On 03/03/2012 04:56 AM, Nathan Hintz wrote:
> On Fri, 2012-03-02 at 21:13 +0100, Hauke Mehrtens wrote:
>> On 03/02/2012 06:01 AM, Nathan Hintz wrote:
>>> The SPROM is initialized in bcm47xx_fill_sprom, so move the call to
>>> bcm47xx_fill_sprom_ethernet after it.
>>>
>>> Signed-off-by: Nathan Hintz <nlhi...@hotmail.com>
>>>
>>> Index:
>>> target/linux/brcm47xx/patches-3.2/199-MIPS-BCM47XX-provide-sprom-to-bcma-bus.patch
>>> ===================================================================
>>> ---
>>> target/linux/brcm47xx/patches-3.2/199-MIPS-BCM47XX-provide-sprom-to-bcma-bus.patch
>>> (revision 30770)
>>> +++
>>> target/linux/brcm47xx/patches-3.2/199-MIPS-BCM47XX-provide-sprom-to-bcma-bus.patch
>>> (working copy)
>>> @@ -66,13 +66,13 @@
>>> + bcm47xx_fill_sprom(out, prefix);
>>> + return 0;
>>> + } else if (bus->hosttype == BCMA_HOSTTYPE_SOC) {
>>> -+ bcm47xx_fill_sprom_ethernet(out, NULL);
>>> + core = bcma_find_core(bus, BCMA_CORE_80211);
>>> + if (core) {
>>> + snprintf(prefix, sizeof(prefix), "sb/%u/",
>>> + core->core_index);
>>> + bcm47xx_fill_sprom(out, prefix);
>>> + }
>>> ++ bcm47xx_fill_sprom_ethernet(out, NULL);
>>> + return 0;
>>> + } else {
>>> + printk(KERN_WARNING "bcm47xx: unable to fill SPROM for given
>>> bustype.\n");
>>
>> Why do you want to move this call? If there is a mac address stored with
>> the prefix (sb/1/) for this core we should take it in favor of the
>> general mac address without a prefix. In my device both addresses are
>> the same so it would not make a difference, but I haven't looked into
>> all nvrams.
>>
>> Hauke
>>
> Hi Hauke,
>
> I'm sorry for not being clear. The problem I was trying to address was
> that the first thing "bcm47xx_fill_sprom" does is to zero fill the
> entire sprom, effectively wiping out anything that might have been
> filled in by the previous call to "bcm47xx_fill_sprom_ethernet". For
> Linksys E3000, et0macaddr and et0phyaddr don't have a prefix in NVRAM,
> and therefore ended up not being set. I've attached an NVRAM dump. How
> would you like to proceed?
>
> Thanks,
>
> Nathan
Hi Nathan,
Thanks for the clarification. I would more like the memset being moved
to the functions calling bcm47xx_fill_sprom() then change the order of
the functions being called. I do not think it is even needed in most
cases, as struct ssb_sprom should be already been set to zero.
Hauke
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
