FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Debian > Debian Kernel

 
 
LinkBack Thread Tools
 
Old 12-09-2008, 09:54 AM
martin f krafft
 
Default Bug#506419: Driver from Supermicro fixes issue

tags 506419 patch
thanks

The attached patch against linux-2.6 2.6.26-11 seems to fix the
original problem for me, but I have not had a chance to verify the
second ('downstream') problem I identified in the bug report.

The patch is based on the driver source I obtained via Transtec from
Supermicro, the board manufacturer. It's quite ugly and big and
I don't have the time to dissect it right now, but if you want me to
try out particulars, I will.

The new driver has one shortcoming: after loading, the LAN card
becomes unusable, even just bridging to the IPMI card fails. Once
you `ip link set eth0 up`, IPMI starts working again after 3-5
seconds.

To address the problem, I have created
/etc/modprobe.d/local-forcedeth with the following contents:

install forcedeth modprobe --ignore-install forcedeth; ip link set eth0 up
options forcedeth debug=1

This causes a problem with IPv6 RA if eth0 is actually part of
a bridge, in which case I advise to just add a pre-up hook to the
bridge iface stanza to down the iface again. Downing the iface after
IPMI has started working again does *not* break it again.

I'd still really like this fixed because having such a hack in place
to keep IPMI working doesn't really make me more relaxed about
remote kernel upgrades.

Cheers,

--
.'`. martin f. krafft <madduck@d.o> Related projects:
: :' : proud Debian developer http://debiansystem.info
`. `'` http://people.debian.org/~madduck http://vcs-pkg.org
`- Debian - when you have better things to do than fixing systems
 
Old 12-09-2008, 10:11 AM
Bastian Blank
 
Default Bug#506419: Driver from Supermicro fixes issue

tags 506419 -patch
thanks

On Tue, Dec 09, 2008 at 11:54:03AM +0100, martin f krafft wrote:
> The attached patch against linux-2.6 2.6.26-11 seems to fix the
> original problem for me, but I have not had a chance to verify the
> second ('downstream') problem I identified in the bug report.

This is no patch which suffice our guidelines. Please show submittions
to the Linux maintainers and/or commit ids.

Bastian

--
There's coffee in that nebula!
-- Capt. Kathryn Janeway, Star Trek: Voyager, "The Cloud"



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-09-2008, 10:36 AM
martin f krafft
 
Default Bug#506419: Driver from Supermicro fixes issue

also sprach Bastian Blank <waldi@debian.org> [2008.12.09.1211 +0100]:
> This is no patch which suffice our guidelines. Please show
> submittions to the Linux maintainers and/or commit ids.

As stated in my message, I don't have the time to do this right now,
but I will try to get Transtec/Supermicro to push this stuff into
the kernel.

It's still a patch. I don't mind if you remove the patch tag, but it
might still be useful to some people. I, for one, don't expect this
to be fixed for lenny, or even any time soon, unfortunately.

Cheers,

--
.'`. martin f. krafft <madduck@d.o> Related projects:
: :' : proud Debian developer http://debiansystem.info
`. `'` http://people.debian.org/~madduck http://vcs-pkg.org
`- Debian - when you have better things to do than fixing systems
 
Old 12-09-2008, 12:13 PM
Bjørn Mork
 
Default Bug#506419: Driver from Supermicro fixes issue

martin f krafft <madduck@debian.org> writes:

> The attached patch against linux-2.6 2.6.26-11 seems to fix the
> original problem for me, but I have not had a chance to verify the
> second ('downstream') problem I identified in the bug report.
>
> The patch is based on the driver source I obtained via Transtec from
> Supermicro, the board manufacturer. It's quite ugly and big and
> I don't have the time to dissect it right now, but if you want me to
> try out particulars, I will.


Hmm, this part of your patch looks a lot like half of
edcfe5f7e307846e578fb88d69fa27051fded0ab which Bastian already has
pointed you to:

> @@ -5417,11 +6978,15 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i
> if (id->driver_data & DEV_HAS_LARGEDESC)
> np->pkt_limit = NV_PKTLIMIT_2;
>
> + dev->mtu = ETH_DATA_LEN;
> +
> if (id->driver_data & DEV_HAS_CHECKSUM) {
> np->rx_csum = 1;
> np->txrxctl_bits |= NVREG_TXRXCTL_RXCHECK;
> - dev->features |= NETIF_F_HW_CSUM | NETIF_F_SG;
> + dev->features |= NETIF_F_IP_CSUM|NETIF_F_SG;
> +#ifdef NETIF_F_TSO
> dev->features |= NETIF_F_TSO;
> +#endif
> }
>
> np->vlanctl_bits = 0;


This should not apply against linux-2.6 2.6.26-11 since it is already a
part of debian/patches/bugfix/all/stable/2.6.26.4.patch. Are you sure
your patch really does apply? If it does, then I'd suggest trying
*just* edcfe5f7e307846e578fb88d69fa27051fded0ab instead.


