irclog2html for blob on 2001.07.23

01:51:24sammythat's greet ....!! what's amazing ....!!
01:56:19sammyRuss : BZFlag == Tim Riker .... ?
01:56:24Russyah
01:57:25sammynow I am watching the page about his little child , so cute ....!!
08:17:24erikmgood morning
08:17:43sammyGood Morning erikm ...!!
08:38:58sammyhi !! lxrbot good to see U again ....!!
08:40:18erikmlxrbot: have a botsnack
08:40:19lxrbotthanks erikm :)
11:35:52erikmwoohoo! blob correctly detects the 64MB expansion board:
11:35:55erikmMemory map:                                                  
11:35:55erikm  0x00800000 (8 MB) @ 0xC0000000                              
11:35:55erikm  0x00800000 (8 MB) @ 0xC1000000                              
11:35:55erikm  0x00800000 (8 MB) @ 0xC8000000                              
11:35:55erikm  0x00800000 (8 MB) @ 0xC9000000                              
11:35:55erikm  0x04000000 (64 MB) @ 0xD0000000              
13:24:37Russerikm: yah
13:25:06RussI was all set to write that so I could figure out the tuxscreen memory may
13:25:25Russerikm: the ramdisks are messed up and I don't know why though
13:26:33Russjust hangs at decompressing ramdisk
13:29:04erikmhi Russ
13:31:32erikmRuss: hmm. is the memory setup correct?
13:31:48erikmRuss: look at the brutus setup, it also has 4MB in each bank
13:44:09RussMemory map:                                                        
13:44:09Russ  0x00400000 (4 MB) @ 0xC0000000                                    
13:44:09Russ  0x00400000 (4 MB) @ 0xC8000000                                    
13:44:09Russ  0x00400000 (4 MB) @ 0xD0000000                                    
13:44:09Russ  0x00400000 (4 MB) @ 0xD8000000                                    
13:44:29Russ        /* The Shannon has 2 banks of 4M on board */
13:44:30Russ        SET_BANK( 0, 0xc0000000, 4*1024*1024 );
13:44:30Russ        SET_BANK( 1, 0xc8000000, 4*1024*1024 );
13:44:30Russ        /* And an add-in card with 2 banks of 4M */
13:44:30Russ        SET_BANK( 2, 0xd0000000, 4*1024*1024 );
13:44:30Russ        SET_BANK( 3, 0xd8000000, 4*1024*1024 );
13:44:32Russ        /* Set to 2 if you don't have the RAM card in */
13:44:35Russ        mi->nr_banks = 4;
13:44:43Russ        ROOT_DEV = MKDEV(RAMDISK_MAJOR,0);
13:44:43Russ        setup_ramdisk(1, 0, 0, 8192);
13:44:43Russ        setup_initrd(0xc8000000, 3*1024*1024);
13:45:05RussBOOT_MEM(0xc0000000, 0x80000000, 0xf8000000)
13:45:45Russthe brutus setup differs slightly
13:45:52Russsetup_initrd( __phys_to_virt(0xd8000000), 3*1024*1024 );
13:46:05Russwhy is the __phy_to_virt there?
13:46:46erikmthat's what I asked myself as well
13:47:39erikmwait. it just hangs at compressing the ramdisk?
13:47:51RussRAMDISK: Compressed image found at block 0
13:48:06erikmtry to relax the memory timings
13:48:25Russits already pretty relaxed
13:48:32erikmthe decompress stuff stresses the memory bus quite seerely
13:48:52Russthe kernel decompresses fine
13:49:15Russbtw, I should use ttySA0, correct?
13:49:36erikmI think so yes
13:49:38erikmhmm, wait
13:49:41erikmthat can be it
13:49:58erikmif you're using the wrong tty you won't see anuthing at all and the system appears to hang
13:50:19Russcould just all be puts
13:50:27Russ(maybe)
13:50:51RussMemory: 11720KB available (1003K code, 219K data, 44K init)
13:50:54Russ...
13:51:04Russplenty for an 8M ramdisk
13:52:34Russand sysrq doesn't want to work
13:52:58erikmwhat kernel?
