* [PATCH 4/4] ehea: Kconfig and Makefile
From: Jan-Bernd Themann @ 2006-06-07 17:06 UTC (permalink / raw)
To: netdev; +Cc: meder, raisch, themann, tklein
Signed-off-by: Jan-Bernd Themann <themann@de.ibm.com>
drivers/net/ehea/Kconfig | 6 ++++++
drivers/net/ehea/Makefile | 45 +++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 51 insertions(+)
--- linux-2.6.16-rc5-orig/drivers/net/ehea/Makefile 1969-12-31 16:00:00.000000000 -0800
+++ kernel/drivers/net/ehea/Makefile 2006-06-07 07:01:14.634178704 -0700
@@ -0,0 +1,45 @@
+# Server eHEA ethernet device driver for Linux on POWER
+#
+#
+# linux/drivers/net/ehea/Makefile
+#
+# eHEA ethernet device driver for IBM eServer System p
+#
+# (C) Copyright IBM Corp. 2006
+#
+# Authors:
+# Christoph Raisch <raisch@de.ibm.com>
+# Jan-Bernd Themann <themann@de.ibm.com>
+# Heiko-Joerg Schick <schickhj@de.ibm.com>
+# Thomas Klein <tklein@de.ibm.com>
+#
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2, or (at your option)
+# any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+
+
+
+
+# To compile module type:
+# make -C /root/linux-2.6.5-7.244 SUBDIRS=`pwd` V=1 modules CONFIG_EHEA=m
+# make -C /usr/src/linux-2.6.17-rc5 SUBDIRS=`pwd` V=1 modules CONFIG_EHEA=m
+
+ehea_mod-objs = ehea_main.o ehea_phyp.o ehea_qmr.o ehea_ethtool.o ehea_phyp.o
+
+obj-$(CONFIG_EHEA) += ehea_mod.o
+
+#CFLAGS += -Werror
+
+
+
--- linux-2.6.16-rc5-orig/drivers/net/ehea/Kconfig 1969-12-31 16:00:00.000000000 -0800
+++ kernel/drivers/net/ehea/Kconfig 2006-06-07 07:01:14.637178248 -0700
@@ -0,0 +1,6 @@
+config EHEA
+ tristate "eHEA Ethernet support"
+ depends on IBMEBUS
+ ---help---
+ This driver supports the IBM pSeries
+ Ethernet adapter
^ permalink raw reply
* Re: Problem authenticating using WPA with bcm43xx-softmac
From: Larry Finger @ 2006-06-07 17:07 UTC (permalink / raw)
To: Johannes Berg; +Cc: Sam Leffler, netdev
In-Reply-To: <1149696400.3925.13.camel@johannes>
Johannes Berg wrote:
> On Wed, 2006-06-07 at 09:01 -0700, Sam Leffler wrote:
>
>> Beware of the order of IE's in the management frames; some AP's are
>> touchy about this.
>
> Uh oh. I have no idea where the ieee80211 layer sticks that one,
> probably at the end.
I moved the WPA IE from the end forward without any effect. I have dumped that IE and find the
following:
SoftMAC: Added WPA_IE of 24 bytes to association request
Contents of WPA_IE: 0xdd 0x16 0x00 0x50 0xf2 0x01 0x01 0x00 0x00 0x50
0xf2 0x02 0x01 0x00 0x00 0x50 0xf2 0x02 0x01 0x00
0x00 0x50 0xf2 0x02
I think I got the right thing, but I'm having trouble interpreting it. The dump was made by
modifying the part of ieee80211softmac_assoc_req in ieee80211softmac_io.c where the WPA IE is added:
/* Add WPA IE */
if (mac->wpa.IElen && mac->wpa.IE) {
int i;
memcpy(data, mac->wpa.IE, mac->wpa.IElen);
data += mac->wpa.IElen;
printk(KERN_INFO PFX "Added WPA_IE of %d bytes to association request\n"
"Contents of WPA_IE: ", mac->wpa.IElen);
for (i=0; i<mac->wpa.IElen; i++)
printk("0x%02x ",mac->wpa.IE[i]);
printk("\n");
}
Larry
^ permalink raw reply
* Re: [PATCH 0/4] ehea: IBM eHEA Ethernet Device Driver
From: Stephen Hemminger @ 2006-06-07 17:24 UTC (permalink / raw)
To: Jan-Bernd Themann; +Cc: netdev, raisch, themann, tklein, meder, schickhj
In-Reply-To: <4486ECED.9080608@de.ibm.com>
On Wed, 07 Jun 2006 17:12:45 +0200
Jan-Bernd Themann <ossthema@de.ibm.com> wrote:
> Hello,
>
> this is the first version of the IBM eHEA Ethernet Device Driver.
> It supports a new IBM ethernet chip on System p.
>
> Main functionality including broadcast multicast and some parts of
> ethtool already work.
>
> Things we're still working on:
> - vlan support
> - performance improvements in SMP systems
> - kernel coding style
Please:
- Get rid of the debug macro's with 9 arguments...
^ permalink raw reply
* Re: Problem authenticating using WPA with bcm43xx-softmac
From: Dan Williams @ 2006-06-07 17:29 UTC (permalink / raw)
To: Johannes Berg; +Cc: Larry Finger, netdev, Jouni Malinen
In-Reply-To: <1149695859.3925.11.camel@johannes>
On Wed, 2006-06-07 at 17:57 +0200, Johannes Berg wrote:
> On Wed, 2006-06-07 at 17:51 +0200, Johannes Berg wrote:
>
> > Well, it should be shown in the 802.11i spec too.
>
> I suppose that it is the association request, and needs to contain the
> RSN described in 7.3.2.25 as per 7.2.3.4 in 802.11i. This is, afaik, the
> 'generic IE' that is added with the wext. Now, it looks like the RSN
> isn't included but the WPA2 info or something? Also, the genIE in your
> log doesn't look correct to me, starting with ffffff?? Jouni, do you
> have any idea what might be going on?
I believe that wpa_supplicant tells the driver what genie to use through
the SIOCSIWGENIE wext call. The IEs match between what the driver
appears to be reporting, and what wpa_supplicant says from the logs.
wpa_supplicant is almost certainly writing the correct IE to the driver
through wext, so I think the debug output from softmac must be
formatting the string incorrectly when printing it out to the logs.
Looking at it further:
struct ieee80211softmac_wpa {
char *IE;
int IElen;
int IEbuflen;
};
from ieee80211softmac_wx.c: ieee80211softmac_wx_set_genie()
memcpy(mac->wpa.IE, extra, wrqu->data.length);
dprintk(KERN_INFO PFX "generic IE set to ");
for (i=0;i<wrqu->data.length;i++)
dprintk("%.2x", mac->wpa.IE[i]);
dprintk("\n");
the dprintk code isn't doing the right thing here, given an array of
bytes. You probably want:
dprintk("%.2hhx", mac->wpa.IE[i]);
(ie, add the "hh" before the x to tell the print that it's a char)
Dan
^ permalink raw reply
* Re: [PATCH 3/4] ehea: queue managment
From: Stephen Hemminger @ 2006-06-07 17:29 UTC (permalink / raw)
To: Jan-Bernd Themann; +Cc: netdev, meder, raisch, themann, tklein
In-Reply-To: <4487073E.1020403@de.ibm.com>
On Wed, 07 Jun 2006 19:05:02 +0200
Jan-Bernd Themann <ossthema@de.ibm.com> wrote:
> Signed-off-by: Jan-Bernd Themann <themann@de.ibm.com>
>
>
> drivers/net/ehea/ehea_qmr.c | 719 ++++++++++++++++++++++++++++++++++++++++++++
> drivers/net/ehea/ehea_qmr.h | 390 +++++++++++++++++++++++
> 2 files changed, 1109 insertions(+)
>
Make it look less like C++
Do you really need to use vmalloc/vfree? Are the data structures that big?
Be careful about global namespace issues. Stick to one prefix like ehea_
for all non static function names. Consider putting all in one file, or
use #include to cause it to be one compilation unit.
^ permalink raw reply
* Re: [PATCH 4/4] ehea: Kconfig and Makefile
From: Randy.Dunlap @ 2006-06-07 17:59 UTC (permalink / raw)
To: Jan-Bernd Themann; +Cc: netdev, meder, raisch, themann, tklein
In-Reply-To: <4487077B.4020405@de.ibm.com>
On Wed, 07 Jun 2006 19:06:03 +0200 Jan-Bernd Themann wrote:
> Signed-off-by: Jan-Bernd Themann <themann@de.ibm.com>
>
>
> drivers/net/ehea/Kconfig | 6 ++++++
> drivers/net/ehea/Makefile | 45 +++++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 51 insertions(+)
>
>
>
> --- linux-2.6.16-rc5-orig/drivers/net/ehea/Makefile 1969-12-31 16:00:00.000000000 -0800
> +++ kernel/drivers/net/ehea/Makefile 2006-06-07 07:01:14.634178704 -0700
> @@ -0,0 +1,45 @@
> +# Server eHEA ethernet device driver for Linux on POWER
> +#
> +#
> +# linux/drivers/net/ehea/Makefile
> +#
> +# eHEA ethernet device driver for IBM eServer System p
> +#
> +# (C) Copyright IBM Corp. 2006
> +#
> +# Authors:
> +# Christoph Raisch <raisch@de.ibm.com>
> +# Jan-Bernd Themann <themann@de.ibm.com>
> +# Heiko-Joerg Schick <schickhj@de.ibm.com>
> +# Thomas Klein <tklein@de.ibm.com>
> +#
> +#
> +# This program is free software; you can redistribute it and/or modify
> +# it under the terms of the GNU General Public License as published by
> +# the Free Software Foundation; either version 2, or (at your option)
> +# any later version.
> +#
> +# This program is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> +# GNU General Public License for more details.
> +#
> +# You should have received a copy of the GNU General Public License
> +# along with this program; if not, write to the Free Software
> +# Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
> +
> +
> +
> +
> +# To compile module type:
> +# make -C /root/linux-2.6.5-7.244 SUBDIRS=`pwd` V=1 modules CONFIG_EHEA=m
> +# make -C /usr/src/linux-2.6.17-rc5 SUBDIRS=`pwd` V=1 modules CONFIG_EHEA=m
> +
> +ehea_mod-objs = ehea_main.o ehea_phyp.o ehea_qmr.o ehea_ethtool.o ehea_phyp.o
> +
> +obj-$(CONFIG_EHEA) += ehea_mod.o
> +
> +#CFLAGS += -Werror
> +
> +
> +
45 lines for a 5-line makefile. :(
You need to integrate the Makefile and Kconfig into the upper-level
directory Makefile and Kconfig files. Don't expect/require people
to enter thing like this:
> +# make -C /root/linux-2.6.5-7.244 SUBDIRS=`pwd` V=1 modules CONFIG_EHEA=m
> +# make -C /usr/src/linux-2.6.17-rc5 SUBDIRS=`pwd` V=1 modules CONFIG_EHEA=m
> --- linux-2.6.16-rc5-orig/drivers/net/ehea/Kconfig 1969-12-31 16:00:00.000000000 -0800
> +++ kernel/drivers/net/ehea/Kconfig 2006-06-07 07:01:14.637178248 -0700
> @@ -0,0 +1,6 @@
> +config EHEA
> + tristate "eHEA Ethernet support"
> + depends on IBMEBUS
> + ---help---
> + This driver supports the IBM pSeries
> + Ethernet adapter
That last sentence can be on one line, please.
---
~Randy
^ permalink raw reply
* Re: Problem authenticating using WPA with bcm43xx-softmac
From: Larry Finger @ 2006-06-07 18:12 UTC (permalink / raw)
To: Dan Williams; +Cc: Johannes Berg, netdev, Jouni Malinen
In-Reply-To: <1149701352.2625.20.camel@localhost.localdomain>
Dan Williams wrote:
> On Wed, 2006-06-07 at 17:57 +0200, Johannes Berg wrote:
>> On Wed, 2006-06-07 at 17:51 +0200, Johannes Berg wrote:
>>
>>> Well, it should be shown in the 802.11i spec too.
>> I suppose that it is the association request, and needs to contain the
>> RSN described in 7.3.2.25 as per 7.2.3.4 in 802.11i. This is, afaik, the
>> 'generic IE' that is added with the wext. Now, it looks like the RSN
>> isn't included but the WPA2 info or something? Also, the genIE in your
>> log doesn't look correct to me, starting with ffffff?? Jouni, do you
>> have any idea what might be going on?
>
> I believe that wpa_supplicant tells the driver what genie to use through
> the SIOCSIWGENIE wext call. The IEs match between what the driver
> appears to be reporting, and what wpa_supplicant says from the logs.
> wpa_supplicant is almost certainly writing the correct IE to the driver
> through wext, so I think the debug output from softmac must be
> formatting the string incorrectly when printing it out to the logs.
>
> Looking at it further:
>
> struct ieee80211softmac_wpa {
> char *IE;
> int IElen;
> int IEbuflen;
> };
>
> from ieee80211softmac_wx.c: ieee80211softmac_wx_set_genie()
>
> memcpy(mac->wpa.IE, extra, wrqu->data.length);
> dprintk(KERN_INFO PFX "generic IE set to ");
> for (i=0;i<wrqu->data.length;i++)
> dprintk("%.2x", mac->wpa.IE[i]);
> dprintk("\n");
>
> the dprintk code isn't doing the right thing here, given an array of
> bytes. You probably want:
>
> dprintk("%.2hhx", mac->wpa.IE[i]);
>
> (ie, add the "hh" before the x to tell the print that it's a char)
>
That doesn't work - the result is
%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx
I changed the line to cast the output byte as a u8 as follows:
dprintk("%.2x", (u8)mac->wpa.IE[i]);
This produces the line
generic IE set to dd160050f20101000050f20201000050f20201000050f202
This is the WPA IE supplied by wpa_supplicant and it matches the one used in the ndiswrapper case.
One mystery solved, but why doesn't it work?
Johannes - should I submit the patch to fix this printout, or would you like to do it?
Larry
^ permalink raw reply
* Re: [PATCH] pcnet32 driver NAPI support
From: Don Fry @ 2006-06-07 18:20 UTC (permalink / raw)
To: Lennart Sorensen; +Cc: linux-kernel, linux-net, netdev
In-Reply-To: <20060607165225.GB7859@csclub.uwaterloo.ca>
On Wed, Jun 07, 2006 at 12:52:25PM -0400, Lennart Sorensen wrote:
> I have added NAPI support to the pcnet32 driver. This has greatly
> improved the responsiveness on my systems (geode GX1 266MHz) when under
> heavy network load. Without this change the system would become
> unresponsive due to interrupts when flooded with traffic, and eventually
> the watchdog would reboot the system due to the watchdog daemon being
> starved for cpu time. With the patch the system is still useable on a
> serial console, although very slow. Network throughput is also higher
> since more time is spend processing packets and getting them sent out,
> instead of only spending time acknowledging interrupts from incoming
> packets.
>
> Now having never actually done a patch submission to the kernel before,
> I will try and see if I can do it right.
>
> The patch adds a PCNET32_NAPI config option to drivers/net/Kconfig, and
> the appropriate code to support the option to drivers/net/pcnet32.c and
> has been tested on many of my systems (allthough they are allmost all
> identical, and require some extra patches to pcnet32 due to not having
> an EEPROM installed), and on an AT-2700TX.
>
> I have made a diff against 2.6.16.20 and 2.6.17-rc6.
>
> Comments would be very welcome.
I am also working on a NAPI version of the pcnet32 driver for many of
the same reasons, and will compare what you have with my own
implementation. I probably won't be able to do much until Friday.
Just a couple of comments. I am adding netdev@vger.kernel.org to the cc
list, as most network driver discussion is done here rather than lkml.
linux-kernel (and linux-net) should be deleted in future replies.
The 2.6.17-rc6 would be the correct source to patch against. Since this
is an enhancement it will not come out till 2.6.18.
I would not change the driver name from pcnet32 to pcnet32napi, but I
would changes the version from 1.32 to 1.33NAPI or something like that.
Some areas of concern that you may have addressed already, I have not
scanned your changes yet, are what happens if the ring size is changed
without bringing down the interface (via ethtool), or if the loopback
test is run in a similar fashion, or a tx timeout occurs.
The lp->lock MUST be held whenever accessing the csr or bcr registers as
this is a multi-step process, and has been the source of problems in the
past. Even on UP systems.
>
> Signed-off-by: Len Sorensen <lsorense@csclub.uwaterloo.ca>
>
> Len Sorensen
--
Don Fry
brazilnut@us.ibm.com
^ permalink raw reply
* Re: 2.6.17-rc6-mm1
From: Dave Jones @ 2006-06-07 18:21 UTC (permalink / raw)
To: Andrew Morton; +Cc: netdev
In-Reply-To: <20060607104724.c5d3d730.akpm@osdl.org>
On Wed, Jun 07, 2006 at 10:47:24AM -0700, Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.17-rc6/2.6.17-rc6-mm1/
Just spotted this typo in a new option.
Signed-off-by: Dave Jones <davej@redhat.com>
--- rc6mm1/net/Kconfig~ 2006-06-07 14:19:44.000000000 -0400
+++ rc6mm1/net/Kconfig 2006-06-07 14:19:58.000000000 -0400
@@ -227,7 +227,7 @@ config NET_TCPPROBE
depends on INET && EXPERIMENTAL && PROC_FS && KPROBES
---help---
This module allows for capturing the changes to TCP connection
- state in response to incoming patckets. It is used for debugging
+ state in response to incoming packets. It is used for debugging
TCP congestion avoidance modules. If you don't understand
what was just said, you don't need it: say N.
--
http://www.codemonkey.org.uk
^ permalink raw reply
* Re: [PATCH 1/2] e1000: fix netpoll with NAPI
From: Auke Kok @ 2006-06-07 18:25 UTC (permalink / raw)
To: Matt Mackall, Garzik, Jeff
Cc: Neil Horman, Jeff Moyer, Mitch Williams, netdev,
Brandeburg, Jesse, Kok, Auke
In-Reply-To: <20060607164801.GX24227@waste.org>
[-- Attachment #1: Type: text/plain, Size: 1982 bytes --]
Matt Mackall wrote:
> On Wed, Jun 07, 2006 at 11:05:22AM -0400, Neil Horman wrote:
>>>
>>>> Matt, any ideas on this?
>>> Not at the moment.
>> how about this for a solution? It doesn't make netpoll any more robust, but I
>> think in the interests of efficiency it would be fair to require that, when
>> netpolled, a driver must receive frames on the same net device for which it was
>> polled. With this patch we detect that condition and handle it accordingly in
>> e1000_intr. This eliminates the need for us to call the clean_rx method from
>> the poll_controller when napi is configured, instead allowing the poll method to
>> be called from napi_poll, as the netpoll model currently does. This fixes the
>> netdump regression, and eliminates the layering violation and the potential race
>> that we've been discussing. I've just tested it with netdump here and it works
>> quite well.
>>
>> Thoughts appreciated.
>
> This looks pretty reasonable, mostly from the perspective that it
> doesn't put any further ugliness in netpoll. We might want to add a
> comment somewhere in netpoll of the new rule we're now observing.
> I'll let the e1000 guys comment on the particulars of the driver change.
Hi,
we're not too happy with this as it puts a branch right in the regular receive
path. We haven't ran the numbers on it yet but it is likely that this will
lower performance significantly during normal receives for something that is
not common use.
Attached below a (revised) patch that adds proper locking around the rx_clean
to prevent the race.
Cheers,
Auke
---
e1000: fix netpoll with NAPI
This fixes netpoll when using NAPI, and protects against a rare race condition
in the netpoll routine, while staying out of our ways from the normal hotpath.
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
---
e1000_main.c | 17 ++++++++++++++++-
1 file changed, 16 insertions(+), 1 deletion(-)
[-- Attachment #2: e1000_netpoll_napi-r2.patch --]
[-- Type: text/x-patch, Size: 1035 bytes --]
diff --git a/drivers/net/e1000/e1000_main.c b/drivers/net/e1000/e1000_main.c
index bd709e5..876251c 100644
--- a/drivers/net/e1000/e1000_main.c
+++ b/drivers/net/e1000/e1000_main.c
@@ -4584,10 +4584,25 @@ static void
e1000_netpoll(struct net_device *netdev)
{
struct e1000_adapter *adapter = netdev_priv(netdev);
+#ifdef CONFIG_E1000_NAPI
+ int budget = 0;
+
+ disable_irq(adapter->pdev->irq);
+ if (likely(netif_rx_schedule_prep(&adapter->polling_netdev[0]))) {
+ if (spin_trylock(&adapter->tx_queue_lock)) {
+ e1000_clean_tx_irq(adapter, &adapter->tx_ring[0]);
+ spin_unlock(&adapter->tx_queue_lock);
+ }
+ adapter->clean_rx(adapter, adapter->rx_ring,
+ &budget, netdev->weight);
+ clear_bit(__LINK_STATE_RX_SCHED,
+ &adapter->polling_netdev[0].state);
+ }
+#else
+
disable_irq(adapter->pdev->irq);
e1000_intr(adapter->pdev->irq, netdev, NULL);
e1000_clean_tx_irq(adapter, adapter->tx_ring);
-#ifndef CONFIG_E1000_NAPI
adapter->clean_rx(adapter, adapter->rx_ring);
#endif
enable_irq(adapter->pdev->irq);
^ permalink raw reply related
* Re: [PATCH 1/2] e1000: fix netpoll with NAPI
From: Jeff Moyer @ 2006-06-07 18:44 UTC (permalink / raw)
To: Auke Kok
Cc: Matt Mackall, Garzik, Jeff, Neil Horman, Mitch Williams, netdev,
Brandeburg, Jesse, Kok, Auke
In-Reply-To: <44871A19.7080805@intel.com>
==> Regarding Re: [PATCH 1/2] e1000: fix netpoll with NAPI; Auke Kok <auke-jan.h.kok@intel.com> adds:
auke-jan.h.kok> Hi,
auke-jan.h.kok> we're not too happy with this as it puts a branch right in
auke-jan.h.kok> the regular receive path. We haven't ran the numbers on it
auke-jan.h.kok> yet but it is likely that this will lower performance
auke-jan.h.kok> significantly during normal receives for something that is
auke-jan.h.kok> not common use.
auke-jan.h.kok> Attached below a (revised) patch that adds proper locking
auke-jan.h.kok> around the rx_clean to prevent the race.
That patch locks around the tx clean routine. As such, it doesn't prevent
the problem.
> + disable_irq(adapter->pdev->irq);
> + if (likely(netif_rx_schedule_prep(&adapter->polling_netdev[0]))) {
> + if (spin_trylock(&adapter->tx_queue_lock)) {
> + e1000_clean_tx_irq(adapter, &adapter->tx_ring[0]);
> + spin_unlock(&adapter->tx_queue_lock);
> + }
> + adapter->clean_rx(adapter, adapter->rx_ring,
> + &budget, netdev->weight);
> + clear_bit(__LINK_STATE_RX_SCHED,
> + &adapter->polling_netdev[0].state);
-Jeff
^ permalink raw reply
* Updated sysctl documentation take #2
From: Diego Calleja @ 2006-06-07 18:53 UTC (permalink / raw)
To: linux-kernel; +Cc: akpm, netdev, linux-xfs, ecki, lkml
Since people didn't like the "many small files" approach, I've moved
it to directories containing index.txt files:
Documentation/sysctl/vm/index.txt
Documentation/sysctl/net/core/index.txt
Documentation/sysctl/net/unix/index.txt
Documentation/sysctl/net/ipv4/index.txt
Documentation/sysctl/net/ipv4/conf/index.txt
Documentation/sysctl/net/ipv4/route/index.txt
Documentation/sysctl/net/ipv4/neigh/index.txt
and so on.
As previously, this moves all sysctl documentation (including
XFS and network) to Documentation/sysctl/*. The patch is
against linus tree and weights more than 200K in size
and is place at: http://www.terra.es/personal/diegocg/sysctl-docs
Comments?
^ permalink raw reply
* Re: [PATCH 1/2] e1000: fix netpoll with NAPI
From: John W. Linville @ 2006-06-07 18:54 UTC (permalink / raw)
To: Auke Kok
Cc: Matt Mackall, Garzik, Jeff, Neil Horman, Jeff Moyer,
Mitch Williams, netdev, Brandeburg, Jesse, Kok, Auke
In-Reply-To: <44871A19.7080805@intel.com>
On Wed, Jun 07, 2006 at 11:25:29AM -0700, Auke Kok wrote:
> @@ -4584,10 +4584,25 @@ static void
> e1000_netpoll(struct net_device *netdev)
> {
> struct e1000_adapter *adapter = netdev_priv(netdev);
> +#ifdef CONFIG_E1000_NAPI
> + int budget = 0;
> +
> + disable_irq(adapter->pdev->irq);
> + if (likely(netif_rx_schedule_prep(&adapter->polling_netdev[0]))) {
> + if (spin_trylock(&adapter->tx_queue_lock)) {
> + e1000_clean_tx_irq(adapter, &adapter->tx_ring[0]);
> + spin_unlock(&adapter->tx_queue_lock);
> + }
> + adapter->clean_rx(adapter, adapter->rx_ring,
> + &budget, netdev->weight);
> + clear_bit(__LINK_STATE_RX_SCHED,
> + &adapter->polling_netdev[0].state);
> + }
> +#else
> +
> disable_irq(adapter->pdev->irq);
Pedantic objection, but I think this would read easier w/o the extra
newline before disable_irq.
John
--
John W. Linville
linville@tuxdriver.com
^ permalink raw reply
* Re: [PATCH 1/2] e1000: fix netpoll with NAPI
From: Neil Horman @ 2006-06-07 19:18 UTC (permalink / raw)
To: Jeff Moyer
Cc: Auke Kok, Matt Mackall, Garzik, Jeff, Mitch Williams, netdev,
Brandeburg, Jesse, Kok, Auke
In-Reply-To: <x49d5dkvm8p.fsf@segfault.boston.redhat.com>
On Wed, Jun 07, 2006 at 02:44:54PM -0400, Jeff Moyer wrote:
> ==> Regarding Re: [PATCH 1/2] e1000: fix netpoll with NAPI; Auke Kok <auke-jan.h.kok@intel.com> adds:
>
> auke-jan.h.kok> Hi,
>
> auke-jan.h.kok> we're not too happy with this as it puts a branch right in
> auke-jan.h.kok> the regular receive path. We haven't ran the numbers on it
> auke-jan.h.kok> yet but it is likely that this will lower performance
> auke-jan.h.kok> significantly during normal receives for something that is
> auke-jan.h.kok> not common use.
>
> auke-jan.h.kok> Attached below a (revised) patch that adds proper locking
> auke-jan.h.kok> around the rx_clean to prevent the race.
>
> That patch locks around the tx clean routine. As such, it doesn't prevent
> the problem.
>
Further to that, do tests on this if you like, but I would certainly think a
properly formed conditional operation is going to provide better performance
than a spin_lock operation in the receive path. Why not just turn the:
if(netpoll_op)
into an
if(unlikely(netpoll_op))
I expect that will reduce the overhead of the conditional to effectively zero
for the normal receive case. The following patch does that, and I expect you
performance won't suffer at all:
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
e1000_main.c | 54 +++++++++++++++++++++++++++++++++++++++++++++---------
1 files changed, 45 insertions(+), 9 deletions(-)
--- linux-2.6.9/drivers/net/e1000/e1000_main.c.neil 2006-06-06 10:37:42.000000000 -0400
+++ linux-2.6.9/drivers/net/e1000/e1000_main.c 2006-06-07 10:48:22.000000000 -0400
@@ -3207,8 +3207,9 @@ e1000_update_stats(struct e1000_adapter
* @pt_regs: CPU registers structure
**/
+
static irqreturn_t
-e1000_intr(int irq, void *data, struct pt_regs *regs)
+__e1000_intr(int irq, void *data, struct pt_regs *regs, int netpoll_op)
{
struct net_device *netdev = data;
struct e1000_adapter *adapter = netdev_priv(netdev);
@@ -3217,6 +3218,7 @@ e1000_intr(int irq, void *data, struct p
#ifndef CONFIG_E1000_NAPI
int i;
#else
+ struct net_device *dev_to_sched;
/* Interrupt Auto-Mask...upon reading ICR,
* interrupts are masked. No need for the
* IMC write, but it does mean we should
@@ -3255,8 +3257,22 @@ e1000_intr(int irq, void *data, struct p
E1000_WRITE_REG(hw, IMC, ~0);
E1000_WRITE_FLUSH(hw);
}
- if (likely(netif_rx_schedule_prep(&adapter->polling_netdev[0])))
- __netif_rx_schedule(&adapter->polling_netdev[0]);
+
+ /*
+ * netpoll operations, in the interests of efficiency
+ * only do napi polling on the device passed to the
+ * poll_controller. Therefore, if we are preforming
+ * a netpoll operation, then we can't schedule a receive
+ * to one of the dummy net devices that exist for sole
+ * purpose of spreading out rx schedules
+ */
+ if (unlikely(netpoll_op))
+ dev_to_sched = netdev;
+ else
+ dev_to_sched = &adapter->polling_netdev[0];
+
+ if (likely(netif_rx_schedule_prep(dev_to_sched)))
+ __netif_rx_schedule(dev_to_sched);
else
e1000_irq_enable(adapter);
#else
@@ -3288,6 +3304,13 @@ e1000_intr(int irq, void *data, struct p
return IRQ_HANDLED;
}
+static irqreturn_t
+e1000_intr(int irq, void *data, struct pt_regs *regs)
+{
+ return __e1000_intr(irq, data, regs, 0);
+}
+
+
#ifdef CONFIG_E1000_NAPI
/**
* e1000_clean - NAPI Rx polling callback
@@ -3300,7 +3323,6 @@ e1000_clean(struct net_device *poll_dev,
struct e1000_adapter *adapter;
int work_to_do = min(*budget, poll_dev->quota);
int tx_cleaned = 0, i = 0, work_done = 0;
-
/* Must NOT use netdev_priv macro here. */
adapter = poll_dev->priv;
@@ -3308,10 +3330,24 @@ e1000_clean(struct net_device *poll_dev,
if (!netif_carrier_ok(adapter->netdev))
goto quit_polling;
- while (poll_dev != &adapter->polling_netdev[i]) {
- i++;
- if (unlikely(i == adapter->num_rx_queues))
- BUG();
+ /*
+ * only search for a matching polling_netdev in the event
+ * that this isn't a real registered net_device
+ * A real net device can be passed in here in the event
+ * that netdump has been activated (this comes through
+ * netpoll_poll_dev). We detect this by virtue of the
+ * fact that each polling_netdev->priv points to the private
+ * data of its parent (registered) netdev. So if:
+ * poll_dev->priv == netdev_priv(poll_dev), its a real device
+ * otherwise its a polling_netdev.
+ */
+ if (likely(adapter != netdev_priv(poll_dev))) {
+ while (poll_dev != &adapter->polling_netdev[i]) {
+ i++;
+ if (unlikely(i == adapter->num_rx_queues))
+ BUG();
+ }
+
}
if (likely(adapter->num_tx_queues == 1)) {
@@ -4624,7 +4660,7 @@ e1000_netpoll(struct net_device *netdev)
{
struct e1000_adapter *adapter = netdev_priv(netdev);
disable_irq(adapter->pdev->irq);
- e1000_intr(adapter->pdev->irq, netdev, NULL);
+ __e1000_intr(adapter->pdev->irq, netdev, NULL, 1);
e1000_clean_tx_irq(adapter, adapter->tx_ring);
#ifndef CONFIG_E1000_NAPI
adapter->clean_rx(adapter, adapter->rx_ring);
^ permalink raw reply
* RE: kernel panic (on DHCP discover?) in sky2 driver of 2.6.17-rc1
From: Guenther Thomsen @ 2006-06-07 19:33 UTC (permalink / raw)
To: Stephen Hemminger; +Cc: John W. Linville, netdev
I was perhaps a bit quick to declare victory. While the results below stand and the machine survived the last few days (idle), it occurred to me only today, to have a look at the kernel's message buffer, where I found following:
--8<--
sky2 eth0: enabling interface
sky2 eth0: Link is up at 1000 Mbps, full duplex, flow control none
sky2 eth1: enabling interface
sky2 eth1: Link is up at 1000 Mbps, full duplex, flow control none
audit(1149379670.514:3): audit_pid=1915 old=0 by auid=4294967295
<unknown>: hw csum failure.
sky2 eth1: rx error, status 0x7ffc0001 length 444
Call Trace: <ffffffff811de741>{__skb_checksum_complete+76}
<ffffffff812030cb>{__tcp_checksum_complete_user+33}
<ffffffff812080d8>{tcp_rcv_established+817} <ffffffff8120f3ee>{tcp_v4_
do_rcv+43}
<ffffffff811da2ee>{sk_wait_data+203} <ffffffff811fe5a8>{tcp_prequeue_p
rocess+121}
<ffffffff811ff71d>{tcp_recvmsg+1104} <ffffffff811d9712>{sock_common_re
cvmsg+48}
<ffffffff811d7d4f>{do_sock_read+209} <ffffffff811d7e7e>{sock_aio_read+
83}
<ffffffff811e2ca1>{dev_queue_xmit+0} <ffffffff8106dce9>{do_sync_read+1
99}
<ffffffff8103d699>{remove_wait_queue+18} <ffffffff8103d530>{autoremove
_wake_function+0}
<ffffffff8106df83>{vfs_read+228} <ffffffff8106ea12>{sys_read+69}
<ffffffff81009b0d>{tracesys+209}
<unknown>: hw csum failure.
sky2 eth1: rx error, status 0x7ffc0001 length 444
Call Trace: <ffffffff811de741>{__skb_checksum_complete+76}
<ffffffff812030cb>{__tcp_checksum_complete_user+33}
<ffffffff812080d8>{tcp_rcv_established+817} <ffffffff8120f3ee>{tcp_v4_
do_rcv+43}
<ffffffff811da2ee>{sk_wait_data+203} <ffffffff811fe5a8>{tcp_prequeue_p
rocess+121}
<ffffffff811ff71d>{tcp_recvmsg+1104} <ffffffff811d9712>{sock_common_re
cvmsg+48}
<ffffffff811d7696>{alloc_sock_iocb+20} <ffffffff811d7d4f>{do_sock_read
+209}
<ffffffff811d7e7e>{sock_aio_read+83} <ffffffff8106dce9>{do_sync_read+1
99}
<ffffffff8103d699>{remove_wait_queue+18} <ffffffff8103d530>{autoremove
_wake_function+0}
<ffffffff8106df83>{vfs_read+228} <ffffffff8106ea12>{sys_read+69}
<ffffffff81009b0d>{tracesys+209}
<unknown>: hw csum failure.
sky2 eth1: rx error, status 0x7ffc0001 length 444
Call Trace: <ffffffff811de741>{__skb_checksum_complete+76}
<ffffffff812030cb>{__tcp_checksum_complete_user+33}
<ffffffff812080d8>{tcp_rcv_established+817} <ffffffff8120f3ee>{tcp_v4_
do_rcv+43}
<ffffffff811da2ee>{sk_wait_data+203} <ffffffff811fe5a8>{tcp_prequeue_p
rocess+121}
<ffffffff811ff71d>{tcp_recvmsg+1104} <ffffffff811d9712>{sock_common_re
cvmsg+48}
<ffffffff811d7d4f>{do_sock_read+209} <ffffffff811d7e7e>{sock_aio_read+
83}
<ffffffff8106dce9>{do_sync_read+199} <ffffffff8103d699>{remove_wait_qu
eue+18}
<ffffffff8103d530>{autoremove_wake_function+0} <ffffffff8102f1a0>{curr
ent_kernel_time+13}
<ffffffff8106df83>{vfs_read+228} <ffffffff8106ea12>{sys_read+69}
<ffffffff81009b0d>{tracesys+209}
sky2 eth0: rx error, status 0x7ffc0001 length 444
sky2 eth0: rx error, status 0x7ffc0001 length 444
sky2 eth1: rx error, status 0x7ffc0001 length 444
sky2 eth1: rx error, status 0x7ffc0001 length 444
sky2 eth1: rx error, status 0x7ffc0001 length 444
-->8--
Looks, like we're almost, but not quite there yet.
cheers
Guenther
-----Original Message-----
From: Guenther Thomsen
Sent: Saturday, June 03, 2006 9:06 PM
To: 'Stephen Hemminger'
Cc: John W. Linville; netdev@vger.kernel.org
Subject: RE: kernel panic (on DHCP discover?) in sky2 driver of
2.6.17-rc1
I received the hardware back and took the opportunity to test with 2.6.17-rc5-git11. So far I did only little tests (ttcp on both interfaces in, out or mixed with some 10e6 packets), but it looks good. No errors (well, 16 overruns in 76574513 packets) and line rate (about 111MB/s) on both channels simultaneously. Hurray!