BTW, I also noticed that linux-2.6 2.6.26-11 includes the patch
linux-2.6-2.6.26/debian/patches/features/all/net-use-gso.patch
which very well may trigger the problem. You might also want to try
building without that patch.

In fact, I'm a little confused as to the reasons that patch is in
Debian. It is probably the correct thing to do, but it is also the kind
of thing that is bound to cause regressions because it will expose
latent bugs in drivers.

But then again, there's your performance requirements. I guess you'd
better get a proper NIC :-)


Bjørn
--
You have the narrow-mindedness of a disease carrier



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-09-2008, 01:06 PM
Bastian Blank
 
Default Bug#506419: Driver from Supermicro fixes issue

On Tue, Dec 09, 2008 at 02:13:36PM +0100, Bjørn Mork wrote:
> BTW, I also noticed that linux-2.6 2.6.26-11 includes the patch
> linux-2.6-2.6.26/debian/patches/features/all/net-use-gso.patch
> which very well may trigger the problem. You might also want to try
> building without that patch.
>
> In fact, I'm a little confused as to the reasons that patch is in
> Debian. It is probably the correct thing to do, but it is also the kind
> of thing that is bound to cause regressions because it will expose
> latent bugs in drivers.

This is in mainline: e5a4a72d4f88f4389e9340d383ca67031d1b8536. I don't
know what the rationale was to add this.

Bastian

--
I'm frequently appalled by the low regard you Earthmen have for life.
-- Spock, "The Galileo Seven", stardate 2822.3



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-09-2008, 01:14 PM
martin f krafft
 
Default Bug#506419: Driver from Supermicro fixes issue

also sprach Bjørn Mork <bjorn@mork.no> [2008.12.09.1413 +0100]:
> This should not apply against linux-2.6 2.6.26-11 since it is already a
> part of debian/patches/bugfix/all/stable/2.6.26.4.patch. Are you sure
> your patch really does apply? If it does, then I'd suggest trying
> *just* edcfe5f7e307846e578fb88d69fa27051fded0ab instead.

My patch does no "apply", it was made against 2.6.26-11 --
Supermicro just sent me their forcedeth.c and I made the patch.

> BTW, I also noticed that linux-2.6 2.6.26-11 includes the patch
> linux-2.6-2.6.26/debian/patches/features/all/net-use-gso.patch
> which very well may trigger the problem. You might also want to
> try building without that patch.

I have not tried -11 on the server, only -8 and -10, which both had
the problems.

How do I find out when net-use-gso.patch was added? The changelog
does not mention it, and I cannot find a repository for the source
(there are also no Vcs-* links in the package).

> But then again, there's your performance requirements. I guess
> you'd better get a proper NIC :-)

As soon as I get a second switch port, I will. While I only have one
switch port, I have no choice...

--
.'`. martin f. krafft <madduck@d.o> Related projects:
: :' : proud Debian developer http://debiansystem.info
`. `'` http://people.debian.org/~madduck http://vcs-pkg.org
`- Debian - when you have better things to do than fixing systems
 
Old 12-09-2008, 03:58 PM
Bjørn Mork
 
Default Bug#506419: Driver from Supermicro fixes issue

martin f krafft <madduck@debian.org> writes:
> also sprach Bjørn Mork <bjorn@mork.no> [2008.12.09.1413 +0100]:
>> This should not apply against linux-2.6 2.6.26-11 since it is already a
>> part of debian/patches/bugfix/all/stable/2.6.26.4.patch. Are you sure
>> your patch really does apply? If it does, then I'd suggest trying
>> *just* edcfe5f7e307846e578fb88d69fa27051fded0ab instead.
>
> My patch does no "apply", it was made against 2.6.26-11 --
> Supermicro just sent me their forcedeth.c and I made the patch.

Probably against the unpatched 2.6.26 then? OK, that will explain the
similarities. But then the real problem must be something else. Sorry
for the confusion.

>> BTW, I also noticed that linux-2.6 2.6.26-11 includes the patch
>> linux-2.6-2.6.26/debian/patches/features/all/net-use-gso.patch
>> which very well may trigger the problem. You might also want to
>> try building without that patch.
>
> I have not tried -11 on the server, only -8 and -10, which both had
> the problems.
>
> How do I find out when net-use-gso.patch was added? The changelog
> does not mention it, and I cannot find a repository for the source
> (there are also no Vcs-* links in the package).

http://packages.debian.org/changelogs/pool/main/l/linux-2.6/linux-2.6_2.6.26-11/changelog#versionversion2.6.26-2

has a

* net: use software GSO for SG+CSUM capable netdevices (Lennert Buytenhek)

line, which matches the subject in the patch.


Bjørn



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-09-2008, 04:47 PM
Martin Michlmayr
 
Default Bug#506419: Driver from Supermicro fixes issue

* Bastian Blank <waldi@debian.org> [2008-12-09 15:06]:
> > In fact, I'm a little confused as to the reasons that patch is in
> > Debian. It is probably the correct thing to do, but it is also the kind
> > of thing that is bound to cause regressions because it will expose
> > latent bugs in drivers.
>
> This is in mainline: e5a4a72d4f88f4389e9340d383ca67031d1b8536. I don't
> know what the rationale was to add this.

I added the patch because it leads to better performances with several
devices. There was one regression with loopback that got fixed
quickly and that patch is also in the Debian package.

I'm currently not aware of any regressions caused by the patch, but if
there are any, we of course have to remove the patch.
--
Martin Michlmayr
http://www.cyrius.com/



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-09-2008, 07:57 PM
martin f krafft
 
Default Bug#506419: Driver from Supermicro fixes issue

also sprach Bjørn Mork <bjorn@mork.no> [2008.12.09.1758 +0100]:
> > My patch does no "apply", it was made against 2.6.26-11 --
> > Supermicro just sent me their forcedeth.c and I made the patch.
>
> Probably against the unpatched 2.6.26 then? OK, that will explain the
> similarities. But then the real problem must be something else. Sorry
> for the confusion.

No, I made it against 2.6.26-11, for no other reason that it was in
Debian unstable at the time, and I had given up hope for targetting
lenny before. I'd be happy to create a patch against -10 instead.

> > How do I find out when net-use-gso.patch was added? The changelog
> > does not mention it, and I cannot find a repository for the source
> > (there are also no Vcs-* links in the package).
>
> http://packages.debian.org/changelogs/pool/main/l/linux-2.6/linux-2.6_2.6.26-11/changelog#versionversion2.6.26-2
>
> has a
>
> * net: use software GSO for SG+CSUM capable netdevices (Lennert Buytenhek)
>
> line, which matches the subject in the patch.

Alright. I admit I didn't look that closely. Isn't there a Git repo
for Debian kernels?

--
.'`. martin f. krafft <madduck@d.o> Related projects:
: :' : proud Debian developer http://debiansystem.info
`. `'` http://people.debian.org/~madduck http://vcs-pkg.org
`- Debian - when you have better things to do than fixing systems

"she was rather too intelligent and competent-looking to be
considered entirely beautiful, but all the more attractive because
of it."
-- george spencer-brown, "a lion's teeth"
 
Old 12-09-2008, 10:05 PM
Bastian Blank
 
Default Bug#506419: Driver from Supermicro fixes issue

On Tue, Dec 09, 2008 at 09:57:19PM +0100, martin f krafft wrote:
> also sprach Bjørn Mork <bjorn@mork.no> [2008.12.09.1758 +0100]:
> > > My patch does no "apply", it was made against 2.6.26-11 --
> > > Supermicro just sent me their forcedeth.c and I made the patch.
> > Probably against the unpatched 2.6.26 then? OK, that will explain the
> > similarities. But then the real problem must be something else. Sorry
> > for the confusion.
> No, I made it against 2.6.26-11, for no other reason that it was in
> Debian unstable at the time, and I had given up hope for targetting
> lenny before. I'd be happy to create a patch against -10 instead.

No, you did not. The lines in question read:

| dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
| dev->features |= NETIF_F_TSO;

> Alright. I admit I didn't look that closely. Isn't there a Git repo
> for Debian kernels?

Nope.

Bastian

--
Murder is contrary to the laws of man and God.
-- M-5 Computer, "The Ultimate Computer", stardate 4731.3



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 

Thread Tools




All times are GMT. The time now is 07:30 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright ©2007 - 2008, www.linux-archive.org