13:53:08Russ2.4.6-rmk2-np1
13:53:24erikmtries sysrq on LART
13:53:31erikmworks
13:53:53Russmaybe whatever happens to the ramdisk is so bad, sysrq breaks
13:54:02erikmmaybe it disabled sysrq
13:54:46Russhmm.....I had it as console=ttySA0,9600
13:55:04RussI changed it to console=/dev/ttySA0,9600, and it stopped working
13:55:30erikmshould be console=ttySA0,9600
13:56:24Russanyway, it gets stuck after the call to gunzip
14:00:13erikmbtw, don't include linux/irq.h
14:00:45erikm(I'm compiling the cs89x0 driver)
14:01:21Russany paticular reason?
14:01:44erikmlinux/irq.h is only for the code in arch/*
14:01:51erikmand ARM just doesn't use it
14:02:00erikmthe header file is never used in drivers
14:02:57Russleaves
14:04:24erikmhohum. kernel oops
14:18:51erikm#ifdef SA1100_ARCH
14:18:51erikmioaddr += PCMCIA_IO_0_BASE;
14:18:51erikm#endif
14:19:00erikmugly hack, Russ.
14:19:24erikmjust tell the driver that you *want* 0xe0008300
16:55:04erikmhi BZFlag
17:14:22BZFlaghey. been a busy weekend.
17:14:30BZFlaggot 4 new Yopys to play with.
17:14:51BZFlaghacked friends systems for fun and pleasure. ;-)
17:15:50erikmhehe
17:15:59erikmpretty lazy weekend over here
17:16:41erikmread lots of newspapers, visited a couple of friends, and decided to do some blob hacking last night
17:18:10BZFlagsend me your mail address and I'll "loan" you a couple of phones to hack on?
17:18:25erikmBZFlag: *blink*
17:18:56erikmBZFlag: it's not that I don't want to pay for it or don't have the money, it's just quite hard to get money to the US
17:19:46BZFlagI know, I'd just like you to get started as soon as possible. ;-) I figure the old saying:
17:20:03BZFlagwhatever is on a linux engineer's desk, will soon run linux.
17:20:41BZFlagI figure I should not let money transfer issues get in the way.
17:20:57BZFlagand this way I can mark the boxes as "gifts" too. ;-)
17:21:28erikmah, ok
17:21:28BZFlagRuss and I have been hacking, I'd just like to get you started too.
17:21:51BZFlagso let's dig in and figure the money out when we can.
17:22:01erikmok, I'll email my address
17:22:04BZFlagthe sooner linux is up, the sooner I get my basement back.
17:22:08erikmhehe
17:22:22BZFlagsweet. tim@rikers.org, but you knew that.
17:26:43erikmlooks for the correct zip code of his work address
17:28:17erikmfound it
17:28:45erikmwork address is easier, otherwise the parcel service will just find a closed door at home ;)
17:29:40BZFlagno prob, whatever you like.
17:30:05erikmmail is on its way
17:38:15BZFlaggot it. 2 units ok? you have stuff for cabling and dongles I presume?
17:38:50BZFlagI figure you want one with inferno still on it and one for blasting blob/linux so you can compare register settings etc.
17:39:04Russ was last seen on #blob 3 hours, 36 minutes and 7 seconds ago, saying: leaves [Mon Jul 23 15:02:57 2001]
17:39:04BZFlagibot: seen Russ
17:40:01erikmBZFlag: sorry, was to the vending machine
17:40:25BZFlagno prob. /me is talking on like 6 channels now.
17:40:50erikmBZFlag: one with inferno on it would be nice, yes.
17:41:08erikmBZFlag: and we have enough stuff over here to make *any* dongle you want :)
17:41:54BZFlagthey will both have inferno, I figure you will jtag and blast one of them till blob is stable anyway. I have like 4 different inferno images too if you want to play with that.
17:42:18erikmI actually never played with inferno
17:42:39BZFlagsweet, I'll get em mailed tomorrow then. probably usps gem (us postal service, global express mail)
17:43:15BZFlaginferno is nice, but I do not have full source, and would much prefer a complete open source solution.
17:43:22erikmok. so they will be here somewhere end of the week or so?
