| 09:04.46 | *** join/#arm-netbook lkcl (~lkcl@nat66.mia.three.co.uk) |
| 09:43.29 | fjp | Backporting from 24.7 to 24.2 is indeed meaningless, but that's not what I've ever said. |
| 09:44.33 | fjp | It looks as if Samsung has provided a patch for .24.2, which is why we're running .24.2. |
| 09:46.03 | fjp | lkcl found a source repo for SmartQ that contains that patch, but they've rebased it on top of .24.7. |
| 09:46.17 | fjp | So that's where the .24.2 and .24.7 come from. |
| 09:47.47 | fjp | What I've said is that if we do get the source from the factory, I'm planning to try to forward port that to current kernels. And try to split up the mega patch into separate changes. |
| 09:51.13 | fjp | bjdooks: Are you familiar with that patch from Samsung? I've put it up here: http://people.debian.org/~fjp/tmp/CT-PC89E/samsung_2.6.24.2.patch.gz. |
| 09:51.28 | fjp | There's a diffstat in the same dir. |
| 09:53.57 | fjp | And a changelog inside the patch. |
| 10:06.30 | bjdooks | every time I see vendor code, I tend to get the urge to hurt the vendor |
| 10:07.30 | fjp | :-) |
| 10:07.54 | bjdooks | a lot of this is now replaced |
| 10:08.30 | fjp | Is the video/s3c/* possibly still relevant? |
| 10:08.47 | bjdooks | probably not, i wrote a new fb driver |
| 10:09.16 | fjp | Right. So all we'd need is support for our LCD with that new driver? |
| 10:10.24 | fjp | That new driver's not yet in .33, correct? |
| 10:16.34 | bjdooks | no, been there for a while |
| 10:16.45 | bjdooks | s3c6410 support was merged several kernels ago |
| 10:17.05 | bjdooks | hmm, what does the CT-PC89E report in /proc/cpuinfo? |
| 10:18.37 | fjp | Not quite complete though. Saw an RTC patch yesterday and hardware sensor's also only 2410. |
| 10:19.23 | fjp | # cat /proc/cpuinfo |
| 10:19.23 | fjp | Processor : ARMv6-compatible processor rev 6 (v6l) |
| 10:19.23 | fjp | BogoMIPS : 719.25 |
| 10:19.24 | fjp | Features : swp half thumb fastmult vfp edsp java |
| 10:19.24 | fjp | CPU implementer : 0x41 |
| 10:19.25 | fjp | CPU architecture: 7 |
| 10:19.25 | fjp | CPU variant : 0x0 |
| 10:19.26 | fjp | CPU part : 0xb76 |
| 10:19.28 | fjp | CPU revision : 6 |
| 10:19.28 | fjp | Cache type : write-back |
| 10:19.30 | fjp | Cache clean : cp15 c7 ops |
| 10:19.30 | fjp | Cache lockdown : format C |
| 10:19.32 | bjdooks | not fully complete, there's a few bits still in rmk's tree |
| 10:19.33 | fjp | Cache format : Harvard |
| 10:19.33 | fjp | I size : 16384 |
| 10:19.36 | fjp | I assoc : 4 |
| 10:19.36 | fjp | I line length : 32 |
| 10:19.38 | fjp | I sets : 128 |
| 10:19.38 | fjp | D size : 16384 |
| 10:19.40 | fjp | D assoc : 4 |
| 10:19.40 | fjp | D line length : 32 |
| 10:19.43 | fjp | D sets : 128 |
| 10:19.43 | fjp | Hardware : SMDK6410 |
| 10:19.45 | fjp | Revision : 0000 |
| 10:19.45 | fjp | Serial : 0000000000000000 |
| 10:19.46 | bjdooks | GAH! |
| 10:19.58 | bjdooks | when will people learn how to set machine types properly! |
| 10:20.08 | bjdooks | IT IS NOT AN SMDK6410! |
| 10:20.21 | bjdooks | HOT FLAMING DEATH to whoever built the bootloader |
| 10:20.33 | fjp | Ugh. So what is it? |
| 10:21.02 | bjdooks | they should have allocated a new machine number and used that |
| 10:22.59 | fjp | So it's 6410 family, but not SMDK. |
| 10:23.10 | bjdooks | yeah |
| 10:23.15 | bjdooks | SMDK6410 is the samsung dev board |
| 10:23.25 | bjdooks | just like this f'ing ubisurder isn't an SMDK2450 |
| 10:25.01 | bjdooks | it might be close to an SMDK6410 |
| 10:25.12 | fjp | AFAICT they're not using mkimage. The kernel that's on flash does not have a header. |
| 10:25.21 | bjdooks | doesn't matter |
| 10:25.33 | bjdooks | whoever the kernel is booted, the machine number should be passed into the kernel |
| 10:26.14 | fjp | Either by u-boot itself or using a header, correct? |
