlikely(!ok))
+ if (unlikely(err))
dev_err(DEV, "drbd_send_block() failed\n");
- return ok;
+ return err;
}
/**
- * w_e_end_rsdata_req() - Worker callback to send a P_RS_DATA_REPLY packet in response to a P_RS_DATA_REQUESTRS
+ * w_e_end_rsdata_req() - Worker callback to send a P_RS_DATA_REPLY packet in response to a P_RS_DATA_REQUEST
* @mdev: DRBD device.
* @w: work object.
* @cancel: The connection will be closed anyways
{
struct drbd_peer_request *peer_req = container_of(w, struct drbd_peer_request, w);
struct drbd_conf *mdev = w->mdev;
- int ok;
+ int err;
if (unlikely(cancel)) {
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
dec_unacked(mdev);
- return 1;
+ return 0;
}
if (get_ldev_if_state(mdev, D_FAILED)) {
}
if (mdev->state.conn == C_AHEAD) {
- ok = drbd_send_ack(mdev, P_RS_CANCEL, peer_req);
+ err = drbd_send_ack(mdev, P_RS_CANCEL, peer_req);
} else if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
if (likely(mdev->state.pdsk >= D_INCONSISTENT)) {
inc_rs_pending(mdev);
- ok = drbd_send_block(mdev, P_RS_DATA_REPLY, peer_req);
+ err = drbd_send_block(mdev, P_RS_DATA_REPLY, peer_req);
} else {
if (__ratelimit(&drbd_ratelimit_state))
dev_err(DEV, "Not sending RSDataReply, "
"partner DISKLESS!\n");
- ok = 1;
+ err = 0;
}
} else {
if (__ratelimit(&drbd_ratelimit_state))
dev_err(DEV, "Sending NegRSDReply. sector %llus.\n",
(unsigned long long)peer_req->i.sector);
- ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, peer_req);
+ err = drbd_send_ack(mdev, P_NEG_RS_DREPLY, peer_req);
/* update resync data with failure */
drbd_rs_failed_io(mdev, peer_req->i.sector, peer_req->i.size);
move_to_net_ee_or_free(mdev, peer_req);
- if (unlikely(!ok))
+ if (unlikely(err))
dev_err(DEV, "drbd_send_block() failed\n");
- return ok;
+ return err;
}
int w_e_end_csum_rs_req(struct drbd_work *w, int cancel)
struct digest_info *di;
int digest_size;
void *digest = NULL;
- int ok, eq = 0;
+ int err, eq = 0;
if (unlikely(cancel)) {
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
dec_unacked(mdev);
- return 1;
+ return 0;
}
if (get_ldev(mdev)) {
/* quick hack to try to avoid a race against reconfiguration.
* a real fix would be much more involved,
* introducing more locking mechanisms */
- if (mdev->csums_tfm) {
- digest_size = crypto_hash_digestsize(mdev->csums_tfm);
+ if (mdev->tconn->csums_tfm) {
+ digest_size = crypto_hash_digestsize(mdev->tconn->csums_tfm);
D_ASSERT(digest_size == di->digest_size);
digest = kmalloc(digest_size, GFP_NOIO);
}
if (digest) {
- drbd_csum_ee(mdev, mdev->csums_tfm, peer_req, digest);
+ drbd_csum_ee(mdev, mdev->tconn->csums_tfm, peer_req, digest);
eq = !memcmp(digest, di->digest, digest_size);
kfree(digest);
}
drbd_set_in_sync(mdev, peer_req->i.sector, peer_req->i.size);
/* rs_same_csums unit is BM_BLOCK_SIZE */
mdev->rs_same_csum += peer_req->i.size >> BM_BLOCK_SHIFT;
- ok = drbd_send_ack(mdev, P_RS_IS_IN_SYNC, peer_req);
+ err = drbd_send_ack(mdev, P_RS_IS_IN_SYNC, peer_req);
} else {
inc_rs_pending(mdev);
peer_req->block_id = ID_SYNCER; /* By setting block_id, digest pointer becomes invalid! */
peer_req->flags &= ~EE_HAS_DIGEST; /* This peer request no longer has a digest pointer */
kfree(di);
- ok = drbd_send_block(mdev, P_RS_DATA_REPLY, peer_req);
+ err = drbd_send_block(mdev, P_RS_DATA_REPLY, peer_req);
