[PATCH 00/13] Policy changes and packaging support

[PATCH 00/13] Policy changes and packaging support

[Jason Gunthorpe]

This is the next batch of rdma-core updates, these 13 patches deal primarily
with how the package is built and the install paths afterwards. Most likely
these changes are mainly of interest to packagers.

The discussion here is informative for some of the patches:
 http://www.spinics.net/lists/linux-rdma/msg39578.html

View the commits on github:

 https://github.com/linux-rdma/rdma-core/pull/3

Please forward Acks and I will update the commits.

One important thing to note is that this series pretty much eliminates all
hard coded paths from the tree. While this is a good thing, it means the
default behavior is to set SYSCONFDIR to /usr/local/etc/, which is probably
not what people testing the daemons really want. Compile with:

 cmake -GNinja .. -DCMAKE_INSTALL_SYSCONFDIR:PATH=/etc

Which is similar to the configure --sysconfdir option

Jason Gunthorpe (13):
  Be consistent about defining NDEBUG
  Switch valgrind memcheck.h to on by default
  Do not produce libtool .la files
  Do not produce static libraries by default
  Have cmake run man pages through text substitution
  Consolidate definitions for paths into config.h
  Pass ancillary files that use paths through cmake substitution
  srp_daemon: Move lock file into /var/run/
  verbs: Move the providers into /usr/lib.../libibverbs by default
  ipathverbs: Move truescale-serdes.cmds to /usr/libexec/
  srp_daemon: Add the debian initscripts as an option
  Document a versioning strategy and use it
  Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

 CMakeLists.txt                                     |  47 +++++-
 Documentation/versioning.md                        | 108 ++++++++++++++
 MAINTAINERS                                        |   1 -
 buildlib/RDMA_BuildType.cmake                      |  41 ++++++
 buildlib/config.h.in                               |  16 +-
 buildlib/rdma_functions.cmake                      | 162 ++++++---------------
 ibacm/CMakeLists.txt                               |   7 +-
 ibacm/ibacm.init.in                                |   2 +-
 ibacm/linux/osd.h                                  |   2 +-
 ibacm/man/{ibacm_prov.7 => ibacm_prov.7.in}        |   4 +-
 ibacm/src/acm.c                                    |   6 +-
 ibacm/src/acme.c                                   |  10 +-
 ibacm/src/libacm.c                                 |   2 +-
 iwpmd/CMakeLists.txt                               |   4 +-
 iwpmd/{iwpmd.1 => iwpmd.1.in}                      |   2 +-
 iwpmd/{iwpmd.conf.5 => iwpmd.conf.5.in}            |   2 +-
 iwpmd/src/iwarp_pm.h                               |   1 -
 iwpmd/src/iwarp_pm_server.c                        |   1 +
 libibcm/src/CMakeLists.txt                         |   4 +-
 libibcm/src/libibcm.map                            |   1 +
 libibumad/src/CMakeLists.txt                       |   4 +-
 libibumad/src/libibumad.map                        |   1 +
 libibverbs/src/CMakeLists.txt                      |   4 +-
 libibverbs/src/init.c                              |  55 +++++--
 libibverbs/src/libibverbs.map                      |   1 +
 libipathverbs/CMakeLists.txt                       |   6 +-
 libipathverbs/dracut_install                       |   2 +-
 libipathverbs/truescale.conf                       |   1 -
 libipathverbs/truescale.conf.in                    |   1 +
 librdmacm/man/CMakeLists.txt                       |   2 +-
 librdmacm/man/{rsocket.7 => rsocket.7.in}          |   4 +-
 librdmacm/src/CMakeLists.txt                       |   4 +-
 librdmacm/src/acm.c                                |   2 +-
 librdmacm/src/cma.h                                |   9 --
 librdmacm/src/librdmacm.map                        |   1 +
 srp_daemon/man/CMakeLists.txt                      |   2 +-
 srp_daemon/man/{srp_daemon.1 => srp_daemon.1.in}   |   8 +-
 srp_daemon/srp_daemon/CMakeLists.txt               |  33 +++--
 srp_daemon/srp_daemon/srp_daemon.c                 |   6 +-
 .../srp_daemon/{srp_daemon.sh => srp_daemon.sh.in} |   4 +-
 srp_daemon/srp_daemon/srpd.in                      |   6 +-
 srp_daemon/srptools.default                        |  14 ++
 srp_daemon/srptools.init                           |  89 +++++++++++
 43 files changed, 468 insertions(+), 214 deletions(-)
 create mode 100644 Documentation/versioning.md
 create mode 100644 buildlib/RDMA_BuildType.cmake
 rename ibacm/man/{ibacm_prov.7 => ibacm_prov.7.in} (95%)
 rename iwpmd/{iwpmd.1 => iwpmd.1.in} (98%)
 rename iwpmd/{iwpmd.conf.5 => iwpmd.conf.5.in} (92%)
 delete mode 100644 libipathverbs/truescale.conf
 create mode 100644 libipathverbs/truescale.conf.in
 rename librdmacm/man/{rsocket.7 => rsocket.7.in} (97%)
 rename srp_daemon/man/{srp_daemon.1 => srp_daemon.1.in} (90%)
 rename srp_daemon/srp_daemon/{srp_daemon.sh => srp_daemon.sh.in} (95%)
 create mode 100644 srp_daemon/srptools.default
 create mode 100644 srp_daemon/srptools.init

-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 01/13] Be consistent about defining NDEBUG

[Jason Gunthorpe]

We setup the two standard CMake build_types so that Release includes
NDEBUG and RelWithDebInfo does not (by default CMake sets it in both).

The recommendation is for packagers to use Release (by setting
-DCMAKE_BUILD_TYPE=Release) and developers use RelWithDebInfo
(the default)

This also replaces the default flags for Release with the RelWithDebInfo,
flags (-O2 -g -DNDEBUG) which is what we consider suitable for packaging.
The CMake default of -O3 is not tested.

Note that all the packaging systems I looked at force NDEBUG into the
CFLAGS.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt                |  5 +++++
 buildlib/RDMA_BuildType.cmake | 41 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 46 insertions(+)
 create mode 100644 buildlib/RDMA_BuildType.cmake

diff --git a/CMakeLists.txt b/CMakeLists.txt
index a829cd477c52..06dd98203349 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -7,6 +7,9 @@
 # Common options passed to cmake are:
 #  -DCMAKE_EXPORT_COMPILE_COMMANDS=1
 #      Write a compile_commands.json file for clang tooling
+#  -DCMAKE_BUILD_TYPE=RelWithDebInfo
+#      Change the optimization level, Debug disables optimization,
+#      Release is for packagers
 #  -DENABLE_VALGRIND=1 (default disabled)
 #      Embed valgrind notations, this has a tiny negative performance impact
 #  -DENABLE_RESOLVE_NEIGH=0 (default enabled)
@@ -58,12 +61,14 @@ include(CheckCCompilerFlag)
 include(CheckIncludeFile)
 include(CheckTypeSize)
 include(RDMA_EnableCStd)
+include(RDMA_BuildType)
 include(RDMA_DoFixup)
 include(publish_headers)
 include(rdma_functions)
 
 #-------------------------
 # Setup the basic C compiler
+RDMA_BuildType()
 include_directories(${BUILD_INCLUDE})
 # FIXME: Eliminate HAVE_CONFIG_H, we always have it.
 add_definitions(-DHAVE_CONFIG_H)
diff --git a/buildlib/RDMA_BuildType.cmake b/buildlib/RDMA_BuildType.cmake
new file mode 100644
index 000000000000..4f7485c397a0
--- /dev/null
+++ b/buildlib/RDMA_BuildType.cmake
@@ -0,0 +1,41 @@
+# COPYRIGHT (c) 2015 Obsidian Research Corporation. See COPYING file
+
+function(RDMA_BuildType)
+  set(build_types Debug Release RelWithDebInfo MinSizeRel)
+
+  # Set the default build type to RelWithDebInfo. Since RDMA is typically used
+  # in performance contexts it doesn't make much sense to have the default build
+  # turn off the optimizer.
+  if(NOT CMAKE_BUILD_TYPE)
+    set(CMAKE_BUILD_TYPE RelWithDebInfo CACHE String
+      "Options are ${build_types}"
+      FORCE
+      )
+    set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS ${build_types})
+  endif()
+
+  # Release should be used by packagers, it is the same as the default RelWithDebInfo,
+  # this means it uses -O2 and -DNDEBUG (not -O3)
+  foreach (language CXX C)
+    set(VAR_TO_MODIFY "CMAKE_${language}_FLAGS_RELEASE")
+    if ("${${VAR_TO_MODIFY}}" STREQUAL "${${VAR_TO_MODIFY}_INIT}")
+      set(${VAR_TO_MODIFY} "${CMAKE_${language}_FLAGS_RELWITHDEBINFO_INIT}"
+	  CACHE STRING "Default flags for Release configuration" FORCE)
+    endif()
+  endforeach()
+
+  # RelWithDebInfo should be used by developers, it is the same as Release but
+  # with the -DNDEBUG removed
+  foreach (language CXX C)
+    set(VAR_TO_MODIFY "CMAKE_${language}_FLAGS_RELWITHDEBINFO")
+    if (${${VAR_TO_MODIFY}} STREQUAL ${${VAR_TO_MODIFY}_INIT})
+	string(REGEX REPLACE "(^| )[/-]D *NDEBUG($| )"
+	  " "
+	  replacement
+	  "${${VAR_TO_MODIFY}}"
+	  )
+	set(${VAR_TO_MODIFY} "${replacement}"
+	    CACHE STRING "Default flags for RelWithDebInfo configuration" FORCE)
+    endif()
+  endforeach()
+endfunction()
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 02/13] Switch valgrind memcheck.h to on by default

[Jason Gunthorpe]

We now recommend that this source be built with valgrind memcheck.h
present, so use it automatically if it is available. Users looking
to remove this tiny overhead can build with -DENABLE_VALGRIND=0

Downstream packagers should ensure the build is done with valgrind
headers available.

NOTE: Fedora/CentOS have shipped with valgrind turn on in their
packaging, so for most users this is no change.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 06dd98203349..be82c2896d75 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -10,8 +10,8 @@
 #  -DCMAKE_BUILD_TYPE=RelWithDebInfo
 #      Change the optimization level, Debug disables optimization,
 #      Release is for packagers
-#  -DENABLE_VALGRIND=1 (default disabled)
-#      Embed valgrind notations, this has a tiny negative performance impact
+#  -DENABLE_VALGRIND=0 (default enabled)
+#      Disable valgrind notations, this has a tiny positive performance impact
 #  -DENABLE_RESOLVE_NEIGH=0 (default enabled)
 #      Do not link to libnl and do not resolve neighbours internally for Ethernet,
 #      and do not build iwpmd.
@@ -189,9 +189,11 @@ RDMA_DoFixup("${HAVE_RDMA_USER_RXE}" "rdma/rdma_user_rxe.h")
 #-------------------------
 # Apply fixups
 
-# FIXME: We should probably always enable memcheck.h, and only selectively
-# turn it off in the real high performance paths. There is no reason umad
-# should ever have memcheck disabled for instance.
+# We prefer to build with valgrind memcheck.h present, but if not, or the user
+# requested valgrind disabled, then replace it with our dummy stub.
+if (NOT DEFINED ENABLE_VALGRIND)
+  set(ENABLE_VALGRIND "ON" CACHE BOOL "Enable use of valgrind annotations")
+endif()
 if (ENABLE_VALGRIND)
   CHECK_INCLUDE_FILE("valgrind/memcheck.h" HAVE_VALGRIND_MEMCHECK)
   CHECK_INCLUDE_FILE("valgrind/drd.h" HAVE_VALGRIND_DRD)
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 03/13] Do not produce libtool .la files

[Jason Gunthorpe]

These were preserved as part of the cmake transition, but no distributor
uses them and we don't need them internally, so time for them to go.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 buildlib/rdma_functions.cmake | 99 -------------------------------------------
 1 file changed, 99 deletions(-)

diff --git a/buildlib/rdma_functions.cmake b/buildlib/rdma_functions.cmake
index ea53f382553f..8fe8828c0e84 100644
--- a/buildlib/rdma_functions.cmake
+++ b/buildlib/rdma_functions.cmake
@@ -151,103 +151,6 @@ function(rdma_alias_man_pages)
   endforeach()
 endfunction()
 
