* Re: [PATCH 6/6] IB: userspace support for RDMA connection manager
From: Roland Dreier @ 2006-03-06 21:33 UTC (permalink / raw)
To: Sean Hefty; +Cc: netdev, linux-kernel, openib-general
In-Reply-To: <ORSMSX4011XvpFVjCRG00000009@orsmsx401.amr.corp.intel.com>
> +struct rdma_ucm_query_route_resp {
> + __u64 node_guid;
> + struct ib_user_path_rec ib_route[2];
> + struct sockaddr_in6 src_addr;
> + struct sockaddr_in6 dst_addr;
> + __u32 num_paths;
> + __u8 port_num;
> + __u8 reserved[3];
> +};
Is there a 32-bit/64-bit compatibility problem here? From a quick
look, struct sockaddr_in6 is not 8-byte aligned.
- R.
^ permalink raw reply
* Re: [PATCH 3/6] net/IB: export ip_dev_find
From: Roland Dreier @ 2006-03-06 21:31 UTC (permalink / raw)
To: Sean Hefty; +Cc: netdev, linux-kernel, openib-general
In-Reply-To: <ORSMSX4011XvpFVjCRG00000006@orsmsx401.amr.corp.intel.com>
Sean> Export ip_dev_find to allow locating a net_device given an
Sean> IP address.
My plan is to queue all of this stuff for merging in 2.6.17.
Is there any objection from netdev or openib-general people?
I just looked back, and the original "unexport ip_dev_find()" patch
was a de-Bunk-ing change. Now that there is a modular user, is there
any problem with re-exporting it?
Thanks,
Roland
^ permalink raw reply
* Re: de2104x: interrupts before interrupt handler is registered
From: Francois Romieu @ 2006-03-06 21:17 UTC (permalink / raw)
To: Martin Michlmayr; +Cc: netdev, linux-kernel
In-Reply-To: <20060306191706.GA6947@deprecation.cyrius.com>
Martin Michlmayr <tbm@cyrius.com> :
[...]
> There's another interrupt related bug in the driver, though. I
> sometimes get a kernel panic when rsycing several 100 megs of data
> across the LAN. A picture showing the call trace can be found at
> http://www.cyrius.com/tmp/de2104x_panic.jpg
Not sure about this one, but...
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
diff --git a/drivers/net/tulip/de2104x.c b/drivers/net/tulip/de2104x.c
index d7fb3ff..49235e2 100644
--- a/drivers/net/tulip/de2104x.c
+++ b/drivers/net/tulip/de2104x.c
@@ -1455,6 +1455,8 @@ static void de_tx_timeout (struct net_de
synchronize_irq(dev->irq);
de_clean_rings(de);
+ de_init_rings(de);
+
de_init_hw(de);
netif_wake_queue(dev);
^ permalink raw reply related
* Re: de2104x: interrupts before interrupt handler is registered
From: Francois Romieu @ 2006-03-06 20:54 UTC (permalink / raw)
To: Martin Michlmayr; +Cc: netdev, linux-kernel
In-Reply-To: <20060306195953.GB10703@deprecation.cyrius.com>
Martin Michlmayr <tbm@cyrius.com> :
[...]
> By the way, I'm getting the following messages in dmesg:
>
> eth0: tx err, status 0x7fffb002
Tx underrun.
Is there anything which could induce a noticeable load on the PCI bus ?
--
Ueimor
^ permalink raw reply
* RE: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
From: Caitlin Bestler @ 2006-03-06 20:32 UTC (permalink / raw)
To: David Stevens; +Cc: Linux Kernel Mailing List, Michael S. Tsirkin, netdev
> -----Original Message-----
> From: David Stevens [mailto:dlstevens@us.ibm.com]
> Sent: Monday, March 06, 2006 12:32 PM
> To: Caitlin Bestler
> Cc: Linux Kernel Mailing List; Michael S. Tsirkin;
> netdev@vger.kernel.org
> Subject: RE: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
>
> IPPROTO_* should match the protocol field on the wire, which
> I gather isn't different. And I'm assuming there is no
> standard API defined already...
>
SDP uses the existing standard sockets API.
That was the intent in its design, and it is
the sole justification for its use. If you are
not using the existing sockets API then your
application would be *far* better off coding
directly to RDMA.
The wire protocol *is* different, it uses RDMA.
There is some justification for the application
knowing this, albeit slight ones. For example
you need to know if the peer supports SDP and
it might effect how intermediate firewalls
need to be configured.
^ permalink raw reply
* RE: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
From: David Stevens @ 2006-03-06 20:31 UTC (permalink / raw)
To: Caitlin Bestler; +Cc: Linux Kernel Mailing List, Michael S. Tsirkin, netdev
In-Reply-To: <54AD0F12E08D1541B826BE97C98F99F12FBF33@NT-SJCA-0751.brcm.ad.broadcom.com>
IPPROTO_* should match the protocol field on the wire, which
I gather isn't different. And I'm assuming there is no standard API
defined already...
What about using a socket option?
+-DLS
^ permalink raw reply
* Re: de2104x: interrupts before interrupt handler is registered
From: Martin Michlmayr @ 2006-03-06 20:29 UTC (permalink / raw)
To: Francois Romieu; +Cc: netdev, linux-kernel
In-Reply-To: <20060306202330.GB15728@electric-eye.fr.zoreil.com>
* Francois Romieu <romieu@fr.zoreil.com> [2006-03-06 21:23]:
> > http://www.cyrius.com/tmp/config-2.6.16-rc5-486
> > By the way, I'm getting the following messages in dmesg:
> netconsole appears enabled. Do you use it ?
It's a standard Debian kernel config so pretty much everything is
enabled as a module. I didn't use netconsole.
--
Martin Michlmayr
http://www.cyrius.com/
^ permalink raw reply
* Re: de2104x: interrupts before interrupt handler is registered
From: Francois Romieu @ 2006-03-06 20:23 UTC (permalink / raw)
To: Martin Michlmayr; +Cc: netdev, linux-kernel
In-Reply-To: <20060306195953.GB10703@deprecation.cyrius.com>
Martin Michlmayr <tbm@cyrius.com> :
[...]
> http://www.cyrius.com/tmp/config-2.6.16-rc5-486
>
> By the way, I'm getting the following messages in dmesg:
netconsole appears enabled. Do you use it ?
--
Ueimor
^ permalink raw reply
* RE: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
From: Caitlin Bestler @ 2006-03-06 20:16 UTC (permalink / raw)
To: David Stevens, Michael S. Tsirkin; +Cc: Linux Kernel Mailing List, netdev
> -----Original Message-----
> From: netdev-owner@vger.kernel.org
> [mailto:netdev-owner@vger.kernel.org] On Behalf Of David Stevens
> Sent: Monday, March 06, 2006 11:49 AM
> To: Michael S. Tsirkin
> Cc: Linux Kernel Mailing List; netdev@vger.kernel.org
> Subject: Re: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
>
> I don't know any details about SDP, but if there are no
> differences at the protocol layer, then neither the address
> family nor the protocol is appropriate. If it's just an API
> change, the socket type is the right selector. So, maybe
> SOCK_DIRECT to go along with SOCK_STREAM, SOCK_DGRAM, etc.
> +-DLS
That wouldn't work either. The whole point of SDP, or TOE,
is that the API is either totally unchanged or at least
essentially unchanged.
Whenever an IP Address is used (SDP/iWARP, TOE and potentially
SDP/IB) changing from AF_INET* is wrong.
For both SDP/iWARP and SDP/IB you could argue that a different
wire protocol is in use so IPPROTO_SDP is acceptable.
That's probably the best answer as long as we are stuck
under the restriction that the selection of an alternate
stack cannot be done in the exact manner that the consumer
wants it done (that is transparently to the application).
There are even some corner case scenarios where the
application might care whether their SOCK_STREAM was
carried over SDP or plain TCP. So a protocol based
distinction is probably the least misleading of all
the explicit selection options.
^ permalink raw reply
* Re: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
From: Michael S. Tsirkin @ 2006-03-06 20:16 UTC (permalink / raw)
To: David Stevens; +Cc: Linux Kernel Mailing List, netdev
In-Reply-To: <OFB3CB1E4D.355F7555-ON88257129.006C1001-88257129.006C836D@us.ibm.com>
Quoting r. David Stevens <dlstevens@us.ibm.com>:
> Subject: Re: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
>
> I don't know any details about SDP, but if there are no differences at the
> protocol layer, then neither the address family nor the protocol is
> appropriate. If it's just an API change, the socket type is the right
> selector. So, maybe SOCK_DIRECT to go along with SOCK_STREAM,
> SOCK_DGRAM, etc.
No, the API SDP implements is the regular SOCK_STREAM semantics.
The difference is in the way connections are established with
infiniband connection management messages, and data is transferred
with infiniband reliable connection send messages.
--
Michael S. Tsirkin
Staff Engineer, Mellanox Technologies
^ permalink raw reply
* Re: de2104x: interrupts before interrupt handler is registered
From: Martin Michlmayr @ 2006-03-06 19:59 UTC (permalink / raw)
To: Francois Romieu; +Cc: netdev, linux-kernel
In-Reply-To: <20060306194821.GA15728@electric-eye.fr.zoreil.com>
* Francois Romieu <romieu@fr.zoreil.com> [2006-03-06 20:48]:
> > There's another interrupt related bug in the driver, though. I
> > sometimes get a kernel panic when rsycing several 100 megs of data
> > across the LAN. A picture showing the call trace can be found at
> > http://www.cyrius.com/tmp/de2104x_panic.jpg
> Can you publish the .config ?
http://www.cyrius.com/tmp/config-2.6.16-rc5-486
By the way, I'm getting the following messages in dmesg:
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb00a
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb00a
eth0: tx err, status 0x7fffb00a
eth0: tx err, status 0x7fffb00a
eth0: tx err, status 0x7fffb002
eth0: tx err, status 0x7fffb032
eth0: tx err, status 0x7fffb002
--
Martin Michlmayr
http://www.cyrius.com/
^ permalink raw reply
* Re: Discourage duplicate symbols in the kernel? [Was: Intel I/O Acc...]
From: Chris Leech @ 2006-03-06 19:56 UTC (permalink / raw)
To: Andrew Morton; +Cc: Sam Ravnborg, jeff, linux-kernel, netdev
In-Reply-To: <20060305011852.368c016e.akpm@osdl.org>
On 3/5/06, Andrew Morton <akpm@osdl.org> wrote:
> Sam Ravnborg <sam@ravnborg.org> wrote:
> >
> > On Sun, Mar 05, 2006 at 12:09:33AM -0800, Andrew Morton wrote:
> > > > +
> > > > +static inline u8 read_reg8(struct cb_device *device, unsigned int offset)
> > > > +{
> > > > + return readb(device->reg_base + offset);
> > > > +}
> > >
> > > These are fairly generic-sounding names. In fact the as-yet-unmerged tiacx
> > > wireless driver is already using these, privately to
> > > drivers/net/wireless/tiacx/pci.c.
> >
> > Do we in general discourage duplicate symbols even if they are static?
>
> Well, it's a bit irritating that it confuses ctags. But in this case, one
> set is in a header file so the risk of collisions is much-increased.
They're in a header file that's specific to a single driver, so I
don't see where a conflict would occur. But I didn't think about
ctags, and these can easily be prefixed so I'll go ahead and change
them.
- Chris
^ permalink raw reply
* Re: RFC: move SDP from AF_INET_SDP to IPPROTO_SDP
From: David Stevens @ 2006-03-06 19:48 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: Linux Kernel Mailing List, netdev
In-Reply-To: <20060306190636.GA14849@mellanox.co.il>
I don't know any details about SDP, but if there are no differences at the
protocol layer, then neither the address family nor the protocol is
appropriate. If it's just an API change, the socket type is the right
selector. So, maybe SOCK_DIRECT to go along with SOCK_STREAM,
SOCK_DGRAM, etc.
+-DLS
^ permalink raw reply
* Re: de2104x: interrupts before interrupt handler is registered
From: Francois Romieu @ 2006-03-06 19:48 UTC (permalink / raw)
To: Martin Michlmayr; +Cc: netdev, linux-kernel
In-Reply-To: <20060306191706.GA6947@deprecation.cyrius.com>
Martin Michlmayr <tbm@cyrius.com> :
[...]
> There's another interrupt related bug in the driver, though. I
> sometimes get a kernel panic when rsycing several 100 megs of data
> across the LAN. A picture showing the call trace can be found at
> http://www.cyrius.com/tmp/de2104x_panic.jpg
Can you publish the .config ?
--
Ueimor
^ permalink raw reply
* Re: [PATCH 1/8] [I/OAT] DMA memcpy subsystem
From: Chris Leech @ 2006-03-06 19:48 UTC (permalink / raw)
To: Benjamin LaHaise; +Cc: linux-kernel, netdev
In-Reply-To: <20060304192030.GA6510@kvack.org>
On 3/4/06, Benjamin LaHaise <bcrl@kvack.org> wrote:
> On Fri, Mar 03, 2006 at 01:42:20PM -0800, Chris Leech wrote:
> > +void dma_async_device_unregister(struct dma_device* device)
> > +{
> ...
> > + kref_put(&device->refcount, dma_async_device_cleanup);
> > + wait_for_completion(&device->done);
> > +}
>
> This looks like a bug: device is dereferenced after it is potentially
> freed.
Actually, this is where the code is waiting to make sure it's safe to
free device. The release function for the kref completes
device->done. Each of the devices channels holds a reference to the
device. When a device is unregistered it's channels are removed from
the clients, which hold a reference for each outstanding transaction.
When all the outstanding transactions complete, the channels kref goes
to 0, and the reference to the device is dropped. When the device
kref goes to 0 the completion is set, and it's now safe to free the
memory for the device and channel structures.
I have a writeup of the locking and reference counting that I'll
finish and add in as a big comment to the code.
-Chris
^ permalink raw reply
* Re: [PATCH 1/8] [I/OAT] DMA memcpy subsystem
From: Chris Leech @ 2006-03-06 19:39 UTC (permalink / raw)
To: David S. Miller; +Cc: linux-kernel, netdev
In-Reply-To: <20060303.174048.14793187.davem@davemloft.net>
On 3/3/06, David S. Miller <davem@davemloft.net> wrote:
> > +static spinlock_t dma_list_lock;
>
> Please use DEFINE_SPINLOCK().
>
> > +static void dma_chan_free_rcu(struct rcu_head *rcu) {
>
> Newline before the brace please.
>
> > +static void dma_async_device_cleanup(struct kref *kref) {
>
> Newline before the brace please.
>
> > +struct dma_chan_percpu
> > +{
>
> Left brace on the same line as "struct dma_chan_percpu" please.
>
> > +struct dma_chan
> > +{
>
> Similarly.
>
> Otherwise this patch looks mostly ok.
Thanks Dave,
I'll apply these and other feedback and get updated patches generated.
