- 08 Nov, 2012 40 commits
-
-
Lars Ellenberg authored
Signed-off-by:
Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by:
Lars Ellenberg <lars.ellenberg@linbit.com>
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
w_restart_write(), run from worker context, calls __drbd_make_request() and further drbd_al_begin_io(, delegate=true), which then potentially deadlocks. The previous patch moved a BUG_ON to expose such call paths, which would now be triggered. Also, if we call __drbd_make_request() from resource worker context, like w_restart_write() did, and that should block for whatever reason (!drbd_state_is_stable(), resource suspended, ...), we potentially deadlock the whole resource, as the worker is needed for state changes and other things. Create a dedicated retry workqueue for this instead. Also make sure that inc_ap_bio()/dec_ap_bio() are properly paired, even if do_retry() needs to retry itself, in case __drbd_make_request() returns != 0. Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
When we have a write request and a state change C_WF_BITMAP_S -> C_SYNC_SOURCE at the same time, and it happens that the line remote = remote && drbd_should_do_remote(s); stills sees C_WF_BITMAP_S, and send_oos = rw == WRITE && drbd_should_send_oos(s); already sees C_SYNC_SOURCE both are 0. This causes the write to not be mirrored, but marked as out-of-sync on the Sync_Source node. Signed-off-by: -
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Without this, iostat frequently sees bogus svctime and >= 100% "utilization". Signed-off-by:
-
Philipp Reisner authored
drbd_accept was modelled after kernel_accept with drbd commit 53eb779 in July 2008. Only, kernel_accept was then broken, and only fixed later with kernel commit 1b08534e in Dec 2008: net: Fix module refcount leak in kernel_accept() Impact: protocol families provided as modules, e.g. ipv6 or ib_sdp, would soon have their reference count become negative, preventing them from being unloaded (likely), or worse, hit zero without actually being unused, allowing them to be unloaded while still in use (unlikely, but if triggered, causing a kernel crash). Signed-off-by:
-
Philipp Reisner authored
...and not all volumes of the resource Signed-off-by:
-
Philipp Reisner authored
If the backing device is already frozen during attach, we failed to recognize that. The current disk-timeout code works on top of the drbd_request objects. During attach we do not allow IO and therefore never generate a drbd_request object but block before that in drbd_make_request(). This patch adds the timeout to all drbd_md_sync_page_io(). Before this patch we used to go from D_ATTACHING directly to D_DISKLESS if IO failed during attach. We can no longer do this since we have to stay in D_FAILED until all IO ops issued to the backing device returned. Signed-off-by:
-
Philipp Reisner authored
Commit d0ef827e (drbd: switch configuration interface from connector to genetlink) introduced a regression by removing the ability to set all bits in the out of sync bitmap and to suspend updates to the activity log of a disconnected device via the invalidate-remote management call. Credits for reporting the issue are going to Arne Redlich. Signed-off-by:
-
Lars Ellenberg authored
We send left-over garbage from the previous packet in P_DATA_REPLY and P_RS_DATA_REPLY packets. That's bad behaviour. Signed-off-by:
-
Philipp Reisner authored
For compatibility reasons 8.4 has to send P_STATE_CHG_REQ (instead of P_CONN_ST_CHG_REQ) when disconnecting. In the receiving code path we missed to convert the old answer (P_STATE_CHG_REPLY) back to 8.4 logic. Therefore the CL_ST_CHG_SUCCESS or CL_ST_CHG_FAIL bit in the flags word of mdev got set, while the state code was waiting for the CONN_WD_ST_CHG_OKAY or CONN_WD_ST_CHG_FAIL bits in tconn. Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Philipp Reisner authored
With Linux-3.2 generic_make_request() will no longer loop over the request function until it finally returns 0. Move this loop into our drbd_make_request() function. Signed-off-by:
-
Philipp Reisner authored
With commit from Mon Mar 28 16:33:12 2011 +0200 "drbd: drbd_connect(): Initialize struct drbd_socket before sending anything" tconn->data.sock and tconn->meta.sock get assigned early, in conn_connect. The early assigning can trigger an OOPS, because it may released the socket without acquiring the mutex protecting the socket. An other thread (worker) might use setsockopt() on the socket while it gets free()ed. Restored the (proven) 8.3 behavior of assigning these sockets after the two connections are established. Credits for reporting the issue are going to Arne Redlich. Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
ap_in_flight only counts writes. NEG_ACKED is an action on a request that might be called for reads and writes. This bug was there forever, but it becomes much more relevant with the read balincing code. Signed-off-by:
-
Philipp Reisner authored
I.e. in C_WF_REPORT_PARAMS or in C_WF_CONNECTION. Sending may already work in these cstates, but the peer still expects the HandShake / ConnectionFeatures packet. Actually triggered by the Testuite on kugel. Signed-off-by:
-
Philipp Reisner authored
If the asender thread, or request_timer_fn(), or some other part of the code, decided to drop the connection (because of timeout or other), but the receiver just now was processing a P_STATE packet, there was a chance that receive_state() would do a hard state change "re-establishing" an already failed connection without additional handshake. Log excerpt: Remote failed to finish a request within ko-count * timeout peer( Secondary -> Unknown ) conn( Connected -> Timeout ) pdsk( UpToDate -> DUnknown ) asender terminated ... peer( Unknown -> Secondary ) conn( Timeout -> Connected ) pdsk( DUnknown -> UpToDate ) peer_isp( 0 -> 1 ) ... Connection closed peer( Secondary -> Unknown ) conn( Connected -> Unconnected ) pdsk( UpToDate -> DUnknown ) peer_isp( 1 -> 0 ) receiver terminated Impact: while the connection state is erroneously "Connected", requests may be queued and even sent, which would never be acknowledged, and may have been missed by the cleanup. These requests would never be completed. The next drbd_suspend_io() will then lock up, waiting forever for these requests to complete. Fixed in several code paths: Make sure the connection state is NetworkFailure or worse before starting the cleanup in drbd_disconnect(). This should make sure the cleanup won't miss any requests. Disallow receive_state() to "upgrade" the connection state from an error state. This will make sure the "illegal" state transition won't happen. For all connection failure states, relax the safe-guard in sanitize_state() again to silently mask out those state changes (e.g. Timeout -> Connected becomes Timeout -> Timeout). Note by Philipp Reisner: The 3rd chunk described as "relax the safe-guard..." is not there in 8.4 as it is relaxed to the maximum in 8.4 already Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
New config option for the disk secition "read-balancing", with the values: prefer-local, prefer-remote, round-robin, when-congested-remote. Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Lars Ellenberg authored
Recent commit drbd: Move write_ordering from mdev to tconn introduced a new idr_for_each loop over all volumes, but did not take necessary rcu locks or krefs. Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Philipp Reisner authored
drbd_try_clear_on_disk_bm() has a sanity check for the number of blocks left to be resynced (rs_left) in the current resync extent. If it detects a mismatch, it complains, and forces a disconnect using drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); Unfortunately, this may be called while holding the req_lock, and drbd_force_state() want's to aquire that lock itself. Deadlock. Don't force a disconnect, but fix up rs_left by recounting and reassigning the number of dirty blocks in that extent. Signed-off-by:
-
Philipp Reisner authored
Wait until IO is drained in all volumes. Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
This is necessary since the transfer_log on the sending is also per tconn. Signed-off-by:
-
Philipp Reisner authored
An epoch object needs a pointer to the mdev it was received for. This is necessary to be able to send the barrier ack packet for the same volume as the original barrier packet was assigned to. This prepares the next step, in which the (receiver side) epoch list is moved from the device (mdev) to the connection (tconn) object. Signed-off-by:
-
