Ebony's UBoot awareness
PATCH 1/1 Ebony's UBoot awareness
Hi,
This adds UBoot support on Ebony board and also sets emac's PHY mode explicitly during initialization as Eugene Surovegin provided.
Regards,
-Shawn.
linux-2.6.10/arch/ppc/platforms/4xx/ebony.c +54 -14
@@ -49,10 +49,14 @@#include <asm/todc.h> #include <asm/bootinfo.h> #include <asm/ppc4xx_pic.h> +#include <asm/ppcboot.h> #include <syslib/gen550.h> +#include "../../../../drivers/net/ibm_emac/ibm_emac_phy.h" +static struct ibm44x_clocks clocks__initdata ; +bd_t __res;/* * Ebony IRQ triggering/polarity settings
@@ -182,6 +186,41 @@ (writel (value, (u32 )pcix_reg_base+offset))/* +* Set mac address for each EMAC +* If uboot sets MAC addresses for each EMAC, use them. +* Otherwise fetch from Ebony VPD. +* + */ +static void __init +ebony_set_emacdata (void ) +{ +unsigned char * vpd_base; +struct ocp_def *def; +struct ocp_func_emac_data *emacdata; + + vpd_base =ioremap64 (EBONY_VPD_BASE ,EBONY_VPD_SIZE ); + + def =ocp_get_one_device (OCP_VENDOR_IBM ,OCP_FUNC_EMAC ,0 ); + emacdata = def->additions; + emacdata->phy_mode =PHY_MODE_RMII ; +if (!strncmp (__res.bi_enetaddr,"\0xFF\0xFF\0xFF\0xFF\0xFF\0xFF" ,6 )) +memcpy (emacdata->mac_addr, __res.bi_enetaddr,6 ); +else +memcpy (emacdata->mac_addr,EBONY_NA0_ADDR (vpd_base),6 ); + + def =ocp_get_one_device (OCP_VENDOR_IBM ,OCP_FUNC_EMAC ,1 ); + emacdata = def->additions; + emacdata->phy_mode =PHY_MODE_RMII ; +if (!strncmp (__res.bi_enet1addr,"\0xFF\0xFF\0xFF\0xFF\0xFF\0xFF" ,6 )) +memcpy (emacdata->mac_addr, __res.bi_enet1addr,6 ); +else +memcpy (emacdata->mac_addr,EBONY_NA1_ADDR (vpd_base),6 ); + +iounmap (vpd_base); + +} + +/* * FIXME: This is only here to "make it work". This will move * to a ibm_pcix.c which will contain a generic IBM PCIX bridge * configuration library. -Matt
@@ -309,19 +348,7 @@static void __init ebony_setup_arch (void ) { -unsigned char * vpd_base; -struct ocp_def *def; -struct ocp_func_emac_data *emacdata; - -/* Set mac_addr for each EMAC */ - vpd_base =ioremap64 (EBONY_VPD_BASE ,EBONY_VPD_SIZE ); - def =ocp_get_one_device (OCP_VENDOR_IBM ,OCP_FUNC_EMAC ,0 ); - emacdata = def->additions; -memcpy (emacdata->mac_addr,EBONY_NA0_ADDR (vpd_base),6 ); - def =ocp_get_one_device (OCP_VENDOR_IBM ,OCP_FUNC_EMAC ,1 ); - emacdata = def->additions; -memcpy (emacdata->mac_addr,EBONY_NA1_ADDR (vpd_base),6 ); -iounmap (vpd_base); +ebony_set_emacdata ();/* * Determine various clocks.
@@ -368,7 +395,20 @@void __init platform_init (unsigned long r3,unsigned long r4,unsigned long r5,unsigned long r6,unsigned long r7) { -parse_bootinfo ((struct bi_record *) (r3 +KERNELBASE )); +parse_bootinfo (find_bootinfo ()); + +/* +* If we were passed in a board information, copy it into the +* residual data area. +*/ +if (r3) + __res = *(bd_t *)(r3 +KERNELBASE ); + +/* Copy the kernel command line arguments to a safe place. */ +if (r6) { + *(char *) (r7 +KERNELBASE ) =0 ; +strcpy (cmd_line, (char *) (r6 +KERNELBASE )); + }ibm44x_platform_init ();
Quoted 2 lines · Shawn Jin
This adds UBoot support on Ebony board and also sets emac's PHY mode explicitly during initialization as Eugene Surovegin provided.