Thanks a lot for you continued efforts.
Guenther
-----Original Message-----
From: Stephen Hemminger [mailto:shemminger@osdl.org]
Sent: Tuesday, May 16, 2006 12:12 PM
To: Guenther Thomsen
Cc: John W. Linville; netdev@vger.kernel.org
Subject: Re: kernel panic (on DHCP discover?) in sky2 driver of
2.6.17-rc1
Could you try the 2.6.17-rc4 version with this patch. It turns out the board
seems to give out of order status responses.
Ignore the vendor sk98lin driver, when I try the stock version it spends it's
life resetting itself because it sets up PCI bus wrong. If I fix that, it spends
it's time getting confused because it can't handle intermixed status reports
properly (checksum et all is per port not per board).
drivers/net/sky2.c | 28 +++++++++++++++++++++-------
1 files changed, 21 insertions(+), 7 deletions(-)
792547bc5e8e4f7d5a1070a168056f429635c254
diff --git a/drivers/net/sky2.c b/drivers/net/sky2.c
index ffd267f..11e7914 100644
--- a/drivers/net/sky2.c
+++ b/drivers/net/sky2.c
@@ -1020,8 +1020,27 @@ static int sky2_up(struct net_device *de
struct sky2_hw *hw = sky2->hw;
unsigned port = sky2->port;
u32 ramsize, rxspace, imask;
- int err = -ENOMEM;
+ int cap, err;
+ struct net_device *otherdev = hw->dev[sky2->port^1];
+ /*
+ * Reduce split transactions (and turn off) rx checksums to
+ * prevent problems with dual ports.
+ */
+ if (otherdev && netif_running(otherdev) &&
+ (cap = pci_find_capability(hw->pdev, PCI_CAP_ID_PCIX))) {
+ struct sky2_port *osky2 = netdev_priv(otherdev);
+ u16 cmd;
+
+ cmd = sky2_pci_read16(hw, cap + PCI_X_CMD);
+ cmd &= ~PCI_X_CMD_MAX_SPLIT;
+ sky2_pci_write16(hw, cap + PCI_X_CMD, cmd);
+
+ sky2->rx_csum = 0;
+ osky2->rx_csum = 0;
+ }
+
+ err = -ENOMEM;
if (netif_msg_ifup(sky2))
printk(KERN_INFO PFX "%s: enabling interface\n", dev->name);
@@ -3067,12 +3086,7 @@ static __devinit struct net_device *sky2
sky2->duplex = -1;
sky2->speed = -1;
sky2->advertising = sky2_supported_modes(hw);
-
- /* Receive checksum disabled for Yukon XL
- * because of observed problems with incorrect
- * values when multiple packets are received in one interrupt
- */
- sky2->rx_csum = (hw->chip_id != CHIP_ID_YUKON_XL);
+ sky2->rx_csum = 1;
spin_lock_init(&sky2->phy_lock);
sky2->tx_pending = TX_DEF_PENDING;
--
1.2.4
^ permalink raw reply related
* Re: [PATCH] pcnet32 driver NAPI support
From: Lennart Sorensen @ 2006-06-07 19:34 UTC (permalink / raw)
To: Don Fry; +Cc: linux-kernel, linux-net, netdev
In-Reply-To: <20060607182040.GA12748@us.ibm.com>
On Wed, Jun 07, 2006 at 11:20:40AM -0700, Don Fry wrote:
> I am also working on a NAPI version of the pcnet32 driver for many of
> the same reasons, and will compare what you have with my own
> implementation. I probably won't be able to do much until Friday.
>
> Just a couple of comments. I am adding netdev@vger.kernel.org to the cc
> list, as most network driver discussion is done here rather than lkml.
> linux-kernel (and linux-net) should be deleted in future replies.
I must have picked the wrong place to cc.
> The 2.6.17-rc6 would be the correct source to patch against. Since this
> is an enhancement it will not come out till 2.6.18.
I thought so. That is why I did it against both 2.6.17-rc6 and 2.6.16
(since I use it with 2.6.16).
> I would not change the driver name from pcnet32 to pcnet32napi, but I
> would changes the version from 1.32 to 1.33NAPI or something like that.
Hmm, perhaps. I just wanted something that made it obvious in dmesg
which driver I was running. I see tulip actually does put it in the
version instead. I don't remember where I got the driver name change
idea from.
> Some areas of concern that you may have addressed already, I have not
> scanned your changes yet, are what happens if the ring size is changed
> without bringing down the interface (via ethtool), or if the loopback
> test is run in a similar fashion, or a tx timeout occurs.
The same thing as if it was done before enabling napi. From a few
messages I have seen, it doesn't work right now, and it won't work any
better with my changes. I have never tried changing the ring size on
the fly, so I don't know.
It appears that the port is stopped before the ring size change is done,
although I can't really tell how it handles things if the queue is not
empty when it stops the port. Does it try to handle anything left in
the ring first or does it just toss those packets? (That I would
consider wrong).
> The lp->lock MUST be held whenever accessing the csr or bcr registers as
> this is a multi-step process, and has been the source of problems in the
> past. Even on UP systems.
Hmm, I just followed what appeared to be in pcnet32_rx and how tulip and
a few other drivers had done their napi conversions. It certainly works
for me the way I did it. Haven't seen any lockups yet. I do see that I
am not holding the lock when I acknowledge IRQs in pcnet32_poll, which
pcnet32_rx doesn't need to worry about since it is called from the
interrupt handler which already holds the lock. That should be fixed
then.
So I can do:
// Clear RX interrupts
spin_lock(&lp->lock);
lp->a.write_csr (ioaddr, 0, 0x1400);
spin_unlock(&lp->lock);
That part seems simple enough to protect.
Is this safe without holding the lock?
} while(lp->a.read_csr (ioaddr, 0) & 0x1400);
Not sure how to wrap a lock around that one without holding the lock for
way too long.
perhaps:
spin_lock(&lp->lock);
state=lp->a.read_csr (ioaddr, 0) & 0x1400;
spin_unlock(&lp->lock);
} while(state);
Does that seem more reasonable?
Len Sorensen
^ permalink raw reply
* Re: Problem authenticating using WPA with bcm43xx-softmac
From: Dan Williams @ 2006-06-07 19:36 UTC (permalink / raw)
To: Larry Finger; +Cc: Johannes Berg, netdev, Jouni Malinen
In-Reply-To: <44871723.3040803@lwfinger.net>
On Wed, 2006-06-07 at 13:12 -0500, Larry Finger wrote:
> Dan Williams wrote:
> >
> > from ieee80211softmac_wx.c: ieee80211softmac_wx_set_genie()
> >
> > memcpy(mac->wpa.IE, extra, wrqu->data.length);
> > dprintk(KERN_INFO PFX "generic IE set to ");
> > for (i=0;i<wrqu->data.length;i++)
> > dprintk("%.2x", mac->wpa.IE[i]);
> > dprintk("\n");
> >
> > the dprintk code isn't doing the right thing here, given an array of
> > bytes. You probably want:
> >
> > dprintk("%.2hhx", mac->wpa.IE[i]);
> >
> > (ie, add the "hh" before the x to tell the print that it's a char)
> >
> That doesn't work - the result is
Weird, does the kernel not do something that fprintf() _does_ do here?
I tested with a short C program that mimics the behavior of this chunk
of code, and "%.2x" didn't work, but "%.2hhx" certainly did. "hh" is
supposed to mean "A following integer conversion corresponds to a signed
char or unsigned char argument". The original conversion was converting
stuff from la-la land after the first 4 bytes (in both softmac and my
testcase), and "hh" solved it in the testcase. I did try casting to
char, but glibc pretty much ignored that.
Dan
> %hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx%hx
>
> I changed the line to cast the output byte as a u8 as follows:
>
> dprintk("%.2x", (u8)mac->wpa.IE[i]);
>
> This produces the line
>
> generic IE set to dd160050f20101000050f20201000050f20201000050f202
>
> This is the WPA IE supplied by wpa_supplicant and it matches the one used in the ndiswrapper case.
> One mystery solved, but why doesn't it work?
>
> Johannes - should I submit the patch to fix this printout, or would you like to do it?
>
> Larry
^ permalink raw reply
* Re: kernel panic (on DHCP discover?) in sky2 driver of 2.6.17-rc1
From: Stephen Hemminger @ 2006-06-07 19:44 UTC (permalink / raw)
To: Guenther Thomsen; +Cc: John W. Linville, netdev
In-Reply-To: <CECD6E8A589E8447BC6E836C8369AFF506828728@us-email.terastack.bluearc.com>
On Wed, 7 Jun 2006 12:33:21 -0700
"Guenther Thomsen" <GThomsen@bluearc.com> wrote:
> I was perhaps a bit quick to declare victory. While the results below stand and the machine survived the last few days (idle), it occurred to me only today, to have a look at the kernel's message buffer, where I found following:
> --8<--
> sky2 eth0: enabling interface
> sky2 eth0: Link is up at 1000 Mbps, full duplex, flow control none
> sky2 eth1: enabling interface
> sky2 eth1: Link is up at 1000 Mbps, full duplex, flow control none
> audit(1149379670.514:3): audit_pid=1915 old=0 by auid=4294967295
> <unknown>: hw csum failure.
> sky2 eth1: rx error, status 0x7ffc0001 length 444
>
> Call Trace: <ffffffff811de741>{__skb_checksum_complete+76}
> <ffffffff812030cb>{__tcp_checksum_complete_user+33}
> <ffffffff812080d8>{tcp_rcv_established+817} <ffffffff8120f3ee>{tcp_v4_
> do_rcv+43}
> <ffffffff811da2ee>{sk_wait_data+203} <ffffffff811fe5a8>{tcp_prequeue_p
> rocess+121}
> <ffffffff811ff71d>{tcp_recvmsg+1104} <ffffffff811d9712>{sock_common_re
> cvmsg+48}
> <ffffffff811d7d4f>{do_sock_read+209} <ffffffff811d7e7e>{sock_aio_read+
> 83}
> <ffffffff811e2ca1>{dev_queue_xmit+0} <ffffffff8106dce9>{do_sync_read+1
> 99}
> <ffffffff8103d699>{remove_wait_queue+18} <ffffffff8103d530>{autoremove
> _wake_function+0}
> <ffffffff8106df83>{vfs_read+228} <ffffffff8106ea12>{sys_read+69}
> <ffffffff81009b0d>{tracesys+209}
> <unknown>: hw csum failure.
> sky2 eth1: rx error, status 0x7ffc0001 length 444
Different problem, I have seen it before. Basically if the receiver gets overloaded, the
packet FIFO gets full. The driver needs to have some kind of recovery logic for this;
probably just shutting down the receiver and restarting.
^ permalink raw reply
* Re: Problem authenticating using WPA with bcm43xx-softmac
From: Larry Finger @ 2006-06-07 19:46 UTC (permalink / raw)
To: Dan Williams; +Cc: netdev
In-Reply-To: <1149708987.2625.114.camel@localhost.localdomain>
Dan Williams wrote:
> Weird, does the kernel not do something that fprintf() _does_ do here?
> I tested with a short C program that mimics the behavior of this chunk
> of code, and "%.2x" didn't work, but "%.2hhx" certainly did. "hh" is
> supposed to mean "A following integer conversion corresponds to a signed
> char or unsigned char argument". The original conversion was converting
> stuff from la-la land after the first 4 bytes (in both softmac and my
> testcase), and "hh" solved it in the testcase. I did try casting to
> char, but glibc pretty much ignored that.
I also found that casting with char gave the same result as no cast, but that a u8 cast worked. I
guess printk is _almost_ like fprintf.
Larry
^ permalink raw reply
* [PATCH v2 0/2][RFC] iWARP Core Support
From: Steve Wise @ 2006-06-07 20:06 UTC (permalink / raw)
To: rdreier, mshefty; +Cc: linux-kernel, netdev, openib-general
This patchset defines the modifications to the Linux infiniband subsystem
to support iWARP devices. We're submitting it for review now with the
goal for inclusion in the 2.6.19 kernel. This code has gone through
several reviews in the openib-general list. Now we are submitting it
for external review by the linux community.
This StGIT patchset is cloned from Roland Dreier's infiniband.git
for-2.6.18 branch. The patchset consists of 2 patches:
1 - New iWARP CM implementation.
2 - Core changes to support iWARP.
I believe I've addressed all the round 1 review comments. Details of
the changes are tracked in each patch comment.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
^ permalink raw reply
* [PATCH v2 1/2] iWARP Connection Manager.
From: Steve Wise @ 2006-06-07 20:06 UTC (permalink / raw)
To: rdreier, mshefty; +Cc: linux-kernel, netdev, openib-general
In-Reply-To: <20060607200600.9003.56328.stgit@stevo-desktop>
This patch provides the new files implementing the iWARP Connection
Manager.
Review Changes:
- sizeof -> sizeof()
- removed printks
- removed TT debug code
- cleaned up lock/unlock around switch statements.
- waitqueue -> completion for destroy path.
---
drivers/infiniband/core/iwcm.c | 877 ++++++++++++++++++++++++++++++++++++++++
include/rdma/iw_cm.h | 254 ++++++++++++
include/rdma/iw_cm_private.h | 62 +++
3 files changed, 1193 insertions(+), 0 deletions(-)
diff --git a/drivers/infiniband/core/iwcm.c b/drivers/infiniband/core/iwcm.c
new file mode 100644
index 0000000..994bc79
--- /dev/null
+++ b/drivers/infiniband/core/iwcm.c
@@ -0,0 +1,877 @@
+/*
+ * Copyright (c) 2004, 2005 Intel Corporation. All rights reserved.
+ * Copyright (c) 2004 Topspin Corporation. All rights reserved.
+ * Copyright (c) 2004, 2005 Voltaire Corporation. All rights reserved.
+ * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved.
+ * Copyright (c) 2005 Network Appliance, Inc. 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/dma-mapping.h>
+#include <linux/err.h>
+#include <linux/idr.h>
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/rbtree.h>
+#include <linux/spinlock.h>
+#include <linux/workqueue.h>
+#include <linux/completion.h>
+#include <rdma/iw_cm.h>
+#include <rdma/iw_cm_private.h>
+#include <rdma/ib_addr.h>
+
+MODULE_AUTHOR("Tom Tucker");
+MODULE_DESCRIPTION("iWARP CM");
+MODULE_LICENSE("Dual BSD/GPL");
+
+static struct workqueue_struct *iwcm_wq;
+struct iwcm_work {
+ struct work_struct work;
+ struct iwcm_id_private *cm_id;
+ struct list_head list;
+ struct iw_cm_event event;
+};
+
+/*
+ * Release a reference on cm_id. If the last reference is being removed
+ * and iw_destroy_cm_id is waiting, wake up the waiting thread.
+ */
+static int iwcm_deref_id(struct iwcm_id_private *cm_id_priv)
+{
+ int ret = 0;
+
+ BUG_ON(atomic_read(&cm_id_priv->refcount)==0);
+ if (atomic_dec_and_test(&cm_id_priv->refcount)) {
+ BUG_ON(!list_empty(&cm_id_priv->work_list));
+ if (waitqueue_active(&cm_id_priv->destroy_comp.wait)) {
+ BUG_ON(cm_id_priv->state != IW_CM_STATE_DESTROYING);
+ BUG_ON(test_bit(IWCM_F_CALLBACK_DESTROY,
+ &cm_id_priv->flags));
+ ret = 1;
+ }
+ complete(&cm_id_priv->destroy_comp);
+ }
+
+ return ret;
+}
+
+static void add_ref(struct iw_cm_id *cm_id)
+{
+ struct iwcm_id_private *cm_id_priv;
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ atomic_inc(&cm_id_priv->refcount);
+}
+
+static void rem_ref(struct iw_cm_id *cm_id)
+{
+ struct iwcm_id_private *cm_id_priv;
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ iwcm_deref_id(cm_id_priv);
+}
+
+static void cm_event_handler(struct iw_cm_id *cm_id, struct iw_cm_event *event);
+
+struct iw_cm_id *iw_create_cm_id(struct ib_device *device,
+ iw_cm_handler cm_handler,
+ void *context)
+{
+ struct iwcm_id_private *cm_id_priv;
+
+ cm_id_priv = kzalloc(sizeof(*cm_id_priv), GFP_KERNEL);
+ if (!cm_id_priv)
+ return ERR_PTR(-ENOMEM);
+
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ cm_id_priv->id.device = device;
+ cm_id_priv->id.cm_handler = cm_handler;
+ cm_id_priv->id.context = context;
+ cm_id_priv->id.event_handler = cm_event_handler;
+ cm_id_priv->id.add_ref = add_ref;
+ cm_id_priv->id.rem_ref = rem_ref;
+ spin_lock_init(&cm_id_priv->lock);
+ atomic_set(&cm_id_priv->refcount, 1);
+ init_waitqueue_head(&cm_id_priv->connect_wait);
+ init_completion(&cm_id_priv->destroy_comp);
+ INIT_LIST_HEAD(&cm_id_priv->work_list);
+
+ return &cm_id_priv->id;
+}
+EXPORT_SYMBOL(iw_create_cm_id);
+
+
+static int iwcm_modify_qp_err(struct ib_qp *qp)
+{
+ struct ib_qp_attr qp_attr;
+
+ if (!qp)
+ return -EINVAL;
+
+ qp_attr.qp_state = IB_QPS_ERR;
+ return ib_modify_qp(qp, &qp_attr, IB_QP_STATE);
+}
+
+/*
+ * This is really the RDMAC CLOSING state. It is most similar to the
+ * IB SQD QP state.
+ */
+static int iwcm_modify_qp_sqd(struct ib_qp *qp)
+{
+ struct ib_qp_attr qp_attr;
+
+ BUG_ON(qp == NULL);
+ qp_attr.qp_state = IB_QPS_SQD;
+ return ib_modify_qp(qp, &qp_attr, IB_QP_STATE);
+}
+
+/*
+ * CM_ID <-- CLOSING
+ *
+ * Block if a passive or active connection is currenlty being processed. Then
+ * process the event as follows:
+ * - If we are ESTABLISHED, move to CLOSING and modify the QP state
+ * based on the abrupt flag
+ * - If the connection is already in the CLOSING or IDLE state, the peer is
+ * disconnecting concurrently with us and we've already seen the
+ * DISCONNECT event -- ignore the request and return 0
+ * - Disconnect on a listening endpoint returns -EINVAL
+ */
+int iw_cm_disconnect(struct iw_cm_id *cm_id, int abrupt)
+{
+ struct iwcm_id_private *cm_id_priv;
+ unsigned long flags;
+ int ret = 0;
+ struct ib_qp *qp = NULL;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ /* Wait if we're currently in a connect or accept downcall */
+ wait_event(cm_id_priv->connect_wait,
+ !test_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags));
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ switch (cm_id_priv->state) {
+ case IW_CM_STATE_ESTABLISHED:
+ cm_id_priv->state = IW_CM_STATE_CLOSING;
+
+ /* QP could be <nul> for user-mode client */
+ if (cm_id_priv->qp)
+ qp = cm_id_priv->qp;
+ else
+ ret = -EINVAL;
+ break;
+ case IW_CM_STATE_LISTEN:
+ ret = -EINVAL;
+ break;
+ case IW_CM_STATE_CLOSING:
+ /* remote peer closed first */
+ case IW_CM_STATE_IDLE:
+ /* accept or connect returned !0 */
+ break;
+ case IW_CM_STATE_CONN_RECV:
+ /*
+ * App called disconnect before/without calling accept after
+ * connect_request event delivered.
+ */
+ break;
+ case IW_CM_STATE_CONN_SENT:
+ /* Can only get here if wait above fails */
+ default:
+ BUG_ON(1);
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ if (qp) {
+ if (abrupt)
+ ret = iwcm_modify_qp_err(qp);
+ else
+ ret = iwcm_modify_qp_sqd(qp);
+
+ /*
+ * If both sides are disconnecting the QP could
+ * already be in ERR or SQD states
+ */
+ ret = 0;
+ }
+
+ return ret;
+}
+EXPORT_SYMBOL(iw_cm_disconnect);
+
+/*
+ * CM_ID <-- DESTROYING
+ *
+ * Clean up all resources associated with the connection and release
+ * the initial reference taken by iw_create_cm_id.
+ */
+static void destroy_cm_id(struct iw_cm_id *cm_id)
+{
+ struct iwcm_id_private *cm_id_priv;
+ unsigned long flags;
+ int ret;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ /* Wait if we're currently in a connect or accept downcall. A
+ * listening endpoint should never block here. */
+ wait_event(cm_id_priv->connect_wait,
+ !test_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags));
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ switch (cm_id_priv->state) {
+ case IW_CM_STATE_LISTEN:
+ cm_id_priv->state = IW_CM_STATE_DESTROYING;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ /* destroy the listening endpoint */
+ ret = cm_id->device->iwcm->destroy_listen(cm_id);
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ break;
+ case IW_CM_STATE_ESTABLISHED:
+ cm_id_priv->state = IW_CM_STATE_DESTROYING;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ /* Abrupt close of the connection */
+ (void)iwcm_modify_qp_err(cm_id_priv->qp);
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ break;
+ case IW_CM_STATE_IDLE:
+ case IW_CM_STATE_CLOSING:
+ cm_id_priv->state = IW_CM_STATE_DESTROYING;
+ break;
+ case IW_CM_STATE_CONN_RECV:
+ /*
+ * App called destroy before/without calling accept after
+ * receiving connection request event notification.
+ */
+ cm_id_priv->state = IW_CM_STATE_DESTROYING;
+ break;
+ case IW_CM_STATE_CONN_SENT:
+ case IW_CM_STATE_DESTROYING:
+ default:
+ BUG_ON(1);
+ break;
+ }
+ if (cm_id_priv->qp) {
+ cm_id_priv->id.device->iwcm->rem_ref(cm_id_priv->qp);
+ cm_id_priv->qp = NULL;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ (void)iwcm_deref_id(cm_id_priv);
+}
+
+/*
+ * This function is only called by the application thread and cannot
+ * be called by the event thread. The function will wait for all
+ * references to be released on the cm_id and then kfree the cm_id
+ * object.
+ */
+void iw_destroy_cm_id(struct iw_cm_id *cm_id)
+{
+ struct iwcm_id_private *cm_id_priv;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ BUG_ON(test_bit(IWCM_F_CALLBACK_DESTROY, &cm_id_priv->flags));
+
+ destroy_cm_id(cm_id);
+
+ wait_for_completion(&cm_id_priv->destroy_comp);
+
+ kfree(cm_id_priv);
+}
+EXPORT_SYMBOL(iw_destroy_cm_id);
+
+/*
+ * CM_ID <-- LISTEN
+ *
+ * Start listening for connect requests. Generates one CONNECT_REQUEST
+ * event for each inbound connect request.
+ */
+int iw_cm_listen(struct iw_cm_id *cm_id, int backlog)
+{
+ struct iwcm_id_private *cm_id_priv;
+ unsigned long flags;
+ int ret = 0;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ switch (cm_id_priv->state) {
+ case IW_CM_STATE_IDLE:
+ cm_id_priv->state = IW_CM_STATE_LISTEN;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ ret = cm_id->device->iwcm->create_listen(cm_id, backlog);
+ if (ret)
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ break;
+ default:
+ ret = -EINVAL;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ return ret;
+}
+EXPORT_SYMBOL(iw_cm_listen);
+
+/*
+ * CM_ID <-- IDLE
+ *
+ * Rejects an inbound connection request. No events are generated.
+ */
+int iw_cm_reject(struct iw_cm_id *cm_id,
+ const void *private_data,
+ u8 private_data_len)
+{
+ struct iwcm_id_private *cm_id_priv;
+ unsigned long flags;
+ int ret;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (cm_id_priv->state != IW_CM_STATE_CONN_RECV) {
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ wake_up_all(&cm_id_priv->connect_wait);
+ return -EINVAL;
+ }
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ ret = cm_id->device->iwcm->reject(cm_id, private_data,
+ private_data_len);
+
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ wake_up_all(&cm_id_priv->connect_wait);
+
+ return ret;
+}
+EXPORT_SYMBOL(iw_cm_reject);
+
+/*
+ * CM_ID <-- ESTABLISHED
+ *
+ * Accepts an inbound connection request and generates an ESTABLISHED
+ * event. Callers of iw_cm_disconnect and iw_destroy_cm_id will block
+ * until the ESTABLISHED event is received from the provider.
+ */
+int iw_cm_accept(struct iw_cm_id *cm_id,
+ struct iw_cm_conn_param *iw_param)
+{
+ struct iwcm_id_private *cm_id_priv;
+ struct ib_qp *qp;
+ unsigned long flags;
+ int ret;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (cm_id_priv->state != IW_CM_STATE_CONN_RECV) {
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ wake_up_all(&cm_id_priv->connect_wait);
+ return -EINVAL;
+ }
+ /* Get the ib_qp given the QPN */
+ qp = cm_id->device->iwcm->get_qp(cm_id->device, iw_param->qpn);
+ if (!qp) {
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ return -EINVAL;
+ }
+ cm_id->device->iwcm->add_ref(qp);
+ cm_id_priv->qp = qp;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ ret = cm_id->device->iwcm->accept(cm_id, iw_param);
+ if (ret) {
+ /* An error on accept precludes provider events */
+ BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_RECV);
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (cm_id_priv->qp) {
+ cm_id->device->iwcm->rem_ref(qp);
+ cm_id_priv->qp = NULL;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ printk("Accept failed, ret=%d\n", ret);
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ wake_up_all(&cm_id_priv->connect_wait);
+ }
+
+ return ret;
+}
+EXPORT_SYMBOL(iw_cm_accept);
+
+/*
+ * Active Side: CM_ID <-- CONN_SENT
+ *
+ * If successful, results in the generation of a CONNECT_REPLY
+ * event. iw_cm_disconnect and iw_cm_destroy will block until the
+ * CONNECT_REPLY event is received from the provider.
+ */
+int iw_cm_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param)
+{
+ struct iwcm_id_private *cm_id_priv;
+ int ret = 0;
+ unsigned long flags;
+ struct ib_qp *qp;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (cm_id_priv->state != IW_CM_STATE_IDLE) {
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ wake_up_all(&cm_id_priv->connect_wait);
+ return -EINVAL;
+ }
+
+ /* Get the ib_qp given the QPN */
+ qp = cm_id->device->iwcm->get_qp(cm_id->device, iw_param->qpn);
+ if (!qp) {
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ return -EINVAL;
+ }
+ cm_id->device->iwcm->add_ref(qp);
+ cm_id_priv->qp = qp;
+ cm_id_priv->state = IW_CM_STATE_CONN_SENT;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ ret = cm_id->device->iwcm->connect(cm_id, iw_param);
+ if (ret) {
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (cm_id_priv->qp) {
+ cm_id->device->iwcm->rem_ref(qp);
+ cm_id_priv->qp = NULL;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_SENT);
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ printk("Connect failed, ret=%d\n", ret);
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ wake_up_all(&cm_id_priv->connect_wait);
+ }
+
+ return ret;
+}
+EXPORT_SYMBOL(iw_cm_connect);
+
+/*
+ * Passive Side: new CM_ID <-- CONN_RECV
+ *
+ * Handles an inbound connect request. The function creates a new
+ * iw_cm_id to represent the new connection and inherits the client
+ * callback function and other attributes from the listening parent.
+ *
+ * The work item contains a pointer to the listen_cm_id and the event. The
+ * listen_cm_id contains the client cm_handler, context and
+ * device. These are copied when the device is cloned. The event
+ * contains the new four tuple.
+ *
+ * An error on the child should not affect the parent, so this
+ * function does not return a value.
+ */
+static void cm_conn_req_handler(struct iwcm_id_private *listen_id_priv,
+ struct iw_cm_event *iw_event)
+{
+ unsigned long flags;
+ struct iw_cm_id *cm_id;
+ struct iwcm_id_private *cm_id_priv;
+ int ret;
+
+ /* The provider should never generate a connection request
+ * event with a bad status.
+ */
+ BUG_ON(iw_event->status);
+
+ /* We could be destroying the listening id. If so, ignore this
+ * upcall. */
+ spin_lock_irqsave(&listen_id_priv->lock, flags);
+ if (listen_id_priv->state != IW_CM_STATE_LISTEN) {
+ spin_unlock_irqrestore(&listen_id_priv->lock, flags);
+ return;
+ }
+ spin_unlock_irqrestore(&listen_id_priv->lock, flags);
+
+ cm_id = iw_create_cm_id(listen_id_priv->id.device,
+ listen_id_priv->id.cm_handler,
+ listen_id_priv->id.context);
+ /* If the cm_id could not be created, ignore the request */
+ if (IS_ERR(cm_id))
+ return;
+
+ cm_id->provider_data = iw_event->provider_data;
+ cm_id->local_addr = iw_event->local_addr;
+ cm_id->remote_addr = iw_event->remote_addr;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ cm_id_priv->state = IW_CM_STATE_CONN_RECV;
+
+ /* Call the client CM handler */
+ ret = cm_id->cm_handler(cm_id, iw_event);
+ if (ret) {
+ set_bit(IWCM_F_CALLBACK_DESTROY, &cm_id_priv->flags);
+ destroy_cm_id(cm_id);
+ if (atomic_read(&cm_id_priv->refcount)==0)
+ kfree(cm_id);
+ }
+}
+
+/*
+ * Passive Side: CM_ID <-- ESTABLISHED
+ *
+ * The provider generated an ESTABLISHED event which means that
+ * the MPA negotion has completed successfully and we are now in MPA
+ * FPDU mode.
+ *
+ * This event can only be received in the CONN_RECV state. If the
+ * remote peer closed, the ESTABLISHED event would be received followed
+ * by the CLOSE event. If the app closes, it will block until we wake
+ * it up after processing this event.
+ */
+static int cm_conn_est_handler(struct iwcm_id_private *cm_id_priv,
+ struct iw_cm_event *iw_event)
+{
+ unsigned long flags;
+ int ret = 0;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+
+ /* We clear the CONNECT_WAIT bit here to allow the callback
+ * function to call iw_cm_disconnect. Calling iw_destroy_cm_id
+ * from a callback handler is not allowed */
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_RECV);
+ cm_id_priv->state = IW_CM_STATE_ESTABLISHED;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, iw_event);
+ wake_up_all(&cm_id_priv->connect_wait);
+
+ return ret;
+}
+
+/*
+ * Active Side: CM_ID <-- ESTABLISHED
+ *
+ * The app has called connect and is waiting for the established event to
+ * post it's requests to the server. This event will wake up anyone
+ * blocked in iw_cm_disconnect or iw_destroy_id.
+ */
+static int cm_conn_rep_handler(struct iwcm_id_private *cm_id_priv,
+ struct iw_cm_event *iw_event)
+{
+ unsigned long flags;
+ int ret = 0;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ /* Clear the connect wait bit so a callback function calling
+ * iw_cm_disconnect will not wait and deadlock this thread */
+ clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags);
+ BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_SENT);
+ if (iw_event->status == IW_CM_EVENT_STATUS_ACCEPTED) {
+ cm_id_priv->id.local_addr = iw_event->local_addr;
+ cm_id_priv->id.remote_addr = iw_event->remote_addr;
+ cm_id_priv->state = IW_CM_STATE_ESTABLISHED;
+ } else {
+ /* REJECTED or RESET */
+ cm_id_priv->id.device->iwcm->rem_ref(cm_id_priv->qp);
+ cm_id_priv->qp = NULL;
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, iw_event);
+
+ /* Wake up waiters on connect complete */
+ wake_up_all(&cm_id_priv->connect_wait);
+
+ return ret;
+}
+
+/*
+ * CM_ID <-- CLOSING
+ *
+ * If in the ESTABLISHED state, move to CLOSING.
+ */
+static void cm_disconnect_handler(struct iwcm_id_private *cm_id_priv,
+ struct iw_cm_event *iw_event)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (cm_id_priv->state == IW_CM_STATE_ESTABLISHED)
+ cm_id_priv->state = IW_CM_STATE_CLOSING;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+}
+
+/*
+ * CM_ID <-- IDLE
+ *
+ * If in the ESTBLISHED or CLOSING states, the QP will have have been
+ * moved by the provider to the ERR state. Disassociate the CM_ID from
+ * the QP, move to IDLE, and remove the 'connected' reference.
+ *
+ * If in some other state, the cm_id was destroyed asynchronously.
+ * This is the last reference that will result in waking up
+ * the app thread blocked in iw_destroy_cm_id.
+ */
+static int cm_close_handler(struct iwcm_id_private *cm_id_priv,
+ struct iw_cm_event *iw_event)
+{
+ unsigned long flags;
+ int ret = 0;
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+
+ if (cm_id_priv->qp) {
+ cm_id_priv->id.device->iwcm->rem_ref(cm_id_priv->qp);
+ cm_id_priv->qp = NULL;
+ }
+ switch (cm_id_priv->state) {
+ case IW_CM_STATE_ESTABLISHED:
+ case IW_CM_STATE_CLOSING:
+ cm_id_priv->state = IW_CM_STATE_IDLE;
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, iw_event);
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ break;
+ case IW_CM_STATE_DESTROYING:
+ break;
+ default:
+ BUG_ON(1);
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ return ret;
+}
+
+static int process_event(struct iwcm_id_private *cm_id_priv,
+ struct iw_cm_event *iw_event)
+{
+ int ret = 0;
+
+ switch (iw_event->event) {
+ case IW_CM_EVENT_CONNECT_REQUEST:
+ cm_conn_req_handler(cm_id_priv, iw_event);
+ break;
+ case IW_CM_EVENT_CONNECT_REPLY:
+ ret = cm_conn_rep_handler(cm_id_priv, iw_event);
+ break;
+ case IW_CM_EVENT_ESTABLISHED:
+ ret = cm_conn_est_handler(cm_id_priv, iw_event);
+ break;
+ case IW_CM_EVENT_DISCONNECT:
+ cm_disconnect_handler(cm_id_priv, iw_event);
+ break;
+ case IW_CM_EVENT_CLOSE:
+ ret = cm_close_handler(cm_id_priv, iw_event);
+ break;
+ default:
+ BUG_ON(1);
+ }
+
+ return ret;
+}
+
+/*
+ * Process events on the work_list for the cm_id. If the callback
+ * function requests that the cm_id be deleted, a flag is set in the
+ * cm_id flags to indicate that when the last reference is
+ * removed, the cm_id is to be destroyed. This is necessary to
+ * distinguish between an object that will be destroyed by the app
+ * thread asleep on the destroy_comp list vs. an object destroyed
+ * here synchronously when the last reference is removed.
+ */
+static void cm_work_handler(void *arg)
+{
+ struct iwcm_work *work = (struct iwcm_work*)arg;
+ struct iwcm_id_private *cm_id_priv = work->cm_id;
+ unsigned long flags;
+ int empty;
+ int ret = 0;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ empty = list_empty(&cm_id_priv->work_list);
+ while (!empty) {
+ work = list_entry(cm_id_priv->work_list.next,
+ struct iwcm_work, list);
+ list_del_init(&work->list);
+ empty = list_empty(&cm_id_priv->work_list);
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+
+ ret = process_event(cm_id_priv, &work->event);
+ kfree(work);
+ if (ret) {
+ set_bit(IWCM_F_CALLBACK_DESTROY, &cm_id_priv->flags);
+ destroy_cm_id(&cm_id_priv->id);
+ }
+ BUG_ON(atomic_read(&cm_id_priv->refcount)==0);
+ if (iwcm_deref_id(cm_id_priv))
+ return;
+
+ if (atomic_read(&cm_id_priv->refcount)==0 &&
+ test_bit(IWCM_F_CALLBACK_DESTROY, &cm_id_priv->flags)) {
+ kfree(cm_id_priv);
+ return;
+ }
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+}
+
+/*
+ * This function is called on interrupt context. Schedule events on
+ * the iwcm_wq thread to allow callback functions to downcall into
+ * the CM and/or block. Events are queued to a per-CM_ID
+ * work_list. If this is the first event on the work_list, the work
+ * element is also queued on the iwcm_wq thread.
