pool/libfabric
SHA256
2
0
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Accepting request 558744 from home:NMoreyChaisemartin:branches:science:HPC

Browse Source 
- Update to v1.5.3
  - Core
    - Handle malloc failures
    - Ensure global lock is initialized on Windows
    - Fix spelling and formatting errors in man pages
  - PSM
    - Fix print format mismatches
    - Remove 15 second startup delay when no hardware is installed
    - Preserve FI_MR_SCALABLE mode bit for backwards compatability
  - PSM2
    - Fix print format mismatches
    - Allow all to all communication between scalable endpoints
    - Preserve FI_MR_SCALABLE mode bit for backwards compatability
    - Fix reference counting issue with opened domains
    - Fix segfault for RMA/atomic operations to local scalable endpoints
    - Fix resource counting related issues for Tx/Rx contexts
    - Allow completion suppression when fi_context is non-NULL
    - Use correct queue for triggered operations with scalable endpoints
  - Sockets
    - Fix check for invalid connection handle
    - Fix crash in fi_av_remove
  - Util
    - Fix number of bits used for connection index
  - Verbs
    - Fix incorrect CQ entry data for MSG endpoints
    - Properly check for errors from getifaddrs
    - Retry getifaddr on failure because of busy netlink sockets
    - Ack CM events on error paths
- Remove 0001-prov-psm-Eliminate-psm2-compat-library-delay-with-hf.patch
   as it was merged upstream

OBS-URL: https://build.opensuse.org/request/show/558744
OBS-URL: https://build.opensuse.org/package/show/science:HPC/libfabric?expand=0&rev=28
This commit is contained in:
Nicolas Morey-Chaisemartin 2017-12-20 09:03:14 +00:00 committed by Git OBS Bridge
parent ad87934964
commit c053ee0122
6 changed files with 40 additions and 94 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

86
0001-prov-psm-Eliminate-psm2-compat-library-delay-with-hf.patch
View File

@ -1,86 +0,0 @@
From 6c8c40ad84fa790831407e5cd25375af898d929b Mon Sep 17 00:00:00 2001
From: Jianxin Xiong <jianxin.xiong@intel.com>
Date: Mon, 9 Oct 2017 15:10:45 -0700
Subject: [PATCH] prov/psm: Eliminate psm2-compat library delay with hfi
devices missing
The PSM2 library may introduce a 15 second delay at device initialization
time if the hfi devices are missing. This has been handled in the psm2
provider by checking the existence of the device files before initializing
the device.
The psm provider didn't handle this situation because the issue doesn't
exist with native PSM library over TrueScale. However, when PSM is
supported via the psm2-compat library over PSM2, the same delay can be
observed.
Now add the same mechanism to the psm provider.
Signed-off-by: Jianxin Xiong <jianxin.xiong@intel.com>
---
prov/psm/src/psmx_init.c | 34 ++++++++++++++++++++++++++++++++++
1 file changed, 34 insertions(+)
diff --git a/prov/psm/src/psmx_init.c b/prov/psm/src/psmx_init.c
index 118ef1a81..c4b06160c 100644
--- a/prov/psm/src/psmx_init.c
+++ b/prov/psm/src/psmx_init.c
@@ -32,10 +32,12 @@
#include "psmx.h"
#include "prov.h"
+#include <glob.h>
static int psmx_init_count = 0;
static int psmx_lib_initialized = 0;
static pthread_mutex_t psmx_lib_mutex;
+static int psmx_compat_lib = 0;
struct psmx_env psmx_env = {
.name_server = 1,
@@ -103,6 +105,12 @@ static int psmx_init_lib(void)
PSM_VERNO_MAJOR, PSM_VERNO_MINOR, major, minor);
}
+ if (major > 1) {
+ psmx_compat_lib = 1;
+ FI_INFO(&psmx_prov, FI_LOG_CORE,
+ "PSM is supported via the psm2-compat library over PSM2.\n");
+ }
+
psmx_lib_initialized = 1;
out:
@@ -197,6 +205,32 @@ static int psmx_getinfo(uint32_t version, const char *node, const char *service,
if (psmx_init_lib())
return -FI_ENODATA;
+ if (psmx_compat_lib) {
+ /*
+ * native PSM running over TrueScale doesn't have the issue handled
+ * here. it's only present when PSM is supported via the psm2-compat
+ * library, where the PSM functions are just wrappers around the PSM2
+ * counterparts.
+ *
+ * psm2_ep_num_devunits() may wait for 15 seconds before return
+ * when /dev/hfi1_0 is not present. Check the existence of any hfi1
+ * device interface first to avoid this delay. Note that the devices
+ * don't necessarily appear consecutively so we need to check all
+ * possible device names before returning "no device found" error.
+ * This also means if "/dev/hfi1_0" doesn't exist but other devices
+ * exist, we are still going to see the delay; but that's a rare case.
+ */
+ glob_t glob_buf;
+
+ if ((glob("/dev/hfi1_[0-9]", 0, NULL, &glob_buf) != 0) &&
+ (glob("/dev/hfi1_[0-9][0-9]", GLOB_APPEND, NULL, &glob_buf) != 0)) {
+ FI_INFO(&psmx_prov, FI_LOG_CORE,
+ "no hfi1 device is found.\n");
+ return -FI_ENODATA;
+ }
+ globfree(&glob_buf);
+ }
+
if (psm_ep_num_devunits(&cnt) || !cnt) {
FI_INFO(&psmx_prov, FI_LOG_CORE,
"no PSM device is found.\n");