-# For compatability write out a libtool .la file. This is only meaningful if
-# the end user is statically linking, and only if the library has dependent
-# libraries.
-
-# FIXME: it isn't clear how this is actually useful for provider libraries and
-# libibverbs itself, the user must do some trick to get the constructor to run
-# in the provider, at least how to do that should be documented someplace..
-function(rdma_make_libtool_la SHARED STATIC LIBS)
-  get_property(LIB TARGET ${STATIC} PROPERTY OUTPUT_NAME SET)
-  if (LIB)
-    get_target_property(LIB ${STATIC} OUTPUT_NAME)
-  else()
-    set(LIB ${STATIC})
-  endif()
-
-  set(BARE_LAFN "${CMAKE_STATIC_LIBRARY_PREFIX}${LIB}.la")
-  set(BARE_LIBFN "${CMAKE_STATIC_LIBRARY_PREFIX}${LIB}${CMAKE_STATIC_LIBRARY_SUFFIX}")
-
-  get_property(SOLIB TARGET ${SHARED} PROPERTY OUTPUT_NAME SET)
-  if (SOLIB)
-    get_target_property(SOLIB ${SHARED} OUTPUT_NAME)
-  else()
-    set(SOLIB ${SHARED})
-  endif()
-
-  set(DLNAME "${CMAKE_SHARED_LIBRARY_PREFIX}${SOLIB}${CMAKE_SHARED_LIBRARY_SUFFIX}")
-  get_property(TMP TARGET ${SHARED} PROPERTY SOVERSION SET)
-  if (TMP)
-    get_target_property(VERSION ${SHARED} VERSION)
-    get_target_property(SOVERSION ${SHARED} SOVERSION)
-    set(NAMES "${DLNAME}.${VERSION} ${DLNAME}.${SOVERSION} ${DLNAME}")
-    set(DLNAME "${DLNAME}.${SOVERSION}")
-  else()
-    set(NAMES "${DLNAME}")
-    set(DLNAME "${CMAKE_SHARED_LIBRARY_PREFIX}${SOLIB}${CMAKE_SHARED_LIBRARY_SUFFIX}")
-  endif()
-
-  if (LIBS)
-    list(REMOVE_DUPLICATES LIBS)
-    foreach(I ${LIBS})
-      if (I MATCHES "^-l")
-	list(APPEND DEPS "${I}")
-      else()
-	list(APPEND DEPS "-l${I}")
-      endif()
-    endforeach()
-    string(REPLACE ";" " " DEPS "${DEPS}")
-  endif()
-
-  set(LAFN "${BUILD_LIB}/${BARE_LAFN}")
-  file(WRITE ${LAFN}
-    "# ${BARE_LAFN} - a libtool library file\n"
-    "# Generated by cmake\n"
-    "#\n"
-    "# Please DO NOT delete this file!\n"
-    "# It is necessary for linking the library.\n"
-    "\n"
-    "# The name that we can dlopen(3).\n"
-    "dlname='${DLNAME}'\n"
-    "\n"
-    "# Names of this library.\n"
-    "library_names='${NAMES}'\n"
-    "\n"
-    "# The name of the static archive.\n"
-    "old_library='${BARE_LIBFN}'\n"
-    "\n"
-    "# Linker flags that can not go in dependency_libs.\n"
-    "inherited_linker_flags=''\n"
-    "\n"
-    "# Libraries that this one depends upon.\n"
-    "dependency_libs='${DEPS}'\n"
-    "\n"
-    "# Names of additional weak libraries provided by this library\n"
-    "weak_library_names=''\n"
-    "\n"
-    "# Version information for ${CMAKE_STATIC_LIBRARY_PREFIX}${LIB}.\n"
-    # We don't try very hard to emulate this, it isn't used for static linking anyhow
-    "current=${SOVERSION}\n"
-    "age=0\n"
-    "revision=0\n"
-    "\n"
-    "# Is this an already installed library?\n"
-    "installed=yes\n"
-    "\n"
-    "# Should we warn about portability when linking against -modules?\n"
-    "shouldnotlink=no\n"
-    "\n"
-    "# Files to dlopen/dlpreopen\n"
-    "dlopen=''\n"
-    "dlpreopen=''\n"
-    "\n"
-    "# Directory that this library needs to be installed in:\n"
-    "libdir='${CMAKE_INSTALL_FULL_LIBDIR}'\n"
-    )
-  install(FILES ${LAFN} DESTINATION "${CMAKE_INSTALL_LIBDIR}")
-endfunction()
-
 # Finalize the setup of the static libraries by copying the meta information
 # from the shared and setting up the libtool .la files.
 function(rdma_finalize_libs)
@@ -274,7 +177,5 @@ function(rdma_finalize_libs)
       set_target_properties(${STATIC} PROPERTIES LINK_LIBRARIES "${TMP}")
       list(APPEND LIBS "${TMP}")
     endif()
-
-    rdma_make_libtool_la(${SHARED} ${STATIC} "${LIBS}")
   endforeach()
 endfunction()
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 04/13] Do not produce static libraries by default

[Jason Gunthorpe]

We no longer recommend that static libraries are distributed, this
never worked sanely for libibverbs.

Use:
 cmake -DENABLE_STATIC=1

To restore the old behaviour

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt                |  6 ++++++
 buildlib/rdma_functions.cmake | 34 +++++++++++++++++++++-------------
 2 files changed, 27 insertions(+), 13 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index be82c2896d75..9b9e34e5cd9d 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -15,6 +15,8 @@
 #  -DENABLE_RESOLVE_NEIGH=0 (default enabled)
 #      Do not link to libnl and do not resolve neighbours internally for Ethernet,
 #      and do not build iwpmd.
+#  -DENABLE_STATIC=1 (default disabled)
+#      Produce static libraries along with the usual shared libraries.
 
 cmake_minimum_required(VERSION 2.8.11 FATAL_ERROR)
 project(RDMA C)
@@ -66,6 +68,10 @@ include(RDMA_DoFixup)
 include(publish_headers)
 include(rdma_functions)
 
+if (NOT DEFINED ENABLE_STATIC)
+  set(ENABLE_STATIC "OFF" CACHE BOOL "Produce static linking libraries as well as shared libraries.")
+endif()
+
 #-------------------------
 # Setup the basic C compiler
 RDMA_BuildType()
diff --git a/buildlib/rdma_functions.cmake b/buildlib/rdma_functions.cmake
index 8fe8828c0e84..a4ecd7d4f2fe 100644
--- a/buildlib/rdma_functions.cmake
+++ b/buildlib/rdma_functions.cmake
@@ -47,14 +47,16 @@ endfunction()
 # Basic function to produce a standard libary with a GNU LD version script.
 function(rdma_library DEST VERSION_SCRIPT SOVERSION VERSION)
   # Create a static library
-  add_library(${DEST}-static STATIC ${ARGN})
-  set_target_properties(${DEST}-static PROPERTIES
-    OUTPUT_NAME ${DEST}
-    LIBRARY_OUTPUT_DIRECTORY "${BUILD_LIB}")
-  install(TARGETS ${DEST}-static DESTINATION "${CMAKE_INSTALL_LIBDIR}")
-
-  list(APPEND RDMA_STATIC_LIBS ${DEST} ${DEST}-static)
-  set(RDMA_STATIC_LIBS "${RDMA_STATIC_LIBS}" CACHE INTERNAL "")
+  if (ENABLE_STATIC)
+    add_library(${DEST}-static STATIC ${ARGN})
+    set_target_properties(${DEST}-static PROPERTIES
+      OUTPUT_NAME ${DEST}
+      LIBRARY_OUTPUT_DIRECTORY "${BUILD_LIB}")
+    install(TARGETS ${DEST}-static DESTINATION "${CMAKE_INSTALL_LIBDIR}")
+
+    list(APPEND RDMA_STATIC_LIBS ${DEST} ${DEST}-static)
+    set(RDMA_STATIC_LIBS "${RDMA_STATIC_LIBS}" CACHE INTERNAL "")
+  endif()
 
   # Create a shared library
   add_library(${DEST} SHARED ${ARGN})
@@ -85,12 +87,14 @@ function(rdma_provider DEST)
   # FIXME: This is probably pointless, the provider library has no symbols so
   # what good is it? Presumably it should be used with -Wl,--whole-archive,
   # but we don't have any directions on how to make static linking work..
-  add_library(${DEST} STATIC ${ARGN})
-  set_target_properties(${DEST} PROPERTIES LIBRARY_OUTPUT_DIRECTORY "${BUILD_LIB}")
-  install(TARGETS ${DEST} DESTINATION "${CMAKE_INSTALL_LIBDIR}")
+  if (ENABLE_STATIC)
+    add_library(${DEST} STATIC ${ARGN})
+    set_target_properties(${DEST} PROPERTIES LIBRARY_OUTPUT_DIRECTORY "${BUILD_LIB}")
+    install(TARGETS ${DEST} DESTINATION "${CMAKE_INSTALL_LIBDIR}")
 
-  list(APPEND RDMA_STATIC_LIBS ${DEST}-rdmav2 ${DEST})
-  set(RDMA_STATIC_LIBS "${RDMA_STATIC_LIBS}" CACHE INTERNAL "")
+    list(APPEND RDMA_STATIC_LIBS ${DEST}-rdmav2 ${DEST})
+    set(RDMA_STATIC_LIBS "${RDMA_STATIC_LIBS}" CACHE INTERNAL "")
+  endif()
 
   # Create the plugin shared library
   set(DEST ${DEST}-rdmav2)
@@ -155,6 +159,10 @@ endfunction()
 # from the shared and setting up the libtool .la files.
 function(rdma_finalize_libs)
   list(LENGTH RDMA_STATIC_LIBS LEN)
+  if (LEN LESS 2)
+    return()
+  endif()
+
   math(EXPR LEN ${LEN}-1)
   foreach(I RANGE 0 ${LEN} 2)
     list(GET RDMA_STATIC_LIBS ${I} SHARED)
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 05/13] Have cmake run man pages through text substitution

[Jason Gunthorpe]

rsocket.7 had an errant text substitution that never worked, it is
a good idea to have the man pages use the correct paths, so let
us have cmake run them through.

Any man page ending in '.in' will be substituted automatically.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 buildlib/rdma_functions.cmake             | 13 +++++++++++--
 librdmacm/man/CMakeLists.txt              |  2 +-
 librdmacm/man/{rsocket.7 => rsocket.7.in} |  4 ++--
 3 files changed, 14 insertions(+), 5 deletions(-)
 rename librdmacm/man/{rsocket.7 => rsocket.7.in} (97%)

diff --git a/buildlib/rdma_functions.cmake b/buildlib/rdma_functions.cmake
index a4ecd7d4f2fe..260dd18d78df 100644
--- a/buildlib/rdma_functions.cmake
+++ b/buildlib/rdma_functions.cmake
@@ -135,8 +135,17 @@ endfunction()
 # filename
 function(rdma_man_pages)
   foreach(I ${ARGN})
-    string(REGEX REPLACE "^.+[.](.+)$" "\\1" MAN_SECT ${I})
-    install(FILES ${I} DESTINATION "${CMAKE_INSTALL_MANDIR}/man${MAN_SECT}/")
+    if ("${I}" MATCHES "\\.in$")
+      string(REGEX REPLACE "^.+[.](.+)\\.in$" "\\1" MAN_SECT "${I}")
+      string(REGEX REPLACE "^(.+)\\.in$" "\\1" BASE_NAME "${I}")
+      get_filename_component(BASE_NAME "${BASE_NAME}" NAME)
+      rdma_subst_install(FILES "${I}"
+	DESTINATION "${CMAKE_INSTALL_MANDIR}/man${MAN_SECT}/"
+	RENAME "${BASE_NAME}")
+    else()
+      string(REGEX REPLACE "^.+[.](.+)$" "\\1" MAN_SECT "${I}")
+      install(FILES "${I}" DESTINATION "${CMAKE_INSTALL_MANDIR}/man${MAN_SECT}/")
+    endif()
   endforeach()
 endfunction()
 
diff --git a/librdmacm/man/CMakeLists.txt b/librdmacm/man/CMakeLists.txt
index 791c98265ad0..d6cc106751a4 100644
--- a/librdmacm/man/CMakeLists.txt
+++ b/librdmacm/man/CMakeLists.txt
@@ -58,7 +58,7 @@ rdma_man_pages(
   riostream.1
   rping.1
   # FIXME: rsocket has a text substitution but nothing ever filled it in.
-  rsocket.7
+  rsocket.7.in
   rstream.1
   ucmatose.1
   udaddy.1
diff --git a/librdmacm/man/rsocket.7 b/librdmacm/man/rsocket.7.in
similarity index 97%
rename from librdmacm/man/rsocket.7
rename to librdmacm/man/rsocket.7.in
index dfb9804d849a..6adf156f4062 100644
--- a/librdmacm/man/rsocket.7
+++ b/librdmacm/man/rsocket.7.in
@@ -129,7 +129,7 @@ fork off a process to handle the new connection.
 .P
 rsockets uses configuration files that give an administrator control
 over the default settings used by rsockets.  Use files under
-%sysconfig%/rdma/rsocket as shown:
+@CMAKE_INSTALL_FULL_SYSCONFDIR@/rdma/rsocket as shown:
 .P
 .P
 mem_default - default size of receive buffer(s)
@@ -149,7 +149,7 @@ polling_time - default number of microseconds to poll for data before waiting
 All configuration files should contain a single integer value.  Values may
 be set by issuing a command similar to the following example.
 .P
-echo 1000000 > /etc/rdma/rsocket/mem_default
+echo 1000000 > @CMAKE_INSTALL_FULL_SYSCONFDIR@/rdma/rsocket/mem_default
 .P
 If configuration files are not available, rsockets uses internal defaults.
 Applications can override default values programmatically through the
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 06/13] Consolidate definitions for paths into config.h

[Jason Gunthorpe]

config.h is the only place we pass through cmake substitution,
so it is the only place that can define the various filesystem
paths.

This patch handles the C code portions that use paths.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt                     | 10 ++++++++++
 buildlib/config.h.in               | 13 +++++++++++--
 ibacm/CMakeLists.txt               |  5 -----
 ibacm/linux/osd.h                  |  2 +-
 ibacm/src/acm.c                    |  6 +++---
 ibacm/src/acme.c                   | 10 +++++-----
 ibacm/src/libacm.c                 |  2 +-
 iwpmd/src/iwarp_pm.h               |  1 -
 iwpmd/src/iwarp_pm_server.c        |  1 +
 librdmacm/src/acm.c                |  2 +-
 librdmacm/src/cma.h                |  9 ---------
 srp_daemon/srp_daemon/srp_daemon.c |  4 ++--
 12 files changed, 35 insertions(+), 30 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 9b9e34e5cd9d..713788f188e3 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -53,6 +53,16 @@ set(CMAKE_INSTALL_SYSTEMD_SERVICEDIR "${CMAKE_INSTALL_PREFIX}/lib/systemd"
 set(ACM_PROVIDER_DIR "${CMAKE_INSTALL_FULL_LIBDIR}/ibacm"
   CACHE PATH "Location for ibacm provider plugin shared library files.")
 