17:44:06BZFlagI think they say like 2-4 working days. I mail ya after I send em with a better estimate.
17:44:12erikmok
18:35:34Russ|werkBZFlag: did you notice blob now works and flashes?
18:35:46Russ|werkand I also have kernel diffs up?
18:36:25erikmRuss|werk: your IRQ autoprobing in the cs89x0 driver can't work
18:36:32Russ|werkerikm: could you change the mach-types on arm.linux.org.uk for tuxscreen from ARCH_SHANNON to SA1100_SHANNON?
18:36:35Russ|werkerikm: why not?
18:36:45erikmRuss|werk: no, I can't. ask rmk
18:37:43erikmRuss|werk: because you first need to make the GPIO an input and make it a rising edge sensitive IRQ
18:38:02Russ|werkright, and I do that for each listed possible IRQ
18:38:45erikmoh, wait that gets compiled
18:38:52Russ|werkerikm: how does the kernel know where to decompress to?
18:39:13erikmRuss|werk: rmk just posted some nice story about that on linux-arm-kernel
18:39:25Russ|werkerikm: the adding of PCMCIA_IO_0_BASE came from the diffs you passed me
18:40:06erikmoh, that's why. I already changed that in my version
18:40:39Russ|werkeasy to fix
18:40:55erikmI already did so
18:41:08erikmhmm, the driver doesn't listen to the irq= argument
18:41:17Russ|werkI just figured there might be a good reason for it, like the networking subsystem not being u32 for ioport in all places
18:41:20erikmcs89x0: No EEPROM, relying on command line....
18:41:21erikm IRQ 892627570, programmed I/O, MAC 00:00:00:00:00:00
18:41:54erikmthe IRQ part scares me
18:42:02Russ|werkerikm: I don't know if I've even specified an IRQ...
18:42:17Russ|werkerikm: I always just let it probe one
18:42:39erikmRuss|werk: that doesn't work, so I just said irq=0, and that gives the same results
18:43:31Russ|werk{ IRQ_GPIO0, IRQ_GPIO1, IRQ_GPIO2, IRQ_GPIO3, 0 };
18:43:37Russ|werkstill 0 terminated?
18:43:52erikmand IRQ_GPIO0 == 0
18:44:03Russ|werkyou sure its not ~0 or something?
18:44:33erikm#define SA1100_IRQ(x)(0 + (x))
18:44:33erikm#defineIRQ_GPIO0SA1100_IRQ(0)
18:45:20erikmso yes, IRQ_GPIO0 == 0
18:45:39erikmI'll make it an -1 terminated list
18:45:50Russ|werkgood idea
18:46:47erikmactually, the IRQ list is a little bit different from what you think it is
18:47:23erikmit is not a list of IRQs for all cards, but a list of four IRQs for a single card
18:47:23Russ|werkI know
18:47:48Russ|werkIts completely different from teh with eeprom setup
18:48:07erikmso just making it count from 0 to 4 should also fix it
18:49:02Russ|werkI was at the time thinking stop after 2 chips, so I added the zero, then the lart config just extended on that
18:49:52erikmhmm
18:50:00erikmwhatever we do, it will be broken
18:50:39Russ|werkIt is possible that the embedded cs89x0 driver should be pulled off and seperate from the ISA card cs89x0 driver
18:51:17Russ|werkI do like the mii-tool hack though
18:51:42Russ|werkwhere'd BZFlag go?
18:52:01erikmchatting on other channels ;)
18:58:15Russ|werkerikm: news on rmk's site?
18:58:22Russ|werkthe demise of rebel.com?
18:58:48erikmRuss|werk: yeah, I knew about that
19:08:43Russ|werkerikm: how does the kernel know where to decompress to?
19:10:21BZFlagRuss|werk: I'm here now. you booted!?
19:10:28Russ|werkwonders why someone always complains when their bogomips value drops
19:10:40Russ|werkBZFlag: it gets stuck at decompressing ramdisk
19:10:56Russ|werkBZFlag: but blob is fully functional, and I have kernel patches
19:11:38erikmRuss|werk: rmk just posted some nice story about that on linux-arm-kernel
19:11:48Russ|werkthe first problem I noticed was that kernel start was at 0xc0008000
19:11:52BZFlagsweet! reg blob or ts blob?