}
} else {
- ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, peer_req);
+ err = drbd_send_ack(mdev, P_NEG_RS_DREPLY, peer_req);
if (__ratelimit(&drbd_ratelimit_state))
dev_err(DEV, "Sending NegDReply. I guess it gets messy.\n");
}
dec_unacked(mdev);
move_to_net_ee_or_free(mdev, peer_req);
- if (unlikely(!ok))
+ if (unlikely(err))
dev_err(DEV, "drbd_send_block/ack() failed\n");
- return ok;
+ return err;
}
int w_e_end_ov_req(struct drbd_work *w, int cancel)
unsigned int size = peer_req->i.size;
int digest_size;
void *digest;
- int ok = 1;
+ int err = 0;
if (unlikely(cancel))
goto out;
- digest_size = crypto_hash_digestsize(mdev->verify_tfm);
+ digest_size = crypto_hash_digestsize(mdev->tconn->verify_tfm);
digest = kmalloc(digest_size, GFP_NOIO);
if (!digest) {
- ok = 0; /* terminate the connection in case the allocation failed */
+ err = 1; /* terminate the connection in case the allocation failed */
goto out;
}
if (likely(!(peer_req->flags & EE_WAS_ERROR)))
- drbd_csum_ee(mdev, mdev->verify_tfm, peer_req, digest);
+ drbd_csum_ee(mdev, mdev->tconn->verify_tfm, peer_req, digest);
else
memset(digest, 0, digest_size);
* In case we block on congestion, we could otherwise run into
* some distributed deadlock, if the other side blocks on
* congestion as well, because our receiver blocks in
- * drbd_pp_alloc due to pp_in_use > max_buffers. */
- drbd_free_ee(mdev, peer_req);
+ * drbd_alloc_pages due to pp_in_use > max_buffers. */
+ drbd_free_peer_req(mdev, peer_req);
peer_req = NULL;
inc_rs_pending(mdev);
- ok = drbd_send_drequest_csum(mdev, sector, size,
- digest, digest_size,
- P_OV_REPLY);
- if (!ok)
+ err = drbd_send_drequest_csum(mdev, sector, size, digest, digest_size, P_OV_REPLY);
+ if (err)
dec_rs_pending(mdev);
kfree(digest);
out:
if (peer_req)
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
dec_unacked(mdev);
- return ok;
+ return err;
}
-void drbd_ov_oos_found(struct drbd_conf *mdev, sector_t sector, int size)
+void drbd_ov_out_of_sync_found(struct drbd_conf *mdev, sector_t sector, int size)
{
if (mdev->ov_last_oos_start + mdev->ov_last_oos_size == sector) {
mdev->ov_last_oos_size += size>>9;
sector_t sector = peer_req->i.sector;
unsigned int size = peer_req->i.size;
int digest_size;
- int ok, eq = 0;
+ int err, eq = 0;
if (unlikely(cancel)) {
- drbd_free_ee(mdev, peer_req);
+ drbd_free_peer_req(mdev, peer_req);
dec_unacked(mdev);
- return 1;
+ return 0;
}
/* after "cancel", because after drbd_disconnect/drbd_rs_cancel_all
di = peer_req->digest;
if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
- digest_size = crypto_hash_digestsize(mdev->verify_tfm);
+ digest_size = crypto_hash_digestsize(mdev->tconn->verify_tfm);
digest = kmalloc(digest_size, GFP_NOIO);
if (digest) {
- drbd_csum_ee(mdev, mdev->verify_tfm, peer_req, digest);
+ drbd_csum_ee(mdev, mdev->tconn->verify_tfm, peer_req, digest);
D_ASSERT(digest_size == di->digest_size);
eq = !memcmp(digest, di->digest, digest_size);
}
}
- /* Free e and pages before send.
- * In case we block on congestion, we could otherwise run into
- * some distributed deadlock, if the other side blocks on
- * congestion as well, because our receiver blocks in
- * drbd_pp_alloc due to pp_in_use > max_buffers. */
- drbd_free_ee(mdev, peer_req);
+ /* Free peer_req and pages before send.
