[ewg] Re: [PATCH RFCv2] RDMA/CMA: Allocate PS_TCP ports from the host TCP port space.
Steve Wise wrote: From: Steve Wise <[EMAIL PROTECTED]> This is needed for iwarp providers that support native and rdma connections over the same interface. Optionally turned on by the new rdma_cm unify_tcp_port_space module option. OK, I guess this can get it, as least from my side, thanks for fixing it. --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -809,6 +815,8 @@ static void cma_release_port(struct rdma_id_private *id_priv) kfree(bind_list); } mutex_unlock(&lock); + if (id_priv->sock) + sock_release(id_priv->sock); } any specific reason to put this here and not in rdma_destroy_id? Or. ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] Re: [GIT PULL ofed-1.4] cxgb3 lro backport fixes
Steve Wise wrote: Vlad, Please pull from: git://git.openfabrics.org/~swise/ofed-1.4 stevo This commit adds backport patches to remove LRO from cxgb3 for older kernels. Thanks, Steve. Done, Regards, Vladimir ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] Re: [GIT PULL ofed-1.4] NFSRDMA Fast Memory Registration Patches
Steve Wise wrote: Hey Vlad, Please pull the NFSRDMA Fast Register Support patches from: git://git.openfabrics.org/~swise/ofed-1.4 ofed_2_6_27 There are 12 patches that have been posted upstream for review and are pulled into ofed-1.4 as files: kernel_patches/fixes/nfsrdma_0100_frmr*.patch. I'll update them as they get refined and merged upstream, but I'd like them in ofed-1.4-beta so we can get some testing done. - If anyone is interested, here are the patch postings from the linux-nfs mailing list: Bug fix needed for fastreg support (patch 1): http://marc.info/?l=linux-nfs&m=121864357315113&w=2 Server Fast Memory Registration Support (patches 2-10): http://marc.info/?l=linux-nfs&m=121864364015273&w=2 Client Fast Memory Registration Support (patches 11-12): http://marc.info/?l=linux-nfs&m=121864523218276&w=2 Steve. Done, Regards, Vladimir ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
Re: [ewg] Re: [PATCH RFCv2] RDMA/CMA: Allocate PS_TCP ports from the host TCP port space.
Or Gerlitz wrote: Steve Wise wrote: From: Steve Wise <[EMAIL PROTECTED]> This is needed for iwarp providers that support native and rdma connections over the same interface. Optionally turned on by the new rdma_cm unify_tcp_port_space module option. OK, I guess this can get it, as least from my side, thanks for fixing it. Or & Steve, Please close the final version of the patch by end of the week and Vlad will take it on Sunday Thanks Tziporet ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] Re: [PATCH RFCv2] RDMA/CMA: Allocate PS_TCP ports from the host TCP port space.
Or Gerlitz wrote: Steve Wise wrote: From: Steve Wise <[EMAIL PROTECTED]> This is needed for iwarp providers that support native and rdma connections over the same interface. Optionally turned on by the new rdma_cm unify_tcp_port_space module option. OK, I guess this can get it, as least from my side, thanks for fixing it. --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -809,6 +815,8 @@ static void cma_release_port(struct rdma_id_private *id_priv) kfree(bind_list); } mutex_unlock(&lock); +if (id_priv->sock) +sock_release(id_priv->sock); } any specific reason to put this here and not in rdma_destroy_id? Or. Seemed like cma_release_port() is where I should release the port. And for sockets, the only way to release the port is to release the socket. Steve. ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
Re: [ewg] Re: [PATCH RFCv2] RDMA/CMA: Allocate PS_TCP ports from the host TCP port space.
Tziporet Koren wrote: Or Gerlitz wrote: Steve Wise wrote: From: Steve Wise <[EMAIL PROTECTED]> This is needed for iwarp providers that support native and rdma connections over the same interface. Optionally turned on by the new rdma_cm unify_tcp_port_space module option. OK, I guess this can get it, as least from my side, thanks for fixing it. Or & Steve, Please close the final version of the patch by end of the week and Vlad will take it on Sunday Thanks Tziporet Will do. Vlad, I will post a GIT PULL email for you by Saturday. ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] IB/ehca: Support for OFED-1.4
Hi Tziporet, here is summary of our test plan for OFED-1.4: platform: - ppc64 OS: - SLES10-SP1 - SLES10-SP2 - RHEL-4.6 - RHEL-4.7 - RHEL-5.1 - RHEL-5.2 kernel: - 2.6.24 - 2.6.25 - 2.6.26 upl: - IPoIB - sdp - udapl regards Stefan ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] $69.95 Viagra (Sildenafil) 100mg x 10 pills buy now
US $ 129.95 buy now Viagra (Sildenafil) 100mg x 60 pills http://anatja.bay.livefilestore.com/y1pU7DAWoz1NDMdFBwhgo6YCpJ4Wb4iSa3wIXQZ2_A_85DlK5nWnA6auPx-sDihScuqhFRtLr-1gQhCdLLS01f7Sw/index.html ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] OFED 1.4 beta release is delayed for next week
Hi, Since we have not completed all requested changes we will delay the beta release for early next week. The only fix we waiting for is the CMA fix from Steve. BTW - I am on vacation next week, so Vlad will coordinate the release Tziporet ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] No test, No class, buy yourself Bacheelor/MasteerMBA/Doctoraate dip1omas, VALID in all countries
Bachellor, MassterMBA, and Doctoorate certification available in YOUR field of choice. Yes, you can become a Doc and get all the benefits alongside it! Our Diiploma/Certiificates are recognised worldwide. No need for exams, or testing, classrooms, tests, even interviewing. We turn no one down We guarantee confidentiality. CALL US 24 HOURS A DAY, 7 DAYS A WEEK For US: 1-781-634-7970 Outside US: +1-781-634-7970 "Just leave your NAME & PHONE NO. (with CountryCode)" in the voicemail our staff will get back to you in next few days ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] [GIT PULL ofed-1.4] RDMA/CMA: Add unified ports patch.
