| 01:53.51 | *** join/#arm-netbook fjp` (~fjp@ip4da6b463.direct-adsl.nl) |
| 11:51.49 | *** join/#arm-netbook lkcl (~lkcl@nat66.mia.three.co.uk) |
| 12:42.22 | fjp | lkcl: Fixed the RAM issue. |
| 12:56.32 | mnemoc | :D |
| 13:08.41 | lkcl | <PROTECTED> |
| 13:08.46 | lkcl | yaay |
| 13:11.36 | lkcl | what was it? |
| 13:48.29 | fjp | lkcl: pc89e_fixup() in mach-smdk6410.c - hardcoded to 128MB... |
| 13:48.38 | lkcl | ohhh of course |
| 13:49.39 | lkcl | fjp: i'm doing a test with my new shiny oscilloscope, on the GL850G, to see what frequency it's being asked to run at. the spec say 12mhz, 27 or 48mhz on piiin.... 6! |
| 13:49.40 | fjp | I've got a patch that allows us to use a separate mach-ctpc89e.c file. |
| 13:49.51 | lkcl | cross fingers, ok? |
| 13:50.08 | fjp | Don't burn your fingers ;-) |
| 13:50.14 | lkcl | do you want to set up a git repository somewhere, e.g. github or better? how's the alioth thing coming? |
| 13:50.35 | fjp | I'm waiting for alioth approval. |
| 13:50.39 | lkcl | ack |
| 13:50.51 | fjp | Will publish the repo I'm working in now when I have that. |
| 14:26.13 | lkcl | fjp: distinctively odd: i'm measuring 0.4 microseconds which is only 2.5mhz |
| 14:27.31 | lkcl | beh?? |
| 14:50.23 | lkcl | fjp: ok - the GL850G isn't operational on the mer-smartq 2.6.24.7 kernel. |
| 14:50.40 | lkcl | which means that there's a GPIO which is connected to RESET. |
| 14:54.23 | fjp | That could explain why we get more hubs with our kernels. |
| 14:55.36 | fjp | 9 instead of 5 |
| 14:56.07 | fjp | And it looks as if the first one is the one we don't want. |
| 14:57.29 | fjp | Eh, no. The first one is the one we want (1-1 gets adresses 1-5; 1-2 gets 6-9) |
| 14:58.58 | fjp | So the arm chip itself is running USB? |
| 15:00.50 | lkcl | ok i looked at the circuit around pin 13 (RESET) and it's connected to ground, via a resistor. GL850G datasheet says this should be a 10k. |
| 15:01.05 | lkcl | so there is _not_ a GPIO involved, not on RESET anyway. |
| 15:01.24 | lkcl | might be on power... *sigh* :) |
| 15:30.05 | fjp | Hmm. This is something to try for D-I: combine the initrd and kernel into a single "kernel" image... |
| 15:54.58 | fjp | Should work. The only thing needed then is to override the kernel cmdline passed by u-boot, but that should be hackable. |
| 16:05.17 | fjp | And I guess that could be how ChiTech is doing it and explains why they have init=/init in their u-boot (although AFAICT that's redundant)... |
| 16:05.44 | fjp | s/it/doing the flashing of the root FS/ |
| 16:06.30 | fjp | Hmm. Cool ibot feature that. I typed 's/it/doing the flashing of the root FS/' :-) |
| 16:17.10 | bjdooks | mkimage will take a kernel and initrd into a format uboot understands |
| 16:17.21 | bjdooks | is back, with new adsl router |
| 16:17.27 | fjp | :-) |
| 16:18.06 | fjp | Would wraplinux work as well for arm or is it better to stick with mkimage? |
| 16:18.13 | bjdooks | wraplinux? |
| 16:18.30 | fjp | Tool that does the same from hpa |
| 16:18.50 | bjdooks | never tried it |
| 16:22.42 | fjp | We have 7.2MB between start of kernel and u-boot in flash. Should be able to make that fit fairly comfortably. |
| 16:26.16 | bjdooks | right, i should test -rc1-git to see if stuff is still working |
| 16:26.39 | lkcl | is going to look at IDA disassembler vmlinux.txt, see if there's some GPIOs to pull for the GL850s |
| 16:26.43 | lkcl | bjdooks: good luck |
| 16:30.54 | bjdooks | I ended up desoldering an couple of iPAQS to find out |