+ * In case we block on congestion, we could otherwise run into
+ * some distributed deadlock, if the other side blocks on
+ * congestion as well, because our receiver blocks in
+ * drbd_alloc_pages due to pp_in_use > max_buffers. */
+ drbd_free_peer_req(mdev, peer_req);
if (!eq)
- drbd_ov_oos_found(mdev, sector, size);
+ drbd_ov_out_of_sync_found(mdev, sector, size);
else
- ov_oos_print(mdev);
+ ov_out_of_sync_print(mdev);
- ok = drbd_send_ack_ex(mdev, P_OV_RESULT, sector, size,
- eq ? ID_IN_SYNC : ID_OUT_OF_SYNC);
+ err = drbd_send_ack_ex(mdev, P_OV_RESULT, sector, size,
+ eq ? ID_IN_SYNC : ID_OUT_OF_SYNC);
dec_unacked(mdev);
drbd_advance_rs_marks(mdev, mdev->ov_left);
if (mdev->ov_left == 0) {
- ov_oos_print(mdev);
+ ov_out_of_sync_print(mdev);
drbd_resync_finished(mdev);
}
- return ok;
+ return err;
}
int w_prev_work_done(struct drbd_work *w, int cancel)
struct drbd_wq_barrier *b = container_of(w, struct drbd_wq_barrier, w);
complete(&b->done);
- return 1;
+ return 0;
}
int w_send_barrier(struct drbd_work *w, int cancel)
{
+ struct drbd_socket *sock;
struct drbd_tl_epoch *b = container_of(w, struct drbd_tl_epoch, w);
struct drbd_conf *mdev = w->mdev;
- struct p_barrier *p = &mdev->tconn->data.sbuf.barrier;
- int ok = 1;
+ struct p_barrier *p;
/* really avoid racing with tl_clear. w.cb may have been referenced
* just before it was reassigned and re-queued, so double check that.
cancel = 1;
spin_unlock_irq(&mdev->tconn->req_lock);
if (cancel)
- return 1;
-
- if (!drbd_get_data_sock(mdev->tconn))
return 0;
+
+ sock = &mdev->tconn->data;
+ p = drbd_prepare_command(mdev, sock);
+ if (!p)
+ return -EIO;
p->barrier = b->br_number;
/* inc_ap_pending was done where this was queued.
* dec_ap_pending will be done in got_BarrierAck
* or (on connection loss) in w_clear_epoch. */
- ok = _drbd_send_cmd(mdev, mdev->tconn->data.socket, P_BARRIER,
- &p->head, sizeof(*p), 0);
- drbd_put_data_sock(mdev->tconn);
-
- return ok;
+ return drbd_send_command(mdev, sock, P_BARRIER, sizeof(*p), NULL, 0);
}
int w_send_write_hint(struct drbd_work *w, int cancel)
{
struct drbd_conf *mdev = w->mdev;
+ struct drbd_socket *sock;
+
if (cancel)
- return 1;
- return drbd_send_short_cmd(mdev, P_UNPLUG_REMOTE);
+ return 0;
+ sock = &mdev->tconn->data;
+ if (!drbd_prepare_command(mdev, sock))
+ return -EIO;
+ return drbd_send_command(mdev, sock, P_UNPLUG_REMOTE, 0, NULL, 0);
}
-int w_send_oos(struct drbd_work *w, int cancel)
+int w_send_out_of_sync(struct drbd_work *w, int cancel)
{
struct drbd_request *req = container_of(w, struct drbd_request, w);
struct drbd_conf *mdev = w->mdev;
- int ok;
+ int err;
if (unlikely(cancel)) {
req_mod(req, SEND_CANCELED);
- return 1;
+ return 0;
}
- ok = drbd_send_oos(mdev, req);
+ err = drbd_send_out_of_sync(mdev, req);
req_mod(req, OOS_HANDED_TO_NETWORK);
- return ok;
+ return err;
}
/**
{
struct drbd_request *req = container_of(w, struct drbd_request, w);
struct drbd_conf *mdev = w->mdev;
- int ok;
+ int err;
if (unlikely(cancel)) {
req_mod(req, SEND_CANCELED);
- return 1;
+ return 0;
}
- ok = drbd_send_dblock(mdev, req);
- req_mod(req, ok ? HANDED_OVER_TO_NETWORK : SEND_FAILED);
+ err = drbd_send_dblock(mdev, req);
+ req_mod(req, err ? SEND_FAILED : HANDED_OVER_TO_NETWORK);
- return ok;
+ return err;
}
/**
{
struct drbd_request *req = container_of(w, struct drbd_request, w);
struct drbd_conf *mdev = w->mdev;
- int ok;
+ int err;
if (unlikely(cancel)) {
req_mod(req, SEND_CANCELED);
- return 1;
+ return 0;
}
- ok = drbd_send_drequest(mdev, P_DATA_REQUEST, req->i.sector, req->i.size,
- (unsigned long)req);
+ err = drbd_send_drequest(mdev, P_DATA_REQUEST, req->i.sector, req->i.size,
+ (unsigned long)req);
- if (!ok) {
- /* ?? we set C_TIMEOUT or C_BROKEN_PIPE in drbd_send();
- * so this is probably redundant */
- if (mdev->state.conn >= C_CONNECTED)
- drbd_force_state(mdev, NS(conn, C_NETWORK_FAILURE));
- }
- req_mod(req, ok ? HANDED_OVER_TO_NETWORK : SEND_FAILED);
+ req_mod(req, err ? SEND_FAILED : HANDED_OVER_TO_NETWORK);
- return ok;
+ return err;
}
int w_restart_disk_io(struct drbd_work *w, int cancel)
struct drbd_conf *mdev = w->mdev;
if (bio_data_dir(req->master_bio) == WRITE && req->rq_state & RQ_IN_ACT_LOG)
- drbd_al_begin_io(mdev, req->i.sector);
- /* Calling drbd_al_begin_io() out of the worker might deadlocks
- theoretically. Practically it can not deadlock, since this is
- only used when unfreezing IOs. All the extents of the requests
- that made it into the TL are already active */
+ drbd_al_begin_io(mdev, &req->i);
drbd_req_make_private_bio(req, req->master_bio);
req->private_bio->bi_bdev = mdev->ldev->backing_bdev;
generic_make_request(req->private_bio);
- return 1;
+ return 0;
}
static int _drbd_may_sync_now(struct drbd_conf *mdev)
{
struct drbd_conf *odev = mdev;
+ int resync_after;
while (1) {
- if (odev->sync_conf.after == -1)
+ if (!odev->ldev)
+ return 1;
+ rcu_read_lock();
+ resync_after = rcu_dereference(odev->ldev->disk_conf)->resync_after;
+ rcu_read_unlock();
+ if (resync_after == -1)
return 1;
- odev = minor_to_mdev(odev->sync_conf.after);
+ odev = minor_to_mdev(resync_after);
if (!expect(odev))
return 1;
if ((odev->state.conn >= C_SYNC_SOURCE &&
struct drbd_conf *odev;
int i, rv = 0;
- for (i = 0; i < minor_count; i++) {
- odev = minor_to_mdev(i);
- if (!odev)
- continue;
+ rcu_read_lock();
+ idr_for_each_entry(&minors, odev, i) {
if (odev->state.conn == C_STANDALONE && odev->state.disk == D_DISKLESS)
continue;
if (!_drbd_may_sync_now(odev))
rv |= (__drbd_set_state(_NS(odev, aftr_isp, 1), CS_HARD, NULL)
!= SS_NOTHING_TO_DO);
}
+ rcu_read_unlock();
return rv;
}
struct drbd_conf *odev;
int i, rv = 0;
- for (i = 0; i < minor_count; i++) {
- odev = minor_to_mdev(i);
- if (!odev)
- continue;
+ rcu_read_lock();
+ idr_for_each_entry(&minors, odev, i) {
if (odev->state.conn == C_STANDALONE && odev->state.disk == D_DISKLESS)
continue;
if (odev->state.aftr_isp) {
!= SS_NOTHING_TO_DO) ;
}
}
+ rcu_read_unlock();
return rv;
}
write_unlock_irq(&global_state_lock);
}
-static int sync_after_error(struct drbd_conf *mdev, int o_minor)
+/* caller must hold global_state_lock */
+enum drbd_ret_code drbd_resync_after_valid(struct drbd_conf *mdev, int o_minor)
{
struct drbd_conf *odev;
+ int resync_after;
if (o_minor == -1)
return NO_ERROR;
if (o_minor < -1 || minor_to_mdev(o_minor) == NULL)
- return ERR_SYNC_AFTER;
+ return ERR_RESYNC_AFTER;
/* check for loops */
odev = minor_to_mdev(o_minor);
while (1) {
if (odev == mdev)
- return ERR_SYNC_AFTER_CYCLE;
+ return ERR_RESYNC_AFTER_CYCLE;
+ rcu_read_lock();
+ resync_after = rcu_dereference(odev->ldev->disk_conf)->resync_after;
+ rcu_read_unlock();
/* dependency chain ends here, no cycles. */
- if (odev->sync_conf.after == -1)
+ if (resync_after == -1)
return NO_ERROR;
/* follow the dependency chain */
- odev = minor_to_mdev(odev->sync_conf.after);
+ odev = minor_to_mdev(resync_after);
}
}
-int drbd_alter_sa(struct drbd_conf *mdev, int na)
+/* caller must hold global_state_lock */
+void drbd_resync_after_changed(struct drbd_conf *mdev)
{
int changes;
- int retcode;
- write_lock_irq(&global_state_lock);
- retcode = sync_after_error(mdev, na);
- if (retcode == NO_ERROR) {
- mdev->sync_conf.after = na;
- do {
- changes = _drbd_pause_after(mdev);
- changes |= _drbd_resume_next(mdev);
- } while (changes);
- }
- write_unlock_irq(&global_state_lock);
- return retcode;
+ do {
+ changes = _drbd_pause_after(mdev);
+ changes |= _drbd_resume_next(mdev);
+ } while (changes);
}
void drbd_rs_controller_reset(struct drbd_conf *mdev)
{
+ struct fifo_buffer *plan;
+
atomic_set(&mdev->rs_sect_in, 0);
atomic_set(&mdev->rs_sect_ev, 0);
mdev->rs_in_flight = 0;
- mdev->rs_planed = 0;
- spin_lock(&mdev->peer_seq_lock);
- fifo_set(&mdev->rs_plan_s, 0);
- spin_unlock(&mdev->peer_seq_lock);
+
+ /* Updating the RCU protected object in place is necessary since
+ this function gets called from atomic context.
+ It is valid since all other updates also lead to an completely
+ empty fifo */
+ rcu_read_lock();
+ plan = rcu_dereference(mdev->rs_plan_s);
+ plan->total = 0;
+ fifo_set(plan, 0);
+ rcu_read_unlock();
}
void start_resync_timer_fn(unsigned long data)
dev_warn(DEV, "w_start_resync later...\n");
mdev->start_resync_timer.expires = jiffies + HZ/10;
add_timer(&mdev->start_resync_timer);
- return 1;
+ return 0;
}
drbd_start_resync(mdev, C_SYNC_SOURCE);
- clear_bit(AHEAD_TO_SYNC_SOURCE, &mdev->current_epoch->flags);
- return 1;
+ clear_bit(AHEAD_TO_SYNC_SOURCE, &mdev->flags);
+ return 0;
}
/**
if (r > 0) {
dev_info(DEV, "before-resync-target handler returned %d, "
"dropping connection.\n", r);
- drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
+ conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
return;
}
} else /* C_SYNC_SOURCE */ {
} else {
dev_info(DEV, "before-resync-source handler returned %d, "
"dropping connection.\n", r);
- drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
+ conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
return;
}
}
}
clear_bit(B_RS_H_DONE, &mdev->flags);
+ write_lock_irq(&global_state_lock);
if (!get_ldev_if_state(mdev, D_NEGOTIATING)) {
+ write_unlock_irq(&global_state_lock);
mutex_unlock(mdev->state_mutex);
return;
}
- write_lock_irq(&global_state_lock);
