**** BEGIN LOGGING AT Mon Oct 28 02:59:58 2013 Oct 28 12:12:26 lumag_: I've sent to linux-mtd. Let's hope we get dwmv2 attention Oct 28 12:12:38 ant_work, :) Oct 28 12:12:39 bluelightning: hey, wb Oct 28 12:12:51 Either dwmv2 or artem should probably respond Oct 28 12:12:51 hey ant_work Oct 28 12:14:02 bluelightning: I've seen RP has commitetd the fixes for jffs2. Unfortunately the 3rd patch fixing sum.jffs2 is not, as the last 2 (those really optional) Oct 28 12:15:38 bluelightning: If I have to guess it, RP dislikes new vars so refused the JFFS2_NOERASEBLOCK and JFFS_PADDING. That's ok, we customize EXTRA_IMAGECMD Oct 28 12:15:53 bluelightning: ..but that other patch is needed or we get build failures :/( Oct 28 12:17:51 (talking about patchwork oe-core #60281) Oct 28 12:19:47 ant_work: don't know why it hasn't been merged, you'll need to talk to RP Oct 28 12:20:34 ok, let him decompress some hours more on the first day of the week ;) Oct 28 12:28:24 lumag_: btw even if it was late, I've tried to add more debug to cfi_util tryng to see how Q R Y were mapped... Oct 28 12:28:27 boom Oct 28 12:28:58 priktk -> kernel oops because on first run cfi->interleave and cfi->device_type were NULL Oct 28 12:29:10 Hmm. Oct 28 12:29:17 it was really too lat but I didn't expect null values Oct 28 12:30:43 I'll do in gen_probe as you said Oct 28 14:03:06 ant_work, I never understood flash code in qemu, so I can't debug mtd in qemu. All my zaurii are at work now and I'm at home babysitting. Oct 28 14:03:19 So probably I can't directly help you :( Oct 28 14:05:53 you're helping me understanding/reading the hw specs ;) Oct 28 14:07:35 I've still the little doubt that for some reason Sharp could have customized this setup and have wired the chips in 8bit more. Hopefully the old Sharp driver doing 4x8 was just copying old stuff from the PRE-CFI models Oct 28 14:08:18 Things might be different for sl-5000/sl-5500/sl-5000d etc. Oct 28 14:08:36 on the serv.man there is that little reference at page 5 or 6 Oct 28 14:08:57 iirc you see #BYTE connected to Vcc3 and Vpp Oct 28 14:09:15 ..but on the spec those chips have no #BYTE pin ... Oct 28 14:09:53 it was the old Strataflash doing 16/8, like the one on ipaq h3600 Oct 28 14:11:57 most of the CFI code is done with that Intel flash in mind (i.e. 28F640J3) Oct 28 14:16:16 heh, on this latter A0 is used only in 8 bit mode (the A0 input buffer is turned off when BYTE# is high) Oct 28 14:16:18 so input addresses are Oct 28 14:16:32 32M A[21:0] 64M [A22:0] 128M [A23:0] Oct 28 14:17:00 maybe all the CFI code need just a shift of one 'address' for our flash... Oct 28 14:28:00 Hmm. Looks like Artem has stepped aside from co-maintaining mtd. Rather sad I'd say. Oct 28 14:29:32 "Tizen IVI maintainer' :((( Oct 28 14:36:34 ah :/ Oct 28 15:01:49 bluelightning: wrt meta-hh, we can remove the udev .bbappend and rburton even said we don't need the xserver-xf86-config.bbappend Oct 28 15:02:30 ant_work: right, I see Chase's patch got merged, yay :) Oct 28 15:02:57 iirc Tartarus patch even superseded it Oct 28 15:41:49 ant_work, regarding collie: Oct 28 15:42:19 I have