- Chris
^ permalink raw reply
* [git patches] net driver fixes
From: Jeff Garzik @ 2006-03-06 19:39 UTC (permalink / raw)
To: Andrew Morton, Linus Torvalds; +Cc: netdev, linux-kernel
Please pull from 'upstream-fixes' branch of
master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6.git
to receive the following updates:
drivers/net/chelsio/espi.c | 4 +---
drivers/net/s2io.c | 1 +
2 files changed, 2 insertions(+), 3 deletions(-)
Eric Sesterhenn:
chelsio: fix kmalloc failure in t1_espi_create
Jeff Garzik:
s2io: set_multicast_list bug
diff --git a/drivers/net/chelsio/espi.c b/drivers/net/chelsio/espi.c
index 2306425..e824aca 100644
--- a/drivers/net/chelsio/espi.c
+++ b/drivers/net/chelsio/espi.c
@@ -296,9 +296,7 @@ void t1_espi_destroy(struct peespi *espi
struct peespi *t1_espi_create(adapter_t *adapter)
{
- struct peespi *espi = kmalloc(sizeof(*espi), GFP_KERNEL);
-
- memset(espi, 0, sizeof(*espi));
+ struct peespi *espi = kzalloc(sizeof(*espi), GFP_KERNEL);
if (espi)
espi->adapter = adapter;
diff --git a/drivers/net/s2io.c b/drivers/net/s2io.c
index 49b597c..b7f00d6 100644
--- a/drivers/net/s2io.c
+++ b/drivers/net/s2io.c
@@ -4092,6 +4092,7 @@ static void s2io_set_multicast(struct ne
i++, mclist = mclist->next) {
memcpy(sp->usr_addrs[i].addr, mclist->dmi_addr,
ETH_ALEN);
+ mac_addr = 0;
for (j = 0; j < ETH_ALEN; j++) {
mac_addr |= mclist->dmi_addr[j];
mac_addr <<= 8;
^ permalink raw reply related
* Re: [PATCH 8/8] [I/OAT] TCP recv offload to I/OAT
From: Chris Leech @ 2006-03-06 19:36 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-kernel, netdev
In-Reply-To: <20060305004534.1d94b3cf.akpm@osdl.org>
On 3/5/06, Andrew Morton <akpm@osdl.org> wrote:
> Chris Leech <christopher.leech@intel.com> wrote:
> >
> > +#ifdef CONFIG_NET_DMA
> > + tp->ucopy.dma_chan = NULL;
> > + if ((len > sysctl_tcp_dma_copybreak) && !(flags & MSG_PEEK) && !sysctl_tcp_low_latency && __get_cpu_var(softnet_data.net_dma))
> > + dma_lock_iovec_pages(msg->msg_iov, len, &tp->ucopy.locked_list);
> > +#endif
>
> The __get_cpu_var() here will run smp_processor_id() from preemptible
> context. You'll get a big warning if the correct debug options are set.
>
> The reason for this is that preemption could cause this code to hop between
> CPUs.
I've been playing with different models of where to select which DMA
channel to use in order to reduce cache thrash and lock contention in
the driver. It's not a clean per-cpu issue because per I/O there are
potentially operations happening in both the process syscall and the
netrx softirq context.
Right now the code delays selection of a DMA channel until the first
offload copy is ready to go, so the __get_cpu_var() you point out is
just checking to see if any hardware exists for I/OAT at this point
before doing the page pinning. Before anything is done with the
channel the per-cpu pointer is re-read safely with preemption disabled
and a reference count is incremented.
- Chris
^ permalink raw reply
* Re: RE: [PATCH 2/6] IB: match connection requests based on private data
From: Sean Hefty @ 2006-03-06 19:31 UTC (permalink / raw)
To: Caitlin Bestler; +Cc: netdev, Roland Dreier, openib-general, linux-kernel
In-Reply-To: <54AD0F12E08D1541B826BE97C98F99F12FBF19@NT-SJCA-0751.brcm.ad.broadcom.com>
Caitlin Bestler wrote:
> The term "private data" is intended to convey the
> intent that the data is private to the application
> layer and is opaque to middleware and the network.
The private data area is for the use of whatever client resides above the
Infiniband CM only. There is no assumption about whether that client is
middleware or an application.
> By what mechanism does the listening application
> delegate how much of the private data for use by
> the CM for sub-dividing a listen? What does an
> application do if it wishes to retain full ownership
> of the private data?
An application that interfaces directly with the Infiniband CM always retains
full control of any private data. Applications that interface to middleware are
restricted by the limitations of that middleware layer.
- Sean
^ permalink raw reply
* Re: [PATCH 8/8] [I/OAT] TCP recv offload to I/OAT
From: Chris Leech @ 2006-03-06 19:28 UTC (permalink / raw)
To: Greg KH; +Cc: linux-kernel, netdev
In-Reply-To: <20060304231842.GA3103@kroah.com>
> #ifdef is not needed here (try not to put #ifdef in .c files.) I think
> a few of your other usages of #ifdef in this file can also be removed
> with judicious use of inline functions in a .h file.
ACK on all the ifdef comments. I may have gone a little ifdef crazy
making sure I could get to a zero impact state with these patches
applied but CONFIG_NET_DMA turned off. I'll get these cleaned up.
- Chris
^ permalink raw reply
* [PATCH 6/6] IB: userspace support for RDMA connection manager
From: Sean Hefty @ 2006-03-06 19:21 UTC (permalink / raw)
To: 'Roland Dreier', netdev, linux-kernel; +Cc: openib-general
In-Reply-To: <adaslpz2l9p.fsf@cisco.com>
Kernel component necessary to support the userspace RDMA connection management
library.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
---
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/drivers/infiniband/core/Makefile
linux-2.6.ib/drivers/infiniband/core/Makefile
--- linux-2.6.git/drivers/infiniband/core/Makefile 2006-01-16 16:58:58.000000000 -0800
+++ linux-2.6.ib/drivers/infiniband/core/Makefile 2006-01-16 16:55:25.000000000 -0800
@@ -1,5 +1,5 @@
obj-$(CONFIG_INFINIBAND) += ib_core.o ib_mad.o ib_sa.o \
- ib_cm.o ib_addr.o rdma_cm.o
+ ib_cm.o ib_addr.o rdma_cm.o rdma_ucm.o
obj-$(CONFIG_INFINIBAND_USER_MAD) += ib_umad.o
obj-$(CONFIG_INFINIBAND_USER_ACCESS) += ib_uverbs.o ib_ucm.o
@@ -14,6 +14,8 @@ ib_cm-y := cm.o
rdma_cm-y := cma.o
+rdma_ucm-y := ucma.o
+
ib_addr-y := addr.o
ib_umad-y := user_mad.o
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/drivers/infiniband/core/ucma.c
linux-2.6.ib/drivers/infiniband/core/ucma.c
--- linux-2.6.git/drivers/infiniband/core/ucma.c 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.ib/drivers/infiniband/core/ucma.c 2006-01-16 16:54:31.000000000 -0800
@@ -0,0 +1,788 @@
+/*
+ * Copyright (c) 2005 Intel Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/poll.h>
+#include <linux/idr.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/miscdevice.h>
+
+#include <rdma/rdma_user_cm.h>
+#include <rdma/ib_marshall.h>
+#include <rdma/rdma_cm.h>
+
+MODULE_AUTHOR("Sean Hefty");
+MODULE_DESCRIPTION("RDMA Userspace Connection Manager Access");
+MODULE_LICENSE("Dual BSD/GPL");
+
+enum {
+ UCMA_MAX_BACKLOG = 128
+};
+
+struct ucma_file {
+ struct semaphore mutex;
+ struct file *filp;
+ struct list_head ctxs;
+ struct list_head events;
+ wait_queue_head_t poll_wait;
+};
+
+struct ucma_context {
+ int id;
+ wait_queue_head_t wait;
+ atomic_t ref;
+ int events_reported;
+ int backlog;
+
+ struct ucma_file *file;
+ struct rdma_cm_id *cm_id;
+ __u64 uid;
+
+ struct list_head events; /* list of pending events. */
+ struct list_head file_list; /* member in file ctx list */
+};
+
+struct ucma_event {
+ struct ucma_context *ctx;
+ struct list_head file_list; /* member in file event list */
+ struct list_head ctx_list; /* member in ctx event list */
+ struct rdma_cm_id *cm_id;
+ struct rdma_ucm_event_resp resp;
+};
+
+static DECLARE_MUTEX(ctx_mutex);
+static DEFINE_IDR(ctx_idr);
+
+static struct ucma_context* ucma_get_ctx(struct ucma_file *file, int id)
+{
+ struct ucma_context *ctx;
+
+ down(&ctx_mutex);
+ ctx = idr_find(&ctx_idr, id);
+ if (!ctx)
+ ctx = ERR_PTR(-ENOENT);
+ else if (ctx->file != file)
+ ctx = ERR_PTR(-EINVAL);
+ else
+ atomic_inc(&ctx->ref);
+ up(&ctx_mutex);
+
+ return ctx;
+}
+
+static void ucma_put_ctx(struct ucma_context *ctx)
+{
+ if (atomic_dec_and_test(&ctx->ref))
+ wake_up(&ctx->wait);
+}
+
+static void ucma_cleanup_events(struct ucma_context *ctx)
+{
+ struct ucma_event *uevent;
+
+ down(&ctx->file->mutex);
+ list_del(&ctx->file_list);
+ while (!list_empty(&ctx->events)) {
+
+ uevent = list_entry(ctx->events.next, struct ucma_event,
+ ctx_list);
+ list_del(&uevent->file_list);
+ list_del(&uevent->ctx_list);
+
+ /* clear incoming connections. */
+ if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST)
+ rdma_destroy_id(uevent->cm_id);
+
+ kfree(uevent);
+ }
+ up(&ctx->file->mutex);
+}
+
+static struct ucma_context* ucma_alloc_ctx(struct ucma_file *file)
+{
+ struct ucma_context *ctx;
+ int ret;
+
+ ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
+ if (!ctx)
+ return NULL;
+
+ atomic_set(&ctx->ref, 1);
+ init_waitqueue_head(&ctx->wait);
+ ctx->file = file;
+ INIT_LIST_HEAD(&ctx->events);
+
+ do {
+ ret = idr_pre_get(&ctx_idr, GFP_KERNEL);
+ if (!ret)
+ goto error;
+
+ down(&ctx_mutex);
+ ret = idr_get_new(&ctx_idr, ctx, &ctx->id);
+ up(&ctx_mutex);
+ } while (ret == -EAGAIN);
+
+ if (ret)
+ goto error;
+
+ list_add_tail(&ctx->file_list, &file->ctxs);
+ return ctx;
+
+error:
+ kfree(ctx);
+ return NULL;
+}
+
+static int ucma_event_handler(struct rdma_cm_id *cm_id,
+ struct rdma_cm_event *event)
+{
+ struct ucma_event *uevent;
+ struct ucma_context *ctx = cm_id->context;
+ int ret = 0;
+
+ uevent = kzalloc(sizeof(*uevent), GFP_KERNEL);
+ if (!uevent)
+ return event->event == RDMA_CM_EVENT_CONNECT_REQUEST;
+
+ uevent->ctx = ctx;
+ uevent->cm_id = cm_id;
+ uevent->resp.uid = ctx->uid;
+ uevent->resp.id = ctx->id;
+ uevent->resp.event = event->event;
+ uevent->resp.status = event->status;
+ if ((uevent->resp.private_data_len = event->private_data_len))
+ memcpy(uevent->resp.private_data, event->private_data,
+ event->private_data_len);
+
+ down(&ctx->file->mutex);
+ if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST) {
+ if (!ctx->backlog) {
+ ret = -EDQUOT;
+ goto out;
+ }
+ ctx->backlog--;
+ }
+ list_add_tail(&uevent->file_list, &ctx->file->events);
+ list_add_tail(&uevent->ctx_list, &ctx->events);
+ wake_up_interruptible(&ctx->file->poll_wait);
+out:
+ up(&ctx->file->mutex);
+ return ret;
+}
+
+static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct ucma_context *ctx;
+ struct rdma_ucm_get_event cmd;
+ struct ucma_event *uevent;
+ int ret = 0;
+ DEFINE_WAIT(wait);
+
+ if (out_len < sizeof(struct rdma_ucm_event_resp))
+ return -ENOSPC;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ down(&file->mutex);
+ while (list_empty(&file->events)) {
+ if (file->filp->f_flags & O_NONBLOCK) {
+ ret = -EAGAIN;
+ break;
+ }
+
+ if (signal_pending(current)) {
+ ret = -ERESTARTSYS;
+ break;
+ }
+
+ prepare_to_wait(&file->poll_wait, &wait, TASK_INTERRUPTIBLE);
+ up(&file->mutex);
+ schedule();
+ down(&file->mutex);
+ finish_wait(&file->poll_wait, &wait);
+ }
+
+ if (ret)
+ goto done;
+
+ uevent = list_entry(file->events.next, struct ucma_event, file_list);
+
+ if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST) {
+ ctx = ucma_alloc_ctx(file);
+ if (!ctx) {
+ ret = -ENOMEM;
+ goto done;
+ }
+ uevent->ctx->backlog++;
+ ctx->cm_id = uevent->cm_id;
+ ctx->cm_id->context = ctx;
+ uevent->resp.id = ctx->id;
+ }
+
+ if (copy_to_user((void __user *)(unsigned long)cmd.response,
+ &uevent->resp, sizeof(uevent->resp))) {
+ ret = -EFAULT;
+ goto done;
+ }
+
+ list_del(&uevent->file_list);
+ list_del(&uevent->ctx_list);
+ uevent->ctx->events_reported++;
+ kfree(uevent);
+done:
+ up(&file->mutex);
+ return ret;
+}
+
+static ssize_t ucma_create_id(struct ucma_file *file,
+ const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_create_id cmd;
+ struct rdma_ucm_create_id_resp resp;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (out_len < sizeof(resp))
+ return -ENOSPC;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ down(&file->mutex);
+ ctx = ucma_alloc_ctx(file);
+ up(&file->mutex);
+ if (!ctx)
+ return -ENOMEM;
+
+ ctx->uid = cmd.uid;
+ ctx->cm_id = rdma_create_id(ucma_event_handler, ctx, RDMA_PS_TCP);
+ if (IS_ERR(ctx->cm_id)) {
+ ret = PTR_ERR(ctx->cm_id);
+ goto err1;
+ }
+
+ resp.id = ctx->id;
+ if (copy_to_user((void __user *)(unsigned long)cmd.response,
+ &resp, sizeof(resp))) {
+ ret = -EFAULT;
+ goto err2;
+ }
+ return 0;
+
+err2:
+ rdma_destroy_id(ctx->cm_id);
+err1:
+ down(&ctx_mutex);
+ idr_remove(&ctx_idr, ctx->id);
+ up(&ctx_mutex);
+ kfree(ctx);
+ return ret;
+}
+
+static ssize_t ucma_destroy_id(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_destroy_id cmd;
+ struct rdma_ucm_destroy_id_resp resp;
+ struct ucma_context *ctx;
+ int ret = 0;
+
+ if (out_len < sizeof(resp))
+ return -ENOSPC;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ down(&ctx_mutex);
+ ctx = idr_find(&ctx_idr, cmd.id);
+ if (!ctx)
+ ctx = ERR_PTR(-ENOENT);
+ else if (ctx->file != file)
+ ctx = ERR_PTR(-EINVAL);
+ else
+ idr_remove(&ctx_idr, ctx->id);
+ up(&ctx_mutex);
+
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ atomic_dec(&ctx->ref);
+ wait_event(ctx->wait, !atomic_read(&ctx->ref));
+
+ /* No new events will be generated after destroying the id. */
+ rdma_destroy_id(ctx->cm_id);
+ /* Cleanup events not yet reported to the user. */
+ ucma_cleanup_events(ctx);
+
+ resp.events_reported = ctx->events_reported;
+ if (copy_to_user((void __user *)(unsigned long)cmd.response,
+ &resp, sizeof(resp)))
+ ret = -EFAULT;
+
+ kfree(ctx);
+ return ret;
+}
+
+static ssize_t ucma_bind_addr(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_bind_addr cmd;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_resolve_addr(struct ucma_file *file,
+ const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_resolve_addr cmd;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr,
+ (struct sockaddr *) &cmd.dst_addr,
+ cmd.timeout_ms);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_resolve_route(struct ucma_file *file,
+ const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_resolve_route cmd;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ret = rdma_resolve_route(ctx->cm_id, cmd.timeout_ms);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static void ucma_copy_ib_route(struct rdma_ucm_query_route_resp *resp,
+ struct rdma_route *route)
+{
+ struct rdma_dev_addr *dev_addr;
+
+ resp->num_paths = route->num_paths;
+ switch (route->num_paths) {
+ case 0:
+ dev_addr = &route->addr.dev_addr;
+ memcpy(&resp->ib_route[0].dgid, ib_addr_get_dgid(dev_addr),
+ sizeof(union ib_gid));
+ memcpy(&resp->ib_route[0].sgid, ib_addr_get_sgid(dev_addr),
+ sizeof(union ib_gid));
+ resp->ib_route[0].pkey = cpu_to_be16(ib_addr_get_pkey(dev_addr));
+ break;
+ case 2:
+ ib_copy_path_rec_to_user(&resp->ib_route[1],
+ &route->path_rec[1]);
+ /* fall through */
+ case 1:
+ ib_copy_path_rec_to_user(&resp->ib_route[0],
+ &route->path_rec[0]);
+ break;
+ default:
+ break;
+ }
+}
+
+static ssize_t ucma_query_route(struct ucma_file *file,
+ const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_query_route cmd;
+ struct rdma_ucm_query_route_resp resp;
+ struct ucma_context *ctx;
+ struct sockaddr *addr;
+ int ret = 0;
+
+ if (out_len < sizeof(resp))
+ return -ENOSPC;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ if (!ctx->cm_id->device) {
+ ret = -ENODEV;
+ goto out;
+ }
+
+ addr = &ctx->cm_id->route.addr.src_addr;
+ memcpy(&resp.src_addr, addr, addr->sa_family == AF_INET ?
