2013/9/12 Hauke Mehrtens <[email protected]>:
> On 09/12/2013 01:23 PM, Rafał Miłecki wrote:
>> Network on BCM5357* devices isn't working since ever. It seems there
>> is some problem with bgmac I don't understand.
>>
>> First of all, when using interface without VLAN enabled and without
>> bridging it - no packets appear. That can be easily workarounded with
>> "ifconfig eth0 promisc". Then I can see packets appearing, however
>> they don't have a valid MAC addresses in src/dest fields.
>>
>> I decided to dump packets in bgmac driver, that allow me to include
>> Broadcom's extra RX header.
>>
>> Does it give any hint for anyone? Do you have any idea what can be wrong?
>>
>
> Hi,
>
> the switch in the BCM5357 needs a special configuration. Does your SoC
> have the chipid 0x5356? Then the robo core rev is 3.

Not exactly a 0x5356

E900:
[ 0.000000] bcma: bus0: Found chip with id 0xD144, rev 0x01 and package 0x08
[ 4.060000] b53_common: found switch: BCM5325, rev 0
[ 4.064000] (unregistered net_device): attached PHY driver [Broadcom
B53 (2)] (mii_bus:phy_addr=bgmac-0-0:1e)

No-name board:
[ 0.000000] bcma: bus0: Found chip with id 0x5357, rev 0x01 and package 0x08
[ 4.608000] b53_common: found switch: BCM5325, rev 0
[ 4.616000] (unregistered net_device): attached PHY driver [Broadcom
B53 (2)] (mii_bus:phy_addr=bgmac-0-0:1e)

I forgot you mention one important fact.
As you already noticed, enabling switch doesn't add VLAN tags to the
packets. Another story is that with the default b53 I don't get any RX
interrupts (and packets). I has to hack b53 to don't set
VC5_DROP_VTABLE_MISS.

Hauke: I'll try your patch, but it seems to be for some higher BCM5325
rev, isn't it?
Secondly, you're patching b53. As I said, even without switch enabled
I get corrupted packets too :(

-- 
Rafał
_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel

Reply via email to