+ *
+ * Each event holds a reference on the cm_id. Until the last posted
+ * event has been delivered and processed, the cm_id cannot be
+ * deleted.
+ */
+static void cm_event_handler(struct iw_cm_id *cm_id,
+ struct iw_cm_event *iw_event)
+{
+ struct iwcm_work *work;
+ struct iwcm_id_private *cm_id_priv;
+ unsigned long flags;
+
+ work = kmalloc(sizeof(*work), GFP_ATOMIC);
+ if (!work)
+ return;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ atomic_inc(&cm_id_priv->refcount);
+
+ INIT_WORK(&work->work, cm_work_handler, work);
+ work->cm_id = cm_id_priv;
+ work->event = *iw_event;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ if (list_empty(&cm_id_priv->work_list)) {
+ list_add_tail(&work->list, &cm_id_priv->work_list);
+ queue_work(iwcm_wq, &work->work);
+ } else
+ list_add_tail(&work->list, &cm_id_priv->work_list);
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+}
+
+static int iwcm_init_qp_init_attr(struct iwcm_id_private *cm_id_priv,
+ struct ib_qp_attr *qp_attr,
+ int *qp_attr_mask)
+{
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ switch (cm_id_priv->state) {
+ case IW_CM_STATE_IDLE:
+ case IW_CM_STATE_CONN_SENT:
+ case IW_CM_STATE_CONN_RECV:
+ case IW_CM_STATE_ESTABLISHED:
+ *qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS;
+ qp_attr->qp_access_flags = IB_ACCESS_LOCAL_WRITE |
+ IB_ACCESS_REMOTE_WRITE|
+ IB_ACCESS_REMOTE_READ;
+ ret = 0;
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ return ret;
+}
+
+static int iwcm_init_qp_rts_attr(struct iwcm_id_private *cm_id_priv,
+ struct ib_qp_attr *qp_attr,
+ int *qp_attr_mask)
+{
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&cm_id_priv->lock, flags);
+ switch (cm_id_priv->state) {
+ case IW_CM_STATE_IDLE:
+ case IW_CM_STATE_CONN_SENT:
+ case IW_CM_STATE_CONN_RECV:
+ case IW_CM_STATE_ESTABLISHED:
+ *qp_attr_mask = 0;
+ ret = 0;
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }
+ spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+ return ret;
+}
+
+int iw_cm_init_qp_attr(struct iw_cm_id *cm_id,
+ struct ib_qp_attr *qp_attr,
+ int *qp_attr_mask)
+{
+ struct iwcm_id_private *cm_id_priv;
+ int ret;
+
+ cm_id_priv = container_of(cm_id, struct iwcm_id_private, id);
+ switch (qp_attr->qp_state) {
+ case IB_QPS_INIT:
+ case IB_QPS_RTR:
+ ret = iwcm_init_qp_init_attr(cm_id_priv,
+ qp_attr, qp_attr_mask);
+ break;
+ case IB_QPS_RTS:
+ ret = iwcm_init_qp_rts_attr(cm_id_priv,
+ qp_attr, qp_attr_mask);
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }
+ return ret;
+}
+EXPORT_SYMBOL(iw_cm_init_qp_attr);
+
+static int __init iw_cm_init(void)
+{
+ iwcm_wq = create_singlethread_workqueue("iw_cm_wq");
+ if (!iwcm_wq)
+ return -ENOMEM;
+
+ return 0;
+}
+
+static void __exit iw_cm_cleanup(void)
+{
+ destroy_workqueue(iwcm_wq);
+}
+
+module_init(iw_cm_init);
+module_exit(iw_cm_cleanup);
diff --git a/include/rdma/iw_cm.h b/include/rdma/iw_cm.h
new file mode 100644
index 0000000..0752a94
--- /dev/null
+++ b/include/rdma/iw_cm.h
@@ -0,0 +1,254 @@
+/*
+ * Copyright (c) 2005 Network Appliance, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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.
+ */
+#if !defined(IW_CM_H)
+#define IW_CM_H
+
+#include <linux/in.h>
+#include <rdma/ib_cm.h>
+
+struct iw_cm_id;
+
+enum iw_cm_event_type {
+ IW_CM_EVENT_CONNECT_REQUEST = 1, /* connect request received */
+ IW_CM_EVENT_CONNECT_REPLY, /* reply from active connect request */
+ IW_CM_EVENT_ESTABLISHED, /* passive side accept successful */
+ IW_CM_EVENT_DISCONNECT, /* orderly shutdown */
+ IW_CM_EVENT_CLOSE /* close complete */
+};
+enum iw_cm_event_status {
+ IW_CM_EVENT_STATUS_OK = 0, /* request successful */
+ IW_CM_EVENT_STATUS_ACCEPTED = 0, /* connect request accepted */
+ IW_CM_EVENT_STATUS_REJECTED, /* connect request rejected */
+ IW_CM_EVENT_STATUS_TIMEOUT, /* the operation timed out */
+ IW_CM_EVENT_STATUS_RESET, /* reset from remote peer */
+ IW_CM_EVENT_STATUS_EINVAL, /* asynchronous failure for bad parm */
+};
+struct iw_cm_event {
+ enum iw_cm_event_type event;
+ enum iw_cm_event_status status;
+ struct sockaddr_in local_addr;
+ struct sockaddr_in remote_addr;
+ void *private_data;
+ u8 private_data_len;
+ void* provider_data;
+};
+
+/**
+ * iw_cm_handler - Function to be called by the IW CM when delivering events
+ * to the client.
+ *
+ * @cm_id: The IW CM identifier associated with the event.
+ * @event: Pointer to the event structure.
+ */
+typedef int (*iw_cm_handler)(struct iw_cm_id *cm_id,
+ struct iw_cm_event *event);
+
+/**
+ * iw_event_handler - Function called by the provider when delivering provider
+ * events to the IW CM.
+ *
+ * @cm_id: The IW CM identifier associated with the event.
+ * @event: Pointer to the event structure.
+ */
+typedef void (*iw_event_handler)(struct iw_cm_id *cm_id,
+ struct iw_cm_event *event);
+struct iw_cm_id {
+ iw_cm_handler cm_handler; /* client callback function */
+ void *context; /* client cb context */
+ struct ib_device *device;
+ struct sockaddr_in local_addr;
+ struct sockaddr_in remote_addr;
+ void *provider_data; /* provider private data */
+ iw_event_handler event_handler; /* cb for provider
+ events */
+ /* Used by provider to add and remove refs on IW cm_id */
+ void (*add_ref)(struct iw_cm_id *);
+ void (*rem_ref)(struct iw_cm_id *);
+};
+
+struct iw_cm_conn_param {
+ const void *private_data;
+ u16 private_data_len;
+ u32 ord;
+ u32 ird;
+ u32 qpn;
+};
+
+struct iw_cm_verbs {
+ void (*add_ref)(struct ib_qp *qp);
+
+ void (*rem_ref)(struct ib_qp *qp);
+
+ struct ib_qp * (*get_qp)(struct ib_device *device,
+ int qpn);
+
+ int (*connect)(struct iw_cm_id *cm_id,
+ struct iw_cm_conn_param *conn_param);
+
+ int (*accept)(struct iw_cm_id *cm_id,
+ struct iw_cm_conn_param *conn_param);
+
+ int (*reject)(struct iw_cm_id *cm_id,
+ const void *pdata, u8 pdata_len);
+
+ int (*create_listen)(struct iw_cm_id *cm_id,
+ int backlog);
+
+ int (*destroy_listen)(struct iw_cm_id *cm_id);
+};
+
+/**
+ * iw_create_cm_id - Create an IW CM identifier.
+ *
+ * @device: The IB device on which to create the IW CM identier.
+ * @event_handler: User callback invoked to report events associated with the
+ * returned IW CM identifier.
+ * @context: User specified context associated with the id.
+ */
+struct iw_cm_id *iw_create_cm_id(struct ib_device *device,
+ iw_cm_handler cm_handler, void *context);
+
+/**
+ * iw_destroy_cm_id - Destroy an IW CM identifier.
+ *
+ * @cm_id: The previously created IW CM identifier to destroy.
+ *
+ * The client can assume that no events will be delivered for the CM ID after
+ * this function returns.
+ */
+void iw_destroy_cm_id(struct iw_cm_id *cm_id);
+
+/**
+ * iw_cm_bind_qp - Unbind the specified IW CM identifier and QP
+ *
+ * @cm_id: The IW CM idenfier to unbind from the QP.
+ * @qp: The QP
+ *
+ * This is called by the provider when destroying the QP to ensure
+ * that any references held by the IWCM are released. It may also
+ * be called by the IWCM when destroying a CM_ID to that any
+ * references held by the provider are released.
+ */
+void iw_cm_unbind_qp(struct iw_cm_id *cm_id, struct ib_qp *qp);
+
+/**
+ * iw_cm_get_qp - Return the ib_qp associated with a QPN
+ *
+ * @ib_device: The IB device
+ * @qpn: The queue pair number
+ */
+struct ib_qp *iw_cm_get_qp(struct ib_device *device, int qpn);
+
+/**
+ * iw_cm_listen - Listen for incoming connection requests on the
+ * specified IW CM id.
+ *
+ * @cm_id: The IW CM identifier.
+ * @backlog: The maximum number of outstanding un-accepted inbound listen
+ * requests to queue.
+ *
+ * The source address and port number are specified in the IW CM identifier
+ * structure.
+ */
+int iw_cm_listen(struct iw_cm_id *cm_id, int backlog);
+
+/**
+ * iw_cm_accept - Called to accept an incoming connect request.
+ *
+ * @cm_id: The IW CM identifier associated with the connection request.
+ * @iw_param: Pointer to a structure containing connection establishment
+ * parameters.
+ *
+ * The specified cm_id will have been provided in the event data for a
+ * CONNECT_REQUEST event. Subsequent events related to this connection will be
+ * delivered to the specified IW CM identifier prior and may occur prior to
+ * the return of this function. If this function returns a non-zero value, the
+ * client can assume that no events will be delivered to the specified IW CM
+ * identifier.
+ */
+int iw_cm_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param);
+
+/**
+ * iw_cm_reject - Reject an incoming connection request.
+ *
+ * @cm_id: Connection identifier associated with the request.
+ * @private_daa: Pointer to data to deliver to the remote peer as part of the
+ * reject message.
+ * @private_data_len: The number of bytes in the private_data parameter.
+ *
+ * The client can assume that no events will be delivered to the specified IW
+ * CM identifier following the return of this function. The private_data
+ * buffer is available for reuse when this function returns.
+ */
+int iw_cm_reject(struct iw_cm_id *cm_id, const void *private_data,
+ u8 private_data_len);
+
+/**
+ * iw_cm_connect - Called to request a connection to a remote peer.
+ *
+ * @cm_id: The IW CM identifier for the connection.
+ * @iw_param: Pointer to a structure containing connection establishment
+ * parameters.
+ *
+ * Events may be delivered to the specified IW CM identifier prior to the
+ * return of this function. If this function returns a non-zero value, the
+ * client can assume that no events will be delivered to the specified IW CM
+ * identifier.
+ */
+int iw_cm_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param);
+
+/**
+ * iw_cm_disconnect - Close the specified connection.
+ *
+ * @cm_id: The IW CM identifier to close.
+ * @abrupt: If 0, the connection will be closed gracefully, otherwise, the
+ * connection will be reset.
+ *
+ * The IW CM identifier is still active until the IW_CM_EVENT_CLOSE event is
+ * delivered.
+ */
+int iw_cm_disconnect(struct iw_cm_id *cm_id, int abrupt);
+
+/**
+ * iw_cm_init_qp_attr - Called to initialize the attributes of the QP
+ * associated with a IW CM identifier.
+ *
+ * @cm_id: The IW CM identifier associated with the QP
+ * @qp_attr: Pointer to the QP attributes structure.
+ * @qp_attr_mask: Pointer to a bit vector specifying which QP attributes are
+ * valid.
+ */
+int iw_cm_init_qp_attr(struct iw_cm_id *cm_id, struct ib_qp_attr *qp_attr,
+ int *qp_attr_mask);
+
+#endif /* IW_CM_H */
diff --git a/include/rdma/iw_cm_private.h b/include/rdma/iw_cm_private.h
new file mode 100644
index 0000000..aba8cb2
--- /dev/null
+++ b/include/rdma/iw_cm_private.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2005 Network Appliance, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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.
+ */
+#if !defined(IW_CM_PRIVATE_H)
+#define IW_CM_PRIVATE_H
+
+#include <rdma/iw_cm.h>
+
+enum iw_cm_state {
+ IW_CM_STATE_IDLE, /* unbound, inactive */
+ IW_CM_STATE_LISTEN, /* listen waiting for connect */
+ IW_CM_STATE_CONN_RECV, /* inbound waiting for user accept */
+ IW_CM_STATE_CONN_SENT, /* outbound waiting for peer accept */
+ IW_CM_STATE_ESTABLISHED, /* established */
+ IW_CM_STATE_CLOSING, /* disconnect */
+ IW_CM_STATE_DESTROYING /* object being deleted */
+};
+
+struct iwcm_id_private {
+ struct iw_cm_id id;
+ enum iw_cm_state state;
+ unsigned long flags;
+ struct ib_qp *qp;
+ struct completion destroy_comp;
+ wait_queue_head_t connect_wait;
+ struct list_head work_list;
+ spinlock_t lock;
+ atomic_t refcount;
+};
+#define IWCM_F_CALLBACK_DESTROY 1
+#define IWCM_F_CONNECT_WAIT 2
+
+#endif /* IW_CM_PRIVATE_H */
^ permalink raw reply related
* [PATCH v2 2/2] iWARP Core Changes.
From: Steve Wise @ 2006-06-07 20:06 UTC (permalink / raw)
To: rdreier, mshefty; +Cc: linux-kernel, netdev, openib-general
In-Reply-To: <20060607200600.9003.56328.stgit@stevo-desktop>
This patch contains modifications to the existing rdma header files,
core files, drivers, and ulp files to support iWARP.
Review updates:
- copy_addr() -> rdma_copy_addr()
- dst_dev_addr param in rdma_copy_addr to const.
- various spacing nits with recasting
- include linux/inetdevice.h to get ip_dev_find() prototype.
---
drivers/infiniband/core/Makefile | 4
drivers/infiniband/core/addr.c | 19 +
drivers/infiniband/core/cache.c | 8 -
drivers/infiniband/core/cm.c | 3
drivers/infiniband/core/cma.c | 353 +++++++++++++++++++++++---
drivers/infiniband/core/device.c | 6
drivers/infiniband/core/mad.c | 11 +
drivers/infiniband/core/sa_query.c | 5
drivers/infiniband/core/smi.c | 18 +
drivers/infiniband/core/sysfs.c | 18 +
drivers/infiniband/core/ucm.c | 5
drivers/infiniband/core/user_mad.c | 9 -
drivers/infiniband/hw/ipath/ipath_verbs.c | 2
drivers/infiniband/hw/mthca/mthca_provider.c | 2
drivers/infiniband/ulp/ipoib/ipoib_main.c | 8 +
drivers/infiniband/ulp/srp/ib_srp.c | 2
include/rdma/ib_addr.h | 15 +
include/rdma/ib_verbs.h | 39 +++
18 files changed, 435 insertions(+), 92 deletions(-)
diff --git a/drivers/infiniband/core/Makefile b/drivers/infiniband/core/Makefile
index 68e73ec..163d991 100644
--- a/drivers/infiniband/core/Makefile
+++ b/drivers/infiniband/core/Makefile
@@ -1,7 +1,7 @@
infiniband-$(CONFIG_INFINIBAND_ADDR_TRANS) := ib_addr.o rdma_cm.o
obj-$(CONFIG_INFINIBAND) += ib_core.o ib_mad.o ib_sa.o \
- ib_cm.o $(infiniband-y)
+ ib_cm.o iw_cm.o $(infiniband-y)
obj-$(CONFIG_INFINIBAND_USER_MAD) += ib_umad.o
obj-$(CONFIG_INFINIBAND_USER_ACCESS) += ib_uverbs.o ib_ucm.o
@@ -14,6 +14,8 @@ ib_sa-y := sa_query.o
ib_cm-y := cm.o
+iw_cm-y := iwcm.o
+
rdma_cm-y := cma.o
ib_addr-y := addr.o
diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c
index d294bbc..83f84ef 100644
--- a/drivers/infiniband/core/addr.c
+++ b/drivers/infiniband/core/addr.c
@@ -32,6 +32,7 @@ #include <linux/mutex.h>
#include <linux/inetdevice.h>
#include <linux/workqueue.h>
#include <linux/if_arp.h>
+#include <linux/inetdevice.h>
#include <net/arp.h>
#include <net/neighbour.h>
#include <net/route.h>
@@ -60,12 +61,15 @@ static LIST_HEAD(req_list);
static DECLARE_WORK(work, process_req, NULL);
static struct workqueue_struct *addr_wq;
-static int copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
- unsigned char *dst_dev_addr)
+int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
+ const unsigned char *dst_dev_addr)
{
switch (dev->type) {
case ARPHRD_INFINIBAND:
- dev_addr->dev_type = IB_NODE_CA;
+ dev_addr->dev_type = RDMA_NODE_IB_CA;
+ break;
+ case ARPHRD_ETHER:
+ dev_addr->dev_type = RDMA_NODE_RNIC;
break;
default:
return -EADDRNOTAVAIL;
@@ -77,6 +81,7 @@ static int copy_addr(struct rdma_dev_add
memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
return 0;
}
+EXPORT_SYMBOL(rdma_copy_addr);
int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
{
@@ -88,7 +93,7 @@ int rdma_translate_ip(struct sockaddr *a
if (!dev)
return -EADDRNOTAVAIL;
- ret = copy_addr(dev_addr, dev, NULL);
+ ret = rdma_copy_addr(dev_addr, dev, NULL);
dev_put(dev);
return ret;
}
@@ -160,7 +165,7 @@ static int addr_resolve_remote(struct so
/* If the device does ARP internally, return 'done' */
if (rt->idev->dev->flags & IFF_NOARP) {
- copy_addr(addr, rt->idev->dev, NULL);
+ rdma_copy_addr(addr, rt->idev->dev, NULL);
goto put;
}
@@ -180,7 +185,7 @@ static int addr_resolve_remote(struct so
src_in->sin_addr.s_addr = rt->rt_src;
}
- ret = copy_addr(addr, neigh->dev, neigh->ha);
+ ret = rdma_copy_addr(addr, neigh->dev, neigh->ha);
release:
neigh_release(neigh);
put:
@@ -244,7 +249,7 @@ static int addr_resolve_local(struct soc
if (ZERONET(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);
+ ret = rdma_copy_addr(addr, dev, dev->dev_addr);
} else if (LOOPBACK(src_ip)) {
ret = rdma_translate_ip((struct sockaddr *)dst_in, addr);
if (!ret)
diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c
index e05ca2c..061858c 100644
--- a/drivers/infiniband/core/cache.c
+++ b/drivers/infiniband/core/cache.c
@@ -32,13 +32,12 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: cache.c 1349 2004-12-16 21:09:43Z roland $
+ * $Id: cache.c 6885 2006-05-03 18:22:02Z sean.hefty $
*/
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/slab.h>
-#include <linux/sched.h> /* INIT_WORK, schedule_work(), flush_scheduled_work() */
#include <rdma/ib_cache.h>
@@ -62,12 +61,13 @@ struct ib_update_work {
static inline int start_port(struct ib_device *device)
{
- return device->node_type == IB_NODE_SWITCH ? 0 : 1;
+ return (device->node_type == RDMA_NODE_IB_SWITCH) ? 0 : 1;
}
static inline int end_port(struct ib_device *device)
{
- return device->node_type == IB_NODE_SWITCH ? 0 : device->phys_port_cnt;
+ return (device->node_type == RDMA_NODE_IB_SWITCH) ?
+ 0 : device->phys_port_cnt;
}
int ib_get_cached_gid(struct ib_device *device,
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 1c7463b..cf43ccb 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -3253,6 +3253,9 @@ static void cm_add_one(struct ib_device
int ret;
u8 i;
+ if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+ return;
+
cm_dev = kmalloc(sizeof(*cm_dev) + sizeof(*port) *
device->phys_port_cnt, GFP_KERNEL);
if (!cm_dev)
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index 94555d2..414600c 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -35,6 +35,7 @@ #include <linux/in6.h>
#include <linux/mutex.h>
#include <linux/random.h>
#include <linux/idr.h>
+#include <linux/inetdevice.h>
#include <net/tcp.h>
@@ -43,6 +44,7 @@ #include <rdma/rdma_cm_ib.h>
#include <rdma/ib_cache.h>
#include <rdma/ib_cm.h>
#include <rdma/ib_sa.h>
+#include <rdma/iw_cm.h>
MODULE_AUTHOR("Sean Hefty");
MODULE_DESCRIPTION("Generic RDMA CM Agent");
@@ -124,6 +126,7 @@ struct rdma_id_private {
int query_id;
union {
struct ib_cm_id *ib;
+ struct iw_cm_id *iw;
} cm_id;
u32 seq_num;
@@ -259,13 +262,23 @@ static void cma_detach_from_dev(struct r
id_priv->cma_dev = NULL;
}
-static int cma_acquire_ib_dev(struct rdma_id_private *id_priv)
+static int cma_acquire_dev(struct rdma_id_private *id_priv)
{
+ enum rdma_node_type dev_type = id_priv->id.route.addr.dev_addr.dev_type;
struct cma_device *cma_dev;
union ib_gid *gid;
int ret = -ENODEV;
- gid = ib_addr_get_sgid(&id_priv->id.route.addr.dev_addr);
+ switch (rdma_node_get_transport(dev_type)) {
+ case RDMA_TRANSPORT_IB:
+ gid = ib_addr_get_sgid(&id_priv->id.route.addr.dev_addr);
+ break;
+ case RDMA_TRANSPORT_IWARP:
+ gid = iw_addr_get_sgid(&id_priv->id.route.addr.dev_addr);
+ break;
+ default:
+ return -ENODEV;
+ }
mutex_lock(&lock);
list_for_each_entry(cma_dev, &dev_list, list) {
@@ -280,16 +293,6 @@ static int cma_acquire_ib_dev(struct rdm
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))
@@ -347,6 +350,16 @@ static int cma_init_ib_qp(struct rdma_id
IB_QP_PKEY_INDEX | IB_QP_PORT);
}
+static int cma_init_iw_qp(struct rdma_id_private *id_priv, struct ib_qp *qp)
+{
+ struct ib_qp_attr qp_attr;
+
+ qp_attr.qp_state = IB_QPS_INIT;
+ qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE;
+
+ return ib_modify_qp(qp, &qp_attr, IB_QP_STATE | IB_QP_ACCESS_FLAGS);
+}
+
int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd,
struct ib_qp_init_attr *qp_init_attr)
{
@@ -362,10 +375,13 @@ int rdma_create_qp(struct rdma_cm_id *id
if (IS_ERR(qp))
return PTR_ERR(qp);
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
ret = cma_init_ib_qp(id_priv, qp);
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = cma_init_iw_qp(id_priv, qp);
+ break;
default:
ret = -ENOSYS;
break;
@@ -451,13 +467,17 @@ int rdma_init_qp_attr(struct rdma_cm_id
int ret;
id_priv = container_of(id, struct rdma_id_private, id);
- switch (id_priv->id.device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id_priv->id.device->node_type)) {
+ case RDMA_TRANSPORT_IB:
ret = ib_cm_init_qp_attr(id_priv->cm_id.ib, qp_attr,
qp_attr_mask);
if (qp_attr->qp_state == IB_QPS_RTR)
qp_attr->rq_psn = id_priv->seq_num;
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = iw_cm_init_qp_attr(id_priv->cm_id.iw, qp_attr,
+ qp_attr_mask);
+ break;
default:
ret = -ENOSYS;
break;
@@ -590,8 +610,8 @@ static int cma_notify_user(struct rdma_i
static void cma_cancel_route(struct rdma_id_private *id_priv)
{
- switch (id_priv->id.device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id_priv->id.device->node_type)) {
+ case RDMA_TRANSPORT_IB:
if (id_priv->query)
ib_sa_cancel_query(id_priv->query_id, id_priv->query);
break;
@@ -611,11 +631,15 @@ static void cma_destroy_listen(struct rd
cma_exch(id_priv, CMA_DESTROYING);
if (id_priv->cma_dev) {
- switch (id_priv->id.device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id_priv->id.device->node_type)) {
+ case RDMA_TRANSPORT_IB:
if (id_priv->cm_id.ib && !IS_ERR(id_priv->cm_id.ib))
ib_destroy_cm_id(id_priv->cm_id.ib);
break;
+ case RDMA_TRANSPORT_IWARP:
+ if (id_priv->cm_id.iw && !IS_ERR(id_priv->cm_id.iw))
+ iw_destroy_cm_id(id_priv->cm_id.iw);
+ break;
default:
break;
}
@@ -690,11 +714,15 @@ void rdma_destroy_id(struct rdma_cm_id *
cma_cancel_operation(id_priv, state);
if (id_priv->cma_dev) {
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
if (id_priv->cm_id.ib && !IS_ERR(id_priv->cm_id.ib))
ib_destroy_cm_id(id_priv->cm_id.ib);
break;
+ case RDMA_TRANSPORT_IWARP:
+ if (id_priv->cm_id.iw && !IS_ERR(id_priv->cm_id.iw))
+ iw_destroy_cm_id(id_priv->cm_id.iw);
+ break;
default:
break;
}
@@ -868,7 +896,7 @@ static struct rdma_id_private *cma_new_i
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;
+ rt->addr.dev_addr.dev_type = RDMA_NODE_IB_CA;
id_priv = container_of(id, struct rdma_id_private, id);
id_priv->state = CMA_CONNECT;
@@ -897,7 +925,7 @@ static int cma_req_handler(struct ib_cm_
}
atomic_inc(&conn_id->dev_remove);
- ret = cma_acquire_ib_dev(conn_id);
+ ret = cma_acquire_dev(conn_id);
if (ret) {
ret = -ENODEV;
cma_release_remove(conn_id);
@@ -981,6 +1009,123 @@ static void cma_set_compare_data(enum rd
}
}
+static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event)
+{
+ struct rdma_id_private *id_priv = iw_id->context;
+ enum rdma_cm_event_type event = 0;
+ struct sockaddr_in *sin;
+ int ret = 0;
+
+ atomic_inc(&id_priv->dev_remove);
+
+ switch (iw_event->event) {
+ case IW_CM_EVENT_CLOSE:
+ event = RDMA_CM_EVENT_DISCONNECTED;
+ break;
+ case IW_CM_EVENT_CONNECT_REPLY:
+ sin = (struct sockaddr_in *) &id_priv->id.route.addr.src_addr;
+ *sin = iw_event->local_addr;
+ sin = (struct sockaddr_in *) &id_priv->id.route.addr.dst_addr;
+ *sin = iw_event->remote_addr;
+ if (iw_event->status)
+ event = RDMA_CM_EVENT_REJECTED;
+ else
+ event = RDMA_CM_EVENT_ESTABLISHED;
+ break;
+ case IW_CM_EVENT_ESTABLISHED:
+ event = RDMA_CM_EVENT_ESTABLISHED;
+ break;
+ default:
+ BUG_ON(1);
+ }
+
+ ret = cma_notify_user(id_priv, event, iw_event->status,
+ iw_event->private_data,
+ iw_event->private_data_len);
+ if (ret) {
+ /* Destroy the CM ID by returning a non-zero value. */
+ id_priv->cm_id.iw = NULL;
+ cma_exch(id_priv, CMA_DESTROYING);
+ cma_release_remove(id_priv);
+ rdma_destroy_id(&id_priv->id);
+ return ret;
+ }
+
+ cma_release_remove(id_priv);
+ return ret;
+}
+
+static int iw_conn_req_handler(struct iw_cm_id *cm_id,
+ struct iw_cm_event *iw_event)
+{
+ struct rdma_cm_id *new_cm_id;
+ struct rdma_id_private *listen_id, *conn_id;
+ struct sockaddr_in *sin;
+ struct net_device *dev;
+ int ret;
+
+ listen_id = cm_id->context;
+ atomic_inc(&listen_id->dev_remove);
+ if (!cma_comp(listen_id, CMA_LISTEN)) {
+ ret = -ECONNABORTED;
+ goto out;
+ }
+
+ /* Create a new RDMA id for the new IW CM ID */
+ new_cm_id = rdma_create_id(listen_id->id.event_handler,
+ listen_id->id.context,
+ RDMA_PS_TCP);
+ if (!new_cm_id) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ conn_id = container_of(new_cm_id, struct rdma_id_private, id);
+ atomic_inc(&conn_id->dev_remove);
+ conn_id->state = CMA_CONNECT;
+
+ dev = ip_dev_find(iw_event->local_addr.sin_addr.s_addr);
+ if (!dev) {
+ ret = -EADDRNOTAVAIL;
+ rdma_destroy_id(new_cm_id);
+ goto out;
+ }
+ ret = rdma_copy_addr(&conn_id->id.route.addr.dev_addr, dev, NULL);
+ if (ret) {
+ rdma_destroy_id(new_cm_id);
+ goto out;
+ }
+
+ ret = cma_acquire_dev(conn_id);
+ if (ret) {
+ rdma_destroy_id(new_cm_id);
+ goto out;
+ }
+
+ conn_id->cm_id.iw = cm_id;
+ cm_id->context = conn_id;
+ cm_id->cm_handler = cma_iw_handler;
+
+ sin = (struct sockaddr_in *) &new_cm_id->route.addr.src_addr;
+ *sin = iw_event->local_addr;
+ sin = (struct sockaddr_in *) &new_cm_id->route.addr.dst_addr;
+ *sin = iw_event->remote_addr;
+
+ ret = cma_notify_user(conn_id, RDMA_CM_EVENT_CONNECT_REQUEST, 0,
+ iw_event->private_data,
+ iw_event->private_data_len);
+ if (ret) {
+ /* User wants to destroy the CM ID */
+ conn_id->cm_id.iw = 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 int cma_ib_listen(struct rdma_id_private *id_priv)
{
struct ib_cm_compare_data compare_data;
@@ -1010,6 +1155,30 @@ static int cma_ib_listen(struct rdma_id_
return ret;
}
+static int cma_iw_listen(struct rdma_id_private *id_priv, int backlog)
+{
+ int ret;
+ struct sockaddr_in *sin;
+
+ id_priv->cm_id.iw = iw_create_cm_id(id_priv->id.device,
+ iw_conn_req_handler,
+ id_priv);
+ if (IS_ERR(id_priv->cm_id.iw))
+ return PTR_ERR(id_priv->cm_id.iw);
+
+ sin = (struct sockaddr_in *) &id_priv->id.route.addr.src_addr;
+ id_priv->cm_id.iw->local_addr = *sin;
+
+ ret = iw_cm_listen(id_priv->cm_id.iw, backlog);
+
+ if (ret) {
+ iw_destroy_cm_id(id_priv->cm_id.iw);
+ id_priv->cm_id.iw = NULL;
+ }
+
+ return ret;
+}
+
static int cma_listen_handler(struct rdma_cm_id *id,
struct rdma_cm_event *event)
{
@@ -1085,12 +1254,17 @@ int rdma_listen(struct rdma_cm_id *id, i
return -EINVAL;
if (id->device) {
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
ret = cma_ib_listen(id_priv);
if (ret)
goto err;
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = cma_iw_listen(id_priv, backlog);
+ if (ret)
+ goto err;
+ break;
default:
ret = -ENOSYS;
goto err;
@@ -1229,6 +1403,23 @@ err:
}
EXPORT_SYMBOL(rdma_set_ib_paths);
+static int cma_resolve_iw_route(struct rdma_id_private *id_priv, int timeout_ms)
+{
+ struct cma_work *work;
+
+ work = kzalloc(sizeof *work, GFP_KERNEL);
+ if (!work)
+ return -ENOMEM;
+
+ work->id = id_priv;
+ INIT_WORK(&work->work, cma_work_handler, work);
+ work->old_state = CMA_ROUTE_QUERY;
+ work->new_state = CMA_ROUTE_RESOLVED;
+ work->event.event = RDMA_CM_EVENT_ROUTE_RESOLVED;
+ queue_work(cma_wq, &work->work);
+ return 0;
+}
+
int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms)
{
struct rdma_id_private *id_priv;
@@ -1239,10 +1430,13 @@ int rdma_resolve_route(struct rdma_cm_id
return -EINVAL;
atomic_inc(&id_priv->refcount);
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
ret = cma_resolve_ib_route(id_priv, timeout_ms);
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = cma_resolve_iw_route(id_priv, timeout_ms);
+ break;
default:
ret = -ENOSYS;
break;
@@ -1354,8 +1548,8 @@ static int cma_resolve_loopback(struct r
ib_addr_get_sgid(&id_priv->id.route.addr.dev_addr));
if (cma_zero_addr(&id_priv->id.route.addr.src_addr)) {
- src_in = (struct sockaddr_in *)&id_priv->id.route.addr.src_addr;
- dst_in = (struct sockaddr_in *)&id_priv->id.route.addr.dst_addr;
+ src_in = (struct sockaddr_in *) &id_priv->id.route.addr.src_addr;
+ dst_in = (struct sockaddr_in *) &id_priv->id.route.addr.dst_addr;
src_in->sin_family = dst_in->sin_family;
src_in->sin_addr.s_addr = dst_in->sin_addr.s_addr;
}
@@ -1646,6 +1840,47 @@ out:
return ret;
}
+static int cma_connect_iw(struct rdma_id_private *id_priv,
+ struct rdma_conn_param *conn_param)
+{
+ struct iw_cm_id *cm_id;
+ struct sockaddr_in* sin;
+ int ret;
+ struct iw_cm_conn_param iw_param;
+
+ cm_id = iw_create_cm_id(id_priv->id.device, cma_iw_handler, id_priv);
+ if (IS_ERR(cm_id)) {
+ ret = PTR_ERR(cm_id);
+ goto out;
+ }
+
+ id_priv->cm_id.iw = cm_id;
+
+ sin = (struct sockaddr_in*) &id_priv->id.route.addr.src_addr;
+ cm_id->local_addr = *sin;
+
+ sin = (struct sockaddr_in*) &id_priv->id.route.addr.dst_addr;
+ cm_id->remote_addr = *sin;
+
+ ret = cma_modify_qp_rtr(&id_priv->id);
+ if (ret) {
+ iw_destroy_cm_id(cm_id);
+ return ret;
+ }
+
+ iw_param.ord = conn_param->initiator_depth;
+ iw_param.ird = conn_param->responder_resources;
+ iw_param.private_data = conn_param->private_data;
+ iw_param.private_data_len = conn_param->private_data_len;
+ if (id_priv->id.qp)
+ iw_param.qpn = id_priv->qp_num;
+ else
+ iw_param.qpn = conn_param->qp_num;
+ ret = iw_cm_connect(cm_id, &iw_param);
+out:
+ return ret;
+}
+
int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
{
struct rdma_id_private *id_priv;
@@ -1661,10 +1896,13 @@ int rdma_connect(struct rdma_cm_id *id,
id_priv->srq = conn_param->srq;
}
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
ret = cma_connect_ib(id_priv, conn_param);
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = cma_connect_iw(id_priv, conn_param);
+ break;
default:
ret = -ENOSYS;
break;
@@ -1705,6 +1943,28 @@ static int cma_accept_ib(struct rdma_id_
return ib_send_cm_rep(id_priv->cm_id.ib, &rep);
}
+static int cma_accept_iw(struct rdma_id_private *id_priv,
+ struct rdma_conn_param *conn_param)
+{
+ struct iw_cm_conn_param iw_param;
+ int ret;
+
+ ret = cma_modify_qp_rtr(&id_priv->id);
+ if (ret)
+ return ret;
+
+ iw_param.ord = conn_param->initiator_depth;
+ iw_param.ird = conn_param->responder_resources;
+ iw_param.private_data = conn_param->private_data;
+ iw_param.private_data_len = conn_param->private_data_len;
+ if (id_priv->id.qp) {
+ iw_param.qpn = id_priv->qp_num;
+ } else
+ iw_param.qpn = conn_param->qp_num;
+
+ return iw_cm_accept(id_priv->cm_id.iw, &iw_param);
+}
+
int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
{
struct rdma_id_private *id_priv;
@@ -1720,13 +1980,16 @@ int rdma_accept(struct rdma_cm_id *id, s
id_priv->srq = conn_param->srq;
}
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
if (conn_param)
ret = cma_accept_ib(id_priv, conn_param);
else
ret = cma_rep_recv(id_priv);
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = cma_accept_iw(id_priv, conn_param);
+ break;
default:
ret = -ENOSYS;
break;
@@ -1753,12 +2016,16 @@ int rdma_reject(struct rdma_cm_id *id, c
if (!cma_comp(id_priv, CMA_CONNECT))
return -EINVAL;
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
ret = ib_send_cm_rej(id_priv->cm_id.ib,
IB_CM_REJ_CONSUMER_DEFINED, NULL, 0,
private_data, private_data_len);
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = iw_cm_reject(id_priv->cm_id.iw,
+ private_data, private_data_len);
+ break;
default:
ret = -ENOSYS;
break;
@@ -1777,16 +2044,18 @@ int rdma_disconnect(struct rdma_cm_id *i
!cma_comp(id_priv, CMA_DISCONNECT))
return -EINVAL;
- ret = cma_modify_qp_err(id);
- if (ret)
- goto out;
-
- switch (id->device->node_type) {
- case IB_NODE_CA:
+ switch (rdma_node_get_transport(id->device->node_type)) {
+ case RDMA_TRANSPORT_IB:
+ ret = cma_modify_qp_err(id);
+ if (ret)
+ goto out;
/* Initiate or respond to a disconnect. */
if (ib_send_cm_dreq(id_priv->cm_id.ib, NULL, 0))
ib_send_cm_drep(id_priv->cm_id.ib, NULL, 0);
break;
+ case RDMA_TRANSPORT_IWARP:
+ ret = iw_cm_disconnect(id_priv->cm_id.iw, 0);
+ break;
default:
break;
}
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index b2f3cb9..7318fba 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -30,7 +30,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: device.c 1349 2004-12-16 21:09:43Z roland $
+ * $Id: device.c 5943 2006-03-22 00:58:04Z roland $
*/
#include <linux/module.h>
@@ -505,7 +505,7 @@ int ib_query_port(struct ib_device *devi
u8 port_num,
struct ib_port_attr *port_attr)
{
- if (device->node_type == IB_NODE_SWITCH) {
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
if (port_num)
return -EINVAL;
} else if (port_num < 1 || port_num > device->phys_port_cnt)
@@ -580,7 +580,7 @@ int ib_modify_port(struct ib_device *dev
u8 port_num, int port_modify_mask,
struct ib_port_modify *port_modify)
{
- if (device->node_type == IB_NODE_SWITCH) {
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
if (port_num)
return -EINVAL;
} else if (port_num < 1 || port_num > device->phys_port_cnt)
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index b38e02a..a928ecf 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2004, 2005 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2004-2006 Voltaire, Inc. All rights reserved.
