Merge branch 'for-jens' of git://git.drbd.org/linux-drbd into for-3.5/drivers
Philipp writes:
This are the updates we have in the drbd-8.3 tree. They are intended
for your "for-3.5/drivers" drivers branch.
These changes include one new feature:
* Allow detach from frozen backing devices with the new --force option;
configurable timeout for backing devices by the new disk-timeout option
And huge number of bug fixes:
* Fixed a write ordering problem on SyncTarget nodes for a write
to a block that gets resynced at the same time. The bug can
only be triggered with a device that has a firmware that
actually reorders writes to the same block
* Fixed a race between disconnect and receive_state, that could cause
a IO lockup
* Fixed resend/resubmit for requests with disk or network timeout
* Make sure that hard state changed do not disturb the connection
establishing process (I.e. detach due to an IO error). When the
bug was triggered it caused a retry in the connect process
* Postpone soft state changes to no disturb the connection
establishing process (I.e. becoming primary). When the bug
was triggered it could cause both nodes going into SyncSource state
* Fixed a refcount leak that could cause failures when trying to
unload a protocol family modules, that was used by DRBD
* Dedicated page pool for meta data IOs
* Deny normal detach (as opposed to --forced) if the user tries
to detach from the last UpToDate disk in the resource
* Fixed a possible protocol error that could be caused by
"unusual" BIOs.
* Enforce the disk-timeout option also on meta-data IO operations
* Implemented stable bitmap pages when we do a full write out of
the bitmap
* Fixed a rare compatibility issue with DRBD's older than 8.3.7
when negotiating the bio_size
* Fixed a rare race condition where an empty resync could stall with
if pause/unpause events happen in parallel
* Made the re-establishing of connections quicker, if it got a broken pipe
once. Previously there was a bug in the code caused it to waste the first
successful established connection after a broken pipe event.
PS: I am postponing the drbd-8.4 for mainline for one or two kernel
development cycles more (the ~400 patchets set).
diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c
index cf0e63d..e54e31b 100644
--- a/drivers/block/drbd/drbd_actlog.c
+++ b/drivers/block/drbd/drbd_actlog.c
@@ -65,39 +65,80 @@
int w_al_write_transaction(struct drbd_conf *, struct drbd_work *, int);
+void *drbd_md_get_buffer(struct drbd_conf *mdev)
+{
+ int r;
+
+ wait_event(mdev->misc_wait,
+ (r = atomic_cmpxchg(&mdev->md_io_in_use, 0, 1)) == 0 ||
+ mdev->state.disk <= D_FAILED);
+
+ return r ? NULL : page_address(mdev->md_io_page);
+}
+
+void drbd_md_put_buffer(struct drbd_conf *mdev)
+{
+ if (atomic_dec_and_test(&mdev->md_io_in_use))
+ wake_up(&mdev->misc_wait);
+}
+
+static bool md_io_allowed(struct drbd_conf *mdev)
+{
+ enum drbd_disk_state ds = mdev->state.disk;
+ return ds >= D_NEGOTIATING || ds == D_ATTACHING;
+}
+
+void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
+ unsigned int *done)
+{
+ long dt = bdev->dc.disk_timeout * HZ / 10;
+ if (dt == 0)
+ dt = MAX_SCHEDULE_TIMEOUT;
+
+ dt = wait_event_timeout(mdev->misc_wait, *done || !md_io_allowed(mdev), dt);
+ if (dt == 0)
+ dev_err(DEV, "meta-data IO operation timed out\n");
+}
+
static int _drbd_md_sync_page_io(struct drbd_conf *mdev,
struct drbd_backing_dev *bdev,
struct page *page, sector_t sector,
int rw, int size)
{
struct bio *bio;
- struct drbd_md_io md_io;
int ok;
- md_io.mdev = mdev;
- init_completion(&md_io.event);
- md_io.error = 0;
+ mdev->md_io.done = 0;
+ mdev->md_io.error = -ENODEV;
if ((rw & WRITE) && !test_bit(MD_NO_FUA, &mdev->flags))
rw |= REQ_FUA | REQ_FLUSH;
rw |= REQ_SYNC;
- bio = bio_alloc(GFP_NOIO, 1);
+ bio = bio_alloc_drbd(GFP_NOIO);
bio->bi_bdev = bdev->md_bdev;
bio->bi_sector = sector;
ok = (bio_add_page(bio, page, size, 0) == size);
if (!ok)
goto out;
- bio->bi_private = &md_io;
+ bio->bi_private = &mdev->md_io;
bio->bi_end_io = drbd_md_io_complete;
bio->bi_rw = rw;
+ if (!get_ldev_if_state(mdev, D_ATTACHING)) { /* Corresponding put_ldev in drbd_md_io_complete() */
+ dev_err(DEV, "ASSERT FAILED: get_ldev_if_state() == 1 in _drbd_md_sync_page_io()\n");
+ ok = 0;
+ goto out;
+ }
+
+ bio_get(bio); /* one bio_put() is in the completion handler */
+ atomic_inc(&mdev->md_io_in_use); /* drbd_md_put_buffer() is in the completion handler */
if (drbd_insert_fault(mdev, (rw & WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD))
bio_endio(bio, -EIO);
else
submit_bio(rw, bio);
- wait_for_completion(&md_io.event);
- ok = bio_flagged(bio, BIO_UPTODATE) && md_io.error == 0;
+ wait_until_done_or_disk_failure(mdev, bdev, &mdev->md_io.done);
+ ok = bio_flagged(bio, BIO_UPTODATE) && mdev->md_io.error == 0;
out:
bio_put(bio);
@@ -111,7 +152,7 @@
int offset = 0;
struct page *iop = mdev->md_io_page;
- D_ASSERT(mutex_is_locked(&mdev->md_io_mutex));
+ D_ASSERT(atomic_read(&mdev->md_io_in_use) == 1);
BUG_ON(!bdev->md_bdev);
@@ -328,8 +369,13 @@
return 1;
}
- mutex_lock(&mdev->md_io_mutex); /* protects md_io_buffer, al_tr_cycle, ... */
- buffer = (struct al_transaction *)page_address(mdev->md_io_page);
+ buffer = drbd_md_get_buffer(mdev); /* protects md_io_buffer, al_tr_cycle, ... */
+ if (!buffer) {
+ dev_err(DEV, "disk failed while waiting for md_io buffer\n");
+ complete(&((struct update_al_work *)w)->event);
+ put_ldev(mdev);
+ return 1;
+ }
buffer->magic = __constant_cpu_to_be32(DRBD_MAGIC);
buffer->tr_number = cpu_to_be32(mdev->al_tr_number);
@@ -374,7 +420,7 @@
D_ASSERT(mdev->al_tr_pos < MD_AL_MAX_SIZE);
mdev->al_tr_number++;
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
complete(&((struct update_al_work *)w)->event);
put_ldev(mdev);
@@ -443,8 +489,9 @@
/* lock out all other meta data io for now,
* and make sure the page is mapped.
*/
- mutex_lock(&mdev->md_io_mutex);
- buffer = page_address(mdev->md_io_page);
+ buffer = drbd_md_get_buffer(mdev);
+ if (!buffer)
+ return 0;
/* Find the valid transaction in the log */
for (i = 0; i <= mx; i++) {
@@ -452,7 +499,7 @@
if (rv == 0)
continue;
if (rv == -1) {
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
return 0;
}
cnr = be32_to_cpu(buffer->tr_number);
@@ -478,7 +525,7 @@
if (!found_valid) {
dev_warn(DEV, "No usable activity log found.\n");
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
return 1;
}
@@ -493,7 +540,7 @@
rv = drbd_al_read_tr(mdev, bdev, buffer, i);
ERR_IF(rv == 0) goto cancel;
if (rv == -1) {
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
return 0;
}
@@ -534,7 +581,7 @@
mdev->al_tr_pos = 0;
/* ok, we are done with it */
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
dev_info(DEV, "Found %d transactions (%d active extents) in activity log.\n",
transactions, active_extents);
@@ -671,16 +718,20 @@
else
ext->rs_failed += count;
if (ext->rs_left < ext->rs_failed) {
- dev_err(DEV, "BAD! sector=%llus enr=%u rs_left=%d "
- "rs_failed=%d count=%d\n",
+ dev_warn(DEV, "BAD! sector=%llus enr=%u rs_left=%d "
+ "rs_failed=%d count=%d cstate=%s\n",
(unsigned long long)sector,
ext->lce.lc_number, ext->rs_left,
- ext->rs_failed, count);
- dump_stack();
+ ext->rs_failed, count,
+ drbd_conn_str(mdev->state.conn));
- lc_put(mdev->resync, &ext->lce);
- drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
- return;
+ /* We don't expect to be able to clear more bits
+ * than have been set when we originally counted
+ * the set bits to cache that value in ext->rs_left.
+ * Whatever the reason (disconnect during resync,
+ * delayed local completion of an application write),
+ * try to fix it up by recounting here. */
+ ext->rs_left = drbd_bm_e_weight(mdev, enr);
}
} else {
/* Normally this element should be in the cache,
@@ -1192,6 +1243,7 @@
put_ldev(mdev);
}
spin_unlock_irq(&mdev->al_lock);
+ wake_up(&mdev->al_wait);
return 0;
}
diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c
index 3030201..b5c5ff5 100644
--- a/drivers/block/drbd/drbd_bitmap.c
+++ b/drivers/block/drbd/drbd_bitmap.c
@@ -205,7 +205,7 @@
static void bm_store_page_idx(struct page *page, unsigned long idx)
{
BUG_ON(0 != (idx & ~BM_PAGE_IDX_MASK));
- page_private(page) |= idx;
+ set_page_private(page, idx);
}
static unsigned long bm_page_to_idx(struct page *page)
@@ -886,12 +886,21 @@
struct bm_aio_ctx {
struct drbd_conf *mdev;
atomic_t in_flight;
- struct completion done;
+ unsigned int done;
unsigned flags;
#define BM_AIO_COPY_PAGES 1
int error;
+ struct kref kref;
};
+static void bm_aio_ctx_destroy(struct kref *kref)
+{
+ struct bm_aio_ctx *ctx = container_of(kref, struct bm_aio_ctx, kref);
+
+ put_ldev(ctx->mdev);
+ kfree(ctx);
+}
+
/* bv_page may be a copy, or may be the original */
static void bm_async_io_complete(struct bio *bio, int error)
{
@@ -930,20 +939,21 @@
bm_page_unlock_io(mdev, idx);
- /* FIXME give back to page pool */
if (ctx->flags & BM_AIO_COPY_PAGES)
- put_page(bio->bi_io_vec[0].bv_page);
+ mempool_free(bio->bi_io_vec[0].bv_page, drbd_md_io_page_pool);
bio_put(bio);
- if (atomic_dec_and_test(&ctx->in_flight))
- complete(&ctx->done);
+ if (atomic_dec_and_test(&ctx->in_flight)) {
+ ctx->done = 1;
+ wake_up(&mdev->misc_wait);
+ kref_put(&ctx->kref, &bm_aio_ctx_destroy);
+ }
}
static void bm_page_io_async(struct bm_aio_ctx *ctx, int page_nr, int rw) __must_hold(local)
{
- /* we are process context. we always get a bio */
- struct bio *bio = bio_alloc(GFP_KERNEL, 1);
+ struct bio *bio = bio_alloc_drbd(GFP_NOIO);
struct drbd_conf *mdev = ctx->mdev;
struct drbd_bitmap *b = mdev->bitmap;
struct page *page;
@@ -966,10 +976,8 @@
bm_set_page_unchanged(b->bm_pages[page_nr]);
if (ctx->flags & BM_AIO_COPY_PAGES) {
- /* FIXME alloc_page is good enough for now, but actually needs
- * to use pre-allocated page pool */
void *src, *dest;
- page = alloc_page(__GFP_HIGHMEM|__GFP_WAIT);
+ page = mempool_alloc(drbd_md_io_page_pool, __GFP_HIGHMEM|__GFP_WAIT);
dest = kmap_atomic(page);
src = kmap_atomic(b->bm_pages[page_nr]);
memcpy(dest, src, PAGE_SIZE);
@@ -981,6 +989,8 @@
bio->bi_bdev = mdev->ldev->md_bdev;
bio->bi_sector = on_disk_sector;
+ /* bio_add_page of a single page to an empty bio will always succeed,
+ * according to api. Do we want to assert that? */
bio_add_page(bio, page, len, 0);
bio->bi_private = ctx;
bio->bi_end_io = bm_async_io_complete;
@@ -999,14 +1009,9 @@
/*
* bm_rw: read/write the whole bitmap from/to its on disk location.
