[ewg] Re: [PATCH] NFS-RDMA backport for RHEL 5.2
Yes. Thanks from me as well. Although I see much overlap between your RHEL5.2 patch and my SLES10SP2 patches, I will start with your work in order to unify the two backports as much as possible. Your backport also answers a few questions I've been having when reviewing my backport. Hopefully the SLES10SP2 backport is not too far off. Thanks again. Yes, good idea. You should be able to start with the RH5.2 (2.6.18) and add to it to get to SLES10SP2 (2.6.16). Hollar at Jon if you run into any issues. Steve. ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] kernel-ib / sles 10sp2 / zypper install issue
Hi, I ran into the following issue on SLES 10 SP2: Recompiling OFED 1.4 went just fine I am able to install kernel-ib using rpm When I set up a repo in order to use zipper (native SLES tool) to perform the install, I get the following: testsles:~ # zypper in kernel-ib Restoring system sources... Parsing metadata for SUSE Linux Enterprise Server 10 SP2... Parsing RPM database... 1 Problems: Problem: kernel-ib-1.4-2.6.16.60_0.21_smp.x86_64[SUSE Linux Enterprise Server 10 SP2] cannot be installed due to missing dependencies Problem: kernel-ib-1.4-2.6.16.60_0.21_smp.x86_64[SUSE Linux Enterprise Server 10 SP2] cannot be installed due to missing dependencies There are no installable providers of kernel(net_netlink) == ed7265d4fe405c95 for kernel-ib-1.4-2.6.16.60_0.21_smp.x86_64[SUSE Linux Enterprise Server 10 SP2] Solution 1: do not install kernel-ib do not install kernel-ib-1.4-2.6.16.60_0.21_smp.x86_64[SUSE Linux Enterprise Server 10 SP2] Solution 2: Ignore this requirement just here number, (r)etry or (c)ancel> c But if I try to install libibverbs using zypper, everything is fine. I checked on the web, no luck. I could not find any reported bug in bugzilla so I was wondering if this kind of error was ever reported / hit by anybody. On both nodes (the one I used for compilation and the one I used for the install) I have kernel/kernel-source/kernel-syms installed and the /proc/config.gz are identical. I also tried to install it on the same node it was recompiled, same problem. The repo was created using yast. The output of uname -a is : testsles:~ # uname -a Linux testsles 2.6.16.60-0.21-smp #1 SMP Tue May 6 12:41:02 UTC 2008 x86_64 x86_64 x86_64 GNU/Linux And for /etc/SuSE-release : testsles:~ # cat /etc/SuSE-release SUSE Linux Enterprise Server 10 (x86_64) VERSION = 10 PATCHLEVEL = 2 testsles:~ # Thanks, Mehdi ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] Re: [PATCH] NFS-RDMA backport for RHEL 5.2
Steve Wise wrote: Jon Mason wrote: Hey Vlad, I wanted to get you the NFS-RDMA backport patches I have queued up prior to rc1 being built. I have this patch (RHEL5.2), as well as 2.6.22, and 2.6.25 (which I will be sending in separate e-mails). Thanks Jon and Tom for doing this work! Yes. Thanks from me as well. Although I see much overlap between your RHEL5.2 patch and my SLES10SP2 patches, I will start with your work in order to unify the two backports as much as possible. Your backport also answers a few questions I've been having when reviewing my backport. Hopefully the SLES10SP2 backport is not too far off. Thanks again. -jeff This patch contains the changes necessary to get NFS-RDMA 99% working on RHEL5.2. It 100% passes of the connectathon testcases when using RDMA. When using TCP, 1 testcase hangs (for which I've found a code hack to work around, but not a fix). The testcase is the 30MB transfer in the "special" tests, and is 100% reproducible. With the attached patch, you can now enable the nfs-rdma-mod by default and run it through your nightly build tests. Just to clarify: Enable nfs-rdma-mod on rhel5.2, 2.6.22, and 2.6.18 only. Also we'll have to address any hw platform issues that come up during your builds. There is one problem with nfs-utils, which will necessitate including it in the OFED distribution. Currently, there is a check for the kernel version you are running when running the mount.nfs command. This check is to verify if one is running a kernel greater than 2.6.22. If not, then it will not allow a mount with RDMA as an option. In earlier versions, there was an override option of "-i". This has since been removed for an unknown reason (but if you run `mount.nfs`, you can see that they forgot to remove 'i' from "usage" options). I will send them a patch to fix this in the latest version, but this will be a problem that will require it to be built locally until such a time as that is pulled in. We'll need to ship nfs-utils that works for nfsrdma on the older kernels as part of ofed-1.4.1. Stay tuned for this. Steve. ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] Re: [PATCH] NFS-RDMA backport for RHEL 5.2