| 10:26.22 | bjdooks | no, by uboot |
| 10:26.35 | bjdooks | the kernel itself doesn't use mkimage |
| 10:26.42 | fjp | Right. |
| 10:26.44 | bjdooks | mkimage is purely for uboot's benefit |
| 10:27.14 | fjp | also has a QNAP NAS |
| 10:27.33 | fjp | But that's all my arm knowledge. |
| 10:27.55 | bjdooks | I have a variety of SMDKs, several third party S3C arm boards, a couple of Kirkwood boards, and a whole pile of stuff built locally |
| 10:35.10 | mnemoc | expectable :) |
| 10:37.41 | bjdooks | so it is pretty much start with an SMDK and work downwards |
| 10:37.58 | bjdooks | I expect the only hard thing will be sorting out the correct LCD panel timings |
| 10:38.15 | bjdooks | anyway, real work to do. maybe around later |
| 10:38.23 | fjp | Thanks so far |
| 16:19.05 | lkcl | bjdooks: that's why i'm looking into this with IDAPro (precisely to get the LCD panel timings) the chitech factory have _one_ thoroughly overloaded software engineer. adam is not getting through to the sales lady that they need to take up merely 2 minutes of that overloaded engineer's time with a "cp .../linux-2.6.24.2 /mnt/sdcard" command and that's _it_. |
| 16:19.25 | lkcl | so he's going to arrange another visit, get through to the factory owner. |
| 17:51.22 | *** join/#arm-netbook SQlvpapir (~teis@188.177.95.62) |
| 17:58.11 | fjp | lkcl: Ask adam to do a cp for u-boot at the same time :-) |
| 18:07.49 | lkcl | ack :) |
| 19:29.14 | lkcl | ok i believe i have the right LCD initialisation parameters, now |
| 19:29.59 | lkcl | the next thing is to get the GPIO sequence, which is a complete fuck-up: they're bypassing the standard s3c gpio setpar functions |
| 20:39.42 | bjdooks | nothing like good coding |
| 20:39.56 | bjdooks | you might want to add that datasheet url to a wiki or something |
| 20:42.01 | lkcl | ack. |
| 20:42.03 | lkcl | good point |
| 20:42.39 | bjdooks | do you have detailed board piccies? and a wikki? |
| 20:43.23 | lkcl | http://elinux.org/CT-PC89E |
| 20:43.47 | lkcl | http://lkcl.net/arm_systems/CT-PC89E/photos/ |
| 20:44.43 | bjdooks | interesting, they've used a DIMM socket for memory? |
| 20:45.54 | lkcl | nope - a DIMM socket for the CPU, memory and NAND flash! |
| 20:45.59 | bjdooks | ah |
| 20:46.04 | lkcl | the entire machine's "upgradeable" |
| 20:46.20 | lkcl | michelle konzack is looking to do a marvell 1.2ghz SO-DIMM |
| 20:46.21 | bjdooks | sort of like: http://www.simtec.co.uk/products/IM2440D20/intro.html |
| 20:46.30 | lkcl | and i have another ... yeah, just like that |
| 20:47.06 | lkcl | another hardware engineer guy i know who is willing to try doing an S5PC100 SO-DIMM, based on the odroid schematics it should be really straightforward |
| 20:47.43 | bjdooks | it depends on how much power you can get to it |
| 20:47.49 | lkcl | _theeen_... michelle's _also_ found a factory in china that has a 10.2in chassis/motherboard (no SO-DIMM) for $USD 48 |
| 20:47.55 | bjdooks | the S5PC100 requires ext. memory so may be more difficult to get routed |
| 20:48.02 | bjdooks | and DDR/DDR2 is a git to route |
| 20:48.25 | bjdooks | hmm, so two usb hubs, an lvds encoded on the top |
| 20:48.33 | lkcl | yep. |
| 20:48.48 | lkcl | DM9000 in the top left corner |
| 20:48.52 | bjdooks | and an DM9000B |
| 20:49.00 | lkcl | WM9713 in the bottom right |
| 20:49.20 | lkcl | "unidentified" and highly dodgy-looking keyboard controller bottom centre.... :) |
| 20:49.24 | bjdooks | hmm, since when davicom relax their output capacitabce |
| 20:49.38 | lkcl | DM = davicom? |
| 20:49.42 | bjdooks | yes |
| 20:49.53 | lkcl | ahh... you spotted a design error? |
| 20:50.18 | bjdooks | all their schematics claim you need 220uF on the output regulators |
| 20:50.28 | lkcl | ok. |
| 20:50.30 | bjdooks | so far, we seem to the only people to put this on |