[snip]
Quoted 1 line
+#include "
../../../../drivers/net/ibm_emac/ibm_emac_phy.h"
This will not work with O= for example. The correct method is to modify the Makefile with a CFLAGS_foo.o += -Idrivers/net/ibm_emac
Signature · 2 lines
Tom Rini http://gate.crashing.org/~trini/
Quoted 4 lines · Tom Rini
Quoted 1 line
+#include "
../../../../drivers/net/ibm_emac/ibm_emac_phy.h"This will not work with O= for example. The correct method is to modify the
Makefilewith a CFLAGS_foo.o += -Idrivers/net/ibm_emac
I just tested and it does work even with O= :).
OTOH, I agree this is ugly hack (it was copied from ocotea.c, so blame Matt for it :) and should be fixed correctly by moving those PHY defines to some common include.
--
Eugene
Quoted 11 lines · Eugene Surovegin
Quoted 4 lines · Tom Rini
Quoted 1 line
+#include "
../../../../drivers/net/ibm_emac/ibm_emac_phy.h"This will not work with O= for example. The correct method is to modify the
Makefilewith a CFLAGS_foo.o += -Idrivers/net/ibm_emacI just tested and it does work even with O= :).
OTOH, I agree this is ugly hack (it was copied from
ocotea.c, so blame Matt for it :) and should be fixed correctly by moving those PHY defines to some common include.
Ugh, yes...in my defense I clearly stated it was an "ugly hack". :)
-Matt
Quoted 4 lines · Shawn Jin
Hi,
This adds UBoot support on Ebony board and also sets emac's PHY mode explicitly during initialization as Eugene Surovegin provided.
<snip>
linux-2.6.10/arch/ppc/platforms/4xx/ebony.c line 209
Quoted 4 lines
linux-2.6.10/arch/ppc/platforms/4xx/ebony.c +4
-0+if (!strncmp (__res.bi_enetaddr,"\0xFF\0xFF\0xFF\0xFF\0xFF\0xFF" ,6 )) +memcpy (emacdata->mac_addr, __res.bi_enetaddr,6 ); +else +memcpy (emacdata->mac_addr,EBONY_NA0_ADDR (vpd_base),6 );
Can you use the ocotea/luan approach and create an openbios stub in arch/ppc/boot to create the bi_enetaddr infos? The PIBS stub in arch/ppc/boot does this so we can have one path in the platform file that parses the bootinfos.
-Matt
Hi Matt,
Quoted 4 lines
Can you use the ocotea/luan approach and create an openbios stub in
arch/ppc/bootto create thebi_enetaddrinfos? The PIBS stub inarch/ppc/bootdoes this so we can have one path in the platform file that parses the bootinfos.
I'd like to help if I can follow you. What does PIBS stand for? What is the ocotea/luan approach you were talking about? I thought I followed the ocotea approach to set up mac addresses.
Thanks,
-Shawn.
Quoted 10 lines · Shawn Jin
Hi Matt,
Quoted 4 lines
Can you use the ocotea/luan approach and create an openbios stub in
arch/ppc/bootto create thebi_enetaddrinfos? The PIBS stub inarch/ppc/bootdoes this so we can have one path in the platform file that parses the bootinfos.I'd like to help if I can follow you. What does PIBS stand for? What is the ocotea/luan approach you were talking about? I thought I followed the ocotea approach to set up mac addresses.
PIBS is the firmware on all post 440GP ref brds from IBM/AMCC. Take a look at how both the stock PIBS f/w and U-Boot are supported on Ocotea by reviewing arch/ppc/boot/simple/pibs.c and arch/ppc/platforms/4xx/ocotea.c
In this implementation, ocotea.c always expects the enetaddrs in a birec. Whether they come direct from U-Boot or from some massaging by the arch/ppc/boot/simple/pibs.c shim doesn't matter. I suggest following that model which requires creating something like arch/ppc/boot/simple/openbios.c with a strong symboled load_kernel() routine that builds the bi-rec for the stock firmware case.
This hides the ugliness in the boot wrapper glue.
-Matt
Quoted 10 lines · Shawn Jin
Hi Matt,
Quoted 4 lines
Can you use the ocotea/luan approach and create an openbios stub in
arch/ppc/bootto create thebi_enetaddrinfos? The PIBS stub inarch/ppc/bootdoes this so we can have one path in the platform file that parses the bootinfos.I'd like to help if I can follow you. What does PIBS stand for? What is the ocotea/luan approach you were talking about? I thought I followed the ocotea approach to set up mac addresses.
FYI, the implementation by Gerhard Jaeger for OpenBIOS/U-Boot Ebony support was sent upstream.
-Matt