Vlad,
Please pull this patch from:
git://git.openfabrics.org/~swise/ofed-1.4 ofed_kernel
Thanks,
Steve.
Original Message
Subject:[PATCH] RDMA/CMA: Add unified ports patch.
Date: Thu, 14 Aug 2008 17:35:36 -0500
From: Steve Wise <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
From: Steve Wise <[EMAIL PROTECTED]>
Signed-off-by: Steve Wise <[EMAIL PROTECTED]>
---
.../fixes/cma_0100_unified_tcp_ports.patch | 96
+++
1 files changed, 96 insertions(+), 0 deletions(-)
diff --git a/kernel_patches/fixes/cma_0100_unified_tcp_ports.patch
b/kernel_patches/fixes/cma_0100_unified_tcp_ports.patch
new file mode 100644
index 000..f164a29
--- /dev/null
+++ b/kernel_patches/fixes/cma_0100_unified_tcp_ports.patch
@@ -0,0 +1,96 @@
+RDMA/CMA: Allocate PS_TCP ports from the host TCP port space.
+
+From: Steve Wise <[EMAIL PROTECTED]>
+
+This is needed for iwarp providers that support native and rdma
+connections over the same interface.
+
+Optionally turned on by the new rdma_cm unify_tcp_port_space module
+option.
+
+Signed-off-by: Steve Wise <[EMAIL PROTECTED]>
+---
+
+ drivers/infiniband/core/cma.c | 34 +-
+ 1 files changed, 33 insertions(+), 1 deletions(-)
+
+diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
+index e980ff3..bffc8df 100644
+--- a/drivers/infiniband/core/cma.c
b/drivers/infiniband/core/cma.c
+@@ -54,6 +54,11 @@ MODULE_AUTHOR("Sean Hefty");
+ MODULE_DESCRIPTION("Generic RDMA CM Agent");
+ MODULE_LICENSE("Dual BSD/GPL");
+
++int unify_tcp_port_space = 0;
++module_param(unify_tcp_port_space, int, 0644);
++MODULE_PARM_DESC(unify_tcp_port_space, "Unify the host TCP and RDMA port "
++ "space allocation (default=0)");
++
+ #define CMA_CM_RESPONSE_TIMEOUT 20
+ #define CMA_MAX_CM_RETRIES 15
+ #define CMA_CM_MRA_SETTING (IB_CM_MRA_FLAG_DELAY | 24)
+@@ -117,6 +122,7 @@ struct rdma_id_private {
+ struct rdma_cm_id id;
+
+ struct rdma_bind_list *bind_list;
++ struct socket *sock;
+ struct hlist_node node;
+ struct list_headlist; /* listen_any_list or cma_device.list */
+ struct list_headlisten_list; /* per device listens */
+@@ -809,6 +815,8 @@ static void cma_release_port(struct rdma_id_private
*id_priv)
+ kfree(bind_list);
+ }
+ mutex_unlock(&lock);
++ if (id_priv->sock)
++ sock_release(id_priv->sock);
+ }
+
+ static void cma_leave_mc_groups(struct rdma_id_private *id_priv)
+@@ -2037,6 +2045,25 @@ static int cma_use_port(struct idr *ps, struct
rdma_id_private *id_priv)
+ return 0;
+ }
+
++static int cma_get_tcp_port(struct rdma_id_private *id_priv)
++{
++ int ret;
++ struct socket *sock;
++
++ ret = sock_create_kern(AF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
++ if (ret)
++ return ret;
++ ret = sock->ops->bind(sock,
++ (struct sockaddr *)&id_priv->id.route.addr.src_addr,
++ ip_addr_size(&id_priv->id.route.addr.src_addr));
++ if (ret) {
++ sock_release(sock);
++ return ret;
++ }
++ id_priv->sock = sock;
++ return 0;
++}
++
+ static int cma_get_port(struct rdma_id_private *id_priv)
+ {
+ struct idr *ps;
+@@ -2048,6 +2075,11 @@ static int cma_get_port(struct rdma_id_private *id_priv)
+ break;
+ case RDMA_PS_TCP:
+ ps = &tcp_ps;
++ if (unify_tcp_port_space) {
++ ret = cma_get_tcp_port(id_priv);
++ if (ret)
++ goto out;
++ }
+ break;
+ case RDMA_PS_UDP:
+ ps = &udp_ps;
+@@ -2065,7 +2097,7 @@ static int cma_get_port(struct rdma_id_private *id_priv)
+ else
+ ret = cma_use_port(ps, id_priv);
+ mutex_unlock(&lock);
+-
++out:
+ return ret;
+ }
+
___
ewg mailing list
ewg@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
[ewg] issues with OFED-1.4-20080814-0600