checked collie.c. I'm quite surprized that it contains flash definitions in two banks Oct 28 15:42:45 Maybe it won't help your issue, but I think those two regions should be merged. Oct 28 15:43:48 Because according to service manual cs1 is not used at all. Oct 28 15:44:27 I'm talking about collie_io_desc Oct 28 15:45:06 Played a bit with device tree. Oct 28 15:45:21 For PXA. Oct 28 15:46:09 It looks like our precious pxa devices have too many connections and dependencies to be expressed as a simple dts Oct 28 15:46:38 I'm especially unsure about all the crazyness in defining and handling scoop/pcmcia devices Oct 28 15:53:01 It looks like even sa1100 can be simpler to migrate Oct 28 15:53:21 (scoop/pcmcia problem is still present). Oct 28 15:55:44 good to know Oct 28 15:56:04 about collie, yes, it looks like some dev-devices had 32M Oct 28 15:56:17 I asked once about the map sa1100-0: Found different chip or no chip at all (mfr 0x4, id 0x3001) at 0x1000000 Oct 28 15:57:51 this address is CS2 in sa-1100.h Oct 28 16:01:13 ah, no, those are the second 16M Oct 28 16:01:13 on CS0 Oct 28 16:02:22 #define SA1100_CS0_PHYS 0x00000000 Oct 28 16:02:22 30 #define SA1100_CS1_PHYS 0x08000000 . . .CS5 Oct 28 16:02:43 #define SA1100_CS2_PHYS 0x10000000 Oct 28 16:32:10 bbl Oct 28 21:23:50 ant_home, according to hpc.ru (sorry ;)) sl-5000d had 32 Mb RAM and 16 Mb ROM. sl-5500 had 64 Mb RAM and 16 Mb ROM. Oct 28 21:23:58 Thus I don't understand collie_io_desc. Oct 28 21:24:18 Does it describe RAM? Then comments are wrong... Oct 28 21:25:11 it seems there is an empty connector on the back of the chips Oct 28 21:25:54 http://www.tekprosystems.com/images/zaurus/zguts1.jpg Oct 28 21:30:12 http://tuxmobil.org/pics/sharp_sl5000_block.jpg Oct 28 21:31:24 ant_home, isn't it the display connector? Oct 28 21:31:49 err... don't tell me.. seems like an old RAM slot on printers Oct 28 21:34:46 See http://www.chipfind.ru/datasheet/pdf/toshiba/tc59sm716.pdf Oct 28 21:35:04 The board has two tc59sm716 chips (in the left top corner). Oct 28 21:35:23 Each is 2M * 4 * 16bit/8 = 16 Mbytes of RAM Oct 28 21:35:40 And page 38 of servman. Oct 28 21:35:51 It shows exactly this connector for the display. Oct 28 21:35:53 :( Oct 28 21:36:05 http://www.tekprosystems.com/images/zaurus/zguts4.jpg Oct 28 21:36:52 well, there were 3 models: SL 5000 SL 5500 SL 5000D Oct 28 21:39:00 OpenZaurus said The 5500 and 5000D are the same, but for the memory size (64 vs 32MB) Oct 28 21:39:30 I confirm 5500 has 64MB RAM and 16MB NOR Oct 28 21:40:03 5000D has 32 RAM/16 NOR Oct 28 21:40:14 I ain't sure there was difference between 5000 and 5000d Oct 28 21:40:35 iirc the -D was cut-down for devs 32MB ;) Oct 28 21:41:21 Yes, but I thought that is was cut of 5500 Oct 28 21:41:56 wikipedia lists only 5000d and 5500 Oct 28 21:43:42 sharp kernel has 5000d, 5500, 5500g and 5500up defconfigs Oct 28 21:44:18 ok, then it was "SL-5000 Developers version were generally available" Oct 28 21:45:26 strangely some sites list 3 models Oct 28 21:46:53 Looking at kernel internals, there were 4 earlier versions - TS, TR0, TR1, DEV Oct 28 21:47:03 omg Oct 28 