On Wed, Nov 12, 2014 at 12:22:22PM +0100, Sebastian Hesselbarth wrote: > On 11/12/2014 11:56 AM, Uwe Kleine-König wrote: > >Hello again, > > > >here come the recent insights. > [...] > > > >It seems to be not possible to easily dump the register space in both > >U-Boot and barebox for comparison. md 0xf1074000+0x4000 just hangs > >somewhere in the middle. > > > >A difference between U-Boot and barebox is the location where the > >internal registers are mapped. Maybe something that depends on U-Boot's > >memory layout leaks into barebox because I do 2nd stage booting? > > > >Out of ideas at the moment. :-( > > Uwe, > > Nice comparison, but did you double check caches are disabled? There is > no support for Dcache on mvebu SoCs in barebox atm. I would expect that U-Boot disables caches on go. But I remember there was a bug in that area some time ago. Now I saw a different behaviour: Marvell>> dhcp BOOTP broadcast 1 *** Unhandled DHCP Option in OFFER/ACK: 28 *** Unhandled DHCP Option in OFFER/ACK: 28 DHCP client bound to address 192.168.77.133 Marvell>> tftp barebox-netgear-rn104-2nd.img Using egiga1 device TFTP from server 192.168.77.157; our IP address is 192.168.77.133 Filename 'barebox-netgear-rn104-2nd.img'. Load address: 0x2000000 Loading: #################### done Bytes transferred = 292299 (475cb hex) Marvell>> md 0x01f15000 That is where the descriptor ... 01f15000: ffffffff ffffffff ffffffff ffffffff ................ 01f15010: ffffffff ffffffff ffffffff ffffffff ................ 01f15020: ffffffff ffffffff ffffffff ffffffff ................ 01f15030: ffffffff ffffffff ffffffff ffffffff ................ 01f15040: ffffffff ffffffff ffffffff ffffffff ................ 01f15050: ffffffff ffffffff ffffffff ffffffff ................ 01f15060: ffffffff ffffffff ffffffff ffffffff ................ 01f15070: ffffffff ffffffff ffffffff ffffffff ................ 01f15080: ffffffff ffffffff ffffffff ffffffff ................ 01f15090: ffffffff ffffffff ffffffff ffffffff ................ 01f150a0: ffffffff ffffffff ffffffff ffffffff ................ 01f150b0: ffffffff ffffffff ffffffff ffffffff ................ 01f150c0: ffffffff ffffffff ffffffff ffffffff ................ 01f150d0: ffffffff ffffffff ffffffff ffffffff ................ 01f150e0: ffffffff ffffffff ffffffff ffffffff ................ 01f150f0: ffffffff ffffffff ffffffff ffffffff ................ Marvell>> md 0x01f12040 ... and the actual data will be located by barebox: 01f12040: ffffffff ffffffff ffffffff ffffffff ................ 01f12050: ffffffff ffffffff ffffffff ffffffff ................ 01f12060: ffffffff ffffffff ffffffff ffffffff ................ 01f12070: ffffffff ffffffff ffffffff ffffffff ................ 01f12080: ffffffff ffffffff ffffffff ffffffff ................ 01f12090: ffffffff ffffffff ffffffff ffffffff ................ 01f120a0: ffffffff ffffffff ffffffff ffffffff ................ 01f120b0: ffffffff ffffffff ffffffff ffffffff ................ 01f120c0: ffffffff ffffffff ffffffff ffffffff ................ 01f120d0: ffffffff ffffffff ffffffff ffffffff ................ 01f120e0: ffffffff ffffffff ffffffff ffffffff ................ 01f120f0: ffffffff ffffffff ffffffff ffffffff ................ 01f12100: ffffffff ffffffff ffffffff ffffffff ................ 01f12110: ffffffff ffffffff ffffffff ffffffff ................ 01f12120: ffffffff ffffffff ffffffff ffffffff ................ 01f12130: ffffffff ffffffff ffffffff ffffffff ................ Marvell>> go $loadaddr ## Starting application at 0x02000000 ... barebox 2014.11.0-00127-g263044f25b47-dirty #14 Thu Nov 13 09:26:54 CET 2014 Board: NETGEAR ReadyNAS 104 SoC: Marvell 6710 rev 1 mdio_bus: miibus0: probed mvneta_setup_tx_rx: port = 0x01f11210, regbase = f1070000 mvneta_setup_tx_rx: port = 0x01f11520, regbase = f1074000 eth1: got preset MAC address: 28:c6:8e:36:df:57 of_get_named_gpio_flags: unable to get gpio num of device pca95540: -19 of_get_named_gpio_flags: unable to get gpio num of device pca95540: -19 of_get_named_gpio_flags: unable to get gpio num of device pca95540: -19 of_get_named_gpio_flags: unable to get gpio num of device pca95540: -19 malloc space: 0x01f00000 -> 0x03dfffff (size 31 MiB) environment load /dev/env0: No such file or directory Maybe you have to create the partition. no valid environment found on /dev/env0. Using default environment running /env/bin/init... /env/bin/init not found barebox:/ ethact eth1 barebox:/ eth1.ipaddr=192.168.77.133 barebox:/ eth1.netmask=255.255.255.0 barebox:/ ping 192.168.77.1 eth1: 1000Mbps full duplex link detected Send a packet (len = 42, txdesc=0x01f15000): 01f12040: ff ff ff ff ff ff 28 c6 8e 36 df 57 08 06 00 01 ......(..6.W.... 01f12050: 08 00 06 04 00 01 28 c6 8e 36 df 57 c0 a8 4d 85 ......(..6.W..M. 01f12060: 00 00 00 00 00 00 c0 a8 4d 01 ........M. eth1: transmit error 3 ping failed: I/O error Hmm, never saw an I/O error before. What does error 3 mean? barebox:/ md 0x01f15000 01f15000: ffffffff 002affff 01f12040 ffffffff ......*.@ ...... 01f15010: ffffffff ffffffff ffffffff 01f14ff8 .............O.. 01f15020: 00000028 03e303f0 03e30408 00000000 (............... 01f15030: 00000000 01f11fd8 00000000 03e616fc ................ 01f15040: 00000000 01f15070 01f11fbc 00000028 ....pP......(... 01f15050: 03e303f0 03e30408 00000000 00000000 ................ 01f15060: 01f1507c 00000000 03e616fc 00000000 |P.............. 01f15070: 01f150b0 01f15044 00000010 746f6f62 .P..DP......boot 01f15080: 666f2e6d 65657274 01f15000 00000028 m.oftree.P..(... 01f15090: 03e303f0 03e30408 00000000 00000000 ................ 01f150a0: 01f150bc 00000000 03e616fc 00000000 .P.............. 01f150b0: 01f150f0 01f15070 00000010 746f6f62 .P..pP......boot 01f150c0: 6e692e6d 64727469 01f15000 00000028 m.initrd.P..(... 01f150d0: 03e303f0 03e30408 00000000 00000000 ................ 01f150e0: 01f150fc 00000000 03e616fc 00000000 .P.............. 01f150f0: 01f15224 01f150b0 00000018 746f6f62 $R...P......boot barebox:/ md 0x01f12040 01f12040: ffffffff c628ffff 57df368e 01000608 ......(..6.W.... 01f12050: 04060008 c6280100 57df368e 854da8c0 ......(..6.W..M. 01f12060: 00000000 a8c00000 ffff014d ffffffff ........M....... 01f12070: ffffffff ffffffff ffffffff ffffffff ................ 01f12080: ffffffff ffffffff ffffffff ffffffff ................ 01f12090: ffffffff ffffffff ffffffff ffffffff ................ 01f120a0: ffffffff ffffffff ffffffff ffffffff ................ 01f120b0: ffffffff ffffffff ffffffff ffffffff ................ 01f120c0: ffffffff ffffffff ffffffff ffffffff ................ 01f120d0: ffffffff ffffffff ffffffff ffffffff ................ 01f120e0: ffffffff ffffffff ffffffff ffffffff ................ 01f120f0: ffffffff ffffffff ffffffff ffffffff ................ 01f12100: ffffffff ffffffff ffffffff ffffffff ................ 01f12110: ffffffff ffffffff ffffffff ffffffff ................ 01f12120: ffffffff ffffffff ffffffff ffffffff ................ 01f12130: ffffffff ffffffff ffffffff ffffffff ................ That looks ok. And when I warm reset and look at these memory locations using U-Boot again, they are still holding that arp packet. So if it's really a cache problem, the cache is very insistent and U-Boot has a problem, too. Furthermore Sascha interpreted barebox:/ cpuinfo implementer: Marvell Semiconductor Inc. architecture: v7 core: unknown r1p1 cache: 512 bytes (linelen = 64) Control register: A W P D L Z I DT IT U XP as: The MMU is off (as there is no M in the Control register line) and so caches must be off, too. Looking into the vendor U-Boot sources and they didn't change drivers/net/kirkwood_egiga.c compared to v2009.08. Best regards Uwe -- Pengutronix e.K. | Uwe Kleine-König | Industrial Linux Solutions | http://www.pengutronix.de/ | _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