| 16:31.06 | bjdooks | you could always go and see what GPIOs are set output in the configuration] |
| 16:31.39 | fjp | is considering how to repartition his external and laptop harddrives. Banning Windows to the external drive :-) |
| 16:32.00 | bjdooks | I'm going to be testing -rc1-git sometime today |
| 16:32.59 | bjdooks | and then see if they're set to 1/0, etc |
| 16:44.57 | fjp | bjdooks: Hmm. AFAICT mkimage will only add a header to an initramfs or kernel, but does not have functionality to combine them into a single image. |
| 16:45.35 | fjp | And the funny part is that the kernel u-boot now boots from flash does not have any mkimage header. It's a plain kernel image. |
| 16:48.16 | lkcl | fjp: i found the function "ds2431_dm9000mac_read" |
| 16:48.38 | lkcl | so yes they _are_ using the DS2431 EEPROM for storing the MAC address. |
| 16:49.08 | lkcl | there are some saad functions nearby, called AuthenticateDS2431onSecret4_app |
| 16:49.13 | lkcl | ohh deearie me... |
| 16:54.14 | fjp | lkcl: Can we read what's on that from Linux? |
| 16:55.26 | lkcl | with a patch for w1 EEPROM, yes. |
| 16:55.29 | lkcl | found it earlier. |
| 16:55.47 | lkcl | w1_DS2431.c by Bernhard Weirich |
| 16:56.54 | lkcl | http://tomoyo.sourceforge.jp/cgi-bin/lxr/source/drivers/w1/slaves/w1_ds2431.c |
| 16:58.04 | lkcl | http://lkcl.net/arm_systems/CT-PC89E/ds2431.patch |
| 16:58.07 | lkcl | fjp: ^ |
| 16:59.04 | lkcl | but what i want to know is wtF those "authenticate secret" functions are, and what SHA-1 is being used for |
| 17:00.42 | fjp | Hmm. Posted on lkml in 2008, but never merged looks like. |
| 17:01.35 | lkcl | well, we'll need it. |
| 17:01.44 | fjp | :-) |
| 17:05.10 | fjp | What have you done with the indentation? |
| 17:06.31 | fjp | takes patch from lkml instead |
| 17:11.32 | fjp | Included in my repo. |
| 17:15.59 | fjp | Hmm. Looks as if Kconfig and Makefile bits are missing. Added. |
| 17:18.14 | fjp | What's the master? DS2490 USB or DS2482 I2C or DS1WM 1-wire? |
| 17:43.36 | lkcl | fjp: in usb_hcd_s3c2410_probe, try this: |
| 17:43.55 | lkcl | usb_host_cli_en(S3C_USB_CLKSRC_48M, 0x01); |
| 17:44.09 | lkcl | ^usb_host_cli_en^usb_host_clk_en |
| 17:45.28 | lkcl | fjp: re DS2490 i have nooo idea. |
| 17:45.39 | lkcl | i'll try to look around. |
| 17:48.10 | lkcl | interestingly, it's on the SO-DIMM (the SD2431 is) |
| 17:49.02 | lkcl | if it's on the SO-DIMM then it's unlikely to be USB |
| 17:51.17 | fjp | enables all 3 for now as modules |
| 17:54.19 | fjp | lkcl: Before or after the gpio calls? |
| 17:54.59 | lkcl | take a look at usb_hcd_s3c2410_probe: due to the #ifdef's being {null} the first function called is actually usb_host_cli_en |
| 17:55.00 | fjp | Or just before the clk = |
| 17:55.06 | lkcl | and the 2nd param is actually 0x2 |
| 17:55.34 | lkcl | OTGH_PHY_CLK_VALUE is #defined to 0x2 (UTMI Intreface, Oscillator) |
| 17:55.50 | lkcl | but the chitech kernel is setting OTGH_PHY_CLK_VALUE to 0x1 |
| 17:56.04 | lkcl | purpose unknown... but who _gives_ a shit :) |
| 17:56.16 | fjp | K. |
| 17:56.34 | lkcl | so USB_HOST_PORT2_EN is _definitely_ #undef |
| 17:57.41 | fjp | building an image now |
| 18:04.52 | lkcl | ack |
| 18:08.14 | fjp | Hmm. Looks like you can even flash the kernel without flashing u-boot |
| 18:09.08 | lkcl | really? that's better |
| 18:09.52 | fjp | I renamed the uboot bin and got a blue screen saying it was flashing the kernel. |
