On Tue, 27 Jan 2009 16:07:50 -0800 (PST)
David Miller wrote:
> From: Roland Dreier
> Date: Tue, 27 Jan 2009 15:53:16 -0800
>
> > but actually I still don't see how it's safe for a net driver to
> > call skb_linearize() from its transmit routine, since there's a
> > chance that that will uncondi
From: Roland Dreier
Date: Tue, 27 Jan 2009 15:53:16 -0800
> but actually I still don't see how it's safe for a net driver to
> call skb_linearize() from its transmit routine, since there's a
> chance that that will unconditionally enable BHs?
It's simply not allowed.
dev_queue_xmit() at a highe
Interesting... looks like an unfortunate interaction with unclear
locking rules. See below for full explanation.
BTW, what workload are you running to hit this?
I assume you have CONFIG_HIGHMEM set?
> WARNING: at kernel/softirq.c:136 local_bh_enable+0x9b/0xa0()
I assume this is
WARN_
On Tue, Jan 27, 2009 at 1:56 PM, Evgeniy Polyakov wrote:
>> > RDS_ prefix?
>>
>> Even needed for statics?
>
> It confuses tags and the like otherwise, and looks more consistent with
> the rest of the code. Likely it is not a must, but just better look.
Yup, will do, just was curious.
Regards --
On Tue, Jan 27, 2009 at 11:36:37AM -0800, Andrew Grover (andy.gro...@gmail.com)
wrote:
> On Tue, Jan 27, 2009 at 5:18 AM, Evgeniy Polyakov wrote:
> > On Mon, Jan 26, 2009 at 06:17:41PM -0800, Andy Grover
> > (andy.gro...@oracle.com) wrote:
> >> +static LIST_HEAD(transports);
> >> +static DECLARE
On Tue, Jan 27, 2009 at 5:18 AM, Evgeniy Polyakov wrote:
> On Mon, Jan 26, 2009 at 06:17:41PM -0800, Andy Grover
> (andy.gro...@oracle.com) wrote:
>> +static LIST_HEAD(transports);
>> +static DECLARE_RWSEM(trans_sem);
>> +
>
> RDS_ prefix?
Even needed for statics?
>> +int rds_trans_register(str
On Mon, Jan 26, 2009 at 11:27 PM, Rémi Denis-Courmont
wrote:
> You also need to add lock class declaration to net/core/sock.c, I believe.
Very true, thanks.
Regards -- Andy
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics
On Tue, Jan 27, 2009 at 7:34 AM, Steve Wise wrote:
> Hey Andy,
>
> Why didn't you include the iWARP transport as well?
Hi Steve,
As I mentioned on IRC, there are some ib/iw coexistence issues and
other minor bugs to resolve, and then I will include the iWARP code.
Regards -- Andy
__
On Mon, Jan 26, 2009 at 11:34 PM, Rémi Denis-Courmont
wrote:
>> +#ifndef PF_RDS
>> +#define PF_RDS AF_RDS
>> +#endif
>
> You should probably remove that and put the last patch of your series ahead of
> this one.
Yup will do.
>> +#ifndef SOL_RDS
>> +#define SOL_RDS 272
>> +#endi
Andrew Grover wrote:
On Tue, Jan 27, 2009 at 5:05 AM, Evgeniy Polyakov wrote:
+#define RDS_PORT 18634
+
What will happen if some application already uses that port?
RDS errors out.
Yeah we're going to want to get an assigned port at some point, I guess.
You should s
On Tue, Jan 27, 2009 at 5:05 AM, Evgeniy Polyakov wrote:
>> +#define RDS_PORT 18634
>> +
>
> What will happen if some application already uses that port?
RDS errors out.
Yeah we're going to want to get an assigned port at some point, I guess.
Regards -- Andy
On Mon, Jan 26, 2009 at 7:48 PM, Stephen Hemminger
wrote:
> So this is starting to look like another "Oracle special" like AIO
> and HugeTLB. That has lots of caveat restrictions on the application.
Yep it's a datacenter-centric protocol.
Regards -- Andy
_
On Tue, Jan 27, 2009 at 5:10 AM, Evgeniy Polyakov wrote:
> On Mon, Jan 26, 2009 at 06:17:40PM -0800, Andy Grover
> (andy.gro...@oracle.com) wrote:
>> +/*
>> + * Yes, a global lock. It's used so infrequently that it's worth keeping it
>> + * global to simplify the locking. It's only used in the
Ah ... ok. That worked. I was not setting any parameter on the server.
On Jan 27, 2009, at 1:45 PM, Steve Wise wrote:
Are you setting the -S 65 on the server side as sell? IE: the
parameters must match on the client and server.
