Adrian, Thank you for your reply. I agree w/ you that the modification needs to add (as opposed to replace) the BIOS recognition capability of the original code. My modification was more of a very simple test than anything else. I was really trying to understand what components were necessary for the /dev/led/error devices to be created.
Subsequent to this test, I noticed that even when using the CPU_GEODE and CPU_SOEKRIS directives to compile a kernel w/ the original geode.c source, I did NOT see the /dev/led devices when booting on a net 4801 (I have both 4801 & 5501 for my testing). No doubt, I'm doing something wrong! Are there any other prerequisites or compilation options required for the led device functionality? Any other ideas on what I might be doing wrong? FYI, I'm building a FreeBSD 6.2 kernel on a deployment server using the method described by David Courtney at ultradesic.com. -----Original Message----- From: Adrian Steinmann [mailto:[EMAIL PROTECTED] Sent: Wednesday, August 29, 2007 3:35 AM To: Tom Mazzotta Cc: Poul-Henning Kamp; [email protected] Subject: Re: [Soekris] Controlling Error LED's on 5501 > In message <[EMAIL PROTECTED]>, Tom Mazzotta writes: > >I would like to use the led command to control the red error led > >>on the 5501. I have compiled my kernel w/ the "options CPU_GEODE" > >and "options CPU_SOEKRIS" directives. However, the system does not > >recognize a /dev/led/error device. What else needs to be done for > >the system to recognize this device? to which phk@ responded: > The bios recognition string in src/sys/i386/i386/geode.c needs updated > to find the 55xx series also. then, on Sat, Aug 25, 2007 at 04:50:30PM -0400, Tom Mazzotta responded: > $ diff geode.c geode.c.original > 46c46 > < { "net5", 0, 8 }, /* net5xxx */ > --- > > { "net4", 0, 8 }, /* net45xx */ > You should *add* and not replace new strings to the list of oem signatures (most specific first). The original patch had a documentation of the central routine used to do this BIOS string groping. You can read it in the i386/doc PR I just filed at http://www.freebsd.org/cgi/query-pr.cgi?pr=115918 a side effect of bios_oem_strings(9) is to print out human readable dmesg of the specific BIOS strings right before the platform-specific initializations are done. You can see examples of it for Advantech, PCEngines, and Soekris in FreeBSD sys/i386/i386/ directory. Adrian _______________________________________________ Soekris-tech mailing list [email protected] http://lists.soekris.com/mailman/listinfo/soekris-tech