+ sizeof(struct sockaddr_in) :
+ sizeof(struct sockaddr_in6));
+ addr = &ctx->cm_id->route.addr.dst_addr;
+ memcpy(&resp.dst_addr, addr, addr->sa_family == AF_INET ?
+ sizeof(struct sockaddr_in) :
+ sizeof(struct sockaddr_in6));
+ resp.node_guid = ctx->cm_id->device->node_guid;
+ resp.port_num = ctx->cm_id->port_num;
+ switch (ctx->cm_id->device->node_type) {
+ case IB_NODE_CA:
+ ucma_copy_ib_route(&resp, &ctx->cm_id->route);
+ default:
+ break;
+ }
+
+ if (copy_to_user((void __user *)(unsigned long)cmd.response,
+ &resp, sizeof(resp)))
+ ret = -EFAULT;
+
+out:
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static void ucma_copy_conn_param(struct rdma_conn_param *dst_conn,
+ struct rdma_ucm_conn_param *src_conn)
+{
+ dst_conn->private_data = src_conn->private_data;
+ dst_conn->private_data_len = src_conn->private_data_len;
+ dst_conn->responder_resources =src_conn->responder_resources;
+ dst_conn->initiator_depth = src_conn->initiator_depth;
+ dst_conn->flow_control = src_conn->flow_control;
+ dst_conn->retry_count = src_conn->retry_count;
+ dst_conn->rnr_retry_count = src_conn->rnr_retry_count;
+ dst_conn->srq = src_conn->srq;
+ dst_conn->qp_num = src_conn->qp_num;
+ dst_conn->qp_type = src_conn->qp_type;
+}
+
+static ssize_t ucma_connect(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_connect cmd;
+ struct rdma_conn_param conn_param;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ if (!cmd.conn_param.valid)
+ return -EINVAL;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ucma_copy_conn_param(&conn_param, &cmd.conn_param);
+ ret = rdma_connect(ctx->cm_id, &conn_param);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_listen(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_listen cmd;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ctx->backlog = cmd.backlog > 0 && cmd.backlog < UCMA_MAX_BACKLOG ?
+ cmd.backlog : UCMA_MAX_BACKLOG;
+ ret = rdma_listen(ctx->cm_id, ctx->backlog);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_accept cmd;
+ struct rdma_conn_param conn_param;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ if (cmd.conn_param.valid) {
+ ctx->uid = cmd.uid;
+ ucma_copy_conn_param(&conn_param, &cmd.conn_param);
+ ret = rdma_accept(ctx->cm_id, &conn_param);
+ } else
+ ret = rdma_accept(ctx->cm_id, NULL);
+
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_reject(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_reject cmd;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ret = rdma_reject(ctx->cm_id, cmd.private_data, cmd.private_data_len);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_disconnect(struct ucma_file *file, const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_disconnect cmd;
+ struct ucma_context *ctx;
+ int ret;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ ret = rdma_disconnect(ctx->cm_id);
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t ucma_init_qp_attr(struct ucma_file *file,
+ const char __user *inbuf,
+ int in_len, int out_len)
+{
+ struct rdma_ucm_init_qp_attr cmd;
+ struct ib_uverbs_qp_attr resp;
+ struct ucma_context *ctx;
+ struct ib_qp_attr qp_attr;
+ int ret;
+
+ if (out_len < sizeof(resp))
+ return -ENOSPC;
+
+ if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+ return -EFAULT;
+
+ ctx = ucma_get_ctx(file, cmd.id);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ resp.qp_attr_mask = 0;
+ memset(&qp_attr, 0, sizeof qp_attr);
+ qp_attr.qp_state = cmd.qp_state;
+ ret = rdma_init_qp_attr(ctx->cm_id, &qp_attr, &resp.qp_attr_mask);
+ if (ret)
+ goto out;
+
+ ib_copy_qp_attr_to_user(&resp, &qp_attr);
+ if (copy_to_user((void __user *)(unsigned long)cmd.response,
+ &resp, sizeof(resp)))
+ ret = -EFAULT;
+
+out:
+ ucma_put_ctx(ctx);
+ return ret;
+}
+
+static ssize_t (*ucma_cmd_table[])(struct ucma_file *file,
+ const char __user *inbuf,
+ int in_len, int out_len) = {
+ [RDMA_USER_CM_CMD_CREATE_ID] = ucma_create_id,
+ [RDMA_USER_CM_CMD_DESTROY_ID] = ucma_destroy_id,
+ [RDMA_USER_CM_CMD_BIND_ADDR] = ucma_bind_addr,
+ [RDMA_USER_CM_CMD_RESOLVE_ADDR] = ucma_resolve_addr,
+ [RDMA_USER_CM_CMD_RESOLVE_ROUTE]= ucma_resolve_route,
+ [RDMA_USER_CM_CMD_QUERY_ROUTE] = ucma_query_route,
+ [RDMA_USER_CM_CMD_CONNECT] = ucma_connect,
+ [RDMA_USER_CM_CMD_LISTEN] = ucma_listen,
+ [RDMA_USER_CM_CMD_ACCEPT] = ucma_accept,
+ [RDMA_USER_CM_CMD_REJECT] = ucma_reject,
+ [RDMA_USER_CM_CMD_DISCONNECT] = ucma_disconnect,
+ [RDMA_USER_CM_CMD_INIT_QP_ATTR] = ucma_init_qp_attr,
+ [RDMA_USER_CM_CMD_GET_EVENT] = ucma_get_event
+};
+
+static ssize_t ucma_write(struct file *filp, const char __user *buf,
+ size_t len, loff_t *pos)
+{
+ struct ucma_file *file = filp->private_data;
+ struct rdma_ucm_cmd_hdr hdr;
+ ssize_t ret;
+
+ if (len < sizeof(hdr))
+ return -EINVAL;
+
+ if (copy_from_user(&hdr, buf, sizeof(hdr)))
+ return -EFAULT;
+
+ if (hdr.cmd < 0 || hdr.cmd >= ARRAY_SIZE(ucma_cmd_table))
+ return -EINVAL;
+
+ if (hdr.in + sizeof(hdr) > len)
+ return -EINVAL;
+
+ ret = ucma_cmd_table[hdr.cmd](file, buf + sizeof(hdr), hdr.in, hdr.out);
+ if (!ret)
+ ret = len;
+
+ return ret;
+}
+
+static unsigned int ucma_poll(struct file *filp, struct poll_table_struct *wait)
+{
+ struct ucma_file *file = filp->private_data;
+ unsigned int mask = 0;
+
+ poll_wait(filp, &file->poll_wait, wait);
+
+ down(&file->mutex);
+ if (!list_empty(&file->events))
+ mask = POLLIN | POLLRDNORM;
+ up(&file->mutex);
+
+ return mask;
+}
+
+static int ucma_open(struct inode *inode, struct file *filp)
+{
+ struct ucma_file *file;
+
+ file = kmalloc(sizeof *file, GFP_KERNEL);
+ if (!file)
+ return -ENOMEM;
+
+ INIT_LIST_HEAD(&file->events);
+ INIT_LIST_HEAD(&file->ctxs);
+ init_waitqueue_head(&file->poll_wait);
+ init_MUTEX(&file->mutex);
+
+ filp->private_data = file;
+ file->filp = filp;
+ return 0;
+}
+
+static int ucma_close(struct inode *inode, struct file *filp)
+{
+ struct ucma_file *file = filp->private_data;
+ struct ucma_context *ctx;
+
+ down(&file->mutex);
+ while (!list_empty(&file->ctxs)) {
+ ctx = list_entry(file->ctxs.next, struct ucma_context,
+ file_list);
+ up(&file->mutex);
+
+ down(&ctx_mutex);
+ idr_remove(&ctx_idr, ctx->id);
+ up(&ctx_mutex);
+
+ rdma_destroy_id(ctx->cm_id);
+ ucma_cleanup_events(ctx);
+ kfree(ctx);
+
+ down(&file->mutex);
+ }
+ up(&file->mutex);
+ kfree(file);
+ return 0;
+}
+
+static struct file_operations ucma_fops = {
+ .owner = THIS_MODULE,
+ .open = ucma_open,
+ .release = ucma_close,
+ .write = ucma_write,
+ .poll = ucma_poll,
+};
+
+static struct miscdevice ucma_misc = {
+ .minor = MISC_DYNAMIC_MINOR,
+ .name = "rdma_cm",
+ .fops = &ucma_fops,
+};
+
+static int __init ucma_init(void)
+{
+ return misc_register(&ucma_misc);
+}
+
+static void __exit ucma_cleanup(void)
+{
+ misc_deregister(&ucma_misc);
+ idr_destroy(&ctx_idr);
+}
+
+module_init(ucma_init);
+module_exit(ucma_cleanup);
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/include/rdma/rdma_user_cm.h
linux-2.6.ib/include/rdma/rdma_user_cm.h
--- linux-2.6.git/include/rdma/rdma_user_cm.h 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.ib/include/rdma/rdma_user_cm.h 2006-01-16 16:54:55.000000000 -0800
@@ -0,0 +1,186 @@
+/*
+ * Copyright (c) 2005 Intel Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef RDMA_USER_CM_H
+#define RDMA_USER_CM_H
+
+#include <linux/types.h>
+#include <linux/in6.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_user_sa.h>
+
+#define RDMA_USER_CM_ABI_VERSION 1
+
+#define RDMA_MAX_PRIVATE_DATA 256
+
+enum {
+ RDMA_USER_CM_CMD_CREATE_ID,
+ RDMA_USER_CM_CMD_DESTROY_ID,
+ RDMA_USER_CM_CMD_BIND_ADDR,
+ RDMA_USER_CM_CMD_RESOLVE_ADDR,
+ RDMA_USER_CM_CMD_RESOLVE_ROUTE,
+ RDMA_USER_CM_CMD_QUERY_ROUTE,
+ RDMA_USER_CM_CMD_CONNECT,
+ RDMA_USER_CM_CMD_LISTEN,
+ RDMA_USER_CM_CMD_ACCEPT,
+ RDMA_USER_CM_CMD_REJECT,
+ RDMA_USER_CM_CMD_DISCONNECT,
+ RDMA_USER_CM_CMD_INIT_QP_ATTR,
+ RDMA_USER_CM_CMD_GET_EVENT
+};
+
+/*
+ * command ABI structures.
+ */
+struct rdma_ucm_cmd_hdr {
+ __u32 cmd;
+ __u16 in;
+ __u16 out;
+};
+
+struct rdma_ucm_create_id {
+ __u64 uid;
+ __u64 response;
+};
+
+struct rdma_ucm_create_id_resp {
+ __u32 id;
+};
+
+struct rdma_ucm_destroy_id {
+ __u64 response;
+ __u32 id;
+ __u32 reserved;
+};
+
+struct rdma_ucm_destroy_id_resp {
+ __u32 events_reported;
+};
+
+struct rdma_ucm_bind_addr {
+ __u64 response;
+ struct sockaddr_in6 addr;
+ __u32 id;
+};
+
+struct rdma_ucm_resolve_addr {
+ struct sockaddr_in6 src_addr;
+ struct sockaddr_in6 dst_addr;
+ __u32 id;
+ __u32 timeout_ms;
+};
+
+struct rdma_ucm_resolve_route {
+ __u32 id;
+ __u32 timeout_ms;
+};
+
+struct rdma_ucm_query_route {
+ __u64 response;
+ __u32 id;
+ __u32 reserved;
+};
+
+struct rdma_ucm_query_route_resp {
+ __u64 node_guid;
+ struct ib_user_path_rec ib_route[2];
+ struct sockaddr_in6 src_addr;
+ struct sockaddr_in6 dst_addr;
+ __u32 num_paths;
+ __u8 port_num;
+ __u8 reserved[3];
+};
+
+struct rdma_ucm_conn_param {
+ __u32 qp_num;
+ __u32 qp_type;
+ __u8 private_data[RDMA_MAX_PRIVATE_DATA];
+ __u8 private_data_len;
+ __u8 srq;
+ __u8 responder_resources;
+ __u8 initiator_depth;
+ __u8 flow_control;
+ __u8 retry_count;
+ __u8 rnr_retry_count;
+ __u8 valid;
+};
+
+struct rdma_ucm_connect {
+ struct rdma_ucm_conn_param conn_param;
+ __u32 id;
+ __u32 reserved;
+};
+
+struct rdma_ucm_listen {
+ __u32 id;
+ __u32 backlog;
+};
+
+struct rdma_ucm_accept {
+ __u64 uid;
+ struct rdma_ucm_conn_param conn_param;
+ __u32 id;
+ __u32 reserved;
+};
+
+struct rdma_ucm_reject {
+ __u32 id;
+ __u8 private_data_len;
+ __u8 reserved[3];
+ __u8 private_data[RDMA_MAX_PRIVATE_DATA];
+};
+
+struct rdma_ucm_disconnect {
+ __u32 id;
+};
+
+struct rdma_ucm_init_qp_attr {
+ __u64 response;
+ __u32 id;
+ __u32 qp_state;
+};
+
+struct rdma_ucm_get_event {
+ __u64 response;
+};
+
+struct rdma_ucm_event_resp {
+ __u64 uid;
+ __u32 id;
+ __u32 event;
+ __u32 status;
+ __u8 private_data_len;
+ __u8 reserved[3];
+ __u8 private_data[RDMA_MAX_PRIVATE_DATA];
+};
+
+#endif /* RDMA_USER_CM_H */
^ permalink raw reply
* RE: [PATCH 2/6] IB: match connection requests based on private data
From: Caitlin Bestler @ 2006-03-06 19:20 UTC (permalink / raw)
To: Sean Hefty, Roland Dreier; +Cc: netdev, linux-kernel, openib-general
> -----Original Message-----
> From: netdev-owner@vger.kernel.org
> [mailto:netdev-owner@vger.kernel.org] On Behalf Of Sean Hefty
> Sent: Monday, March 06, 2006 11:04 AM
> To: 'Roland Dreier'
> Cc: netdev@vger.kernel.org; linux-kernel@vger.kernel.org;
> openib-general@openib.org
> Subject: [PATCH 2/6] IB: match connection requests based on
> private data
>
> Extend matching connection requests to listens in the
> Infiniband CM to include private data checks.
>
> This allows applications to listen on the same service
> identifier, with private data directing the request to the
> appropriate application.
>
> Signed-off-by: Sean Hefty <sean.hefty@intel.com>
>
The term "private data" is intended to convey the
intent that the data is private to the application
layer and is opaque to middleware and the network.
By what mechanism does the listening application
delegate how much of the private data for use by
the CM for sub-dividing a listen? What does an
application do if it wishes to retain full ownership
of the private data?
^ permalink raw reply
* [PATCH 5/6] IB: IP address based RDMA connection manager
From: Sean Hefty @ 2006-03-06 19:18 UTC (permalink / raw)
To: 'Roland Dreier', netdev, linux-kernel; +Cc: openib-general
In-Reply-To: <adaslpz2l9p.fsf@cisco.com>
Kernel mode connection management agent over Infiniband that connects based
on IP addresses. The agent defines a generic RDMA connection abstraction
to support clients wanting to connect over different RDMA devices.
Agent also handles RDMA device hotplug events on behalf of clients.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
---
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/drivers/infiniband/core/cma.c
linux-2.6.ib/drivers/infiniband/core/cma.c
--- linux-2.6.git/drivers/infiniband/core/cma.c 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.ib/drivers/infiniband/core/cma.c 2006-01-16 16:17:34.000000000 -0800
@@ -0,0 +1,1639 @@
+/*
+ * Copyright (c) 2005 Voltaire Inc. All rights reserved.
+ * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved.
+ * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved.
+ * Copyright (c) 2005 Intel Corporation. All rights reserved.
+ *
+ * This Software is licensed under one of the following licenses:
+ *
+ * 1) under the terms of the "Common Public License 1.0" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/cpl.php.
+ *
+ * 2) under the terms of the "The BSD License" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/bsd-license.php.
+ *
+ * 3) under the terms of the "GNU General Public License (GPL) Version 2" a
+ * copy of which is available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/gpl-license.php.
+ *
+ * Licensee has the right to choose one of the above licenses.
+ *
+ * Redistributions of source code must retain the above copyright
+ * notice and one of the license notices.
+ *
+ * Redistributions in binary form must reproduce both the above copyright
+ * notice, one of the license notices in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ */
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/random.h>
+#include <rdma/rdma_cm.h>
+#include <rdma/ib_cache.h>
+#include <rdma/ib_cm.h>
+#include <rdma/ib_sa.h>
+
+MODULE_AUTHOR("Guy German");
+MODULE_DESCRIPTION("Generic RDMA CM Agent");
+MODULE_LICENSE("Dual BSD/GPL");
+
+#define CMA_CM_RESPONSE_TIMEOUT 20
+#define CMA_MAX_CM_RETRIES 3
+
+static void cma_add_one(struct ib_device *device);
+static void cma_remove_one(struct ib_device *device);
+
+static struct ib_client cma_client = {
+ .name = "cma",
+ .add = cma_add_one,
+ .remove = cma_remove_one
+};
+
+static LIST_HEAD(dev_list);
+static LIST_HEAD(listen_any_list);
+static DECLARE_MUTEX(mutex);
+
+struct cma_device {
+ struct list_head list;
+ struct ib_device *device;
+ __be64 node_guid;
+ wait_queue_head_t wait;
+ atomic_t refcount;
+ struct list_head id_list;
+};
+
+enum cma_state {
+ CMA_IDLE,
+ CMA_ADDR_QUERY,
+ CMA_ADDR_RESOLVED,
+ CMA_ROUTE_QUERY,
+ CMA_ROUTE_RESOLVED,
+ CMA_CONNECT,
+ CMA_ADDR_BOUND,
+ CMA_LISTEN,
+ CMA_DEVICE_REMOVAL,
+ CMA_DESTROYING
+};
+
+/*
+ * Device removal can occur at anytime, so we need extra handling to
+ * serialize notifying the user of device removal with other callbacks.
+ * We do this by disabling removal notification while a callback is in process,
+ * and reporting it after the callback completes.
+ */
+struct rdma_id_private {
+ struct rdma_cm_id id;
+
+ struct list_head list;
+ struct list_head listen_list;
+ struct cma_device *cma_dev;
+
+ enum cma_state state;
+ spinlock_t lock;
+ wait_queue_head_t wait;
+ atomic_t refcount;
+ wait_queue_head_t wait_remove;
+ atomic_t dev_remove;
+
+ int backlog;
+ int timeout_ms;
+ struct ib_sa_query *query;
+ int query_id;
+ struct ib_cm_id *cm_id;
+
+ u32 seq_num;
+ u32 qp_num;
+ enum ib_qp_type qp_type;
+ u8 srq;
+};
+
+struct cma_work {
+ struct work_struct work;
+ struct rdma_id_private *id;
+};
+
+union cma_ip_addr {
+ struct in6_addr ip6;
+ struct {
+ __u32 pad[3];
+ __u32 addr;
+ } ip4;
+};
+
+struct cma_hdr {
+ u8 cma_version;
+ u8 ip_version; /* IP version: 7:4 */
+ __u16 port;
+ union cma_ip_addr src_addr;
+ union cma_ip_addr dst_addr;
+};
+
+struct sdp_hh {
+ u8 sdp_version;
+ u8 ip_version; /* IP version: 7:4 */
+ u8 sdp_specific1[10];
+ __u16 port;
+ __u16 sdp_specific2;
+ union cma_ip_addr src_addr;
+ union cma_ip_addr dst_addr;
+};
+
+#define CMA_VERSION 0x10
+#define SDP_VERSION 0x22
+
+static int cma_comp(struct rdma_id_private *id_priv, enum cma_state comp)
+{
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&id_priv->lock, flags);
+ ret = (id_priv->state == comp);
+ spin_unlock_irqrestore(&id_priv->lock, flags);
+ return ret;
+}
+
+static int cma_comp_exch(struct rdma_id_private *id_priv,
+ enum cma_state comp, enum cma_state exch)
+{
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&id_priv->lock, flags);
+ if ((ret = (id_priv->state == comp)))
+ id_priv->state = exch;
+ spin_unlock_irqrestore(&id_priv->lock, flags);
+ return ret;
+}
+
+static enum cma_state cma_exch(struct rdma_id_private *id_priv,
+ enum cma_state exch)
+{
+ unsigned long flags;
+ enum cma_state old;
+
+ spin_lock_irqsave(&id_priv->lock, flags);
+ old = id_priv->state;
+ id_priv->state = exch;
+ spin_unlock_irqrestore(&id_priv->lock, flags);
+ return old;
+}
+
+static inline u8 cma_get_ip_ver(struct cma_hdr *hdr)
+{
+ return hdr->ip_version >> 4;
+}
+
+static inline void cma_set_ip_ver(struct cma_hdr *hdr, u8 ip_ver)
+{
+ hdr->ip_version = (ip_ver << 4) | (hdr->ip_version & 0xF);
+}
+
+static inline u8 sdp_get_ip_ver(struct sdp_hh *hh)
+{
+ return hh->ip_version >> 4;
+}
+
+static inline void sdp_set_ip_ver(struct sdp_hh *hh, u8 ip_ver)
+{
+ hh->ip_version = (ip_ver << 4) | (hh->ip_version & 0xF);
+}
+
+static void cma_attach_to_dev(struct rdma_id_private *id_priv,
+ struct cma_device *cma_dev)
+{
+ atomic_inc(&cma_dev->refcount);
+ id_priv->cma_dev = cma_dev;
+ id_priv->id.device = cma_dev->device;
+ list_add_tail(&id_priv->list, &cma_dev->id_list);
+}
+
+static void cma_detach_from_dev(struct rdma_id_private *id_priv)
+{
+ list_del(&id_priv->list);
+ if (atomic_dec_and_test(&id_priv->cma_dev->refcount))
+ wake_up(&id_priv->cma_dev->wait);
+ id_priv->cma_dev = NULL;
+}
+
+static int cma_acquire_ib_dev(struct rdma_id_private *id_priv)
+{
+ struct cma_device *cma_dev;
+ union ib_gid *gid;
+ int ret = -ENODEV;
+
+ gid = ib_addr_get_sgid(&id_priv->id.route.addr.dev_addr);
+
+ down(&mutex);
+ list_for_each_entry(cma_dev, &dev_list, list) {
+ ret = ib_find_cached_gid(cma_dev->device, gid,
+ &id_priv->id.port_num, NULL);
+ if (!ret) {
+ cma_attach_to_dev(id_priv, cma_dev);
+ break;
+ }
+ }
+ up(&mutex);
+ return ret;
+}
+
+static int cma_acquire_dev(struct rdma_id_private *id_priv)
+{
+ switch (id_priv->id.route.addr.dev_addr.dev_type) {
+ case IB_NODE_CA:
+ return cma_acquire_ib_dev(id_priv);
+ default:
+ return -ENODEV;
+ }
+}
+
+static void cma_deref_id(struct rdma_id_private *id_priv)
+{
+ if (atomic_dec_and_test(&id_priv->refcount))
+ wake_up(&id_priv->wait);
+}
+
+static void cma_release_remove(struct rdma_id_private *id_priv)
+{
+ if (atomic_dec_and_test(&id_priv->dev_remove))
+ wake_up(&id_priv->wait_remove);
+}
+
+struct rdma_cm_id* rdma_create_id(rdma_cm_event_handler event_handler,
+ void *context, enum rdma_port_space ps)
+{
+ struct rdma_id_private *id_priv;
+
+ id_priv = kzalloc(sizeof *id_priv, GFP_KERNEL);
+ if (!id_priv)
+ return ERR_PTR(-ENOMEM);
+
+ id_priv->state = CMA_IDLE;
+ id_priv->id.context = context;
+ id_priv->id.event_handler = event_handler;
+ id_priv->id.ps = ps;
+ spin_lock_init(&id_priv->lock);
+ init_waitqueue_head(&id_priv->wait);
+ atomic_set(&id_priv->refcount, 1);
+ init_waitqueue_head(&id_priv->wait_remove);
+ atomic_set(&id_priv->dev_remove, 0);
+ INIT_LIST_HEAD(&id_priv->listen_list);
+ get_random_bytes(&id_priv->seq_num, sizeof id_priv->seq_num);
+
+ return &id_priv->id;
+}
+EXPORT_SYMBOL(rdma_create_id);
+
+static int cma_init_ib_qp(struct rdma_id_private *id_priv, struct ib_qp *qp)
+{
+ struct ib_qp_attr qp_attr;
+ struct rdma_dev_addr *dev_addr;
+ int ret;
+
+ dev_addr = &id_priv->id.route.addr.dev_addr;
+ ret = ib_find_cached_pkey(id_priv->id.device, id_priv->id.port_num,
+ ib_addr_get_pkey(dev_addr),
+ &qp_attr.pkey_index);
+ if (ret)
+ return ret;
+
+ qp_attr.qp_state = IB_QPS_INIT;
+ qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE;
+ qp_attr.port_num = id_priv->id.port_num;
+ return ib_modify_qp(qp, &qp_attr, IB_QP_STATE | IB_QP_ACCESS_FLAGS |
+ IB_QP_PKEY_INDEX | IB_QP_PORT);
+}
+
+int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd,
+ struct ib_qp_init_attr *qp_init_attr)
+{
+ struct rdma_id_private *id_priv;
+ struct ib_qp *qp;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (id->device != pd->device)
+ return -EINVAL;
+
+ qp = ib_create_qp(pd, qp_init_attr);
+ if (IS_ERR(qp))
+ return PTR_ERR(qp);
+
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ ret = cma_init_ib_qp(id_priv, qp);
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+
+ if (ret)
+ goto err;
+
+ id->qp = qp;
+ id_priv->qp_num = qp->qp_num;
+ id_priv->qp_type = qp->qp_type;
+ id_priv->srq = (qp->srq != NULL);
+ return 0;
+err:
+ ib_destroy_qp(qp);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_create_qp);
+
+void rdma_destroy_qp(struct rdma_cm_id *id)
+{
+ ib_destroy_qp(id->qp);
+}
+EXPORT_SYMBOL(rdma_destroy_qp);
+
+static int cma_modify_qp_rtr(struct rdma_cm_id *id)
+{
+ struct ib_qp_attr qp_attr;
+ int qp_attr_mask, ret;
+
+ if (!id->qp)
+ return 0;
+
+ /* Need to update QP attributes from default values. */
+ qp_attr.qp_state = IB_QPS_INIT;
+ ret = rdma_init_qp_attr(id, &qp_attr, &qp_attr_mask);
+ if (ret)
+ return ret;
+
+ ret = ib_modify_qp(id->qp, &qp_attr, qp_attr_mask);
+ if (ret)
+ return ret;
+
+ qp_attr.qp_state = IB_QPS_RTR;
+ ret = rdma_init_qp_attr(id, &qp_attr, &qp_attr_mask);
+ if (ret)
+ return ret;
+
+ return ib_modify_qp(id->qp, &qp_attr, qp_attr_mask);
+}
+
+static int cma_modify_qp_rts(struct rdma_cm_id *id)
+{
+ struct ib_qp_attr qp_attr;
+ int qp_attr_mask, ret;
+
+ if (!id->qp)
+ return 0;
+
+ qp_attr.qp_state = IB_QPS_RTS;
+ ret = rdma_init_qp_attr(id, &qp_attr, &qp_attr_mask);
+ if (ret)
+ return ret;
+
+ return ib_modify_qp(id->qp, &qp_attr, qp_attr_mask);
+}
+
+static int cma_modify_qp_err(struct rdma_cm_id *id)
+{
+ struct ib_qp_attr qp_attr;
+
+ if (!id->qp)
+ return 0;
+
+ qp_attr.qp_state = IB_QPS_ERR;
+ return ib_modify_qp(id->qp, &qp_attr, IB_QP_STATE);
+}
+
+int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr,
+ int *qp_attr_mask)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ switch (id_priv->id.device->node_type) {
+ case IB_NODE_CA:
+ ret = ib_cm_init_qp_attr(id_priv->cm_id, qp_attr,
+ qp_attr_mask);
+ if (qp_attr->qp_state == IB_QPS_RTR)
+ qp_attr->rq_psn = id_priv->seq_num;
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+
+ return ret;
+}
+EXPORT_SYMBOL(rdma_init_qp_attr);
+
+static inline int cma_any_addr(struct sockaddr *addr)
+{
+ struct in6_addr *ip6;
+
+ if (addr->sa_family == AF_INET)
+ return ((struct sockaddr_in *) addr)->sin_addr.s_addr ==
+ INADDR_ANY;
+ else {
+ ip6 = &((struct sockaddr_in6 *) addr)->sin6_addr;
+ return (ip6->s6_addr32[0] | ip6->s6_addr32[1] |
+ ip6->s6_addr32[3] | ip6->s6_addr32[4]) == 0;
+ }
+}
+
+static inline int cma_loopback_addr(struct sockaddr *addr)
+{
+ return ((struct sockaddr_in *) addr)->sin_addr.s_addr ==
+ ntohl(INADDR_LOOPBACK);
+}
+
+static int cma_get_net_info(void *hdr, enum rdma_port_space ps,
+ u8 *ip_ver, __u16 *port,
+ union cma_ip_addr **src, union cma_ip_addr **dst)
+{
+ switch (ps) {
+ case RDMA_PS_SDP:
+ if (((struct sdp_hh *) hdr)->sdp_version != SDP_VERSION)
+ return -EINVAL;
+
+ *ip_ver = sdp_get_ip_ver(hdr);
+ *port = ((struct sdp_hh *) hdr)->port;
+ *src = &((struct sdp_hh *) hdr)->src_addr;
+ *dst = &((struct sdp_hh *) hdr)->dst_addr;
+ break;
+ default:
+ if (((struct cma_hdr *) hdr)->cma_version != CMA_VERSION)
+ return -EINVAL;
+
+ *ip_ver = cma_get_ip_ver(hdr);
+ *port = ((struct cma_hdr *) hdr)->port;
+ *src = &((struct cma_hdr *) hdr)->src_addr;
+ *dst = &((struct cma_hdr *) hdr)->dst_addr;
+ break;
+ }
+ return 0;
+}
+
+static void cma_save_net_info(struct rdma_addr *addr,
+ struct rdma_addr *listen_addr,
+ u8 ip_ver, __u16 port,
+ union cma_ip_addr *src, union cma_ip_addr *dst)
+{
+ struct sockaddr_in *listen4, *ip4;
+ struct sockaddr_in6 *listen6, *ip6;
+
+ switch (ip_ver) {
+ case 4:
+ listen4 = (struct sockaddr_in *) &listen_addr->src_addr;
+ ip4 = (struct sockaddr_in *) &addr->src_addr;
+ ip4->sin_family = listen4->sin_family;
+ ip4->sin_addr.s_addr = dst->ip4.addr;
+ ip4->sin_port = listen4->sin_port;
+
+ ip4 = (struct sockaddr_in *) &addr->dst_addr;
+ ip4->sin_family = listen4->sin_family;
+ ip4->sin_addr.s_addr = src->ip4.addr;
+ ip4->sin_port = port;
+ break;
+ case 6:
+ listen6 = (struct sockaddr_in6 *) &listen_addr->src_addr;
+ ip6 = (struct sockaddr_in6 *) &addr->src_addr;
+ ip6->sin6_family = listen6->sin6_family;
+ ip6->sin6_addr = dst->ip6;
+ ip6->sin6_port = listen6->sin6_port;
+
+ ip6 = (struct sockaddr_in6 *) &addr->dst_addr;
+ ip6->sin6_family = listen6->sin6_family;
+ ip6->sin6_addr = src->ip6;
+ ip6->sin6_port = port;
+ break;
+ default:
+ break;
+ }
+}
+
+static inline int cma_user_data_offset(enum rdma_port_space ps)
+{
+ switch (ps) {
+ case RDMA_PS_SDP:
+ return 0;
+ default:
+ return sizeof(struct cma_hdr);
+ }
+}
+
+static int cma_notify_user(struct rdma_id_private *id_priv,