| 18:12.23 | lkcl | excellent. start a campaign. "save our nand" |
| 18:15.52 | fjp | But the kernel I have looks old... |
| 18:16.44 | fjp | In fact, it's so old that I wonder what I've been testing yesterday and today... |
| 18:16.55 | fjp | # cat /proc/version |
| 18:16.55 | fjp | Linux version 2.6.24.7-s3c (root@aragorn) (gcc version 4.1.3 20080704 (prerelease) (Debian 4.1.2-25)) #0 PREEMPT Sat Mar 13 06:45:02 CET 2010 |
| 18:17.28 | fjp | This makes no sense... |
| 18:20.09 | lkcl | hehe |
| 18:20.22 | lkcl | beh? |
| 18:21.22 | lkcl | compiled... erk |
| 18:23.14 | lkcl | i get that same hanging that you did... gonna leave it for a while, see what happens... |
| 18:26.42 | fjp | Still getting -62 errors. |
| 18:28.05 | lkcl | i ain't. i get a hang. |
| 18:28.24 | lkcl | suspected sloowwww but couldn't wait around that long to find out |
| 18:28.41 | fjp | Still the old kernel maybe. Dunno. |
| 18:28.54 | fjp | Think I'll flash original to check. |
| 18:30.30 | fjp | OK. It's definitely flashing the kernel. |
| 18:30.57 | fjp | (without u-boot) |
| 18:37.47 | lkcl | ack |
| 18:38.15 | lkcl | ok, usb_host_cli_en is actually in mach-smdk6410.c |
| 18:38.42 | fjp | adds printk in usb_host_clk_en() |
| 18:39.19 | lkcl | i've confirmed that it's doing the right thing: case 0x01 is correct |
| 18:40.27 | lkcl | oh... wait... |
| 18:40.32 | lkcl | belay that. |
| 18:40.50 | lkcl | wrong switch statement :) |
| 18:42.30 | lkcl | okaaay. otg_phy_init() _is_ being called... |
| 18:42.33 | lkcl | checking that... |
| 18:44.45 | bjdooks | the otg phy also provides the 48MHz osc into the 48M clock |
| 18:45.42 | bjdooks | if you want to source the ohci clock from that, the phy has to be started even if it isn't being used |
| 18:47.59 | lkcl | bjdooks: 0x01 is bbeing written to S3C_USBORT_PHYCLK |
| 18:48.03 | lkcl | bjdooks: 0x01 is bbeing written to S3C_USBOTG_PHYCLK |
| 18:48.34 | lkcl | but that results in "hanging" rather than just errors |
| 18:48.49 | lkcl | time forr.... some printks! |
| 18:48.50 | fjp | lkcl: usb_host_clk_en is never called here? |
| 18:49.30 | fjp | Dinner first. |
| 18:49.52 | fjp | Or my kernel message level is incorrect... |
| 18:50.20 | lkcl | fjp: ah? oops. |
| 18:50.56 | lkcl | fjp: is CONFIG_PLAY_S3C64XX defined? is usb_hcd_s3c2410_probe being called? |
| 18:52.08 | fjp | guesses PLAT |
| 18:52.10 | fjp | yes |
| 18:52.12 | lkcl | yep |
| 18:52.21 | fjp | dunno |
| 19:01.20 | fjp | Weird. I do get the printk from usb_hcd_s3c2410_probe, but not from usb_host_clk_en |
| 19:01.45 | fjp | and the first is immediately before the call to the second |
| 19:02.20 | fjp | Ah! I know. |
| 19:02.25 | lkcl | beh? :) |
| 19:03.04 | fjp | I'm using arch/arm/mach-s3c6410/mach-ctpc89e.c, not mach-smdk6410.c |
| 19:03.12 | fjp | Been changing the wrong file :-/ |
| 19:03.31 | lkcl | loovely he he |
| 19:04.35 | fjp | Really dinner |
| 19:06.21 | lkcl | ah... |
| 19:06.22 | lkcl | go! :) |
| 19:06.32 | lkcl | i sort out some GPIOs it looks like i missed out |
| 19:10.09 | fjp | lkcl: Can you look at poweroff too? For me it shuts down correctly, but does not power off. |
| 19:10.25 | fjp | Would be nice not to have to reset all the time. |
| 19:13.08 | lkcl | fjp: yes, likewise. there's a function called poweroff, i thought it maybe was a lie |
| 19:13.16 | lkcl | but we need the button to work, first. |
| 19:13.26 | lkcl | and that means getting midfun-keys |