* Copyright (c) 2005 Intel Corporation. All rights reserved.
* Copyright (c) 2005 Mellanox Technologies Ltd. All rights reserved.
*
@@ -31,7 +31,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: mad.c 5596 2006-03-03 01:00:07Z sean.hefty $
+ * $Id: mad.c 7294 2006-05-17 18:12:30Z roland $
*/
#include <linux/dma-mapping.h>
#include <rdma/ib_cache.h>
@@ -2877,7 +2877,10 @@ static void ib_mad_init_device(struct ib
{
int start, end, i;
- if (device->node_type == IB_NODE_SWITCH) {
+ if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+ return;
+
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
start = 0;
end = 0;
} else {
@@ -2924,7 +2927,7 @@ static void ib_mad_remove_device(struct
{
int i, num_ports, cur_port;
- if (device->node_type == IB_NODE_SWITCH) {
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
num_ports = 1;
cur_port = 0;
} else {
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index 501cc05..4230277 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -887,7 +887,10 @@ static void ib_sa_add_one(struct ib_devi
struct ib_sa_device *sa_dev;
int s, e, i;
- if (device->node_type == IB_NODE_SWITCH)
+ if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+ return;
+
+ if (device->node_type == RDMA_NODE_IB_SWITCH)
s = e = 0;
else {
s = 1;
diff --git a/drivers/infiniband/core/smi.c b/drivers/infiniband/core/smi.c
index 35852e7..b81b2b9 100644
--- a/drivers/infiniband/core/smi.c
+++ b/drivers/infiniband/core/smi.c
@@ -34,7 +34,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: smi.c 1389 2004-12-27 22:56:47Z roland $
+ * $Id: smi.c 5258 2006-02-01 20:32:40Z sean.hefty $
*/
#include <rdma/ib_smi.h>
@@ -64,7 +64,7 @@ int smi_handle_dr_smp_send(struct ib_smp
/* C14-9:2 */
if (hop_ptr && hop_ptr < hop_cnt) {
- if (node_type != IB_NODE_SWITCH)
+ if (node_type != RDMA_NODE_IB_SWITCH)
return 0;
/* smp->return_path set when received */
@@ -77,7 +77,7 @@ int smi_handle_dr_smp_send(struct ib_smp
if (hop_ptr == hop_cnt) {
/* smp->return_path set when received */
smp->hop_ptr++;
- return (node_type == IB_NODE_SWITCH ||
+ return (node_type == RDMA_NODE_IB_SWITCH ||
smp->dr_dlid == IB_LID_PERMISSIVE);
}
@@ -95,7 +95,7 @@ int smi_handle_dr_smp_send(struct ib_smp
/* C14-13:2 */
if (2 <= hop_ptr && hop_ptr <= hop_cnt) {
- if (node_type != IB_NODE_SWITCH)
+ if (node_type != RDMA_NODE_IB_SWITCH)
return 0;
smp->hop_ptr--;
@@ -107,7 +107,7 @@ int smi_handle_dr_smp_send(struct ib_smp
if (hop_ptr == 1) {
smp->hop_ptr--;
/* C14-13:3 -- SMPs destined for SM shouldn't be here */
- return (node_type == IB_NODE_SWITCH ||
+ return (node_type == RDMA_NODE_IB_SWITCH ||
smp->dr_slid == IB_LID_PERMISSIVE);
}
@@ -142,7 +142,7 @@ int smi_handle_dr_smp_recv(struct ib_smp
/* C14-9:2 -- intermediate hop */
if (hop_ptr && hop_ptr < hop_cnt) {
- if (node_type != IB_NODE_SWITCH)
+ if (node_type != RDMA_NODE_IB_SWITCH)
return 0;
smp->return_path[hop_ptr] = port_num;
@@ -156,7 +156,7 @@ int smi_handle_dr_smp_recv(struct ib_smp
smp->return_path[hop_ptr] = port_num;
/* smp->hop_ptr updated when sending */
- return (node_type == IB_NODE_SWITCH ||
+ return (node_type == RDMA_NODE_IB_SWITCH ||
smp->dr_dlid == IB_LID_PERMISSIVE);
}
@@ -175,7 +175,7 @@ int smi_handle_dr_smp_recv(struct ib_smp
/* C14-13:2 */
if (2 <= hop_ptr && hop_ptr <= hop_cnt) {
- if (node_type != IB_NODE_SWITCH)
+ if (node_type != RDMA_NODE_IB_SWITCH)
return 0;
/* smp->hop_ptr updated when sending */
@@ -190,7 +190,7 @@ int smi_handle_dr_smp_recv(struct ib_smp
return 1;
}
/* smp->hop_ptr updated when sending */
- return (node_type == IB_NODE_SWITCH);
+ return (node_type == RDMA_NODE_IB_SWITCH);
}
/* C14-13:4 -- hop_ptr = 0 -> give to SM */
diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c
index 21f9282..cfd2c06 100644
--- a/drivers/infiniband/core/sysfs.c
+++ b/drivers/infiniband/core/sysfs.c
@@ -31,7 +31,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: sysfs.c 1349 2004-12-16 21:09:43Z roland $
+ * $Id: sysfs.c 6940 2006-05-04 17:04:55Z roland $
*/
#include "core_priv.h"
@@ -589,10 +589,16 @@ static ssize_t show_node_type(struct cla
return -ENODEV;
switch (dev->node_type) {
- case IB_NODE_CA: return sprintf(buf, "%d: CA\n", dev->node_type);
- case IB_NODE_SWITCH: return sprintf(buf, "%d: switch\n", dev->node_type);
- case IB_NODE_ROUTER: return sprintf(buf, "%d: router\n", dev->node_type);
- default: return sprintf(buf, "%d: <unknown>\n", dev->node_type);
+ case RDMA_NODE_IB_CA:
+ return sprintf(buf, "%d: CA\n", dev->node_type);
+ case RDMA_NODE_RNIC:
+ return sprintf(buf, "%d: RNIC\n", dev->node_type);
+ case RDMA_NODE_IB_SWITCH:
+ return sprintf(buf, "%d: switch\n", dev->node_type);
+ case RDMA_NODE_IB_ROUTER:
+ return sprintf(buf, "%d: router\n", dev->node_type);
+ default:
+ return sprintf(buf, "%d: <unknown>\n", dev->node_type);
}
}
@@ -708,7 +714,7 @@ int ib_device_register_sysfs(struct ib_d
if (ret)
goto err_put;
- if (device->node_type == IB_NODE_SWITCH) {
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
ret = add_port(device, 0);
if (ret)
goto err_put;
diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c
index 67caf36..ad2e417 100644
--- a/drivers/infiniband/core/ucm.c
+++ b/drivers/infiniband/core/ucm.c
@@ -30,7 +30,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: ucm.c 4311 2005-12-05 18:42:01Z sean.hefty $
+ * $Id: ucm.c 7119 2006-05-11 16:40:38Z sean.hefty $
*/
#include <linux/completion.h>
@@ -1248,7 +1248,8 @@ static void ib_ucm_add_one(struct ib_dev
{
struct ib_ucm_device *ucm_dev;
- if (!device->alloc_ucontext)
+ if (!device->alloc_ucontext ||
+ rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
return;
ucm_dev = kzalloc(sizeof *ucm_dev, GFP_KERNEL);
diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index afe70a5..0cbd692 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -1,6 +1,6 @@
/*
* Copyright (c) 2004 Topspin Communications. All rights reserved.
- * Copyright (c) 2005 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2005-2006 Voltaire, Inc. All rights reserved.
* Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
*
* This software is available to you under a choice of one of two
@@ -31,7 +31,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: user_mad.c 5596 2006-03-03 01:00:07Z sean.hefty $
+ * $Id: user_mad.c 6041 2006-03-27 21:06:00Z halr $
*/
#include <linux/module.h>
@@ -967,7 +967,10 @@ static void ib_umad_add_one(struct ib_de
struct ib_umad_device *umad_dev;
int s, e, i;
- if (device->node_type == IB_NODE_SWITCH)
+ if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+ return;
+
+ if (device->node_type == RDMA_NODE_IB_SWITCH)
s = e = 0;
else {
s = 1;
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
index 28fdbda..e4b45d7 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c
@@ -984,7 +984,7 @@ static void *ipath_register_ib_device(in
(1ull << IB_USER_VERBS_CMD_QUERY_SRQ) |
(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ) |
(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
- dev->node_type = IB_NODE_CA;
+ dev->node_type = RDMA_NODE_IB_CA;
dev->phys_port_cnt = 1;
dev->dma_device = ipath_layer_get_device(dd);
dev->class_dev.dev = dev->dma_device;
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index a2eae8a..5c31819 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -1273,7 +1273,7 @@ int mthca_register_device(struct mthca_d
(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ) |
(1ull << IB_USER_VERBS_CMD_QUERY_SRQ) |
(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ);
- dev->ib_dev.node_type = IB_NODE_CA;
+ dev->ib_dev.node_type = RDMA_NODE_IB_CA;
dev->ib_dev.phys_port_cnt = dev->limits.num_ports;
dev->ib_dev.dma_device = &dev->pdev->dev;
dev->ib_dev.class_dev.dev = &dev->pdev->dev;
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index 1c6ea1c..262427f 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -1084,13 +1084,16 @@ static void ipoib_add_one(struct ib_devi
struct ipoib_dev_priv *priv;
int s, e, p;
+ if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+ return;
+
dev_list = kmalloc(sizeof *dev_list, GFP_KERNEL);
if (!dev_list)
return;
INIT_LIST_HEAD(dev_list);
- if (device->node_type == IB_NODE_SWITCH) {
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
s = 0;
e = 0;
} else {
@@ -1114,6 +1117,9 @@ static void ipoib_remove_one(struct ib_d
struct ipoib_dev_priv *priv, *tmp;
struct list_head *dev_list;
+ if (rdma_node_get_transport(device->node_type) != RDMA_TRANSPORT_IB)
+ return;
+
dev_list = ib_get_client_data(device, &ipoib_client);
list_for_each_entry_safe(priv, tmp, dev_list, list) {
diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index f1401e1..bba2956 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -1845,7 +1845,7 @@ static void srp_add_one(struct ib_device
if (IS_ERR(srp_dev->fmr_pool))
srp_dev->fmr_pool = NULL;
- if (device->node_type == IB_NODE_SWITCH) {
+ if (device->node_type == RDMA_NODE_IB_SWITCH) {
s = 0;
e = 0;
} else {
diff --git a/include/rdma/ib_addr.h b/include/rdma/ib_addr.h
index fcb5ba8..d95d3eb 100644
--- a/include/rdma/ib_addr.h
+++ b/include/rdma/ib_addr.h
@@ -40,7 +40,7 @@ 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;
+ enum rdma_node_type dev_type;
};
/**
@@ -72,6 +72,9 @@ int rdma_resolve_ip(struct sockaddr *src
void rdma_addr_cancel(struct rdma_dev_addr *addr);
+int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
+ const unsigned char *dst_dev_addr);
+
static inline int ip_addr_size(struct sockaddr *addr)
{
return addr->sa_family == AF_INET6 ?
@@ -111,4 +114,14 @@ static inline void ib_addr_set_dgid(stru
memcpy(dev_addr->dst_dev_addr + 4, gid, sizeof *gid);
}
+static inline union ib_gid* iw_addr_get_sgid(struct rdma_dev_addr* rda)
+{
+ return (union ib_gid *) rda->src_dev_addr;
+}
+
+static inline union ib_gid* iw_addr_get_dgid(struct rdma_dev_addr* rda)
+{
+ return (union ib_gid *) rda->dst_dev_addr;
+}
+
#endif /* IB_ADDR_H */
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index aeb4fcd..eac2d8f 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -35,7 +35,7 @@
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
- * $Id: ib_verbs.h 1349 2004-12-16 21:09:43Z roland $
+ * $Id: ib_verbs.h 6885 2006-05-03 18:22:02Z sean.hefty $
*/
#if !defined(IB_VERBS_H)
@@ -56,12 +56,35 @@ union ib_gid {
} global;
};
-enum ib_node_type {
- IB_NODE_CA = 1,
- IB_NODE_SWITCH,
- IB_NODE_ROUTER
+enum rdma_node_type {
+ /* IB values map to NodeInfo:NodeType. */
+ RDMA_NODE_IB_CA = 1,
+ RDMA_NODE_IB_SWITCH,
+ RDMA_NODE_IB_ROUTER,
+ RDMA_NODE_RNIC
};
+enum rdma_transport_type {
+ RDMA_TRANSPORT_IB,
+ RDMA_TRANSPORT_IWARP
+};
+
+static inline enum rdma_transport_type
+rdma_node_get_transport(enum rdma_node_type node_type)
+{
+ switch (node_type) {
+ case RDMA_NODE_IB_CA:
+ case RDMA_NODE_IB_SWITCH:
+ case RDMA_NODE_IB_ROUTER:
+ return RDMA_TRANSPORT_IB;
+ case RDMA_NODE_RNIC:
+ return RDMA_TRANSPORT_IWARP;
+ default:
+ BUG();
+ return 0;
+ }
+}
+
enum ib_device_cap_flags {
IB_DEVICE_RESIZE_MAX_WR = 1,
IB_DEVICE_BAD_PKEY_CNTR = (1<<1),
@@ -78,6 +101,9 @@ enum ib_device_cap_flags {
IB_DEVICE_RC_RNR_NAK_GEN = (1<<12),
IB_DEVICE_SRQ_RESIZE = (1<<13),
IB_DEVICE_N_NOTIFY_CQ = (1<<14),
+ IB_DEVICE_ZERO_STAG = (1<<15),
+ IB_DEVICE_SEND_W_INV = (1<<16),
+ IB_DEVICE_MEM_WINDOW = (1<<17)
};
enum ib_atomic_cap {
@@ -830,6 +856,7 @@ struct ib_cache {
u8 *lmc_cache;
};
+struct iw_cm_verbs;
struct ib_device {
struct device *dma_device;
@@ -846,6 +873,8 @@ struct ib_device {
u32 flags;
+ struct iw_cm_verbs *iwcm;
+
int (*query_device)(struct ib_device *device,
struct ib_device_attr *device_attr);
int (*query_port)(struct ib_device *device,
^ permalink raw reply related
* [PATCH v2 0/7][RFC] Ammasso 1100 iWARP Driver
From: Steve Wise @ 2006-06-07 20:06 UTC (permalink / raw)
To: rdreier, mshefty; +Cc: linux-kernel, netdev, openib-general
This patchset implements the iWARP provider driver for the Ammasso
1100 RNIC. It is dependent on the "iWARP Core Support" patch set. We're
submitting it for review with the goal for inclusion in the 2.6.19 kernel.
This code has gone through several reviews in the openib-general list.
Now we are submitting it for external review by the linux community.
This StGIT patchset is cloned from Roland Dreier's infiniband.git
for-2.6.18 branch. The patchset consists of 7 patches:
1 - Low-level device interface and native stack support
2 - Work request definitions
3 - Provider interface
4 - Memory management
5 - User mode message queue implementation
6 - Verbs queue implementation
7 - Kconfig and Makefile
I believe I've addressed all the round 1 review comments. Details of
the changes are tracked in each patch comment.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
^ permalink raw reply
* [PATCH v2 1/7] AMSO1100 Low Level Driver.
From: Steve Wise @ 2006-06-07 20:06 UTC (permalink / raw)
To: rdreier, mshefty; +Cc: netdev, linux-kernel, openib-general
In-Reply-To: <20060607200646.9259.24588.stgit@stevo-desktop>
This is the core of the driver and includes the hardware probe, low-level
device interfaces and native Ethernet support.
Review Changes:
- sizeof -> sizeof()
- dprintk() -> pr_debug()
- removed useless asserts
- assert() -> BUG_ON()
- C2_DEBUG -> DEBUG
- removed debug netevent code
- removed arp request squelch code from intr handler, replacing it
with setting arp_ignore when the c2 netdev is brought up.
- removed c2_set_mac_addr().
---
drivers/infiniband/hw/amso1100/c2.c | 1255 ++++++++++++++++++++++++++++++
drivers/infiniband/hw/amso1100/c2.h | 555 +++++++++++++
drivers/infiniband/hw/amso1100/c2_ae.c | 359 +++++++++
drivers/infiniband/hw/amso1100/c2_intr.c | 209 +++++
drivers/infiniband/hw/amso1100/c2_rnic.c | 631 +++++++++++++++
5 files changed, 3009 insertions(+), 0 deletions(-)
diff --git a/drivers/infiniband/hw/amso1100/c2.c b/drivers/infiniband/hw/amso1100/c2.c
new file mode 100644
index 0000000..4fdbd80
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2.c
@@ -0,0 +1,1255 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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/module.h>
+#include <linux/moduleparam.h>
+#include <linux/pci.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/inetdevice.h>
+#include <linux/delay.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/if_vlan.h>
+#include <linux/crc32.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/init.h>
+#include <linux/dma-mapping.h>
+
+#include <asm/io.h>
+#include <asm/irq.h>
+#include <asm/byteorder.h>
+
+#include <rdma/ib_smi.h>
+#include "c2.h"
+#include "c2_provider.h"
+
+MODULE_AUTHOR("Tom Tucker <tom@opengridcomputing.com>");
+MODULE_DESCRIPTION("Ammasso AMSO1100 Low-level iWARP Driver");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_VERSION(DRV_VERSION);
+
+static const u32 default_msg = NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK
+ | NETIF_MSG_IFUP | NETIF_MSG_IFDOWN;
+
+static int debug = -1; /* defaults above */
+module_param(debug, int, 0);
+MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
+
+static int c2_up(struct net_device *netdev);
+static int c2_down(struct net_device *netdev);
+static int c2_xmit_frame(struct sk_buff *skb, struct net_device *netdev);
+static void c2_tx_interrupt(struct net_device *netdev);
+static void c2_rx_interrupt(struct net_device *netdev);
+static irqreturn_t c2_interrupt(int irq, void *dev_id, struct pt_regs *regs);
+static void c2_tx_timeout(struct net_device *netdev);
+static int c2_change_mtu(struct net_device *netdev, int new_mtu);
+static void c2_reset(struct c2_port *c2_port);
+static struct net_device_stats *c2_get_stats(struct net_device *netdev);
+
+static struct pci_device_id c2_pci_table[] = {
+ {0x18b8, 0xb001, PCI_ANY_ID, PCI_ANY_ID},
+ {0}
+};
+
+MODULE_DEVICE_TABLE(pci, c2_pci_table);
+
+static void c2_print_macaddr(struct net_device *netdev)
+{
+ pr_debug("%s: MAC %02X:%02X:%02X:%02X:%02X:%02X, "
+ "IRQ %u\n", netdev->name,
+ netdev->dev_addr[0], netdev->dev_addr[1], netdev->dev_addr[2],
+ netdev->dev_addr[3], netdev->dev_addr[4], netdev->dev_addr[5],
+ netdev->irq);
+}
+
+static void c2_set_rxbufsize(struct c2_port *c2_port)
+{
+ struct net_device *netdev = c2_port->netdev;
+
+ if (netdev->mtu > RX_BUF_SIZE)
+ c2_port->rx_buf_size =
+ netdev->mtu + ETH_HLEN + sizeof(struct c2_rxp_hdr) +
+ NET_IP_ALIGN;
+ else
+ c2_port->rx_buf_size = sizeof(struct c2_rxp_hdr) + RX_BUF_SIZE;
+}
+
+/*
+ * Allocate TX ring elements and chain them together.
+ * One-to-one association of adapter descriptors with ring elements.
+ */
+static int c2_tx_ring_alloc(struct c2_ring *tx_ring, void *vaddr,
+ dma_addr_t base, void __iomem * mmio_txp_ring)
+{
+ struct c2_tx_desc *tx_desc;
+ struct c2_txp_desc __iomem *txp_desc;
+ struct c2_element *elem;
+ int i;
+
+ tx_ring->start = kmalloc(sizeof(*elem) * tx_ring->count, GFP_KERNEL);
+ if (!tx_ring->start)
+ return -ENOMEM;
+
+ elem = tx_ring->start;
+ tx_desc = vaddr;
+ txp_desc = mmio_txp_ring;
+ for (i = 0; i < tx_ring->count; i++, elem++, tx_desc++, txp_desc++) {
+ tx_desc->len = 0;
+ tx_desc->status = 0;
+
+ /* Set TXP_HTXD_UNINIT */
+ __raw_writeq(cpu_to_be64(0x1122334455667788ULL),
+ (void __iomem *) txp_desc + C2_TXP_ADDR);
+ __raw_writew(0, (void __iomem *) txp_desc + C2_TXP_LEN);
+ __raw_writew(cpu_to_be16(TXP_HTXD_UNINIT),
+ (void __iomem *) txp_desc + C2_TXP_FLAGS);
+
+ elem->skb = NULL;
+ elem->ht_desc = tx_desc;
+ elem->hw_desc = txp_desc;
+
+ if (i == tx_ring->count - 1) {
+ elem->next = tx_ring->start;
+ tx_desc->next_offset = base;
+ } else {
+ elem->next = elem + 1;
+ tx_desc->next_offset =
+ base + (i + 1) * sizeof(*tx_desc);
+ }
+ }
+
+ tx_ring->to_use = tx_ring->to_clean = tx_ring->start;
+
+ return 0;
+}
+
+/*
+ * Allocate RX ring elements and chain them together.
+ * One-to-one association of adapter descriptors with ring elements.
+ */
+static int c2_rx_ring_alloc(struct c2_ring *rx_ring, void *vaddr,
+ dma_addr_t base, void __iomem * mmio_rxp_ring)
+{
+ struct c2_rx_desc *rx_desc;
+ struct c2_rxp_desc __iomem *rxp_desc;
+ struct c2_element *elem;
+ int i;
+
+ rx_ring->start = kmalloc(sizeof(*elem) * rx_ring->count, GFP_KERNEL);
+ if (!rx_ring->start)
+ return -ENOMEM;
+
+ elem = rx_ring->start;
+ rx_desc = vaddr;
+ rxp_desc = mmio_rxp_ring;
+ for (i = 0; i < rx_ring->count; i++, elem++, rx_desc++, rxp_desc++) {
+ rx_desc->len = 0;
+ rx_desc->status = 0;
+
+ /* Set RXP_HRXD_UNINIT */
+ __raw_writew(cpu_to_be16(RXP_HRXD_OK),
+ (void __iomem *) rxp_desc + C2_RXP_STATUS);
+ __raw_writew(0, (void __iomem *) rxp_desc + C2_RXP_COUNT);
+ __raw_writew(0, (void __iomem *) rxp_desc + C2_RXP_LEN);
+ __raw_writeq(cpu_to_be64(0x99aabbccddeeffULL),
+ (void __iomem *) rxp_desc + C2_RXP_ADDR);
+ __raw_writew(cpu_to_be16(RXP_HRXD_UNINIT),
+ (void __iomem *) rxp_desc + C2_RXP_FLAGS);
+
+ elem->skb = NULL;
+ elem->ht_desc = rx_desc;
+ elem->hw_desc = rxp_desc;
+
+ if (i == rx_ring->count - 1) {
+ elem->next = rx_ring->start;
+ rx_desc->next_offset = base;
+ } else {
+ elem->next = elem + 1;
+ rx_desc->next_offset =
+ base + (i + 1) * sizeof(*rx_desc);
+ }
+ }
+
+ rx_ring->to_use = rx_ring->to_clean = rx_ring->start;
+
+ return 0;
+}
+
+/* Setup buffer for receiving */
+static inline int c2_rx_alloc(struct c2_port *c2_port, struct c2_element *elem)
+{
+ struct c2_dev *c2dev = c2_port->c2dev;
+ struct c2_rx_desc *rx_desc = elem->ht_desc;
+ struct sk_buff *skb;
+ dma_addr_t mapaddr;
+ u32 maplen;
+ struct c2_rxp_hdr *rxp_hdr;
+
+ skb = dev_alloc_skb(c2_port->rx_buf_size);
+ if (unlikely(!skb)) {
+ pr_debug("%s: out of memory for receive\n",
+ c2_port->netdev->name);
+ return -ENOMEM;
+ }
+
+ /* Zero out the rxp hdr in the sk_buff */
+ memset(skb->data, 0, sizeof(*rxp_hdr));
+
+ skb->dev = c2_port->netdev;
+
+ maplen = c2_port->rx_buf_size;
+ mapaddr =
+ pci_map_single(c2dev->pcidev, skb->data, maplen,
+ PCI_DMA_FROMDEVICE);
+
+ /* Set the sk_buff RXP_header to RXP_HRXD_READY */
+ rxp_hdr = (struct c2_rxp_hdr *) skb->data;
+ rxp_hdr->flags = RXP_HRXD_READY;
+
+ __raw_writew(0, elem->hw_desc + C2_RXP_STATUS);
+ __raw_writew(cpu_to_be16((u16) maplen - sizeof(*rxp_hdr)),
+ elem->hw_desc + C2_RXP_LEN);
+ __raw_writeq(cpu_to_be64(mapaddr), elem->hw_desc + C2_RXP_ADDR);
+ __raw_writew(cpu_to_be16(RXP_HRXD_READY), elem->hw_desc + C2_RXP_FLAGS);
+
+ elem->skb = skb;
+ elem->mapaddr = mapaddr;
+ elem->maplen = maplen;
+ rx_desc->len = maplen;
+
+ return 0;
+}
+
+/*
+ * Allocate buffers for the Rx ring
+ * For receive: rx_ring.to_clean is next received frame
+ */
+static int c2_rx_fill(struct c2_port *c2_port)
+{
+ struct c2_ring *rx_ring = &c2_port->rx_ring;
+ struct c2_element *elem;
+ int ret = 0;
+
+ elem = rx_ring->start;
+ do {
+ if (c2_rx_alloc(c2_port, elem)) {
+ ret = 1;
+ break;
+ }
+ } while ((elem = elem->next) != rx_ring->start);
+
+ rx_ring->to_clean = rx_ring->start;
+ return ret;
+}
+
+/* Free all buffers in RX ring, assumes receiver stopped */
+static void c2_rx_clean(struct c2_port *c2_port)
+{
+ struct c2_dev *c2dev = c2_port->c2dev;
+ struct c2_ring *rx_ring = &c2_port->rx_ring;
+ struct c2_element *elem;
+ struct c2_rx_desc *rx_desc;
+
+ elem = rx_ring->start;
+ do {
+ rx_desc = elem->ht_desc;
+ rx_desc->len = 0;
+
+ __raw_writew(0, elem->hw_desc + C2_RXP_STATUS);
+ __raw_writew(0, elem->hw_desc + C2_RXP_COUNT);
+ __raw_writew(0, elem->hw_desc + C2_RXP_LEN);
+ __raw_writeq(cpu_to_be64(0x99aabbccddeeffULL),
+ elem->hw_desc + C2_RXP_ADDR);
+ __raw_writew(cpu_to_be16(RXP_HRXD_UNINIT),
+ elem->hw_desc + C2_RXP_FLAGS);
+
+ if (elem->skb) {
+ pci_unmap_single(c2dev->pcidev, elem->mapaddr,
+ elem->maplen, PCI_DMA_FROMDEVICE);
+ dev_kfree_skb(elem->skb);
+ elem->skb = NULL;
+ }
+ } while ((elem = elem->next) != rx_ring->start);
+}
+
+static inline int c2_tx_free(struct c2_dev *c2dev, struct c2_element *elem)
+{
+ struct c2_tx_desc *tx_desc = elem->ht_desc;
+
+ tx_desc->len = 0;
+
+ pci_unmap_single(c2dev->pcidev, elem->mapaddr, elem->maplen,
+ PCI_DMA_TODEVICE);
+
+ if (elem->skb) {
+ dev_kfree_skb_any(elem->skb);
+ elem->skb = NULL;
+ }
+
+ return 0;
+}
+
+/* Free all buffers in TX ring, assumes transmitter stopped */
+static void c2_tx_clean(struct c2_port *c2_port)
+{
+ struct c2_ring *tx_ring = &c2_port->tx_ring;
+ struct c2_element *elem;
+ struct c2_txp_desc txp_htxd;
+ int retry;
+ unsigned long flags;
+
+ spin_lock_irqsave(&c2_port->tx_lock, flags);
+
+ elem = tx_ring->start;
+
+ do {
+ retry = 0;
+ do {
+ txp_htxd.flags =
+ readw(elem->hw_desc + C2_TXP_FLAGS);
+
+ if (txp_htxd.flags == TXP_HTXD_READY) {
+ retry = 1;
+ __raw_writew(0,
+ elem->hw_desc + C2_TXP_LEN);
+ __raw_writeq(0,
+ elem->hw_desc + C2_TXP_ADDR);
+ __raw_writew(cpu_to_be16(TXP_HTXD_DONE),
+ elem->hw_desc + C2_TXP_FLAGS);
+ c2_port->netstats.tx_dropped++;
+ break;
+ } else {
+ __raw_writew(0,
+ elem->hw_desc + C2_TXP_LEN);
+ __raw_writeq(cpu_to_be64(0x1122334455667788ULL),
+ elem->hw_desc + C2_TXP_ADDR);
+ __raw_writew(cpu_to_be16(TXP_HTXD_UNINIT),
+ elem->hw_desc + C2_TXP_FLAGS);
+ }
+
+ c2_tx_free(c2_port->c2dev, elem);
+
+ } while ((elem = elem->next) != tx_ring->start);
+ } while (retry);
+
+ c2_port->tx_avail = c2_port->tx_ring.count - 1;
+ c2_port->c2dev->cur_tx = tx_ring->to_use - tx_ring->start;
+
+ if (c2_port->tx_avail > MAX_SKB_FRAGS + 1)
+ netif_wake_queue(c2_port->netdev);
+
+ spin_unlock_irqrestore(&c2_port->tx_lock, flags);
+}
+
+/*
+ * Process transmit descriptors marked 'DONE' by the firmware,
+ * freeing up their unneeded sk_buffs.
+ */
+static void c2_tx_interrupt(struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+ struct c2_dev *c2dev = c2_port->c2dev;
+ struct c2_ring *tx_ring = &c2_port->tx_ring;
+ struct c2_element *elem;
+ struct c2_txp_desc txp_htxd;
+
+ spin_lock(&c2_port->tx_lock);
+
+ for (elem = tx_ring->to_clean; elem != tx_ring->to_use;
+ elem = elem->next) {
+ txp_htxd.flags =
+ be16_to_cpu(readw(elem->hw_desc + C2_TXP_FLAGS));
+
+ if (txp_htxd.flags != TXP_HTXD_DONE)
+ break;
+
+ if (netif_msg_tx_done(c2_port)) {
+ /* PCI reads are expensive in fast path */
+ txp_htxd.len =
+ be16_to_cpu(readw(elem->hw_desc + C2_TXP_LEN));
+ pr_debug("%s: tx done slot %3Zu status 0x%x len "
+ "%5u bytes\n",
+ netdev->name, elem - tx_ring->start,
+ txp_htxd.flags, txp_htxd.len);
+ }
+
+ c2_tx_free(c2dev, elem);
+ ++(c2_port->tx_avail);
+ }
+
+ tx_ring->to_clean = elem;
+
+ if (netif_queue_stopped(netdev)
+ && c2_port->tx_avail > MAX_SKB_FRAGS + 1)
+ netif_wake_queue(netdev);
+
+ spin_unlock(&c2_port->tx_lock);
+}
+
+static void c2_rx_error(struct c2_port *c2_port, struct c2_element *elem)
+{
+ struct c2_rx_desc *rx_desc = elem->ht_desc;
+ struct c2_rxp_hdr *rxp_hdr = (struct c2_rxp_hdr *) elem->skb->data;
+
+ if (rxp_hdr->status != RXP_HRXD_OK ||
+ rxp_hdr->len > (rx_desc->len - sizeof(*rxp_hdr))) {
+ pr_debug("BAD RXP_HRXD\n");
+ pr_debug(" rx_desc : %p\n", rx_desc);
+ pr_debug(" index : %Zu\n",
+ elem - c2_port->rx_ring.start);
+ pr_debug(" len : %u\n", rx_desc->len);
+ pr_debug(" rxp_hdr : %p [PA %p]\n", rxp_hdr,
+ (void *) __pa((unsigned long) rxp_hdr));
+ pr_debug(" flags : 0x%x\n", rxp_hdr->flags);
+ pr_debug(" status: 0x%x\n", rxp_hdr->status);
+ pr_debug(" len : %u\n", rxp_hdr->len);
+ pr_debug(" rsvd : 0x%x\n", rxp_hdr->rsvd);
+ }
+
+ /* Setup the skb for reuse since we're dropping this pkt */
+ elem->skb->tail = elem->skb->data = elem->skb->head;
+
+ /* Zero out the rxp hdr in the sk_buff */
+ memset(elem->skb->data, 0, sizeof(*rxp_hdr));
+
+ /* Write the descriptor to the adapter's rx ring */
+ __raw_writew(0, elem->hw_desc + C2_RXP_STATUS);
+ __raw_writew(0, elem->hw_desc + C2_RXP_COUNT);
+ __raw_writew(cpu_to_be16((u16) elem->maplen - sizeof(*rxp_hdr)),
+ elem->hw_desc + C2_RXP_LEN);
+ __raw_writeq(cpu_to_be64(elem->mapaddr), elem->hw_desc + C2_RXP_ADDR);
+ __raw_writew(cpu_to_be16(RXP_HRXD_READY), elem->hw_desc + C2_RXP_FLAGS);
+
+ pr_debug("packet dropped\n");
+ c2_port->netstats.rx_dropped++;
+}
+
+static void c2_rx_interrupt(struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+ struct c2_dev *c2dev = c2_port->c2dev;
+ struct c2_ring *rx_ring = &c2_port->rx_ring;
+ struct c2_element *elem;
+ struct c2_rx_desc *rx_desc;
+ struct c2_rxp_hdr *rxp_hdr;
+ struct sk_buff *skb;
+ dma_addr_t mapaddr;
+ u32 maplen, buflen;
+ unsigned long flags;
+
+ spin_lock_irqsave(&c2dev->lock, flags);
+
+ /* Begin where we left off */
+ rx_ring->to_clean = rx_ring->start + c2dev->cur_rx;
+
+ for (elem = rx_ring->to_clean; elem->next != rx_ring->to_clean;
+ elem = elem->next) {
+ rx_desc = elem->ht_desc;
+ mapaddr = elem->mapaddr;
+ maplen = elem->maplen;
+ skb = elem->skb;
+ rxp_hdr = (struct c2_rxp_hdr *) skb->data;
+
+ if (rxp_hdr->flags != RXP_HRXD_DONE)
+ break;
+ buflen = rxp_hdr->len;
+
+ /* Sanity check the RXP header */
+ if (rxp_hdr->status != RXP_HRXD_OK ||
+ buflen > (rx_desc->len - sizeof(*rxp_hdr))) {
+ c2_rx_error(c2_port, elem);
+ continue;
+ }
+
+ /*
+ * Allocate and map a new skb for replenishing the host
+ * RX desc
+ */
+ if (c2_rx_alloc(c2_port, elem)) {
+ c2_rx_error(c2_port, elem);
+ continue;
+ }
+
+ /* Unmap the old skb */
+ pci_unmap_single(c2dev->pcidev, mapaddr, maplen,
+ PCI_DMA_FROMDEVICE);
+
+ prefetch(skb->data);
+
+ /*
+ * Skip past the leading 8 bytes comprising of the
+ * "struct c2_rxp_hdr", prepended by the adapter
+ * to the usual Ethernet header ("struct ethhdr"),
+ * to the start of the raw Ethernet packet.