+# Allow the 'run' dir to be configurable, this historically has been /var/run, but
+# some systems now use /run/
+set(CMAKE_INSTALL_RUNDIR "var/run"
+  CACHE PATH "Location for runtime information, typically /var/run, or /run")
+if(NOT IS_ABSOLUTE ${CMAKE_INSTALL_RUNDIR})
+  set(CMAKE_INSTALL_FULL_RUNDIR "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_RUNDIR}")
+else()
+  set(CMAKE_INSTALL_FULL_RUNDIR "${CMAKE_INSTALL_RUNDIR}")
+endif()
+
 #-------------------------
 # Load CMake components
 set(BUILDLIB "${CMAKE_SOURCE_DIR}/buildlib")
diff --git a/buildlib/config.h.in b/buildlib/config.h.in
index 984d51bfaa09..bbc279d3c9c3 100644
--- a/buildlib/config.h.in
+++ b/buildlib/config.h.in
@@ -17,9 +17,18 @@
 #define HAVE_VALGRIND_MEMCHECK_H 1
 #define INCLUDE_VALGRIND 1
 
-#define SYSCONFDIR "@CMAKE_INSTALL_FULL_SYSCONFDIR@"
-
 #define IBV_CONFIG_DIR "@CONFIG_DIR@"
+#define RS_CONF_DIR "@CMAKE_INSTALL_FULL_SYSCONFDIR@/rdma/rsocket"
+#define IWPM_CONFIG_FILE "@CMAKE_INSTALL_FULL_SYSCONFDIR@/iwpmd.conf"
+
+#define SRP_DEAMON_CONFIG_FILE "@CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf"
+
+#define ACM_CONF_DIR "@CMAKE_INSTALL_FULL_SYSCONFDIR@/rdma"
+#define IBACM_LIB_PATH "@ACM_PROVIDER_DIR@"
+#define IBACM_BIN_PATH "@CMAKE_INSTALL_FULL_BINDIR@"
+#define IBACM_PID_FILE "@CMAKE_INSTALL_FULL_RUNDIR@/ibacm.pid"
+#define IBACM_PORT_FILE "@CMAKE_INSTALL_FULL_RUNDIR@/ibacm.port"
+#define IBACM_LOG_FILE "@CMAKE_INSTALL_FULL_LOCALSTATEDIR@/log/ibacm.log"
 
 // FIXME This has been supported in compilers forever, we should just fail to build on such old systems.
 #cmakedefine HAVE_FUNC_ATTRIBUTE_ALWAYS_INLINE 1
diff --git a/ibacm/CMakeLists.txt b/ibacm/CMakeLists.txt
index 8887c13af463..857a6e82320a 100644
--- a/ibacm/CMakeLists.txt
+++ b/ibacm/CMakeLists.txt
@@ -3,11 +3,6 @@ publish_headers(infiniband
   include/infiniband/acm_prov.h
   )
 
-# FIXME: To config.h
-add_definitions("-DIBACM_CONFIG_PATH=\"${CMAKE_INSTALL_FULL_SYSCONFDIR}/rdma\"")
-add_definitions("-DIBACM_LIB_PATH=\"${ACM_PROVIDER_DIR}\"")
-add_definitions("-DIBACM_BIN_PATH=\"${CMAKE_INSTALL_FULL_BINDIR}\"")
-
 # FIXME: Fixup the include scheme to not require all these -Is
 include_directories("include")
 include_directories("src")
diff --git a/ibacm/linux/osd.h b/ibacm/linux/osd.h
index 5ca4c6f554c5..83d31b3589c5 100644
--- a/ibacm/linux/osd.h
+++ b/ibacm/linux/osd.h
@@ -31,6 +31,7 @@
 #if !defined(OSD_H)
 #define OSD_H
 
+#include <config.h>
 #include <stdlib.h>
 #include <string.h>
 #include <stdio.h>
@@ -46,7 +47,6 @@
 #include <sys/time.h>
 #include <netinet/in.h>
 
-#define ACM_CONF_DIR  IBACM_CONFIG_PATH
 #define ACM_ADDR_FILE "ibacm_addr.cfg"
 #define ACM_OPTS_FILE "ibacm_opts.cfg"
 
diff --git a/ibacm/src/acm.c b/ibacm/src/acm.c
index ab1269f71d02..4650421b81b7 100644
--- a/ibacm/src/acm.c
+++ b/ibacm/src/acm.c
@@ -226,9 +226,9 @@ static struct sa_data {
 static char *acme = IBACM_BIN_PATH "/ib_acme -A";
 static char *opts_file = ACM_CONF_DIR "/" ACM_OPTS_FILE;
 static char *addr_file = ACM_CONF_DIR "/" ACM_ADDR_FILE;
-static char log_file[128] = "/var/log/ibacm.log";
+static char log_file[128] = IBACM_LOG_FILE;
 static int log_level = 0;
-static char lock_file[128] = "/var/run/ibacm.pid";
+static char lock_file[128] = IBACM_PID_FILE;
 static short server_port = 6125;
 static int support_ips_in_addr_cfg = 0;
 static char prov_lib_path[256] = IBACM_LIB_PATH;
@@ -578,7 +578,7 @@ static void acm_init_server(void)
 		atomic_init(&client_array[i].refcnt);
 	}
 
-	if (!(f = fopen("/var/run/ibacm.port", "w"))) {
+	if (!(f = fopen(IBACM_PORT_FILE, "w"))) {
 		acm_log(0, "notice - cannot publish ibacm port number\n");
 		return;
 	}
diff --git a/ibacm/src/acme.c b/ibacm/src/acme.c
index f1b0d010625e..4d9003047506 100644
--- a/ibacm/src/acme.c
+++ b/ibacm/src/acme.c
@@ -130,9 +130,9 @@ static void gen_opts_temp(FILE *f)
 	fprintf(f, "# Examples:\n");
 	fprintf(f, "# log_file stdout\n");
 	fprintf(f, "# log_file stderr\n");
-	fprintf(f, "# log_file /var/log/ibacm.log\n");
+	fprintf(f, "# log_file %s\n", IBACM_LOG_FILE);
 	fprintf(f, "\n");
-	fprintf(f, "log_file /var/log/ibacm.log\n");
+	fprintf(f, "log_file %s\n", IBACM_LOG_FILE);
 	fprintf(f, "\n");
 	fprintf(f, "# log_level:\n");
 	fprintf(f, "# Indicates the amount of detailed data written to the log file.  Log levels\n");
@@ -147,7 +147,7 @@ static void gen_opts_temp(FILE *f)
 	fprintf(f, "# Specifies the location of the ACM lock file used to ensure that only a\n");
 	fprintf(f, "# single instance of ACM is running.\n");
 	fprintf(f, "\n");
-	fprintf(f, "lock_file /var/run/ibacm.pid\n");
+	fprintf(f, "lock_file %s\n", IBACM_PID_FILE);
 	fprintf(f, "\n");
 	fprintf(f, "# addr_prot:\n");
 	fprintf(f, "# Default resolution protocol to resolve IP addresses into IB GIDs.\n");
@@ -276,7 +276,7 @@ static void gen_opts_temp(FILE *f)
 	fprintf(f, "# the ACM cache.  This option is only valid if route_preload\n");
 	fprintf(f, "# indicates that routing data should be read from a file.\n");
 	fprintf(f, "# Default is %s/ibacm_route.data\n", ACM_CONF_DIR);
-	fprintf(f, "# route_data_file /etc/rdma/ibacm_route.data\n");
+	fprintf(f, "# route_data_file %s/ibacm_route.data\n", ACM_CONF_DIR);
 	fprintf(f, "\n");
 	fprintf(f, "# addr_preload:\n");
 	fprintf(f, "# Specifies if the ACM address cache should be preloaded, or built on demand.\n");
@@ -292,7 +292,7 @@ static void gen_opts_temp(FILE *f)
 	fprintf(f, "# the ACM cache.  This option is only valid if addr_preload\n");
 	fprintf(f, "# indicates that address data should be read from a file.\n");
 	fprintf(f, "# Default is %s/ibacm_hosts.data\n", ACM_CONF_DIR);
-	fprintf(f, "# addr_data_file /etc/rdma/ibacm_hosts.data\n");
+	fprintf(f, "# addr_data_file %s/ibacm_hosts.data\n", ACM_CONF_DIR);
 	fprintf(f, "\n");
 	fprintf(f, "# support_ips_in_addr_cfg:\n");
 	fprintf(f, "# If 1 continue to read IP addresses from ibacm_addr.cfg\n");
diff --git a/ibacm/src/libacm.c b/ibacm/src/libacm.c
index 95e562deb031..3ad1db1d307b 100644
--- a/ibacm/src/libacm.c
+++ b/ibacm/src/libacm.c
@@ -48,7 +48,7 @@ static void acm_set_server_port(void)
 {
 	FILE *f;
 
-	if ((f = fopen("/var/run/ibacm.port", "r"))) {
+	if ((f = fopen(IBACM_PORT_FILE, "r"))) {
 		if (fscanf(f, "%hu", (unsigned short *) &server_port) != 1) 
 			printf("Failed to read server port\n"); 
 		fclose(f);
diff --git a/iwpmd/src/iwarp_pm.h b/iwpmd/src/iwarp_pm.h
index bcf88ca30433..5b7c6136f386 100644
--- a/iwpmd/src/iwarp_pm.h
+++ b/iwpmd/src/iwarp_pm.h
@@ -85,7 +85,6 @@
 #define IWPM_IFNAME_SIZE   16
 #define IWPM_IPADDR_SIZE   16
 
-#define IWPM_CONFIG_FILE "/etc/iwpmd.conf"
 #define IWPM_PARAM_NUM 1
 #define IWPM_PARAM_NAME_LEN 64
 
diff --git a/iwpmd/src/iwarp_pm_server.c b/iwpmd/src/iwarp_pm_server.c
index e1c2a9c660bb..39e0aa2c3a36 100644
--- a/iwpmd/src/iwarp_pm_server.c
+++ b/iwpmd/src/iwarp_pm_server.c
@@ -31,6 +31,7 @@
  *
  */
 
+#include "config.h"
 #include "iwarp_pm.h"
 
 const char iwpm_ulib_name [] = "iWarpPortMapperUser";
diff --git a/librdmacm/src/acm.c b/librdmacm/src/acm.c
index f0da01e6d286..75d9d8cf487e 100644
--- a/librdmacm/src/acm.c
+++ b/librdmacm/src/acm.c
@@ -120,7 +120,7 @@ static int ucma_set_server_port(void)
 {
 	FILE *f;
 
-	if ((f = fopen("/var/run/ibacm.port", "r" STREAM_CLOEXEC))) {
+	if ((f = fopen(IBACM_PORT_FILE, "r" STREAM_CLOEXEC))) {
 		fscanf(f, "%" SCNu16, &server_port);
 		fclose(f);
 	}
diff --git a/librdmacm/src/cma.h b/librdmacm/src/cma.h
index 98eba8dc21e1..091d104655c5 100644
--- a/librdmacm/src/cma.h
+++ b/librdmacm/src/cma.h
@@ -179,13 +179,4 @@ struct ib_connect_hdr {
 #define cma_dst_ip6 dst_addr[0]
 };
 
-#ifndef SYSCONFDIR
-#define SYSCONFDIR "/etc"
-#endif
-#ifndef RDMADIR
-#define RDMADIR "rdma"
-#endif
-#define RDMA_CONF_DIR  SYSCONFDIR "/" RDMADIR
-#define RS_CONF_DIR RDMA_CONF_DIR "/rsocket"
-
 #endif /* CMA_H */
diff --git a/srp_daemon/srp_daemon/srp_daemon.c b/srp_daemon/srp_daemon/srp_daemon.c
index dfc976b8f6db..70764e00f851 100644
--- a/srp_daemon/srp_daemon/srp_daemon.c
+++ b/srp_daemon/srp_daemon/srp_daemon.c
@@ -226,7 +226,7 @@ static void usage(const char *argv0)
 	fprintf(stderr, "-R <rescan time>	perform complete Rescan every <rescan time> seconds\n");
 	fprintf(stderr, "-T <retry timeout>	Retries to connect to existing target after Timeout of <retry timeout> seconds\n");
 	fprintf(stderr, "-l <tl_retry timeout>	Transport retry count before failing IO. should be in range [2..7], (default 2)\n");
-	fprintf(stderr, "-f <rules file>	use rules File to set to which target(s) to connect (default: /etc/srp_daemon.conf\n");
+	fprintf(stderr, "-f <rules file>	use rules File to set to which target(s) to connect (default: " SRP_DEAMON_CONFIG_FILE ")\n");
 	fprintf(stderr, "-t <timeout>		Timeout for mad response in milliseconds\n");
 	fprintf(stderr, "-r <retries>		number of send Retries for each mad\n");
 	fprintf(stderr, "-n 			New connection command format - use also initiator extension\n");
@@ -1622,7 +1622,7 @@ static int get_config(struct config_t *conf, int argc, char *argv[])
 	conf->retry_timeout 		= 20;
 	conf->add_target_file  		= NULL;
 	conf->print_initiator_ext	= 0;
-	conf->rules_file		= "/etc/srp_daemon.conf";
+	conf->rules_file		= SRP_DEAMON_CONFIG_FILE;
 	conf->rules			= NULL;
 	conf->tl_retry_count		= 0;
 
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 07/13] Pass ancillary files that use paths through cmake substitution

[Jason Gunthorpe]