Jaiwant Mulik wrote:
Hi all,
Here is the config I am running:
Below patch removes following compilation warnings :
drivers/infiniband/hw/nes/nes_cm.c:781: warning: unused variable 'tmp_addr'
drivers/infiniband/hw/nes/nes_cm.c:820: warning: unused variable 'tmp_addr'
Signed-off-by: Manish Katiyar
---
drivers/infiniband/hw/nes/nes_cm.c |4
1 files
Are you setting the -S 65 on the server side as sell? IE: the
parameters must match on the client and server.
Jaiwant Mulik wrote:
Hi all,
Here is the config I am running:
--
OS: Linux iwarp2 2.6.20-1.2320.fc5 #1 SMP Tue Jun 12 18:50:49 EDT 2007
x86_64 x86_64 x86
Hi all,
Here is the config I am running:
--
OS: Linux iwarp2 2.6.20-1.2320.fc5 #1 SMP Tue Jun 12 18:50:49 EDT 2007
x86_64 x86_64 x86_64 GNU/Linux
OFED: 1.3.1
librdmacm: 1.0.7
hw: Chelsio 302X
---
ulimit -l is set to 128
rping does not work
Hi Hal,
On 11:10 Tue 27 Jan , Hal Rosenstock wrote:
>
> Just because the class allows RMPP doesn't mean all operations use it
Ok, agree. But obviously it doesn't mean that RMPP should not be used.
So I think we are fine now.
> so no this wasn't a bug IMO. It was done that way for extensibil
Oliver Neukum wrote:
Am Tuesday 27 January 2009 14:34:19 schrieb Evgeniy Polyakov:
On Mon, Jan 26, 2009 at 06:17:43PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
+static inline int rds_conn_is_sending(struct rds_connection *conn)
+{
+ int ret = 0;
+
+ if (!mutex_tryl
Sasha,
On Tue, Jan 27, 2009 at 7:19 AM, Sasha Khapyorsky wrote:
> Hi Hal,
>
> On 17:14 Mon 26 Jan , Hal Rosenstock wrote:
>> On Mon, Jan 26, 2009 at 4:26 PM, Sasha Khapyorsky
>> wrote:
>> >
>> > This patch adds support for bigger than (256 - vendor2 data offset) data
>> > sending by ibsysst
On Tue, Jan 27, 2009 at 02:47:27PM +0100, Oliver Neukum (oli...@neukum.org)
wrote:
> > > +static inline int rds_conn_is_sending(struct rds_connection *conn)
> > > +{
> > > + int ret = 0;
> > > +
> > > + if (!mutex_trylock(&conn->c_send_lock))
> > > + ret = 1;
> > > + else
> > > + m
Am Tuesday 27 January 2009 14:34:19 schrieb Evgeniy Polyakov:
> On Mon, Jan 26, 2009 at 06:17:43PM -0800, Andy Grover
> (andy.gro...@oracle.com) wrote:
> > +static inline int rds_conn_is_sending(struct rds_connection *conn)
> > +{
> > + int ret = 0;
> > +
> > + if (!mutex_trylock(&conn->c_send
On Mon, Jan 26, 2009 at 06:17:43PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
> +static inline int rds_conn_is_sending(struct rds_connection *conn)
> +{
> + int ret = 0;
> +
> + if (!mutex_trylock(&conn->c_send_lock))
> + ret = 1;
> + else
> + mutex_unlo
On Mon, Jan 26, 2009 at 06:17:42PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
> +void rds_info_register_func(int optname, rds_info_func func)
> +{
> + int offset = optname - RDS_INFO_FIRST;
> +
> + BUG_ON(optname < RDS_INFO_FIRST || optname > RDS_INFO_LAST);
> +
> + spin_lock(&
On Mon, Jan 26, 2009 at 06:17:41PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
> +static LIST_HEAD(transports);
> +static DECLARE_RWSEM(trans_sem);
> +
RDS_ prefix?
> +int rds_trans_register(struct rds_transport *trans)
> +{
> + BUG_ON(strlen(trans->t_name) + 1 >
> +sizeof
On Mon, Jan 26, 2009 at 06:17:40PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
> +/*
> + * Yes, a global lock. It's used so infrequently that it's worth keeping it
> + * global to simplify the locking. It's only used in the following
> + * circumstances:
> + *
> + * - on connection build
Hi.
On Mon, Jan 26, 2009 at 06:17:39PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
> +/*
> + * XXX randomly chosen, but at least seems to be unused:
> + * # 18464-18768 Unassigned
> + * We should do better. We want a reserved port to discourage unpriv'ed
> + * userspace from
Hi Andy.