+ *
+ * Fix up the various fields in the sk_buff before
+ * passing it up to netif_rx(). The transfer size
+ * (in bytes) specified by the adapter len field of
+ * the "struct rxp_hdr_t" does NOT include the
+ * "sizeof(struct c2_rxp_hdr)".
+ */
+ skb->data += sizeof(*rxp_hdr);
+ skb->tail = skb->data + buflen;
+ skb->len = buflen;
+ skb->dev = netdev;
+ skb->protocol = eth_type_trans(skb, netdev);
+
+ netif_rx(skb);
+
+ netdev->last_rx = jiffies;
+ c2_port->netstats.rx_packets++;
+ c2_port->netstats.rx_bytes += buflen;
+ }
+
+ /* Save where we left off */
+ rx_ring->to_clean = elem;
+ c2dev->cur_rx = elem - rx_ring->start;
+ C2_SET_CUR_RX(c2dev, c2dev->cur_rx);
+
+ spin_unlock_irqrestore(&c2dev->lock, flags);
+}
+
+/*
+ * Handle netisr0 TX & RX interrupts.
+ */
+static irqreturn_t c2_interrupt(int irq, void *dev_id, struct pt_regs *regs)
+{
+ unsigned int netisr0, dmaisr;
+ int handled = 0;
+ struct c2_dev *c2dev = (struct c2_dev *) dev_id;
+
+ /* Process CCILNET interrupts */
+ netisr0 = readl(c2dev->regs + C2_NISR0);
+ if (netisr0) {
+
+ /*
+ * There is an issue with the firmware that always
+ * provides the status of RX for both TX & RX
+ * interrupts. So process both queues here.
+ */
+ c2_rx_interrupt(c2dev->netdev);
+ c2_tx_interrupt(c2dev->netdev);
+
+ /* Clear the interrupt */
+ writel(netisr0, c2dev->regs + C2_NISR0);
+ handled++;
+ }
+
+ /* Process RNIC interrupts */
+ dmaisr = readl(c2dev->regs + C2_DISR);
+ if (dmaisr) {
+ writel(dmaisr, c2dev->regs + C2_DISR);
+ c2_rnic_interrupt(c2dev);
+ handled++;
+ }
+
+ if (handled) {
+ return IRQ_HANDLED;
+ } else {
+ return IRQ_NONE;
+ }
+}
+
+static int c2_up(struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+ struct c2_dev *c2dev = c2_port->c2dev;
+ struct c2_element *elem;
+ struct c2_rxp_hdr *rxp_hdr;
+ struct in_device *in_dev;
+ size_t rx_size, tx_size;
+ int ret, i;
+ unsigned int netimr0;
+
+ if (netif_msg_ifup(c2_port))
+ pr_debug("%s: enabling interface\n", netdev->name);
+
+ /* Set the Rx buffer size based on MTU */
+ c2_set_rxbufsize(c2_port);
+
+ /* Allocate DMA'able memory for Tx/Rx host descriptor rings */
+ rx_size = c2_port->rx_ring.count * sizeof(struct c2_rx_desc);
+ tx_size = c2_port->tx_ring.count * sizeof(struct c2_tx_desc);
+
+ c2_port->mem_size = tx_size + rx_size;
+ c2_port->mem = pci_alloc_consistent(c2dev->pcidev, c2_port->mem_size,
+ &c2_port->dma);
+ if (c2_port->mem == NULL) {
+ pr_debug("Unable to allocate memory for "
+ "host descriptor rings\n");
+ return -ENOMEM;
+ }
+
+ memset(c2_port->mem, 0, c2_port->mem_size);
+
+ /* Create the Rx host descriptor ring */
+ if ((ret =
+ c2_rx_ring_alloc(&c2_port->rx_ring, c2_port->mem, c2_port->dma,
+ c2dev->mmio_rxp_ring))) {
+ pr_debug("Unable to create RX ring\n");
+ goto bail0;
+ }
+
+ /* Allocate Rx buffers for the host descriptor ring */
+ if (c2_rx_fill(c2_port)) {
+ pr_debug("Unable to fill RX ring\n");
+ goto bail1;
+ }
+
+ /* Create the Tx host descriptor ring */
+ if ((ret = c2_tx_ring_alloc(&c2_port->tx_ring, c2_port->mem + rx_size,
+ c2_port->dma + rx_size,
+ c2dev->mmio_txp_ring))) {
+ pr_debug("Unable to create TX ring\n");
+ goto bail1;
+ }
+
+ /* Set the TX pointer to where we left off */
+ c2_port->tx_avail = c2_port->tx_ring.count - 1;
+ c2_port->tx_ring.to_use = c2_port->tx_ring.to_clean =
+ c2_port->tx_ring.start + c2dev->cur_tx;
+
+ /* missing: Initialize MAC */
+
+ BUG_ON(c2_port->tx_ring.to_use != c2_port->tx_ring.to_clean);
+
+ /* Reset the adapter, ensures the driver is in sync with the RXP */
+ c2_reset(c2_port);
+
+ /* Reset the READY bit in the sk_buff RXP headers & adapter HRXDQ */
+ for (i = 0, elem = c2_port->rx_ring.start; i < c2_port->rx_ring.count;
+ i++, elem++) {
+ rxp_hdr = (struct c2_rxp_hdr *) elem->skb->data;
+ rxp_hdr->flags = 0;
+ __raw_writew(cpu_to_be16(RXP_HRXD_READY),
+ elem->hw_desc + C2_RXP_FLAGS);
+ }
+
+ /* Enable network packets */
+ netif_start_queue(netdev);
+
+ /* Enable IRQ */
+ writel(0, c2dev->regs + C2_IDIS);
+ netimr0 = readl(c2dev->regs + C2_NIMR0);
+ netimr0 &= ~(C2_PCI_HTX_INT | C2_PCI_HRX_INT);
+ writel(netimr0, c2dev->regs + C2_NIMR0);
+
+ /* Tell the stack to ignore arp requests for ipaddrs bound to
+ * other interfaces. This is needed to prevent the host stack
+ * from responding to arp requests to the ipaddr bound on the
+ * rdma interface.
+ */
+ in_dev = in_dev_get(netdev);
+ in_dev->cnf.arp_ignore = 1;
+ in_dev_put(in_dev);
+
+ return 0;
+
+ bail1:
+ c2_rx_clean(c2_port);
+ kfree(c2_port->rx_ring.start);
+
+ bail0:
+ pci_free_consistent(c2dev->pcidev, c2_port->mem_size, c2_port->mem,
+ c2_port->dma);
+
+ return ret;
+}
+
+static int c2_down(struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+ struct c2_dev *c2dev = c2_port->c2dev;
+
+ if (netif_msg_ifdown(c2_port))
+ pr_debug("%s: disabling interface\n",
+ netdev->name);
+
+ /* Wait for all the queued packets to get sent */
+ c2_tx_interrupt(netdev);
+
+ /* Disable network packets */
+ netif_stop_queue(netdev);
+
+ /* Disable IRQs by clearing the interrupt mask */
+ writel(1, c2dev->regs + C2_IDIS);
+ writel(0, c2dev->regs + C2_NIMR0);
+
+ /* missing: Stop transmitter */
+
+ /* missing: Stop receiver */
+
+ /* Reset the adapter, ensures the driver is in sync with the RXP */
+ c2_reset(c2_port);
+
+ /* missing: Turn off LEDs here */
+
+ /* Free all buffers in the host descriptor rings */
+ c2_tx_clean(c2_port);
+ c2_rx_clean(c2_port);
+
+ /* Free the host descriptor rings */
+ kfree(c2_port->rx_ring.start);
+ kfree(c2_port->tx_ring.start);
+ pci_free_consistent(c2dev->pcidev, c2_port->mem_size, c2_port->mem,
+ c2_port->dma);
+
+ return 0;
+}
+
+static void c2_reset(struct c2_port *c2_port)
+{
+ struct c2_dev *c2dev = c2_port->c2dev;
+ unsigned int cur_rx = c2dev->cur_rx;
+
+ /* Tell the hardware to quiesce */
+ C2_SET_CUR_RX(c2dev, cur_rx | C2_PCI_HRX_QUI);
+
+ /*
+ * The hardware will reset the C2_PCI_HRX_QUI bit once
+ * the RXP is quiesced. Wait 2 seconds for this.
+ */
+ ssleep(2);
+
+ cur_rx = C2_GET_CUR_RX(c2dev);
+
+ if (cur_rx & C2_PCI_HRX_QUI)
+ pr_debug("c2_reset: failed to quiesce the hardware!\n");
+
+ cur_rx &= ~C2_PCI_HRX_QUI;
+
+ c2dev->cur_rx = cur_rx;
+
+ pr_debug("Current RX: %u\n", c2dev->cur_rx);
+}
+
+static int c2_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+ struct c2_dev *c2dev = c2_port->c2dev;
+ struct c2_ring *tx_ring = &c2_port->tx_ring;
+ struct c2_element *elem;
+ dma_addr_t mapaddr;
+ u32 maplen;
+ unsigned long flags;
+ unsigned int i;
+
+ spin_lock_irqsave(&c2_port->tx_lock, flags);
+
+ if (unlikely(c2_port->tx_avail < (skb_shinfo(skb)->nr_frags + 1))) {
+ netif_stop_queue(netdev);
+ spin_unlock_irqrestore(&c2_port->tx_lock, flags);
+
+ pr_debug("%s: Tx ring full when queue awake!\n",
+ netdev->name);
+ return NETDEV_TX_BUSY;
+ }
+
+ maplen = skb_headlen(skb);
+ mapaddr =
+ pci_map_single(c2dev->pcidev, skb->data, maplen, PCI_DMA_TODEVICE);
+
+ elem = tx_ring->to_use;
+ elem->skb = skb;
+ elem->mapaddr = mapaddr;
+ elem->maplen = maplen;
+
+ /* Tell HW to xmit */
+ __raw_writeq(cpu_to_be64(mapaddr), elem->hw_desc + C2_TXP_ADDR);
+ __raw_writew(cpu_to_be16(maplen), elem->hw_desc + C2_TXP_LEN);
+ __raw_writew(cpu_to_be16(TXP_HTXD_READY), elem->hw_desc + C2_TXP_FLAGS);
+
+ c2_port->netstats.tx_packets++;
+ c2_port->netstats.tx_bytes += maplen;
+
+ /* Loop thru additional data fragments and queue them */
+ if (skb_shinfo(skb)->nr_frags) {
+ for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+ skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+ maplen = frag->size;
+ mapaddr =
+ pci_map_page(c2dev->pcidev, frag->page,
+ frag->page_offset, maplen,
+ PCI_DMA_TODEVICE);
+
+ elem = elem->next;
+ elem->skb = NULL;
+ elem->mapaddr = mapaddr;
+ elem->maplen = maplen;
+
+ /* Tell HW to xmit */
+ __raw_writeq(cpu_to_be64(mapaddr),
+ elem->hw_desc + C2_TXP_ADDR);
+ __raw_writew(cpu_to_be16(maplen),
+ elem->hw_desc + C2_TXP_LEN);
+ __raw_writew(cpu_to_be16(TXP_HTXD_READY),
+ elem->hw_desc + C2_TXP_FLAGS);
+
+ c2_port->netstats.tx_packets++;
+ c2_port->netstats.tx_bytes += maplen;
+ }
+ }
+
+ tx_ring->to_use = elem->next;
+ c2_port->tx_avail -= (skb_shinfo(skb)->nr_frags + 1);
+
+ if (c2_port->tx_avail <= MAX_SKB_FRAGS + 1) {
+ netif_stop_queue(netdev);
+ if (netif_msg_tx_queued(c2_port))
+ pr_debug("%s: transmit queue full\n",
+ netdev->name);
+ }
+
+ spin_unlock_irqrestore(&c2_port->tx_lock, flags);
+
+ netdev->trans_start = jiffies;
+
+ return NETDEV_TX_OK;
+}
+
+static struct net_device_stats *c2_get_stats(struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+
+ return &c2_port->netstats;
+}
+
+static void c2_tx_timeout(struct net_device *netdev)
+{
+ struct c2_port *c2_port = netdev_priv(netdev);
+
+ if (netif_msg_timer(c2_port))
+ pr_debug("%s: tx timeout\n", netdev->name);
+
+ c2_tx_clean(c2_port);
+}
+
+static int c2_change_mtu(struct net_device *netdev, int new_mtu)
+{
+ int ret = 0;
+
+ if (new_mtu < ETH_ZLEN || new_mtu > ETH_JUMBO_MTU)
+ return -EINVAL;
+
+ netdev->mtu = new_mtu;
+
+ if (netif_running(netdev)) {
+ c2_down(netdev);
+
+ c2_up(netdev);
+ }
+
+ return ret;
+}
+
+/* Initialize network device */
+static struct net_device *c2_devinit(struct c2_dev *c2dev,
+ void __iomem * mmio_addr)
+{
+ struct c2_port *c2_port = NULL;
+ struct net_device *netdev = alloc_etherdev(sizeof(*c2_port));
+
+ if (!netdev) {
+ pr_debug("c2_port etherdev alloc failed");
+ return NULL;
+ }
+
+ SET_MODULE_OWNER(netdev);
+ SET_NETDEV_DEV(netdev, &c2dev->pcidev->dev);
+
+ netdev->open = c2_up;
+ netdev->stop = c2_down;
+ netdev->hard_start_xmit = c2_xmit_frame;
+ netdev->get_stats = c2_get_stats;
+ netdev->tx_timeout = c2_tx_timeout;
+ netdev->change_mtu = c2_change_mtu;
+ netdev->watchdog_timeo = C2_TX_TIMEOUT;
+ netdev->irq = c2dev->pcidev->irq;
+
+ c2_port = netdev_priv(netdev);
+ c2_port->netdev = netdev;
+ c2_port->c2dev = c2dev;
+ c2_port->msg_enable = netif_msg_init(debug, default_msg);
+ c2_port->tx_ring.count = C2_NUM_TX_DESC;
+ c2_port->rx_ring.count = C2_NUM_RX_DESC;
+
+ spin_lock_init(&c2_port->tx_lock);
+
+ /* Copy our 48-bit ethernet hardware address */
+ memcpy_fromio(netdev->dev_addr, mmio_addr + C2_REGS_ENADDR, 6);
+
+ /* Validate the MAC address */
+ if (!is_valid_ether_addr(netdev->dev_addr)) {
+ pr_debug("Invalid MAC Address\n");
+ c2_print_macaddr(netdev);
+ free_netdev(netdev);
+ return NULL;
+ }
+
+ c2dev->netdev = netdev;
+
+ return netdev;
+}
+
+static int __devinit c2_probe(struct pci_dev *pcidev,
+ const struct pci_device_id *ent)
+{
+ int ret = 0, i;
+ unsigned long reg0_start, reg0_flags, reg0_len;
+ unsigned long reg2_start, reg2_flags, reg2_len;
+ unsigned long reg4_start, reg4_flags, reg4_len;
+ unsigned kva_map_size;
+ struct net_device *netdev = NULL;
+ struct c2_dev *c2dev = NULL;
+ void __iomem *mmio_regs = NULL;
+
+ printk(KERN_INFO PFX "AMSO1100 Gigabit Ethernet driver v%s loaded\n",
+ DRV_VERSION);
+
+ /* Enable PCI device */
+ ret = pci_enable_device(pcidev);
+ if (ret) {
+ printk(KERN_ERR PFX "%s: Unable to enable PCI device\n",
+ pci_name(pcidev));
+ goto bail0;
+ }
+
+ reg0_start = pci_resource_start(pcidev, BAR_0);
+ reg0_len = pci_resource_len(pcidev, BAR_0);
+ reg0_flags = pci_resource_flags(pcidev, BAR_0);
+
+ reg2_start = pci_resource_start(pcidev, BAR_2);
+ reg2_len = pci_resource_len(pcidev, BAR_2);
+ reg2_flags = pci_resource_flags(pcidev, BAR_2);
+
+ reg4_start = pci_resource_start(pcidev, BAR_4);
+ reg4_len = pci_resource_len(pcidev, BAR_4);
+ reg4_flags = pci_resource_flags(pcidev, BAR_4);
+
+ pr_debug("BAR0 size = 0x%lX bytes\n", reg0_len);
+ pr_debug("BAR2 size = 0x%lX bytes\n", reg2_len);
+ pr_debug("BAR4 size = 0x%lX bytes\n", reg4_len);
+
+ /* Make sure PCI base addr are MMIO */
+ if (!(reg0_flags & IORESOURCE_MEM) ||
+ !(reg2_flags & IORESOURCE_MEM) || !(reg4_flags & IORESOURCE_MEM)) {
+ printk(KERN_ERR PFX "PCI regions not an MMIO resource\n");
+ ret = -ENODEV;
+ goto bail1;
+ }
+
+ /* Check for weird/broken PCI region reporting */
+ if ((reg0_len < C2_REG0_SIZE) ||
+ (reg2_len < C2_REG2_SIZE) || (reg4_len < C2_REG4_SIZE)) {
+ printk(KERN_ERR PFX "Invalid PCI region sizes\n");
+ ret = -ENODEV;
+ goto bail1;
+ }
+
+ /* Reserve PCI I/O and memory resources */
+ ret = pci_request_regions(pcidev, DRV_NAME);
+ if (ret) {
+ printk(KERN_ERR PFX "%s: Unable to request regions\n",
+ pci_name(pcidev));
+ goto bail1;
+ }
+
+ if ((sizeof(dma_addr_t) > 4)) {
+ ret = pci_set_dma_mask(pcidev, DMA_64BIT_MASK);
+ if (ret < 0) {
+ printk(KERN_ERR PFX "64b DMA configuration failed\n");
+ goto bail2;
+ }
+ } else {
+ ret = pci_set_dma_mask(pcidev, DMA_32BIT_MASK);
+ if (ret < 0) {
+ printk(KERN_ERR PFX "32b DMA configuration failed\n");
+ goto bail2;
+ }
+ }
+
+ /* Enables bus-mastering on the device */
+ pci_set_master(pcidev);
+
+ /* Remap the adapter PCI registers in BAR4 */
+ mmio_regs = ioremap_nocache(reg4_start + C2_PCI_REGS_OFFSET,
+ sizeof(struct c2_adapter_pci_regs));
+ if (mmio_regs == 0UL) {
+ printk(KERN_ERR PFX
+ "Unable to remap adapter PCI registers in BAR4\n");
+ ret = -EIO;
+ goto bail2;
+ }
+
+ /* Validate PCI regs magic */
+ for (i = 0; i < sizeof(c2_magic); i++) {
+ if (c2_magic[i] != readb(mmio_regs + C2_REGS_MAGIC + i)) {
+ printk(KERN_ERR PFX "Downlevel Firmware boot loader "
+ "[%d/%Zd: got 0x%x, exp 0x%x]. Use the cc_flash "
+ "utility to update your boot loader\n",
+ i + 1, sizeof(c2_magic),
+ readb(mmio_regs + C2_REGS_MAGIC + i),
+ c2_magic[i]);
+ printk(KERN_ERR PFX "Adapter not claimed\n");
+ iounmap(mmio_regs);
+ ret = -EIO;
+ goto bail2;
+ }
+ }
+
+ /* Validate the adapter version */
+ if (be32_to_cpu(readl(mmio_regs + C2_REGS_VERS)) != C2_VERSION) {
+ printk(KERN_ERR PFX "Version mismatch "
+ "[fw=%u, c2=%u], Adapter not claimed\n",
+ be32_to_cpu(readl(mmio_regs + C2_REGS_VERS)),
+ C2_VERSION);
+ ret = -EINVAL;
+ iounmap(mmio_regs);
+ goto bail2;
+ }
+
+ /* Validate the adapter IVN */
+ if (be32_to_cpu(readl(mmio_regs + C2_REGS_IVN)) != C2_IVN) {
+ printk(KERN_ERR PFX "Downlevel FIrmware level. You should be using "
+ "the OpenIB device support kit. "
+ "[fw=0x%x, c2=0x%x], Adapter not claimed\n",
+ be32_to_cpu(readl(mmio_regs + C2_REGS_IVN)),
+ C2_IVN);
+ ret = -EINVAL;
+ iounmap(mmio_regs);
+ goto bail2;
+ }
+
+ /* Allocate hardware structure */
+ c2dev = (struct c2_dev *) ib_alloc_device(sizeof(*c2dev));
+ if (!c2dev) {
+ printk(KERN_ERR PFX "%s: Unable to alloc hardware struct\n",
+ pci_name(pcidev));
+ ret = -ENOMEM;
+ iounmap(mmio_regs);
+ goto bail2;
+ }
+
+ memset(c2dev, 0, sizeof(*c2dev));
+ spin_lock_init(&c2dev->lock);
+ c2dev->pcidev = pcidev;
+ c2dev->cur_tx = 0;
+
+ /* Get the last RX index */
+ c2dev->cur_rx =
+ (be32_to_cpu(readl(mmio_regs + C2_REGS_HRX_CUR)) -
+ 0xffffc000) / sizeof(struct c2_rxp_desc);
+
+ /* Request an interrupt line for the driver */
+ ret = request_irq(pcidev->irq, c2_interrupt, SA_SHIRQ, DRV_NAME, c2dev);
+ if (ret) {
+ printk(KERN_ERR PFX "%s: requested IRQ %u is busy\n",
+ pci_name(pcidev), pcidev->irq);
+ iounmap(mmio_regs);
+ goto bail3;
+ }
+
+ /* Set driver specific data */
+ pci_set_drvdata(pcidev, c2dev);
+
+ /* Initialize network device */
+ if ((netdev = c2_devinit(c2dev, mmio_regs)) == NULL) {
+ iounmap(mmio_regs);
+ goto bail4;
+ }
+
+ /* Save off the actual size prior to unmapping mmio_regs */
+ kva_map_size = be32_to_cpu(readl(mmio_regs + C2_REGS_PCI_WINSIZE));
+
+ /* Unmap the adapter PCI registers in BAR4 */
+ iounmap(mmio_regs);
+
+ /* Register network device */
+ ret = register_netdev(netdev);
+ if (ret) {
+ printk(KERN_ERR PFX "Unable to register netdev, ret = %d\n",
+ ret);
+ goto bail5;
+ }
+
+ /* Disable network packets */
+ netif_stop_queue(netdev);
+
+ /* Remap the adapter HRXDQ PA space to kernel VA space */
+ c2dev->mmio_rxp_ring = ioremap_nocache(reg4_start + C2_RXP_HRXDQ_OFFSET,
+ C2_RXP_HRXDQ_SIZE);
+ if (c2dev->mmio_rxp_ring == 0UL) {
+ printk(KERN_ERR PFX "Unable to remap MMIO HRXDQ region\n");
+ ret = -EIO;
+ goto bail6;
+ }
+
+ /* Remap the adapter HTXDQ PA space to kernel VA space */
+ c2dev->mmio_txp_ring = ioremap_nocache(reg4_start + C2_TXP_HTXDQ_OFFSET,
+ C2_TXP_HTXDQ_SIZE);
+ if (c2dev->mmio_txp_ring == 0UL) {
+ printk(KERN_ERR PFX "Unable to remap MMIO HTXDQ region\n");
+ ret = -EIO;
+ goto bail7;
+ }
+
+ /* Save off the current RX index in the last 4 bytes of the TXP Ring */
+ C2_SET_CUR_RX(c2dev, c2dev->cur_rx);
+
+ /* Remap the PCI registers in adapter BAR0 to kernel VA space */
+ c2dev->regs = ioremap_nocache(reg0_start, reg0_len);
+ if (c2dev->regs == 0UL) {
+ printk(KERN_ERR PFX "Unable to remap BAR0\n");
+ ret = -EIO;
+ goto bail8;
+ }
+
+ /* Remap the PCI registers in adapter BAR4 to kernel VA space */
+ c2dev->pa = reg4_start + C2_PCI_REGS_OFFSET;
+ c2dev->kva = ioremap_nocache(reg4_start + C2_PCI_REGS_OFFSET,
+ kva_map_size);
+ if (c2dev->kva == 0UL) {
+ printk(KERN_ERR PFX "Unable to remap BAR4\n");
+ ret = -EIO;
+ goto bail9;
+ }
+
+ /* Print out the MAC address */
+ c2_print_macaddr(netdev);
+
+ ret = c2_rnic_init(c2dev);
+ if (ret) {
+ printk(KERN_ERR PFX "c2_rnic_init failed: %d\n", ret);
+ goto bail10;
+ }
+
+ c2_register_device(c2dev);
+
+ return 0;
+
+ bail10:
+ iounmap(c2dev->kva);
+
+ bail9:
+ iounmap(c2dev->regs);
+
+ bail8:
+ iounmap(c2dev->mmio_txp_ring);
+
+ bail7:
+ iounmap(c2dev->mmio_rxp_ring);
+
+ bail6:
+ unregister_netdev(netdev);
+
+ bail5:
+ free_netdev(netdev);
+
+ bail4:
+ free_irq(pcidev->irq, c2dev);
+
+ bail3:
+ ib_dealloc_device(&c2dev->ibdev);
+
+ bail2:
+ pci_release_regions(pcidev);
+
+ bail1:
+ pci_disable_device(pcidev);
+
+ bail0:
+ return ret;
+}
+
+static void __devexit c2_remove(struct pci_dev *pcidev)
+{
+ struct c2_dev *c2dev = pci_get_drvdata(pcidev);
+ struct net_device *netdev = c2dev->netdev;
+
+ /* Unregister with OpenIB */
+ c2_unregister_device(c2dev);
+
+ /* Clean up the RNIC resources */
+ c2_rnic_term(c2dev);
+
+ /* Remove network device from the kernel */
+ unregister_netdev(netdev);
+
+ /* Free network device */
+ free_netdev(netdev);
+
+ /* Free the interrupt line */
+ free_irq(pcidev->irq, c2dev);
+
+ /* missing: Turn LEDs off here */
+
+ /* Unmap adapter PA space */
+ iounmap(c2dev->kva);
+ iounmap(c2dev->regs);
+ iounmap(c2dev->mmio_txp_ring);
+ iounmap(c2dev->mmio_rxp_ring);
+
+ /* Free the hardware structure */
+ ib_dealloc_device(&c2dev->ibdev);
+
+ /* Release reserved PCI I/O and memory resources */
+ pci_release_regions(pcidev);
+
+ /* Disable PCI device */
+ pci_disable_device(pcidev);
+
+ /* Clear driver specific data */
+ pci_set_drvdata(pcidev, NULL);
+}
+
+static struct pci_driver c2_pci_driver = {
+ .name = DRV_NAME,
+ .id_table = c2_pci_table,
+ .probe = c2_probe,
+ .remove = __devexit_p(c2_remove),
+};
+
+static int __init c2_init_module(void)
+{
+ return pci_module_init(&c2_pci_driver);
+}
+
+static void __exit c2_exit_module(void)
+{
+ pci_unregister_driver(&c2_pci_driver);
+}
+
+module_init(c2_init_module);
+module_exit(c2_exit_module);
diff --git a/drivers/infiniband/hw/amso1100/c2.h b/drivers/infiniband/hw/amso1100/c2.h
new file mode 100644
index 0000000..3251e8f
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2.h
@@ -0,0 +1,555 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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 __C2_H
+#define __C2_H
+
+#include <linux/netdevice.h>
+#include <linux/spinlock.h>
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <asm/semaphore.h>
+
+#include "c2_provider.h"
+#include "c2_mq.h"
+#include "c2_status.h"
+
+#define DRV_NAME "c2"
+#define DRV_VERSION "1.1"
+#define PFX DRV_NAME ": "
+
+#define BAR_0 0
+#define BAR_2 2
+#define BAR_4 4
+
+#define RX_BUF_SIZE (1536 + 8)
+#define ETH_JUMBO_MTU 9000
+#define C2_MAGIC "CEPHEUS"
+#define C2_VERSION 4
+#define C2_IVN (18 & 0x7fffffff)
+
+#define C2_REG0_SIZE (16 * 1024)
+#define C2_REG2_SIZE (2 * 1024 * 1024)
+#define C2_REG4_SIZE (256 * 1024 * 1024)
+#define C2_NUM_TX_DESC 341
+#define C2_NUM_RX_DESC 256
+#define C2_PCI_REGS_OFFSET (0x10000)
+#define C2_RXP_HRXDQ_OFFSET (((C2_REG4_SIZE)/2))
+#define C2_RXP_HRXDQ_SIZE (4096)
+#define C2_TXP_HTXDQ_OFFSET (((C2_REG4_SIZE)/2) + C2_RXP_HRXDQ_SIZE)
+#define C2_TXP_HTXDQ_SIZE (4096)
+#define C2_TX_TIMEOUT (6*HZ)
+
+/* CEPHEUS */
+static const u8 c2_magic[] = {
+ 0x43, 0x45, 0x50, 0x48, 0x45, 0x55, 0x53
+};
+
+enum adapter_pci_regs {
+ C2_REGS_MAGIC = 0x0000,
+ C2_REGS_VERS = 0x0008,
+ C2_REGS_IVN = 0x000C,
+ C2_REGS_PCI_WINSIZE = 0x0010,
+ C2_REGS_Q0_QSIZE = 0x0014,
+ C2_REGS_Q0_MSGSIZE = 0x0018,
+ C2_REGS_Q0_POOLSTART = 0x001C,
+ C2_REGS_Q0_SHARED = 0x0020,
+ C2_REGS_Q1_QSIZE = 0x0024,
+ C2_REGS_Q1_MSGSIZE = 0x0028,
+ C2_REGS_Q1_SHARED = 0x0030,
+ C2_REGS_Q2_QSIZE = 0x0034,
+ C2_REGS_Q2_MSGSIZE = 0x0038,
+ C2_REGS_Q2_SHARED = 0x0040,
+ C2_REGS_ENADDR = 0x004C,
+ C2_REGS_RDMA_ENADDR = 0x0054,
+ C2_REGS_HRX_CUR = 0x006C,
+};
+
+struct c2_adapter_pci_regs {
+ char reg_magic[8];
+ u32 version;
+ u32 ivn;
+ u32 pci_window_size;
+ u32 q0_q_size;
+ u32 q0_msg_size;
+ u32 q0_pool_start;
+ u32 q0_shared;
+ u32 q1_q_size;
+ u32 q1_msg_size;
+ u32 q1_pool_start;
+ u32 q1_shared;
+ u32 q2_q_size;
+ u32 q2_msg_size;
+ u32 q2_pool_start;
+ u32 q2_shared;
+ u32 log_start;
+ u32 log_size;
+ u8 host_enaddr[8];
+ u8 rdma_enaddr[8];
+ u32 crash_entry;
+ u32 crash_ready[2];
+ u32 fw_txd_cur;
+ u32 fw_hrxd_cur;
+ u32 fw_rxd_cur;
+};
+
+enum pci_regs {
+ C2_HISR = 0x0000,
+ C2_DISR = 0x0004,
+ C2_HIMR = 0x0008,
+ C2_DIMR = 0x000C,
+ C2_NISR0 = 0x0010,
+ C2_NISR1 = 0x0014,
+ C2_NIMR0 = 0x0018,
+ C2_NIMR1 = 0x001C,
+ C2_IDIS = 0x0020,
+};
+
+enum {
+ C2_PCI_HRX_INT = 1 << 8,
+ C2_PCI_HTX_INT = 1 << 17,
+ C2_PCI_HRX_QUI = 1 << 31,
+};
+
+/*
+ * Cepheus registers in BAR0.
+ */
+struct c2_pci_regs {
+ u32 hostisr;
+ u32 dmaisr;
+ u32 hostimr;
+ u32 dmaimr;
+ u32 netisr0;
+ u32 netisr1;
+ u32 netimr0;
+ u32 netimr1;
+ u32 int_disable;
+};
+
+/* TXP flags */
+enum c2_txp_flags {
+ TXP_HTXD_DONE = 0,
+ TXP_HTXD_READY = 1 << 0,
+ TXP_HTXD_UNINIT = 1 << 1,
+};
+
+/* RXP flags */
+enum c2_rxp_flags {
+ RXP_HRXD_UNINIT = 0,
+ RXP_HRXD_READY = 1 << 0,
+ RXP_HRXD_DONE = 1 << 1,
+};
+
+/* RXP status */
+enum c2_rxp_status {
+ RXP_HRXD_ZERO = 0,
+ RXP_HRXD_OK = 1 << 0,
+ RXP_HRXD_BUF_OV = 1 << 1,
+};
+
+/* TXP descriptor fields */
+enum txp_desc {
+ C2_TXP_FLAGS = 0x0000,
+ C2_TXP_LEN = 0x0002,
+ C2_TXP_ADDR = 0x0004,
+};
+
+/* RXP descriptor fields */
+enum rxp_desc {
+ C2_RXP_FLAGS = 0x0000,
+ C2_RXP_STATUS = 0x0002,
+ C2_RXP_COUNT = 0x0004,
+ C2_RXP_LEN = 0x0006,
+ C2_RXP_ADDR = 0x0008,
+};
+
+struct c2_txp_desc {
+ u16 flags;
+ u16 len;
+ u64 addr;
+} __attribute__ ((packed));
+
+struct c2_rxp_desc {
+ u16 flags;
+ u16 status;
+ u16 count;
+ u16 len;
+ u64 addr;
+} __attribute__ ((packed));
+
+struct c2_rxp_hdr {
+ u16 flags;
+ u16 status;
+ u16 len;
+ u16 rsvd;
+} __attribute__ ((packed));
+
+struct c2_tx_desc {
+ u32 len;
+ u32 status;
+ dma_addr_t next_offset;
+};
+
+struct c2_rx_desc {
+ u32 len;
+ u32 status;
+ dma_addr_t next_offset;
+};
+
+struct c2_alloc {
+ u32 last;
+ u32 max;
+ spinlock_t lock;
+ unsigned long *table;
+};
+
+struct c2_array {
+ struct {
+ void **page;
+ int used;
+ } *page_list;
+};
+
+/*
+ * The MQ shared pointer pool is organized as a linked list of
+ * chunks. Each chunk contains a linked list of free shared pointers
+ * that can be allocated to a given user mode client.