21:47:19 They all slightly differ. Oct 28 21:47:58 First three most probably were technical samples, never intended to be sold. Oct 28 21:48:42 It is typical to produce "beta" versions of devices to be used internally for testing. Then various hardware bugs pop up. Oct 28 21:49:22 about sa1100_flash, I was briefly looking in mach-sa1100 and i.e. ipaq h3600 uses higher CS# Oct 28 21:49:33 3 of them iirc Oct 28 21:49:33 E.g. TR0 uses audio chip gpio (instead of Scoop gpio) to control charger Oct 28 21:50:48 Ah, so comments are definitely wrong and collie_io_desc is about RAM and not flash. Oct 28 21:51:59 static struct resource h3xxx_flash_resource = Oct 28 21:51:59 DEFINE_RES_MEM(SA1100_CS0_PHYS, SZ_32M); Oct 28 21:52:34 So does Jornada Oct 28 21:52:49 CS=_PHYS is 0x0 on collie afaik Oct 28 21:52:54 CS0 Oct 28 21:53:30 I'0m sure you can better compare and sneak out bugs ;) Oct 28 22:01:38 Hmm. Very strange CS0 is 0x0, so it is flash. Oct 28 22:02:23 But I completely fail to understand 1) Why it is 32M? 2) Where comes second flash from? Oct 28 22:04:15 maybe the mtd-ram ? Oct 28 22:04:17 CS1 goes into locomo (main_ga) Oct 28 22:04:22 no-no-no Oct 28 22:04:54 mtd-ram means that you devote part of RAM (RAM is at physical 0xc0000000 - ...) to behave like mtd Oct 28 22:05:05 But it can not change physical memory maps. Oct 28 22:05:29 ok, I never used that Oct 28 22:06:11 fwiw Oct 28 22:06:13 root@collie:~# cat /proc/iomem Oct 28 22:06:13 00000000-01ffffff : sa1100-mtd Oct 28 22:06:13 20000000-2fffffff : PCMCIA socket 0 Oct 28 22:06:13 20000000-23ffffff : io Oct 28 22:06:13 28000000-2bffffff : attribute Oct 28 22:06:14 2c000000-2fffffff : memory Oct 28 22:06:16 40000000-40001fff : locomo.0 Oct 28 22:06:18 40800000-40800fff : sharp-scoop Oct 28 22:06:20 80000000-8000ffff : sa11x0-udc Oct 28 22:06:22 80010000-8001ffff : sa11x0-uart.1 Oct 28 22:06:24 80010000-80010023 : sa11x0-uart Oct 28 22:06:26 80050000-8005ffff : sa11x0-uart.3 Oct 28 22:06:28 80050000-80050023 : sa11x0-uart Oct 28 22:06:32 80060000-8006ffff : sa11x0-mcp Oct 28 22:06:34 80060000-8006ffff : sa11x0-mcp Oct 28 22:06:36 80070000-8007ffff : sa11x0-ssp Oct 28 22:06:38 Ah-ha. Oct 28 22:06:38 90010000-9001003f : sa1100-rtc Oct 28 22:06:40 90050000-9005ffff : irqs Oct 28 22:06:42 90060030-90060033 : sa11x0-mcp Oct 28 22:06:44 90060030-90060033 : sa11x0-mcp Oct 28 22:06:46 b0000000-b00000bf : sa11x0-dma Oct 28 22:06:48 b0100000-b010ffff : sa11x0-fb Oct 28 22:06:50 b0100000-b010ffff : LCD Oct 28 22:06:52 c0000000-c3ffffff : System RAM Oct 28 22:06:54 c0008000-c080dc17 : Kernel code Oct 28 22:06:56 c087c000-c0966cff : Kernel data Oct 28 22:06:57 #if defined(CONFIG_COLLIE_TS) || defined(CONFIG_COLLIE_TR0) || \ Oct 28 22:06:57 defined(CONFIG_COLLIE_TR1) || defined(CONFIG_COLLIE_DEV) Oct 28 22:06:57 SET_BANK( 0, 0xc0000000, 32*1024*1024 ); Oct 28 22:06:57 #else Oct 28 22:06:57 SET_BANK( 0, 0xc0000000, 64*1024*1024 ); Oct 28 22:06:58 #endif Oct 28 22:06:58 c4848040-c484804f : locomokbd Oct 28 22:07:02 root@collie:~# Oct 28 22:07:20 those had 2x Oct 28 22:07:45 No, this is about RAM> Oct 28 22:07:54 developer models 32 Mb, normal one - 64 Oct 28 22:08:02 Next part is funnier. Oct 28 22:08:05 ok, dev vs. production I meant Oct 28 22:08:12 #ifdef CONFIG_COLLIE_TS Oct 28 22:08:12 #ifdef CONFIG_XIP_ROM Oct 28 22:08:13 { 0xe8000000, 0x08000000, 0x02000000, DOMAIN_IO, 1, 1, 1, 0 }, /* 32M main flash (cs1) */ Oct 28 22:08:13 #else Oct 28 22:08:13 { 0xe8000000, 0x08000000, 0x02000000, DOMAIN_IO, 1, 1, 0, 0 }, /* 32M main flash (cs1) */ Oct 28 22:08:13 #endif Oct 28 22:08:17 { 0xea000000, 0x00000000, 0x02000000, DOMAIN_IO, 1, 1, 1, 0 }, /* 32M boot flash (cs0) */ Oct 28 22:08:19 { 0xec000000, 0x10000000, 0x02000000, DOMAIN_IO, 1, 1, 0, 0 }, /* 32M telios flash (cs2) */ Oct 28 22:08:21 { 0xf2000000, 0x18000000, 0x00001000, DOMAIN_IO, 1, 1, 0, 0 }, /* 4kb debug LED (cs3) */ Oct 28 22:08:23 { 0xf3000000, 0x48000000, 0x00001000, DOMAIN_IO, 1, 1, 0, 0 }, /* 4kb CF buf ctrl (cs5) */ Oct 28 22:08:25 #else Oct 28 22:08:27 { 0xe8000000, 0x00000000, 0x02000000, DOMAIN_IO, 1, 1, 0, 0 }, /* 32M main flash (cs0) */ Oct 28 22:08:29 { 0xea000000, 0x08000000, 0x02000000, DOMAIN_IO, 1, 1, 0, 0 }, /* 32M boot flash (cs1) */ Oct 28 22:08:31 { 0xf2000000, 0xC2000000, 0x02000000, DOMAIN_IO, 1, 1, 0, 0 }, /* 32M Special Case */ Oct 28 22:08:33 #endif Oct 28 22:08:35 { 0xf0000000, 0x40000000, 0x01000000, DOMAIN_IO, 1, 1, 0, 0 }, /* 16M LOCOMO & SCOOP (cs4) */ Oct 28 22:08:53 Which nails down to: Oct 28 22:10:29 32Mb main flash at 0x0 (cs0), 32Mb boot flash at cs1, cs2, cs3, cs5 not used. cs4 - locomo + scoop Oct 28 22:11:16 But the schematics doesn't have any special 'boot flash'. Oct 28 22:11:37 maybe that boot flash was on a daughterboard for devs Oct 28 22:12:07 some external part, because there is no place on the pcb... Oct 28 22:12:16 afais Oct 28 22:12:17 This is not a devs board config. Oct 28 22:12:37 Or at least it is used for production 5500 Oct 28 22:13:48 here all the pics of disassembled 5500 Oct 28 22:13:50 http://www.tekprosystems.com/zaurus/internals.htm Oct 28 22:15:01 locomo detects rom width and reports that back to sa1110 via romsel pin Oct 28 22:15:06 which is sampled during reset Oct 28 22:15:13 to select 32-bit or 16-bit bus access Oct 28 22:16:00 how does it detect? the chips seems not having the #BYTE pin Oct 28 22:17:33 but on Service MAn on page 6 it is represented...not grounded, high I'd say Oct 28 22:18:49 It should just check for a presence of a chip on D16-D32. Oct 28 22:19:06 That is the easiest way Oct 28 22:20:08 Either pins are floating, or not. This can be easily implemented in hw. Oct 28 22:23:29 hm, at page 29 they say ROMP capacity Oct 28 22:23:29 (x32 bus fixed) Oct 28 22:23:47 nCS0 3 16MB Oct 28 22:23:47 GA Oct 28 22:23:47 CLK64fs Oct 28 22:23:47 SCLK MCLK Oct 28 22:23:47 SCK Oct 28 22:23:48 SFRM Oct 28 22:23:50 LRCLK Oct 28 22:23:52 SDOUT Oct 28 22:23:54 nCS1 3 Not Use Oct 28 22:23:56 oops Oct 28 22:24:04 see, it lists the cs as you said Oct 28 22:25:35 yes Oct 28 22:26:00 just with 16MB on 5500 Oct 28 22:26:30 that's the 'ghost flash' detected by jedec_probe Oct 28 22:26:49 it doesn't harm anyway Oct 28 22:27:27 Yes. Oct 28 22:27:33 sa1100-0: Found different chip or no chip at all (mfr 0x1001, id 