[ofa-general] [ANNOUNCE] libibverbs 1.1-rc1 released

[Roland Dreier]

I just tagged the 1.1-rc1 release of libibverbs and pushed it out to
my git tree on kernel.org:

    git://git.kernel.org/pub/scm/libs/infiniband/libibverbs.git

(the name of the tag is libibverbs-1.1-rc1).

I've also copied a tarball into my home directory on openfabrics.org,
with sha1sum:

    f92b71268eacfffa858a05582b0ed71ce37e760c  libibverbs-1.1-rc1.tar.gz

I would appreciated it if someone with access could move this into the
right directory to appear in <http://www.openfabrics.org//downloads/>

This release is the beginning of a major release cycle for libibverbs,
so full compatibility with earlier libibverbs 1.0 releases is not
preserved.  Low-level device drivers will need to be rebuilt to work
with libibverbs 1.1.  However, a versioned ABI is provided so that
applications dynamically linked with libibverbs 1.0 should work.  I
don't know of any major source level incompatibilities that would
prevent an application that compiles against libibverbs 1.0 from
building and working with libibverbs 1.1.

I believe that libibverbs 1.1 is quite stable, since prereleases of
the 1.1 tree have been shipped in the OFED 1.2 prereleases without any
significant bug reports, and no major changes have gone into the tree
for quite some time.  Therefore I think a realistic schedule would be
libibverbs 1.1-rc2 in one week, and libibverbs 1.1 final one week
after that (April 11).

A git shortlog of the changes on the 1.1 branch is below:

Dotan Barak (4):
      Fix some memory leaks in read_config() error path
      Add resource cleanup at end of pingpong tests
      Fix memory leak on ibv_fork_init() error path
      Man page updates

Jack Morgenstein (2):
      Return sq_draining properly from query_qp
      Delete man3 symbolic links before creating them during install

Jeff Squyres (1):
      Add README notes about Valgrind memcheck support

Leonid Arsh (1):
      Add IBV_EVENT_CLIENT_REREGISTER to libibverbs

Ralph Campbell (1):
      Add response handling to ibv_cmd_resize_cq()

Roland Dreier (34):
      Branch a libibverbs-1.0 tree for maintenance
      Fix update to Debian policy 3.7.2
      Fix minor memory leaks
      Fix ibv_get_device_list() to really NULL-terminate the array
      Fix libibverbs definition of mb() for sparc
      Make fork() work for verbs consumers
      Simplify Debian package version
      Fix formatting of pingpong man pages slightly
      Debian packaging improvements
      Fix alignment of work request structures
      Update libibverbs man pages so they don't refer to "OpenIB"
      Add node_type and transport_type members to struct ibv_device
      Add Valgrind annotations
      Fix up configure test for <valgrind/memcheck.h>
      Update ChangeLogs to give credit for Valgrind annotations
      Add handling of --with-valgrind=<valgrind dir>
      Add rmb() and wmb() to <infiniband/arch.h>
      Minor cleanups
      Fix previous sq_draining change so it actually builds
      Rewrite test for linker script to get rid of Makefile conditionals
      Fix rewritten test for linker script support
      Implement new method for finding and loading device-specific drivers
      Revert "Pass driver data through ibv_cmd_req_notify_cq()"
      Don't use d_type member of struct dirent
      Fix caching of --version-script check
      Add ABI compatibility for apps linked against libibverbs 1.0
      Rename Debian package back to libibverbs1
      Fix unset context breakage when a low-level driver does kernel bypass
      Update Debian changelog
      Start adding libibverbs manpages
      Add remaining libibverbs manpages
      Add low-level driver hooks for reregister MR and memory windows
      Reference count completion channels
      Roll libibverbs 1.1-rc1 release

Sean Hefty (1):
      Add some helper functions to simplify using UD QPs

Steve Wise (5):
      Add async_event callback function to struct ibv_context_ops
      Support provider response data in reg_mr command
      Pass driver data through ibv_cmd_req_notify_cq()
      Don't lose devices when multiple RDMA devices are present
      The ibv_cmd_* create functions need to set context