+ enum rdma_cm_event_type type, int status,
+ void *data, u8 data_len)
+{
+ struct rdma_cm_event event;
+
+ event.event = type;
+ event.status = status;
+ event.private_data = data;
+ event.private_data_len = data_len;
+
+ return id_priv->id.event_handler(&id_priv->id, &event);
+}
+
+static void cma_cancel_addr(struct rdma_id_private *id_priv)
+{
+ switch (id_priv->id.device->node_type) {
+ case IB_NODE_CA:
+ rdma_addr_cancel(&id_priv->id.route.addr.dev_addr);
+ break;
+ default:
+ break;
+ }
+}
+
+static void cma_cancel_route(struct rdma_id_private *id_priv)
+{
+ switch (id_priv->id.device->node_type) {
+ case IB_NODE_CA:
+ ib_sa_cancel_query(id_priv->query_id, id_priv->query);
+ break;
+ default:
+ break;
+ }
+}
+
+static inline int cma_internal_listen(struct rdma_id_private *id_priv)
+{
+ return (id_priv->state == CMA_LISTEN) && id_priv->cma_dev &&
+ cma_any_addr(&id_priv->id.route.addr.src_addr);
+}
+
+static void cma_destroy_listen(struct rdma_id_private *id_priv)
+{
+ cma_exch(id_priv, CMA_DESTROYING);
+
+ if (id_priv->cm_id && !IS_ERR(id_priv->cm_id))
+ ib_destroy_cm_id(id_priv->cm_id);
+
+ list_del(&id_priv->listen_list);
+ if (id_priv->cma_dev)
+ cma_detach_from_dev(id_priv);
+
+ atomic_dec(&id_priv->refcount);
+ wait_event(id_priv->wait, !atomic_read(&id_priv->refcount));
+
+ kfree(id_priv);
+}
+
+static void cma_cancel_listens(struct rdma_id_private *id_priv)
+{
+ struct rdma_id_private *dev_id_priv;
+
+ down(&mutex);
+ list_del(&id_priv->list);
+
+ while (!list_empty(&id_priv->listen_list)) {
+ dev_id_priv = list_entry(id_priv->listen_list.next,
+ struct rdma_id_private, listen_list);
+ cma_destroy_listen(dev_id_priv);
+ }
+ up(&mutex);
+}
+
+static void cma_cancel_operation(struct rdma_id_private *id_priv,
+ enum cma_state state)
+{
+ switch (state) {
+ case CMA_ADDR_QUERY:
+ cma_cancel_addr(id_priv);
+ break;
+ case CMA_ROUTE_QUERY:
+ cma_cancel_route(id_priv);
+ break;
+ case CMA_LISTEN:
+ if (cma_any_addr(&id_priv->id.route.addr.src_addr) &&
+ !id_priv->cma_dev)
+ cma_cancel_listens(id_priv);
+ break;
+ default:
+ break;
+ }
+}
+
+void rdma_destroy_id(struct rdma_cm_id *id)
+{
+ struct rdma_id_private *id_priv;
+ enum cma_state state;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ state = cma_exch(id_priv, CMA_DESTROYING);
+ cma_cancel_operation(id_priv, state);
+
+ if (id_priv->cm_id && !IS_ERR(id_priv->cm_id))
+ ib_destroy_cm_id(id_priv->cm_id);
+
+ if (id_priv->cma_dev) {
+ down(&mutex);
+ cma_detach_from_dev(id_priv);
+ up(&mutex);
+ }
+
+ atomic_dec(&id_priv->refcount);
+ wait_event(id_priv->wait, !atomic_read(&id_priv->refcount));
+
+ kfree(id_priv->id.route.path_rec);
+ kfree(id_priv);
+}
+EXPORT_SYMBOL(rdma_destroy_id);
+
+static int cma_rep_recv(struct rdma_id_private *id_priv)
+{
+ int ret;
+
+ ret = cma_modify_qp_rtr(&id_priv->id);
+ if (ret)
+ goto reject;
+
+ ret = cma_modify_qp_rts(&id_priv->id);
+ if (ret)
+ goto reject;
+
+ ret = ib_send_cm_rtu(id_priv->cm_id, NULL, 0);
+ if (ret)
+ goto reject;
+
+ return 0;
+reject:
+ cma_modify_qp_err(&id_priv->id);
+ ib_send_cm_rej(id_priv->cm_id, IB_CM_REJ_CONSUMER_DEFINED,
+ NULL, 0, NULL, 0);
+ return ret;
+}
+
+static int cma_rtu_recv(struct rdma_id_private *id_priv)
+{
+ int ret;
+
+ ret = cma_modify_qp_rts(&id_priv->id);
+ if (ret)
+ goto reject;
+
+ return 0;
+reject:
+ cma_modify_qp_err(&id_priv->id);
+ ib_send_cm_rej(id_priv->cm_id, IB_CM_REJ_CONSUMER_DEFINED,
+ NULL, 0, NULL, 0);
+ return ret;
+}
+
+static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
+{
+ struct rdma_id_private *id_priv = cm_id->context;
+ enum rdma_cm_event_type event;
+ u8 private_data_len = 0;
+ int ret = 0, status = 0;
+
+ if (!cma_comp(id_priv, CMA_CONNECT))
+ return 0;
+
+ atomic_inc(&id_priv->dev_remove);
+ switch (ib_event->event) {
+ case IB_CM_REQ_ERROR:
+ case IB_CM_REP_ERROR:
+ event = RDMA_CM_EVENT_UNREACHABLE;
+ status = -ETIMEDOUT;
+ break;
+ case IB_CM_REP_RECEIVED:
+ if (id_priv->id.qp) {
+ status = cma_rep_recv(id_priv);
+ event = status ? RDMA_CM_EVENT_CONNECT_ERROR :
+ RDMA_CM_EVENT_ESTABLISHED;
+ } else
+ event = RDMA_CM_EVENT_CONNECT_RESPONSE;
+ private_data_len = IB_CM_REP_PRIVATE_DATA_SIZE;
+ break;
+ case IB_CM_RTU_RECEIVED:
+ status = cma_rtu_recv(id_priv);
+ event = status ? RDMA_CM_EVENT_CONNECT_ERROR :
+ RDMA_CM_EVENT_ESTABLISHED;
+ break;
+ case IB_CM_DREQ_ERROR:
+ status = -ETIMEDOUT; /* fall through */
+ case IB_CM_DREQ_RECEIVED:
+ case IB_CM_DREP_RECEIVED:
+ event = RDMA_CM_EVENT_DISCONNECTED;
+ break;
+ case IB_CM_TIMEWAIT_EXIT:
+ case IB_CM_MRA_RECEIVED:
+ /* ignore event */
+ goto out;
+ case IB_CM_REJ_RECEIVED:
+ cma_modify_qp_err(&id_priv->id);
+ status = ib_event->param.rej_rcvd.reason;
+ event = RDMA_CM_EVENT_REJECTED;
+ break;
+ default:
+ printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d",
+ ib_event->event);
+ goto out;
+ }
+
+ ret = cma_notify_user(id_priv, event, status, ib_event->private_data,
+ private_data_len);
+ if (ret) {
+ /* Destroy the CM ID by returning a non-zero value. */
+ id_priv->cm_id = NULL;
+ cma_exch(id_priv, CMA_DESTROYING);
+ cma_release_remove(id_priv);
+ rdma_destroy_id(&id_priv->id);
+ return ret;
+ }
+out:
+ cma_release_remove(id_priv);
+ return ret;
+}
+
+static struct rdma_id_private* cma_new_id(struct rdma_cm_id *listen_id,
+ struct ib_cm_event *ib_event)
+{
+ struct rdma_id_private *id_priv;
+ struct rdma_cm_id *id;
+ struct rdma_route *rt;
+ union cma_ip_addr *src, *dst;
+ __u16 port;
+ u8 ip_ver;
+
+ id = rdma_create_id(listen_id->event_handler, listen_id->context,
+ listen_id->ps);
+ if (IS_ERR(id))
+ return NULL;
+
+ rt = &id->route;
+ rt->num_paths = ib_event->param.req_rcvd.alternate_path ? 2 : 1;
+ rt->path_rec = kmalloc(sizeof *rt->path_rec * rt->num_paths, GFP_KERNEL);
+ if (!rt->path_rec)
+ goto err;
+
+ if (cma_get_net_info(ib_event->private_data, listen_id->ps,
+ &ip_ver, &port, &src, &dst))
+ goto err;
+
+ cma_save_net_info(&id->route.addr, &listen_id->route.addr,
+ ip_ver, port, src, dst);
+ rt->path_rec[0] = *ib_event->param.req_rcvd.primary_path;
+ if (rt->num_paths == 2)
+ rt->path_rec[1] = *ib_event->param.req_rcvd.alternate_path;
+
+ ib_addr_set_sgid(&rt->addr.dev_addr, &rt->path_rec[0].sgid);
+ ib_addr_set_dgid(&rt->addr.dev_addr, &rt->path_rec[0].dgid);
+ ib_addr_set_pkey(&rt->addr.dev_addr, be16_to_cpu(rt->path_rec[0].pkey));
+ rt->addr.dev_addr.dev_type = IB_NODE_CA;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ id_priv->state = CMA_CONNECT;
+ return id_priv;
+err:
+ rdma_destroy_id(id);
+ return NULL;
+}
+
+static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
+{
+ struct rdma_id_private *listen_id, *conn_id;
+ int offset, ret;
+
+ listen_id = cm_id->context;
+ atomic_inc(&listen_id->dev_remove);
+ if (!cma_comp(listen_id, CMA_LISTEN)) {
+ ret = -ECONNABORTED;
+ goto out;
+ }
+
+ conn_id = cma_new_id(&listen_id->id, ib_event);
+ if (!conn_id) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ atomic_inc(&conn_id->dev_remove);
+ ret = cma_acquire_ib_dev(conn_id);
+ if (ret) {
+ ret = -ENODEV;
+ cma_release_remove(conn_id);
+ rdma_destroy_id(&conn_id->id);
+ goto out;
+ }
+
+ conn_id->cm_id = cm_id;
+ cm_id->context = conn_id;
+ cm_id->cm_handler = cma_ib_handler;
+
+ offset = cma_user_data_offset(listen_id->id.ps);
+ ret = cma_notify_user(conn_id, RDMA_CM_EVENT_CONNECT_REQUEST, 0,
+ ib_event->private_data + offset,
+ IB_CM_REQ_PRIVATE_DATA_SIZE - offset);
+ if (ret) {
+ /* Destroy the CM ID by returning a non-zero value. */
+ conn_id->cm_id = NULL;
+ cma_exch(conn_id, CMA_DESTROYING);
+ cma_release_remove(conn_id);
+ rdma_destroy_id(&conn_id->id);
+ }
+out:
+ cma_release_remove(listen_id);
+ return ret;
+}
+
+static __be64 cma_get_service_id(enum rdma_port_space ps, struct sockaddr *addr)
+{
+ return cpu_to_be64(((u64)ps << 16) +
+ ((struct sockaddr_in *) addr)->sin_port);
+}
+
+static void cma_set_compare_data(struct sockaddr *addr,
+ struct ib_cm_private_data_compare *compare)
+{
+ struct cma_hdr *data, *mask;
+
+ memset(compare, 0, sizeof *compare);
+ data = (void *) compare->data;
+ mask = (void *) compare->mask;
+
+ switch (addr->sa_family) {
+ case AF_INET:
+ cma_set_ip_ver(data, 4);
+ cma_set_ip_ver(mask, 0xF);
+ data->dst_addr.ip4.addr = ((struct sockaddr_in *) addr)->
+ sin_addr.s_addr;
+ mask->dst_addr.ip4.addr = ~0;
+ break;
+ case AF_INET6:
+ cma_set_ip_ver(data, 6);
+ cma_set_ip_ver(mask, 0xF);
+ data->dst_addr.ip6 = ((struct sockaddr_in6 *) addr)->
+ sin6_addr;
+ memset(&mask->dst_addr.ip6, 1, sizeof mask->dst_addr.ip6);
+ break;
+ default:
+ break;
+ }
+}
+
+static int cma_ib_listen(struct rdma_id_private *id_priv)
+{
+ struct ib_cm_private_data_compare compare_data;
+ struct sockaddr *addr;
+ __be64 svc_id;
+ int ret;
+
+ id_priv->cm_id = ib_create_cm_id(id_priv->id.device, cma_req_handler,
+ id_priv);
+ if (IS_ERR(id_priv->cm_id))
+ return PTR_ERR(id_priv->cm_id);
+
+ addr = &id_priv->id.route.addr.src_addr;
+ svc_id = cma_get_service_id(id_priv->id.ps, addr);
+ if (cma_any_addr(addr))
+ ret = ib_cm_listen(id_priv->cm_id, svc_id, 0, NULL);
+ else {
+ cma_set_compare_data(addr, &compare_data);
+ ret = ib_cm_listen(id_priv->cm_id, svc_id, 0, &compare_data);
+ }
+
+ if (ret) {
+ ib_destroy_cm_id(id_priv->cm_id);
+ id_priv->cm_id = NULL;
+ }
+
+ return ret;
+}
+
+static int cma_duplicate_listen(struct rdma_id_private *id_priv)
+{
+ struct rdma_id_private *cur_id_priv;
+ struct sockaddr_in *cur_addr, *new_addr;
+
+ new_addr = (struct sockaddr_in *) &id_priv->id.route.addr.src_addr;
+ list_for_each_entry(cur_id_priv, &listen_any_list, listen_list) {
+ cur_addr = (struct sockaddr_in *)
+ &cur_id_priv->id.route.addr.src_addr;
+ if (cur_addr->sin_port == new_addr->sin_port)
+ return -EADDRINUSE;
+ }
+ return 0;
+}
+
+static int cma_listen_handler(struct rdma_cm_id *id,
+ struct rdma_cm_event *event)
+{
+ struct rdma_id_private *id_priv = id->context;
+
+ id->context = id_priv->id.context;
+ id->event_handler = id_priv->id.event_handler;
+ return id_priv->id.event_handler(id, event);
+}
+
+static void cma_listen_on_dev(struct rdma_id_private *id_priv,
+ struct cma_device *cma_dev)
+{
+ struct rdma_id_private *dev_id_priv;
+ struct rdma_cm_id *id;
+ int ret;
+
+ id = rdma_create_id(cma_listen_handler, id_priv, id_priv->id.ps);
+ if (IS_ERR(id))
+ return;
+
+ dev_id_priv = container_of(id, struct rdma_id_private, id);
+ ret = rdma_bind_addr(id, &id_priv->id.route.addr.src_addr);
+ if (ret)
+ goto err;
+
+ cma_attach_to_dev(dev_id_priv, cma_dev);
+ list_add_tail(&dev_id_priv->listen_list, &id_priv->listen_list);
+
+ ret = rdma_listen(id, id_priv->backlog);
+ if (ret)
+ goto err;
+
+ return;
+err:
+ cma_destroy_listen(dev_id_priv);
+}
+
+static int cma_listen_on_all(struct rdma_id_private *id_priv)
+{
+ struct cma_device *cma_dev;
+ int ret;
+
+ down(&mutex);
+ ret = cma_duplicate_listen(id_priv);
+ if (ret)
+ goto out;
+
+ list_add_tail(&id_priv->list, &listen_any_list);
+ list_for_each_entry(cma_dev, &dev_list, list)
+ cma_listen_on_dev(id_priv, cma_dev);
+out:
+ up(&mutex);
+ return ret;
+}
+
+int rdma_listen(struct rdma_cm_id *id, int backlog)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp_exch(id_priv, CMA_ADDR_BOUND, CMA_LISTEN))
+ return -EINVAL;
+
+ if (id->device) {
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ ret = cma_ib_listen(id_priv);
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+ } else
+ ret = cma_listen_on_all(id_priv);
+
+ if (ret)
+ goto err;
+
+ id_priv->backlog = backlog;
+ return 0;
+err:
+ cma_comp_exch(id_priv, CMA_LISTEN, CMA_ADDR_BOUND);
+ return ret;
+};
+EXPORT_SYMBOL(rdma_listen);
+
+static void cma_query_handler(int status, struct ib_sa_path_rec *path_rec,
+ void *context)
+{
+ struct rdma_id_private *id_priv = context;
+ struct rdma_route *route = &id_priv->id.route;
+ enum rdma_cm_event_type event = RDMA_CM_EVENT_ROUTE_RESOLVED;
+
+ atomic_inc(&id_priv->dev_remove);
+ if (!status) {
+ route->path_rec = kmalloc(sizeof *route->path_rec, GFP_KERNEL);
+ if (route->path_rec) {
+ route->num_paths = 1;
+ *route->path_rec = *path_rec;
+ if (!cma_comp_exch(id_priv, CMA_ROUTE_QUERY,
+ CMA_ROUTE_RESOLVED)) {
+ kfree(route->path_rec);
+ goto out;
+ }
+ } else
+ status = -ENOMEM;
+ }
+
+ if (status) {
+ if (!cma_comp_exch(id_priv, CMA_ROUTE_QUERY, CMA_ADDR_RESOLVED))
+ goto out;
+ event = RDMA_CM_EVENT_ROUTE_ERROR;
+ }
+
+ if (cma_notify_user(id_priv, event, status, NULL, 0)) {
+ cma_exch(id_priv, CMA_DESTROYING);
+ cma_release_remove(id_priv);
+ cma_deref_id(id_priv);
+ rdma_destroy_id(&id_priv->id);
+ return;
+ }
+out:
+ cma_release_remove(id_priv);
+ cma_deref_id(id_priv);
+}
+
+static int cma_resolve_ib_route(struct rdma_id_private *id_priv, int timeout_ms)
+{
+ struct rdma_dev_addr *addr = &id_priv->id.route.addr.dev_addr;
+ struct ib_sa_path_rec path_rec;
+
+ memset(&path_rec, 0, sizeof path_rec);
+ path_rec.sgid = *ib_addr_get_sgid(addr);
+ path_rec.dgid = *ib_addr_get_dgid(addr);
+ path_rec.pkey = cpu_to_be16(ib_addr_get_pkey(addr));
+ path_rec.numb_path = 1;
+
+ id_priv->query_id = ib_sa_path_rec_get(id_priv->id.device,
+ id_priv->id.port_num, &path_rec,
+ IB_SA_PATH_REC_DGID | IB_SA_PATH_REC_SGID |
+ IB_SA_PATH_REC_PKEY | IB_SA_PATH_REC_NUMB_PATH,
+ timeout_ms, GFP_KERNEL,
+ cma_query_handler, id_priv, &id_priv->query);
+
+ return (id_priv->query_id < 0) ? id_priv->query_id : 0;
+}
+
+int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp_exch(id_priv, CMA_ADDR_RESOLVED, CMA_ROUTE_QUERY))
+ return -EINVAL;
+
+ atomic_inc(&id_priv->refcount);
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ ret = cma_resolve_ib_route(id_priv, timeout_ms);
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+ if (ret)
+ goto err;
+
+ return 0;
+err:
+ cma_comp_exch(id_priv, CMA_ROUTE_QUERY, CMA_ADDR_RESOLVED);
+ cma_deref_id(id_priv);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_resolve_route);
+
+static int cma_bind_loopback(struct rdma_id_private *id_priv)
+{
+ struct cma_device *cma_dev;
+ union ib_gid *gid;
+ u16 pkey;
+ int ret;
+
+ down(&mutex);
+ if (list_empty(&dev_list)) {
+ ret = -ENODEV;
+ goto out;
+ }
+
+ cma_dev = list_entry(dev_list.next, struct cma_device, list);
+ gid = ib_addr_get_sgid(&id_priv->id.route.addr.dev_addr);
+ ret = ib_get_cached_gid(cma_dev->device, 1, 0, gid);
+ if (ret)
+ goto out;
+
+ ret = ib_get_cached_pkey(cma_dev->device, 1, 0, &pkey);
+ if (ret)
+ goto out;
+
+ ib_addr_set_pkey(&id_priv->id.route.addr.dev_addr, pkey);
+ id_priv->id.port_num = 1;
+ cma_attach_to_dev(id_priv, cma_dev);
+out:
+ up(&mutex);
+ return ret;
+}
+
+static void addr_handler(int status, struct sockaddr *src_addr,
+ struct rdma_dev_addr *dev_addr, void *context)
+{
+ struct rdma_id_private *id_priv = context;
+ enum rdma_cm_event_type event;
+ enum cma_state old_state;
+
+ atomic_inc(&id_priv->dev_remove);
+ if (!id_priv->cma_dev) {
+ old_state = CMA_IDLE;
+ if (!status)
+ status = cma_acquire_dev(id_priv);
+ } else
+ old_state = CMA_ADDR_BOUND;
+
+ if (status) {
+ if (!cma_comp_exch(id_priv, CMA_ADDR_QUERY, old_state))
+ goto out;
+ event = RDMA_CM_EVENT_ADDR_ERROR;
+ } else {
+ if (!cma_comp_exch(id_priv, CMA_ADDR_QUERY, CMA_ADDR_RESOLVED))
+ goto out;
+ memcpy(&id_priv->id.route.addr.src_addr, src_addr,
+ ip_addr_size(src_addr));
+ event = RDMA_CM_EVENT_ADDR_RESOLVED;
+ }
+
+ if (cma_notify_user(id_priv, event, status, NULL, 0)) {
+ cma_exch(id_priv, CMA_DESTROYING);
+ cma_release_remove(id_priv);
+ cma_deref_id(id_priv);
+ rdma_destroy_id(&id_priv->id);
+ return;
+ }
+out:
+ cma_release_remove(id_priv);
+ cma_deref_id(id_priv);
+}
+
+static void loopback_addr_handler(void *data)
+{
+ struct cma_work *work = data;
+ struct rdma_id_private *id_priv = work->id;
+
+ kfree(work);
+ atomic_inc(&id_priv->dev_remove);
+
+ if (!cma_comp_exch(id_priv, CMA_ADDR_QUERY, CMA_ADDR_RESOLVED))
+ goto out;
+
+ if (cma_notify_user(id_priv, RDMA_CM_EVENT_ADDR_RESOLVED, 0, NULL, 0)) {
+ cma_exch(id_priv, CMA_DESTROYING);
+ cma_release_remove(id_priv);
+ cma_deref_id(id_priv);
+ rdma_destroy_id(&id_priv->id);
+ return;
+ }
+out:
+ cma_release_remove(id_priv);
+ cma_deref_id(id_priv);
+}
+
+static int cma_resolve_loopback(struct rdma_id_private *id_priv,
+ struct sockaddr *src_addr, enum cma_state state)
+{
+ struct cma_work *work;
+ struct rdma_dev_addr *dev_addr;
+ int ret;
+
+ work = kmalloc(sizeof *work, GFP_KERNEL);
+ if (!work)
+ return -ENOMEM;
+
+ if (state == CMA_IDLE) {
+ ret = cma_bind_loopback(id_priv);
+ if (ret)
+ goto err;
+ dev_addr = &id_priv->id.route.addr.dev_addr;
+ ib_addr_set_dgid(dev_addr, ib_addr_get_sgid(dev_addr));
+ if (!src_addr || cma_any_addr(src_addr))
+ src_addr = &id_priv->id.route.addr.dst_addr;
+ memcpy(&id_priv->id.route.addr.src_addr, src_addr,
+ ip_addr_size(src_addr));
+ }
+
+ work->id = id_priv;
+ INIT_WORK(&work->work, loopback_addr_handler, work);
+ queue_work(rdma_wq, &work->work);
+ return 0;
+err:
+ kfree(work);
+ return ret;
+}
+
+int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
+ struct sockaddr *dst_addr, int timeout_ms)
+{
+ struct rdma_id_private *id_priv;
+ enum cma_state expected_state;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (id_priv->cma_dev) {
+ expected_state = CMA_ADDR_BOUND;
+ src_addr = &id->route.addr.src_addr;
+ } else
+ expected_state = CMA_IDLE;
+
+ if (!cma_comp_exch(id_priv, expected_state, CMA_ADDR_QUERY))
+ return -EINVAL;
+
+ atomic_inc(&id_priv->refcount);
+ memcpy(&id->route.addr.dst_addr, dst_addr, ip_addr_size(dst_addr));
+ if (cma_loopback_addr(dst_addr))
+ ret = cma_resolve_loopback(id_priv, src_addr, expected_state);
+ else
+ ret = rdma_resolve_ip(src_addr, dst_addr,
+ &id->route.addr.dev_addr,
+ timeout_ms, addr_handler, id_priv);
+ if (ret)
+ goto err;
+
+ return 0;
+err:
+ cma_comp_exch(id_priv, CMA_ADDR_QUERY, expected_state);
+ cma_deref_id(id_priv);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_resolve_addr);
+
+int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
+{
+ struct rdma_id_private *id_priv;
+ struct rdma_dev_addr *dev_addr;
+ int ret;
+
+ if (addr->sa_family != AF_INET)
+ return -EINVAL;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp_exch(id_priv, CMA_IDLE, CMA_ADDR_BOUND))
+ return -EINVAL;
+
+ if (cma_any_addr(addr)) {
+ ret = 0;
+ } else if (cma_loopback_addr(addr)) {
+ ret = cma_bind_loopback(id_priv);
+ } else {
+ dev_addr = &id->route.addr.dev_addr;
+ ret = rdma_translate_ip(addr, dev_addr);
+ if (!ret)
+ ret = cma_acquire_dev(id_priv);
+ }
+
+ if (ret)
+ goto err;
+
+ memcpy(&id->route.addr.src_addr, addr, ip_addr_size(addr));
+ return 0;
+err:
+ cma_comp_exch(id_priv, CMA_ADDR_BOUND, CMA_IDLE);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_bind_addr);
+
+static void cma_format_hdr(void *hdr, enum rdma_port_space ps,
+ struct rdma_route *route)
+{
+ struct sockaddr_in *src4, *dst4;
+ struct cma_hdr *cma_hdr;
+ struct sdp_hh *sdp_hdr;
+
+ src4 = (struct sockaddr_in *) &route->addr.src_addr;
+ dst4 = (struct sockaddr_in *) &route->addr.dst_addr;
+
+ switch (ps) {
+ case RDMA_PS_SDP:
+ sdp_hdr = hdr;
+ sdp_hdr->sdp_version = SDP_VERSION;
+ sdp_set_ip_ver(sdp_hdr, 4);
+ sdp_hdr->src_addr.ip4.addr = src4->sin_addr.s_addr;
+ sdp_hdr->dst_addr.ip4.addr = dst4->sin_addr.s_addr;
+ sdp_hdr->port = src4->sin_port;
+ break;
+ default:
+ cma_hdr = hdr;
+ cma_hdr->cma_version = CMA_VERSION;
+ cma_set_ip_ver(cma_hdr, 4);
+ cma_hdr->src_addr.ip4.addr = src4->sin_addr.s_addr;
+ cma_hdr->dst_addr.ip4.addr = dst4->sin_addr.s_addr;
+ cma_hdr->port = src4->sin_port;
+ break;
+ }
+}
+
+static int cma_connect_ib(struct rdma_id_private *id_priv,
+ struct rdma_conn_param *conn_param)
+{
+ struct ib_cm_req_param req;
+ struct rdma_route *route;
+ void *private_data;
+ int offset, ret;
+
+ memset(&req, 0, sizeof req);
+ offset = cma_user_data_offset(id_priv->id.ps);
+ req.private_data_len = offset + conn_param->private_data_len;
+ private_data = kzalloc(req.private_data_len, GFP_ATOMIC);
+ if (!private_data)
+ return -ENOMEM;
+
+ if (conn_param->private_data && conn_param->private_data_len)
+ memcpy(private_data + offset, conn_param->private_data,
+ conn_param->private_data_len);
+
+ id_priv->cm_id = ib_create_cm_id(id_priv->id.device, cma_ib_handler,
+ id_priv);
+ if (IS_ERR(id_priv->cm_id)) {
+ ret = PTR_ERR(id_priv->cm_id);
+ goto out;
+ }
+
+ route = &id_priv->id.route;
+ cma_format_hdr(private_data, id_priv->id.ps, route);
+ req.private_data = private_data;
+
+ req.primary_path = &route->path_rec[0];
+ if (route->num_paths == 2)
+ req.alternate_path = &route->path_rec[1];
+
+ req.service_id = cma_get_service_id(id_priv->id.ps,
+ &route->addr.dst_addr);
+ req.qp_num = id_priv->qp_num;
+ req.qp_type = id_priv->qp_type;
+ req.starting_psn = id_priv->seq_num;
+ req.responder_resources = conn_param->responder_resources;
+ req.initiator_depth = conn_param->initiator_depth;
+ req.flow_control = conn_param->flow_control;
+ req.retry_count = conn_param->retry_count;
+ req.rnr_retry_count = conn_param->rnr_retry_count;
+ req.remote_cm_response_timeout = CMA_CM_RESPONSE_TIMEOUT;
+ req.local_cm_response_timeout = CMA_CM_RESPONSE_TIMEOUT;
+ req.max_cm_retries = CMA_MAX_CM_RETRIES;
+ req.srq = id_priv->srq ? 1 : 0;
+
+ ret = ib_send_cm_req(id_priv->cm_id, &req);
+out:
+ kfree(private_data);
+ return ret;
+}
+
+int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp_exch(id_priv, CMA_ROUTE_RESOLVED, CMA_CONNECT))
+ return -EINVAL;
+
+ if (!id->qp) {
+ id_priv->qp_num = conn_param->qp_num;
+ id_priv->qp_type = conn_param->qp_type;
+ id_priv->srq = conn_param->srq;
+ }
+
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ ret = cma_connect_ib(id_priv, conn_param);
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+ if (ret)
+ goto err;
+
+ return 0;
+err:
+ cma_comp_exch(id_priv, CMA_CONNECT, CMA_ROUTE_RESOLVED);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_connect);
+
+static int cma_accept_ib(struct rdma_id_private *id_priv,
+ struct rdma_conn_param *conn_param)
+{
+ struct ib_cm_rep_param rep;
+ int ret;
+
+ ret = cma_modify_qp_rtr(&id_priv->id);
+ if (ret)
+ return ret;
+
+ memset(&rep, 0, sizeof rep);
+ rep.qp_num = id_priv->qp_num;
+ rep.starting_psn = id_priv->seq_num;
+ rep.private_data = conn_param->private_data;
+ rep.private_data_len = conn_param->private_data_len;
+ rep.responder_resources = conn_param->responder_resources;
+ rep.initiator_depth = conn_param->initiator_depth;
+ rep.target_ack_delay = CMA_CM_RESPONSE_TIMEOUT;
+ rep.failover_accepted = 0;
+ rep.flow_control = conn_param->flow_control;
+ rep.rnr_retry_count = conn_param->rnr_retry_count;
+ rep.srq = id_priv->srq ? 1 : 0;
+
+ return ib_send_cm_rep(id_priv->cm_id, &rep);
+}
+
+int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp(id_priv, CMA_CONNECT))
+ return -EINVAL;
+
+ if (!id->qp && conn_param) {
+ id_priv->qp_num = conn_param->qp_num;
+ id_priv->qp_type = conn_param->qp_type;
+ id_priv->srq = conn_param->srq;
+ }
+
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ if (conn_param)
+ ret = cma_accept_ib(id_priv, conn_param);
+ else
+ ret = cma_rep_recv(id_priv);
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+
+ if (ret)
+ goto reject;
+
+ return 0;
+reject:
+ cma_modify_qp_err(id);
+ rdma_reject(id, NULL, 0);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_accept);
+
+int rdma_reject(struct rdma_cm_id *id, const void *private_data,
+ u8 private_data_len)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp(id_priv, CMA_CONNECT))
+ return -EINVAL;
+
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ ret = ib_send_cm_rej(id_priv->cm_id, IB_CM_REJ_CONSUMER_DEFINED,
+ NULL, 0, private_data, private_data_len);
+ break;
+ default:
+ ret = -ENOSYS;
+ break;
+ }
+ return ret;
+};
+EXPORT_SYMBOL(rdma_reject);
+
+int rdma_disconnect(struct rdma_cm_id *id)
+{
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ id_priv = container_of(id, struct rdma_id_private, id);
+ if (!cma_comp(id_priv, CMA_CONNECT))
+ return -EINVAL;
+
+ ret = cma_modify_qp_err(id);
+ if (ret)
+ goto out;
+
+ switch (id->device->node_type) {
+ case IB_NODE_CA:
+ /* Initiate or respond to a disconnect. */
+ if (ib_send_cm_dreq(id_priv->cm_id, NULL, 0))
+ ib_send_cm_drep(id_priv->cm_id, NULL, 0);
+ break;
+ default:
+ break;
+ }
+out:
+ return ret;
+}
+EXPORT_SYMBOL(rdma_disconnect);
+
+static void cma_add_one(struct ib_device *device)
+{
+ struct cma_device *cma_dev;
+ struct rdma_id_private *id_priv;
+
+ cma_dev = kmalloc(sizeof *cma_dev, GFP_KERNEL);
+ if (!cma_dev)
+ return;
+
+ cma_dev->device = device;
+ cma_dev->node_guid = device->node_guid;
+ if (!cma_dev->node_guid)
+ goto err;
+
+ init_waitqueue_head(&cma_dev->wait);
+ atomic_set(&cma_dev->refcount, 1);
+ INIT_LIST_HEAD(&cma_dev->id_list);
+ ib_set_client_data(device, &cma_client, cma_dev);
+
+ down(&mutex);
+ list_add_tail(&cma_dev->list, &dev_list);
+ list_for_each_entry(id_priv, &listen_any_list, list)
+ cma_listen_on_dev(id_priv, cma_dev);
+ up(&mutex);
+ return;
+err:
+ kfree(cma_dev);
+}
+
+static int cma_remove_id_dev(struct rdma_id_private *id_priv)
+{
+ enum cma_state state;
+
+ /* Record that we want to remove the device */