0x2827) at 0x1000000 Oct 28 22:27:49 However maybe changing DEFINE_RES_MEM(SA1100_CS0_PHYS, SZ_32M), Oct 28 22:27:52 interestingly the values change on every read Oct 28 22:28:04 to reference 16M can help cfi/whatever Oct 28 22:28:27 I don't think the 32MB version was ever sold Oct 28 22:28:33 so yes Oct 28 22:28:53 that is why it talks about "no chip at all" - it just reads static charge on the pins at that time. Oct 28 22:29:15 one variant was Oct 28 22:29:16 sa1100-0: Found different chip or no chip at all (mfr 0x0, id 0x3001) at 0x1000000 Oct 28 22:29:23 so it is static... Oct 28 22:29:43 static means static electricity - some results from previous "transaction" Oct 28 22:29:46 on data bus Oct 28 22:30:36 ok Oct 28 22:35:58 that's wht reset ix 0xf0 (or 0xff on some Intel) so you have a plane lead for the waveform Oct 28 22:36:12 (I humbly guess ;) Oct 28 22:39:23 btw, what about sound on collie? Does it work at all? Oct 28 22:39:47 tbh I did not test sound on any Z since long... Oct 28 22:39:55 too many other issues ;) Oct 28 22:40:22 ok Oct 28 22:40:31 now OE metadata is rather stable...we'll concentrate on userspace Oct 28 22:40:42 I was just looking onto ucb and mcp drivers.\ Oct 28 22:41:06 ah, yes, I've sent these little fixes Oct 28 22:41:34 My tose is controlled by opie Oct 28 22:41:44 at least all compiles w/out warnings Oct 28 22:41:53 :) Oct 28 22:42:05 (apart some well known introduced by Yocto :p) Oct 28 22:42:35 I promised to bluelightning to test opie soon Oct 28 22:42:43 * ant_home hides Oct 28 22:43:25 lumag_, we /I messed up with the modules so much is built in kernel but many modules could be missing Oct 28 22:43:38 for sound especially Oct 28 22:44:11 atm there is any module-list to autoload for Zaurus, we have to fix that Oct 28 22:44:36 the modules are built, however Oct 28 22:44:38 ant_home, I found kexec for 2.4 kernels, so I might have a chance to also look onto collie. Oct 28 22:45:02 ah, great Oct 28 22:45:05 If everything is correct (cross your fingers) you just have to install those modules and kernel will autoload them. Oct 28 22:45:18 First priority is backlight though. Oct 28 22:45:38 Oh, by the way. Oct 28 22:45:47 bluelightning, could you please merge https://gitorious.org/opie/opie/merge_requests/3 Oct 28 22:46:59 backlight on tosa is really brain damaged. Oct 28 22:47:29 You should turn on power in 3 different drivers, before dac chip stops grounding i2c bus Oct 28 22:53:39 And to complete crazyness those 3 drivers also have some dependencies on the order. Oct 28 22:54:15 lumag_: merged, thanks Oct 28 22:54:25 bluelightning, :) Oct 28 23:05:20 ant_home, what is your collie userspace? Oct 28 23:05:25 command line ? Oct 28 23:05:27 bluelightning,: I've sent now the last round of jffs2 fixes Oct 28 23:06:36 lumag_, root=/dev/sda1 rootfstype=ext2 rootwait fbcon=rotate:1 console=ttySA0,115200n8 console=tty1 mem=64M fbcon=rotate:1 debug Oct 28 23:06:52 this is done by kexecboot ofc Oct 28 23:10:54 bluelightning, argh. ..typo Oct 28 23:10:55 ant_home, No I mean what userspace do you use on collie Oct 28 23:11:25 bluelightning, can you fix