| 19:13.37 | fjp | Normally poweroff from ssh works too. |
| 19:14.02 | fjp | I saw the function. |
| 19:14.03 | lkcl | yes |
| 19:14.06 | lkcl | go eat! |
| 19:14.08 | lkcl | :) |
| 19:14.14 | fjp | It's in the oven... |
| 19:14.37 | lkcl | ahh |
| 19:14.49 | lkcl | ok. there's 3 sets of GPIO init needed |
| 19:15.11 | lkcl | and a whooole stack of crap. |
| 19:16.12 | lkcl | http://pastebin.ca/1839822 |
| 19:16.35 | lkcl | so that's cfgpin 1 setpin 0 on 0x1AD |
| 19:16.53 | lkcl | cfgpin pullup setpin 1 0 1 on 0x201 |
| 19:17.10 | lkcl | and cfgpin pullup setpin 1 0 1 on 0x200 |
| 19:17.20 | lkcl | theeeeNnn... there's a stack of crap not yet decoded, at the bottom |
| 19:17.30 | lkcl | after smdk_machine_init() |
| 19:18.05 | lkcl | brb |
| 19:45.45 | lkcl | back, probably briefly |
| 21:43.31 | lkcl | fjp: the smdk6410_hsmmc_init() and smdk6410_set_qos() functions are valid |
| 21:44.41 | lkcl | need to be added after smdk_machine_init() |
| 22:55.12 | fjp | lkcl: What exactly do you mean by "after"? |
| 22:55.45 | fjp | I already have those enabled here. |
| 22:55.56 | fjp | <PROTECTED> |
| 22:55.56 | fjp | <PROTECTED> |
| 22:55.56 | fjp | <PROTECTED> |
| 22:57.06 | fjp | (I've renamed the functions in my mach- file.) |
| 22:57.19 | fjp | . |
| 22:58.08 | fjp | had to rescue his laptop after a Windows XP installation to and external USB drive modified the partition table on the internal drive..... |
| 22:58.44 | fjp | And with root in encrypted LVM that's not quite trivial. |
| 23:03.41 | fjp | It also wrote the bootloader to the MBR of the internal disk instead of the external one. |
| 23:03.47 | fjp | And the damn XP install won't even boot... |
| 23:04.11 | fjp | Maybe if I chainload it from grub. |
| 23:05.57 | lkcl | loovely |
| 23:06.18 | lkcl | fjp: yep, that's it - smdk__ pc89e_msmmc.... setqos yep. |
| 23:06.45 | lkcl | you do need those cfgpin setpin etc. on 0x1ad, 0x201 and 0x200 |
| 23:07.03 | lkcl | and i've added smdk6410_dm9k_set() as well |
| 23:08.10 | lkcl | fjp: http://lkcl.net/arm_systems/CT-PC89E/stuff.patch |
| 23:08.57 | fjp | 404 |
| 23:09.03 | lkcl | 1sec... |
| 23:09.21 | lkcl | oh, i have to press return :) |
| 23:09.29 | lkcl | fjp: try again |
| 23:12.52 | fjp | Got it. And committed. |
| 23:13.11 | fjp | How much of that is trial and how much makes USB work now? ;-) |
| 23:14.49 | lkcl | pffh. it's none of it "trial", it's direct from the IDApro disassembly. |
| 23:15.06 | lkcl | and nooo it doesn't make USB work bleh :) |
| 23:17.20 | fjp | pity |
| 23:18.10 | lkcl | ok.... what about the timer values? |
| 23:18.47 | lkcl | timer tcon=00600d00, tcnt 15f8f tcfg 0000450,00000040 usec 0000e39 |
| 23:18.48 | lkcl | ? |
| 23:18.55 | lkcl | and on the "working" kernel it's: |
| 23:19.17 | lkcl | timer tcon 00600d00, tcnt 439a, tcfg 0001950,00000040, usec 00049f5 |
| 23:19.25 | lkcl | what's _that_ all about? |
| 23:21.49 | fjp | reboots to XP for a bit (or tries to) |
| 23:22.29 | lkcl | bye bye fjp :) |
| 23:22.33 | lkcl | see you next week ha ha |
| 23:30.48 | lkcl | whaat the fuuuck is going on??? s3c6410_init_clocks() |
| 23:32.50 | lkcl | greeeeat. loovely. |
| 23:33.46 | fjp | no dice with XP |
| 23:34.40 | lkcl | RRRIGHT, i think i have a clue |
| 23:34.50 | fjp | Cool |
| 23:34.58 | lkcl | fjp: take a look at arch/arm/mach-s3c6410/clock.c |
| 23:35.02 | lkcl | grep for "usb-host" |
| 23:36.07 | lkcl | i think there's something different |
| 23:37.31 | lkcl | also sclk_48m that's.... err... |