severe regression in alx ethernet driver

There is a serious regression in the alx ethernet driver.  The driver
stopped working after upgrading the kernel from 4.0.x to 4.1.x.
Please see https://bugzilla.redhat.com/show_bug.cgi?id=1251434 and
https://bugzilla.kernel.org/show_bug.cgi?id=70761 This regression is
urgent, as I cannot update my kernel to include the latest security
fixes.

The severe regression in alx ethernet driver is still present in the
latest released kernel.  Patches have been posted in
https://bugzilla.kernel.org/show_bug.cgi?id=70761 There are reports
that these patches have resolved the problem.  Could you please review
these patches and see if they can be included upstream?  I have not
been able to update my kernel in almost six months.  Many other users
are suffering from this same problem.

On Mon, Sep 14, 2015 at 6:47 PM, Ldap Tester [off-list ref] wrote:

There is a serious regression in the alx ethernet driver.  The driver
stopped working after upgrading the kernel from 4.0.x to 4.1.x.
Please see https://bugzilla.redhat.com/show_bug.cgi?id=1251434 and
https://bugzilla.kernel.org/show_bug.cgi?id=70761 This regression is
urgent, as I cannot update my kernel to include the latest security
fixes.

From: Ldap Tester <redacted>
Date: Mon, 23 Nov 2015 11:00:08 -0300

https://bugzilla.kernel.org/show_bug.cgi?id=70761

Patches sitting in a bugzilla entry will simply rot.

For upstream inclusion someone must formally submit them to the netdev
mailing list, with proper commit message, signoffs, etc. and then the
patch must be fully reviewed by the community.

Only then will it be applied and make it upstream.

From: David Miller <davem@davemloft.net>
Date: Mon, 23 Nov 2015 11:15:30 -0500 (EST)

From: Ldap Tester <redacted>
Date: Mon, 23 Nov 2015 11:00:08 -0300

quoted

https://bugzilla.kernel.org/show_bug.cgi?id=70761

Patches sitting in a bugzilla entry will simply rot.

For upstream inclusion someone must formally submit them to the netdev
mailing list, with proper commit message, signoffs, etc. and then the
patch must be fully reviewed by the community.

Only then will it be applied and make it upstream.

And amusingly, the final comment in that bugzilla is someone asking
the patch author to do exactly this.

On Mon, Nov 23, 2015 at 11:16:10AM -0500, David Miller wrote:

From: David Miller <davem@davemloft.net>
Date: Mon, 23 Nov 2015 11:15:30 -0500 (EST)

quoted

From: Ldap Tester <redacted>
Date: Mon, 23 Nov 2015 11:00:08 -0300

quoted

https://bugzilla.kernel.org/show_bug.cgi?id=70761

Patches sitting in a bugzilla entry will simply rot.

For upstream inclusion someone must formally submit them to the netdev
mailing list, with proper commit message, signoffs, etc. and then the
patch must be fully reviewed by the community.

Only then will it be applied and make it upstream.

And amusingly, the final comment in that bugzilla is someone asking
the patch author to do exactly this.

They do at least have a signed-off-by in the patches attached to the bug,
so I'm working on touching up the descriptions and formatting, regression
testing them on my laptop that has an alx-driven E2200 in it (which isn't
affected by this bug), and then I can ship them up for proper review.

-- 
Jarod Wilson
jarod@redhat.com

Jarod Wilson [off-list ref] :
[...]

They do at least have a signed-off-by in the patches attached to the bug,
so I'm working on touching up the descriptions and formatting, regression
testing them on my laptop that has an alx-driven E2200 in it (which isn't
affected by this bug), and then I can ship them up for proper review.

You can forget the last refill-me-harder patch and save a roundtrip.
The driver should 1) prohibit holes in its receive ring, 2) allocate before
pushing data up in the stack and 3) drop packets when it can't allocate a
fresh buffer.

-- 
Ueimor

From: Francois Romieu <romieu@fr.zoreil.com>
Date: Wed, 25 Nov 2015 23:47:08 +0100

Jarod Wilson [off-list ref] :
[...]

quoted

They do at least have a signed-off-by in the patches attached to the bug,
so I'm working on touching up the descriptions and formatting, regression
testing them on my laptop that has an alx-driven E2200 in it (which isn't
affected by this bug), and then I can ship them up for proper review.

You can forget the last refill-me-harder patch and save a roundtrip.
The driver should 1) prohibit holes in its receive ring, 2) allocate before
pushing data up in the stack and 3) drop packets when it can't allocate a
fresh buffer.

+1