+ state = cma_exch(id_priv, CMA_DEVICE_REMOVAL);
+ if (state == CMA_DESTROYING)
+ return 0;
+
+ cma_cancel_operation(id_priv, state);
+ wait_event(id_priv->wait_remove, !atomic_read(&id_priv->dev_remove));
+
+ /* Check for destruction from another callback. */
+ if (!cma_comp(id_priv, CMA_DEVICE_REMOVAL))
+ return 0;
+
+ return cma_notify_user(id_priv, RDMA_CM_EVENT_DEVICE_REMOVAL,
+ 0, NULL, 0);
+}
+
+static void cma_process_remove(struct cma_device *cma_dev)
+{
+ struct list_head remove_list;
+ struct rdma_id_private *id_priv;
+ int ret;
+
+ INIT_LIST_HEAD(&remove_list);
+
+ down(&mutex);
+ while (!list_empty(&cma_dev->id_list)) {
+ id_priv = list_entry(cma_dev->id_list.next,
+ struct rdma_id_private, list);
+
+ if (cma_internal_listen(id_priv)) {
+ cma_destroy_listen(id_priv);
+ continue;
+ }
+
+ list_del(&id_priv->list);
+ list_add_tail(&id_priv->list, &remove_list);
+ atomic_inc(&id_priv->refcount);
+ up(&mutex);
+
+ ret = cma_remove_id_dev(id_priv);
+ cma_deref_id(id_priv);
+ if (ret)
+ rdma_destroy_id(&id_priv->id);
+
+ down(&mutex);
+ }
+ up(&mutex);
+
+ atomic_dec(&cma_dev->refcount);
+ wait_event(cma_dev->wait, !atomic_read(&cma_dev->refcount));
+}
+
+static void cma_remove_one(struct ib_device *device)
+{
+ struct cma_device *cma_dev;
+
+ cma_dev = ib_get_client_data(device, &cma_client);
+ if (!cma_dev)
+ return;
+
+ down(&mutex);
+ list_del(&cma_dev->list);
+ up(&mutex);
+
+ cma_process_remove(cma_dev);
+ kfree(cma_dev);
+}
+
+static int cma_init(void)
+{
+ return ib_register_client(&cma_client);
+}
+
+static void cma_cleanup(void)
+{
+ ib_unregister_client(&cma_client);
+}
+
+module_init(cma_init);
+module_exit(cma_cleanup);
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/drivers/infiniband/core/Makefile
linux-2.6.ib/drivers/infiniband/core/Makefile
--- linux-2.6.git/drivers/infiniband/core/Makefile 2006-01-16 16:16:18.000000000 -0800
+++ linux-2.6.ib/drivers/infiniband/core/Makefile 2006-01-16 16:35:48.000000000 -0800
@@ -1,5 +1,5 @@
obj-$(CONFIG_INFINIBAND) += ib_core.o ib_mad.o ib_sa.o \
- ib_cm.o ib_addr.o
+ ib_cm.o ib_addr.o rdma_cm.o
obj-$(CONFIG_INFINIBAND_USER_MAD) += ib_umad.o
obj-$(CONFIG_INFINIBAND_USER_ACCESS) += ib_uverbs.o ib_ucm.o
@@ -12,6 +12,8 @@ ib_sa-y := sa_query.o
ib_cm-y := cm.o
+rdma_cm-y := cma.o
+
ib_addr-y := addr.o
ib_umad-y := user_mad.o
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/include/rdma/rdma_cm.h
linux-2.6.ib/include/rdma/rdma_cm.h
--- linux-2.6.git/include/rdma/rdma_cm.h 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.ib/include/rdma/rdma_cm.h 2006-01-16 16:19:12.000000000 -0800
@@ -0,0 +1,255 @@
+/*
+ * Copyright (c) 2005 Voltaire Inc. All rights reserved.
+ * Copyright (c) 2005 Intel Corporation. All rights reserved.
+ *
+ * This Software is licensed under one of the following licenses:
+ *
+ * 1) under the terms of the "Common Public License 1.0" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/cpl.php.
+ *
+ * 2) under the terms of the "The BSD License" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/bsd-license.php.
+ *
+ * 3) under the terms of the "GNU General Public License (GPL) Version 2" a
+ * copy of which is available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/gpl-license.php.
+ *
+ * Licensee has the right to choose one of the above licenses.
+ *
+ * Redistributions of source code must retain the above copyright
+ * notice and one of the license notices.
+ *
+ * Redistributions in binary form must reproduce both the above copyright
+ * notice, one of the license notices in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ */
+
+#if !defined(RDMA_CM_H)
+#define RDMA_CM_H
+
+#include <linux/socket.h>
+#include <linux/in6.h>
+#include <rdma/ib_addr.h>
+#include <rdma/ib_sa.h>
+
+/*
+ * Upon receiving a device removal event, users must destroy the associated
+ * RDMA identifier and release all resources allocated with the device.
+ */
+enum rdma_cm_event_type {
+ RDMA_CM_EVENT_ADDR_RESOLVED,
+ RDMA_CM_EVENT_ADDR_ERROR,
+ RDMA_CM_EVENT_ROUTE_RESOLVED,
+ RDMA_CM_EVENT_ROUTE_ERROR,
+ RDMA_CM_EVENT_CONNECT_REQUEST,
+ RDMA_CM_EVENT_CONNECT_RESPONSE,
+ RDMA_CM_EVENT_CONNECT_ERROR,
+ RDMA_CM_EVENT_UNREACHABLE,
+ RDMA_CM_EVENT_REJECTED,
+ RDMA_CM_EVENT_ESTABLISHED,
+ RDMA_CM_EVENT_DISCONNECTED,
+ RDMA_CM_EVENT_DEVICE_REMOVAL,
+};
+
+enum rdma_port_space {
+ RDMA_PS_SDP = 0x0001,
+ RDMA_PS_TCP = 0x0106,
+ RDMA_PS_UDP = 0x0111,
+ RDMA_PS_SCTP = 0x0183
+};
+
+struct rdma_addr {
+ struct sockaddr src_addr;
+ u8 src_pad[sizeof(struct sockaddr_in6) -
+ sizeof(struct sockaddr)];
+ struct sockaddr dst_addr;
+ u8 dst_pad[sizeof(struct sockaddr_in6) -
+ sizeof(struct sockaddr)];
+ struct rdma_dev_addr dev_addr;
+};
+
+struct rdma_route {
+ struct rdma_addr addr;
+ struct ib_sa_path_rec *path_rec;
+ int num_paths;
+};
+
+struct rdma_cm_event {
+ enum rdma_cm_event_type event;
+ int status;
+ void *private_data;
+ u8 private_data_len;
+};
+
+struct rdma_cm_id;
+
+/**
+ * rdma_cm_event_handler - Callback used to report user events.
+ *
+ * Notes: Users may not call rdma_destroy_id from this callback to destroy
+ * the passed in id, or a corresponding listen id. Returning a
+ * non-zero value from the callback will destroy the corresponding id.
+ */
+typedef int (*rdma_cm_event_handler)(struct rdma_cm_id *id,
+ struct rdma_cm_event *event);
+
+struct rdma_cm_id {
+ struct ib_device *device;
+ void *context;
+ struct ib_qp *qp;
+ rdma_cm_event_handler event_handler;
+ struct rdma_route route;
+ enum rdma_port_space ps;
+ u8 port_num;
+};
+
+/**
+ * rdma_create_id - Create an RDMA identifier.
+ *
+ * @event_handler: User callback invoked to report events associated with the
+ * returned rdma_id.
+ * @context: User specified context associated with the id.
+ * @ps: RDMA port space.
+ */
+struct rdma_cm_id* rdma_create_id(rdma_cm_event_handler event_handler,
+ void *context, enum rdma_port_space ps);
+
+void rdma_destroy_id(struct rdma_cm_id *id);
+
+/**
+ * rdma_bind_addr - Bind an RDMA identifier to a source address and
+ * associated RDMA device, if needed.
+ *
+ * @id: RDMA identifier.
+ * @addr: Local address information. Wildcard values are permitted.
+ *
+ * This associates a source address with the RDMA identifier before calling
+ * rdma_listen. If a specific local address is given, the RDMA identifier will
+ * be bound to a local RDMA device.
+ */
+int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr);
+
+/**
+ * rdma_resolve_addr - Resolve destination and optional source addresses
+ * from IP addresses to an RDMA address. If successful, the specified
+ * rdma_cm_id will be bound to a local device.
+ *
+ * @id: RDMA identifier.
+ * @src_addr: Source address information. This parameter may be NULL.
+ * @dst_addr: Destination address information.
+ * @timeout_ms: Time to wait for resolution to complete.
+ */
+int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
+ struct sockaddr *dst_addr, int timeout_ms);
+
+/**
+ * rdma_resolve_route - Resolve the RDMA address bound to the RDMA identifier
+ * into route information needed to establish a connection.
+ *
+ * This is called on the client side of a connection.
+ * Users must have first called rdma_resolve_addr to resolve a dst_addr
+ * into an RDMA address before calling this routine.
+ */
+int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms);
+
+/**
+ * rdma_create_qp - Allocate a QP and associate it with the specified RDMA
+ * identifier.
+ *
+ * QPs allocated to an rdma_cm_id will automatically be transitioned by the CMA
+ * through their states.
+ */
+int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd,
+ struct ib_qp_init_attr *qp_init_attr);
+
+/**
+ * rdma_destroy_qp - Deallocate the QP associated with the specified RDMA
+ * identifier.
+ *
+ * Users must destroy any QP associated with an RDMA identifier before
+ * destroying the RDMA ID.
+ */
+void rdma_destroy_qp(struct rdma_cm_id *id);
+
+/**
+ * rdma_init_qp_attr - Initializes the QP attributes for use in transitioning
+ * to a specified QP state.
+ * @id: Communication identifier associated with the QP attributes to
+ * initialize.
+ * @qp_attr: On input, specifies the desired QP state. On output, the
+ * mandatory and desired optional attributes will be set in order to
+ * modify the QP to the specified state.
+ * @qp_attr_mask: The QP attribute mask that may be used to transition the
+ * QP to the specified state.
+ *
+ * Users must set the @qp_attr->qp_state to the desired QP state. This call
+ * will set all required attributes for the given transition, along with
+ * known optional attributes. Users may override the attributes returned from
+ * this call before calling ib_modify_qp.
+ *
+ * Users that wish to have their QP automatically transitioned through its
+ * states can associate a QP with the rdma_cm_id by calling rdma_create_qp().
+ */
+int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr,
+ int *qp_attr_mask);
+
+struct rdma_conn_param {
+ const void *private_data;
+ u8 private_data_len;
+ u8 responder_resources;
+ u8 initiator_depth;
+ u8 flow_control;
+ u8 retry_count; /* ignored when accepting */
+ u8 rnr_retry_count;
+ /* Fields below ignored if a QP is created on the rdma_cm_id. */
+ u8 srq;
+ u32 qp_num;
+ enum ib_qp_type qp_type;
+};
+
+/**
+ * rdma_connect - Initiate an active connection request.
+ *
+ * Users must have resolved a route for the rdma_cm_id to connect with
+ * by having called rdma_resolve_route before calling this routine.
+ */
+int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param);
+
+/**
+ * rdma_listen - This function is called by the passive side to
+ * listen for incoming connection requests.
+ *
+ * Users must have bound the rdma_cm_id to a local address by calling
+ * rdma_bind_addr before calling this routine.
+ */
+int rdma_listen(struct rdma_cm_id *id, int backlog);
+
+/**
+ * rdma_accept - Called to accept a connection request or response.
+ * @id: Connection identifier associated with the request.
+ * @conn_param: Information needed to establish the connection. This must be
+ * provided if accepting a connection request. If accepting a connection
+ * response, this parameter must be NULL.
+ *
+ * Typically, this routine is only called by the listener to accept a connection
+ * request. It must also be called on the active side of a connection if the
+ * user is performing their own QP transitions.
+ */
+int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param);
+
+/**
+ * rdma_reject - Called on the passive side to reject a connection request.
+ */
+int rdma_reject(struct rdma_cm_id *id, const void *private_data,
+ u8 private_data_len);
+
+/**
+ * rdma_disconnect - This function disconnects the associated QP.
+ */
+int rdma_disconnect(struct rdma_cm_id *id);
+
+#endif /* RDMA_CM_H */
+
^ permalink raw reply
* Re: de2104x: interrupts before interrupt handler is registered
From: Martin Michlmayr @ 2006-03-06 19:17 UTC (permalink / raw)
To: Francois Romieu; +Cc: netdev, linux-kernel
In-Reply-To: <20060306143512.GI23669@deprecation.cyrius.com>
* Martin Michlmayr <tbm@cyrius.com> [2006-03-06 14:35]:
> Thanks a lot for your quick response, Francois. I can confirm that
> this patch fixes the problem for me.
There's another interrupt related bug in the driver, though. I
sometimes get a kernel panic when rsycing several 100 megs of data
across the LAN. A picture showing the call trace can be found at
http://www.cyrius.com/tmp/de2104x_panic.jpg
--
Martin Michlmayr
http://www.cyrius.com/
^ permalink raw reply
* [PATCH 4/6] IB: address translation to map IP to IB addresses (GIDs)
From: Sean Hefty @ 2006-03-06 19:10 UTC (permalink / raw)
To: 'Roland Dreier', linux-kernel, netdev; +Cc: openib-general
In-Reply-To: <adaslpz2l9p.fsf@cisco.com>
Add an address translation service that maps IP addresses to Infiniband
GID addresses using IPoIB.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
---
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/drivers/infiniband/core/addr.c
linux-2.6.ib/drivers/infiniband/core/addr.c
--- linux-2.6.git/drivers/infiniband/core/addr.c 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.ib/drivers/infiniband/core/addr.c 2006-01-16 16:14:24.000000000 -0800
@@ -0,0 +1,356 @@
+/*
+ * Copyright (c) 2005 Voltaire Inc. All rights reserved.
+ * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved.
+ * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved.
+ * Copyright (c) 2005 Intel Corporation. All rights reserved.
