- 14 Oct, 2011 32 commits
-
-
Andreas Gruenbacher authored
Signed-off-by:
Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by:
Lars Ellenberg <lars.ellenberg@linbit.com>
-
Andreas Gruenbacher authored
Instead of artificially enlarging the command decoding arrays to P_MAX_CMD entries, check if an index is within the valid range using the ARRAY_SIZE() macro. Signed-off-by:
-
Andreas Gruenbacher authored
Opening a device only succeeds on a primary node, or when explicitly setting the allow_oos module parameter to allow opening the device read-only on a secondary node. There is no other way that a request can get into drbd_make_request(), so this code cannot trigger. Signed-off-by:
-
Andreas Gruenbacher authored
The previous algorithm for dealing with overlapping concurrent writes was generating unnecessary warnings for scenarios which could be legitimate, and did not always handle partially overlapping requests correctly. Improve it algorithm as follows: * While local or remote write requests are in progress, conflicting new local write requests will be delayed (commit 82172f7). * When a conflict between a local and remote write request is detected, the node with the discard flag decides how to resolve the conflict: It will ask its peer to discard conflicting requests which are fully contained in the local request and retry requests which overlap only partially. This involves a protocol change. Signed-off-by:
-
Andreas Gruenbacher authored
When the node with the discard flag resolves write conflicts in dual-primary mode, it may determine that its peer has sent ack packets on the metadata socket which did not arrive, yet. Wait for the next ack with ping-timeout instead of a hard-coded 30 seconds. Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Andreas Gruenbacher authored
Commit 9b1e63e changed the concurrent write detection algorithm to only insert peer requests into write_requests tree after determining that there is no conflict. With this change, new conflicting local requests could be added while the algorithm runs, but this case was not handled correctly. Instead of making the algorithm deal with this case, switch back to adding peer requests to the write_requests tree immediately: this improves fairness. When a peer request is discarded, remove that request from the write_requests Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
Summary log messages meant for global bitmap IO should not be printed for bitmap IO caused by activity log transactions. Signed-off-by:
-
Lars Ellenberg authored
Use a new on-disk transaction format for the activity log, which allows for multiple changes to the active set per transaction. Using 4k transaction blocks, we can now get rid of the work-around code to deal with devices not supporting 512 byte logical block size. Signed-off-by:
-
Lars Ellenberg authored
Allow multiple changes to the active set of elements in lru_cache. The only current user of lru_cache, drbd, is driving this generalisation. Signed-off-by:
-
Lars Ellenberg authored
We are about to allow several changes to the active set in one activity log transaction. We have to write out the corresponding bitmap pages as well, if changed. Introduce drbd_bm_mark_for_writeout(), then re-use the existing bitmap writeout path to submit all marked pages in one go. Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
For some time we contemplated calling the "struct lru_cache" a "struct tracked_set", and some comments kept the ts_ prefix. Fix those to match the member field names. Signed-off-by:
-
Lars Ellenberg authored
Some open-coded clear_bit(); smp_mb__after_clear_bit(); should in fact have been smp_mb__before_clear_bit(); clear_bit(); Instead, use clear_bit_unlock() to annotate the intention, and have it do the right thing. Signed-off-by:
-
Lars Ellenberg authored
As using an empty activity log is the whole point of the excercise, make sure it is still empty when setting AL_SUSPENDED. Signed-off-by:
-
Lars Ellenberg authored
Signed-off-by:
-
Lars Ellenberg authored
fold if (x >= (N+1)) return 0; if (x < N) return 0; into if (x != N) return 0; Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Andreas Gruenbacher authored
Add a macro and helper function for doing that. Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Andreas Gruenbacher authored
Signed-off-by:
-
Philipp Reisner authored
That is used for graceful disconnect only Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
In compatibility mode with old DRBDs, use that as the state_mutex as well. Signed-off-by:
-
Philipp Reisner authored
The lock they constructed is only taken when the state_mutex was already taken. It is superficial. Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
- 28 Sep, 2011 8 commits
-
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
And removed the unused mdev parameter, and made the order of the state parameters: os, ns Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
Signed-off-by:
-
Philipp Reisner authored
No longer work callbacks must operate on a mdev. From now on they can also operate on a tconn. Signed-off-by:
-
