BSD Commit Log Search

   Add O and V loops in formats for options and environment.

   Add floating panes key bindings and menu options, from Dane Jensen.

   Fix some spelling and indentation.

   Add a quoted pane_start_command_list which does not lose argument
   information.

   Add -E to new-window and respawn-pane and respawn-window to create empty
   panes.

   Add additional pane sort orders.

   Add manual width and height window formats.

   Add single quote operator.

   iked: vroute: Validate routing socket replies

   When parsing malformed RTM replies from the kernel vroute_process()
   might walk beyond the response buffer.  Therefore check msglen and
   rtm_msglen cover the header, and bound every sa_len against the
   remaining message and sizeof(struct sockaddr_storage) before copying.

   From Andrew Griffiths, thanks!

   rework how mbufs share references to external storage

   previously mbufs that shared a reference to the same external storage
   were linked together with a pair of lists, but operations on these
   lists have to be serialised because different cpus can be working
   on on mbufs that share the same external storage. this serialisation
   was provided by a single global mutex, which has now become contended
   when the kernel is doing a lot of work that relies on shared external
   storage.

   this diff replaces the links with an m_ext_refs struct that "proxies"
   the external storage information on mbufs when that storage is first
   shared between two mbufs. that struct contains a refcnt that's
   increased if the external storage is shared again, and only when
   the refcnt drops to zero is the external storage actually released
   with the original free handler. these per shared storage refcnts
   replace the global mutex for coordinating the "finalisation" and
   actual free of the external storage.


    [9 lines not shown]

   aldap_parse(): simplify error handling

   looks good to me. jmatthew@

   Allow join-pane to tile floating panes, from Dane Jensen.

   Attach to BCM5745x devices, which should work as far as I can tell from the
   vendor driver.

   regen

   Add missing BCM5745x devices, pointed out by Brad

   use the same rwlock for queue kstats as other kstat instances

   ok dlg@

   Enable swap-pane on floating panes again but not -U/-D. From Dane
   Jensen.

   Use message-style as default for message-format again. GitHub issue 5315.

   Do not double free or leak pane on failure, from Uzair Aftab in GitHub
   issue 5316.

   cal: make specifying a month conflict with -y

   ok jca@

   Don't hardcode the disklabel partition of the EFI System Partition in the
   code that mounts it to copy over the Apple firmware.  use chicken scratches
   provided by deraadt@ instead to find it.

   Fixes a problem with copying the firmware when doing an install as reported
   bu caspar@ (who also tested an earlier version of the chicken scratches).

   ok deraadt@

   add support for wakeup interrupts to bytgpio(4)

   from John Troy who tested on a Lenovo IdeaPad 100S-11IBY
   resuming from S0ix suspend with a lid open
   ok kettenis@

   move bounds test in proc_accept() before first use
   found with smatch, ok rsadowski@ martijn@

   Copy UTF-8 data correctly.

   Fix bug in prefix_evaluate handling of new prefixes.

   If new is not eligible then prefix_peer(new) will blow up since new is NULL.
   If the new prefix is eligible then clear old and old_pathid_tx and set
   peer - the new path will be recorded in rde_enqueue_updates().
   If the new prefix is not eligible then clear new but keep old and
   old_pathid_tx. If old is also NULL then nothing changed and an early exit
   happens. Else old sets peer and rde_enqueue_updates() records the withdraw.

   Found by jsg@, OK tb@

   Add a raw grid output to capture-pane for debugging and regression
   tests.

   Do not crash if no client.

   Implement a simple rib entry pending paths queue.

   Elements are added to the queue when rib entries are enqueued in
   rde_enqueue_updates() for the EVAL_DEFAULT and EVAL_ALL cases.
   The queue is then flushed once processed by peer_process_updates().
   In rib_remove() assert that the queue is empty before free. This is
   implied by the re_is_queued() check but that check only looks at the
   pq_mode.

   A specific path can only be enqueued once, only the last version matters.
   In other words rapid withdraw and update actions will be aggregated into
   a single action.

   Nothing is currently using this data. This will follow next.
   OK tb@

   Cope with recent rde_enqueue_updates() API changes.

   correct indentation of cd9660_ihashget()