This removes hardwired paths from the documentation and broadly
makes the documentation and scripts match what the C code is now
doing.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 ibacm/CMakeLists.txt                                      | 2 +-
 ibacm/ibacm.init.in                                       | 2 +-
 ibacm/man/{ibacm_prov.7 => ibacm_prov.7.in}               | 4 ++--
 iwpmd/CMakeLists.txt                                      | 4 ++--
 iwpmd/{iwpmd.1 => iwpmd.1.in}                             | 2 +-
 iwpmd/{iwpmd.conf.5 => iwpmd.conf.5.in}                   | 2 +-
 srp_daemon/man/CMakeLists.txt                             | 2 +-
 srp_daemon/man/{srp_daemon.1 => srp_daemon.1.in}          | 8 ++++----
 srp_daemon/srp_daemon/CMakeLists.txt                      | 4 +++-
 srp_daemon/srp_daemon/{srp_daemon.sh => srp_daemon.sh.in} | 4 ++--
 srp_daemon/srp_daemon/srpd.in                             | 6 +++---
 11 files changed, 21 insertions(+), 19 deletions(-)
 rename ibacm/man/{ibacm_prov.7 => ibacm_prov.7.in} (95%)
 rename iwpmd/{iwpmd.1 => iwpmd.1.in} (98%)
 rename iwpmd/{iwpmd.conf.5 => iwpmd.conf.5.in} (92%)
 rename srp_daemon/man/{srp_daemon.1 => srp_daemon.1.in} (90%)
 rename srp_daemon/srp_daemon/{srp_daemon.sh => srp_daemon.sh.in} (95%)

diff --git a/ibacm/CMakeLists.txt b/ibacm/CMakeLists.txt
index 857a6e82320a..376a3a89ddae 100644
--- a/ibacm/CMakeLists.txt
+++ b/ibacm/CMakeLists.txt
@@ -52,7 +52,7 @@ rdma_man_pages(
   man/ib_acme.1
   man/ibacm.1
   man/ibacm.7
-  man/ibacm_prov.7
+  man/ibacm_prov.7.in
   )
 
 # FIXME: update the .init.in
diff --git a/ibacm/ibacm.init.in b/ibacm/ibacm.init.in
index aea262b6768a..0fb0011f3953 100644
--- a/ibacm/ibacm.init.in
+++ b/ibacm/ibacm.init.in
@@ -19,7 +19,7 @@
 #	host route lookups.
 ### END INIT INFO
 
-pidfile=/var/run/ibacm.pid
+pidfile=@CMAKE_INSTALL_FULL_RUNDIR@/ibacm.pid
 subsys=/var/lock/subsys/ibacm
 
 daemon() { /sbin/daemon   ${1+"$@"}; }
diff --git a/ibacm/man/ibacm_prov.7 b/ibacm/man/ibacm_prov.7.in
similarity index 95%
rename from ibacm/man/ibacm_prov.7
rename to ibacm/man/ibacm_prov.7.in
index d04617e837c8..9e79d58f639e 100644
--- a/ibacm/man/ibacm_prov.7
+++ b/ibacm/man/ibacm_prov.7.in
@@ -12,7 +12,7 @@ To add a provider to the ibacm core service, the provider must
 .TP
 1. be implemented as a shared library;
 .TP
-2. be installed under a configured directory, eg., /usr/lib64/ibacm/;
+2. be installed under a configured directory, eg., @ACM_PROVIDER_DIR@;
 .TP
 3  export a function provider_query() that returns a pointer to its provider info 
 and version info.
@@ -66,7 +66,7 @@ Non-related sections should be ignored.
 .P
 Some helper functions are also exported by the ibacm core. For example, the
 acm_log define (or the acm_write() function) can be used to log messages into
-ibacm's log file (default /var/log/ibacm.log).  For details, refer to
+ibacm's log file (default @CMAKE_INSTALL_FULL_LOCALSTATEDIR@/log/ibacm.log).  For details, refer to
 the acm_prov.h file.
 .SH "NOTES"
 A provider should always set the version in its provider info structure as the
diff --git a/iwpmd/CMakeLists.txt b/iwpmd/CMakeLists.txt
index 21208dd5b82a..89a715128076 100644
--- a/iwpmd/CMakeLists.txt
+++ b/iwpmd/CMakeLists.txt
@@ -1,4 +1,4 @@
 rdma_man_pages(
-  iwpmd.1
-  iwpmd.conf.5
+  iwpmd.1.in
+  iwpmd.conf.5.in
   )
diff --git a/iwpmd/iwpmd.1 b/iwpmd/iwpmd.1.in
similarity index 98%
rename from iwpmd/iwpmd.1
rename to iwpmd/iwpmd.1.in
index a4948c509f3b..88eece7f6522 100644
--- a/iwpmd/iwpmd.1
+++ b/iwpmd/iwpmd.1.in
@@ -51,6 +51,6 @@ to the system message log.
 .P
 SIGTERM/SIGHUP will force iwpmd to exit.
 .SH "FILES"
-/etc/iwpmd.conf
+@CMAKE_INSTALL_FULL_SYSCONFDIR@/iwpmd.conf
 .SH "SEE ALSO"
 rdma_cm(7)
diff --git a/iwpmd/iwpmd.conf.5 b/iwpmd/iwpmd.conf.5.in
similarity index 92%
rename from iwpmd/iwpmd.conf.5
rename to iwpmd/iwpmd.conf.5.in
index 9dfe137d3fab..1baae82b76da 100644
--- a/iwpmd/iwpmd.conf.5
+++ b/iwpmd/iwpmd.conf.5.in
@@ -15,6 +15,6 @@ to communicate with the kernel port map client.  The default is 400MB.
 .SH "EXAMPLES"
 nl_sock_rbuf_size=419430400
 .SH "FILES"
-/etc/iwpmd.conf
+@CMAKE_INSTALL_FULL_SYSCONFDIR@/iwpmd.conf
 .SH "SEE ALSO"
 iwpmd(1)
diff --git a/srp_daemon/man/CMakeLists.txt b/srp_daemon/man/CMakeLists.txt
index f05de60b97aa..cd644a6bd836 100644
--- a/srp_daemon/man/CMakeLists.txt
+++ b/srp_daemon/man/CMakeLists.txt
@@ -1,4 +1,4 @@
 rdma_man_pages(
   ibsrpdm.1
-  srp_daemon.1
+  srp_daemon.1.in
   )
diff --git a/srp_daemon/man/srp_daemon.1 b/srp_daemon/man/srp_daemon.1.in
similarity index 90%
rename from srp_daemon/man/srp_daemon.1
rename to srp_daemon/man/srp_daemon.1.in
index 6117e3679a25..6ee114284780 100644
--- a/srp_daemon/man/srp_daemon.1
+++ b/srp_daemon/man/srp_daemon.1.in
@@ -15,7 +15,7 @@ Each srp_daemon instance operates on one local port. Upon boot it performs a ful
 
 When a new machine joins the fabric, srp_daemon checks if it is a target. When there is a change of capabilities, srp_daemon checks if the machine has turned into a target. When there is an SA change or a timeout expiration, srp_daemon performs a full rescan of the fabric.
 
-For each target srp_daemon finds, it checks if it should connect to this target according to its rules (default rules file is /etc/srp_daemon.conf) and if it is already connected to the local port. If it should connect to this target and if it is not connected yet, srp_daemon can either print the target details or connect to it.
+For each target srp_daemon finds, it checks if it should connect to this target according to its rules (default rules file is @CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf) and if it is already connected to the local port. If it should connect to this target and if it is not connected yet, srp_daemon can either print the target details or connect to it.
 
 .SH OPTIONS
 
@@ -57,7 +57,7 @@ Retries to connect to existing target after \fIretry-Timeout\fR seconds. If -R i
 .TP
 \fB\-f\fR \fIrules-File\fR
 Decide to which targets to connect according to the rules in \fIrules-File\fR. 
-If \fB\-f\fR is not specified, uses the default rules file /etc/srp_daemon.conf.
+If \fB\-f\fR is not specified, uses the default rules file @CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf.
 Each line in the \fIrules-File\fR is a rule which can be either an allow connection or a disallow connection according to 
 the first character in the line (a or d accordingly). The rest of the line is values for id_ext, ioc_guid, dgid, 
 service_id. Please take a look at the example section for an example of the file. srp_daemon decide whether to allow or disallow each target according  to first rule that match the target. If no rule matches the target, the target is allowed and will be connected. In an allow rule it is possible to set attributes for the connection to the target. Supported attributes are max_cmd_per_lun and max_sect.
@@ -72,7 +72,7 @@ Perform \fIretries\fR retries on each send to MAD (default: 3 retries).
 New format - use also initiator_ext in the connection command.
 
 .SH FILES
-/etc/srp_daemon.conf -
+@CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf -
 Default rules configuration file that indicates to which targets to connect. Can be overridden using the \fB\-f\fR \fIrules-File\fR option. 
 Each line in this file is a rule which can be either an allow connection or a disallow connection according to 
 the first character in the line (a or d accordingly). The rest of the line is values for id_ext, ioc_guid, dgid, 
@@ -86,7 +86,7 @@ srp_daemon -o -c -a		    (Prints the connection commands for the targets in the
 srp_daemon -e -f rules.txt	    (Connects to the targets allowed in the rules file rules.txt)
 
 .nf
-An example for a rules configuration file (such as /etc/srp_daemon.conf)
+An example for a rules configuration file (such as @CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf)
 ------------------------------------------------------------------------
 # Rules file example
 # This is a comment
diff --git a/srp_daemon/srp_daemon/CMakeLists.txt b/srp_daemon/srp_daemon/CMakeLists.txt
index 84d74044186a..fe6b41811b92 100644
--- a/srp_daemon/srp_daemon/CMakeLists.txt
+++ b/srp_daemon/srp_daemon/CMakeLists.txt
@@ -15,7 +15,9 @@ target_link_libraries(srp_daemon
 rdma_install_symlink(srp_daemon "${CMAKE_INSTALL_SBINDIR}/ibsrpdm")
 # FIXME: Why?
 rdma_install_symlink(srp_daemon "${CMAKE_INSTALL_SBINDIR}/run_srp_daemon")
-install(FILES srp_daemon.sh DESTINATION "${CMAKE_INSTALL_SBINDIR}")
+rdma_subst_install(FILES "srp_daemon.sh.in"
+  DESTINATION "${CMAKE_INSTALL_SBINDIR}"
+  RENAME "srp_daemon.sh")
 
 install(FILES logrotate-srp_daemon DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/logrotate.d" RENAME "srp_daemon")
 install(FILES rsyslog-srp_daemon.conf DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/rsyslog.d" RENAME "srp_daemon.conf")
diff --git a/srp_daemon/srp_daemon/srp_daemon.sh b/srp_daemon/srp_daemon/srp_daemon.sh.in
similarity index 95%
rename from srp_daemon/srp_daemon/srp_daemon.sh
rename to srp_daemon/srp_daemon/srp_daemon.sh.in
index 3981a0d4b586..cb0b81efe6e6 100755
--- a/srp_daemon/srp_daemon/srp_daemon.sh
+++ b/srp_daemon/srp_daemon/srp_daemon.sh.in
@@ -30,12 +30,12 @@
 
 shopt -s nullglob
 
-prog=/usr/sbin/srp_daemon
+prog=@CMAKE_INSTALL_FULL_SBINDIR@/srp_daemon
 params=$@
 ibdir="/sys/class/infiniband"
 rescan_interval=60
 pids=""
-pidfile=/var/run/srp_daemon.sh.pid
+pidfile=@CMAKE_INSTALL_FULL_RUNDIR@/srp_daemon.sh.pid
 mypid=$$
 
 trap_handler()
diff --git a/srp_daemon/srp_daemon/srpd.in b/srp_daemon/srp_daemon/srpd.in
index 10692e7e0f05..d77579170a3b 100755
--- a/srp_daemon/srp_daemon/srpd.in
+++ b/srp_daemon/srp_daemon/srpd.in
@@ -4,7 +4,7 @@
 #
 # chkconfig: - 25 75
 # description: Starts/Stops InfiniBand SRP client service
-# config:	/etc/srp_daemon.conf
+# config:	@CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf
 #
 ### BEGIN INIT INFO
 # Provides:       srpd
@@ -32,8 +32,8 @@ fi
 if [ -f $RDMA_CONFIG ]; then
     . $RDMA_CONFIG
 fi
-pidfile=/var/run/srp_daemon.sh.pid
-prog=/usr/sbin/srp_daemon.sh
+pidfile=@CMAKE_INSTALL_FULL_RUNDIR@/srp_daemon.sh.pid
+prog=@CMAKE_INSTALL_FULL_SBINDIR@/srp_daemon.sh
 
 checkpid() {
     [ -e "/proc/$1" ]
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 08/13] srp_daemon: Move lock file into /var/run/

[Jason Gunthorpe]

'/var/tmp' is an inappropriate places for lock files of this nature,
they belong in /var/run. /var/lock does not seem suitable because
this lock is not against a basic device node.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 buildlib/config.h.in               | 1 +
 srp_daemon/srp_daemon/srp_daemon.c | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/buildlib/config.h.in b/buildlib/config.h.in
index bbc279d3c9c3..78994a39d7ad 100644
--- a/buildlib/config.h.in
+++ b/buildlib/config.h.in
@@ -22,6 +22,7 @@
 #define IWPM_CONFIG_FILE "@CMAKE_INSTALL_FULL_SYSCONFDIR@/iwpmd.conf"
 
 #define SRP_DEAMON_CONFIG_FILE "@CMAKE_INSTALL_FULL_SYSCONFDIR@/srp_daemon.conf"
+#define SRP_DEAMON_LOCK_PREFIX "@CMAKE_INSTALL_FULL_RUNDIR@/srp_daemon"
 
 #define ACM_CONF_DIR "@CMAKE_INSTALL_FULL_SYSCONFDIR@/rdma"
 #define IBACM_LIB_PATH "@ACM_PROVIDER_DIR@"
diff --git a/srp_daemon/srp_daemon/srp_daemon.c b/srp_daemon/srp_daemon/srp_daemon.c
index 70764e00f851..f16674dbfafa 100644
--- a/srp_daemon/srp_daemon/srp_daemon.c
+++ b/srp_daemon/srp_daemon/srp_daemon.c
@@ -131,7 +131,7 @@ static int check_process_uniqueness(struct config_t *conf)
 	char path[256];
 	int fd;
 
-	snprintf(path, sizeof(path), "/var/tmp/srp_daemon_%s_%d",
+	snprintf(path, sizeof(path), SRP_DEAMON_LOCK_PREFIX "_%s_%d",
 		 conf->dev_name, conf->port_num);
 
 	if ((fd = open(path, O_CREAT|O_RDWR,
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 09/13] verbs: Move the providers into /usr/lib.../libibverbs by default

[Jason Gunthorpe]

The Debian packaging has always used this path, provide
official support for this configuration so Debian does not
rely on the absolute path in the .driver file, which breaks biarch.