*/
-static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_idx) __must_hold(local)
+static int bm_rw(struct drbd_conf *mdev, int rw, unsigned flags, unsigned lazy_writeout_upper_idx) __must_hold(local)
{
- struct bm_aio_ctx ctx = {
- .mdev = mdev,
- .in_flight = ATOMIC_INIT(1),
- .done = COMPLETION_INITIALIZER_ONSTACK(ctx.done),
- .flags = lazy_writeout_upper_idx ? BM_AIO_COPY_PAGES : 0,
- };
+ struct bm_aio_ctx *ctx;
struct drbd_bitmap *b = mdev->bitmap;
int num_pages, i, count = 0;
unsigned long now;
@@ -1021,7 +1026,27 @@
* For lazy writeout, we don't care for ongoing changes to the bitmap,
* as we submit copies of pages anyways.
*/
- if (!ctx.flags)
+
+ ctx = kmalloc(sizeof(struct bm_aio_ctx), GFP_NOIO);
+ if (!ctx)
+ return -ENOMEM;
+
+ *ctx = (struct bm_aio_ctx) {
+ .mdev = mdev,
+ .in_flight = ATOMIC_INIT(1),
+ .done = 0,
+ .flags = flags,
+ .error = 0,
+ .kref = { ATOMIC_INIT(2) },
+ };
+
+ if (!get_ldev_if_state(mdev, D_ATTACHING)) { /* put is in bm_aio_ctx_destroy() */
+ dev_err(DEV, "ASSERT FAILED: get_ldev_if_state() == 1 in bm_rw()\n");
+ kfree(ctx);
+ return -ENODEV;
+ }
+
+ if (!ctx->flags)
WARN_ON(!(BM_LOCKED_MASK & b->bm_flags));
num_pages = b->bm_number_of_pages;
@@ -1046,29 +1071,38 @@
continue;
}
}
- atomic_inc(&ctx.in_flight);
- bm_page_io_async(&ctx, i, rw);
+ atomic_inc(&ctx->in_flight);
+ bm_page_io_async(ctx, i, rw);
++count;
cond_resched();
}
/*
- * We initialize ctx.in_flight to one to make sure bm_async_io_complete
- * will not complete() early, and decrement / test it here. If there
+ * We initialize ctx->in_flight to one to make sure bm_async_io_complete
+ * will not set ctx->done early, and decrement / test it here. If there
* are still some bios in flight, we need to wait for them here.
+ * If all IO is done already (or nothing had been submitted), there is
+ * no need to wait. Still, we need to put the kref associated with the
+ * "in_flight reached zero, all done" event.
*/
- if (!atomic_dec_and_test(&ctx.in_flight))
- wait_for_completion(&ctx.done);
+ if (!atomic_dec_and_test(&ctx->in_flight))
+ wait_until_done_or_disk_failure(mdev, mdev->ldev, &ctx->done);
+ else
+ kref_put(&ctx->kref, &bm_aio_ctx_destroy);
+
dev_info(DEV, "bitmap %s of %u pages took %lu jiffies\n",
rw == WRITE ? "WRITE" : "READ",
count, jiffies - now);
- if (ctx.error) {
+ if (ctx->error) {
dev_alert(DEV, "we had at least one MD IO ERROR during bitmap IO\n");
drbd_chk_io_error(mdev, 1, true);
- err = -EIO; /* ctx.error ? */
+ err = -EIO; /* ctx->error ? */
}
+ if (atomic_read(&ctx->in_flight))
+ err = -EIO; /* Disk failed during IO... */
+
now = jiffies;
if (rw == WRITE) {
drbd_md_flush(mdev);
@@ -1082,6 +1116,7 @@
dev_info(DEV, "%s (%lu bits) marked out-of-sync by on disk bit-map.\n",
ppsize(ppb, now << (BM_BLOCK_SHIFT-10)), now);
+ kref_put(&ctx->kref, &bm_aio_ctx_destroy);
return err;
}
@@ -1091,7 +1126,7 @@
*/
int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local)
{
- return bm_rw(mdev, READ, 0);
+ return bm_rw(mdev, READ, 0, 0);
}
/**
@@ -1102,7 +1137,7 @@
*/
int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local)
{
- return bm_rw(mdev, WRITE, 0);
+ return bm_rw(mdev, WRITE, 0, 0);
}
/**
@@ -1112,7 +1147,23 @@
*/
int drbd_bm_write_lazy(struct drbd_conf *mdev, unsigned upper_idx) __must_hold(local)
{
- return bm_rw(mdev, WRITE, upper_idx);
+ return bm_rw(mdev, WRITE, BM_AIO_COPY_PAGES, upper_idx);
+}
+
+/**
+ * drbd_bm_write_copy_pages() - Write the whole bitmap to its on disk location.
+ * @mdev: DRBD device.
+ *
+ * Will only write pages that have changed since last IO.
+ * In contrast to drbd_bm_write(), this will copy the bitmap pages
+ * to temporary writeout pages. It is intended to trigger a full write-out
+ * while still allowing the bitmap to change, for example if a resync or online
+ * verify is aborted due to a failed peer disk, while local IO continues, or
+ * pending resync acks are still being processed.
+ */
+int drbd_bm_write_copy_pages(struct drbd_conf *mdev) __must_hold(local)
+{
+ return bm_rw(mdev, WRITE, BM_AIO_COPY_PAGES, 0);
}
@@ -1130,28 +1181,45 @@
*/
int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local)
{
- struct bm_aio_ctx ctx = {
- .mdev = mdev,
- .in_flight = ATOMIC_INIT(1),
- .done = COMPLETION_INITIALIZER_ONSTACK(ctx.done),
- .flags = BM_AIO_COPY_PAGES,
- };
+ struct bm_aio_ctx *ctx;
+ int err;
if (bm_test_page_unchanged(mdev->bitmap->bm_pages[idx])) {
dynamic_dev_dbg(DEV, "skipped bm page write for idx %u\n", idx);
return 0;
}
- bm_page_io_async(&ctx, idx, WRITE_SYNC);
- wait_for_completion(&ctx.done);
+ ctx = kmalloc(sizeof(struct bm_aio_ctx), GFP_NOIO);
+ if (!ctx)
+ return -ENOMEM;
- if (ctx.error)
+ *ctx = (struct bm_aio_ctx) {
+ .mdev = mdev,
+ .in_flight = ATOMIC_INIT(1),
+ .done = 0,
+ .flags = BM_AIO_COPY_PAGES,
+ .error = 0,
+ .kref = { ATOMIC_INIT(2) },
+ };
+
+ if (!get_ldev_if_state(mdev, D_ATTACHING)) { /* put is in bm_aio_ctx_destroy() */
+ dev_err(DEV, "ASSERT FAILED: get_ldev_if_state() == 1 in drbd_bm_write_page()\n");
+ kfree(ctx);
+ return -ENODEV;
+ }
+
+ bm_page_io_async(ctx, idx, WRITE_SYNC);
+ wait_until_done_or_disk_failure(mdev, mdev->ldev, &ctx->done);
+
+ if (ctx->error)
drbd_chk_io_error(mdev, 1, true);
/* that should force detach, so the in memory bitmap will be
* gone in a moment as well. */
mdev->bm_writ_cnt++;
- return ctx.error;
+ err = atomic_read(&ctx->in_flight) ? -EIO : ctx->error;
+ kref_put(&ctx->kref, &bm_aio_ctx_destroy);
+ return err;
}
/* NOTE
diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 8d68056..02f013a 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -712,7 +712,6 @@
struct list_head tl_requests; /* ring list in the transfer log */
struct bio *master_bio; /* master bio pointer */
unsigned long rq_state; /* see comments above _req_mod() */
- int seq_num;
unsigned long start_time;
};
@@ -851,6 +850,7 @@
NEW_CUR_UUID, /* Create new current UUID when thawing IO */
AL_SUSPENDED, /* Activity logging is currently suspended. */
AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
+ STATE_SENT, /* Do not change state/UUIDs while this is set */
};
struct drbd_bitmap; /* opaque for drbd_conf */
@@ -862,31 +862,30 @@
BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
/* currently locked for bulk operation */
- BM_LOCKED_MASK = 0x7,
+ BM_LOCKED_MASK = 0xf,
/* in detail, that is: */
BM_DONT_CLEAR = 0x1,
BM_DONT_SET = 0x2,
BM_DONT_TEST = 0x4,
+ /* so we can mark it locked for bulk operation,
+ * and still allow all non-bulk operations */
+ BM_IS_LOCKED = 0x8,
+
/* (test bit, count bit) allowed (common case) */
- BM_LOCKED_TEST_ALLOWED = 0x3,
+ BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
/* testing bits, as well as setting new bits allowed, but clearing bits
* would be unexpected. Used during bitmap receive. Setting new bits
* requires sending of "out-of-sync" information, though. */
- BM_LOCKED_SET_ALLOWED = 0x1,
+ BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
- /* clear is not expected while bitmap is locked for bulk operation */
+ /* for drbd_bm_write_copy_pages, everything is allowed,
+ * only concurrent bulk operations are locked out. */
+ BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
};
-
-/* TODO sort members for performance
- * MAYBE group them further */
-
-/* THINK maybe we actually want to use the default "event/%s" worker threads
- * or similar in linux 2.6, which uses per cpu data and threads.