On Mon, Jan 26, 2009 at 06:17:38PM -0800, Andy Grover (andy.gro...@oracle.com)
wrote:
> +/* this is just used for stats gathering :/ */
Shouldn't this be some kind of per-cpu data?
> +static DEFINE_SPINLOCK(rds_sock_lock);
> +static unsigned long rds_sock_count;
> +static LIST_HEAD(rds
Hello,
I'm using "iw_nes" driver for NetEffect 10Gbit card under linux
2.6.27.10. I run it on PC with 8 processors.
>From time to time kernel shows following warning:
WARNING: at kernel/softirq.c:136 local_bh_enable+0x9b/0xa0()
Modules linked in: iscsi_trgt st sg scst_vdisk scst drbd twofish
two
Hello,
I'm using "iw_nes" driver for NetEffect 10Gbit card under linux
2.6.27.10. I run it on PC with 8 processors.
>From time to time kernel shows following warning:
WARNING: at kernel/softirq.c:136 local_bh_enable+0x9b/0xa0()
Modules linked in: iscsi_trgt st sg scst_vdisk scst drbd twofish
two
On Tuesday 27 January 2009 04:17:39 ext Andy Grover, you wrote:
> +/*
> + * XXX randomly chosen, but at least seems to be unused:
> + * # 18464-18768 Unassigned
> + * We should do better. We want a reserved port to discourage unpriv'ed
> + * userspace from listening.
> + */
> +#defin
On Tuesday 27 January 2009 04:17:58 ext Andy Grover, you wrote:
> RDS is a reliable datagram protocol used for IPC on Oracle
> database clusters. This adds address and protocol family numbers
> for it.
>
> Signed-off-by: Andy Grover
> ---
> include/linux/socket.h |4 +++-
> 1 files changed, 3
Hey Andy,
Why didn't you include the iWARP transport as well?
Andy Grover wrote:
Hi Roland,
This patchset adds support for RDS as an Infiniband ULP. RDS is an
Oracle-originated protocol used to send IPC datagrams (up to 1MB) reliably,
and is used currently in Oracle RAC and Exadata products.
Hi,
RHEL5.3 support was added to OFED-1.4 daily builds, starting from
OFED-1.4-20090127-0600.
OFED-1.4 daily builds are available under:
http://www.openfabrics.org/downloads/OFED/ofed-1.4-daily/
Regards,
Vladimir
___
general mailing list
general
On 14:19 Tue 27 Jan , Sasha Khapyorsky wrote:
>
> diff --git a/infiniband-diags/src/ibsysstat.c
> b/infiniband-diags/src/ibsysstat.c
> index c20a6f0..a145daf 100644
> --- a/infiniband-diags/src/ibsysstat.c
> +++ b/infiniband-diags/src/ibsysstat.c
> @@ -169,6 +174,11 @@ static char *ibsystat_s
Hi Hal,
On 17:14 Mon 26 Jan , Hal Rosenstock wrote:
> On Mon, Jan 26, 2009 at 4:26 PM, Sasha Khapyorsky wrote:
> >
> > This patch adds support for bigger than (256 - vendor2 data offset) data
> > sending by ibsysstat server using RMPP. It fixes bug#1237 - where server
> > output was truncated
Hello,
I'm using "iw_nes" driver for NetEffect 10Gbit card under linux
2.6.27.10. I run it on PC with 8 processors.
>From time to time kernel shows following warning:
WARNING: at kernel/softirq.c:136 local_bh_enable+0x9b/0xa0()
Modules linked in: iscsi_trgt st sg scst_vdisk scst drbd twofish
tw
This email was generated automatically, please do not reply
git_url: git://git.openfabrics.org/ofed_1_4/linux-2.6.git
git_branch: ofed_kernel
Common build parameters:
Passed:
Passed on i686 with linux-2.6.16
Passed on i686 with linux-2.6.17
Passed on i686 with linux-2.6.19
Passed on i686 with
On Mon, Jan 26, 2009 at 11:08 PM, Adit Ranadive wrote:
> Hello,
> I have been looking at doing some low level work with the OFED library 1.1
> in terms of figuring out how many bytes have been sent by an IB
> application.
If you have the source code, you can add a counter when posting send reques
On Monday 26 January 2009 19:00, Yossi Etigin wrote:
> There's a patch of mine in OFED that's probably exposing a bug in ipoib.
> The bug is that priv->broadcast can be NULL-ified and join_task does not
> protect the check with the spinlock.
> The patch may expose the bug because it uses rtnl_lock
40 matches
Mail list logo