+ *
+ * This Software is licensed under one of the following licenses:
+ *
+ * 1) under the terms of the "Common Public License 1.0" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/cpl.php.
+ *
+ * 2) under the terms of the "The BSD License" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/bsd-license.php.
+ *
+ * 3) under the terms of the "GNU General Public License (GPL) Version 2" a
+ * copy of which is available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/gpl-license.php.
+ *
+ * Licensee has the right to choose one of the above licenses.
+ *
+ * Redistributions of source code must retain the above copyright
+ * notice and one of the license notices.
+ *
+ * Redistributions in binary form must reproduce both the above copyright
+ * notice, one of the license notices in the documentation
+ * and/or other materials provided with the distribution.
+ */
+#include <linux/inetdevice.h>
+#include <linux/workqueue.h>
+#include <net/arp.h>
+#include <net/neighbour.h>
+#include <net/route.h>
+#include <rdma/ib_addr.h>
+
+MODULE_AUTHOR("Sean Hefty");
+MODULE_DESCRIPTION("IB Address Translation");
+MODULE_LICENSE("Dual BSD/GPL");
+
+struct addr_req {
+ struct list_head list;
+ struct sockaddr src_addr;
+ struct sockaddr dst_addr;
+ struct rdma_dev_addr *addr;
+ void *context;
+ void (*callback)(int status, struct sockaddr *src_addr,
+ struct rdma_dev_addr *addr, void *context);
+ unsigned long timeout;
+ int status;
+};
+
+static void process_req(void *data);
+
+static DECLARE_MUTEX(mutex);
+static LIST_HEAD(req_list);
+static DECLARE_WORK(work, process_req, NULL);
+struct workqueue_struct *rdma_wq;
+EXPORT_SYMBOL(rdma_wq);
+
+static int copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
+ unsigned char *dst_dev_addr)
+{
+ switch (dev->type) {
+ case ARPHRD_INFINIBAND:
+ dev_addr->dev_type = IB_NODE_CA;
+ break;
+ default:
+ return -EADDRNOTAVAIL;
+ }
+
+ memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
+ memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN);
+ if (dst_dev_addr)
+ memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
+ return 0;
+}
+
+int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
+{
+ struct net_device *dev;
+ u32 ip = ((struct sockaddr_in *) addr)->sin_addr.s_addr;
+ int ret;
+
+ dev = ip_dev_find(ip);
+ if (!dev)
+ return -EADDRNOTAVAIL;
+
+ ret = copy_addr(dev_addr, dev, NULL);
+ dev_put(dev);
+ return ret;
+}
+EXPORT_SYMBOL(rdma_translate_ip);
+
+static void set_timeout(unsigned long time)
+{
+ unsigned long delay;
+
+ cancel_delayed_work(&work);
+
+ delay = time - jiffies;
+ if ((long)delay <= 0)
+ delay = 1;
+
+ queue_delayed_work(rdma_wq, &work, delay);
+}
+
+static void queue_req(struct addr_req *req)
+{
+ struct addr_req *temp_req;
+
+ down(&mutex);
+ list_for_each_entry_reverse(temp_req, &req_list, list) {
+ if (time_after(req->timeout, temp_req->timeout))
+ break;
+ }
+
+ list_add(&req->list, &temp_req->list);
+
+ if (req_list.next == &req->list)
+ set_timeout(req->timeout);
+ up(&mutex);
+}
+
+static void addr_send_arp(struct sockaddr_in *dst_in)
+{
+ struct rtable *rt;
+ struct flowi fl;
+ u32 dst_ip = dst_in->sin_addr.s_addr;
+
+ memset(&fl, 0, sizeof fl);
+ fl.nl_u.ip4_u.daddr = dst_ip;
+ if (ip_route_output_key(&rt, &fl))
+ return;
+
+ arp_send(ARPOP_REQUEST, ETH_P_ARP, rt->rt_gateway, rt->idev->dev,
+ rt->rt_src, NULL, rt->idev->dev->dev_addr, NULL);
+ ip_rt_put(rt);
+}
+
+static int addr_resolve_remote(struct sockaddr_in *src_in,
+ struct sockaddr_in *dst_in,
+ struct rdma_dev_addr *addr)
+{
+ u32 src_ip = src_in->sin_addr.s_addr;
+ u32 dst_ip = dst_in->sin_addr.s_addr;
+ struct flowi fl;
+ struct rtable *rt;
+ struct neighbour *neigh;
+ int ret;
+
+ memset(&fl, 0, sizeof fl);
+ fl.nl_u.ip4_u.daddr = dst_ip;
+ fl.nl_u.ip4_u.saddr = src_ip;
+ ret = ip_route_output_key(&rt, &fl);
+ if (ret)
+ goto out;
+
+ neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, rt->idev->dev);
+ if (!neigh) {
+ ret = -ENODATA;
+ goto err1;
+ }
+
+ if (!(neigh->nud_state & NUD_VALID)) {
+ ret = -ENODATA;
+ goto err2;
+ }
+
+ if (!src_ip) {
+ src_in->sin_family = dst_in->sin_family;
+ src_in->sin_addr.s_addr = rt->rt_src;
+ }
+
+ ret = copy_addr(addr, neigh->dev, neigh->ha);
+err2:
+ neigh_release(neigh);
+err1:
+ ip_rt_put(rt);
+out:
+ return ret;
+}
+
+static void process_req(void *data)
+{
+ struct addr_req *req, *temp_req;
+ struct sockaddr_in *src_in, *dst_in;
+ struct list_head done_list;
+
+ INIT_LIST_HEAD(&done_list);
+
+ down(&mutex);
+ list_for_each_entry_safe(req, temp_req, &req_list, list) {
+ if (req->status) {
+ src_in = (struct sockaddr_in *) &req->src_addr;
+ dst_in = (struct sockaddr_in *) &req->dst_addr;
+ req->status = addr_resolve_remote(src_in, dst_in,
+ req->addr);
+ }
+ if (req->status && time_after(jiffies, req->timeout))
+ req->status = -ETIMEDOUT;
+ else if (req->status == -ENODATA)
+ continue;
+
+ list_del(&req->list);
+ list_add_tail(&req->list, &done_list);
+ }
+
+ if (!list_empty(&req_list)) {
+ req = list_entry(req_list.next, struct addr_req, list);
+ set_timeout(req->timeout);
+ }
+ up(&mutex);
+
+ list_for_each_entry_safe(req, temp_req, &done_list, list) {
+ list_del(&req->list);
+ req->callback(req->status, &req->src_addr, req->addr,
+ req->context);
+ kfree(req);
+ }
+}
+
+static int addr_resolve_local(struct sockaddr_in *src_in,
+ struct sockaddr_in *dst_in,
+ struct rdma_dev_addr *addr)
+{
+ struct net_device *dev;
+ u32 src_ip = src_in->sin_addr.s_addr;
+ u32 dst_ip = dst_in->sin_addr.s_addr;
+ int ret;
+
+ dev = ip_dev_find(dst_ip);
+ if (!dev)
+ return -EADDRNOTAVAIL;
+
+ if (!src_ip) {
+ src_in->sin_family = dst_in->sin_family;
+ src_in->sin_addr.s_addr = dst_ip;
+ ret = copy_addr(addr, dev, dev->dev_addr);
+ } else {
+ ret = rdma_translate_ip((struct sockaddr *)src_in, addr);
+ if (!ret)
+ memcpy(addr->dst_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
+ }
+
+ dev_put(dev);
+ return ret;
+}
+
+int rdma_resolve_ip(struct sockaddr *src_addr, struct sockaddr *dst_addr,
+ struct rdma_dev_addr *addr, int timeout_ms,
+ void (*callback)(int status, struct sockaddr *src_addr,
+ struct rdma_dev_addr *addr, void *context),
+ void *context)
+{
+ struct sockaddr_in *src_in, *dst_in;
+ struct addr_req *req;
+ int ret = 0;
+
+ req = kmalloc(sizeof *req, GFP_KERNEL);
+ if (!req)
+ return -ENOMEM;
+ memset(req, 0, sizeof *req);
+
+ if (src_addr)
+ memcpy(&req->src_addr, src_addr, ip_addr_size(src_addr));
+ memcpy(&req->dst_addr, dst_addr, ip_addr_size(dst_addr));
+ req->addr = addr;
+ req->callback = callback;
+ req->context = context;
+
+ src_in = (struct sockaddr_in *) &req->src_addr;
+ dst_in = (struct sockaddr_in *) &req->dst_addr;
+
+ req->status = addr_resolve_local(src_in, dst_in, addr);
+ if (req->status == -EADDRNOTAVAIL)
+ req->status = addr_resolve_remote(src_in, dst_in, addr);
+
+ switch (req->status) {
+ case 0:
+ req->timeout = jiffies;
+ queue_req(req);
+ break;
+ case -ENODATA:
+ req->timeout = msecs_to_jiffies(timeout_ms) + jiffies;
+ queue_req(req);
+ addr_send_arp(dst_in);
+ break;
+ default:
+ ret = req->status;
+ kfree(req);
+ break;
+ }
+ return ret;
+}
+EXPORT_SYMBOL(rdma_resolve_ip);
+
+void rdma_addr_cancel(struct rdma_dev_addr *addr)
+{
+ struct addr_req *req, *temp_req;
+
+ up(&mutex);
+ list_for_each_entry_safe(req, temp_req, &req_list, list) {
+ if (req->addr == addr) {
+ req->status = -ECANCELED;
+ req->timeout = jiffies;
+ list_del(&req->list);
+ list_add(&req->list, &req_list);
+ set_timeout(req->timeout);
+ break;
+ }
+ }
+ up(&mutex);
+}
+EXPORT_SYMBOL(rdma_addr_cancel);
+
+static int addr_arp_recv(struct sk_buff *skb, struct net_device *dev,
+ struct packet_type *pkt, struct net_device *orig_dev)
+{
+ struct arphdr *arp_hdr;
+
+ arp_hdr = (struct arphdr *) skb->nh.raw;
+
+ if (dev->type == ARPHRD_INFINIBAND &&
+ (arp_hdr->ar_op == __constant_htons(ARPOP_REQUEST) ||
+ arp_hdr->ar_op == __constant_htons(ARPOP_REPLY)))
+ set_timeout(jiffies);
+
+ kfree_skb(skb);
+ return 0;
+}
+
+static struct packet_type addr_arp = {
+ .type = __constant_htons(ETH_P_ARP),
+ .func = addr_arp_recv,
+ .af_packet_priv = (void*) 1,
+};
+
+static int addr_init(void)
+{
+ rdma_wq = create_singlethread_workqueue("rdma_wq");
+ if (!rdma_wq)
+ return -ENOMEM;
+
+ dev_add_pack(&addr_arp);
+ return 0;
+}
+
+static void addr_cleanup(void)
+{
+ dev_remove_pack(&addr_arp);
+ destroy_workqueue(rdma_wq);
+}
+
+module_init(addr_init);
+module_exit(addr_cleanup);
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/drivers/infiniband/core/Makefile
linux-2.6.ib/drivers/infiniband/core/Makefile
--- linux-2.6.git/drivers/infiniband/core/Makefile 2006-01-16 16:03:08.000000000 -0800
+++ linux-2.6.ib/drivers/infiniband/core/Makefile 2006-01-16 16:14:24.000000000 -0800
@@ -1,5 +1,5 @@
obj-$(CONFIG_INFINIBAND) += ib_core.o ib_mad.o ib_sa.o \
- ib_cm.o
+ ib_cm.o ib_addr.o
obj-$(CONFIG_INFINIBAND_USER_MAD) += ib_umad.o
obj-$(CONFIG_INFINIBAND_USER_ACCESS) += ib_uverbs.o ib_ucm.o
@@ -12,6 +12,8 @@ ib_sa-y := sa_query.o
ib_cm-y := cm.o
+ib_addr-y := addr.o
+
ib_umad-y := user_mad.o
ib_ucm-y := ucm.o
diff -uprN -X linux-2.6.git/Documentation/dontdiff
linux-2.6.git/include/rdma/ib_addr.h
linux-2.6.ib/include/rdma/ib_addr.h
--- linux-2.6.git/include/rdma/ib_addr.h 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.ib/include/rdma/ib_addr.h 2006-01-16 16:14:24.000000000 -0800
@@ -0,0 +1,117 @@
+/*
+ * Copyright (c) 2005 Voltaire Inc. All rights reserved.
+ * Copyright (c) 2005 Intel Corporation. All rights reserved.
+ *
+ * This Software is licensed under one of the following licenses:
+ *
+ * 1) under the terms of the "Common Public License 1.0" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/cpl.php.
+ *
+ * 2) under the terms of the "The BSD License" a copy of which is
+ * available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/bsd-license.php.
+ *
+ * 3) under the terms of the "GNU General Public License (GPL) Version 2" a
+ * copy of which is available from the Open Source Initiative, see
+ * http://www.opensource.org/licenses/gpl-license.php.
+ *
+ * Licensee has the right to choose one of the above licenses.
+ *
+ * Redistributions of source code must retain the above copyright
+ * notice and one of the license notices.
+ *
+ * Redistributions in binary form must reproduce both the above copyright
+ * notice, one of the license notices in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ */
+
+#if !defined(IB_ADDR_H)
+#define IB_ADDR_H
+
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/netdevice.h>
+#include <linux/socket.h>
+#include <rdma/ib_verbs.h>
+
+extern struct workqueue_struct *rdma_wq;
+
+struct rdma_dev_addr {
+ unsigned char src_dev_addr[MAX_ADDR_LEN];
+ unsigned char dst_dev_addr[MAX_ADDR_LEN];
+ unsigned char broadcast[MAX_ADDR_LEN];
+ enum ib_node_type dev_type;
+};
+
+/**
+ * rdma_translate_ip - Translate a local IP address to an RDMA hardware
+ * address.
+ */
+int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr);
+
+/**
+ * rdma_resolve_ip - Resolve source and destination IP addresses to
+ * RDMA hardware addresses.
+ * @src_addr: An optional source address to use in the resolution. If a
+ * source address is not provided, a usable address will be returned via
+ * the callback.
+ * @dst_addr: The destination address to resolve.
+ * @addr: A reference to a data location that will receive the resolved
+ * addresses. The data location must remain valid until the callback has
+ * been invoked.
+ * @timeout_ms: Amount of time to wait for the address resolution to complete.
+ * @callback: Call invoked once address resolution has completed, timed out,
+ * or been canceled. A status of 0 indicates success.
+ * @context: User-specified context associated with the call.
+ */
+int rdma_resolve_ip(struct sockaddr *src_addr, struct sockaddr *dst_addr,
+ struct rdma_dev_addr *addr, int timeout_ms,
+ void (*callback)(int status, struct sockaddr *src_addr,
+ struct rdma_dev_addr *addr, void *context),
+ void *context);
+
+void rdma_addr_cancel(struct rdma_dev_addr *addr);
+
+static inline int ip_addr_size(struct sockaddr *addr)
+{
+ return addr->sa_family == AF_INET6 ?
+ sizeof(struct sockaddr_in6) : sizeof(struct sockaddr_in);
+}
+
+static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
+{
+ return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
+}
+
+static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
+{
+ dev_addr->broadcast[8] = pkey >> 8;
+ dev_addr->broadcast[9] = (unsigned char) pkey;
+}
+
+static inline union ib_gid* ib_addr_get_sgid(struct rdma_dev_addr *dev_addr)
+{
+ return (union ib_gid *) (dev_addr->src_dev_addr + 4);
+}
+
+static inline void ib_addr_set_sgid(struct rdma_dev_addr *dev_addr,
+ union ib_gid *gid)
+{
+ memcpy(dev_addr->src_dev_addr + 4, gid, sizeof *gid);
+}
+
+static inline union ib_gid* ib_addr_get_dgid(struct rdma_dev_addr *dev_addr)
+{
+ return (union ib_gid *) (dev_addr->dst_dev_addr + 4);
+}
+
+static inline void ib_addr_set_dgid(struct rdma_dev_addr *dev_addr,
+ union ib_gid *gid)
+{
+ memcpy(dev_addr->dst_dev_addr + 4, gid, sizeof *gid);
+}
+
+#endif /* IB_ADDR_H */
+
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox