Re: Fixes for kern/40018 -- any chance of getting these pulled into the -current and 5.x trees?

[buhrow%lothlorien.nfbcal.org@localhost (Brian Buhrow)]

        Hello Jonathan.  In researching this patch, I read the FreeBSD and
Linux drivers extensively to see what they did. As a final checked, I just
looked at the OpenBSD driver.  This patch now brings our driver into
conformance with all three other drivers with respect to which mode it
drives the chips in.
        With respect to the 5700 chip, it looks like we already consider the
Rev B version of that chip broken and don't even enable hardware
checksumming on it, which matches the behavior of the other drivers.  I
believe I understand the reasons you chose to do things the way you did,
and I believe, at the time, it was a reasonable choice.  Now, however, I
believe no other drivers are taking advantage of the chip's ability to not
calculate the pseudo header checksum, and, worse, the chips now randomly
switch modes while runing.  Rather than trying to figure out which chips
change modes and when they do it, it seemed easier to me to let the chips
run in full checksum calculation mode, and tell our stack of the
change.  And, since this is what all the other drivers I've looked at do,
and none of them have notes about having to worry about different revisions
of chips handling checksums on incoming packets differently, except for
those chips which are completely broken and which we already account for, I
feel pretty certain that this patch is correct and will put a lot of our
bge(4) problems, at least those related to checksum calculations, to bed.

-Brian
On Sep 21, 10:29am, Jonathan Stone wrote:
} Subject: Re: Fixes for kern/40018 -- any chance of getting these pulled in
} --0-272765047-1285090164=:24001
} Content-Type: text/plain; charset=iso-8859-1
} Content-Transfer-Encoding: quoted-printable
} 
} Brian,
} 
} I suspect I'm the original author of that code.=A0=A0=A0 My recollection is=
}  that the patch you suggest will cause the same symptoms, when used with a =
} BCM5700.
} Can you find a volunteer to check with a bcm5700 before getting these pulle=
} d up?
} 
} If I can find a copy of my driver-source tree, I'll check to see what i did=
}  to handle the bcm5721. I still have a couple of those to check against.
} 
} cheers,
} --Jonathan
} 
} --- On Tue, 9/21/10, Brian Buhrow <buhrow%lothlorien.nfbcal.org@localhost> 
wrote:
} 
} From: Brian Buhrow <buhrow%lothlorien.nfbcal.org@localhost>
} Subject: Fixes for kern/40018 -- any chance of getting these pulled into th=
} e -current and 5.x trees?
} To: tech-kern%netbsd.org@localhost
} Cc: buhrow%lothlorien.nfbcal.org@localhost
} Date: Tuesday, September 21, 2010, 10:05 AM
} 
} =A0=A0=A0 Hello.=A0 I submitted a couple of patches for kern/40018, hardwar=
} e
} checksum failures on bge(4) controlled chips, which fix the problem.=A0 The=
} se
} patches have been running on a number of high volume production servers for
} a week with babsolutely no problems and all hardware capabilities=A0 enable=
} d.
} You can look=A0 at the e-mail thread on current-users at:
} 
} http://mail-index.NetBSD.org/current-users/2010/09/13/msg014299.html
} 
} =A0=A0=A0 I've included the patches, and some of the thread in the bug repo=
} rt.
} 
} =A0=A0=A0 If someone could pull these patches into the -current tree and re=
} quest
} a pullup into the 5.x branches, that would be wondderful.=A0 The patch is
} small, and I've included versions for the 5.x tree as well as -current.
} 
} -thanks
} -Brian
} =0A=0A=0A      
} --0-272765047-1285090164=:24001
} Content-Type: text/html; charset=iso-8859-1
} Content-Transfer-Encoding: quoted-printable
} 
} <table cellspacing=3D"0" cellpadding=3D"0" border=3D"0" ><tr><td valign=3D"=
} top" style=3D"font: inherit;">Brian,<br><br>I suspect I'm the original auth=
} or of that code.&nbsp;&nbsp;&nbsp; My recollection is that the patch you su=
} ggest will cause the same symptoms, when used with a BCM5700.<br>Can you fi=
} nd a volunteer to check with a bcm5700 before getting these pulled up?<br><=
} br>If I can find a copy of my driver-source tree, I'll check to see what i =
} did to handle the bcm5721. I still have a couple of those to check against.=
} <br><br>cheers,<br>--Jonathan<br><br>--- On <b>Tue, 9/21/10, Brian Buhrow <=
} i>&lt;buhrow%lothlorien.nfbcal.org@localhost&gt;</i></b> 
wrote:<br><blockquote style=
} =3D"border-left: 2px solid rgb(16, 16, 255); margin-left: 5px; padding-left=
} : 5px;"><br>From: Brian Buhrow 
&lt;buhrow%lothlorien.nfbcal.org@localhost&gt;<br>Subj=
} ect: Fixes for kern/40018 -- any chance of getting these pulled into the -c=
} urrent and 5.x trees?<br>To: tech-kern%netbsd.org@localhost<br>Cc:
}  buhrow%lothlorien.nfbcal.org@localhost<br>Date: Tuesday, September 21, 2010, 
10:05 A=
} M<br><br><div class=3D"plainMail">&nbsp;&nbsp;&nbsp; Hello.&nbsp; I submitt=
} ed a couple of patches for kern/40018, hardware<br>checksum failures on bge=
} (4) controlled chips, which fix the problem.&nbsp; These<br>patches have be=
} en running on a number of high volume production servers for<br>a week with=
}  babsolutely no problems and all hardware capabilities&nbsp; enabled.<br>Yo=
} u can look&nbsp; at the e-mail thread on current-users at:<br><br><a href=
} =3D"http://mail-index.NetBSD.org/current-users/2010/09/13/msg014299.html"; t=
} arget=3D"_blank">http://mail-index.NetBSD.org/current-users/2010/09/13/msg0=
} 14299.html</a><br><br>&nbsp;&nbsp;&nbsp; I've included the patches, and som=
} e of the thread in the bug report.<br><br>&nbsp;&nbsp;&nbsp; If someone cou=
} ld pull these patches into the -current tree and request<br>a pullup into t=
} he 5.x branches, that would be wondderful.&nbsp; The patch is<br>small, and
}  I've included versions for the 5.x tree as well as -current.<br><br>-thank=
} s<br>-Brian<br></div></blockquote></td></tr></table><br>=0A=0A=0A=0A=0A=0A=
} =0A=0A      
} --0-272765047-1285090164=:24001--
>-- End of excerpt from Jonathan Stone