2
_service
View File

@ -8,7 +8,7 @@
<param name="versionformat">@PARENT_TAG@.@TAG_OFFSET@.%h</param>
<param name="versionrewrite-pattern">v(.*)</param>
<param name="versionrewrite-replacement">\1</param>
<param name="revision">480a6db351fbe8ee38077902c8df875e3cd13205</param>
<param name="revision">09aaaf78c239df7fb57f19de44cc96b25701bcb0</param>
</service>
<service name="recompress" mode="disabled">
<param name="file">libfabric*.tar</param>

3
libfabric-1.5.2.0.480a6db3.tar.bz2
View File

@ -1,3 +0,0 @@
version https://git-lfs.github.com/spec/v1
oid sha256:c2ebdb6376c35b4c88935648843e633bb0a2549eed96d8e610d4ce7ea39d2754
size 1068927

3
libfabric-1.5.3.0.09aaaf78c239.tar.bz2 Normal file
View File

@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:eeca796583059a9d3454648b585a98e790ea9c16f8ebe75872e824e410eb8010
size 1072867

34
libfabric.changes
View File

@ -1,3 +1,37 @@
-------------------------------------------------------------------
Wed Dec 20 08:49:03 UTC 2017 - nmoreychaisemartin@suse.com
- Update to v1.5.3
- Core
- Handle malloc failures
- Ensure global lock is initialized on Windows
- Fix spelling and formatting errors in man pages
- PSM
- Fix print format mismatches
- Remove 15 second startup delay when no hardware is installed
- Preserve FI_MR_SCALABLE mode bit for backwards compatability
- PSM2
- Fix print format mismatches
- Allow all to all communication between scalable endpoints
- Preserve FI_MR_SCALABLE mode bit for backwards compatability
- Fix reference counting issue with opened domains
- Fix segfault for RMA/atomic operations to local scalable endpoints
- Fix resource counting related issues for Tx/Rx contexts
- Allow completion suppression when fi_context is non-NULL
- Use correct queue for triggered operations with scalable endpoints
- Sockets
- Fix check for invalid connection handle
- Fix crash in fi_av_remove
- Util
- Fix number of bits used for connection index
- Verbs
- Fix incorrect CQ entry data for MSG endpoints
- Properly check for errors from getifaddrs
- Retry getifaddr on failure because of busy netlink sockets
- Ack CM events on error paths
- Remove 0001-prov-psm-Eliminate-psm2-compat-library-delay-with-hf.patch
as it was merged upstream
-------------------------------------------------------------------
Mon Nov 20 16:27:13 UTC 2017 - nmoreychaisemartin@suse.com

6
libfabric.spec
View File

@ -17,18 +17,17 @@
#
%define git_ver .0.480a6db3
%define git_ver .0.09aaaf78c239
Name: libfabric
Summary: User-space RDMA Fabric Interfaces
License: GPL-2.0 or BSD-2-Clause
Group: Development/Libraries/C and C++
Version: 1.5.2
Version: 1.5.3
Release: 0
Source: %{name}-%{version}%{git_ver}.tar.bz2
Source1: baselibs.conf
Patch0: libfabric-libtool.patch
Patch1: 0001-prov-psm-Eliminate-psm2-compat-library-delay-with-hf.patch
Url: http://www.github.com/ofiwg/libfabric
Prefix: ${_prefix}
BuildRequires: autoconf
@ -72,7 +71,6 @@ services, such as RDMA. This package contains the development files.
%prep
%setup -q -n %{name}-%{version}%{git_ver}
%patch0 -p1
%patch1 -p1
%build
rm -f config/libtool.m4