+ *
+ */
+struct sp_chunk {
+ struct sp_chunk *next;
+ gfp_t gfp_mask;
+ u16 head;
+ u16 shared_ptr[0];
+};
+
+struct c2_pd_table {
+ struct c2_alloc alloc;
+ struct c2_array pd;
+};
+
+struct c2_qp_table {
+ struct c2_alloc alloc;
+ spinlock_t lock;
+ struct c2_array qp;
+ struct c2_qp** map;
+};
+
+struct c2_element {
+ struct c2_element *next;
+ void *ht_desc; /* host descriptor */
+ void __iomem *hw_desc; /* hardware descriptor */
+ struct sk_buff *skb;
+ dma_addr_t mapaddr;
+ u32 maplen;
+};
+
+struct c2_ring {
+ struct c2_element *to_clean;
+ struct c2_element *to_use;
+ struct c2_element *start;
+ unsigned long count;
+};
+
+struct c2_dev {
+ struct ib_device ibdev;
+ void __iomem *regs;
+ void __iomem *mmio_txp_ring; /* remapped adapter memory for hw rings */
+ void __iomem *mmio_rxp_ring;
+ spinlock_t lock;
+ struct pci_dev *pcidev;
+ struct net_device *netdev;
+ struct net_device *pseudo_netdev;
+ unsigned int cur_tx;
+ unsigned int cur_rx;
+ u32 adapter_handle;
+ int device_cap_flags;
+ void __iomem *kva; /* KVA device memory */
+ unsigned long pa; /* PA device memory */
+ void **qptr_array;
+
+ kmem_cache_t *host_msg_cache;
+
+ struct list_head cca_link; /* adapter list */
+ struct list_head eh_wakeup_list; /* event wakeup list */
+ wait_queue_head_t req_vq_wo;
+
+ /* Cached RNIC properties */
+ struct ib_device_attr props;
+
+ struct c2_pd_table pd_table;
+ struct c2_qp_table qp_table;
+ int ports; /* num of GigE ports */
+ int devnum;
+ spinlock_t vqlock; /* sync vbs req MQ */
+
+ /* Verbs Queues */
+ struct c2_mq req_vq; /* Verbs Request MQ */
+ struct c2_mq rep_vq; /* Verbs Reply MQ */
+ struct c2_mq aeq; /* Async Events MQ */
+
+ /* Kernel client MQs */
+ struct sp_chunk *kern_mqsp_pool;
+
+ /* Device updates these values when posting messages to a host
+ * target queue */
+ u16 req_vq_shared;
+ u16 rep_vq_shared;
+ u16 aeq_shared;
+ u16 irq_claimed;
+
+ /*
+ * Shared host target pages for user-accessible MQs.
+ */
+ int hthead; /* index of first free entry */
+ void *htpages; /* kernel vaddr */
+ int htlen; /* length of htpages memory */
+ void *htuva; /* user mapped vaddr */
+ spinlock_t htlock; /* serialize allocation */
+
+ u64 adapter_hint_uva; /* access to the activity FIFO */
+
+ // spinlock_t aeq_lock;
+ // spinlock_t rnic_lock;
+
+ u16 hint_count;
+ u16 hints_read;
+
+ int init; /* TRUE if it's ready */
+ char ae_cache_name[16];
+ char vq_cache_name[16];
+};
+
+struct c2_port {
+ u32 msg_enable;
+ struct c2_dev *c2dev;
+ struct net_device *netdev;
+
+ spinlock_t tx_lock;
+ u32 tx_avail;
+ struct c2_ring tx_ring;
+ struct c2_ring rx_ring;
+
+ void *mem; /* PCI memory for host rings */
+ dma_addr_t dma;
+ unsigned long mem_size;
+
+ u32 rx_buf_size;
+
+ struct net_device_stats netstats;
+};
+
+/*
+ * Activity FIFO registers in BAR0.
+ */
+#define PCI_BAR0_HOST_HINT 0x100
+#define PCI_BAR0_ADAPTER_HINT 0x2000
+
+/*
+ * Ammasso PCI vendor id and Cepheus PCI device id.
+ */
+#define CQ_ARMED 0x01
+#define CQ_WAIT_FOR_DMA 0x80
+
+/*
+ * The format of a hint is as follows:
+ * Lower 16 bits are the count of hints for the queue.
+ * Next 15 bits are the qp_index
+ * Upper most bit depends on who reads it:
+ * If read by producer, then it means Full (1) or Not-Full (0)
+ * If read by consumer, then it means Empty (1) or Not-Empty (0)
+ */
+#define C2_HINT_MAKE(q_index, hint_count) (((q_index) << 16) | hint_count)
+#define C2_HINT_GET_INDEX(hint) (((hint) & 0x7FFF0000) >> 16)
+#define C2_HINT_GET_COUNT(hint) ((hint) & 0x0000FFFF)
+
+
+/*
+ * The following defines the offset in SDRAM for the c2_adapter_pci_regs_t
+ * struct.
+ */
+#define C2_ADAPTER_PCI_REGS_OFFSET 0x10000
+
+#ifndef readq
+static inline u64 readq(const void __iomem * addr)
+{
+ u64 ret = readl(addr + 4);
+ ret <<= 32;
+ ret |= readl(addr);
+
+ return ret;
+}
+#endif
+
+#ifndef __raw_writeq
+static inline void __raw_writeq(u64 val, void __iomem * addr)
+{
+ __raw_writel((u32) (val), addr);
+ __raw_writel((u32) (val >> 32), (addr + 4));
+}
+#endif
+
+#define C2_SET_CUR_RX(c2dev, cur_rx) \
+ __raw_writel(cpu_to_be32(cur_rx), c2dev->mmio_txp_ring + 4092)
+
+#define C2_GET_CUR_RX(c2dev) \
+ be32_to_cpu(readl(c2dev->mmio_txp_ring + 4092))
+
+static inline struct c2_dev *to_c2dev(struct ib_device *ibdev)
+{
+ return container_of(ibdev, struct c2_dev, ibdev);
+}
+
+static inline int c2_errno(void *reply)
+{
+ switch (c2_wr_get_result(reply)) {
+ case C2_OK:
+ return 0;
+ case CCERR_NO_BUFS:
+ case CCERR_INSUFFICIENT_RESOURCES:
+ case CCERR_ZERO_RDMA_READ_RESOURCES:
+ return -ENOMEM;
+ case CCERR_MR_IN_USE:
+ case CCERR_QP_IN_USE:
+ return -EBUSY;
+ case CCERR_ADDR_IN_USE:
+ return -EADDRINUSE;
+ case CCERR_ADDR_NOT_AVAIL:
+ return -EADDRNOTAVAIL;
+ case CCERR_CONN_RESET:
+ return -ECONNRESET;
+ case CCERR_NOT_IMPLEMENTED:
+ case CCERR_INVALID_WQE:
+ return -ENOSYS;
+ case CCERR_QP_NOT_PRIVILEGED:
+ return -EPERM;
+ case CCERR_STACK_ERROR:
+ return -EPROTO;
+ case CCERR_ACCESS_VIOLATION:
+ case CCERR_BASE_AND_BOUNDS_VIOLATION:
+ return -EFAULT;
+ case CCERR_STAG_STATE_NOT_INVALID:
+ case CCERR_INVALID_ADDRESS:
+ case CCERR_INVALID_CQ:
+ case CCERR_INVALID_EP:
+ case CCERR_INVALID_MODIFIER:
+ case CCERR_INVALID_MTU:
+ case CCERR_INVALID_PD_ID:
+ case CCERR_INVALID_QP:
+ case CCERR_INVALID_RNIC:
+ case CCERR_INVALID_STAG:
+ return -EINVAL;
+ default:
+ return -EAGAIN;
+ }
+}
+
+/* Device */
+extern int c2_register_device(struct c2_dev *c2dev);
+extern void c2_unregister_device(struct c2_dev *c2dev);
+extern int c2_rnic_init(struct c2_dev *c2dev);
+extern void c2_rnic_term(struct c2_dev *c2dev);
+extern void c2_rnic_interrupt(struct c2_dev *c2dev);
+extern int c2_rnic_query(struct c2_dev *c2dev, struct ib_device_attr *props);
+extern int c2_del_addr(struct c2_dev *c2dev, u32 inaddr, u32 inmask);
+extern int c2_add_addr(struct c2_dev *c2dev, u32 inaddr, u32 inmask);
+
+/* QPs */
+extern int c2_alloc_qp(struct c2_dev *c2dev, struct c2_pd *pd,
+ struct ib_qp_init_attr *qp_attrs, struct c2_qp *qp);
+extern void c2_free_qp(struct c2_dev *c2dev, struct c2_qp *qp);
+extern struct ib_qp *c2_get_qp(struct ib_device *device, int qpn);
+extern int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
+ struct ib_qp_attr *attr, int attr_mask);
+extern int c2_qp_set_read_limits(struct c2_dev *c2dev, struct c2_qp *qp,
+ int ord, int ird);
+extern int c2_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
+ struct ib_send_wr **bad_wr);
+extern int c2_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr,
+ struct ib_recv_wr **bad_wr);
+extern int __devinit c2_init_qp_table(struct c2_dev *c2dev);
+extern void __devexit c2_cleanup_qp_table(struct c2_dev *c2dev);
+extern void c2_set_qp_state(struct c2_qp *, int);
+
+/* PDs */
+extern int c2_pd_alloc(struct c2_dev *c2dev, int privileged, struct c2_pd *pd);
+extern void c2_pd_free(struct c2_dev *c2dev, struct c2_pd *pd);
+extern int __devinit c2_init_pd_table(struct c2_dev *c2dev);
+extern void __devexit c2_cleanup_pd_table(struct c2_dev *c2dev);
+
+/* CQs */
+extern int c2_init_cq(struct c2_dev *c2dev, int entries,
+ struct c2_ucontext *ctx, struct c2_cq *cq);
+extern void c2_free_cq(struct c2_dev *c2dev, struct c2_cq *cq);
+extern void c2_cq_event(struct c2_dev *c2dev, u32 mq_index);
+extern void c2_cq_clean(struct c2_dev *c2dev, struct c2_qp *qp, u32 mq_index);
+extern int c2_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
+extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify);
+
+/* CM */
+extern int c2_llp_connect(struct iw_cm_id *cm_id,
+ struct iw_cm_conn_param *iw_param);
+extern int c2_llp_accept(struct iw_cm_id *cm_id,
+ struct iw_cm_conn_param *iw_param);
+extern int c2_llp_reject(struct iw_cm_id *cm_id, const void *pdata,
+ u8 pdata_len);
+extern int c2_llp_service_create(struct iw_cm_id *cm_id, int backlog);
+extern int c2_llp_service_destroy(struct iw_cm_id *cm_id);
+
+/* MM */
+extern int c2_nsmr_register_phys_kern(struct c2_dev *c2dev, u64 *addr_list,
+ int page_size, int pbl_depth, u32 length,
+ u32 off, u64 *va, enum c2_acf acf,
+ struct c2_mr *mr);
+extern int c2_stag_dealloc(struct c2_dev *c2dev, u32 stag_index);
+
+/* AE */
+extern void c2_ae_event(struct c2_dev *c2dev, u32 mq_index);
+
+/* Allocators */
+extern u32 c2_alloc(struct c2_alloc *alloc);
+extern void c2_free(struct c2_alloc *alloc, u32 obj);
+extern int c2_alloc_init(struct c2_alloc *alloc, u32 num, u32 reserved);
+extern void c2_alloc_cleanup(struct c2_alloc *alloc);
+extern int c2_init_mqsp_pool(gfp_t gfp_mask, struct sp_chunk **root);
+extern void c2_free_mqsp_pool(struct sp_chunk *root);
+extern u16 *c2_alloc_mqsp(struct sp_chunk *head);
+extern void c2_free_mqsp(u16 * mqsp);
+extern void c2_array_cleanup(struct c2_array *array, int nent);
+extern int c2_array_init(struct c2_array *array, int nent);
+extern void c2_array_clear(struct c2_array *array, int index);
+extern int c2_array_set(struct c2_array *array, int index, void *value);
+extern void *c2_array_get(struct c2_array *array, int index);
+
+#endif
diff --git a/drivers/infiniband/hw/amso1100/c2_ae.c b/drivers/infiniband/hw/amso1100/c2_ae.c
new file mode 100644
index 0000000..c979ef6
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2_ae.c
@@ -0,0 +1,359 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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 "c2.h"
+#include <rdma/iw_cm.h>
+#include "c2_status.h"
+#include "c2_ae.h"
+
+static int c2_convert_cm_status(u32 c2_status)
+{
+ switch (c2_status) {
+ case C2_CONN_STATUS_SUCCESS:
+ return 0;
+ case C2_CONN_STATUS_REJECTED:
+ return -ENETRESET;
+ case C2_CONN_STATUS_REFUSED:
+ return -ECONNREFUSED;
+ case C2_CONN_STATUS_TIMEDOUT:
+ return -ETIMEDOUT;
+ case C2_CONN_STATUS_NETUNREACH:
+ return -ENETUNREACH;
+ case C2_CONN_STATUS_HOSTUNREACH:
+ return -EHOSTUNREACH;
+ case C2_CONN_STATUS_INVALID_RNIC:
+ return -EINVAL;
+ case C2_CONN_STATUS_INVALID_QP:
+ return -EINVAL;
+ case C2_CONN_STATUS_INVALID_QP_STATE:
+ return -EINVAL;
+ case C2_CONN_STATUS_ADDR_NOT_AVAIL:
+ return -EADDRNOTAVAIL;
+ default:
+ printk(KERN_ERR PFX
+ "%s - Unable to convert CM status: %d\n",
+ __FUNCTION__, c2_status);
+ return -EIO;
+ }
+}
+
+#ifdef DEBUG
+static const char* to_event_str(int event)
+{
+ static const char* event_str[] = {
+ "CCAE_REMOTE_SHUTDOWN",
+ "CCAE_ACTIVE_CONNECT_RESULTS",
+ "CCAE_CONNECTION_REQUEST",
+ "CCAE_LLP_CLOSE_COMPLETE",
+ "CCAE_TERMINATE_MESSAGE_RECEIVED",
+ "CCAE_LLP_CONNECTION_RESET",
+ "CCAE_LLP_CONNECTION_LOST",
+ "CCAE_LLP_SEGMENT_SIZE_INVALID",
+ "CCAE_LLP_INVALID_CRC",
+ "CCAE_LLP_BAD_FPDU",
+ "CCAE_INVALID_DDP_VERSION",
+ "CCAE_INVALID_RDMA_VERSION",
+ "CCAE_UNEXPECTED_OPCODE",
+ "CCAE_INVALID_DDP_QUEUE_NUMBER",
+ "CCAE_RDMA_READ_NOT_ENABLED",
+ "CCAE_RDMA_WRITE_NOT_ENABLED",
+ "CCAE_RDMA_READ_TOO_SMALL",
+ "CCAE_NO_L_BIT",
+ "CCAE_TAGGED_INVALID_STAG",
+ "CCAE_TAGGED_BASE_BOUNDS_VIOLATION",
+ "CCAE_TAGGED_ACCESS_RIGHTS_VIOLATION",
+ "CCAE_TAGGED_INVALID_PD",
+ "CCAE_WRAP_ERROR",
+ "CCAE_BAD_CLOSE",
+ "CCAE_BAD_LLP_CLOSE",
+ "CCAE_INVALID_MSN_RANGE",
+ "CCAE_INVALID_MSN_GAP",
+ "CCAE_IRRQ_OVERFLOW",
+ "CCAE_IRRQ_MSN_GAP",
+ "CCAE_IRRQ_MSN_RANGE",
+ "CCAE_IRRQ_INVALID_STAG",
+ "CCAE_IRRQ_BASE_BOUNDS_VIOLATION",
+ "CCAE_IRRQ_ACCESS_RIGHTS_VIOLATION",
+ "CCAE_IRRQ_INVALID_PD",
+ "CCAE_IRRQ_WRAP_ERROR",
+ "CCAE_CQ_SQ_COMPLETION_OVERFLOW",
+ "CCAE_CQ_RQ_COMPLETION_ERROR",
+ "CCAE_QP_SRQ_WQE_ERROR",
+ "CCAE_QP_LOCAL_CATASTROPHIC_ERROR",
+ "CCAE_CQ_OVERFLOW",
+ "CCAE_CQ_OPERATION_ERROR",
+ "CCAE_SRQ_LIMIT_REACHED",
+ "CCAE_QP_RQ_LIMIT_REACHED",
+ "CCAE_SRQ_CATASTROPHIC_ERROR",
+ "CCAE_RNIC_CATASTROPHIC_ERROR"
+ };
+
+ if (event < CCAE_REMOTE_SHUTDOWN ||
+ event > CCAE_RNIC_CATASTROPHIC_ERROR)
+ return "<invalid event>";
+
+ event -= CCAE_REMOTE_SHUTDOWN;
+ return event_str[event];
+}
+
+const char *to_qp_state_str(int state)
+{
+ switch (state) {
+ case C2_QP_STATE_IDLE:
+ return "C2_QP_STATE_IDLE";
+ case C2_QP_STATE_CONNECTING:
+ return "C2_QP_STATE_CONNECTING";
+ case C2_QP_STATE_RTS:
+ return "C2_QP_STATE_RTS";
+ case C2_QP_STATE_CLOSING:
+ return "C2_QP_STATE_CLOSING";
+ case C2_QP_STATE_TERMINATE:
+ return "C2_QP_STATE_TERMINATE";
+ case C2_QP_STATE_ERROR:
+ return "C2_QP_STATE_ERROR";
+ default:
+ return "<invalid QP state>";
+ };
+}
+#endif
+
+void c2_ae_event(struct c2_dev *c2dev, u32 mq_index)
+{
+ struct c2_mq *mq = c2dev->qptr_array[mq_index];
+ union c2wr *wr;
+ void *resource_user_context;
+ struct iw_cm_event cm_event;
+ struct ib_event ib_event;
+ enum c2_resource_indicator resource_indicator;
+ enum c2_event_id event_id;
+ unsigned long flags;
+ u8 *pdata = NULL;
+ int status;
+
+ /*
+ * retreive the message
+ */
+ wr = c2_mq_consume(mq);
+ if (!wr)
+ return;
+
+ memset(&ib_event, 0, sizeof(ib_event));
+ memset(&cm_event, 0, sizeof(cm_event));
+
+ event_id = c2_wr_get_id(wr);
+ resource_indicator = be32_to_cpu(wr->ae.ae_generic.resource_type);
+ resource_user_context =
+ (void *) (unsigned long) wr->ae.ae_generic.user_context;
+
+ status = cm_event.status = c2_convert_cm_status(c2_wr_get_result(wr));
+
+ pr_debug("event received c2_dev=%p, event_id=%d, "
+ "resource_indicator=%d, user_context=%p, status = %d\n",
+ c2dev, event_id, resource_indicator, resource_user_context,
+ status);
+
+ switch (resource_indicator) {
+ case C2_RES_IND_QP:{
+
+ struct c2_qp *qp = (struct c2_qp *)resource_user_context;
+ struct iw_cm_id *cm_id = qp->cm_id;
+ struct c2wr_ae_active_connect_results *res;
+
+ if (!cm_id) {
+ pr_debug("event received, but cm_id is <nul>, qp=%p!\n",
+ qp);
+ goto ignore_it;
+ }
+ pr_debug("%s: event = %s, user_context=%llx, "
+ "resource_type=%x, "
+ "resource=%x, qp_state=%s\n",
+ __FUNCTION__,
+ to_event_str(event_id),
+ be64_to_cpu(wr->ae.ae_generic.user_context),
+ be32_to_cpu(wr->ae.ae_generic.resource_type),
+ be32_to_cpu(wr->ae.ae_generic.resource),
+ to_qp_state_str(be32_to_cpu(wr->ae.ae_generic.qp_state)));
+
+ c2_set_qp_state(qp, be32_to_cpu(wr->ae.ae_generic.qp_state));
+
+ switch (event_id) {
+ case CCAE_ACTIVE_CONNECT_RESULTS:
+ res = &wr->ae.ae_active_connect_results;
+ cm_event.event = IW_CM_EVENT_CONNECT_REPLY;
+ cm_event.local_addr.sin_addr.s_addr = res->laddr;
+ cm_event.remote_addr.sin_addr.s_addr = res->raddr;
+ cm_event.local_addr.sin_port = res->lport;
+ cm_event.remote_addr.sin_port = res->rport;
+ if (status == 0) {
+ cm_event.private_data_len =
+ be32_to_cpu(res->private_data_length);
+ } else {
+ spin_lock_irqsave(&qp->lock, flags);
+ if (qp->cm_id) {
+ qp->cm_id->rem_ref(qp->cm_id);
+ qp->cm_id = NULL;
+ }
+ spin_unlock_irqrestore(&qp->lock, flags);
+ cm_event.private_data_len = 0;
+ cm_event.private_data = NULL;
+ }
+ if (cm_event.private_data_len) {
+ /* copy private data */
+ pdata =
+ kmalloc(cm_event.private_data_len,
+ GFP_ATOMIC);
+ if (!pdata) {
+ /* Ignore the request, maybe the
+ * remote peer will retry */
+ pr_debug ("Ignored connect request -- "
+ "no memory for pdata"
+ "private_data_len=%d\n",
+ cm_event.private_data_len);
+ goto ignore_it;
+ }
+
+ memcpy(pdata, res->private_data,
+ cm_event.private_data_len);
+
+ cm_event.private_data = pdata;
+ }
+ if (cm_id->event_handler)
+ cm_id->event_handler(cm_id, &cm_event);
+ break;
+ case CCAE_TERMINATE_MESSAGE_RECEIVED:
+ case CCAE_CQ_SQ_COMPLETION_OVERFLOW:
+ ib_event.device = &c2dev->ibdev;
+ ib_event.element.qp = &qp->ibqp;
+ ib_event.event = IB_EVENT_QP_REQ_ERR;
+
+ if (qp->ibqp.event_handler)
+ qp->ibqp.event_handler(&ib_event,
+ qp->ibqp.
+ qp_context);
+ break;
+ case CCAE_BAD_CLOSE:
+ case CCAE_LLP_CLOSE_COMPLETE:
+ case CCAE_LLP_CONNECTION_RESET:
+ case CCAE_LLP_CONNECTION_LOST:
+ BUG_ON(cm_id->event_handler==(void*)0x6b6b6b6b);
+
+ spin_lock_irqsave(&qp->lock, flags);
+ if (qp->cm_id) {
+ qp->cm_id->rem_ref(qp->cm_id);
+ qp->cm_id = NULL;
+ }
+ spin_unlock_irqrestore(&qp->lock, flags);
+ cm_event.event = IW_CM_EVENT_CLOSE;
+ cm_event.status = 0;
+ if (cm_id->event_handler)
+ cm_id->event_handler(cm_id, &cm_event);
+ break;
+ default:
+ BUG_ON(1);
+ pr_debug("%s:%d Unexpected event_id=%d on QP=%p, "
+ "CM_ID=%p\n",
+ __FUNCTION__, __LINE__,
+ event_id, qp, cm_id);
+ break;
+ }
+ break;
+ }
+
+ case C2_RES_IND_EP:{
+
+ struct c2wr_ae_connection_request *req =
+ &wr->ae.ae_connection_request;
+ struct iw_cm_id *cm_id =
+ (struct iw_cm_id *)resource_user_context;
+
+ pr_debug("C2_RES_IND_EP event_id=%d\n", event_id);
+ if (event_id != CCAE_CONNECTION_REQUEST) {
+ pr_debug("%s: Invalid event_id: %d\n",
+ __FUNCTION__, event_id);
+ break;
+ }
+ cm_event.event = IW_CM_EVENT_CONNECT_REQUEST;
+ cm_event.provider_data = (void*)(unsigned long)req->cr_handle;
+ cm_event.local_addr.sin_addr.s_addr = req->laddr;
+ cm_event.remote_addr.sin_addr.s_addr = req->raddr;
+ cm_event.local_addr.sin_port = req->lport;
+ cm_event.remote_addr.sin_port = req->rport;
+ cm_event.private_data_len =
+ be32_to_cpu(req->private_data_length);
+
+ if (cm_event.private_data_len) {
+ pdata =
+ kmalloc(cm_event.private_data_len,
+ GFP_ATOMIC);
+ if (!pdata) {
+ /* Ignore the request, maybe the remote peer
+ * will retry */
+ pr_debug ("Ignored connect request -- "
+ "no memory for pdata"
+ "private_data_len=%d\n",
+ cm_event.private_data_len);
+ goto ignore_it;
+ }
+ memcpy(pdata,
+ req->private_data,
+ cm_event.private_data_len);
+
+ cm_event.private_data = pdata;
+ }
+ if (cm_id->event_handler)
+ cm_id->event_handler(cm_id, &cm_event);
+ break;
+ }
+
+ case C2_RES_IND_CQ:{
+ struct c2_cq *cq =
+ (struct c2_cq *) resource_user_context;
+
+ pr_debug("IB_EVENT_CQ_ERR\n");
+ ib_event.device = &c2dev->ibdev;
+ ib_event.element.cq = &cq->ibcq;
+ ib_event.event = IB_EVENT_CQ_ERR;
+
+ if (cq->ibcq.event_handler)
+ cq->ibcq.event_handler(&ib_event,
+ cq->ibcq.cq_context);
+ }
+
+ default:
+ printk("Bad resource indicator = %d\n",
+ resource_indicator);
+ break;
+ }
+
+ ignore_it:
+ c2_mq_free(mq);
+}
diff --git a/drivers/infiniband/hw/amso1100/c2_intr.c b/drivers/infiniband/hw/amso1100/c2_intr.c
new file mode 100644
index 0000000..75bb18c
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2_intr.c
@@ -0,0 +1,209 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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 "c2.h"
+#include <rdma/iw_cm.h>
+#include "c2_vq.h"
+
+static void handle_mq(struct c2_dev *c2dev, u32 index);
+static void handle_vq(struct c2_dev *c2dev, u32 mq_index);
+
+/*
+ * Handle RNIC interrupts
+ */
+void c2_rnic_interrupt(struct c2_dev *c2dev)
+{
+ unsigned int mq_index;
+
+ while (c2dev->hints_read != be16_to_cpu(c2dev->hint_count)) {
+ mq_index = readl(c2dev->regs + PCI_BAR0_HOST_HINT);
+ if (mq_index & 0x80000000) {
+ break;
+ }
+
+ c2dev->hints_read++;
+ handle_mq(c2dev, mq_index);
+ }
+
+}
+
+/*
+ * Top level MQ handler
+ */
+static void handle_mq(struct c2_dev *c2dev, u32 mq_index)
+{
+ if (c2dev->qptr_array[mq_index] == NULL) {
+ pr_debug(KERN_INFO "handle_mq: stray activity for mq_index=%d\n",
+ mq_index);
+ return;
+ }
+
+ switch (mq_index) {
+ case (0):
+ /*
+ * An index of 0 in the activity queue
+ * indicates the req vq now has messages
+ * available...
+ *
+ * Wake up any waiters waiting on req VQ
+ * message availability.
+ */
+ wake_up(&c2dev->req_vq_wo);
+ break;
+ case (1):
+ handle_vq(c2dev, mq_index);
+ break;
+ case (2):
+ /* We have to purge the VQ in case there are pending
+ * accept reply requests that would result in the
+ * generation of an ESTABLISHED event. If we don't
+ * generate these first, a CLOSE event could end up
+ * being delivered before the ESTABLISHED event.
+ */
+ handle_vq(c2dev, 1);
+
+ c2_ae_event(c2dev, mq_index);
+ break;
+ default:
+ /* There is no event synchronization between CQ events
+ * and AE or CM events. In fact, CQE could be
+ * delivered for all of the I/O up to and including the
+ * FLUSH for a peer disconenct prior to the ESTABLISHED
+ * event being delivered to the app. The reason for this
+ * is that CM events are delivered on a thread, while AE
+ * and CM events are delivered on interrupt context.
+ */
+ c2_cq_event(c2dev, mq_index);
+ break;
+ }
+
+ return;
+}
+
+/*
+ * Handles verbs WR replies.
+ */
+static void handle_vq(struct c2_dev *c2dev, u32 mq_index)
+{
+ void *adapter_msg, *reply_msg;
+ struct c2wr_hdr *host_msg;
+ struct c2wr_hdr tmp;
+ struct c2_mq *reply_vq;
+ struct c2_vq_req *req;
+ struct iw_cm_event cm_event;
+ int err;
+
+ reply_vq = (struct c2_mq *) c2dev->qptr_array[mq_index];
+
+ /*
+ * get next msg from mq_index into adapter_msg.
+ * don't free it yet.
+ */
+ adapter_msg = c2_mq_consume(reply_vq);
+ if (adapter_msg == NULL) {
+ return;
+ }
+
+ host_msg = vq_repbuf_alloc(c2dev);
+
+ /*
+ * If we can't get a host buffer, then we'll still
+ * wakeup the waiter, we just won't give him the msg.
+ * It is assumed the waiter will deal with this...
+ */
+ if (!host_msg) {
+ pr_debug("handle_vq: no repbufs!\n");
+
+ /*
+ * just copy the WR header into a local variable.
+ * this allows us to still demux on the context
+ */
+ host_msg = &tmp;
+ memcpy(host_msg, adapter_msg, sizeof(tmp));
+ reply_msg = NULL;
+ } else {
+ memcpy(host_msg, adapter_msg, reply_vq->msg_size);
+ reply_msg = host_msg;
+ }
+
+ /*
+ * consume the msg from the MQ
+ */
+ c2_mq_free(reply_vq);
+
+ /*
+ * wakeup the waiter.
+ */
+ req = (struct c2_vq_req *) (unsigned long) host_msg->context;
+ if (req == NULL) {
+ /*
+ * We should never get here, as the adapter should
+ * never send us a reply that we're not expecting.
+ */
+ vq_repbuf_free(c2dev, host_msg);
+ pr_debug("handle_vq: UNEXPECTEDLY got NULL req\n");
+ return;
+ }
+
+ err = c2_errno(reply_msg);
+ if (!err) switch (req->event) {
+ case IW_CM_EVENT_ESTABLISHED:
+ c2_set_qp_state(req->qp,
+ C2_QP_STATE_RTS);
+ case IW_CM_EVENT_CLOSE:
+
+ /*
+ * Move the QP to RTS if this is
+ * the established event
+ */
+ cm_event.event = req->event;
+ cm_event.status = 0;
+ cm_event.local_addr = req->cm_id->local_addr;
+ cm_event.remote_addr = req->cm_id->remote_addr;
+ cm_event.private_data = NULL;
+ cm_event.private_data_len = 0;
+ req->cm_id->event_handler(req->cm_id, &cm_event);
+ break;
+ default:
+ break;
+ }
+
+ req->reply_msg = (u64) (unsigned long) (reply_msg);
+ atomic_set(&req->reply_ready, 1);
+ wake_up(&req->wait_object);
+
+ /*
+ * If the request was cancelled, then this put will
+ * free the vq_req memory...and reply_msg!!!