Jon Mason wrote: Hey Vlad, I wanted to get you the NFS-RDMA backport patches I have queued up prior to rc1 being built. I have this patch (RHEL5.2), as well as 2.6.22, and 2.6.25 (which I will be sending in separate e-mails). Thanks Jon and Tom for doing this work! This patch contains the changes necessary to get NFS-RDMA 99% working on RHEL5.2. It 100% passes of the connectathon testcases when using RDMA. When using TCP, 1 testcase hangs (for which I've found a code hack to work around, but not a fix). The testcase is the 30MB transfer in the "special" tests, and is 100% reproducible. With the attached patch, you can now enable the nfs-rdma-mod by default and run it through your nightly build tests. Just to clarify: Enable nfs-rdma-mod on rhel5.2, 2.6.22, and 2.6.18 only. Also we'll have to address any hw platform issues that come up during your builds. There is one problem with nfs-utils, which will necessitate including it in the OFED distribution. Currently, there is a check for the kernel version you are running when running the mount.nfs command. This check is to verify if one is running a kernel greater than 2.6.22. If not, then it will not allow a mount with RDMA as an option. In earlier versions, there was an override option of "-i". This has since been removed for an unknown reason (but if you run `mount.nfs`, you can see that they forgot to remove 'i' from "usage" options). I will send them a patch to fix this in the latest version, but this will be a problem that will require it to be built locally until such a time as that is pulled in. We'll need to ship nfs-utils that works for nfsrdma on the older kernels as part of ofed-1.4.1. Stay tuned for this. Steve. ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] [ANNOUNCE] management tarballs release
Hi, There is a new release of the management (OpenSM and infiniband diagnostics) tarballs available in: http://www.openfabrics.org/downloads/management/ md5sum: 97b2609f5eaaf4320b39f44a50500b70 libibumad-1.3.1.tar.gz e60b1c787d7cd2768967ca4766238210 libibmad-1.3.1.tar.gz 8c8c153f21d9f6cee51fc3d501c54fe7 opensm-3.3.1.tar.gz 6b6c87ed01291a2a3322b0ff696c5a11 infiniband-diags-1.5.1.tar.gz All component versions are from recent master branch. Full change log is below. Sasha Arlin Davis (3): libibmad: add os dependent definitions. libibmad: remove c99 definitions within the ib_mad_f structure libibmad: minor changes to source to allow portability to WinOF. David McMillen (1): infiniband-diags/src/ibnetdiscover.c missing LID information on --ports Eli Dorfman (1): opensm/osm_inform.c report IB traps to plugin Eli Dorfman (Voltaire) (10): opensm/osm_subnet.c Fix memory leak for QOS string parameters. libibmad add PortXmitWait and CounterSelect2 to fields. opensm: Add new partition keyword for all hca, switches and routers docs update documenatation about new partition keywords infiniband-diags support PortXmitWait get and set opensm/osm_log.c save log_max_size in subnet opt in MB opensm/osm_subnet.c support subnet configuration rescan and update libibmad/src/dump.c fix dump functions for big endian machines opensm/osm_subnet.c enable log_max_size opt update opensm/osm_subnet.c fix parse functions for big endian machines Hal Rosenstock (23): opensm/libvendor/osm_vendor_sa_api.h: Fix commentary typo opensm/osm_inform.c: Eliminate compile warning opensm/osm_perfmgr_db.h: Remove unused typedef opensm/osm_perfmgr.c: In osm_perfmgr_init, eliminate memory leak on error libibmad/(mad.h fields.c): Add support for PerfMgt ClassPortInfo opensm/include/iba/ib_types.h: Add xmit_wait for PortCounters opensm/PerfMgr: Mainly cosmetic changes opensm/osm_node.h: Fix osm_node_get_num_physp description opensm/PerfMgr: Primarily fix enhanced switch port 0 perf manager operation opensm/doc/perf-manager-arch.txt: Fix some commentary typos opensm/PerfMgr: Add copyrights libibmad: lid print format changed to unsigned libibumad/umad.c: Change lid print format to unsigned infiniband-diags/perfquery: Change option name for extended counters opensm/osm_inform.c: Fix sense of zero GID compare in __match_inf_rec management/libibmad.txt: Remove madrpc_lock/unlock opensm/man/opensm.8.in: Indicate ROUTER_EXP obsoleted opensm/osm_console.c: Improve perfmgr print_counters error message infiniband-diags/smpdump.c: Fix usage examples infiniband-diags/smpdump.c: Release umad resources on exit opensm/console: Enhance perfmgr print_counters for better nodenames libibmad/fields.c: Dump LIDs as unsigned decimal infiniband-diags/saquery.c: Convert more LID prints to unsigned decimal Ira Weiny (3): opensm/opensm/osm_console.c: move reporting of plugins to "status" command. OpenSM: update osmeventplugin example for the new TRAP event. libibmad: Use enum types for function parameters Mike Heinz (1): opensm/osm_vendor_*_sa: fix incompatibility with QLogic SM Nicolas Morey Chaisemartin (4): Corrected incoherency in __osm_ftree_fabric_route_to_non_cns comments opensm/osm_ucast_ftree.c: Fixed bug on index port incrementation opensm/osm_ucast_ftree.c Fixed bad init value for down port index opensm/osm_console.c : Added dump_portguid function to console to generate a list of port guids matching one or more regexps Ralph Campbell (2): libibumad: get_ca() can call release_ca() with uninitialized data opensm: fix structure definition for trap 257-258 Robert Pearson (10): mesh analysis - skeleton mesh analysis - mesh_t data structure mesh analysis - node and link structures mesh analysis - matrix/polynomial routines mesh analysis - local geometry mesh analysis - mesh info table mesh analysis - induce global geometry mesh analysis - reorder links mesh analysis - lash preparation mesh analysis - integrate into lash core Sasha Khapyorsky (111): opensm: remove some unused variables and funcs opensm/osm_ucast_mgr: indentation fix infiniband-diags/saquery: indentation fixes infiniband-diabs/saquery: unify SA queries processors infiniband-diags/saquery: separate queries and commands infiniband-diags/saquery: PortInfoRecord query infinabd-diags: convert type uint -> unsigned int opensm: remove unused header osm_pkey_mgr.h opensm/osm_sm.c: fix MC group creation in race condition opensm/osm_sa_mcmember_record: improve __cleanup_mgrp() opensm/multicast: remove some unused parameters. opensm/osm_subnet: consolidate some duplicated code ope