Since there is no reason for the providers to be in the system
library search path (they export no symbols, and have no soname)
make this the default configuration.

The old behaviour can be restored by using:

 cmake -DVERBS_PROVIDER_DIR=''

This continues to support out-of-tree drivers by searching both
the provider path and the system library path if an unqualified
name is given in the .driver file.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt                |  6 +++++
 buildlib/config.h.in          |  2 ++
 buildlib/rdma_functions.cmake | 16 ++++++++-----
 libibverbs/src/init.c         | 55 ++++++++++++++++++++++++++++++-------------
 4 files changed, 57 insertions(+), 22 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 713788f188e3..7ae236f88f67 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -17,6 +17,8 @@
 #      and do not build iwpmd.
 #  -DENABLE_STATIC=1 (default disabled)
 #      Produce static libraries along with the usual shared libraries.
+#  -DVERBS_PROVIDER_DIR='' (default /usr/lib.../libibverbs)
+#      Use the historical search path for providers, in the standard system library.
 
 cmake_minimum_required(VERSION 2.8.11 FATAL_ERROR)
 project(RDMA C)
@@ -53,6 +55,10 @@ set(CMAKE_INSTALL_SYSTEMD_SERVICEDIR "${CMAKE_INSTALL_PREFIX}/lib/systemd"
 set(ACM_PROVIDER_DIR "${CMAKE_INSTALL_FULL_LIBDIR}/ibacm"
   CACHE PATH "Location for ibacm provider plugin shared library files.")
 
+# Location to find the provider plugin shared library files
+set(VERBS_PROVIDER_DIR "${CMAKE_INSTALL_FULL_LIBDIR}/libibverbs"
+  CACHE PATH "Location for provider plugin shared library files. If set to empty the system search path is used.")
+
 # Allow the 'run' dir to be configurable, this historically has been /var/run, but
 # some systems now use /run/
 set(CMAKE_INSTALL_RUNDIR "var/run"
diff --git a/buildlib/config.h.in b/buildlib/config.h.in
index 78994a39d7ad..8ae2a4e536c0 100644
--- a/buildlib/config.h.in
+++ b/buildlib/config.h.in
@@ -31,6 +31,8 @@
 #define IBACM_PORT_FILE "@CMAKE_INSTALL_FULL_RUNDIR@/ibacm.port"
 #define IBACM_LOG_FILE "@CMAKE_INSTALL_FULL_LOCALSTATEDIR@/log/ibacm.log"
 
+#define VERBS_PROVIDER_DIR "@VERBS_PROVIDER_DIR@"
+
 // FIXME This has been supported in compilers forever, we should just fail to build on such old systems.
 #cmakedefine HAVE_FUNC_ATTRIBUTE_ALWAYS_INLINE 1
 
diff --git a/buildlib/rdma_functions.cmake b/buildlib/rdma_functions.cmake
index 260dd18d78df..06ccd518c81d 100644
--- a/buildlib/rdma_functions.cmake
+++ b/buildlib/rdma_functions.cmake
@@ -78,11 +78,6 @@ function(rdma_provider DEST)
   file(MAKE_DIRECTORY "${BUILD_LIB}/libibverbs.d/")
   file(WRITE "${BUILD_LIB}/libibverbs.d/${DEST}.driver" "driver ${BUILD_LIB}/${DEST}\n")
 
-  # FIXME: This symlink is provided for compat with the old build, but it
-  # never should have existed in the first place, nothing should use this
-  # name, we can probably remove it.
-  rdma_install_symlink("lib${DEST}-rdmav2.so" "${CMAKE_INSTALL_LIBDIR}/lib${DEST}.so")
-
   # Create a static provider library
   # FIXME: This is probably pointless, the provider library has no symbols so
   # what good is it? Presumably it should be used with -Wl,--whole-archive,
@@ -108,7 +103,16 @@ function(rdma_provider DEST)
   # Provider Plugins do not use SONAME versioning, there is no reason to
   # create the usual symlinks.
 
-  install(TARGETS ${DEST} DESTINATION "${CMAKE_INSTALL_LIBDIR}")
+  if (VERBS_PROVIDER_DIR)
+    install(TARGETS ${DEST} DESTINATION "${VERBS_PROVIDER_DIR}")
+  else()
+    install(TARGETS ${DEST} DESTINATION "${CMAKE_INSTALL_LIBDIR}")
+
+    # FIXME: This symlink is provided for compat with the old build, but it
+    # never should have existed in the first place, nothing should use this
+    # name, we can probably remove it.
+    rdma_install_symlink("lib${DEST}-rdmav2.so" "${CMAKE_INSTALL_LIBDIR}/lib${DEST}.so")
+  endif()
 endfunction()
 
  # Create an installed executable
diff --git a/libibverbs/src/init.c b/libibverbs/src/init.c
index bca0e02d11e1..7ae0fc87d332 100644
--- a/libibverbs/src/init.c
+++ b/libibverbs/src/init.c
@@ -190,33 +190,56 @@ void verbs_register_driver(const char *name, verbs_driver_init_func init_func)
 	register_driver(name, NULL, init_func);
 }
 
+#define __IBV_QUOTE(x)	#x
+#define IBV_QUOTE(x)	__IBV_QUOTE(x)
+#define DLOPEN_TRAILER "-" IBV_QUOTE(IBV_DEVICE_LIBRARY_EXTENSION) ".so"
+
 static void load_driver(const char *name)
 {
 	char *so_name;
 	void *dlhandle;
 
-#define __IBV_QUOTE(x)	#x
-#define IBV_QUOTE(x)	__IBV_QUOTE(x)
-
-	if (asprintf(&so_name,
-		     name[0] == '/' ?
-		     "%s-" IBV_QUOTE(IBV_DEVICE_LIBRARY_EXTENSION) ".so" :
-		     "lib%s-" IBV_QUOTE(IBV_DEVICE_LIBRARY_EXTENSION) ".so",
-		     name) < 0) {
-		fprintf(stderr, PFX "Warning: couldn't load driver '%s'.\n",
-			name);
+	/* If the name is an absolute path then open that path after appending
+	   the trailer suffix */
+	if (name[0] == '/') {
+		if (asprintf(&so_name, "%s" DLOPEN_TRAILER, name) < 0)
+			goto out_asprintf;
+		dlhandle = dlopen(so_name, RTLD_NOW);
+		if (!dlhandle)
+			goto out_dlopen;
+		free(so_name);
 		return;
 	}
 
-	dlhandle = dlopen(so_name, RTLD_NOW);
-	if (!dlhandle) {
-		fprintf(stderr, PFX "Warning: couldn't load driver '%s': %s\n",
-			name, dlerror());
-		goto out;
+	/* If configured with a provider plugin path then try that next */
+	if (sizeof(VERBS_PROVIDER_DIR) >= 1) {
+		if (asprintf(&so_name, VERBS_PROVIDER_DIR "/lib%s" DLOPEN_TRAILER, name) <
+		    0)
+			goto out_asprintf;
+		dlhandle = dlopen(so_name, RTLD_NOW);
+		free(so_name);
+		if (dlhandle)
+			return;
 	}
 
-out:
+	/* Otherwise use the system libary search path. This is the historical
+	   behavior of libibverbs */
+	if (asprintf(&so_name, "lib%s" DLOPEN_TRAILER, name) < 0)
+		goto out_asprintf;
+	dlhandle = dlopen(so_name, RTLD_NOW);
+	if (!dlhandle)
+		goto out_dlopen;
+	free(so_name);
+	return;
+
+out_asprintf:
+	fprintf(stderr, PFX "Warning: couldn't load driver '%s'.\n", name);
+	return;
+out_dlopen:
+	fprintf(stderr, PFX "Warning: couldn't load driver '%s': %s\n", so_name,
+		dlerror());
 	free(so_name);
+	return;
 }
 
 static void load_drivers(void)
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 10/13] ipathverbs: Move truescale-serdes.cmds to /usr/libexec/

[Jason Gunthorpe]

This is the FHS defined place for non-user runnable helper
programs.

Debian forbids the use of /usr/libexec/ so we provide
substitution support to let cmake customize this.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 libipathverbs/CMakeLists.txt    | 6 ++++--
 libipathverbs/dracut_install    | 2 +-
 libipathverbs/truescale.conf    | 1 -
 libipathverbs/truescale.conf.in | 1 +
 4 files changed, 6 insertions(+), 4 deletions(-)
 delete mode 100644 libipathverbs/truescale.conf
 create mode 100644 libipathverbs/truescale.conf.in

diff --git a/libipathverbs/CMakeLists.txt b/libipathverbs/CMakeLists.txt
index 359438ef0e49..bde405d0d776 100644
--- a/libipathverbs/CMakeLists.txt
+++ b/libipathverbs/CMakeLists.txt
@@ -1,4 +1,6 @@
-install(FILES truescale.conf DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/modprobe.d/")
+rdma_subst_install(FILES "truescale.conf.in"
+  DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/modprobe.d/"
+  RENAME "truescale.conf")
 install(FILES truescale-serdes.cmds
-  DESTINATION "${CMAKE_INSTALL_SBINDIR}"
+  DESTINATION "${CMAKE_INSTALL_LIBEXECDIR}"
   PERMISSIONS OWNER_WRITE OWNER_READ GROUP_READ WORLD_READ OWNER_EXECUTE GROUP_EXECUTE WORLD_EXECUTE)
diff --git a/libipathverbs/dracut_install b/libipathverbs/dracut_install
index 131adfa036b8..a7ef490ea40f 100644
--- a/libipathverbs/dracut_install
+++ b/libipathverbs/dracut_install
@@ -1,7 +1,7 @@
 #!/bin/bash
 
 inst /etc/modprobe.d/truescale.conf
-inst /usr/sbin/truescale-serdes.cmds
+inst /usr/libexec/truescale-serdes.cmds
 
 # All files needed by truescale-serdes.cmds need to be present here
 inst /sbin/lspci
diff --git a/libipathverbs/truescale.conf b/libipathverbs/truescale.conf
deleted file mode 100644
index 8ed227c91f86..000000000000
diff --git a/libipathverbs/truescale.conf.in b/libipathverbs/truescale.conf.in
new file mode 100644
index 000000000000..e2827d956015
--- /dev/null
+++ b/libipathverbs/truescale.conf.in
@@ -0,0 +1 @@
+install ib_qib modprobe -i ib_qib $CMDLINE_OPTS && @CMAKE_INSTALL_FULL_LIBEXECDIR@/truescale-serdes.cmds start
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Jason Gunthorpe]

Necessary to reproduce the Debian packaging.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt                       |  3 ++
 srp_daemon/srp_daemon/CMakeLists.txt | 29 ++++++++----
 srp_daemon/srptools.default          | 14 ++++++
 srp_daemon/srptools.init             | 89 ++++++++++++++++++++++++++++++++++++
 4 files changed, 126 insertions(+), 9 deletions(-)
 create mode 100644 srp_daemon/srptools.default
 create mode 100644 srp_daemon/srptools.init

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 7ae236f88f67..805b25bb8669 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -69,6 +69,9 @@ else()
   set(CMAKE_INSTALL_FULL_RUNDIR "${CMAKE_INSTALL_RUNDIR}")
 endif()
 
+set(DISTRO_FLAVOUR "None" CACHE
+  STRING "Flavour of distribution to install for. This primarily impacts the init.d scripts installed.")
+
 #-------------------------
 # Load CMake components
 set(BUILDLIB "${CMAKE_SOURCE_DIR}/buildlib")
diff --git a/srp_daemon/srp_daemon/CMakeLists.txt b/srp_daemon/srp_daemon/CMakeLists.txt
index fe6b41811b92..4ad4edaf8c33 100644
--- a/srp_daemon/srp_daemon/CMakeLists.txt
+++ b/srp_daemon/srp_daemon/CMakeLists.txt
@@ -23,12 +23,23 @@ install(FILES logrotate-srp_daemon DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/logr
 install(FILES rsyslog-srp_daemon.conf DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/rsyslog.d" RENAME "srp_daemon.conf")
 install(FILES srp_daemon.conf DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}")
 