the last line, for simpad? Oct 28 23:11:58 .jffs2 -> _jffs2 +EXTRA_IMAGECMD_jffs2 Oct 28 23:12:41 lumag_, I'm still in console (core-image-base) Oct 28 23:12:51 framebuffer is unusable atm Oct 28 23:13:01 Hmm. Oct 28 23:13:09 I can take a pic Oct 28 23:13:15 This sounds more and more interesting. Oct 28 23:13:18 seems driven with wrong freq Oct 28 23:13:34 wow, ppl, you rocks :) Oct 28 23:13:36 Check if comadj is programmed correctly Oct 28 23:13:53 deep dive into collie architecture :) Oct 28 23:13:56 and I don't remember if it has phase adjust Oct 28 23:14:09 Jay7, hello Oct 28 23:14:21 hi, lumag_, ant_home, bluelightning Oct 28 23:14:32 lumag: is ok with sharp flash AND with kexecboot 2.6.31 Oct 28 23:14:45 is broken after kexec Oct 28 23:16:08 Jay7, hi Oct 28 23:16:38 ant_home, Probably kexec trashes sharpsl params structure. Oct 28 23:16:43 Jay7, my kid is now at school (1st year) so he get to sleep earlier and I have a bit of time ;) Oct 28 23:17:08 We should add a way to pass them through command line args. Oct 28 23:17:32 procfs (or sysfs) interface for reading them Oct 28 23:17:43 lumag_, hm.. I've sent that last patch for sharpsl_param Oct 28 23:17:52 params are passed on pxa Oct 28 23:18:01 (on a virtual address) Oct 28 23:18:05 Yes. Oct 28 23:18:08 on sa1100 on phys address Oct 28 23:18:16 I wander if kernel/userspace overrides that address. Oct 28 23:18:25 on sa1100 you mean? Oct 28 23:18:31 Add a few checks to sharpsl_params Oct 28 23:18:32 yes Oct 28 23:19:23 s/checks/debug prints/ Oct 28 23:20:29 bluelightning, ok, sent v2 Oct 28 23:20:50 OK, it's 3:20 am already and I will get up at 7:10-7:15, Oct 28 23:21:07 lumag, afair there was one message like 'the logo is bigger than fb' Oct 28 23:21:19 lumag_: omg Oct 28 23:21:21 goodbye Oct 28 23:21:23 gn Oct 28 23:21:26 thx thx Oct 28 23:21:30 ant_home, when? During bootup? Oct 28 23:21:41 that is also strange... Oct 28 23:21:49 yes, don't remember with kernel though Oct 28 23:21:54 which Oct 28 23:21:58 let you know Oct 28 23:22:13 don't bother. I never tried 2.6/3.x on my collie. Oct 28 23:22:42 kexecboot site has a 'good' 2.6.31, I used that Oct 28 23:22:50 I used it as a small FBreader till battery could not hold long enough. Oct 28 23:23:05 Are any good battery resellers on eBay? Oct 28 23:23:21 I tried one, but looks like his batteries are even more dead. Oct 28 23:23:40 heh, my readings are 0v ... Oct 28 23:23:50 Or I just should build an adapter for nokia batteries. Oct 28 23:23:50 hopefully is just the driver Oct 28 23:25:08 What should be next after we squeeze all bugs in collie kernel? Oct 28 23:25:13 SL-A300? ;) Oct 28 23:27:54 Sharp IS01 Snapdragon-powered ? hard to find... Oct 28 23:29:39 http://ezaurus.com/lineup/sl/sla300/sla300_spec.html Oct 28 23:30:29 ah, that's really old stuff :/ Oct 28 23:30:55 bluelightning, I was porting libsdl-qpe patches to current sdl and dropped all the hacks for A300. I hope nobody will notice that? Oct 28 23:35:31 ok, night Oct 28 23:36:10 'nite Oct 28 23:37:29 lumag_: I don't think anyone will mind :) Oct 28 23:37:32 night **** ENDING LOGGING AT Tue Oct 29 02:59:58 2013