19:11:59Russ|werkand length was 0xf8000
19:12:16Russ|werkthen the stack was put at 0xc0010000 - 4....
19:12:23Russ|werkBZFlag: regular blob
19:12:24erikmRuss|werk: it tells exactly where the kernel decompresses ramdisks, etc
19:12:29Russ|werkBZFlag: no code from AMD
19:12:45Russ|werkerikm: still getting through my email box
19:12:51BZFlagnice! in cvs now?
19:12:59Russ|werkyup
19:13:08erikmthinks BZFlag should subscribe to blob-cvs-commit
19:13:17erikmBZFlag: it has been a busy weekend for blob
19:13:20Russ|werkthen I noticed that BLOB_RAM_BASE was at 0xc0000000 + 16M...
19:13:28BZFlaghehe, I suppose I could procmail it... ;-)
19:13:34erikmBZFlag: it also has memory probing
19:13:42Russ|werkthen, erikm commited some memory map detection code, and I saw the memory map
19:13:50Russ|werkwas going to do that code next
19:13:56BZFlagyou two are awesome!
19:14:08erikmBZFlag: here is your procmail rule:
19:14:09erikm:0:
19:14:10erikm* ^X-BeenThere:\ blob-cvs-commit@lists\.sourceforge\.net
19:14:10erikmblob-cvs-commit
19:14:13Russ|werkso I was able to move everything around to the memory map
19:14:29Russ|werkBZFlag: oh. you don't get blob-cvs-commit?
19:14:47erikmthe memory map code also correctly detects the 64MB expansion board
19:15:04erikmgets the dmm
19:16:08Russ|werkBZFlag: more work to be done on the kernel patches, keyboard, lcd, ucb, handling of a mem= paramater, etc
19:16:55Russ|werkalso, I asked rmk to change ARCH_SHANNON to SA1100_SHANNON
19:17:49BZFlagRuss|werk: whatever is the norm for naming is fine by me.
19:18:16erikmRuss|werk: ucb sound support should be easy to do
19:18:37Russ|werkerikm: I figure, I saw only two ifdef LART's in the driver
19:18:49Russ|werkit uses the same serial port on all sa1100's?
19:18:57erikmyes
19:19:31erikmRuss|werk: look in a recent kernel for drivers/char/ucb1200_sa1100.c
19:19:45erikmRuss|werk: I added LART a week ago over there
19:20:00erikmdamn. farnell gave us the wrong connectors
19:20:23erikmwants the bloody male connectors
19:20:54Russ|werkerikm: I know, I grepped the kernel source for a few of the SA1100 boards with similar hardware
19:21:02Russ|werk(2.4.6-rmk2-np1)
19:21:16erikmdamn
19:21:17Russ|werkerikm: connector to what?
19:21:19erikmaargh!
19:21:24erikmetherboard
19:21:34erikmI took the wrong example ether board!
19:21:42erikmthe connectors are wrong *again*
19:21:56erikms/connectors/resistors/
19:22:02erikmno wonder it doesn't work
19:23:15Russ|werkrouted wrong?
19:23:21Russ|werkoh, resistors
19:23:32erikmsilk screen is wrong
19:23:44Russ|werknot wrong, "misleading"
19:23:50erikmehm, yes
19:24:06erikmthe 10k and 4k99 were mixed up
19:24:13Russ|werkbut it depends on what your definition of the word 'is' is
19:27:52Russ|werkjust curious, are navaho code talkers well known outside of arizona?
19:28:23erikmnot really. never heard about them over here
19:30:00Russ|werkthe US military used indians as code talkers in the pacific theatre, because all other forms of code were quickly broken
19:30:11Russ|werknavaho was never broken
19:31:12erikmah, yes, I remember
19:32:09Russ|werkis it well known that the US kept japanese citizens in camps?
19:32:40BZFlagwonders why the ww2 conversation all of a sudden?
19:33:04erikmRuss|werk: japanese citizens? or US residents of japanese origin?
