On 05/02/2015 04:54, Weiny, Ira wrote:
On Thu, Jan 29, 2015 at 09:50:38PM +0100, Yann Droneaud wrote:
Anyway, I recognize that uverb way of abusing write() syscall is
borderline (at best) regarding other Linux subsystems and Unix
paradigm in general. But it's not enough to screw it more.
While commit 7e36ef8205ff (IB/core: Temporarily disable ex_query_device uverb)
is correct as it make the extended QUERY_DEVICE uverb as integrated as
part of commit 5a77abf9a97a (IB/core: Add support for extended query device
caps)
and commit 860f10a799c8 (IB/core: Add flags for on demand paging
Hi Trond,
The following changes since commit a0a1d50cd1e80652142af5cddcde500d06c71bdd:
xprtrdma: Update the GFP flags used in xprt_rdma_allocate() (2015-01-30
12:18:48 -0500)
are available in the git repository at:
git://git.linux-nfs.org/projects/anna/nfs-rdma.git
From: SF Markus Elfring elfr...@users.sourceforge.net
Date: Wed, 04 Feb 2015 15:22:33 +0100
From: Markus Elfring elfr...@users.sourceforge.net
Date: Wed, 4 Feb 2015 15:17:00 +0100
The vunmap() function performs also input parameter validation.
Thus the test around the call is not needed.
On 02/05/15 13:26, SF Markus Elfring wrote:
From: Markus Elfring elfr...@users.sourceforge.net
Date: Thu, 5 Feb 2015 13:20:42 +0100
The srp_destroy_fr_pool() function tests whether its argument is NULL and then
returns immediately. Thus the test around the call is not needed.
This issue
Subject: [PATCH RE-RESEND V2 for-next 4/5] IB/core: Make sure that the PSN
does not overflow
Is there a particular bug which this fixes?
Ira
From: Majd Dibbiny m...@mellanox.com
The rq/sq-psn is 24 bits as defined in the IB spec, therefore ULPs and User
space applications shouldn't
The pr_* macros already supply the relevant KERN_* level; any extra
just ends up as garbage in the output.
Signed-off-by: Rasmus Villemoes li...@rasmusvillemoes.dk
---
drivers/infiniband/hw/mlx4/main.c | 3 +--
drivers/infiniband/hw/mlx5/qp.c | 2 +-
2 files changed, 2 insertions(+), 3
The macro nes_debug already ensures that __func__ (and __LINE__) is
part of the output, so there's no reason to duplicate the function
name in the format string itself.
Signed-off-by: Rasmus Villemoes li...@rasmusvillemoes.dk
---
drivers/infiniband/hw/nes/nes.c | 2 +-
The macro isert_dbg already ensures that __func__ is part of the
output, so there's no reason to duplicate the function name in the
format string itself.
Signed-off-by: Rasmus Villemoes li...@rasmusvillemoes.dk
---
drivers/infiniband/ulp/isert/ib_isert.c | 7 +++
1 file changed, 3
PDBG prepends MOD (which expands to iw_cxgb4 :) to the format
string. So the KERN_INFO just ends up in the middle where it just
garbles the output. At the same place, there's obviously no reason to
add MOD explicitly.
The two other are straight-forward fixes of KERN_ERR KERN_ERR, where
the second
On Thu, Feb 05, 2015 at 12:12:16PM +0200, Or Gerlitz wrote:
diff --git a/drivers/infiniband/hw/mlx4/mad.c
b/drivers/infiniband/hw/mlx4/mad.c
index 82a7dd8..6be0d2c 100644
+++ b/drivers/infiniband/hw/mlx4/mad.c
@@ -907,7 +907,7 @@ int mlx4_ib_mad_init(struct mlx4_ib_dev *dev)
err:
From: Majd Dibbiny m...@mellanox.com
Send with invalidate can be used to invalidate a remote memory region or memory
window. The complete details are specified in the InfiniBand Architecture
Specifications, section 9.4.1.1.
According to the specs, the following objects can be invalidated:
1.
From: Majd Dibbiny m...@mellanox.com
Memory Windows(MW) is a method to raise the remote privileges of memory range
within a Memory Region(MR).
A MW is allocated using ibv_alloc_mw, and for it to be useful it should be
bound to a MR using ibv_bind_mw.
The bind operation generates a new R_key
From: Majd Dibbiny m...@mellanox.com
Overview:
Memory Windows(MW) is a method to raise the remote privileges of memory range
within a Memory Region(MR).
A MW is allocated using ibv_alloc_mw, and for it to be useful it should be
bound to a MR.
A MW can be bound to a MR according to it's type:
1.
From: Dan Ben Yosef da...@mellanox.com
Add new flag --enable-rdma-ndd (default=yes)
Signed-off-by: Dan Ben Yosef da...@mellanox.com
---
diff --git a/Makefile.am b/Makefile.am
index 4e08c2b..63c4b48 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -15,9 +15,11 @@ sbin_PROGRAMS = src/ibaddr
From: Majd Dibbiny m...@mellanox.com
The rq/sq-psn is 24 bits as defined in the IB spec, therefore ULPs and User
space applications shouldn't use the 8 most significant bits in the 32 bits
variables to avoid overflow in modify_qp.
Fixed the PSN generation by the RDMA-CM to mask out the 8 most
From: Majd Dibbiny m...@mellanox.com
Fixed off-by-one bug, we need to decrement the port number only after we
released the resources to the current port.
Call ib_unregister_mad_agent only for cases where ib_register_mad_agent
succeeded.
Separate the ib_register_event_handler() call error flow
From: Moshe Lazer mos...@mellanox.com
The deadlock occurs on __uverbs_modify_qp, we take a lock (idr_read_qp)
and in case of failure in ib_resolve_eth_l2_attrs we don't release
it (put_qp_read). Fix that.