| 20:50.32 | lkcl | whoops :) |
| 20:50.58 | bjdooks | http://www.fluff.org/ben/pics/im-s3e-mk1/img_0472.jpg |
| 20:51.08 | lkcl | that would explain the problems i'm seeing on ethernet. |
| 20:51.15 | bjdooks | what sort of problems? |
| 20:51.16 | lkcl | disconnect and reconnect, and it's screwed. |
| 20:51.31 | lkcl | a reboot's needed |
| 20:51.38 | bjdooks | it could be the nasty socket theu've used |
| 20:51.47 | bjdooks | oh, looks like there's a 47uF tant near the transformer |
| 20:51.51 | lkcl | yes, that too. i've found that gravity has to lend a hand |
| 20:52.13 | bjdooks | however, they've crossed the tx/rx pair over on the output side and not put a gnd plan inbetween |
| 20:52.16 | fjp | I've had some problems with eth0 too. Seems to be the plug. |
| 20:52.33 | bjdooks | I've done quite a bit of work with teh DM9000 driver |
| 20:52.50 | bjdooks | that img_0472.jpg is my own 2 layer quickie |
| 20:52.54 | fjp | I also have had multiple times that kbd and touchpad are dead after boot. |
| 20:53.07 | bjdooks | the DM9000 works quite nicely |
| 20:53.16 | bjdooks | even if I couldn't get the FPGA to solder |
| 20:54.33 | bjdooks | do you have a closeup of the kbd controller? |
| 20:54.47 | lkcl | fjp: there are some GPIO cock-ups - direct access to GPIO registers - that could be messing things up. they compile _all_ drivers into the kernel. |
| 20:55.05 | lkcl | bjdooks: the "unidentified" chip? no, that hi-res photo is "it" for now. |
| 20:56.07 | lkcl | i see quite a few s3c64xx arch's use the DM9000. |
| 20:56.21 | bjdooks | they're the best we've found |
| 20:56.25 | bjdooks | small package |
| 20:56.29 | bjdooks | paralell bus |
| 20:56.35 | bjdooks | and easy to drive |
| 20:58.26 | lkcl | btw - you do hardware designs, i take it? |
| 20:58.49 | lkcl | using e.g. eagle or something. |
| 20:58.57 | bjdooks | that one was eagle |
| 20:59.00 | lkcl | ok. |
| 20:59.07 | bjdooks | our hardware designed uses proper tools |
| 20:59.46 | lkcl | the guy who's volunteered to do the S5PC100 board (as a free software project), is there any way that you'd be able to double-check what he does? |
| 20:59.55 | lkcl | S5PC100 SO-DIMM that is |
| 21:02.29 | lkcl | basically i'm not going to wait around for the "big" manufacturers to start creating ARM netbooks, so, stuff it, i'm coordinating and putting together people who happen to have some time or resources |
| 21:03.25 | bjdooks | i'd say send me one of these things, but i'm busy this week and will be packing next week for a flight |
| 21:03.36 | lkcl | or who could be persuaded to take a vested interest in some way (S5PC100 SODIMM is _useful_!) |
| 21:03.45 | lkcl | no problem. |
| 21:03.56 | bjdooks | trouble is the pinout, as noted on the list |
| 21:04.01 | lkcl | there are.... eek, 25 left of the samples they knocked up. |
| 21:04.39 | lkcl | yehhs, i knooow. adam's going to ask them for it. if they don't provide, we just simply create a 10.2in netbook using the chassis and motherboard that michelle found, instead. |
| 21:04.43 | lkcl | and drop chitech. |
| 21:05.14 | lkcl | much as i love their little machine. |
| 21:05.27 | lkcl | i'm not one for hanging about. |
| 21:06.12 | bjdooks | we designed the s3c2440 DIMM for someone elses netbook |
| 21:06.16 | lkcl | well, it will take a while to get one of the CT-PC89E samples from chitech, anyway. |
| 21:06.29 | lkcl | oh? cool |
| 21:06.36 | bjdooks | note, legally they're obliged to give you source |
| 21:07.19 | lkcl | i knowww, ben. adam's going to go down to the factory with the list of bugs and other issues. the factory owner wasn't there when he first went there |
| 21:07.39 | bjdooks | the threat of lawyers might get them to move |
| 21:07.58 | lkcl | yehh, i'd rather it was "cooperative" at this stage. |