-# FIXME: The ib init.d file should really be included in rdma-core as well.
-set(RDMA_SERVICE "openibd" CACHE STRING "init.d file service name to order srpd after")
-# NOTE: These defaults are for CentOS, packagers should override.
-set(SRP_DEFAULT_START "2 3 4 5" CACHE STRING "Default-Start service data for srpd")
-set(SRP_DEFAULT_STOP "0 1 6" CACHE STRING "Default-Stop service data for srpd")
-configure_file(srpd.in "${CMAKE_CURRENT_BINARY_DIR}/srpd")
-install(FILES "${CMAKE_CURRENT_BINARY_DIR}/srpd"
-  DESTINATION "${CMAKE_INSTALL_INITDDIR}"
-  PERMISSIONS OWNER_WRITE OWNER_READ GROUP_READ WORLD_READ OWNER_EXECUTE GROUP_EXECUTE WORLD_EXECUTE)
+if ("${DISTRO_FLAVOUR}" STREQUAL "Debian")
+  # Debian version of the initscript system
+  install(FILES "srptools.init"
+    DESTINATION "${CMAKE_INSTALL_INITDDIR}"
+    RENAME "srptools"
+    PERMISSIONS OWNER_WRITE OWNER_READ GROUP_READ WORLD_READ OWNER_EXECUTE GROUP_EXECUTE WORLD_EXECUTE)
+  install(FILES "srptools.default"
+    DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}/default/"
+    RENAME "srptools")
+else()
+  # FIXME: The ib init.d file should really be included in rdma-core as well.
+  set(RDMA_SERVICE "openibd" CACHE STRING "init.d file service name to order srpd after")
+  # NOTE: These defaults are for CentOS, packagers should override.
+  set(SRP_DEFAULT_START "2 3 4 5" CACHE STRING "Default-Start service data for srpd")
+  set(SRP_DEFAULT_STOP "0 1 6" CACHE STRING "Default-Stop service data for srpd")
+  configure_file(srpd.in "${CMAKE_CURRENT_BINARY_DIR}/srpd")
+  install(FILES "${CMAKE_CURRENT_BINARY_DIR}/srpd"
+    DESTINATION "${CMAKE_INSTALL_INITDDIR}"
+    PERMISSIONS OWNER_WRITE OWNER_READ GROUP_READ WORLD_READ OWNER_EXECUTE GROUP_EXECUTE WORLD_EXECUTE)
+endif()
diff --git a/srp_daemon/srptools.default b/srp_daemon/srptools.default
new file mode 100644
index 000000000000..81e84f4a5cba
--- /dev/null
+++ b/srp_daemon/srptools.default
@@ -0,0 +1,14 @@
+#How often should srpdeamon  rescan the fabric (seconds)
+RETRIES=60
+
+#Where should srp-deamon log to
+LOG=/var/log/srp_daemon.log
+
+# What ports should srp-deamon be started on.
+# Format is CA:port
+# ALL or NONE will run on all ports on none
+# respectively
+
+PORTS=NONE
+#PORTS=ALL
+#PORTS="mthca0:1 mlx4_0:2"
diff --git a/srp_daemon/srptools.init b/srp_daemon/srptools.init
new file mode 100644
index 000000000000..2c1a140ccbc3
--- /dev/null
+++ b/srp_daemon/srptools.init
@@ -0,0 +1,89 @@
+#!/bin/bash
+### BEGIN INIT INFO
+# Provides:          srptools
+# Required-Start:    $remote_fs $syslog
+# Required-Stop:     $remote_fs $syslog
+# Default-Start:     2 3 4 5
+# Default-Stop:      0 1 6
+# Short-Description: Discovers SRP scsi targets.
+# Description:       Discovers SRP scsi over infiniband targets.
+### END INIT INFO
+
+[ -x /usr/sbin/srp_daemon ] || exit 0
+
+IBDIR=/sys/class/infiniband
+
+PORTS=""
+RETRIES=""
+LOG=""
+
+[ -f /etc/default/srptools ] &&  . /etc/default/srptools
+
+start_daemon () {
+
+if [ "$PORTS" = "NONE" ] ; then
+echo "srptools disabled."
+exit 0
+fi
+
+
+if [ "$PORTS" = "ALL" ]  ; then
+    for HCA_ID in `/bin/ls -1 ${IBDIR}`
+      do
+      for PORT in `/bin/ls -1 ${IBDIR}/${HCA_ID}/ports/`
+        do
+        run_daemon
+      done
+    done
+fi
+
+
+for ADAPTER in $PORTS ; do
+    HCA_ID=`echo $ADAPTER | awk -F: '{print $1}'`
+    PORT=`echo $ADAPTER | awk -F:  '{print $2}'`
+    run_daemon
+done
+}
+
+
+run_daemon() {
+# SRP deamon wedges if we start it on a port which is not up
+
+        STATUS=`/usr/sbin/ibstat $HCA_ID $PORT | grep "State:"`
+
+        if [ "$STATUS" = "State: Active" ] ; then
+            echo "Starting srp on $HCA_ID $PORT"
+
+# srp does not background itself; using the start-stop-daemon background function
+# causes us to lose stdout, which is where it logs to
+            nohup start-stop-daemon --start --quiet -m --pidfile /var/run/srp_daemon.${HCA_ID}.${PORT} \
+            --exec  /usr/sbin/srp_daemon -- -e -c -n -i ${HCA_ID} -p ${PORT} -R ${RETRIES}   >> $LOG 2>&1 &
+            RETVAL=$?
+        fi
+}
+
+stop_daemon () {
+     for HCA_ID in `/bin/ls -1 ${IBDIR}`
+      do
+      for PORT in `/bin/ls -1 ${IBDIR}/${HCA_ID}/ports/`
+        do
+        start-stop-daemon --stop --quiet --oknodo -m --pidfile /var/run/srp_daemon.${HCA_ID}.${PORT}
+        RETVAL=$?
+      done
+    done
+}
+
+
+case "$1" in
+
+start)
+start_daemon
+;;
+stop)
+stop_daemon
+;;
+restart | reload | force-reload )
+stop_daemon
+start_daemon
+;;
+esac
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 12/13] Document a versioning strategy and use it

[Jason Gunthorpe]

This is particularly important for the three shared libraries, and
we haven't been doing it right historically, perhaps due to libtool
braindamage.

The names of the shlibs are updated to:
  libibcm 1.0.11
  libibumad 3.1.11
  libibverbs 1.3.11
  librdmacm 1.1.11

The SONAME remains the same.

The overall package release is set to 11 due to libibumad having got up
to a .10 release.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 CMakeLists.txt                |   5 +-
 Documentation/versioning.md   | 108 ++++++++++++++++++++++++++++++++++++++++++
 libibcm/src/CMakeLists.txt    |   4 +-
 libibcm/src/libibcm.map       |   1 +
 libibumad/src/CMakeLists.txt  |   4 +-
 libibumad/src/libibumad.map   |   1 +
 libibverbs/src/CMakeLists.txt |   4 +-
 libibverbs/src/libibverbs.map |   1 +
 librdmacm/src/CMakeLists.txt  |   4 +-
 librdmacm/src/librdmacm.map   |   1 +
 10 files changed, 127 insertions(+), 6 deletions(-)
 create mode 100644 Documentation/versioning.md

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 805b25bb8669..8d58db0b9c07 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -32,8 +32,9 @@ else()
 endif()
 
 set(PACKAGE_NAME "RDMA")
-# FIXME versioning strategy?
-set(PACKAGE_VERSION "1")
+
+# See Documentation/versioning.md
+set(PACKAGE_VERSION "11")
 
 #-------------------------
 # Basic standard paths
diff --git a/Documentation/versioning.md b/Documentation/versioning.md
new file mode 100644
index 000000000000..0cc542aa6b79
--- /dev/null
+++ b/Documentation/versioning.md
@@ -0,0 +1,108 @@
+# Overall Package Version
+
+This version number is set in the top level CMakeLists.txt:
+
+```sh
+set(PACKAGE_VERSION "11")
+````
+
+For upstream releases this is a single integer showing the release
+ordering. We do not attempt to encode any 'ABI' information in this version.
+
+Branched stabled releases can append an additional counter eg `11.2`.
+
+Unofficial releases should include a distributor tag, eg '11.vendor2'.
+
+When the PACKAGE_VERSION is changed, the packaging files should be updated:
+
+```diff
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 389feee1e0f9..63854fe8f07f 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -26,7 +26,7 @@ project(RDMA C)
+ set(PACKAGE_NAME "RDMA")
+ 
+ # See Documentation/versioning.md
+-set(PACKAGE_VERSION "11")
++set(PACKAGE_VERSION "12")
+ 
+ #-------------------------
+ # Basic standard paths
+```
+
+# Shared Library Versions
+
+The shared libraries use the typical semantic versioning scheme, eg
+*libibumad* has a version like `3.1.11`.
+
+The version number is broken up into three fields:
+- '3' is called the SONAME and is embedded into the ELF:
+   ```sh
+   $ readelf -ds build/lib/libibumad.so.3.1.11
+    0x000000000000000e (SONAME)             Library soname: [libibumad.so.3]
+   ```
+
+   We do not expect this value to ever change for our libraries. It indicates
+   the overall ABI, changing it means the library will not dynamically to old
+   programs link anymore.
+
+- '1' is called the ABI level and is used within the ELF as the last component
+   symbol version tag.  This version must be changed every time a new symbol
+   is introduced. It allows the user to see what version of the ABI the
+   library provides.
+
+- '11' is the overall release number and is copied from `PACKAGE_VERSION` This
+  version increases with every package release, even if the library code did
+  not change. It allows the user to see what upstream source was used to build
+  the library.
+
+This version is encoded into the filename `build/lib/libibumad.so.3.1.11` and
+a symlink from `libibumad.so.3` to `build/lib/libibumad.so.3.1.11` is created.
+
+## Shared Library Symbol Versions
+
+Symbol versions are a linker technique that lets the library author provide
+two symbols with different ABIs that have the same API name. The linker
+differentiates the two cases internally. This allows the library author to
+change the ABI that the API uses. This project typically does not make use of
+this feature.
+
+As a secondary feature, the symbol version is also used by package managers
+like RPM to manage the ABI level. To make this work properly the ABI level
+must be correctly encoded into the symbol version.
+
+## Adding a new symbol
+
+First, increase the ABI level of the library. It is safe to re-use the ABI
+level for multiple new functions within a single release, but once a release
+is tagged the ABI level becomes *immutable*. The maintainer can provide
+guidence on what ABI level to use for each series.
+
+```diff
+ rdma_library(ibumad libibumad.map
+   # See Documentation/versioning.md
+-  3 3.1.${PACKAGE_VERSION}
++  3 3.2.${PACKAGE_VERSION}
+```
+
+Next, add your new symbol to the symbol version file:
+
+```diff
++ IBUMAD_3.2 {
++ 	global:
++ 		umad_new_symbol;
++ } IBUMAD_1.0;
+```
+
+NOTE: Once a release is made the stanzas in the map file are *immutable* and
+cannot be changed. Do not add your new symbol to old stanzas.
+
+The new symbol should appear in the ELF:
+
+```sh
+$ readelf -s build/lib/libibumad.so.3.1.11
+ 35: 00000000000031e0   450 FUNC    GLOBAL DEFAULT   12 umad_new_symbol@@IBUMAD_3.2
+```
+
+Finally update the `debian/libibumad3.symbols` file.
diff --git a/libibcm/src/CMakeLists.txt b/libibcm/src/CMakeLists.txt
index 2479886c6238..66b3362ec7a0 100644
--- a/libibcm/src/CMakeLists.txt
+++ b/libibcm/src/CMakeLists.txt
@@ -3,7 +3,9 @@ publish_headers(infiniband
   ../include/infiniband/cm_abi.h
   )
 
-rdma_library(ibcm libibcm.map 1 1.0.0
+rdma_library(ibcm libibcm.map
+  # See Documentation/versioning.md
+  1 1.0.${PACKAGE_VERSION}
   cm.c
   )
 target_link_libraries(ibcm LINK_PUBLIC ibverbs)
diff --git a/libibcm/src/libibcm.map b/libibcm/src/libibcm.map
index 3d83e481a6bb..c94e420a6cc8 100644
--- a/libibcm/src/libibcm.map
+++ b/libibcm/src/libibcm.map
@@ -1,3 +1,4 @@
+/* Do not change this file without reading Documentation/versioning.md */
 IBCM_1.0 {
 	global:
 		ib_cm_open_device;
diff --git a/libibumad/src/CMakeLists.txt b/libibumad/src/CMakeLists.txt
index b7b5d03bc53e..fbd15893d973 100644
--- a/libibumad/src/CMakeLists.txt
+++ b/libibumad/src/CMakeLists.txt
@@ -7,7 +7,9 @@ publish_headers(infiniband
   ../include/infiniband/umad_types.h
   )
 
-rdma_library(ibumad libibumad.map 3 3.1.0
+rdma_library(ibumad libibumad.map
+  # See Documentation/versioning.md
+  3 3.1.${PACKAGE_VERSION}
   sysfs.c
   umad.c
   umad_str.c
diff --git a/libibumad/src/libibumad.map b/libibumad/src/libibumad.map
index e42dc799d69e..8bf474e26ae2 100644
--- a/libibumad/src/libibumad.map
+++ b/libibumad/src/libibumad.map
@@ -1,3 +1,4 @@
+/* Do not change this file without reading Documentation/versioning.md */
 IBUMAD_1.0 {
 	global:
 		umad_init;
diff --git a/libibverbs/src/CMakeLists.txt b/libibverbs/src/CMakeLists.txt
index e1a54345c537..6989f7730c5e 100644
--- a/libibverbs/src/CMakeLists.txt
+++ b/libibverbs/src/CMakeLists.txt
@@ -17,7 +17,9 @@ else()
   set(NEIGH "")
 endif()
 
-rdma_library(ibverbs libibverbs.map 1 1.0.0
+rdma_library(ibverbs libibverbs.map
+  # See Documentation/versioning.md
+  1 1.3.${PACKAGE_VERSION}
   cmd.c
   compat-1_0.c
   device.c
diff --git a/libibverbs/src/libibverbs.map b/libibverbs/src/libibverbs.map
index 46744e551068..33cd6b63917b 100644
--- a/libibverbs/src/libibverbs.map
+++ b/libibverbs/src/libibverbs.map
@@ -1,3 +1,4 @@
+/* Do not change this file without reading Documentation/versioning.md */
 IBVERBS_1.0 {
 	global:
 		ibv_get_device_list;
diff --git a/librdmacm/src/CMakeLists.txt b/librdmacm/src/CMakeLists.txt
index 5324f625616d..0a60786c5f7a 100644
--- a/librdmacm/src/CMakeLists.txt
+++ b/librdmacm/src/CMakeLists.txt
@@ -8,7 +8,9 @@ publish_headers(infiniband
   ../include/infiniband/ib.h
   )
 
-rdma_library(rdmacm librdmacm.map 1 1.0.0
+rdma_library(rdmacm librdmacm.map
+  # See Documentation/versioning.md
+  1 1.1.${PACKAGE_VERSION}
   acm.c
   addrinfo.c
   cma.c
diff --git a/librdmacm/src/librdmacm.map b/librdmacm/src/librdmacm.map
index ffbd199d3119..65c049211c65 100644
--- a/librdmacm/src/librdmacm.map
+++ b/librdmacm/src/librdmacm.map
@@ -1,3 +1,4 @@
+/* Do not change this file without reading Documentation/versioning.md */
 RDMACM_1.0 {
 	global:
 		rdma_create_event_channel;
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 13/13] Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

[Jason Gunthorpe]

Turns out this is not a mailing list.

Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
---
 MAINTAINERS | 1 -
 1 file changed, 1 deletion(-)

diff --git a/MAINTAINERS b/MAINTAINERS
index b39529feec9b..7a31376056e0 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -69,7 +69,6 @@ F:	ibacm/*
 IPATH/QIB USERSPACE PROVIDER (for ib_qib.ko)
 M:	Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
 S:	Supported
-L:	ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org
 F:	libipathverbs/
 
 IWARP PORT MAPPER DAEMON (for iwarp kernel providers)
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH 06/13] Consolidate definitions for paths into config.h

[Hefty, Sean]

> config.h is the only place we pass through cmake substitution,
> so it is the only place that can define the various filesystem
> paths.
> 
> This patch handles the C code portions that use paths.
> 
> Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>

Acked-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH 05/13] Have cmake run man pages through text substitution

[Hefty, Sean]

> rsocket.7 had an errant text substitution that never worked, it is
> a good idea to have the man pages use the correct paths, so let
> us have cmake run them through.
> 
> Any man page ending in '.in' will be substituted automatically.
> 
> Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>

Acked-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

FWIW, I'm in favor of replacing the man pages with markdown and then generating the man pages from that.  I volunteer someone else to do that work.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 05/13] Have cmake run man pages through text substitution

[Jason Gunthorpe]

On Fri, Sep 23, 2016 at 09:23:42PM +0000, Hefty, Sean wrote:
> > rsocket.7 had an errant text substitution that never worked, it is
> > a good idea to have the man pages use the correct paths, so let
> > us have cmake run them through.
> > 
> > Any man page ending in '.in' will be substituted automatically.
> > 
> > Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> 
> Acked-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> 
> FWIW, I'm in favor of replacing the man pages with markdown and then
> generating the man pages from that.  I volunteer someone else to do
> that work.

I concur, and someone else should do the work ..

We could begin by requiring all new man pages to be in markdown
format. Do you do this in libfabric? What conversion tool do you like?

I've used rst for this in the past with reasonable results..

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH 05/13] Have cmake run man pages through text substitution

[Hefty, Sean]

> We could begin by requiring all new man pages to be in markdown
> format. Do you do this in libfabric? What conversion tool do you like?

We do this in libfabric.  I believe we use Jekyll.  Since someone else set that up, I was happy and didn't really care which conversion tool was used.  :)
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH 05/13] Have cmake run man pages through text substitution

[Hefty, Sean]

> We do this in libfabric.  I believe we use Jekyll.  Since someone else
> set that up, I was happy and didn't really care which conversion tool
> was used.  :)

Or maybe Jekyll does the html, and pandoc does the nroff...
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 13/13] Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

[Dalessandro, Dennis]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="utf-8", Size: 961 bytes --]

On Fri, 2016-09-23 at 13:17 -0600, Jason Gunthorpe wrote:
> Turns out this is not a mailing list.
> 
> Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
> ---
>  MAINTAINERS | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index b39529feec9b..7a31376056e0 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -69,7 +69,6 @@ F:	ibacm/*
>  IPATH/QIB USERSPACE PROVIDER (for ib_qib.ko)
>  M:	Mike Marciniszyn <mike.marciniszyn@intel.com>
>  S:	Supported
> -L:	ibsupport@intel.com
>  F:	libipathverbs/
>  
>  IWARP PORT MAPPER DAEMON (for iwarp kernel providers)

This is the right one to use, while you are at it you can tack my name
onto here as well.

https://github.com/01org/opa-libhfi1verbs/commit/377b68888a0b885fc2a44d
df7a2ec33f2fcf217b

N‹§²æìr¸›yúèšØb²X¬¶Ç§vØ^–)Þº{.nÇ+‰·¥Š{±­ÙšŠ{ayº\x1dʇڙë,j\a­¢f£¢·hš‹»öì\x17/oSc¾™Ú³9˜uÀ¦æå‰È&jw¨®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿïêäz¹Þ–Šàþf£¢·hšˆ§~ˆmš

Re: [PATCH 13/13] Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

[Jason Gunthorpe]

On Sat, Sep 24, 2016 at 01:43:25PM +0000, Dalessandro, Dennis wrote:
> On Fri, 2016-09-23 at 13:17 -0600, Jason Gunthorpe wrote:
> > Turns out this is not a mailing list.
> > 
> > Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> > ??MAINTAINERS | 1 -
> > ??1 file changed, 1 deletion(-)
> > 
> > diff --git a/MAINTAINERS b/MAINTAINERS
> > index b39529feec9b..7a31376056e0 100644
> > +++ b/MAINTAINERS
> > @@ -69,7 +69,6 @@ F:	ibacm/*
> > ??IPATH/QIB USERSPACE PROVIDER (for ib_qib.ko)
> > ??M:	Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> > ??S:	Supported
> > -L:	ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org
> > ??F:	libipathverbs/
> > ??
> > ??IWARP PORT MAPPER DAEMON (for iwarp kernel providers)
> 
> This is the right one to use, while you are at it you can tack my name
> onto here as well.
> 
> https://github.com/01org/opa-libhfi1verbs/commit/377b68888a0b885fc2a44ddf7a2ec33f2fcf217b

intel-opa is the address you want for for QIB too? It is already in the
HFI stanza

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 13/13] Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

[Dalessandro, Dennis]

On Sat, 2016-09-24 at 10:56 -0600, Jason Gunthorpe wrote:
> On Sat, Sep 24, 2016 at 01:43:25PM +0000, Dalessandro, Dennis wrote:
> > On Fri, 2016-09-23 at 13:17 -0600, Jason Gunthorpe wrote:
> > > Turns out this is not a mailing list.
> > > 
> > > Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
> > > ??MAINTAINERS | 1 -
> > > ??1 file changed, 1 deletion(-)
> > > 
> > > diff --git a/MAINTAINERS b/MAINTAINERS
> > > index b39529feec9b..7a31376056e0 100644
> > > +++ b/MAINTAINERS
> > > @@ -69,7 +69,6 @@ F:	ibacm/*
> > > ??IPATH/QIB USERSPACE PROVIDER (for ib_qib.ko)
> > > ??M:	Mike Marciniszyn <mike.marciniszyn@intel.com>
> > > ??S:	Supported
> > > -L:	ibsupport@intel.com
> > > ??F:	libipathverbs/
> > > ??
> > > ??IWARP PORT MAPPER DAEMON (for iwarp kernel providers)
> > 
> > This is the right one to use, while you are at it you can tack my
> > name
> > onto here as well.
> > 
> > https://github.com/01org/opa-libhfi1verbs/commit/377b68888a0b885fc2
> > a44ddf7a2ec33f2fcf217b
> 
> intel-opa is the address you want for for QIB too? It is already in
> the
> HFI stanza

Actually on second thought, maybe better to use:

infinipath <infinipath@intel.com>

-Denny


> Jason
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" 
> in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH 05/13] Have cmake run man pages through text substitution

[Weiny, Ira]

> 
> On Fri, Sep 23, 2016 at 09:23:42PM +0000, Hefty, Sean wrote:
> > > rsocket.7 had an errant text substitution that never worked, it is a
> > > good idea to have the man pages use the correct paths, so let us
> > > have cmake run them through.
> > >
> > > Any man page ending in '.in' will be substituted automatically.
> > >
> > > Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> >
> > Acked-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> >
> > FWIW, I'm in favor of replacing the man pages with markdown and then
> > generating the man pages from that.  I volunteer someone else to do
> > that work.
> 
> I concur, and someone else should do the work ..
> 
> We could begin by requiring all new man pages to be in markdown format.
> Do you do this in libfabric? What conversion tool do you like?
> 
> I've used rst for this in the past with reasonable results..

Infiniband-diags uses rst through rst2man.  But if I had to do it over I would probably do it different.  rst2man is not always installed and has caused build issues for some people.  Whatever tool we use I would recommend it be something common.  I'm not too familiar with markdown.  Is this really common now?

Ira

> 
> Jason
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the
> body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at
> http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 05/13] Have cmake run man pages through text substitution

[Leon Romanovsky]

[-- Attachment #1: Type: text/plain, Size: 1862 bytes --]

On Mon, Sep 26, 2016 at 06:30:21AM +0000, Weiny, Ira wrote:
> >
> > On Fri, Sep 23, 2016 at 09:23:42PM +0000, Hefty, Sean wrote:
> > > > rsocket.7 had an errant text substitution that never worked, it is a
> > > > good idea to have the man pages use the correct paths, so let us
> > > > have cmake run them through.
> > > >
> > > > Any man page ending in '.in' will be substituted automatically.
> > > >
> > > > Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> > >
> > > Acked-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> > >
> > > FWIW, I'm in favor of replacing the man pages with markdown and then
> > > generating the man pages from that.  I volunteer someone else to do
> > > that work.
> >
> > I concur, and someone else should do the work ..
> >
> > We could begin by requiring all new man pages to be in markdown format.
> > Do you do this in libfabric? What conversion tool do you like?
> >
> > I've used rst for this in the past with reasonable results..
>
> Infiniband-diags uses rst through rst2man.  But if I had to do it over I would probably do it different.  rst2man is not always installed and has caused build issues for some people.  Whatever tool we use I would recommend it be something common.  I'm not too familiar with markdown.  Is this really common now?

Among hipsters - yes :)

>
> Ira
>
> >
> > Jason
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the
> > body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at
> > http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

RE: [PATCH 05/13] Have cmake run man pages through text substitution

[Hefty, Sean]

> Infiniband-diags uses rst through rst2man.  But if I had to do it over
> I would probably do it different.  rst2man is not always installed and
> has caused build issues for some people.  Whatever tool we use I would
> recommend it be something common.  I'm not too familiar with markdown.
> Is this really common now?

Only by those who want to provide actual documentation.
 
Libfabric uses a daemon that checks for updates to the markdown pages.  If it finds any, it generates the man page and commits those changes to the source tree.  So the tar file always has the actual man page.  Our tool selection was based on what worked well with github and web page generation.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 05/13] Have cmake run man pages through text substitution

[Jason Gunthorpe]

On Mon, Sep 26, 2016 at 03:19:53PM +0000, Hefty, Sean wrote:
> > Infiniband-diags uses rst through rst2man.  But if I had to do it over
> > I would probably do it different.  rst2man is not always installed and
> > has caused build issues for some people.  Whatever tool we use I would
> > recommend it be something common.  I'm not too familiar with markdown.
> > Is this really common now?
> 
> Only by those who want to provide actual documentation.

Markdown is very popular on github and the integrated processing of
the web UI is why I picked it for the docs I started. But rst started
the trend of structued ascii..

IMHO, rst is more widely available in distros today than pandoc.

rst2man is incredibly widely distributed: it comes with
python-docutils which is widely included in every distribution. I have
no idea why people would complain to you on that point. If something
comes in-box in all the distros then it is fair game to depend upon,
IMHO.

> Libfabric uses a daemon that checks for updates to the markdown
> pages.  If it finds any, it generates the man page and commits those
> changes to the source tree.  So the tar file always has the actual
> man page.  Our tool selection was based on what worked well with
> github and web page generation.

I generally frown on checking in built files, but yeah, that is an
option.

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 13/13] Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

[Jason Gunthorpe]

On Sat, Sep 24, 2016 at 07:18:42PM +0000, Dalessandro, Dennis wrote:
> Actually on second thought, maybe better to use:
> 
> infinipath <infinipath-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

This Ok?

https://github.com/jgunthorpe/rdma-plumbing/commit/b8436b9b21bc9f8432bba6b49573ea31db945101

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH 05/13] Have cmake run man pages through text substitution

[Weiny, Ira]

> On Mon, Sep 26, 2016 at 03:19:53PM +0000, Hefty, Sean wrote:
> > > Infiniband-diags uses rst through rst2man.  But if I had to do it
> > > over I would probably do it different.  rst2man is not always
> > > installed and has caused build issues for some people.  Whatever
> > > tool we use I would recommend it be something common.  I'm not too
> familiar with markdown.
> > > Is this really common now?
> >
> > Only by those who want to provide actual documentation.
> 
> Markdown is very popular on github and the integrated processing of the
> web UI is why I picked it for the docs I started. But rst started the trend of
> structued ascii..
> 
> IMHO, rst is more widely available in distros today than pandoc.
> 
> rst2man is incredibly widely distributed: it comes with python-docutils which
> is widely included in every distribution. I have no idea why people would
> complain to you on that point. If something comes in-box in all the distros
> then it is fair game to depend upon, IMHO.