Issue: 355606
Fixes: ed4c54e5b4ba (IB/core: Resolve Ethernet L2 addresses when modifying
-- resent due to Gerrit IDs that were left by mistake, sorry for spamming you
over
Hi Roland,
This is a batch with small IB core fixes, please apply for 3.20
There's one fix which touches few HW drivers to make sure
they invoke ib_unregister_mad_agent() only for valid cases
in their error
From: Ilya Nelkenbaum il...@mellanox.com
When marsheling a user path to the kernel struct ib_sa_path, need
to zero smac, dmac and set the vlan id to the no vlan value.
This is to ensure that Ethernet attributes are not used with
InfiniBand QPs.
Fixes: dd5f03beb4f7 (IB/core: Ethernet L2
From: Majd Dibbiny m...@mellanox.com
In some error flows, ib_mad_unregister_agent is being invoked also in cases
where
the ib_mad_register_agent call failed (resulting in an illegal pointer in the
agent field). This causes a kernel crash in the error flow.
Fix this by calling
From: Ilya Nelkenbaum il...@mellanox.com
When marsheling a user path to the kernel struct ib_sa_path, need
to zero smac, dmac and set the vlan id to the no vlan value.
This is to ensure that Ethernet attributes are not used with
InfiniBand QPs.
Fixes: dd5f03beb4f7 (IB/core: Ethernet L2
From: Majd Dibbiny m...@mellanox.com
The rq/sq-psn is 24 bits as defined in the IB spec, therefore ULPs and User
space applications shouldn't use the 8 most significant bits in the 32 bits
variables to avoid overflow in modify_qp.
Fixed the PSN generation by the RDMA-CM to mask out the 8 most
From: Majd Dibbiny m...@mellanox.com
Fixed off-by-one bug, we need to decrement the port number only after we
released the resources to the current port.
Call ib_unregister_mad_agent only for cases where ib_register_mad_agent
succeeded.
Separate the ib_register_event_handler() call error flow
From: Majd Dibbiny m...@mellanox.com
In some error flows, ib_mad_unregister_agent is being invoked also in cases
where
the ib_mad_register_agent call failed (resulting in an illegal pointer in the
agent field). This causes a kernel crash in the error flow.
Fix this by calling
Implement the following libibvebrs Memory Window verbs:
1. ibv_alloc_mw
2. ibv_dealloc_mw
3. ibv_bind_mw
Signed-off-by: Majd Dibbiny m...@mellanox.com
---
src/cq.c|3 ++
src/mlx4.c |3 ++
src/mlx4.h |5
src/qp.c| 54 +
Send with invalidate can be used to invalidate a remote memory region or memory
window. The complete details are specified in the InfiniBand Architecture
Specifications, section 9.4.1.1.
According to the specs, the following objects can be invalidated:
1. Memory Window type 2
2. Physical MR
3.
This patch set implements the Memory Windows support that was introduced in
libibverbs:
http://www.mail-archive.com/linux-rdma@vger.kernel.org/msg22856.html
This patch set includes two patches:
Patch #1: Implements the Memory Windows verbs; ibv_alloc_mw, ibv_bind_mw and
ibv_dealloc_mw.
Patch #2:
From: Majd Dibbiny m...@mellanox.com
The rq/sq-psn is 24 bits as defined in the IB spec, therefore ULPs and User
space applications shouldn't use the 8 most significant bits in the 32 bits
variables to avoid overflow in modify_qp.
Fixed the PSN generation by the RDMA-CM to mask out the 8 most
From: Majd Dibbiny m...@mellanox.com
In some error flows, ib_mad_unregister_agent is being invoked also in cases
where
the ib_mad_register_agent call failed (resulting in an illegal pointer in the
agent field). This causes a kernel crash in the error flow.
Fix this by calling
From: Ilya Nelkenbaum il...@mellanox.com
When marsheling a user path to the kernel struct ib_sa_path, need
to zero smac, dmac and set the vlan id to the no vlan value.
This is to ensure that Ethernet attributes are not used with
InfiniBand QPs.
Fixes: dd5f03beb4f7 (IB/core: Ethernet L2
Hi Roland,
This is a batch with small IB core fixes, please apply for 3.20
There's one fix which touches few HW drivers to make sure
they invoke ib_unregister_mad_agent() only for valid cases
in their error flows.
There's a pending CMA patch we sent for 3.19 and was approved
by Sean, please
From: Majd Dibbiny m...@mellanox.com
Align the ibv_send_wr structure according to the infiniband/verbs.h
Signed-off-by: Majd Dibbiny m...@mellanox.com
---
man/ibv_post_send.3 | 10 ++
1 files changed, 10 insertions(+), 0 deletions(-)
diff --git a/man/ibv_post_send.3
From: Markus Elfring elfr...@users.sourceforge.net
Date: Thu, 5 Feb 2015 13:20:42 +0100
The srp_destroy_fr_pool() function tests whether its argument is NULL and then
returns immediately. Thus the test around the call is not needed.
This issue was detected by using the Coccinelle software.
From: Moshe Lazer mos...@mellanox.com
The deadlock occurs on __uverbs_modify_qp, we take a lock (idr_read_qp)
and in case of failure in ib_resolve_eth_l2_attrs we don't release
it (put_qp_read). Fix that.
Issue: 355606
Fixes: ed4c54e5b4ba (IB/core: Resolve Ethernet L2 addresses when modifying
From: Majd Dibbiny m...@mellanox.com
Fixed off-by-one bug, we need to decrement the port number only after we
released the resources to the current port.
Call ib_unregister_mad_agent only for cases where ib_register_mad_agent
succeeded.
Separate the ib_register_event_handler() call error flow
36 matches
Mail list logo