+ */
+ vq_req_put(c2dev, req);
+}
diff --git a/drivers/infiniband/hw/amso1100/c2_rnic.c b/drivers/infiniband/hw/amso1100/c2_rnic.c
new file mode 100644
index 0000000..49645a9
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2_rnic.c
@@ -0,0 +1,631 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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/module.h>
+#include <linux/moduleparam.h>
+#include <linux/pci.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/delay.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/if_vlan.h>
+#include <linux/crc32.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/init.h>
+#include <linux/dma-mapping.h>
+#include <linux/mm.h>
+#include <linux/inet.h>
+
+#include <linux/route.h>
+
+#include <asm/io.h>
+#include <asm/irq.h>
+#include <asm/byteorder.h>
+#include <rdma/ib_smi.h>
+#include "c2.h"
+#include "c2_vq.h"
+
+/* Device capabilities */
+#define C2_MIN_PAGESIZE 1024
+
+#define C2_MAX_MRS 32768
+#define C2_MAX_QPS 16000
+#define C2_MAX_WQE_SZ 256
+#define C2_MAX_QP_WR ((128*1024)/C2_MAX_WQE_SZ)
+#define C2_MAX_SGES 4
+#define C2_MAX_SGE_RD 1
+#define C2_MAX_CQS 32768
+#define C2_MAX_CQES 4096
+#define C2_MAX_PDS 16384
+
+/*
+ * Send the adapter INIT message to the amso1100
+ */
+static int c2_adapter_init(struct c2_dev *c2dev)
+{
+ struct c2wr_init_req wr;
+ int err;
+
+ memset(&wr, 0, sizeof(wr));
+ c2_wr_set_id(&wr, CCWR_INIT);
+ wr.hdr.context = 0;
+ wr.hint_count = cpu_to_be64(__pa(&c2dev->hint_count));
+ wr.q0_host_shared = cpu_to_be64(__pa(c2dev->req_vq.shared));
+ wr.q1_host_shared = cpu_to_be64(__pa(c2dev->rep_vq.shared));
+ wr.q1_host_msg_pool = cpu_to_be64(__pa(c2dev->rep_vq.msg_pool.host));
+ wr.q2_host_shared = cpu_to_be64(__pa(c2dev->aeq.shared));
+ wr.q2_host_msg_pool = cpu_to_be64(__pa(c2dev->aeq.msg_pool.host));
+
+ /* Post the init message */
+ err = vq_send_wr(c2dev, (union c2wr *) & wr);
+
+ return err;
+}
+
+/*
+ * Send the adapter TERM message to the amso1100
+ */
+static void c2_adapter_term(struct c2_dev *c2dev)
+{
+ struct c2wr_init_req wr;
+
+ memset(&wr, 0, sizeof(wr));
+ c2_wr_set_id(&wr, CCWR_TERM);
+ wr.hdr.context = 0;
+
+ /* Post the init message */
+ vq_send_wr(c2dev, (union c2wr *) & wr);
+ c2dev->init = 0;
+
+ return;
+}
+
+/*
+ * Query the adapter
+ */
+int c2_rnic_query(struct c2_dev *c2dev,
+ struct ib_device_attr *props)
+{
+ struct c2_vq_req *vq_req;
+ struct c2wr_rnic_query_req wr;
+ struct c2wr_rnic_query_rep *reply;
+ int err;
+
+ vq_req = vq_req_alloc(c2dev);
+ if (!vq_req)
+ return -ENOMEM;
+
+ c2_wr_set_id(&wr, CCWR_RNIC_QUERY);
+ wr.hdr.context = (unsigned long) vq_req;
+ wr.rnic_handle = c2dev->adapter_handle;
+
+ vq_req_get(c2dev, vq_req);
+
+ err = vq_send_wr(c2dev, (union c2wr *) &wr);
+ if (err) {
+ vq_req_put(c2dev, vq_req);
+ goto bail1;
+ }
+
+ err = vq_wait_for_reply(c2dev, vq_req);
+ if (err)
+ goto bail1;
+
+ reply =
+ (struct c2wr_rnic_query_rep *) (unsigned long) (vq_req->reply_msg);
+ if (!reply)
+ err = -ENOMEM;
+
+ err = c2_errno(reply);
+ if (err)
+ goto bail2;
+
+ props->fw_ver =
+ ((u64)be32_to_cpu(reply->fw_ver_major) << 32) |
+ ((be32_to_cpu(reply->fw_ver_minor) && 0xFFFF) << 16) |
+ (be32_to_cpu(reply->fw_ver_patch) && 0xFFFF);
+ memcpy(&props->sys_image_guid, c2dev->netdev->dev_addr, 6);
+ props->max_mr_size = 0xFFFFFFFF;
+ props->page_size_cap = ~(C2_MIN_PAGESIZE-1);
+ props->vendor_id = be32_to_cpu(reply->vendor_id);
+ props->vendor_part_id = be32_to_cpu(reply->part_number);
+ props->hw_ver = be32_to_cpu(reply->hw_version);
+ props->max_qp = be32_to_cpu(reply->max_qps);
+ props->max_qp_wr = be32_to_cpu(reply->max_qp_depth);
+ props->device_cap_flags = c2dev->device_cap_flags;
+ props->max_sge = C2_MAX_SGES;
+ props->max_sge_rd = C2_MAX_SGE_RD;
+ props->max_cq = be32_to_cpu(reply->max_cqs);
+ props->max_cqe = be32_to_cpu(reply->max_cq_depth);
+ props->max_mr = be32_to_cpu(reply->max_mrs);
+ props->max_pd = be32_to_cpu(reply->max_pds);
+ props->max_qp_rd_atom = be32_to_cpu(reply->max_qp_ird);
+ props->max_ee_rd_atom = 0;
+ props->max_res_rd_atom = be32_to_cpu(reply->max_global_ird);
+ props->max_qp_init_rd_atom = be32_to_cpu(reply->max_qp_ord);
+ props->max_ee_init_rd_atom = 0;
+ props->atomic_cap = IB_ATOMIC_NONE;
+ props->max_ee = 0;
+ props->max_rdd = 0;
+ props->max_mw = be32_to_cpu(reply->max_mws);
+ props->max_raw_ipv6_qp = 0;
+ props->max_raw_ethy_qp = 0;
+ props->max_mcast_grp = 0;
+ props->max_mcast_qp_attach = 0;
+ props->max_total_mcast_qp_attach = 0;
+ props->max_ah = 0;
+ props->max_fmr = 0;
+ props->max_map_per_fmr = 0;
+ props->max_srq = 0;
+ props->max_srq_wr = 0;
+ props->max_srq_sge = 0;
+ props->max_pkeys = 0;
+ props->local_ca_ack_delay = 0;
+
+ bail2:
+ vq_repbuf_free(c2dev, reply);
+
+ bail1:
+ vq_req_free(c2dev, vq_req);
+ return err;
+}
+
+/*
+ * Add an IP address to the RNIC interface
+ */
+int c2_add_addr(struct c2_dev *c2dev, u32 inaddr, u32 inmask)
+{
+ struct c2_vq_req *vq_req;
+ struct c2wr_rnic_setconfig_req *wr;
+ struct c2wr_rnic_setconfig_rep *reply;
+ struct c2_netaddr netaddr;
+ int err, len;
+
+ vq_req = vq_req_alloc(c2dev);
+ if (!vq_req)
+ return -ENOMEM;
+
+ len = sizeof(struct c2_netaddr);
+ wr = kmalloc(c2dev->req_vq.msg_size, GFP_KERNEL);
+ if (!wr) {
+ err = -ENOMEM;
+ goto bail0;
+ }
+
+ c2_wr_set_id(wr, CCWR_RNIC_SETCONFIG);
+ wr->hdr.context = (unsigned long) vq_req;
+ wr->rnic_handle = c2dev->adapter_handle;
+ wr->option = cpu_to_be32(C2_CFG_ADD_ADDR);
+
+ netaddr.ip_addr = inaddr;
+ netaddr.netmask = inmask;
+ netaddr.mtu = 0;
+
+ memcpy(wr->data, &netaddr, len);
+
+ vq_req_get(c2dev, vq_req);
+
+ err = vq_send_wr(c2dev, (union c2wr *) wr);
+ if (err) {
+ vq_req_put(c2dev, vq_req);
+ goto bail1;
+ }
+
+ err = vq_wait_for_reply(c2dev, vq_req);
+ if (err)
+ goto bail1;
+
+ reply =
+ (struct c2wr_rnic_setconfig_rep *) (unsigned long) (vq_req->reply_msg);
+ if (!reply) {
+ err = -ENOMEM;
+ goto bail1;
+ }
+
+ err = c2_errno(reply);
+ vq_repbuf_free(c2dev, reply);
+
+ bail1:
+ kfree(wr);
+ bail0:
+ vq_req_free(c2dev, vq_req);
+ return err;
+}
+
+/*
+ * Delete an IP address from the RNIC interface
+ */
+int c2_del_addr(struct c2_dev *c2dev, u32 inaddr, u32 inmask)
+{
+ struct c2_vq_req *vq_req;
+ struct c2wr_rnic_setconfig_req *wr;
+ struct c2wr_rnic_setconfig_rep *reply;
+ struct c2_netaddr netaddr;
+ int err, len;
+
+ vq_req = vq_req_alloc(c2dev);
+ if (!vq_req)
+ return -ENOMEM;
+
+ len = sizeof(struct c2_netaddr);
+ wr = kmalloc(c2dev->req_vq.msg_size, GFP_KERNEL);
+ if (!wr) {
+ err = -ENOMEM;
+ goto bail0;
+ }
+
+ c2_wr_set_id(wr, CCWR_RNIC_SETCONFIG);
+ wr->hdr.context = (unsigned long) vq_req;
+ wr->rnic_handle = c2dev->adapter_handle;
+ wr->option = cpu_to_be32(C2_CFG_DEL_ADDR);
+
+ netaddr.ip_addr = inaddr;
+ netaddr.netmask = inmask;
+ netaddr.mtu = 0;
+
+ memcpy(wr->data, &netaddr, len);
+
+ vq_req_get(c2dev, vq_req);
+
+ err = vq_send_wr(c2dev, (union c2wr *) wr);
+ if (err) {
+ vq_req_put(c2dev, vq_req);
+ goto bail1;
+ }
+
+ err = vq_wait_for_reply(c2dev, vq_req);
+ if (err)
+ goto bail1;
+
+ reply =
+ (struct c2wr_rnic_setconfig_rep *) (unsigned long) (vq_req->reply_msg);
+ if (!reply) {
+ err = -ENOMEM;
+ goto bail1;
+ }
+
+ err = c2_errno(reply);
+ vq_repbuf_free(c2dev, reply);
+
+ bail1:
+ kfree(wr);
+ bail0:
+ vq_req_free(c2dev, vq_req);
+ return err;
+}
+
+/*
+ * Open a single RNIC instance to use with all
+ * low level openib calls
+ */
+static int c2_rnic_open(struct c2_dev *c2dev)
+{
+ struct c2_vq_req *vq_req;
+ union c2wr wr;
+ struct c2wr_rnic_open_rep *reply;
+ int err;
+
+ vq_req = vq_req_alloc(c2dev);
+ if (vq_req == NULL) {
+ return -ENOMEM;
+ }
+
+ memset(&wr, 0, sizeof(wr));
+ c2_wr_set_id(&wr, CCWR_RNIC_OPEN);
+ wr.rnic_open.req.hdr.context = (unsigned long) (vq_req);
+ wr.rnic_open.req.flags = cpu_to_be16(RNIC_PRIV_MODE);
+ wr.rnic_open.req.port_num = cpu_to_be16(0);
+ wr.rnic_open.req.user_context = (unsigned long) c2dev;
+
+ vq_req_get(c2dev, vq_req);
+
+ err = vq_send_wr(c2dev, &wr);
+ if (err) {
+ vq_req_put(c2dev, vq_req);
+ goto bail0;
+ }
+
+ err = vq_wait_for_reply(c2dev, vq_req);
+ if (err) {
+ goto bail0;
+ }
+
+ reply = (struct c2wr_rnic_open_rep *) (unsigned long) (vq_req->reply_msg);
+ if (!reply) {
+ err = -ENOMEM;
+ goto bail0;
+ }
+
+ if ((err = c2_errno(reply)) != 0) {
+ goto bail1;
+ }
+
+ c2dev->adapter_handle = reply->rnic_handle;
+
+ bail1:
+ vq_repbuf_free(c2dev, reply);
+ bail0:
+ vq_req_free(c2dev, vq_req);
+ return err;
+}
+
+/*
+ * Close the RNIC instance
+ */
+static int c2_rnic_close(struct c2_dev *c2dev)
+{
+ struct c2_vq_req *vq_req;
+ union c2wr wr;
+ struct c2wr_rnic_close_rep *reply;
+ int err;
+
+ vq_req = vq_req_alloc(c2dev);
+ if (vq_req == NULL) {
+ return -ENOMEM;
+ }
+
+ memset(&wr, 0, sizeof(wr));
+ c2_wr_set_id(&wr, CCWR_RNIC_CLOSE);
+ wr.rnic_close.req.hdr.context = (unsigned long) vq_req;
+ wr.rnic_close.req.rnic_handle = c2dev->adapter_handle;
+
+ vq_req_get(c2dev, vq_req);
+
+ err = vq_send_wr(c2dev, &wr);
+ if (err) {
+ vq_req_put(c2dev, vq_req);
+ goto bail0;
+ }
+
+ err = vq_wait_for_reply(c2dev, vq_req);
+ if (err) {
+ goto bail0;
+ }
+
+ reply = (struct c2wr_rnic_close_rep *) (unsigned long) (vq_req->reply_msg);
+ if (!reply) {
+ err = -ENOMEM;
+ goto bail0;
+ }
+
+ if ((err = c2_errno(reply)) != 0) {
+ goto bail1;
+ }
+
+ c2dev->adapter_handle = 0;
+
+ bail1:
+ vq_repbuf_free(c2dev, reply);
+ bail0:
+ vq_req_free(c2dev, vq_req);
+ return err;
+}
+
+/*
+ * Called by c2_probe to initialize the RNIC. This principally
+ * involves initalizing the various limits and resouce pools that
+ * comprise the RNIC instance.
+ */
+int c2_rnic_init(struct c2_dev *c2dev)
+{
+ int err;
+ u32 qsize, msgsize;
+ void *q1_pages;
+ void *q2_pages;
+ void __iomem *mmio_regs;
+
+ /* Device capabilities */
+ c2dev->device_cap_flags =
+ (IB_DEVICE_RESIZE_MAX_WR |
+ IB_DEVICE_CURR_QP_STATE_MOD |
+ IB_DEVICE_SYS_IMAGE_GUID |
+ IB_DEVICE_ZERO_STAG |
+ IB_DEVICE_SEND_W_INV | IB_DEVICE_MEM_WINDOW);
+
+ /* Allocate the qptr_array */
+ c2dev->qptr_array = vmalloc(C2_MAX_CQS * sizeof(void *));
+ if (!c2dev->qptr_array) {
+ return -ENOMEM;
+ }
+
+ /* Inialize the qptr_array */
+ memset(c2dev->qptr_array, 0, C2_MAX_CQS * sizeof(void *));
+ c2dev->qptr_array[0] = (void *) &c2dev->req_vq;
+ c2dev->qptr_array[1] = (void *) &c2dev->rep_vq;
+ c2dev->qptr_array[2] = (void *) &c2dev->aeq;
+
+ /* Initialize data structures */
+ init_waitqueue_head(&c2dev->req_vq_wo);
+ spin_lock_init(&c2dev->vqlock);
+ spin_lock_init(&c2dev->lock);
+
+ /* Allocate MQ shared pointer pool for kernel clients. User
+ * mode client pools are hung off the user context
+ */
+ err = c2_init_mqsp_pool(GFP_KERNEL, &c2dev->kern_mqsp_pool);
+ if (err) {
+ goto bail0;
+ }
+
+ /* Allocate shared pointers for Q0, Q1, and Q2 from
+ * the shared pointer pool.
+ */
+ c2dev->req_vq.shared = c2_alloc_mqsp(c2dev->kern_mqsp_pool);
+ c2dev->rep_vq.shared = c2_alloc_mqsp(c2dev->kern_mqsp_pool);
+ c2dev->aeq.shared = c2_alloc_mqsp(c2dev->kern_mqsp_pool);
+ if (!c2dev->req_vq.shared ||
+ !c2dev->rep_vq.shared || !c2dev->aeq.shared) {
+ err = -ENOMEM;
+ goto bail1;
+ }
+
+ mmio_regs = c2dev->kva;
+ /* Initialize the Verbs Request Queue */
+ c2_mq_req_init(&c2dev->req_vq, 0,
+ be32_to_cpu(readl(mmio_regs + C2_REGS_Q0_QSIZE)),
+ be32_to_cpu(readl(mmio_regs + C2_REGS_Q0_MSGSIZE)),
+ mmio_regs +
+ be32_to_cpu(readl(mmio_regs + C2_REGS_Q0_POOLSTART)),
+ mmio_regs +
+ be32_to_cpu(readl(mmio_regs + C2_REGS_Q0_SHARED)),
+ C2_MQ_ADAPTER_TARGET);
+
+ /* Initialize the Verbs Reply Queue */
+ qsize = be32_to_cpu(readl(mmio_regs + C2_REGS_Q1_QSIZE));
+ msgsize = be32_to_cpu(readl(mmio_regs + C2_REGS_Q1_MSGSIZE));
+ q1_pages = kmalloc(qsize * msgsize, GFP_KERNEL);
+ if (!q1_pages) {
+ err = -ENOMEM;
+ goto bail1;
+ }
+ c2_mq_rep_init(&c2dev->rep_vq,
+ 1,
+ qsize,
+ msgsize,
+ q1_pages,
+ mmio_regs +
+ be32_to_cpu(readl(mmio_regs + C2_REGS_Q1_SHARED)),
+ C2_MQ_HOST_TARGET);
+
+ /* Initialize the Asynchronus Event Queue */
+ qsize = be32_to_cpu(readl(mmio_regs + C2_REGS_Q2_QSIZE));
+ msgsize = be32_to_cpu(readl(mmio_regs + C2_REGS_Q2_MSGSIZE));
+ q2_pages = kmalloc(qsize * msgsize, GFP_KERNEL);
+ if (!q2_pages) {
+ err = -ENOMEM;
+ goto bail2;
+ }
+ c2_mq_rep_init(&c2dev->aeq,
+ 2,
+ qsize,
+ msgsize,
+ q2_pages,
+ mmio_regs +
+ be32_to_cpu(readl(mmio_regs + C2_REGS_Q2_SHARED)),
+ C2_MQ_HOST_TARGET);
+
+ /* Initialize the verbs request allocator */
+ err = vq_init(c2dev);
+ if (err)
+ goto bail3;
+
+ /* Enable interrupts on the adapter */
+ writel(0, c2dev->regs + C2_IDIS);
+
+ /* create the WR init message */
+ err = c2_adapter_init(c2dev);
+ if (err)
+ goto bail4;
+ c2dev->init++;
+
+ /* open an adapter instance */
+ err = c2_rnic_open(c2dev);
+ if (err)
+ goto bail4;
+
+ /* Initialize cached the adapter limits */
+ if (c2_rnic_query(c2dev, &c2dev->props))
+ goto bail4;
+
+ /* Initialize the PD pool */
+ err = c2_init_pd_table(c2dev);
+ if (err)
+ goto bail5;
+
+ /* Initialize the QP pool */
+ err = c2_init_qp_table(c2dev);
+ if (err)
+ goto bail6;
+ return 0;
+
+ bail6:
+ c2_cleanup_pd_table(c2dev);
+ bail5:
+ c2_rnic_close(c2dev);
+ bail4:
+ vq_term(c2dev);
+ bail3:
+ kfree(q2_pages);
+ bail2:
+ kfree(q1_pages);
+ bail1:
+ c2_free_mqsp_pool(c2dev->kern_mqsp_pool);
+ bail0:
+ vfree(c2dev->qptr_array);
+
+ return err;
+}
+
+/*
+ * Called by c2_remove to cleanup the RNIC resources.
+ */
+void c2_rnic_term(struct c2_dev *c2dev)
+{
+
+ /* Close the open adapter instance */
+ c2_rnic_close(c2dev);
+
+ /* Send the TERM message to the adapter */
+ c2_adapter_term(c2dev);
+
+ /* Disable interrupts on the adapter */
+ writel(1, c2dev->regs + C2_IDIS);
+
+ /* Free the QP pool */
+ c2_cleanup_qp_table(c2dev);
+
+ /* Free the PD pool */
+ c2_cleanup_pd_table(c2dev);
+
+ /* Free the verbs request allocator */
+ vq_term(c2dev);
+
+ /* Free the asynchronus event queue */
+ kfree(c2dev->aeq.msg_pool.host);
+
+ /* Free the verbs reply queue */
+ kfree(c2dev->rep_vq.msg_pool.host);
+
+ /* Free the MQ shared pointer pool */
+ c2_free_mqsp_pool(c2dev->kern_mqsp_pool);
+
+ /* Free the qptr_array */
+ vfree(c2dev->qptr_array);
+
+ return;
+}
^ permalink raw reply related
* [PATCH v2 2/7] AMSO1100 WR / Event Definitions.
From: Steve Wise @ 2006-06-07 20:06 UTC (permalink / raw)
To: rdreier, mshefty; +Cc: netdev, linux-kernel, openib-general
In-Reply-To: <20060607200646.9259.24588.stgit@stevo-desktop>
Review Changes:
C2_DEBUG -> DEBUG
---
drivers/infiniband/hw/amso1100/c2_ae.h | 108 ++
drivers/infiniband/hw/amso1100/c2_status.h | 158 +++
drivers/infiniband/hw/amso1100/c2_wr.h | 1523 ++++++++++++++++++++++++++++
3 files changed, 1789 insertions(+), 0 deletions(-)
diff --git a/drivers/infiniband/hw/amso1100/c2_ae.h b/drivers/infiniband/hw/amso1100/c2_ae.h
new file mode 100644
index 0000000..3a065c3
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2_ae.h
@@ -0,0 +1,108 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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 _C2_AE_H_
+#define _C2_AE_H_
+
+/*
+ * WARNING: If you change this file, also bump C2_IVN_BASE
+ * in common/include/clustercore/c2_ivn.h.
+ */
+
+/*
+ * Asynchronous Event Identifiers
+ *
+ * These start at 0x80 only so it's obvious from inspection that
+ * they are not work-request statuses. This isn't critical.
+ *
+ * NOTE: these event id's must fit in eight bits.
+ */
+enum c2_event_id {
+ CCAE_REMOTE_SHUTDOWN = 0x80,
+ CCAE_ACTIVE_CONNECT_RESULTS,
+ CCAE_CONNECTION_REQUEST,
+ CCAE_LLP_CLOSE_COMPLETE,
+ CCAE_TERMINATE_MESSAGE_RECEIVED,
+ CCAE_LLP_CONNECTION_RESET,
+ CCAE_LLP_CONNECTION_LOST,
+ CCAE_LLP_SEGMENT_SIZE_INVALID,
+ CCAE_LLP_INVALID_CRC,
+ CCAE_LLP_BAD_FPDU,
+ CCAE_INVALID_DDP_VERSION,
+ CCAE_INVALID_RDMA_VERSION,
+ CCAE_UNEXPECTED_OPCODE,
+ CCAE_INVALID_DDP_QUEUE_NUMBER,
+ CCAE_RDMA_READ_NOT_ENABLED,
+ CCAE_RDMA_WRITE_NOT_ENABLED,
+ CCAE_RDMA_READ_TOO_SMALL,
+ CCAE_NO_L_BIT,
+ CCAE_TAGGED_INVALID_STAG,
+ CCAE_TAGGED_BASE_BOUNDS_VIOLATION,
+ CCAE_TAGGED_ACCESS_RIGHTS_VIOLATION,
+ CCAE_TAGGED_INVALID_PD,
+ CCAE_WRAP_ERROR,
+ CCAE_BAD_CLOSE,
+ CCAE_BAD_LLP_CLOSE,
+ CCAE_INVALID_MSN_RANGE,
+ CCAE_INVALID_MSN_GAP,
+ CCAE_IRRQ_OVERFLOW,
+ CCAE_IRRQ_MSN_GAP,
+ CCAE_IRRQ_MSN_RANGE,
+ CCAE_IRRQ_INVALID_STAG,
+ CCAE_IRRQ_BASE_BOUNDS_VIOLATION,
+ CCAE_IRRQ_ACCESS_RIGHTS_VIOLATION,
+ CCAE_IRRQ_INVALID_PD,
+ CCAE_IRRQ_WRAP_ERROR,
+ CCAE_CQ_SQ_COMPLETION_OVERFLOW,
+ CCAE_CQ_RQ_COMPLETION_ERROR,
+ CCAE_QP_SRQ_WQE_ERROR,
+ CCAE_QP_LOCAL_CATASTROPHIC_ERROR,
+ CCAE_CQ_OVERFLOW,
+ CCAE_CQ_OPERATION_ERROR,
+ CCAE_SRQ_LIMIT_REACHED,
+ CCAE_QP_RQ_LIMIT_REACHED,
+ CCAE_SRQ_CATASTROPHIC_ERROR,
+ CCAE_RNIC_CATASTROPHIC_ERROR
+/* WARNING If you add more id's, make sure their values fit in eight bits. */
+};
+
+/*
+ * Resource Indicators and Identifiers
+ */
+enum c2_resource_indicator {
+ C2_RES_IND_QP = 1,
+ C2_RES_IND_EP,
+ C2_RES_IND_CQ,
+ C2_RES_IND_SRQ,
+};
+
+#endif /* _C2_AE_H_ */
diff --git a/drivers/infiniband/hw/amso1100/c2_status.h b/drivers/infiniband/hw/amso1100/c2_status.h
new file mode 100644
index 0000000..6ee4aa9
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2_status.h
@@ -0,0 +1,158 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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 _C2_STATUS_H_
+#define _C2_STATUS_H_
+
+/*
+ * Verbs Status Codes
+ */
+enum c2_status {
+ C2_OK = 0, /* This must be zero */
+ CCERR_INSUFFICIENT_RESOURCES = 1,
+ CCERR_INVALID_MODIFIER = 2,
+ CCERR_INVALID_MODE = 3,
+ CCERR_IN_USE = 4,
+ CCERR_INVALID_RNIC = 5,
+ CCERR_INTERRUPTED_OPERATION = 6,
+ CCERR_INVALID_EH = 7,
+ CCERR_INVALID_CQ = 8,
+ CCERR_CQ_EMPTY = 9,
+ CCERR_NOT_IMPLEMENTED = 10,
+ CCERR_CQ_DEPTH_TOO_SMALL = 11,
+ CCERR_PD_IN_USE = 12,
+ CCERR_INVALID_PD = 13,
+ CCERR_INVALID_SRQ = 14,
+ CCERR_INVALID_ADDRESS = 15,
+ CCERR_INVALID_NETMASK = 16,
+ CCERR_INVALID_QP = 17,
+ CCERR_INVALID_QP_STATE = 18,
+ CCERR_TOO_MANY_WRS_POSTED = 19,
+ CCERR_INVALID_WR_TYPE = 20,
+ CCERR_INVALID_SGL_LENGTH = 21,
+ CCERR_INVALID_SQ_DEPTH = 22,
+ CCERR_INVALID_RQ_DEPTH = 23,
+ CCERR_INVALID_ORD = 24,
+ CCERR_INVALID_IRD = 25,
+ CCERR_QP_ATTR_CANNOT_CHANGE = 26,
+ CCERR_INVALID_STAG = 27,
+ CCERR_QP_IN_USE = 28,
+ CCERR_OUTSTANDING_WRS = 29,
+ CCERR_STAG_IN_USE = 30,
+ CCERR_INVALID_STAG_INDEX = 31,
+ CCERR_INVALID_SGL_FORMAT = 32,
+ CCERR_ADAPTER_TIMEOUT = 33,
+ CCERR_INVALID_CQ_DEPTH = 34,
+ CCERR_INVALID_PRIVATE_DATA_LENGTH = 35,
+ CCERR_INVALID_EP = 36,
+ CCERR_MR_IN_USE = CCERR_STAG_IN_USE,
+ CCERR_FLUSHED = 38,
+ CCERR_INVALID_WQE = 39,
+ CCERR_LOCAL_QP_CATASTROPHIC_ERROR = 40,
+ CCERR_REMOTE_TERMINATION_ERROR = 41,
+ CCERR_BASE_AND_BOUNDS_VIOLATION = 42,
+ CCERR_ACCESS_VIOLATION = 43,
+ CCERR_INVALID_PD_ID = 44,
+ CCERR_WRAP_ERROR = 45,
+ CCERR_INV_STAG_ACCESS_ERROR = 46,
+ CCERR_ZERO_RDMA_READ_RESOURCES = 47,
+ CCERR_QP_NOT_PRIVILEGED = 48,
+ CCERR_STAG_STATE_NOT_INVALID = 49,
+ CCERR_INVALID_PAGE_SIZE = 50,
+ CCERR_INVALID_BUFFER_SIZE = 51,
+ CCERR_INVALID_PBE = 52,
+ CCERR_INVALID_FBO = 53,
+ CCERR_INVALID_LENGTH = 54,
+ CCERR_INVALID_ACCESS_RIGHTS = 55,
+ CCERR_PBL_TOO_BIG = 56,
+ CCERR_INVALID_VA = 57,
+ CCERR_INVALID_REGION = 58,
+ CCERR_INVALID_WINDOW = 59,
+ CCERR_TOTAL_LENGTH_TOO_BIG = 60,
+ CCERR_INVALID_QP_ID = 61,
+ CCERR_ADDR_IN_USE = 62,
+ CCERR_ADDR_NOT_AVAIL = 63,
+ CCERR_NET_DOWN = 64,
+ CCERR_NET_UNREACHABLE = 65,
+ CCERR_CONN_ABORTED = 66,
+ CCERR_CONN_RESET = 67,
+ CCERR_NO_BUFS = 68,
+ CCERR_CONN_TIMEDOUT = 69,
+ CCERR_CONN_REFUSED = 70,
+ CCERR_HOST_UNREACHABLE = 71,
+ CCERR_INVALID_SEND_SGL_DEPTH = 72,
+ CCERR_INVALID_RECV_SGL_DEPTH = 73,
+ CCERR_INVALID_RDMA_WRITE_SGL_DEPTH = 74,
+ CCERR_INSUFFICIENT_PRIVILEGES = 75,
+ CCERR_STACK_ERROR = 76,
+ CCERR_INVALID_VERSION = 77,
+ CCERR_INVALID_MTU = 78,
+ CCERR_INVALID_IMAGE = 79,
+ CCERR_PENDING = 98, /* not an error; user internally by adapter */
+ CCERR_DEFER = 99, /* not an error; used internally by adapter */
+ CCERR_FAILED_WRITE = 100,
+ CCERR_FAILED_ERASE = 101,
+ CCERR_FAILED_VERIFICATION = 102,
+ CCERR_NOT_FOUND = 103,
+
+};
+
+/*
+ * CCAE_ACTIVE_CONNECT_RESULTS status result codes.
+ */
+enum c2_connect_status {
+ C2_CONN_STATUS_SUCCESS = C2_OK,
+ C2_CONN_STATUS_NO_MEM = CCERR_INSUFFICIENT_RESOURCES,
+ C2_CONN_STATUS_TIMEDOUT = CCERR_CONN_TIMEDOUT,
+ C2_CONN_STATUS_REFUSED = CCERR_CONN_REFUSED,
+ C2_CONN_STATUS_NETUNREACH = CCERR_NET_UNREACHABLE,
+ C2_CONN_STATUS_HOSTUNREACH = CCERR_HOST_UNREACHABLE,
+ C2_CONN_STATUS_INVALID_RNIC = CCERR_INVALID_RNIC,
+ C2_CONN_STATUS_INVALID_QP = CCERR_INVALID_QP,
+ C2_CONN_STATUS_INVALID_QP_STATE = CCERR_INVALID_QP_STATE,
+ C2_CONN_STATUS_REJECTED = CCERR_CONN_RESET,
+ C2_CONN_STATUS_ADDR_NOT_AVAIL = CCERR_ADDR_NOT_AVAIL,
+};
+
+/*
+ * Flash programming status codes.
+ */
+enum c2_flash_status {
+ C2_FLASH_STATUS_SUCCESS = 0x0000,
+ C2_FLASH_STATUS_VERIFY_ERR = 0x0002,
+ C2_FLASH_STATUS_IMAGE_ERR = 0x0004,
+ C2_FLASH_STATUS_ECLBS = 0x0400,
+ C2_FLASH_STATUS_PSLBS = 0x0800,
+ C2_FLASH_STATUS_VPENS = 0x1000,
+};
+
+#endif /* _C2_STATUS_H_ */
diff --git a/drivers/infiniband/hw/amso1100/c2_wr.h b/drivers/infiniband/hw/amso1100/c2_wr.h
new file mode 100644
index 0000000..9d6468d
--- /dev/null
+++ b/drivers/infiniband/hw/amso1100/c2_wr.h
@@ -0,0 +1,1523 @@
+/*
+ * Copyright (c) 2005 Ammasso, Inc. All rights reserved.
+ * Copyright (c) 2005 Open Grid Computing, Inc. 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 _C2_WR_H_
+#define _C2_WR_H_
+
+#ifdef CCDEBUG
+#define CCWR_MAGIC 0xb07700b0
+#endif
+
+#define C2_QP_NO_ATTR_CHANGE 0xFFFFFFFF
+
+/* Maximum allowed size in bytes of private_data exchange
+ * on connect.
+ */
+#define C2_MAX_PRIVATE_DATA_SIZE 200
+
+/*
+ * These types are shared among the adapter, host, and CCIL consumer.
+ */
+enum c2_cq_notification_type {
+ C2_CQ_NOTIFICATION_TYPE_NONE = 1,
+ C2_CQ_NOTIFICATION_TYPE_NEXT,
+ C2_CQ_NOTIFICATION_TYPE_NEXT_SE
+};
+
+enum c2_setconfig_cmd {
+ C2_CFG_ADD_ADDR = 1,
+ C2_CFG_DEL_ADDR = 2,
+ C2_CFG_ADD_ROUTE = 3,
+ C2_CFG_DEL_ROUTE = 4
+};
+
+enum c2_getconfig_cmd {
+ C2_GETCONFIG_ROUTES = 1,
+ C2_GETCONFIG_ADDRS
+};
+
+/*
+ * CCIL Work Request Identifiers
+ */
+enum c2wr_ids {
+ CCWR_RNIC_OPEN = 1,
+ CCWR_RNIC_QUERY,
+ CCWR_RNIC_SETCONFIG,
+ CCWR_RNIC_GETCONFIG,
+ CCWR_RNIC_CLOSE,
+ CCWR_CQ_CREATE,
+ CCWR_CQ_QUERY,
+ CCWR_CQ_MODIFY,
+ CCWR_CQ_DESTROY,
+ CCWR_QP_CONNECT,
+ CCWR_PD_ALLOC,
+ CCWR_PD_DEALLOC,
+ CCWR_SRQ_CREATE,
+ CCWR_SRQ_QUERY,
+ CCWR_SRQ_MODIFY,
+ CCWR_SRQ_DESTROY,
+ CCWR_QP_CREATE,
+ CCWR_QP_QUERY,
+ CCWR_QP_MODIFY,
+ CCWR_QP_DESTROY,
+ CCWR_NSMR_STAG_ALLOC,
+ CCWR_NSMR_REGISTER,
+ CCWR_NSMR_PBL,
+ CCWR_STAG_DEALLOC,
+ CCWR_NSMR_REREGISTER,
+ CCWR_SMR_REGISTER,
+ CCWR_MR_QUERY,
+ CCWR_MW_ALLOC,
+ CCWR_MW_QUERY,
+ CCWR_EP_CREATE,
+ CCWR_EP_GETOPT,
+ CCWR_EP_SETOPT,
+ CCWR_EP_DESTROY,
+ CCWR_EP_BIND,
+ CCWR_EP_CONNECT,
+ CCWR_EP_LISTEN,
+ CCWR_EP_SHUTDOWN,
+ CCWR_EP_LISTEN_CREATE,
+ CCWR_EP_LISTEN_DESTROY,
+ CCWR_EP_QUERY,
+ CCWR_CR_ACCEPT,
+ CCWR_CR_REJECT,
+ CCWR_CONSOLE,
+ CCWR_TERM,
+ CCWR_FLASH_INIT,
+ CCWR_FLASH,
+ CCWR_BUF_ALLOC,
+ CCWR_BUF_FREE,
+ CCWR_FLASH_WRITE,
+ CCWR_INIT, /* WARNING: Don't move this ever again! */
+
+
+
+ /* Add new IDs here */
+
+
+
+ /*
+ * WARNING: CCWR_LAST must always be the last verbs id defined!
+ * All the preceding IDs are fixed, and must not change.
+ * You can add new IDs, but must not remove or reorder
+ * any IDs. If you do, YOU will ruin any hope of
+ * compatability between versions.
+ */
+ CCWR_LAST,
+
+ /*
+ * Start over at 1 so that arrays indexed by user wr id's
+ * begin at 1. This is OK since the verbs and user wr id's
+ * are always used on disjoint sets of queues.
+ */
+ /*
+ * The order of the CCWR_SEND_XX verbs must
+ * match the order of the RDMA_OPs
+ */
+ CCWR_SEND = 1,
+ CCWR_SEND_INV,
+ CCWR_SEND_SE,
+ CCWR_SEND_SE_INV,
+ CCWR_RDMA_WRITE,
+ CCWR_RDMA_READ,
+ CCWR_RDMA_READ_INV,
+ CCWR_MW_BIND,
+ CCWR_NSMR_FASTREG,
+ CCWR_STAG_INVALIDATE,
+ CCWR_RECV,
+ CCWR_NOP,
+ CCWR_UNIMPL,
+/* WARNING: This must always be the last user wr id defined! */
+};
+#define RDMA_SEND_OPCODE_FROM_WR_ID(x) (x+2)
+
+/*
+ * SQ/RQ Work Request Types
+ */
+enum c2_wr_type {
+ C2_WR_TYPE_SEND = CCWR_SEND,
+ C2_WR_TYPE_SEND_SE = CCWR_SEND_SE,
+ C2_WR_TYPE_SEND_INV = CCWR_SEND_INV,
+ C2_WR_TYPE_SEND_SE_INV = CCWR_SEND_SE_INV,
+ C2_WR_TYPE_RDMA_WRITE = CCWR_RDMA_WRITE,
+ C2_WR_TYPE_RDMA_READ = CCWR_RDMA_READ,
+ C2_WR_TYPE_RDMA_READ_INV_STAG = CCWR_RDMA_READ_INV,
+ C2_WR_TYPE_BIND_MW = CCWR_MW_BIND,
+ C2_WR_TYPE_FASTREG_NSMR = CCWR_NSMR_FASTREG,
+ C2_WR_TYPE_INV_STAG = CCWR_STAG_INVALIDATE,
+ C2_WR_TYPE_RECV = CCWR_RECV,
+ C2_WR_TYPE_NOP = CCWR_NOP,
+};
+
+struct c2_netaddr {
+ u32 ip_addr;
+ u32 netmask;
+ u32 mtu;
+};
+
+struct c2_route {
+ u32 ip_addr; /* 0 indicates the default route */
+ u32 netmask; /* netmask associated with dst */
+ u32 flags;
+ union {
+ u32 ipaddr; /* address of the nexthop interface */
+ u8 enaddr[6];
+ } nexthop;
+};
+
+/*
+ * A Scatter Gather Entry.
+ */
+struct c2_data_addr {
+ u32 stag;
+ u32 length;
+ u64 to;
+};
+
+/*
+ * MR and MW flags used by the consumer, RI, and RNIC.
+ */
+enum c2_mm_flags {
+ MEM_REMOTE = 0x0001, /* allow mw binds with remote access. */
+ MEM_VA_BASED = 0x0002, /* Not Zero-based */
+ MEM_PBL_COMPLETE = 0x0004, /* PBL array is complete in this msg */
+ MEM_LOCAL_READ = 0x0008, /* allow local reads */
+ MEM_LOCAL_WRITE = 0x0010, /* allow local writes */
+ MEM_REMOTE_READ = 0x0020, /* allow remote reads */
+ MEM_REMOTE_WRITE = 0x0040, /* allow remote writes */
+ MEM_WINDOW_BIND = 0x0080, /* binds allowed */
+ MEM_SHARED = 0x0100, /* set if MR is shared */
+ MEM_STAG_VALID = 0x0200 /* set if STAG is in valid state */
+};
+
+/*
+ * CCIL API ACF flags defined in terms of the low level mem flags.
+ * This minimizes translation needed in the user API
+ */
+enum c2_acf {
+ C2_ACF_LOCAL_READ = MEM_LOCAL_READ,
+ C2_ACF_LOCAL_WRITE = MEM_LOCAL_WRITE,
+ C2_ACF_REMOTE_READ = MEM_REMOTE_READ,
+ C2_ACF_REMOTE_WRITE = MEM_REMOTE_WRITE,
+ C2_ACF_WINDOW_BIND = MEM_WINDOW_BIND
+};
+
+/*
+ * Image types of objects written to flash
+ */
+#define C2_FLASH_IMG_BITFILE 1
+#define C2_FLASH_IMG_OPTION_ROM 2
+#define C2_FLASH_IMG_VPD 3
+
+/*
+ * to fix bug 1815 we define the max size allowable of the
+ * terminate message (per the IETF spec).Refer to the IETF
+ * protocal specification, section 12.1.6, page 64)
+ * The message is prefixed by 20 types of DDP info.