- */
struct drbd_work_queue {
struct list_head q;
struct semaphore s; /* producers up it, worker down()s it */
@@ -938,8 +937,7 @@
};
struct drbd_md_io {
- struct drbd_conf *mdev;
- struct completion event;
+ unsigned int done;
int error;
};
@@ -1022,6 +1020,7 @@
struct drbd_tl_epoch *newest_tle;
struct drbd_tl_epoch *oldest_tle;
struct list_head out_of_sequence_requests;
+ struct list_head barrier_acked_requests;
struct hlist_head *tl_hash;
unsigned int tl_hash_s;
@@ -1056,6 +1055,8 @@
struct crypto_hash *csums_tfm;
struct crypto_hash *verify_tfm;
+ unsigned long last_reattach_jif;
+ unsigned long last_reconnect_jif;
struct drbd_thread receiver;
struct drbd_thread worker;
struct drbd_thread asender;
@@ -1094,7 +1095,8 @@
wait_queue_head_t ee_wait;
struct page *md_io_page; /* one page buffer for md_io */
struct page *md_io_tmpp; /* for logical_block_size != 512 */
- struct mutex md_io_mutex; /* protects the md_io_buffer */
+ struct drbd_md_io md_io;
+ atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */
spinlock_t al_lock;
wait_queue_head_t al_wait;
struct lru_cache *act_log; /* activity log */
@@ -1228,8 +1230,8 @@
extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
-extern int _drbd_send_state(struct drbd_conf *mdev);
-extern int drbd_send_state(struct drbd_conf *mdev);
+extern int drbd_send_state(struct drbd_conf *mdev, union drbd_state s);
+extern int drbd_send_current_state(struct drbd_conf *mdev);
extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
enum drbd_packets cmd, struct p_header80 *h,
size_t size, unsigned msg_flags);
@@ -1461,6 +1463,7 @@
extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
+extern int drbd_bm_write_copy_pages(struct drbd_conf *mdev) __must_hold(local);
extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
unsigned long al_enr);
extern size_t drbd_bm_words(struct drbd_conf *mdev);
@@ -1493,11 +1496,38 @@
extern mempool_t *drbd_request_mempool;
extern mempool_t *drbd_ee_mempool;
-extern struct page *drbd_pp_pool; /* drbd's page pool */
+/* drbd's page pool, used to buffer data received from the peer,
+ * or data requested by the peer.
+ *
+ * This does not have an emergency reserve.
+ *
+ * When allocating from this pool, it first takes pages from the pool.
+ * Only if the pool is depleted will try to allocate from the system.
+ *
+ * The assumption is that pages taken from this pool will be processed,
+ * and given back, "quickly", and then can be recycled, so we can avoid
+ * frequent calls to alloc_page(), and still will be able to make progress even
+ * under memory pressure.
+ */
+extern struct page *drbd_pp_pool;
extern spinlock_t drbd_pp_lock;
extern int drbd_pp_vacant;
extern wait_queue_head_t drbd_pp_wait;
+/* We also need a standard (emergency-reserve backed) page pool
+ * for meta data IO (activity log, bitmap).
+ * We can keep it global, as long as it is used as "N pages at a time".
+ * 128 should be plenty, currently we probably can get away with as few as 1.
+ */
+#define DRBD_MIN_POOL_PAGES 128
+extern mempool_t *drbd_md_io_page_pool;
+
+/* We also need to make sure we get a bio
+ * when we need it for housekeeping purposes */
+extern struct bio_set *drbd_md_io_bio_set;
+/* to allocate from that set */
+extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
+
extern rwlock_t global_state_lock;
extern struct drbd_conf *drbd_new_device(unsigned int minor);
@@ -1536,8 +1566,12 @@
extern void suspend_other_sg(struct drbd_conf *mdev);
extern int drbd_resync_finished(struct drbd_conf *mdev);
/* maybe rather drbd_main.c ? */
+extern void *drbd_md_get_buffer(struct drbd_conf *mdev);
+extern void drbd_md_put_buffer(struct drbd_conf *mdev);
extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
- struct drbd_backing_dev *bdev, sector_t sector, int rw);
+ struct drbd_backing_dev *bdev, sector_t sector, int rw);
+extern void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
+ unsigned int *done);
extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
@@ -1754,19 +1788,6 @@
#define page_chain_for_each_safe(page, n) \
for (; page && ({ n = page_chain_next(page); 1; }); page = n)
-static inline int drbd_bio_has_active_page(struct bio *bio)
-{
- struct bio_vec *bvec;
- int i;
-
- __bio_for_each_segment(bvec, bio, i, 0) {
- if (page_count(bvec->bv_page) > 1)
- return 1;
- }
-
- return 0;
-}
-
static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
{
struct page *page = e->pages;
@@ -1777,7 +1798,6 @@
return 0;
}
-
static inline void drbd_state_lock(struct drbd_conf *mdev)
{
wait_event(mdev->misc_wait,
@@ -2230,7 +2250,7 @@
* Note: currently we don't support such large bitmaps on 32bit
* arch anyways, but no harm done to be prepared for it here.
*/
- unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10;
+ unsigned int shift = mdev->rs_total > UINT_MAX ? 16 : 10;
unsigned long left = *bits_left >> shift;
unsigned long total = 1UL + (mdev->rs_total >> shift);
unsigned long tmp = 1000UL - left * 1000UL/total;
@@ -2306,12 +2326,12 @@
case D_OUTDATED:
case D_CONSISTENT:
case D_UP_TO_DATE:
+ case D_FAILED:
/* disk state is stable as well. */
break;
/* no new io accepted during tansitional states */
case D_ATTACHING:
- case D_FAILED:
case D_NEGOTIATING:
case D_UNKNOWN:
case D_MASK:
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index 211fc44..920ede2 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -139,6 +139,8 @@
struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
mempool_t *drbd_request_mempool;
mempool_t *drbd_ee_mempool;
+mempool_t *drbd_md_io_page_pool;
+struct bio_set *drbd_md_io_bio_set;
/* I do not use a standard mempool, because:
1) I want to hand out the pre-allocated objects first.
@@ -159,7 +161,24 @@
.release = drbd_release,
};
-#define ARRY_SIZE(A) (sizeof(A)/sizeof(A[0]))
+static void bio_destructor_drbd(struct bio *bio)
+{
+ bio_free(bio, drbd_md_io_bio_set);
+}
+
+struct bio *bio_alloc_drbd(gfp_t gfp_mask)
+{
+ struct bio *bio;
+
+ if (!drbd_md_io_bio_set)
+ return bio_alloc(gfp_mask, 1);
+
+ bio = bio_alloc_bioset(gfp_mask, 1, drbd_md_io_bio_set);
+ if (!bio)
+ return NULL;
+ bio->bi_destructor = bio_destructor_drbd;
+ return bio;
+}
#ifdef __CHECKER__
/* When checking with sparse, and this is an inline function, sparse will
@@ -208,6 +227,7 @@
mdev->oldest_tle = b;
mdev->newest_tle = b;
INIT_LIST_HEAD(&mdev->out_of_sequence_requests);
+ INIT_LIST_HEAD(&mdev->barrier_acked_requests);
mdev->tl_hash = NULL;
mdev->tl_hash_s = 0;
@@ -246,9 +266,7 @@
new->n_writes = 0;
newest_before = mdev->newest_tle;
- /* never send a barrier number == 0, because that is special-cased
- * when using TCQ for our write ordering code */
- new->br_number = (newest_before->br_number+1) ?: 1;
+ new->br_number = newest_before->br_number+1;
if (mdev->newest_tle != new) {
mdev->newest_tle->next = new;
mdev->newest_tle = new;
@@ -311,7 +329,7 @@
These have been list_move'd to the out_of_sequence_requests list in
_req_mod(, barrier_acked) above.
*/
- list_del_init(&b->requests);
+ list_splice_init(&b->requests, &mdev->barrier_acked_requests);
nob = b->next;
if (test_and_clear_bit(CREATE_BARRIER, &mdev->flags)) {
@@ -411,6 +429,23 @@
b = tmp;
list_splice(&carry_reads, &b->requests);
}
+
+ /* Actions operating on the disk state, also want to work on
+ requests that got barrier acked. */
+ switch (what) {
+ case fail_frozen_disk_io:
+ case restart_frozen_disk_io:
+ list_for_each_safe(le, tle, &mdev->barrier_acked_requests) {
+ req = list_entry(le, struct drbd_request, tl_requests);
+ _req_mod(req, what);
+ }
+
+ case connection_lost_while_pending:
+ case resend:
+ break;
+ default:
+ dev_err(DEV, "what = %d in _tl_restart()\n", what);
+ }
}
@@ -458,6 +493,38 @@
}
/**
+ * tl_abort_disk_io() - Abort disk I/O for all requests for a certain mdev in the TL
+ * @mdev: DRBD device.
+ */
+void tl_abort_disk_io(struct drbd_conf *mdev)
+{
+ struct drbd_tl_epoch *b;
+ struct list_head *le, *tle;
+ struct drbd_request *req;
+
+ spin_lock_irq(&mdev->req_lock);
+ b = mdev->oldest_tle;
+ while (b) {
+ list_for_each_safe(le, tle, &b->requests) {
+ req = list_entry(le, struct drbd_request, tl_requests);
+ if (!(req->rq_state & RQ_LOCAL_PENDING))
+ continue;
+ _req_mod(req, abort_disk_io);
+ }
+ b = b->next;
+ }
+
+ list_for_each_safe(le, tle, &mdev->barrier_acked_requests) {
+ req = list_entry(le, struct drbd_request, tl_requests);
+ if (!(req->rq_state & RQ_LOCAL_PENDING))
+ continue;
+ _req_mod(req, abort_disk_io);
+ }
+
+ spin_unlock_irq(&mdev->req_lock);
+}
+
+/**
* cl_wide_st_chg() - true if the state change is a cluster wide one
* @mdev: DRBD device.
* @os: old (current) state.
@@ -470,7 +537,7 @@
((os.role != R_PRIMARY && ns.role == R_PRIMARY) ||
(os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) ||
(os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) ||
- (os.disk != D_DISKLESS && ns.disk == D_DISKLESS))) ||
+ (os.disk != D_FAILED && ns.disk == D_FAILED))) ||
(os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) ||
(os.conn == C_CONNECTED && ns.conn == C_VERIFY_S);
}
@@ -509,8 +576,16 @@
static enum drbd_state_rv is_valid_state_transition(struct drbd_conf *,
union drbd_state,
union drbd_state);
+enum sanitize_state_warnings {
+ NO_WARNING,
+ ABORTED_ONLINE_VERIFY,
+ ABORTED_RESYNC,
+ CONNECTION_LOST_NEGOTIATING,
+ IMPLICITLY_UPGRADED_DISK,
+ IMPLICITLY_UPGRADED_PDSK,
+};
static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os,
- union drbd_state ns, const char **warn_sync_abort);
+ union drbd_state ns, enum sanitize_state_warnings *warn);
int drbd_send_state_req(struct drbd_conf *,
union drbd_state, union drbd_state);
@@ -785,6 +860,13 @@
if (ns.conn == os.conn && ns.conn == C_WF_REPORT_PARAMS)
rv = SS_IN_TRANSIENT_STATE;
+ /* While establishing a connection only allow cstate to change.
+ Delay/refuse role changes, detach attach etc... */
+ if (test_bit(STATE_SENT, &mdev->flags) &&
+ !(os.conn == C_WF_REPORT_PARAMS ||
+ (ns.conn == C_WF_REPORT_PARAMS && os.conn == C_WF_CONNECTION)))
+ rv = SS_IN_TRANSIENT_STATE;
+
if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED)
rv = SS_NEED_CONNECTION;
@@ -803,6 +885,21 @@
return rv;
}
+static void print_sanitize_warnings(struct drbd_conf *mdev, enum sanitize_state_warnings warn)
+{
+ static const char *msg_table[] = {
+ [NO_WARNING] = "",
+ [ABORTED_ONLINE_VERIFY] = "Online-verify aborted.",
+ [ABORTED_RESYNC] = "Resync aborted.",
+ [CONNECTION_LOST_NEGOTIATING] = "Connection lost while negotiating, no data!",
+ [IMPLICITLY_UPGRADED_DISK] = "Implicitly upgraded disk",
+ [IMPLICITLY_UPGRADED_PDSK] = "Implicitly upgraded pdsk",
+ };
+
+ if (warn != NO_WARNING)
+ dev_warn(DEV, "%s\n", msg_table[warn]);
+}
+
/**
* sanitize_state() - Resolves implicitly necessary additional changes to a state transition
* @mdev: DRBD device.
@@ -814,11 +911,14 @@
* to D_UNKNOWN. This rule and many more along those lines are in this function.
*/
static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os,
- union drbd_state ns, const char **warn_sync_abort)
+ union drbd_state ns, enum sanitize_state_warnings *warn)
{
enum drbd_fencing_p fp;
enum drbd_disk_state disk_min, disk_max, pdsk_min, pdsk_max;
+ if (warn)
+ *warn = NO_WARNING;
+
fp = FP_DONT_CARE;
if (get_ldev(mdev)) {
fp = mdev->ldev->dc.fencing;
@@ -833,18 +933,13 @@
/* After a network error (+C_TEAR_DOWN) only C_UNCONNECTED or C_DISCONNECTING can follow.
* If you try to go into some Sync* state, that shall fail (elsewhere). */
if (os.conn >= C_TIMEOUT && os.conn <= C_TEAR_DOWN &&
- ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING && ns.conn <= C_TEAR_DOWN)
+ ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING && ns.conn <= C_CONNECTED)
ns.conn = os.conn;
/* we cannot fail (again) if we already detached */
if (ns.disk == D_FAILED && os.disk == D_DISKLESS)
ns.disk = D_DISKLESS;
- /* if we are only D_ATTACHING yet,
- * we can (and should) go directly to D_DISKLESS. */
- if (ns.disk == D_FAILED && os.disk == D_ATTACHING)
- ns.disk = D_DISKLESS;
-
/* After C_DISCONNECTING only C_STANDALONE may follow */
if (os.conn == C_DISCONNECTING && ns.conn != C_STANDALONE)
ns.conn = os.conn;
@@ -863,10 +958,9 @@
/* Abort resync if a disk fails/detaches */
if (os.conn > C_CONNECTED && ns.conn > C_CONNECTED &&
(ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) {
- if (warn_sync_abort)
- *warn_sync_abort =
- os.conn == C_VERIFY_S || os.conn == C_VERIFY_T ?
- "Online-verify" : "Resync";
+ if (warn)
+ *warn = os.conn == C_VERIFY_S || os.conn == C_VERIFY_T ?
+ ABORTED_ONLINE_VERIFY : ABORTED_RESYNC;
ns.conn = C_CONNECTED;
}
@@ -877,7 +971,8 @@
ns.disk = mdev->new_state_tmp.disk;
ns.pdsk = mdev->new_state_tmp.pdsk;
} else {
- dev_alert(DEV, "Connection lost while negotiating, no data!\n");
+ if (warn)
+ *warn = CONNECTION_LOST_NEGOTIATING;
ns.disk = D_DISKLESS;
ns.pdsk = D_UNKNOWN;
}
@@ -959,16 +1054,16 @@
ns.disk = disk_max;
if (ns.disk < disk_min) {
- dev_warn(DEV, "Implicitly set disk from %s to %s\n",
- drbd_disk_str(ns.disk), drbd_disk_str(disk_min));
+ if (warn)
+ *warn = IMPLICITLY_UPGRADED_DISK;
ns.disk = disk_min;
}
if (ns.pdsk > pdsk_max)
ns.pdsk = pdsk_max;
if (ns.pdsk < pdsk_min) {
- dev_warn(DEV, "Implicitly set pdsk from %s to %s\n",
- drbd_disk_str(ns.pdsk), drbd_disk_str(pdsk_min));
+ if (warn)
+ *warn = IMPLICITLY_UPGRADED_PDSK;
ns.pdsk = pdsk_min;
}
@@ -1045,12 +1140,12 @@
{
union drbd_state os;
enum drbd_state_rv rv = SS_SUCCESS;
- const char *warn_sync_abort = NULL;
+ enum sanitize_state_warnings ssw;
struct after_state_chg_work *ascw;
os = mdev->state;
- ns = sanitize_state(mdev, os, ns, &warn_sync_abort);
+ ns = sanitize_state(mdev, os, ns, &ssw);
if (ns.i == os.i)
return SS_NOTHING_TO_DO;
@@ -1076,8 +1171,7 @@
return rv;
}
- if (warn_sync_abort)
- dev_warn(DEV, "%s aborted.\n", warn_sync_abort);
+ print_sanitize_warnings(mdev, ssw);
{
char *pbp, pb[300];
@@ -1243,7 +1337,7 @@
drbd_thread_stop_nowait(&mdev->receiver);
/* Upon network failure, we need to restart the receiver. */
- if (os.conn > C_TEAR_DOWN &&
+ if (os.conn > C_WF_CONNECTION &&
ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT)
drbd_thread_restart_nowait(&mdev->receiver);
@@ -1251,6 +1345,15 @@
if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED)
drbd_resume_al(mdev);
+ /* remember last connect and attach times so request_timer_fn() won't
+ * kill newly established sessions while we are still trying to thaw
+ * previously frozen IO */
+ if (os.conn != C_WF_REPORT_PARAMS && ns.conn == C_WF_REPORT_PARAMS)
+ mdev->last_reconnect_jif = jiffies;
+ if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) &&
+ ns.disk > D_NEGOTIATING)
+ mdev->last_reattach_jif = jiffies;
+
ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC);
if (ascw) {
ascw->os = os;
@@ -1354,12 +1457,16 @@
/* Here we have the actions that are performed after a
state change. This function might sleep */
+ if (os.disk <= D_NEGOTIATING && ns.disk > D_NEGOTIATING)
+ mod_timer(&mdev->request_timer, jiffies + HZ);
+
nsm.i = -1;
if (ns.susp_nod) {
if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED)
what = resend;
- if (os.disk == D_ATTACHING && ns.disk > D_ATTACHING)
+ if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) &&
+ ns.disk > D_NEGOTIATING)
what = restart_frozen_disk_io;
if (what != nothing)
@@ -1408,7 +1515,7 @@
/* Do not change the order of the if above and the two below... */
if (os.pdsk == D_DISKLESS && ns.pdsk > D_DISKLESS) { /* attach on the peer */
drbd_send_uuids(mdev);
- drbd_send_state(mdev);
+ drbd_send_state(mdev, ns);
}
/* No point in queuing send_bitmap if we don't have a connection
* anymore, so check also the _current_ state, not only the new state
@@ -1441,11 +1548,11 @@
}
if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) {
- if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0) {
+ if (os.peer == R_SECONDARY && ns.peer == R_PRIMARY &&
+ mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) {
drbd_uuid_new_current(mdev);
drbd_send_uuids(mdev);
}
-
/* D_DISKLESS Peer becomes secondary */
if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY)
/* We may still be Primary ourselves.
@@ -1473,14 +1580,14 @@
os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) {
drbd_send_sizes(mdev, 0, 0); /* to start sync... */
drbd_send_uuids(mdev);
- drbd_send_state(mdev);
+ drbd_send_state(mdev, ns);
}
/* We want to pause/continue resync, tell peer. */
if (ns.conn >= C_CONNECTED &&
((os.aftr_isp != ns.aftr_isp) ||
(os.user_isp != ns.user_isp)))
- drbd_send_state(mdev);
+ drbd_send_state(mdev, ns);
/* In case one of the isp bits got set, suspend other devices. */
if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) &&
@@ -1490,10 +1597,10 @@
/* Make sure the peer gets informed about eventual state
changes (ISP bits) while we were in WFReportParams. */
if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED)
- drbd_send_state(mdev);
+ drbd_send_state(mdev, ns);
if (os.conn != C_AHEAD && ns.conn == C_AHEAD)
- drbd_send_state(mdev);
+ drbd_send_state(mdev, ns);
/* We are in the progress to start a full sync... */
if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) ||
@@ -1513,33 +1620,38 @@
/* first half of local IO error, failure to attach,
* or administrative detach */
if (os.disk != D_FAILED && ns.disk == D_FAILED) {
- enum drbd_io_error_p eh;
- int was_io_error;
+ enum drbd_io_error_p eh = EP_PASS_ON;
+ int was_io_error = 0;
/* corresponding get_ldev was in __drbd_set_state, to serialize
- * our cleanup here with the transition to D_DISKLESS,
- * so it is safe to dreference ldev here. */
- eh = mdev->ldev->dc.on_io_error;
- was_io_error = test_and_clear_bit(WAS_IO_ERROR, &mdev->flags);
+ * our cleanup here with the transition to D_DISKLESS.
+ * But is is still not save to dreference ldev here, since
+ * we might come from an failed Attach before ldev was set. */
+ if (mdev->ldev) {
+ eh = mdev->ldev->dc.on_io_error;
+ was_io_error = test_and_clear_bit(WAS_IO_ERROR, &mdev->flags);
- /* current state still has to be D_FAILED,
- * there is only one way out: to D_DISKLESS,
- * and that may only happen after our put_ldev below. */
- if (mdev->state.disk != D_FAILED)
- dev_err(DEV,
- "ASSERT FAILED: disk is %s during detach\n",
- drbd_disk_str(mdev->state.disk));
+ /* Immediately allow completion of all application IO, that waits
+ for completion from the local disk. */
+ tl_abort_disk_io(mdev);
- if (drbd_send_state(mdev))
- dev_warn(DEV, "Notified peer that I am detaching my disk\n");
- else
- dev_err(DEV, "Sending state for detaching disk failed\n");
+ /* current state still has to be D_FAILED,
+ * there is only one way out: to D_DISKLESS,
+ * and that may only happen after our put_ldev below. */
+ if (mdev->state.disk != D_FAILED)
+ dev_err(DEV,
+ "ASSERT FAILED: disk is %s during detach\n",
+ drbd_disk_str(mdev->state.disk));
- drbd_rs_cancel_all(mdev);
+ if (ns.conn >= C_CONNECTED)
+ drbd_send_state(mdev, ns);
- /* In case we want to get something to stable storage still,
- * this may be the last chance.
- * Following put_ldev may transition to D_DISKLESS. */
- drbd_md_sync(mdev);
+ drbd_rs_cancel_all(mdev);
+
+ /* In case we want to get something to stable storage still,
+ * this may be the last chance.
+ * Following put_ldev may transition to D_DISKLESS. */
+ drbd_md_sync(mdev);
+ }
put_ldev(mdev);
if (was_io_error && eh == EP_CALL_HELPER)
@@ -1561,16 +1673,17 @@
mdev->rs_failed = 0;
atomic_set(&mdev->rs_pending_cnt, 0);
- if (drbd_send_state(mdev))
- dev_warn(DEV, "Notified peer that I'm now diskless.\n");
+ if (ns.conn >= C_CONNECTED)
+ drbd_send_state(mdev, ns);
+
/* corresponding get_ldev in __drbd_set_state
* this may finally trigger drbd_ldev_destroy. */
put_ldev(mdev);
}
/* Notify peer that I had a local IO error, and did not detached.. */
- if (os.disk == D_UP_TO_DATE && ns.disk == D_INCONSISTENT)
- drbd_send_state(mdev);
+ if (os.disk == D_UP_TO_DATE && ns.disk == D_INCONSISTENT && ns.conn >= C_CONNECTED)
+ drbd_send_state(mdev, ns);
/* Disks got bigger while they were detached */
if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING &&
@@ -1588,7 +1701,13 @@
/* sync target done with resync. Explicitly notify peer, even though
* it should (at least for non-empty resyncs) already know itself. */
if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED)
- drbd_send_state(mdev);
+ drbd_send_state(mdev, ns);
+
+ /* Wake up role changes, that were delayed because of connection establishing */
+ if (os.conn == C_WF_REPORT_PARAMS && ns.conn != C_WF_REPORT_PARAMS) {
+ clear_bit(STATE_SENT, &mdev->flags);
+ wake_up(&mdev->state_wait);
+ }
/* This triggers bitmap writeout of potentially still unwritten pages
* if the resync finished cleanly, or aborted because of peer disk
@@ -1598,8 +1717,8 @@
* No harm done if some bits change during this phase.
*/
if (os.conn > C_CONNECTED && ns.conn <= C_CONNECTED && get_ldev(mdev)) {
- drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL,
- "write from resync_finished", BM_LOCKED_SET_ALLOWED);
+ drbd_queue_bitmap_io(mdev, &drbd_bm_write_copy_pages, NULL,
+ "write from resync_finished", BM_LOCKED_CHANGE_ALLOWED);
put_ldev(mdev);
}
@@ -2057,7 +2176,11 @@
D_ASSERT(mdev->state.disk == D_UP_TO_DATE);
- uuid = mdev->ldev->md.uuid[UI_BITMAP] + UUID_NEW_BM_OFFSET;
+ uuid = mdev->ldev->md.uuid[UI_BITMAP];
+ if (uuid && uuid != UUID_JUST_CREATED)
+ uuid = uuid + UUID_NEW_BM_OFFSET;
+ else
+ get_random_bytes(&uuid, sizeof(u64));
drbd_uuid_set(mdev, UI_BITMAP, uuid);
drbd_print_uuids(mdev, "updated sync UUID");
drbd_md_sync(mdev);
@@ -2089,6 +2212,10 @@
max_bio_size = DRBD_MAX_BIO_SIZE; /* ... multiple BIOs per peer_request */
}
+ /* Never allow old drbd (up to 8.3.7) to see more than 32KiB */
+ if (mdev->agreed_pro_version <= 94)
+ max_bio_size = min_t(int, max_bio_size, DRBD_MAX_SIZE_H80_PACKET);
+
p.d_size = cpu_to_be64(d_size);
p.u_size = cpu_to_be64(u_size);
p.c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(mdev->this_bdev));
@@ -2102,10 +2229,10 @@
}
/**
- * drbd_send_state() - Sends the drbd state to the peer
+ * drbd_send_current_state() - Sends the drbd state to the peer
* @mdev: DRBD device.
*/
-int drbd_send_state(struct drbd_conf *mdev)
+int drbd_send_current_state(struct drbd_conf *mdev)
{
struct socket *sock;
struct p_state p;
@@ -2131,6 +2258,37 @@
return ok;
}
+/**
+ * drbd_send_state() - After a state change, sends the new state to the peer
+ * @mdev: DRBD device.
+ * @state: the state to send, not necessarily the current state.
+ *
+ * Each state change queues an "after_state_ch" work, which will eventually
+ * send the resulting new state to the peer. If more state changes happen
+ * between queuing and processing of the after_state_ch work, we still
+ * want to send each intermediary state in the order it occurred.
+ */
+int drbd_send_state(struct drbd_conf *mdev, union drbd_state state)
+{
+ struct socket *sock;
+ struct p_state p;
+ int ok = 0;
+
+ mutex_lock(&mdev->data.mutex);
+
+ p.state = cpu_to_be32(state.i);
+ sock = mdev->data.socket;
+
+ if (likely(sock != NULL)) {
+ ok = _drbd_send_cmd(mdev, sock, P_STATE,
+ (struct p_header80 *)&p, sizeof(p), 0);
+ }
+
+ mutex_unlock(&mdev->data.mutex);
+
+ return ok;
+}
+
int drbd_send_state_req(struct drbd_conf *mdev,
union drbd_state mask, union drbd_state val)
{
@@ -2615,7 +2773,7 @@
struct bio_vec *bvec;
int i;
/* hint all but last page with MSG_MORE */
- __bio_for_each_segment(bvec, bio, i, 0) {
+ bio_for_each_segment(bvec, bio, i) {
if (!_drbd_no_send_page(mdev, bvec->bv_page,
bvec->bv_offset, bvec->bv_len,
i == bio->bi_vcnt -1 ? 0 : MSG_MORE))
@@ -2629,7 +2787,7 @@
struct bio_vec *bvec;
int i;
/* hint all but last page with MSG_MORE */
- __bio_for_each_segment(bvec, bio, i, 0) {
+ bio_for_each_segment(bvec, bio, i) {
if (!_drbd_send_page(mdev, bvec->bv_page,
bvec->bv_offset, bvec->bv_len,
i == bio->bi_vcnt -1 ? 0 : MSG_MORE))
@@ -2695,8 +2853,7 @@
p.sector = cpu_to_be64(req->sector);
p.block_id = (unsigned long)req;
- p.seq_num = cpu_to_be32(req->seq_num =
- atomic_add_return(1, &mdev->packet_seq));
+ p.seq_num = cpu_to_be32(atomic_add_return(1, &mdev->packet_seq));
dp_flags = bio_flags_to_wire(mdev, req->master_bio->bi_rw);
@@ -2987,8 +3144,8 @@
atomic_set(&mdev->rs_sect_in, 0);
atomic_set(&mdev->rs_sect_ev, 0);
atomic_set(&mdev->ap_in_flight, 0);
+ atomic_set(&mdev->md_io_in_use, 0);
- mutex_init(&mdev->md_io_mutex);
mutex_init(&mdev->data.mutex);
mutex_init(&mdev->meta.mutex);
sema_init(&mdev->data.work.s, 0);
@@ -3126,6 +3283,10 @@
/* D_ASSERT(atomic_read(&drbd_pp_vacant)==0); */
+ if (drbd_md_io_bio_set)
+ bioset_free(drbd_md_io_bio_set);
+ if (drbd_md_io_page_pool)
+ mempool_destroy(drbd_md_io_page_pool);
if (drbd_ee_mempool)
mempool_destroy(drbd_ee_mempool);
if (drbd_request_mempool)
@@ -3139,6 +3300,8 @@
if (drbd_al_ext_cache)
kmem_cache_destroy(drbd_al_ext_cache);
+ drbd_md_io_bio_set = NULL;
+ drbd_md_io_page_pool = NULL;
drbd_ee_mempool = NULL;
drbd_request_mempool = NULL;
drbd_ee_cache = NULL;
@@ -3162,6 +3325,8 @@
drbd_bm_ext_cache = NULL;
drbd_al_ext_cache = NULL;
drbd_pp_pool = NULL;
+ drbd_md_io_page_pool = NULL;
+ drbd_md_io_bio_set = NULL;
/* caches */
drbd_request_cache = kmem_cache_create(
@@ -3185,6 +3350,16 @@
goto Enomem;
/* mempools */
+#ifdef COMPAT_HAVE_BIOSET_CREATE
+ drbd_md_io_bio_set = bioset_create(DRBD_MIN_POOL_PAGES, 0);
+ if (drbd_md_io_bio_set == NULL)
+ goto Enomem;
+#endif
+
+ drbd_md_io_page_pool = mempool_create_page_pool(DRBD_MIN_POOL_PAGES, 0);
+ if (drbd_md_io_page_pool == NULL)
+ goto Enomem;
+
drbd_request_mempool = mempool_create(number,
mempool_alloc_slab, mempool_free_slab, drbd_request_cache);
if (drbd_request_mempool == NULL)
@@ -3262,6 +3437,8 @@
if (!mdev)
return;
+ del_timer_sync(&mdev->request_timer);
+
/* paranoia asserts */
if (mdev->open_cnt != 0)
dev_err(DEV, "open_cnt = %d in %s:%u", mdev->open_cnt,
@@ -3666,8 +3843,10 @@
if (!get_ldev_if_state(mdev, D_FAILED))
return;
- mutex_lock(&mdev->md_io_mutex);
- buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page);
+ buffer = drbd_md_get_buffer(mdev);
+ if (!buffer)
+ goto out;
+
memset(buffer, 0, 512);
buffer->la_size = cpu_to_be64(drbd_get_capacity(mdev->this_bdev));
@@ -3698,7 +3877,8 @@
* since we updated it on metadata. */
mdev->ldev->md.la_size_sect = drbd_get_capacity(mdev->this_bdev);
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
+out:
put_ldev(mdev);
}
@@ -3718,8 +3898,9 @@
if (!get_ldev_if_state(mdev, D_ATTACHING))
return ERR_IO_MD_DISK;
- mutex_lock(&mdev->md_io_mutex);
- buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page);
+ buffer = drbd_md_get_buffer(mdev);
+ if (!buffer)
+ goto out;
if (!drbd_md_sync_page_io(mdev, bdev, bdev->md.md_offset, READ)) {
/* NOTE: can't do normal error processing here as this is
@@ -3780,7 +3961,8 @@
mdev->sync_conf.al_extents = 127;
err:
- mutex_unlock(&mdev->md_io_mutex);
+ drbd_md_put_buffer(mdev);
+ out:
put_ldev(mdev);
return rv;
@@ -4183,12 +4365,11 @@
static char buildtag[38] = "\0uilt-in";
if (buildtag[0] == 0) {
-#ifdef CONFIG_MODULES
- if (THIS_MODULE != NULL)
- sprintf(buildtag, "srcversion: %-24s", THIS_MODULE->srcversion);
- else
+#ifdef MODULE
+ sprintf(buildtag, "srcversion: %-24s", THIS_MODULE->srcversion);
+#else
+ buildtag[0] = 'b';
#endif
- buildtag[0] = 'b';
}
return buildtag;
diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
index abfaaca..867bf1d 100644
--- a/drivers/block/drbd/drbd_nl.c
+++ b/drivers/block/drbd/drbd_nl.c
@@ -289,7 +289,7 @@
*/
spin_lock_irq(&mdev->req_lock);
ns = mdev->state;
- if (ns.conn < C_WF_REPORT_PARAMS) {
+ if (ns.conn < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &mdev->flags)) {
ns.pdsk = nps;
_drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
}
@@ -432,7 +432,7 @@
/* if this was forced, we should consider sync */
if (forced)
drbd_send_uuids(mdev);
- drbd_send_state(mdev);
+ drbd_send_current_state(mdev);
}
drbd_md_sync(mdev);
@@ -845,9 +845,10 @@
Because new from 8.3.8 onwards the peer can use multiple
BIOs for a single peer_request */
if (mdev->state.conn >= C_CONNECTED) {
- if (mdev->agreed_pro_version < 94)
- peer = mdev->peer_max_bio_size;
- else if (mdev->agreed_pro_version == 94)
+ if (mdev->agreed_pro_version < 94) {
+ peer = min_t(int, mdev->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET);
+ /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */
+ } else if (mdev->agreed_pro_version == 94)
peer = DRBD_MAX_SIZE_H80_PACKET;
else /* drbd 8.3.8 onwards */
peer = DRBD_MAX_BIO_SIZE;
@@ -1032,7 +1033,7 @@
dev_err(DEV, "max capacity %llu smaller than disk size %llu\n",
(unsigned long long) drbd_get_max_capacity(nbc),
(unsigned long long) nbc->dc.disk_size);
- retcode = ERR_DISK_TO_SMALL;
+ retcode = ERR_DISK_TOO_SMALL;
goto fail;
}
@@ -1046,7 +1047,7 @@
}
if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) {
- retcode = ERR_MD_DISK_TO_SMALL;
+ retcode = ERR_MD_DISK_TOO_SMALL;
dev_warn(DEV, "refusing attach: md-device too small, "
"at least %llu sectors needed for this meta-disk type\n",
(unsigned long long) min_md_device_sectors);
@@ -1057,7 +1058,7 @@
* (we may currently be R_PRIMARY with no local disk...) */
if (drbd_get_max_capacity(nbc) <
drbd_get_capacity(mdev->this_bdev)) {
- retcode = ERR_DISK_TO_SMALL;
+ retcode = ERR_DISK_TOO_SMALL;
goto fail;
}
@@ -1138,7 +1139,7 @@
if (drbd_md_test_flag(nbc, MDF_CONSISTENT) &&
drbd_new_dev_size(mdev, nbc, 0) < nbc->md.la_size_sect) {
dev_warn(DEV, "refusing to truncate a consistent device\n");
- retcode = ERR_DISK_TO_SMALL;
+ retcode = ERR_DISK_TOO_SMALL;
goto force_diskless_dec;
}
@@ -1336,17 +1337,34 @@
{
enum drbd_ret_code retcode;
int ret;
+ struct detach dt = {};
+
+ if (!detach_from_tags(mdev, nlp->tag_list, &dt)) {
+ reply->ret_code = ERR_MANDATORY_TAG;
+ goto out;
+ }
+
+ if (dt.detach_force) {
+ drbd_force_state(mdev, NS(disk, D_FAILED));
+ reply->ret_code = SS_SUCCESS;
+ goto out;
+ }
+
drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */
+ drbd_md_get_buffer(mdev); /* make sure there is no in-flight meta-data IO */
retcode = drbd_request_state(mdev, NS(disk, D_FAILED));
+ drbd_md_put_buffer(mdev);
/* D_FAILED will transition to DISKLESS. */
ret = wait_event_interruptible(mdev->misc_wait,
mdev->state.disk != D_FAILED);
drbd_resume_io(mdev);
+
if ((int)retcode == (int)SS_IS_DISKLESS)
retcode = SS_NOTHING_TO_DO;
if (ret)
retcode = ERR_INTR;
reply->ret_code = retcode;
+out:
return 0;
}
@@ -1711,7 +1729,7 @@
if (rs.no_resync && mdev->agreed_pro_version < 93) {
retcode = ERR_NEED_APV_93;
- goto fail;
+ goto fail_ldev;
}
if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev))
@@ -1738,6 +1756,10 @@
fail:
reply->ret_code = retcode;
return 0;
+
+ fail_ldev:
+ put_ldev(mdev);
+ goto fail;
}
static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
@@ -1941,6 +1963,7 @@
/* If there is still bitmap IO pending, probably because of a previous
* resync just being finished, wait for it before requesting a new resync. */
+ drbd_suspend_io(mdev);
wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED);
@@ -1959,6 +1982,7 @@
retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
}
+ drbd_resume_io(mdev);
reply->ret_code = retcode;
return 0;
@@ -1980,6 +2004,7 @@
/* If there is still bitmap IO pending, probably because of a previous
* resync just being finished, wait for it before requesting a new resync. */
+ drbd_suspend_io(mdev);
wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED);
@@ -1998,6 +2023,7 @@
} else
retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S));
}
+ drbd_resume_io(mdev);
reply->ret_code = retcode;
return 0;
@@ -2170,11 +2196,13 @@
/* If there is still bitmap IO pending, e.g. previous resync or verify
* just being finished, wait for it before requesting a new resync. */
+ drbd_suspend_io(mdev);
wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
/* w_make_ov_request expects position to be aligned */
mdev->ov_start_sector = args.start_sector & ~BM_SECT_PER_BIT;
reply->ret_code = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
+ drbd_resume_io(mdev);
return 0;
}
diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c
index 2959cdf..869bada 100644
--- a/drivers/block/drbd/drbd_proc.c
+++ b/drivers/block/drbd/drbd_proc.c
@@ -52,7 +52,7 @@
if (unlikely(v >= 1000000)) {
/* cool: > GiByte/s */
seq_printf(seq, "%ld,", v / 1000000);
- v /= 1000000;
+ v %= 1000000;
seq_printf(seq, "%03ld,%03ld", v/1000, v % 1000);
} else if (likely(v >= 1000))
seq_printf(seq, "%ld,%03ld", v/1000, v % 1000);
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 43beaca..1d088c4 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -466,6 +466,7 @@
goto out;
}
(*newsock)->ops = sock->ops;
+ __module_get((*newsock)->ops->owner);
out:
return err;
@@ -750,6 +751,7 @@
{
struct socket *s, *sock, *msock;
int try, h, ok;
+ enum drbd_state_rv rv;
D_ASSERT(!mdev->data.socket);
@@ -888,25 +890,32 @@
}
}
- if (drbd_request_state(mdev, NS(conn, C_WF_REPORT_PARAMS)) < SS_SUCCESS)
- return 0;
-
sock->sk->sk_sndtimeo = mdev->net_conf->timeout*HZ/10;
sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
atomic_set(&mdev->packet_seq, 0);
mdev->peer_seq = 0;
- drbd_thread_start(&mdev->asender);
-
if (drbd_send_protocol(mdev) == -1)
return -1;
+ set_bit(STATE_SENT, &mdev->flags);
drbd_send_sync_param(mdev, &mdev->sync_conf);
drbd_send_sizes(mdev, 0, 0);
drbd_send_uuids(mdev);
- drbd_send_state(mdev);
+ drbd_send_current_state(mdev);
clear_bit(USE_DEGR_WFC_T, &mdev->flags);
clear_bit(RESIZE_PENDING, &mdev->flags);
+
+ spin_lock_irq(&mdev->req_lock);
+ rv = _drbd_set_state(_NS(mdev, conn, C_WF_REPORT_PARAMS), CS_VERBOSE, NULL);
+ if (mdev->state.conn != C_WF_REPORT_PARAMS)
+ clear_bit(STATE_SENT, &mdev->flags);
+ spin_unlock_irq(&mdev->req_lock);
+
+ if (rv < SS_SUCCESS)
+ return 0;
+
+ drbd_thread_start(&mdev->asender);
mod_timer(&mdev->request_timer, jiffies + HZ); /* just start it here. */
return 1;
@@ -957,7 +966,7 @@
rv = blkdev_issue_flush(mdev->ldev->backing_bdev, GFP_KERNEL,
NULL);
if (rv) {
- dev_err(DEV, "local disk flush failed with status %d\n", rv);
+ dev_info(DEV, "local disk flush failed with status %d\n", rv);
/* would rather check on EOPNOTSUPP, but that is not reliable.
* don't try again for ANY return value != 0
* if (rv == -EOPNOTSUPP) */
@@ -1001,13 +1010,14 @@
if (epoch_size != 0 &&
atomic_read(&epoch->active) == 0 &&
- test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags)) {
+ (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags) || ev & EV_CLEANUP)) {
if (!(ev & EV_CLEANUP)) {
spin_unlock(&mdev->epoch_lock);
drbd_send_b_ack(mdev, epoch->barrier_nr, epoch_size);
spin_lock(&mdev->epoch_lock);
}
- dec_unacked(mdev);
+ if (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags))
+ dec_unacked(mdev);
if (mdev->current_epoch != epoch) {
next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list);
@@ -1096,7 +1106,11 @@
/* In most cases, we will only need one bio. But in case the lower
* level restrictions happen to be different at this offset on this
* side than those of the sending peer, we may need to submit the
- * request in more than one bio. */
+ * request in more than one bio.
+ *
+ * Plain bio_alloc is good enough here, this is no DRBD internally
+ * generated bio, but a bio allocated on behalf of the peer.
+ */
next_bio:
bio = bio_alloc(GFP_NOIO, nr_pages);
if (!bio) {
@@ -1583,6 +1597,24 @@
return ok;
}
+static bool overlapping_resync_write(struct drbd_conf *mdev, struct drbd_epoch_entry *data_e)
+{
+
+ struct drbd_epoch_entry *rs_e;
+ bool rv = 0;
+
+ spin_lock_irq(&mdev->req_lock);
+ list_for_each_entry(rs_e, &mdev->sync_ee, w.list) {
+ if (overlaps(data_e->sector, data_e->size, rs_e->sector, rs_e->size)) {
+ rv = 1;
+ break;
+ }
+ }
+ spin_unlock_irq(&mdev->req_lock);
+
+ return rv;
+}
+
/* Called from receive_Data.
* Synchronize packets on sock with packets on msock.
*
@@ -1826,6 +1858,9 @@
list_add(&e->w.list, &mdev->active_ee);
spin_unlock_irq(&mdev->req_lock);
+ if (mdev->state.conn == C_SYNC_TARGET)
+ wait_event(mdev->ee_wait, !overlapping_resync_write(mdev, e));
+
switch (mdev->net_conf->wire_protocol) {
case DRBD_PROT_C:
inc_unacked(mdev);
@@ -2420,7 +2455,7 @@
mdev->p_uuid[UI_BITMAP] = mdev->p_uuid[UI_HISTORY_START];
mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_HISTORY_START + 1];
- dev_info(DEV, "Did not got last syncUUID packet, corrected:\n");
+ dev_info(DEV, "Lost last syncUUID packet, corrected:\n");
drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]);
return -1;
@@ -2806,10 +2841,10 @@
if (apv >= 88) {
if (apv == 88) {
- if (data_size > SHARED_SECRET_MAX) {
- dev_err(DEV, "verify-alg too long, "
- "peer wants %u, accepting only %u byte\n",
- data_size, SHARED_SECRET_MAX);
+ if (data_size > SHARED_SECRET_MAX || data_size == 0) {
+ dev_err(DEV, "verify-alg of wrong size, "
+ "peer wants %u, accepting only up to %u byte\n",
+ data_size, SHARED_SECRET_MAX);
return false;
}
@@ -3168,9 +3203,20 @@
os = ns = mdev->state;
spin_unlock_irq(&mdev->req_lock);
- /* peer says his disk is uptodate, while we think it is inconsistent,
- * and this happens while we think we have a sync going on. */
- if (os.pdsk == D_INCONSISTENT && real_peer_disk == D_UP_TO_DATE &&
+ /* If some other part of the code (asender thread, timeout)
+ * already decided to close the connection again,
+ * we must not "re-establish" it here. */
+ if (os.conn <= C_TEAR_DOWN)
+ return false;
+
+ /* If this is the "end of sync" confirmation, usually the peer disk
+ * transitions from D_INCONSISTENT to D_UP_TO_DATE. For empty (0 bits
+ * set) resync started in PausedSyncT, or if the timing of pause-/
+ * unpause-sync events has been "just right", the peer disk may
+ * transition from D_CONSISTENT to D_UP_TO_DATE as well.
+ */
+ if ((os.pdsk == D_INCONSISTENT || os.pdsk == D_CONSISTENT) &&
+ real_peer_disk == D_UP_TO_DATE &&
os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) {
/* If we are (becoming) SyncSource, but peer is still in sync
* preparation, ignore its uptodate-ness to avoid flapping, it
@@ -3288,7 +3334,7 @@
/* Nowadays only used when forcing a node into primary role and
setting its disk to UpToDate with that */
drbd_send_uuids(mdev);
- drbd_send_state(mdev);
+ drbd_send_current_state(mdev);
}
}
@@ -3776,6 +3822,13 @@
if (mdev->state.conn == C_STANDALONE)
return;
+ /* We are about to start the cleanup after connection loss.
+ * Make sure drbd_make_request knows about that.
+ * Usually we should be in some network failure state already,
+ * but just in case we are not, we fix it up here.
+ */
+ drbd_force_state(mdev, NS(conn, C_NETWORK_FAILURE));
+
/* asender does not clean up anything. it must not interfere, either */
drbd_thread_stop(&mdev->asender);
drbd_free_sock(mdev);
@@ -3803,8 +3856,6 @@
atomic_set(&mdev->rs_pending_cnt, 0);
wake_up(&mdev->misc_wait);
- del_timer(&mdev->request_timer);
-
/* make sure syncer is stopped and w_resume_next_sg queued */
del_timer_sync(&mdev->resync_timer);
resync_timer_fn((unsigned long)mdev);
@@ -4433,7 +4484,7 @@
if (mdev->state.conn == C_AHEAD &&
atomic_read(&mdev->ap_in_flight) == 0 &&
- !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &mdev->current_epoch->flags)) {
+ !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &mdev->flags)) {
mdev->start_resync_timer.expires = jiffies + HZ;
add_timer(&mdev->start_resync_timer);
}
diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c
index 4a0f314..9c5c849 100644
--- a/drivers/block/drbd/drbd_req.c
+++ b/drivers/block/drbd/drbd_req.c
@@ -37,6 +37,7 @@
const int rw = bio_data_dir(bio);
int cpu;
cpu = part_stat_lock();
+ part_round_stats(cpu, &mdev->vdisk->part0);
part_stat_inc(cpu, &mdev->vdisk->part0, ios[rw]);
part_stat_add(cpu, &mdev->vdisk->part0, sectors[rw], bio_sectors(bio));
part_inc_in_flight(&mdev->vdisk->part0, rw);
@@ -214,8 +215,7 @@
{
const unsigned long s = req->rq_state;
struct drbd_conf *mdev = req->mdev;
- /* only WRITES may end up here without a master bio (on barrier ack) */
- int rw = req->master_bio ? bio_data_dir(req->master_bio) : WRITE;
+ int rw = req->rq_state & RQ_WRITE ? WRITE : READ;
/* we must not complete the master bio, while it is
* still being processed by _drbd_send_zc_bio (drbd_send_dblock)
@@ -230,7 +230,7 @@
return;
if (s & RQ_NET_PENDING)
return;
- if (s & RQ_LOCAL_PENDING)
+ if (s & RQ_LOCAL_PENDING && !(s & RQ_LOCAL_ABORTED))
return;
if (req->master_bio) {
@@ -277,6 +277,9 @@
req->master_bio = NULL;
}
+ if (s & RQ_LOCAL_PENDING)
+ return;
+
if ((s & RQ_NET_MASK) == 0 || (s & RQ_NET_DONE)) {
/* this is disconnected (local only) operation,
* or protocol C P_WRITE_ACK,
@@ -429,7 +432,7 @@
break;
case completed_ok:
- if (bio_data_dir(req->master_bio) == WRITE)
+ if (req->rq_state & RQ_WRITE)
mdev->writ_cnt += req->size>>9;
else
mdev->read_cnt += req->size>>9;
@@ -438,7 +441,14 @@
req->rq_state &= ~RQ_LOCAL_PENDING;
_req_may_be_done_not_susp(req, m);
- put_ldev(mdev);
+ break;
+
+ 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;
case write_completed_with_error:
@@ -447,7 +457,6 @@
__drbd_chk_io_error(mdev, false);
_req_may_be_done_not_susp(req, m);
- put_ldev(mdev);
break;
case read_ahead_completed_with_error:
@@ -455,7 +464,6 @@
req->rq_state |= RQ_LOCAL_COMPLETED;
req->rq_state &= ~RQ_LOCAL_PENDING;
_req_may_be_done_not_susp(req, m);
- put_ldev(mdev);
break;
case read_completed_with_error:
@@ -467,7 +475,8 @@
D_ASSERT(!(req->rq_state & RQ_NET_MASK));
__drbd_chk_io_error(mdev, false);
- put_ldev(mdev);
+
+ goto_queue_for_net_read:
/* no point in retrying if there is no good remote data,
* or we have no connection. */
@@ -556,10 +565,8 @@
drbd_queue_work(&mdev->data.work, &req->w);
break;
- case oos_handed_to_network:
- /* actually the same */
+ case read_retry_remote_canceled:
case send_canceled:
- /* treat it the same */
case send_failed:
/* real cleanup will be done from tl_clear. just update flags
* so it is no longer marked as on the worker queue */
@@ -589,17 +596,17 @@
}
req->rq_state &= ~RQ_NET_QUEUED;
req->rq_state |= RQ_NET_SENT;
- /* because _drbd_send_zc_bio could sleep, and may want to
- * dereference the bio even after the "write_acked_by_peer" and
- * "completed_ok" events came in, once we return from
- * _drbd_send_zc_bio (drbd_send_dblock), we have to check
- * whether it is done already, and end it. */
_req_may_be_done_not_susp(req, m);
break;
- case read_retry_remote_canceled:
+ case oos_handed_to_network:
+ /* Was not set PENDING, no longer QUEUED, so is now DONE
+ * as far as this connection is concerned. */
req->rq_state &= ~RQ_NET_QUEUED;
- /* fall through, in case we raced with drbd_disconnect */
+ req->rq_state |= RQ_NET_DONE;
+ _req_may_be_done_not_susp(req, m);
+ break;
+
case connection_lost_while_pending:
/* transfer log cleanup after connection loss */
/* assert something? */
@@ -616,8 +623,6 @@
_req_may_be_done(req, m); /* Allowed while state.susp */
break;
- case write_acked_by_peer_and_sis:
- req->rq_state |= RQ_NET_SIS;
case conflict_discarded_by_peer:
/* for discarded conflicting writes of multiple primaries,
* there is no need to keep anything in the tl, potential
@@ -628,18 +633,15 @@
(unsigned long long)req->sector, req->size);
req->rq_state |= RQ_NET_DONE;
/* fall through */
+ case write_acked_by_peer_and_sis:
case write_acked_by_peer:
+ if (what == write_acked_by_peer_and_sis)
+ req->rq_state |= RQ_NET_SIS;
/* protocol C; successfully written on peer.
- * Nothing to do here.
+ * Nothing more 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. */
+ * for volatile write-back caches on lower level devices. */
- /* this makes it effectively the same as for: */
case recv_acked_by_peer:
/* protocol B; pretends to be successfully written on peer.
