Patchwork 2.6.28-git8: tg3 doesn't work due to firmware not loading (-git7 is ok)

login
register
mail settings
Submitter Valdis.Kletnieks@vt.edu
Date Jan. 11, 2009, 10:29 p.m.
Message ID <99827.1231712979@turing-police.cc.vt.edu>
Download mbox | patch
Permalink /patch/17857/
State RFC
Delegated to: David Miller
Headers show

Comments

Valdis.Kletnieks@vt.edu - Jan. 11, 2009, 10:29 p.m.
On Sun, 11 Jan 2009 13:49:37 PST, David Miller said:
> From: Valdis.Kletnieks@vt.edu
> Date: Sun, 11 Jan 2009 11:48:22 -0500
> 
> > One unanswered question:  What do we expect the system to do if they have this
> > patch, TIGON3=y, FIRMWARE_IN_KERNEL=n, and configure a netconsole for boot
> > messages?  I'm *hoping* the answer is "the netconsole doesn't come up at boot,
> > but can be re-enabled via the /sys/kernel/config/netconsole interface after
> > you've done an 'ifconfig eth0 up' or similar, or do a 'modprobe netconsole'.
> > 
> > Those seem like reasonable semantics to me - anybody got a different opinion?
> 
> Even better, how about nfsroot?  There is no "later", either you
> can mount to root filesystem or you fail.

I don't see any sane way for an nfsroot to work unless you've built the kernel
with TIGON3=y, and FIRMWARE=y.  Anything else is just crazy talk.  For that
case, I'd suggest we just do this:

Signed-off-by: Valdis Kletnieks <valdis.kletnieks@vt.edu>
David Woodhouse - Jan. 11, 2009, 10:46 p.m.
On Sun, 2009-01-11 at 17:29 -0500, Valdis.Kletnieks@vt.edu wrote:
> On Sun, 11 Jan 2009 13:49:37 PST, David Miller said:
> > From: Valdis.Kletnieks@vt.edu
> > Date: Sun, 11 Jan 2009 11:48:22 -0500
> > 
> > > One unanswered question:  What do we expect the system to do if they have this
> > > patch, TIGON3=y, FIRMWARE_IN_KERNEL=n, and configure a netconsole for boot
> > > messages?  I'm *hoping* the answer is "the netconsole doesn't come up at boot,
> > > but can be re-enabled via the /sys/kernel/config/netconsole interface after
> > > you've done an 'ifconfig eth0 up' or similar, or do a 'modprobe netconsole'.
> > > 
> > > Those seem like reasonable semantics to me - anybody got a different opinion?
> > 
> > Even better, how about nfsroot?  There is no "later", either you
> > can mount to root filesystem or you fail.
> 
> I don't see any sane way for an nfsroot to work unless you've built the kernel
> with TIGON3=y, and FIRMWARE=y.  Anything else is just crazy talk.  For that
> case, I'd suggest we just do this:
> 
> Signed-off-by: Valdis Kletnieks <valdis.kletnieks@vt.edu>
> 
> --- linux-2.6.28-mmotm0109/Documentation/filesystems/nfsroot.txt.dist	2008-12-24 18:26:37.000000000 -0500
> +++ linux-2.6.28-mmotm0109/Documentation/filesystems/nfsroot.txt	2009-01-11 17:27:31.000000000 -0500
> @@ -30,6 +30,9 @@ In the networking options, kernel level 
>  along with the types of autoconfiguration to support. Selecting all of
>  DHCP, BOOTP and RARP is safe.
>  
> +In addition, your network interface driver must be selected as built-in,
> +and if the card is a Tigon3 or other card that requires a firmware load
> +to become functional, you need to select FIRMWARE_IN_KERNEL.

Yeah, but it's not FIRMWARE_IN_KERNEL in the general case -- that's just
for a handful of 'speshul' legacy drivers. In the general case it's the
CONFIG_EXTRA_FIRMWARE option which needs to be set to include the
required firmware.

Patch

--- linux-2.6.28-mmotm0109/Documentation/filesystems/nfsroot.txt.dist	2008-12-24 18:26:37.000000000 -0500
+++ linux-2.6.28-mmotm0109/Documentation/filesystems/nfsroot.txt	2009-01-11 17:27:31.000000000 -0500
@@ -30,6 +30,9 @@  In the networking options, kernel level 
 along with the types of autoconfiguration to support. Selecting all of
 DHCP, BOOTP and RARP is safe.
 
+In addition, your network interface driver must be selected as built-in,
+and if the card is a Tigon3 or other card that requires a firmware load
+to become functional, you need to select FIRMWARE_IN_KERNEL.