- ns = mdev->state;
+ ns = drbd_read_state(mdev);
ns.aftr_isp = !_drbd_may_sync_now(mdev);
ns.pdsk = D_INCONSISTENT;
r = __drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
- ns = mdev->state;
+ ns = drbd_read_state(mdev);
if (ns.conn < C_CONNECTED)
r = SS_UNKNOWN_ERROR;
* detect connection loss, then waiting for a ping
* response (implicit in drbd_resync_finished) reduces
* the race considerably, but does not solve it. */
- if (side == C_SYNC_SOURCE)
- schedule_timeout_interruptible(
- mdev->tconn->net_conf->ping_int * HZ +
- mdev->tconn->net_conf->ping_timeo*HZ/9);
+ if (side == C_SYNC_SOURCE) {
+ struct net_conf *nc;
+ int timeo;
+
+ rcu_read_lock();
+ nc = rcu_dereference(mdev->tconn->net_conf);
+ timeo = nc->ping_int * HZ + nc->ping_timeo * HZ / 9;
+ rcu_read_unlock();
+ schedule_timeout_interruptible(timeo);
+ }
drbd_resync_finished(mdev);
}
mutex_unlock(mdev->state_mutex);
}
-static int _worker_dying(int vnr, void *p, void *data)
-{
- struct drbd_conf *mdev = (struct drbd_conf *)p;
-
- D_ASSERT(mdev->state.disk == D_DISKLESS && mdev->state.conn == C_STANDALONE);
- /* _drbd_set_state only uses stop_nowait.
- * wait here for the exiting receiver. */
- drbd_thread_stop(&mdev->tconn->receiver);
- drbd_mdev_cleanup(mdev);
-
- clear_bit(DEVICE_DYING, &mdev->flags);
- clear_bit(CONFIG_PENDING, &mdev->flags);
- wake_up(&mdev->state_wait);
-
- return 0;
-}
-
int drbd_worker(struct drbd_thread *thi)
{
struct drbd_tconn *tconn = thi->tconn;
struct drbd_work *w = NULL;
+ struct drbd_conf *mdev;
+ struct net_conf *nc;
LIST_HEAD(work_list);
- int intr = 0;
+ int vnr, intr = 0;
+ int cork;
while (get_t_state(thi) == RUNNING) {
drbd_thread_current_set_cpu(thi);
if (down_trylock(&tconn->data.work.s)) {
mutex_lock(&tconn->data.mutex);
- if (tconn->data.socket && !tconn->net_conf->no_cork)
+
+ rcu_read_lock();
+ nc = rcu_dereference(tconn->net_conf);
+ cork = nc ? nc->tcp_cork : 0;
+ rcu_read_unlock();
+
+ if (tconn->data.socket && cork)
drbd_tcp_uncork(tconn->data.socket);
mutex_unlock(&tconn->data.mutex);
intr = down_interruptible(&tconn->data.work.s);
mutex_lock(&tconn->data.mutex);
- if (tconn->data.socket && !tconn->net_conf->no_cork)
+ if (tconn->data.socket && cork)
drbd_tcp_cork(tconn->data.socket);
mutex_unlock(&tconn->data.mutex);
}
list_del_init(&w->list);
spin_unlock_irq(&tconn->data.work.q_lock);
- if (!w->cb(w, tconn->cstate < C_WF_REPORT_PARAMS)) {
+ if (w->cb(w, tconn->cstate < C_WF_REPORT_PARAMS)) {
/* dev_warn(DEV, "worker: a callback failed! \n"); */
if (tconn->cstate >= C_WF_REPORT_PARAMS)
conn_request_state(tconn, NS(conn, C_NETWORK_FAILURE), CS_HARD);
*/
spin_unlock_irq(&tconn->data.work.q_lock);
- idr_for_each(&tconn->volumes, _worker_dying, NULL);
+ rcu_read_lock();
+ idr_for_each_entry(&tconn->volumes, mdev, vnr) {
+ D_ASSERT(mdev->state.disk == D_DISKLESS && mdev->state.conn == C_STANDALONE);
+ kref_get(&mdev->kref);
+ rcu_read_unlock();
+ drbd_mdev_cleanup(mdev);
+ kref_put(&mdev->kref, &drbd_minor_destroy);
+ rcu_read_lock();
+ }
+ rcu_read_unlock();
return 0;
}
projects / cascardo / linux.git / blobdiff