I thought so to.  But python-docutils is in the "optional" section of RH.

https://ask.openstack.org/en/question/67176/python-docutils-openstack-installation/

> 
> > Libfabric uses a daemon that checks for updates to the markdown pages.
> > If it finds any, it generates the man page and commits those changes
> > to the source tree.  So the tar file always has the actual man page.
> > Our tool selection was based on what worked well with github and web
> > page generation.
> 
> I generally frown on checking in built files, but yeah, that is an option.

Yep I agree.  But I resorted to this as well.

:-(

Ira

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 05/13] Have cmake run man pages through text substitution

[Jason Gunthorpe]

On Mon, Sep 26, 2016 at 06:17:03PM +0000, Weiny, Ira wrote:

> > rst2man is incredibly widely distributed: it comes with python-docutils which
> > is widely included in every distribution. I have no idea why people would
> > complain to you on that point. If something comes in-box in all the distros
> > then it is fair game to depend upon, IMHO.
> 
> I thought so to.  But python-docutils is in the "optional" section of RH.

I mean it is available via 'yum' with no special repositories. I do
not expect software to build out of the box on any arbitary distro
install. I expect builders to install the required set of
packages. This is why the exect package set is documented in the
README.md

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 13/13] Remove ibsupport-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org from MAINTAINERS

[Dalessandro, Dennis]

On Mon, 2016-09-26 at 12:12 -0600, Jason Gunthorpe wrote:
> On Sat, Sep 24, 2016 at 07:18:42PM +0000, Dalessandro, Dennis wrote:
> > Actually on second thought, maybe better to use:
> > 
> > infinipath <infinipath@intel.com>
> 
> This Ok?
> 
> https://github.com/jgunthorpe/rdma-plumbing/commit/b8436b9b21bc9f8432
> bba6b49573ea31db945101

Works for me. Thanks.

-Denny

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Doug Ledford]

[-- Attachment #1.1: Type: text/plain, Size: 954 bytes --]

On 9/23/16 3:17 PM, Jason Gunthorpe wrote:
> Necessary to reproduce the Debian packaging.
> 
> Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> ---
>  CMakeLists.txt                       |  3 ++
>  srp_daemon/srp_daemon/CMakeLists.txt | 29 ++++++++----
>  srp_daemon/srptools.default          | 14 ++++++
>  srp_daemon/srptools.init             | 89 ++++++++++++++++++++++++++++++++++++

Would it be best to have directories for related install files for a
specific OS?  For instance, srp_daemon/debian/ and srp_daemon/redhat/?
It's not that I expect to put a redhat init script for srpd in place,
but the script for debian is likely to not work properly on redhat, so
having it in an OS directory would at least make that clear.


-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>    GPG Key ID: 0E572FDD
  Red Hat, Inc.
  100 E. Davie St
  Raleigh, NC 27601 USA


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 907 bytes --]

Re: [PATCH 00/13] Policy changes and packaging support

[Doug Ledford]

[-- Attachment #1.1: Type: text/plain, Size: 1169 bytes --]

On 9/23/16 3:16 PM, Jason Gunthorpe wrote:
> This is the next batch of rdma-core updates, these 13 patches deal primarily
> with how the package is built and the install paths afterwards. Most likely
> these changes are mainly of interest to packagers.
> 
> The discussion here is informative for some of the patches:
>  http://www.spinics.net/lists/linux-rdma/msg39578.html
> 
> View the commits on github:
> 
>  https://github.com/linux-rdma/rdma-core/pull/3
> 
> Please forward Acks and I will update the commits.
> 
> One important thing to note is that this series pretty much eliminates all
> hard coded paths from the tree. While this is a good thing, it means the
> default behavior is to set SYSCONFDIR to /usr/local/etc/, which is probably
> not what people testing the daemons really want. Compile with:
> 
>  cmake -GNinja .. -DCMAKE_INSTALL_SYSCONFDIR:PATH=/etc
> 
> Which is similar to the configure --sysconfdir option

Series looks good modulo the few feedback items.


-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>    GPG Key ID: 0E572FDD
  Red Hat, Inc.
  100 E. Davie St
  Raleigh, NC 27601 USA


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 907 bytes --]

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Jason Gunthorpe]

On Wed, Sep 28, 2016 at 02:27:01PM -0400, Doug Ledford wrote:
> On 9/23/16 3:17 PM, Jason Gunthorpe wrote:
> > Necessary to reproduce the Debian packaging.
> > 
> > Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> >  CMakeLists.txt                       |  3 ++
> >  srp_daemon/srp_daemon/CMakeLists.txt | 29 ++++++++----
> >  srp_daemon/srptools.default          | 14 ++++++
> >  srp_daemon/srptools.init             | 89 ++++++++++++++++++++++++++++++++++++
> 
> Would it be best to have directories for related install files for a
> specific OS?  For instance, srp_daemon/debian/ and srp_daemon/redhat/?
> It's not that I expect to put a redhat init script for srpd in place,
> but the script for debian is likely to not work properly on redhat, so
> having it in an OS directory would at least make that clear.

To be clear, there is already an initscript that (perhaps?) is for
RedHat - but it isn't even close to the Debian version. So this patch
introduces two scripts for srp_dameon, which I hated doing..

I'll drop this patch from the series and we can go ahead with the
other patches in the series.

For now I'll put the initscript in the Debian packaging patch and we
can think about what to do later. I think it can be moved to the
debian/ directory as well.

Going forward I think we need to make some decisions..

1) Do we want to do something with the initscripts so distros can use
   them? Is that even possible? I think Debian uses the bundled
   acm initscript, didn't look at suse.

   Is Debian the only major distro that still ships init scripts?
   Does FC/RH exclusively ship systemd unit files now?

   Maybe we should delete the initscripts entirely.

2) I'd like to support cross-distro systemd unit files upstream.
   Is that feasible? We are short unit files, could you
   contribute yours?

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Doug Ledford]

[-- Attachment #1.1: Type: text/plain, Size: 3331 bytes --]

On 9/28/16 2:47 PM, Jason Gunthorpe wrote:
> On Wed, Sep 28, 2016 at 02:27:01PM -0400, Doug Ledford wrote:
>> On 9/23/16 3:17 PM, Jason Gunthorpe wrote:
>>> Necessary to reproduce the Debian packaging.
>>>
>>> Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
>>>  CMakeLists.txt                       |  3 ++
>>>  srp_daemon/srp_daemon/CMakeLists.txt | 29 ++++++++----
>>>  srp_daemon/srptools.default          | 14 ++++++
>>>  srp_daemon/srptools.init             | 89 ++++++++++++++++++++++++++++++++++++
>>
>> Would it be best to have directories for related install files for a
>> specific OS?  For instance, srp_daemon/debian/ and srp_daemon/redhat/?
>> It's not that I expect to put a redhat init script for srpd in place,
>> but the script for debian is likely to not work properly on redhat, so
>> having it in an OS directory would at least make that clear.
> 
> To be clear, there is already an initscript that (perhaps?) is for
> RedHat - but it isn't even close to the Debian version. So this patch
> introduces two scripts for srp_dameon, which I hated doing..
> 
> I'll drop this patch from the series and we can go ahead with the
> other patches in the series.
> 
> For now I'll put the initscript in the Debian packaging patch and we
> can think about what to do later. I think it can be moved to the
> debian/ directory as well.
> 
> Going forward I think we need to make some decisions..
> 
> 1) Do we want to do something with the initscripts so distros can use
>    them? Is that even possible? I think Debian uses the bundled
>    acm initscript, didn't look at suse.

I'm in favor of providing a good, reliable, correct set of startup files
for each of the major distro flavors.  One of my main reasons for that
is it makes it possible for us to try and provide some level of startup
script parity and commonality between the distros.  And allows us to fix
bugs across all the distros as once even if we didn't necessarily hit
the bug on each distro.  Of course, the distros may ignore our scripts,
but we can try.

> 
>    Is Debian the only major distro that still ships init scripts?
>    Does FC/RH exclusively ship systemd unit files now?

We're almost entirely systemd now.  We only have EL 6 that still uses
init scripts, and I doubt we will ever put this package into EL 6.  So,
we might as well be all systemd as far as this repo is concerned.

>    Maybe we should delete the initscripts entirely.

Not if Debian still uses it.  And I'm not opposed to providing a Red Hat
init script in case someone wants to put this package on EL 6 themselves.

> 2) I'd like to support cross-distro systemd unit files upstream.
>    Is that feasible?

Maybe....I'm not entirely sure about that.

> We are short unit files, could you
>    contribute yours?

Absolutely.  I'd like to basically import the entire redhat rdma package
into this, but it will take a little sorting things out to get all of
the files and such in the right place.  And that's a precursor to our
srpd unit file as it lists a specific dependency on the rdma unit file.

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>    GPG Key ID: 0E572FDD
  Red Hat, Inc.
  100 E. Davie St
  Raleigh, NC 27601 USA


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 907 bytes --]

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Bart Van Assche]

On 09/28/2016 11:47 AM, Jason Gunthorpe wrote:
> To be clear, there is already an initscript that (perhaps?) is for
> RedHat - but it isn't even close to the Debian version. So this patch
> introduces two scripts for srp_dameon, which I hated doing..

The existing init script works reliably on multiple versions of RHEL, 
SLES, Fedora and openSuSE. It is not specific to Red Hat distro's. So 
I'm surprised that a separate initscript is needed for Debian?

Bart.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Jason Gunthorpe]

On Wed, Sep 28, 2016 at 01:18:51PM -0700, Bart Van Assche wrote:
> On 09/28/2016 11:47 AM, Jason Gunthorpe wrote:
> >To be clear, there is already an initscript that (perhaps?) is for
> >RedHat - but it isn't even close to the Debian version. So this patch
> >introduces two scripts for srp_dameon, which I hated doing..
> 
> The existing init script works reliably on multiple versions of RHEL, SLES,
> Fedora and openSuSE. It is not specific to Red Hat distro's. So I'm
> surprised that a separate initscript is needed for Debian?

I suspect not being distro specific is the entire problem, eg the
Debian initscript is using /etc/default/srptools to configure
how the daemon is launched and it uses the start-stop-daemon tool.

A good initscript should consistently follow the distro policies for
initscripts..

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Doug Ledford]

[-- Attachment #1.1: Type: text/plain, Size: 1459 bytes --]

On 9/28/16 5:01 PM, Jason Gunthorpe wrote:
> On Wed, Sep 28, 2016 at 01:18:51PM -0700, Bart Van Assche wrote:
>> On 09/28/2016 11:47 AM, Jason Gunthorpe wrote:
>>> To be clear, there is already an initscript that (perhaps?) is for
>>> RedHat - but it isn't even close to the Debian version. So this patch
>>> introduces two scripts for srp_dameon, which I hated doing..
>>
>> The existing init script works reliably on multiple versions of RHEL, SLES,
>> Fedora and openSuSE. It is not specific to Red Hat distro's. So I'm
>> surprised that a separate initscript is needed for Debian?
> 
> I suspect not being distro specific is the entire problem,

Indeed...

> eg the
> Debian initscript is using /etc/default/srptools to configure
> how the daemon is launched and it uses the start-stop-daemon tool.
> 
> A good initscript should consistently follow the distro policies for
> initscripts..

Exactly why we often don't use the ones in the upstream package.  Try to
be generically good, and you will be universally, generically bad for
any specific distro.  If you want to provide one as an example, that's
fine.  If you want to provide one that's actually used, then you need
one per distro (or a means of generating a distro specific one from a
template or some such).

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>    GPG Key ID: 0E572FDD
  Red Hat, Inc.
  100 E. Davie St
  Raleigh, NC 27601 USA


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 907 bytes --]

Re: [PATCH 11/13] srp_daemon: Add the debian initscripts as an option

[Jason Gunthorpe]

On Wed, Sep 28, 2016 at 02:55:11PM -0400, Doug Ledford wrote:

> I'm in favor of providing a good, reliable, correct set of startup files
> for each of the major distro flavors.

Okay, how about a initscripts/{debian,centos6,suse..} directory tree,
and centralize all the initscripts there? Some one who understands
this can then try and eventually make sense of the duplication?

> > We are short unit files, could you
> >    contribute yours?
> 
> Absolutely.  I'd like to basically import the entire redhat rdma package
> into this, but it will take a little sorting things out to get all of
> the files and such in the right place.  And that's a precursor to our
> srpd unit file as it lists a specific dependency on the rdma unit file.

Yeah, that rdma unit file and udev integration really needs to be
upstream and in all distros if we are going to have any hope of having
upstream unit files...

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 05/13] Have cmake run man pages through text substitution

[ira.weiny]

On Mon, Sep 26, 2016 at 12:21:54PM -0600, Jason Gunthorpe wrote:
> On Mon, Sep 26, 2016 at 06:17:03PM +0000, Weiny, Ira wrote:
> 
> > > rst2man is incredibly widely distributed: it comes with python-docutils which
> > > is widely included in every distribution. I have no idea why people would
> > > complain to you on that point. If something comes in-box in all the distros
> > > then it is fair game to depend upon, IMHO.
> > 
> > I thought so to.  But python-docutils is in the "optional" section of RH.
> 
> I mean it is available via 'yum' with no special repositories. I do
> not expect software to build out of the box on any arbitary distro
> install. I expect builders to install the required set of
> packages. This is why the exect package set is documented in the
> README.md

FWIW, I'm supportive of that.  Just saying what my experience was.

Ira

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html