Hi all, Please retain me on replies, I'm not on the mailing list. I bumped into two issues while trying to use the scripts provided with OFED-1.4-20080814-0600. 1) uninstall.sh does not succeed due to unmentioned -devel packages. I worked around this by specifying --nodeps (on a system where I was not afraid to horribly break things), but I think the following packages/issues either need to be listed or accounted for: a) Running rpm -e --allmatches ib-bonding libibverbs libibverbs-devel libmthca libmlx4 libmlx4-devel libehca libcxgb3 libcxgb3-devel libipathverbs libipathverbs-devel libibcm libibcm-devel libibcommon libibcommon-devel libibumad libibumad-devel libibmad libibmad-devel librdmacm librdmacm-devel libsdp libsdp-devel opensm opensm-devel dapl dapl-devel perftest mstflint qlvnictools sdpnetstat srptools rds-tools ibutils infiniband-diags ibvexdmtools opensm opensm-devel libipathverbs libipathverbs-devel libehca libehca-devel dapl dapl-devel libibcm libibcm-devel libibcommon libibcommon-devel libibmad libibmad-devel libibumad libibumad-devel libibverbs libibverbs-devel libipathverbs libipathverbs-devel libmthca libmthca-devel libmlx4 libmlx4-devel libsdp librdmacm librdmacm-devel openmpi openmpi-devel ibutils ibutils-devel error: Failed dependencies: libibverbs.so.1 is needed by (installed) libamso-1.0-4.4.ppc libibverbs.so.1 is needed by (installed) dapl2-2.0.7-0.4.ppc libibverbs.so.1 is needed by (installed) mvapich2-1.0.2-0.4.ppc ... libibverbs-devel is needed by (installed) libamso-devel-1.0-4.4.ppc libibverbs-devel = 1.1.1 is needed by (installed) libibverbs-devel-64bit-1.1.1-9.6.ppc libibverbs-devel is needed by (installed) mvapich2-devel-1.0.2-0.4.ppc libibverbs-devel is needed by (installed) dapl2-devel-2.0.7-0.4.ppc libibverbs-devel is needed by (installed) mvapich-devel-1.0.0_2106-0.3.ppc64 libmlx4-devel = 1.0 is needed by (installed) libmlx4-devel-64bit-1.0-0.4.ppc libcxgb3-devel = 1.1.4 is needed by (installed) libcxgb3-devel-64bit-1.1.4-0.3.ppc libipathverbs-devel = 1.1 is needed by (installed) libipathverbs-devel-64bit-1.1-3.10.ppc libibcm-devel = 1.0.2 is needed by (installed) libibcm-devel-64bit-1.0.2-0.3.ppc libibcommon-devel = 1.0.8 is needed by (installed) libibcommon-devel-64bit-1.0.8-0.3.ppc ... libibumad-devel = 1.1.7 is needed by (installed) libibumad-devel-64bit-1.1.7-0.3.ppc libibmad-devel = 1.1.6 is needed by (installed) libibmad-devel-64bit-1.1.6-0.3.ppc ... librdmacm-devel = 1.0.6 is needed by (installed) librdmacm-devel-64bit-1.0.6-0.3.ppc b) Running rpm -e --allmatches libehca-devel error: Failed dependencies: libehca-devel = 1.2 is needed by (installed) libehca-devel-64bit-1.2-0.3.ppc c) I think there are other examples like b) where the 64bit package also needs to be removed, and will be because of the result of rpm -qa, but the 64bit package depends on the 32-bit one (which honestly seems broken to me, for libraries). Simply adding the packages in for the -devel cases is insufficient, because the specified order matters here (as far as I can tell). 2) Build failure with newer kernels for ofa_kernel Mainline recently moved to allowing architectures to place architecture-specific header files in arch/$SUBARCH/include/asm rather than include/asm (which is a symlink to include/asm-$SUBARCH). But line 150 of the makefile provided with ofa_kernel specifies: $$(if $$(KBUILD_SRC),-Iinclude2 -I$$(srctree)/include) \ I think a " -I$$(srctree)/arch/$$(SUBARCH)" is also needed. I'm not sure if $(SUBARCH) is the right variable to use but it seemed to work in my quick testing (manually). Thanks, Nish -- Nishanth Aravamudan <[EMAIL PROTECTED]> IBM Linux Technology Center ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