+ *
+ * Then the message has 6 bytes for the terminate control
+ * and DDP segment length info plus a DDP header (either
+ * 14 or 18 byts) plus 28 bytes for the RDMA header.
+ * Thus the max size in:
+ * 20 + (6 + 18 + 28) = 72
+ */
+#define C2_MAX_TERMINATE_MESSAGE_SIZE (72)
+
+/*
+ * Build String Length. It must be the same as C2_BUILD_STR_LEN in ccil_api.h
+ */
+#define WR_BUILD_STR_LEN 64
+
+/*
+ * WARNING: All of these structs need to align any 64bit types on
+ * 64 bit boundaries! 64bit types include u64 and u64.
+ */
+
+/*
+ * Clustercore Work Request Header. Be sensitive to field layout
+ * and alignment.
+ */
+struct c2wr_hdr {
+ /* wqe_count is part of the cqe. It is put here so the
+ * adapter can write to it while the wr is pending without
+ * clobbering part of the wr. This word need not be dma'd
+ * from the host to adapter by libccil, but we copy it anyway
+ * to make the memcpy to the adapter better aligned.
+ */
+ u32 wqe_count;
+
+ /* Put these fields next so that later 32- and 64-bit
+ * quantities are naturally aligned.
+ */
+ u8 id;
+ u8 result; /* adapter -> host */
+ u8 sge_count; /* host -> adapter */
+ u8 flags; /* host -> adapter */
+
+ u64 context;
+#ifdef CCMSGMAGIC
+ u32 magic;
+ u32 pad;
+#endif
+} __attribute__((packed));
+
+/*
+ *------------------------ RNIC ------------------------
+ */
+
+/*
+ * WR_RNIC_OPEN
+ */
+
+/*
+ * Flags for the RNIC WRs
+ */
+enum c2_rnic_flags {
+ RNIC_IRD_STATIC = 0x0001,
+ RNIC_ORD_STATIC = 0x0002,
+ RNIC_QP_STATIC = 0x0004,
+ RNIC_SRQ_SUPPORTED = 0x0008,
+ RNIC_PBL_BLOCK_MODE = 0x0010,
+ RNIC_SRQ_MODEL_ARRIVAL = 0x0020,
+ RNIC_CQ_OVF_DETECTED = 0x0040,
+ RNIC_PRIV_MODE = 0x0080
+};
+
+struct c2wr_rnic_open_req {
+ struct c2wr_hdr hdr;
+ u64 user_context;
+ u16 flags; /* See enum c2_rnic_flags */
+ u16 port_num;
+} __attribute__((packed));
+
+struct c2wr_rnic_open_rep {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+} __attribute__((packed));
+
+union c2wr_rnic_open {
+ struct c2wr_rnic_open_req req;
+ struct c2wr_rnic_open_rep rep;
+} __attribute__((packed));
+
+struct c2wr_rnic_query_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+} __attribute__((packed));
+
+/*
+ * WR_RNIC_QUERY
+ */
+struct c2wr_rnic_query_rep {
+ struct c2wr_hdr hdr;
+ u64 user_context;
+ u32 vendor_id;
+ u32 part_number;
+ u32 hw_version;
+ u32 fw_ver_major;
+ u32 fw_ver_minor;
+ u32 fw_ver_patch;
+ char fw_ver_build_str[WR_BUILD_STR_LEN];
+ u32 max_qps;
+ u32 max_qp_depth;
+ u32 max_srq_depth;
+ u32 max_send_sgl_depth;
+ u32 max_rdma_sgl_depth;
+ u32 max_cqs;
+ u32 max_cq_depth;
+ u32 max_cq_event_handlers;
+ u32 max_mrs;
+ u32 max_pbl_depth;
+ u32 max_pds;
+ u32 max_global_ird;
+ u32 max_global_ord;
+ u32 max_qp_ird;
+ u32 max_qp_ord;
+ u32 flags;
+ u32 max_mws;
+ u32 pbe_range_low;
+ u32 pbe_range_high;
+ u32 max_srqs;
+ u32 page_size;
+} __attribute__((packed));
+
+union c2wr_rnic_query {
+ struct c2wr_rnic_query_req req;
+ struct c2wr_rnic_query_rep rep;
+} __attribute__((packed));
+
+/*
+ * WR_RNIC_GETCONFIG
+ */
+
+struct c2wr_rnic_getconfig_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 option; /* see c2_getconfig_cmd_t */
+ u64 reply_buf;
+ u32 reply_buf_len;
+} __attribute__((packed)) ;
+
+struct c2wr_rnic_getconfig_rep {
+ struct c2wr_hdr hdr;
+ u32 option; /* see c2_getconfig_cmd_t */
+ u32 count_len; /* length of the number of addresses configured */
+} __attribute__((packed)) ;
+
+union c2wr_rnic_getconfig {
+ struct c2wr_rnic_getconfig_req req;
+ struct c2wr_rnic_getconfig_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ * WR_RNIC_SETCONFIG
+ */
+struct c2wr_rnic_setconfig_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 option; /* See c2_setconfig_cmd_t */
+ /* variable data and pad. See c2_netaddr and c2_route */
+ u8 data[0];
+} __attribute__((packed)) ;
+
+struct c2wr_rnic_setconfig_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_rnic_setconfig {
+ struct c2wr_rnic_setconfig_req req;
+ struct c2wr_rnic_setconfig_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ * WR_RNIC_CLOSE
+ */
+struct c2wr_rnic_close_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+} __attribute__((packed)) ;
+
+struct c2wr_rnic_close_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_rnic_close {
+ struct c2wr_rnic_close_req req;
+ struct c2wr_rnic_close_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ *------------------------ CQ ------------------------
+ */
+struct c2wr_cq_create_req {
+ struct c2wr_hdr hdr;
+ u64 shared_ht;
+ u64 user_context;
+ u64 msg_pool;
+ u32 rnic_handle;
+ u32 msg_size;
+ u32 depth;
+} __attribute__((packed)) ;
+
+struct c2wr_cq_create_rep {
+ struct c2wr_hdr hdr;
+ u32 mq_index;
+ u32 adapter_shared;
+ u32 cq_handle;
+} __attribute__((packed)) ;
+
+union c2wr_cq_create {
+ struct c2wr_cq_create_req req;
+ struct c2wr_cq_create_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_cq_modify_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 cq_handle;
+ u32 new_depth;
+ u64 new_msg_pool;
+} __attribute__((packed)) ;
+
+struct c2wr_cq_modify_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_cq_modify {
+ struct c2wr_cq_modify_req req;
+ struct c2wr_cq_modify_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_cq_destroy_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 cq_handle;
+} __attribute__((packed)) ;
+
+struct c2wr_cq_destroy_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_cq_destroy {
+ struct c2wr_cq_destroy_req req;
+ struct c2wr_cq_destroy_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ *------------------------ PD ------------------------
+ */
+struct c2wr_pd_alloc_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 pd_id;
+} __attribute__((packed)) ;
+
+struct c2wr_pd_alloc_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_pd_alloc {
+ struct c2wr_pd_alloc_req req;
+ struct c2wr_pd_alloc_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_pd_dealloc_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 pd_id;
+} __attribute__((packed)) ;
+
+struct c2wr_pd_dealloc_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_pd_dealloc {
+ struct c2wr_pd_dealloc_req req;
+ struct c2wr_pd_dealloc_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ *------------------------ SRQ ------------------------
+ */
+struct c2wr_srq_create_req {
+ struct c2wr_hdr hdr;
+ u64 shared_ht;
+ u64 user_context;
+ u32 rnic_handle;
+ u32 srq_depth;
+ u32 srq_limit;
+ u32 sgl_depth;
+ u32 pd_id;
+} __attribute__((packed)) ;
+
+struct c2wr_srq_create_rep {
+ struct c2wr_hdr hdr;
+ u32 srq_depth;
+ u32 sgl_depth;
+ u32 msg_size;
+ u32 mq_index;
+ u32 mq_start;
+ u32 srq_handle;
+} __attribute__((packed)) ;
+
+union c2wr_srq_create {
+ struct c2wr_srq_create_req req;
+ struct c2wr_srq_create_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_srq_destroy_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 srq_handle;
+} __attribute__((packed)) ;
+
+struct c2wr_srq_destroy_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_srq_destroy {
+ struct c2wr_srq_destroy_req req;
+ struct c2wr_srq_destroy_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ *------------------------ QP ------------------------
+ */
+enum c2wr_qp_flags {
+ QP_RDMA_READ = 0x00000001, /* RDMA read enabled? */
+ QP_RDMA_WRITE = 0x00000002, /* RDMA write enabled? */
+ QP_MW_BIND = 0x00000004, /* MWs enabled */
+ QP_ZERO_STAG = 0x00000008, /* enabled? */
+ QP_REMOTE_TERMINATION = 0x00000010, /* remote end terminated */
+ QP_RDMA_READ_RESPONSE = 0x00000020 /* Remote RDMA read */
+ /* enabled? */
+};
+
+struct c2wr_qp_create_req {
+ struct c2wr_hdr hdr;
+ u64 shared_sq_ht;
+ u64 shared_rq_ht;
+ u64 user_context;
+ u32 rnic_handle;
+ u32 sq_cq_handle;
+ u32 rq_cq_handle;
+ u32 sq_depth;
+ u32 rq_depth;
+ u32 srq_handle;
+ u32 srq_limit;
+ u32 flags; /* see enum c2wr_qp_flags */
+ u32 send_sgl_depth;
+ u32 recv_sgl_depth;
+ u32 rdma_write_sgl_depth;
+ u32 ord;
+ u32 ird;
+ u32 pd_id;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_create_rep {
+ struct c2wr_hdr hdr;
+ u32 sq_depth;
+ u32 rq_depth;
+ u32 send_sgl_depth;
+ u32 recv_sgl_depth;
+ u32 rdma_write_sgl_depth;
+ u32 ord;
+ u32 ird;
+ u32 sq_msg_size;
+ u32 sq_mq_index;
+ u32 sq_mq_start;
+ u32 rq_msg_size;
+ u32 rq_mq_index;
+ u32 rq_mq_start;
+ u32 qp_handle;
+} __attribute__((packed)) ;
+
+union c2wr_qp_create {
+ struct c2wr_qp_create_req req;
+ struct c2wr_qp_create_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_query_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 qp_handle;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_query_rep {
+ struct c2wr_hdr hdr;
+ u64 user_context;
+ u32 rnic_handle;
+ u32 sq_depth;
+ u32 rq_depth;
+ u32 send_sgl_depth;
+ u32 rdma_write_sgl_depth;
+ u32 recv_sgl_depth;
+ u32 ord;
+ u32 ird;
+ u16 qp_state;
+ u16 flags; /* see c2wr_qp_flags_t */
+ u32 qp_id;
+ u32 local_addr;
+ u32 remote_addr;
+ u16 local_port;
+ u16 remote_port;
+ u32 terminate_msg_length; /* 0 if not present */
+ u8 data[0];
+ /* Terminate Message in-line here. */
+} __attribute__((packed)) ;
+
+union c2wr_qp_query {
+ struct c2wr_qp_query_req req;
+ struct c2wr_qp_query_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_modify_req {
+ struct c2wr_hdr hdr;
+ u64 stream_msg;
+ u32 stream_msg_length;
+ u32 rnic_handle;
+ u32 qp_handle;
+ u32 next_qp_state;
+ u32 ord;
+ u32 ird;
+ u32 sq_depth;
+ u32 rq_depth;
+ u32 llp_ep_handle;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_modify_rep {
+ struct c2wr_hdr hdr;
+ u32 ord;
+ u32 ird;
+ u32 sq_depth;
+ u32 rq_depth;
+ u32 sq_msg_size;
+ u32 sq_mq_index;
+ u32 sq_mq_start;
+ u32 rq_msg_size;
+ u32 rq_mq_index;
+ u32 rq_mq_start;
+} __attribute__((packed)) ;
+
+union c2wr_qp_modify {
+ struct c2wr_qp_modify_req req;
+ struct c2wr_qp_modify_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_destroy_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 qp_handle;
+} __attribute__((packed)) ;
+
+struct c2wr_qp_destroy_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_qp_destroy {
+ struct c2wr_qp_destroy_req req;
+ struct c2wr_qp_destroy_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ * The CCWR_QP_CONNECT msg is posted on the verbs request queue. It can
+ * only be posted when a QP is in IDLE state. After the connect request is
+ * submitted to the LLP, the adapter moves the QP to CONNECT_PENDING state.
+ * No synchronous reply from adapter to this WR. The results of
+ * connection are passed back in an async event CCAE_ACTIVE_CONNECT_RESULTS
+ * See c2wr_ae_active_connect_results_t
+ */
+struct c2wr_qp_connect_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 qp_handle;
+ u32 remote_addr;
+ u16 remote_port;
+ u16 pad;
+ u32 private_data_length;
+ u8 private_data[0]; /* Private data in-line. */
+} __attribute__((packed)) ;
+
+struct c2wr_qp_connect {
+ struct c2wr_qp_connect_req req;
+ /* no synchronous reply. */
+} __attribute__((packed)) ;
+
+
+/*
+ *------------------------ MM ------------------------
+ */
+
+struct c2wr_nsmr_stag_alloc_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 pbl_depth;
+ u32 pd_id;
+ u32 flags;
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_stag_alloc_rep {
+ struct c2wr_hdr hdr;
+ u32 pbl_depth;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+union c2wr_nsmr_stag_alloc {
+ struct c2wr_nsmr_stag_alloc_req req;
+ struct c2wr_nsmr_stag_alloc_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_register_req {
+ struct c2wr_hdr hdr;
+ u64 va;
+ u32 rnic_handle;
+ u16 flags;
+ u8 stag_key;
+ u8 pad;
+ u32 pd_id;
+ u32 pbl_depth;
+ u32 pbe_size;
+ u32 fbo;
+ u32 length;
+ u32 addrs_length;
+ /* array of paddrs (must be aligned on a 64bit boundary) */
+ u64 paddrs[0];
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_register_rep {
+ struct c2wr_hdr hdr;
+ u32 pbl_depth;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+union c2wr_nsmr_register {
+ struct c2wr_nsmr_register_req req;
+ struct c2wr_nsmr_register_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_pbl_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 flags;
+ u32 stag_index;
+ u32 addrs_length;
+ /* array of paddrs (must be aligned on a 64bit boundary) */
+ u64 paddrs[0];
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_pbl_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_nsmr_pbl {
+ struct c2wr_nsmr_pbl_req req;
+ struct c2wr_nsmr_pbl_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_mr_query_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+struct c2wr_mr_query_rep {
+ struct c2wr_hdr hdr;
+ u8 stag_key;
+ u8 pad[3];
+ u32 pd_id;
+ u32 flags;
+ u32 pbl_depth;
+} __attribute__((packed)) ;
+
+union c2wr_mr_query {
+ struct c2wr_mr_query_req req;
+ struct c2wr_mr_query_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_mw_query_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+struct c2wr_mw_query_rep {
+ struct c2wr_hdr hdr;
+ u8 stag_key;
+ u8 pad[3];
+ u32 pd_id;
+ u32 flags;
+} __attribute__((packed)) ;
+
+union c2wr_mw_query {
+ struct c2wr_mw_query_req req;
+ struct c2wr_mw_query_rep rep;
+} __attribute__((packed)) ;
+
+
+struct c2wr_stag_dealloc_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+struct c2wr_stag_dealloc_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed)) ;
+
+union c2wr_stag_dealloc {
+ struct c2wr_stag_dealloc_req req;
+ struct c2wr_stag_dealloc_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_reregister_req {
+ struct c2wr_hdr hdr;
+ u64 va;
+ u32 rnic_handle;
+ u16 flags;
+ u8 stag_key;
+ u8 pad;
+ u32 stag_index;
+ u32 pd_id;
+ u32 pbl_depth;
+ u32 pbe_size;
+ u32 fbo;
+ u32 length;
+ u32 addrs_length;
+ u32 pad1;
+ /* array of paddrs (must be aligned on a 64bit boundary) */
+ u64 paddrs[0];
+} __attribute__((packed)) ;
+
+struct c2wr_nsmr_reregister_rep {
+ struct c2wr_hdr hdr;
+ u32 pbl_depth;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+union c2wr_nsmr_reregister {
+ struct c2wr_nsmr_reregister_req req;
+ struct c2wr_nsmr_reregister_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_smr_register_req {
+ struct c2wr_hdr hdr;
+ u64 va;
+ u32 rnic_handle;
+ u16 flags;
+ u8 stag_key;
+ u8 pad;
+ u32 stag_index;
+ u32 pd_id;
+} __attribute__((packed)) ;
+
+struct c2wr_smr_register_rep {
+ struct c2wr_hdr hdr;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+union c2wr_smr_register {
+ struct c2wr_smr_register_req req;
+ struct c2wr_smr_register_rep rep;
+} __attribute__((packed)) ;
+
+struct c2wr_mw_alloc_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 pd_id;
+} __attribute__((packed)) ;
+
+struct c2wr_mw_alloc_rep {
+ struct c2wr_hdr hdr;
+ u32 stag_index;
+} __attribute__((packed)) ;
+
+union c2wr_mw_alloc {
+ struct c2wr_mw_alloc_req req;
+ struct c2wr_mw_alloc_rep rep;
+} __attribute__((packed)) ;
+
+/*
+ *------------------------ WRs -----------------------
+ */
+
+struct c2wr_user_hdr {
+ struct c2wr_hdr hdr; /* Has status and WR Type */
+} __attribute__((packed)) ;
+
+enum c2_qp_state {
+ C2_QP_STATE_IDLE = 0x01,
+ C2_QP_STATE_CONNECTING = 0x02,
+ C2_QP_STATE_RTS = 0x04,
+ C2_QP_STATE_CLOSING = 0x08,
+ C2_QP_STATE_TERMINATE = 0x10,
+ C2_QP_STATE_ERROR = 0x20,
+};
+
+/* Completion queue entry. */
+struct c2wr_ce {
+ struct c2wr_hdr hdr; /* Has status and WR Type */
+ u64 qp_user_context; /* c2_user_qp_t * */
+ u32 qp_state; /* Current QP State */
+ u32 handle; /* QPID or EP Handle */
+ u32 bytes_rcvd; /* valid for RECV WCs */
+ u32 stag;
+} __attribute__((packed)) ;
+
+
+/*
+ * Flags used for all post-sq WRs. These must fit in the flags
+ * field of the struct c2wr_hdr (eight bits).
+ */
+enum {
+ SQ_SIGNALED = 0x01,
+ SQ_READ_FENCE = 0x02,
+ SQ_FENCE = 0x04,
+};
+
+/*
+ * Common fields for all post-sq WRs. Namely the standard header and a
+ * secondary header with fields common to all post-sq WRs.
+ */
+struct c2_sq_hdr {
+ struct c2wr_user_hdr user_hdr;
+} __attribute__((packed));
+
+/*
+ * Same as above but for post-rq WRs.
+ */
+struct c2_rq_hdr {
+ struct c2wr_user_hdr user_hdr;
+} __attribute__((packed));
+
+/*
+ * use the same struct for all sends.
+ */
+struct c2wr_send_req {
+ struct c2_sq_hdr sq_hdr;
+ u32 sge_len;
+ u32 remote_stag;
+ u8 data[0]; /* SGE array */
+} __attribute__((packed));
+/* XXX c2wr_send_req_t, c2wr_send_se_req_t, c2wr_send_inv_req_t,
+ c2wr_send_se_inv_req_t;*/
+
+union c2wr_send {
+ struct c2wr_send_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+struct c2wr_rdma_write_req {
+ struct c2_sq_hdr sq_hdr;
+ u64 remote_to;
+ u32 remote_stag;
+ u32 sge_len;
+ u8 data[0]; /* SGE array */
+} __attribute__((packed));
+
+union c2wr_rdma_write {
+ struct c2wr_rdma_write_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+struct c2wr_rdma_read_req {
+ struct c2_sq_hdr sq_hdr;
+ u64 local_to;
+ u64 remote_to;
+ u32 local_stag;
+ u32 remote_stag;
+ u32 length;
+} __attribute__((packed));
+
+union c2wr_rdma_read {
+ struct c2wr_rdma_read_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+struct c2wr_mw_bind_req {
+ struct c2_sq_hdr sq_hdr;
+ u64 va;
+ u8 stag_key;
+ u8 pad[3];
+ u32 mw_stag_index;
+ u32 mr_stag_index;
+ u32 length;
+ u32 flags;
+} __attribute__((packed));
+
+union c2wr_mw_bind {
+ struct c2wr_mw_bind_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+struct c2wr_nsmr_fastreg_req {
+ struct c2_sq_hdr sq_hdr;
+ u64 va;
+ u8 stag_key;
+ u8 pad[3];
+ u32 stag_index;
+ u32 pbe_size;
+ u32 fbo;
+ u32 length;
+ u32 addrs_length;
+ /* array of paddrs (must be aligned on a 64bit boundary) */
+ u64 paddrs[0];
+} __attribute__((packed));
+
+union c2wr_nsmr_fastreg {
+ struct c2wr_nsmr_fastreg_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+struct c2wr_stag_invalidate_req {
+ struct c2_sq_hdr sq_hdr;
+ u8 stag_key;
+ u8 pad[3];
+ u32 stag_index;
+} __attribute__((packed));
+
+union c2wr_stag_invalidate {
+ struct c2wr_stag_invalidate_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+union c2wr_sqwr {
+ struct c2_sq_hdr sq_hdr;
+ struct c2wr_send_req send;
+ struct c2wr_send_req send_se;
+ struct c2wr_send_req send_inv;
+ struct c2wr_send_req send_se_inv;
+ struct c2wr_rdma_write_req rdma_write;
+ struct c2wr_rdma_read_req rdma_read;
+ struct c2wr_mw_bind_req mw_bind;
+ struct c2wr_nsmr_fastreg_req nsmr_fastreg;
+ struct c2wr_stag_invalidate_req stag_inv;
+} __attribute__((packed));
+
+
+/*
+ * RQ WRs
+ */
+struct c2wr_rqwr {
+ struct c2_rq_hdr rq_hdr;
+ u8 data[0]; /* array of SGEs */
+} __attribute__((packed));
+/* XXX c2wr_rqwr_t, c2wr_recv_req_t; */
+
+union c2wr_recv {
+ struct c2wr_rqwr req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+/*
+ * All AEs start with this header. Most AEs only need to convey the
+ * information in the header. Some, like LLP connection events, need
+ * more info. The union typdef c2wr_ae_t has all the possible AEs.
+ *
+ * hdr.context is the user_context from the rnic_open WR. NULL If this
+ * is not affiliated with an rnic
+ *
+ * hdr.id is the AE identifier (eg; CCAE_REMOTE_SHUTDOWN,
+ * CCAE_LLP_CLOSE_COMPLETE)
+ *
+ * resource_type is one of: C2_RES_IND_QP, C2_RES_IND_CQ, C2_RES_IND_SRQ
+ *
+ * user_context is the context passed down when the host created the resource.
+ */
+struct c2wr_ae_hdr {
+ struct c2wr_hdr hdr;
+ u64 user_context; /* user context for this res. */
+ u32 resource_type; /* see enum c2_resource_indicator */
+ u32 resource; /* handle for resource */
+ u32 qp_state; /* current QP State */
+} __attribute__((packed));
+
+/*
+ * After submitting the CCAE_ACTIVE_CONNECT_RESULTS message on the AEQ,
+ * the adapter moves the QP into RTS state
+ */
+struct c2wr_ae_active_connect_results {
+ struct c2wr_ae_hdr ae_hdr;
+ u32 laddr;
+ u32 raddr;
+ u16 lport;
+ u16 rport;
+ u32 private_data_length;
+ u8 private_data[0]; /* data is in-line in the msg. */
+} __attribute__((packed));
+
+/*
+ * When connections are established by the stack (and the private data
+ * MPA frame is received), the adapter will generate an event to the host.
+ * The details of the connection, any private data, and the new connection
+ * request handle is passed up via the CCAE_CONNECTION_REQUEST msg on the
+ * AE queue:
+ */
+struct c2wr_ae_connection_request {
+ struct c2wr_ae_hdr ae_hdr;
+ u32 cr_handle; /* connreq handle (sock ptr) */
+ u32 laddr;
+ u32 raddr;
+ u16 lport;
+ u16 rport;
+ u32 private_data_length;
+ u8 private_data[0]; /* data is in-line in the msg. */
+} __attribute__((packed));
+
+union c2wr_ae {
+ struct c2wr_ae_hdr ae_generic;
+ struct c2wr_ae_active_connect_results ae_active_connect_results;
+ struct c2wr_ae_connection_request ae_connection_request;
+} __attribute__((packed));
+
+struct c2wr_init_req {
+ struct c2wr_hdr hdr;
+ u64 hint_count;
+ u64 q0_host_shared;
+ u64 q1_host_shared;
+ u64 q1_host_msg_pool;
+ u64 q2_host_shared;
+ u64 q2_host_msg_pool;
+} __attribute__((packed));
+
+struct c2wr_init_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed));
+
+union c2wr_init {
+ struct c2wr_init_req req;
+ struct c2wr_init_rep rep;
+} __attribute__((packed));
+
+/*
+ * For upgrading flash.
+ */
+
+struct c2wr_flash_init_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+} __attribute__((packed));
+
+struct c2wr_flash_init_rep {
+ struct c2wr_hdr hdr;
+ u32 adapter_flash_buf_offset;
+ u32 adapter_flash_len;
+} __attribute__((packed));
+
+union c2wr_flash_init {
+ struct c2wr_flash_init_req req;
+ struct c2wr_flash_init_rep rep;
+} __attribute__((packed));
+
+struct c2wr_flash_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 len;
+} __attribute__((packed));
+
+struct c2wr_flash_rep {
+ struct c2wr_hdr hdr;
+ u32 status;
+} __attribute__((packed));
+
+union c2wr_flash {
+ struct c2wr_flash_req req;
+ struct c2wr_flash_rep rep;
+} __attribute__((packed));
+
+struct c2wr_buf_alloc_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 size;
+} __attribute__((packed));
+
+struct c2wr_buf_alloc_rep {
+ struct c2wr_hdr hdr;
+ u32 offset; /* 0 if mem not available */
+ u32 size; /* 0 if mem not available */
+} __attribute__((packed));
+
+union c2wr_buf_alloc {
+ struct c2wr_buf_alloc_req req;
+ struct c2wr_buf_alloc_rep rep;
+} __attribute__((packed));
+
+struct c2wr_buf_free_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 offset; /* Must match value from alloc */
+ u32 size; /* Must match value from alloc */
+} __attribute__((packed));
+
+struct c2wr_buf_free_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed));
+
+union c2wr_buf_free {
+ struct c2wr_buf_free_req req;
+ struct c2wr_ce rep;
+} __attribute__((packed));
+
+struct c2wr_flash_write_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 offset;
+ u32 size;
+ u32 type;
+ u32 flags;
+} __attribute__((packed));
+
+struct c2wr_flash_write_rep {
+ struct c2wr_hdr hdr;
+ u32 status;
+} __attribute__((packed));
+
+union c2wr_flash_write {
+ struct c2wr_flash_write_req req;
+ struct c2wr_flash_write_rep rep;
+} __attribute__((packed));
+
+/*
+ * Messages for LLP connection setup.
+ */
+
+/*
+ * Listen Request. This allocates a listening endpoint to allow passive
+ * connection setup. Newly established LLP connections are passed up
+ * via an AE. See c2wr_ae_connection_request_t
+ */
+struct c2wr_ep_listen_create_req {
+ struct c2wr_hdr hdr;
+ u64 user_context; /* returned in AEs. */
+ u32 rnic_handle;
+ u32 local_addr; /* local addr, or 0 */
+ u16 local_port; /* 0 means "pick one" */
+ u16 pad;
+ u32 backlog; /* tradional tcp listen bl */
+} __attribute__((packed));
+
+struct c2wr_ep_listen_create_rep {
+ struct c2wr_hdr hdr;
+ u32 ep_handle; /* handle to new listening ep */
+ u16 local_port; /* resulting port... */
+ u16 pad;
+} __attribute__((packed));
+
+union c2wr_ep_listen_create {
+ struct c2wr_ep_listen_create_req req;
+ struct c2wr_ep_listen_create_rep rep;
+} __attribute__((packed));
+
+struct c2wr_ep_listen_destroy_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 ep_handle;
+} __attribute__((packed));
+
+struct c2wr_ep_listen_destroy_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed));
+
+union c2wr_ep_listen_destroy {
+ struct c2wr_ep_listen_destroy_req req;
+ struct c2wr_ep_listen_destroy_rep rep;
+} __attribute__((packed));
+
+struct c2wr_ep_query_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 ep_handle;
+} __attribute__((packed));
+
+struct c2wr_ep_query_rep {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 local_addr;
+ u32 remote_addr;
+ u16 local_port;
+ u16 remote_port;
+} __attribute__((packed));
+
+union c2wr_ep_query {
+ struct c2wr_ep_query_req req;
+ struct c2wr_ep_query_rep rep;
+} __attribute__((packed));
+
+
+/*
+ * The host passes this down to indicate acceptance of a pending iWARP
+ * connection. The cr_handle was obtained from the CONNECTION_REQUEST
+ * AE passed up by the adapter. See c2wr_ae_connection_request_t.
+ */
+struct c2wr_cr_accept_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 qp_handle; /* QP to bind to this LLP conn */
+ u32 ep_handle; /* LLP handle to accept */
+ u32 private_data_length;
+ u8 private_data[0]; /* data in-line in msg. */
+} __attribute__((packed));
+
+/*
+ * adapter sends reply when private data is successfully submitted to
+ * the LLP.
+ */
+struct c2wr_cr_accept_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed));
+
+union c2wr_cr_accept {
+ struct c2wr_cr_accept_req req;
+ struct c2wr_cr_accept_rep rep;
+} __attribute__((packed));
+
+/*
+ * The host sends this down if a given iWARP connection request was
+ * rejected by the consumer. The cr_handle was obtained from a
+ * previous c2wr_ae_connection_request_t AE sent by the adapter.
+ */
+struct c2wr_cr_reject_req {
+ struct c2wr_hdr hdr;
+ u32 rnic_handle;
+ u32 ep_handle; /* LLP handle to reject */
+} __attribute__((packed));
+
+/*
+ * Dunno if this is needed, but we'll add it for now. The adapter will
+ * send the reject_reply after the LLP endpoint has been destroyed.
+ */
+struct c2wr_cr_reject_rep {
+ struct c2wr_hdr hdr;
+} __attribute__((packed));
+
+union c2wr_cr_reject {
+ struct c2wr_cr_reject_req req;
+ struct c2wr_cr_reject_rep rep;
+} __attribute__((packed));
+
+/*
+ * console command. Used to implement a debug console over the verbs
+ * request and reply queues.
+ */
+
+/*
+ * Console request message. It contains:
+ * - message hdr with id = CCWR_CONSOLE
+ * - the physaddr/len of host memory to be used for the reply.
+ * - the command string. eg: "netstat -s" or "zoneinfo"
+ */
+struct c2wr_console_req {
+ struct c2wr_hdr hdr; /* id = CCWR_CONSOLE */
+ u64 reply_buf; /* pinned host buf for reply */
+ u32 reply_buf_len; /* length of reply buffer */
+ u8 command[0]; /* NUL terminated ascii string */
+ /* containing the command req */
+} __attribute__((packed));
+
+/*
+ * flags used in the console reply.
+ */
+enum c2_console_flags {
+ CONS_REPLY_TRUNCATED = 0x00000001 /* reply was truncated */
+} __attribute__((packed));
+
+/*
+ * Console reply message.
+ * hdr.result contains the c2_status_t error if the reply was _not_ generated,
+ * or C2_OK if the reply was generated.
+ */
+struct c2wr_console_rep {
+ struct c2wr_hdr hdr; /* id = CCWR_CONSOLE */
+ u32 flags;
+} __attribute__((packed));
+
+union c2wr_console {
+ struct c2wr_console_req req;
+ struct c2wr_console_rep rep;
+} __attribute__((packed));
+
+
+/*
+ * Giant union with all WRs. Makes life easier...
+ */
+union c2wr {
+ struct c2wr_hdr hdr;
+ struct c2wr_user_hdr user_hdr;
+ union c2wr_rnic_open rnic_open;
+ union c2wr_rnic_query rnic_query;
+ union c2wr_rnic_getconfig rnic_getconfig;
+ union c2wr_rnic_setconfig rnic_setconfig;
+ union c2wr_rnic_close rnic_close;
+ union c2wr_cq_create cq_create;
+ union c2wr_cq_modify cq_modify;
+ union c2wr_cq_destroy cq_destroy;
+ union c2wr_pd_alloc pd_alloc;
+ union c2wr_pd_dealloc pd_dealloc;
+ union c2wr_srq_create srq_create;
+ union c2wr_srq_destroy srq_destroy;
+ union c2wr_qp_create qp_create;
+ union c2wr_qp_query qp_query;
+ union c2wr_qp_modify qp_modify;
+ union c2wr_qp_destroy qp_destroy;
+ struct c2wr_qp_connect qp_connect;
+ union c2wr_nsmr_stag_alloc nsmr_stag_alloc;
+ union c2wr_nsmr_register nsmr_register;
+ union c2wr_nsmr_pbl nsmr_pbl;
+ union c2wr_mr_query mr_query;
+ union c2wr_mw_query mw_query;
+ union c2wr_stag_dealloc stag_dealloc;
+ union c2wr_sqwr sqwr;
+ struct c2wr_rqwr rqwr;
+ struct c2wr_ce ce;
+ union c2wr_ae ae;
+ union c2wr_init init;
+ union c2wr_ep_listen_create ep_listen_create;
+ union c2wr_ep_listen_destroy ep_listen_destroy;
+ union c2wr_cr_accept cr_accept;
+ union c2wr_cr_reject cr_reject;
+ union c2wr_console console;
+ union c2wr_flash_init flash_init;
+ union c2wr_flash flash;
+ union c2wr_buf_alloc buf_alloc;
+ union c2wr_buf_free buf_free;
+ union c2wr_flash_write flash_write;
+} __attribute__((packed));
+
+
+/*
+ * Accessors for the wr fields that are packed together tightly to
+ * reduce the wr message size. The wr arguments are void* so that
+ * either a struct c2wr*, a struct c2wr_hdr*, or a pointer to any of the types
+ * in the struct c2wr union can be passed in.
+ */
+static __inline__ u8 c2_wr_get_id(void *wr)
+{
+ return ((struct c2wr_hdr *) wr)->id;
+}
+static __inline__ void c2_wr_set_id(void *wr, u8 id)
+{
+ ((struct c2wr_hdr *) wr)->id = id;
+}
+static __inline__ u8 c2_wr_get_result(void *wr)
+{
+ return ((struct c2wr_hdr *) wr)->result;
+}
+static __inline__ void c2_wr_set_result(void *wr, u8 result)
+{
+ ((struct c2wr_hdr *) wr)->result = result;
+}
+static __inline__ u8 c2_wr_get_flags(void *wr)
+{
+ return ((struct c2wr_hdr *) wr)->flags;
+}
+static __inline__ void c2_wr_set_flags(void *wr, u8 flags)
+{
+ ((struct c2wr_hdr *) wr)->flags = flags;
+}
+static __inline__ u8 c2_wr_get_sge_count(void *wr)
+{
+ return ((struct c2wr_hdr *) wr)->sge_count;
+}
+static __inline__ void c2_wr_set_sge_count(void *wr, u8 sge_count)
+{
+ ((struct c2wr_hdr *) wr)->sge_count = sge_count;
+}
+static __inline__ u32 c2_wr_get_wqe_count(void *wr)
+{
+ return ((struct c2wr_hdr *) wr)->wqe_count;
+}
+static __inline__ void c2_wr_set_wqe_count(void *wr, u32 wqe_count)
+{
+ ((struct c2wr_hdr *) wr)->wqe_count = wqe_count;
+}
+
+#endif /* _C2_WR_H_ */
^ permalink raw reply related
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