IRC logs for #openrisc Tuesday, 2013-12-24

--- Log opened Tue Dec 24 00:00:22 2013
maxpalnstekern: looks like you're onto the right08:34
maxpalnIt wasn't straight forward to disable bursts in the DDR controller so I disabled the cache and it now works08:34
maxpalnhowever, at the console output I am getting these messages which indicate something is wrong at the Ethernet Rx side08:35
maxpalnnet eth0: RX: wrong CRC08:36
maxpalnnet eth0: RX: frame too short08:36
maxpalnnet eth0: RX: dribble nibble08:36
maxpalnhmmm, reading around on the web - not sure if this is a Linux thing or a HW thing. Any thoughs?08:44
stekernhmm, haven't seen that before09:30
maxpalnA ping yields no Rx packets09:30
maxpalnand it's difficult to say whether and are being Tx09:31
maxpalnI think this must be a GMII vs MII thing09:31
stekernbut it *has* to be straight forward to disable bursts in the controller, since that's essentially what you did when you disabled bursts in the master09:31
maxpalnI would agree - but the controller hasn't been written in a way that I can find an easy method to disable the burst09:32
maxpalnit's a kinda crappy implementation if I am being honest09:32
maxpalnit assumes an incremental burst -09:32
maxpalnand waits for the controller to deassert STD to stop the burst09:32
stekernhmm, I see09:32
stekernthat's not really wb compliant behaviour09:33
maxpalnI think a rewrite of the controller (with the Xilinx DDR2 controller as a template) is probably the best method09:33
maxpalnyeah - I would agree with you09:33
maxpalnbut, back on the ethernet09:33
maxpalnthe board has a GMII phy09:33
maxpalnI think the problem must be to do with how the MAC is setting up the PHY09:34
maxpalnThe router is showing the connection as a Gbps link09:34
maxpalnwhich is almost certainly the problem09:34
maxpalnthe MII ethmac only uses four of the eight data bits09:35
maxpalnfor Tx and Rx09:35
maxpalnI need to get the MAC to put the PHy into 10/100 mode09:35
maxpalnprobably need to do a bit of research there09:35
maxpalnalthough, interestingly ifconfig shows a valid IP address09:42
maxpaln# ifconfig09:42
maxpalneth0      Link encap:Ethernet  HWaddr 16:D3:2E:BA:D6:BC09:42
maxpaln          inet addr:  Bcast:  Mask:
maxpaln          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:109:42
maxpaln          RX packets:0 errors:0 dropped:345 overruns:0 frame:76009:42
maxpaln          TX packets:1758 errors:1 dropped:0 overruns:0 carrier:009:42
maxpaln          collisions:0 txqueuelen:100009:42
maxpaln          RX bytes:0 (0.0 B)  TX bytes:73836 (72.1 KiB)09:42
maxpaln          Interrupt:4 Memory:1200000-122000009:42
maxpalnlo        Link encap:Local Loopback09:42
maxpaln          inet addr:  Mask:
maxpaln          UP LOOPBACK RUNNING  MTU:65536  Metric:109:42
maxpalnalthough a LOT of dropped Rx kets09:43
stekernisn't that a static ip though?09:49
stekernbut, yes, if the phy is set to autosense and the router too, then it'll probably autosense to gigabit09:50
stekernwhat phy is it?09:50
maxpalnits a marvell PHY09:54
maxpalnah, you might be right about the static IP09:54
maxpalnit fits in well with my network09:54
maxpalnthe Bcast address is wrontg09:54
maxpalnthe Bcast address is wrong09:54
maxpalnoh, hang on09:55
maxpalnnope it's ok09:55
stekernok, that's the same that's on the atlys board09:55
maxpalnah, ok09:56
stekernbut I only have 100Mbit switches09:56
maxpalnwell, its definitely a static IP - I disconnected from the router and the IP adress remains09:57
maxpalnwell, I've asked our ethernet team to let me know how the PHY should be configured to force it to 10/100 mode09:58
maxpalnI don't have the datasheet here09:59
stekerniirc, it's in the "standard" regs10:08
stekernbut I might remember wrong10:08
stekernyes, it's in reg 0: basic control10:09
maxpalnwhere are you looking?10:11
stekernin any random phy datasheet10:12
maxpalnah, ok10:12
stekernfor instance10:12
maxpalnSO I assume this needs to be implemented in the Linux build somewhere - perhaps in one of the drivers10:13
stekernbut I'm not sure how that is handled on gbit phys, never poked around those in-depth10:13
maxpalnor perhaps even while Linux is running - should be possible to reconfigure the PHY I would have thought10:16
stekernthere's ethtool and mii-tool to access mdio under linux10:16
_franck_web_same thing on a 88E1512 gigabit phy10:16
_franck_web_reg 010:16
stekernbut you can probably hack something into the ethmac driver too10:17
maxpalnin the driver would be better long term10:23
maxpalnbut it would be good to be able to modify in the running Linux to see what settings are needed10:23
maxpalnhmmm, mii-tool and ethtool are not i the vmlinux binary10:25
maxpalnI'd download it and install it but ....10:25
maxpalnok, so I think my lack of Linux admin knowledge is exposing me here [blushes]10:27
maxpalnI'm running out of time today so I think I will have to return to this after the festive period10:28
maxpalnperhaps I'll recruit a Linux expert from our apps team :-)10:28
maxpalnthank you for all your help so far -10:28
maxpalnyou have been very helpful and definitely allowing me to progress a lot quicker than I could have otherwise10:29
maxpalnI'm wondering if drivers for the Marvell phy have actually been included10:32
maxpalnI can't see one in the drivers/net/ethernet folder10:33
stekernwell, you don't need any special phy drivers to change the regs10:34
maxpalnand if I run 'make menuconfig' to run the Linux configuration GUI thing (not really sure what to call it - it's kinda like a GUI) I can see that Marvell PHY drivers are not set for inclusion - just Micrel PHYs10:34
maxpalnright - time to dash10:37
maxpalnI'll pick this up again in a few days10:37
stekernsee you10:38
-!- Netsplit *.net <-> *.split quits: Amadiro, jeremybennett, FreezingCold14:06
-!- Netsplit over, joins: FreezingCold, Amadiro14:11
--- Log closed Wed Dec 25 00:00:24 2013

Generated by 2.15.2 by Marius Gedminas - find it at!