| 21:08.25 | lkcl | we're trying the "we'll do stuff for you" approach. |
| 21:08.38 | lkcl | next will be the "we'll drop you and go with your competitors" approach |
| 21:08.50 | bjdooks | for the DM9000, you'll need to find the hardware address it is connected to |
| 21:08.58 | lkcl | yep, that's in the kern.logs |
| 21:09.12 | bjdooks | f7600300 looks like a virtual address to me |
| 21:09.19 | lkcl | DM9000 is at f7600300,f7600304 IRQ74 |
| 21:09.22 | lkcl | ahh. |
| 21:09.24 | lkcl | oh. yes. |
| 21:09.33 | lkcl | mmm.... |
| 21:09.37 | bjdooks | you might find from /proc/iomem where it came from |
| 21:09.45 | bjdooks | iirc, it should nab the phys address |
| 21:10.21 | lkcl | yep it's f7600300-303 |
| 21:10.39 | lkcl | s3c2410-nand is 70200000 |
| 21:10.44 | lkcl | ok i record this in the wiki |
| 21:10.49 | lkcl | we need it |
| 21:11.24 | lkcl | i gotta go soon. any other suggestions of cat /proc/* to record? |
| 21:12.04 | bjdooks | hmm, the datasheet doesn't have anything for F76... |
| 21:13.59 | bjdooks | hmm, the print on the log is the _virtual_ address |
| 21:14.16 | bjdooks | /proc/iomem should be the phys |
| 21:14.19 | bjdooks | and they shouldn't be the same |
| 21:14.37 | lkcl | hee hee |
| 21:15.09 | bjdooks | and this patch has no invocation of a dm9000 in it |
| 21:16.48 | bjdooks | you might have to read the IO bank configs to work out which one it is in |
| 21:18.10 | lkcl | ok |
| 21:22.18 | lkcl | ok - i gotta get |
| 21:22.43 | fjp | bjdooks: You aware that spi_s3c64xx.c because of '#include <mach/dma.h> |
| 21:22.43 | fjp | ' |
| 21:22.53 | fjp | which does not exist? |
| 21:23.10 | bjdooks | which kernerl? |
| 21:23.23 | fjp | .33 and AFAICT .34 too |
| 21:23.28 | bjdooks | due to the non-merge of rmk's tree, stuff may be borked due to missing stuff |
| 21:23.44 | fjp | Where is his tree? |
| 21:23.56 | bjdooks | somewhere on master.kernel.org |
| 21:24.00 | bjdooks | i'm not sure he exports it |
| 21:24.17 | fjp | looked for it in vain yesterday |
| 21:25.15 | bjdooks | the best you can do currently is my next-samsugn |
| 21:25.23 | fjp | K |
| 21:26.19 | bjdooks | also, mmc0 and mmc1, the address args are pretty much meaningless on that output |
| 21:26.29 | bjdooks | you'll need to find which hsmmc controller they're conected to |
| 21:27.35 | bjdooks | and how the card-detect is configured for hsmmc1 (may be clue in gpio setup) |
| 21:32.29 | fjp | bjdooks: What's the git:// URI for that? It isn't listed in the summary. |
| 21:32.51 | bjdooks | git://git.fluff.org/bjdooks/linux.git |
| 21:34.17 | fjp | thx |
| 21:34.25 | bjdooks | bah, got to go to office tomorrow |
| 21:35.28 | bjdooks | right, pizza ordered |
| 21:35.45 | bjdooks | will be off until tomorrow (possibly tomorrow evening) once it arrives |
| 21:49.38 | SQlvpapir | thats a long time you allocated just for eating pizza |
| 21:50.09 | bjdooks | no, going to watch True Blood that was recorded a couple of days ago, fill bag to take to work with things like laptop and then go to bed |
| 21:50.16 | bjdooks | office means 0645 start |
| 21:50.32 | bjdooks | packing things at that time of the morning generally a recipe for disaster |
| 21:51.59 | SQlvpapir | you know I'm just kidding :) |
| 22:12.37 | fjp | bjdooks: |
| 22:12.39 | fjp | <PROTECTED> |
| 22:12.42 | fjp | /home/fjp/projects/kernel/linux-2.6/drivers/spi/spi_s3c64xx.c:31:22: error: plat/spi.h: No such file or directory |
| 22:13.06 | fjp | That's with your branch merged with .33 |
| 22:26.20 | fjp | lkcl: With Ben's next-samsung branch I get the same result as with vanilla .33. |
| 22:26.42 | fjp | was hoping for some netconsole output |
| 22:27.16 | fjp | Need to test with a different system if I'm listening correctly. Maybe need a hub instead of a switch in between. |