* see also notes above in handed_over_to_network about
@@ -773,6 +775,7 @@
int local, remote, send_oos = 0;
int err = -EIO;
int ret = 0;
+ union drbd_state s;
/* allocate outside of all locks; */
req = drbd_req_new(mdev, bio);
@@ -834,8 +837,9 @@
drbd_al_begin_io(mdev, sector);
}
- remote = remote && drbd_should_do_remote(mdev->state);
- send_oos = rw == WRITE && drbd_should_send_oos(mdev->state);
+ s = mdev->state;
+ remote = remote && drbd_should_do_remote(s);
+ send_oos = rw == WRITE && drbd_should_send_oos(s);
D_ASSERT(!(remote && send_oos));
if (!(local || remote) && !is_susp(mdev->state)) {
@@ -867,7 +871,7 @@
if (is_susp(mdev->state)) {
/* If we got suspended, use the retry mechanism of
- generic_make_request() to restart processing of this
+ drbd_make_request() to restart processing of this
bio. In the next call to drbd_make_request
we sleep in inc_ap_bio() */
ret = 1;
@@ -1091,7 +1095,6 @@
*/
D_ASSERT(bio->bi_size > 0);
D_ASSERT((bio->bi_size & 0x1ff) == 0);
- D_ASSERT(bio->bi_idx == 0);
/* to make some things easier, force alignment of requests within the
* granularity of our hash tables */
@@ -1099,8 +1102,9 @@
e_enr = (bio->bi_sector+(bio->bi_size>>9)-1) >> HT_SHIFT;
if (likely(s_enr == e_enr)) {
- inc_ap_bio(mdev, 1);
- drbd_make_request_common(mdev, bio, start_time);
+ do {
+ inc_ap_bio(mdev, 1);
+ } while (drbd_make_request_common(mdev, bio, start_time));
return;
}
@@ -1196,36 +1200,66 @@
struct drbd_conf *mdev = (struct drbd_conf *) data;
struct drbd_request *req; /* oldest request */
struct list_head *le;
- unsigned long et = 0; /* effective timeout = ko_count * timeout */
+ unsigned long ent = 0, dt = 0, et, nt; /* effective timeout = ko_count * timeout */
+ unsigned long now;
if (get_net_conf(mdev)) {
- et = mdev->net_conf->timeout*HZ/10 * mdev->net_conf->ko_count;
+ if (mdev->state.conn >= C_WF_REPORT_PARAMS)
+ ent = mdev->net_conf->timeout*HZ/10
+ * mdev->net_conf->ko_count;
put_net_conf(mdev);
}
- if (!et || mdev->state.conn < C_WF_REPORT_PARAMS)
+ if (get_ldev(mdev)) { /* implicit state.disk >= D_INCONSISTENT */
+ dt = mdev->ldev->dc.disk_timeout * HZ / 10;
+ put_ldev(mdev);
+ }
+ et = min_not_zero(dt, ent);
+
+ if (!et)
return; /* Recurring timer stopped */
+ now = jiffies;
+
spin_lock_irq(&mdev->req_lock);
le = &mdev->oldest_tle->requests;
if (list_empty(le)) {
spin_unlock_irq(&mdev->req_lock);
- mod_timer(&mdev->request_timer, jiffies + et);
+ mod_timer(&mdev->request_timer, now + et);
return;
}
le = le->prev;
req = list_entry(le, struct drbd_request, tl_requests);
- if (time_is_before_eq_jiffies(req->start_time + et)) {
- if (req->rq_state & RQ_NET_PENDING) {
- dev_warn(DEV, "Remote failed to finish a request within ko-count * timeout\n");
- _drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE, NULL);
- } else {
- dev_warn(DEV, "Local backing block device frozen?\n");
- mod_timer(&mdev->request_timer, jiffies + et);
- }
- } else {
- mod_timer(&mdev->request_timer, req->start_time + et);
- }
+ /* The request is considered timed out, if
+ * - we have some effective timeout from the configuration,
+ * with above state restrictions applied,
+ * - the oldest request is waiting for a response from the network
+ * resp. the local disk,
+ * - the oldest request is in fact older than the effective timeout,
+ * - the connection was established (resp. disk was attached)
+ * for longer than the timeout already.
+ * Note that for 32bit jiffies and very stable connections/disks,
+ * we may have a wrap around, which is catched by
+ * !time_in_range(now, last_..._jif, last_..._jif + timeout).
+ *
+ * Side effect: once per 32bit wrap-around interval, which means every
+ * ~198 days with 250 HZ, we have a window where the timeout would need
+ * to expire twice (worst case) to become effective. Good enough.
+ */
+ if (ent && req->rq_state & RQ_NET_PENDING &&
+ time_after(now, req->start_time + ent) &&
+ !time_in_range(now, mdev->last_reconnect_jif, mdev->last_reconnect_jif + ent)) {
+ dev_warn(DEV, "Remote failed to finish a request within ko-count * timeout\n");
+ _drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE | CS_HARD, NULL);
+ }
+ if (dt && req->rq_state & RQ_LOCAL_PENDING &&
+ time_after(now, req->start_time + dt) &&
+ !time_in_range(now, mdev->last_reattach_jif, mdev->last_reattach_jif + dt)) {
+ dev_warn(DEV, "Local backing device failed to meet the disk-timeout\n");
+ __drbd_chk_io_error(mdev, 1);
+ }
+ nt = (time_after(now, req->start_time + et) ? now : req->start_time) + et;
spin_unlock_irq(&mdev->req_lock);
+ mod_timer(&mdev->request_timer, nt);
}
diff --git a/drivers/block/drbd/drbd_req.h b/drivers/block/drbd/drbd_req.h
index 68a234a..3d21119 100644
--- a/drivers/block/drbd/drbd_req.h
+++ b/drivers/block/drbd/drbd_req.h
@@ -105,6 +105,7 @@
read_completed_with_error,
read_ahead_completed_with_error,
write_completed_with_error,
+ abort_disk_io,
completed_ok,
resend,
fail_frozen_disk_io,
@@ -118,18 +119,21 @@
* same time, so we should hold the request lock anyways.
*/
enum drbd_req_state_bits {
- /* 210
- * 000: no local possible
- * 001: to be submitted
+ /* 3210
+ * 0000: no local possible
+ * 0001: to be submitted
* UNUSED, we could map: 011: submitted, completion still pending
- * 110: completed ok
- * 010: completed with error
+ * 0110: completed ok
+ * 0010: completed with error
+ * 1001: Aborted (before completion)
+ * 1x10: Aborted and completed -> free
*/
__RQ_LOCAL_PENDING,
__RQ_LOCAL_COMPLETED,
__RQ_LOCAL_OK,
+ __RQ_LOCAL_ABORTED,
- /* 76543
+ /* 87654
* 00000: no network possible
* 00001: to be send
* 00011: to be send, on worker queue
@@ -199,8 +203,9 @@
#define RQ_LOCAL_PENDING (1UL << __RQ_LOCAL_PENDING)
#define RQ_LOCAL_COMPLETED (1UL << __RQ_LOCAL_COMPLETED)
#define RQ_LOCAL_OK (1UL << __RQ_LOCAL_OK)
+#define RQ_LOCAL_ABORTED (1UL << __RQ_LOCAL_ABORTED)
-#define RQ_LOCAL_MASK ((RQ_LOCAL_OK << 1)-1) /* 0x07 */
+#define RQ_LOCAL_MASK ((RQ_LOCAL_ABORTED << 1)-1)
#define RQ_NET_PENDING (1UL << __RQ_NET_PENDING)
#define RQ_NET_QUEUED (1UL << __RQ_NET_QUEUED)
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index 4d3e6f6..620c70f 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -70,11 +70,29 @@
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,
@@ -226,6 +244,7 @@
spin_lock_irqsave(&mdev->req_lock, flags);
__req_mod(req, what, &m);
spin_unlock_irqrestore(&mdev->req_lock, flags);
+ put_ldev(mdev);
if (m.bio)
complete_master_bio(mdev, &m);
@@ -290,7 +309,7 @@
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);
}
@@ -728,7 +747,7 @@
}
drbd_start_resync(mdev, C_SYNC_SOURCE);
- clear_bit(AHEAD_TO_SYNC_SOURCE, &mdev->current_epoch->flags);
+ clear_bit(AHEAD_TO_SYNC_SOURCE, &mdev->flags);
return 1;
}
@@ -1519,14 +1538,14 @@
}
drbd_state_lock(mdev);
-
+ write_lock_irq(&global_state_lock);
if (!get_ldev_if_state(mdev, D_NEGOTIATING)) {
+ write_unlock_irq(&global_state_lock);
drbd_state_unlock(mdev);
return;
}
- write_lock_irq(&global_state_lock);
- ns = mdev->state;
+ ns.i = mdev->state.i;
ns.aftr_isp = !_drbd_may_sync_now(mdev);
diff --git a/include/linux/drbd.h b/include/linux/drbd.h
index 9e5f560..47e3d48 100644
--- a/include/linux/drbd.h
+++ b/include/linux/drbd.h
@@ -53,7 +53,7 @@
extern const char *drbd_buildtag(void);
-#define REL_VERSION "8.3.11"
+#define REL_VERSION "8.3.13"
#define API_VERSION 88
#define PRO_VERSION_MIN 86
#define PRO_VERSION_MAX 96
@@ -112,8 +112,8 @@
ERR_OPEN_MD_DISK = 105,
ERR_DISK_NOT_BDEV = 107,
ERR_MD_NOT_BDEV = 108,
- ERR_DISK_TO_SMALL = 111,
- ERR_MD_DISK_TO_SMALL = 112,
+ ERR_DISK_TOO_SMALL = 111,
+ ERR_MD_DISK_TOO_SMALL = 112,
ERR_BDCLAIM_DISK = 114,
ERR_BDCLAIM_MD_DISK = 115,
ERR_MD_IDX_INVALID = 116,
diff --git a/include/linux/drbd_limits.h b/include/linux/drbd_limits.h
index 447c367..fb670bf 100644
--- a/include/linux/drbd_limits.h
+++ b/include/linux/drbd_limits.h
@@ -48,6 +48,11 @@
#define DRBD_TIMEOUT_MAX 600
#define DRBD_TIMEOUT_DEF 60 /* 6 seconds */
+ /* If backing disk takes longer than disk_timeout, mark the disk as failed */
+#define DRBD_DISK_TIMEOUT_MIN 0 /* 0 = disabled */
+#define DRBD_DISK_TIMEOUT_MAX 6000 /* 10 Minutes */
+#define DRBD_DISK_TIMEOUT_DEF 0 /* disabled */
+
/* active connection retries when C_WF_CONNECTION */
#define DRBD_CONNECT_INT_MIN 1
#define DRBD_CONNECT_INT_MAX 120
@@ -60,7 +65,7 @@
/* timeout for the ping packets.*/
#define DRBD_PING_TIMEO_MIN 1
-#define DRBD_PING_TIMEO_MAX 100
+#define DRBD_PING_TIMEO_MAX 300
#define DRBD_PING_TIMEO_DEF 5
/* max number of write requests between write barriers */
diff --git a/include/linux/drbd_nl.h b/include/linux/drbd_nl.h
index ab6159e4..a8706f0 100644
--- a/include/linux/drbd_nl.h
+++ b/include/linux/drbd_nl.h
@@ -31,9 +31,12 @@
NL_INTEGER( 56, T_MAY_IGNORE, max_bio_bvecs)
NL_BIT( 57, T_MAY_IGNORE, no_disk_barrier)
NL_BIT( 58, T_MAY_IGNORE, no_disk_drain)
+ NL_INTEGER( 89, T_MAY_IGNORE, disk_timeout)
)
-NL_PACKET(detach, 4, )
+NL_PACKET(detach, 4,
+ NL_BIT( 88, T_MANDATORY, detach_force)
+)
NL_PACKET(net_conf, 5,
NL_STRING( 8, T_MANDATORY, my_addr, 128)