19:33:06Russ|werkdunno, there is that codetalker movie coming out, or was that a tv movie I dunno
19:33:10Russ|werker
19:33:29Russ|werkyes, US citizens of japanese orign
19:34:44Russ|werkprints rmk's faq and starts using it as wallpaper
19:35:36Russ|werkso __virt_to_phys does nothing if your physical start of ram and virtual start of ram is the same adress
19:35:41Russ|werk(which is the case)
19:38:09erikmthat also explains why some archs use __virt_to_phys() for the base address and others don't
19:38:17Russ|werkerikm: that guy asking about the assabetized lart is the sameone that rewrote my patches in "c++"
19:38:31erikmpatches for what?
19:39:24Russ|werkjffs2, he changed all the coding styles, and changed all the pre/post incremement/decremnts to
19:39:28Russ|werki = i + 1
19:39:36erikmah, that one
19:39:45Russ|werkand made no distinction between pre and post
19:40:01Russ|werkand then emailed me saying he found a bug in my code
19:40:49erikmproblem over here is that the students only learn C++. and doing low-level UNIX things requires quite some C knowledge. *very* steep learning curve ahead
19:42:16Russ|werkthe comp.os.c faq is very handy for such students
19:43:03Russ|werkc++ is being replaced by java in schools
19:44:04erikmhmm, I'll point them to it next time
19:44:08erikmgood idea
19:44:22erikmthey usually ask me because I am the C guru over here
19:44:30erikmand the unix guru
19:45:01Russ|werkthe gcc.gnu.org manuals are also helpfull, as well as the glibc manual/info pages
19:55:15Russ|werkhow come the strongarm processor I have isn't as pretty as the ones who have
19:55:42Russ|werkmine has very small, barely visable lettering
19:56:01Russ|werkthe ones I see in other pictures have big bright white letters...
20:01:03Russ|werkok, machine registry is updated
20:04:46erikmgood
20:05:27Russ|werksuppose I should hand out cvs write access to my kernel?
20:05:34Russ|werkor does that not work well with kernels
20:06:22erikmRuss|werk: what do you mean?
20:06:41Russ|werkI have a cvs kernel that I make my diffs from
20:06:49Russ|werkfor the tuxscreen
20:07:04Russ|werkI suppose submitting patches is the traditional method
20:07:05erikmand you want to give us write access to it?
20:07:11Russ|werkyad
20:07:12Russ|werker
20:07:13Russ|werkyah
20:07:33erikmRuss|werk: well, rmk and nico are certainly not going to pull patches from CVS trees
20:07:46Russ|werk?
20:08:01Russ|werkI use the tree to make -tux patches
20:08:05erikmRuss|werk: rmk does for two CVS trees: cpufreq and mtd
20:08:13Russ|werkat -rmk2-np1-tux1 right now
20:08:37erikmRuss|werk: don't make it as ugly as the handhelds CVS tree...
20:08:48Russ|werkhows that?
20:09:04erikmhttp://www.arm.linux.org.uk/developer/patches/?action=viewpatch&id=597/1
20:09:53erikmyes, that's a PCMCIA driver with two integrated MTD drivers
20:19:27Russ|werkadds || defined NESA to the arm flash code
20:20:56erikmethernet works *much* better with the resistors on the correct place
20:21:13BZFlagRuss: so I never heard what the jtag issue was on your end.
20:21:16erikmapt-get upgrades the LART
20:23:47Russ|werkBZFlag: still dunno
20:24:01Russ|werkBZFlag: got it working with jflash though
20:26:43Russ|werknow I don't need jtag
20:27:03erikmRuss|werk: because blob flashes itself ;)
20:27:09Russ|werkat least until I screw up blob
20:27:24Russ|werkthe jumping to a downloaded blob is a cool idea
20:28:00Russ|werkthe startup.S stuff would have to be position independant
20:28:09erikmRuss|werk: *very* hard
20:28:17erikmRuss|werk: I tried it
20:28:23Russ|werker, start.S
20:28:33Russ|werkthe rest of the stuff wouldn't have to be
20:28:49Russ|werkbecause start.S would recopy everything to the correct location
20:28:59erikmRuss|werk: jus make a separate start.S that copies blob-rest to 0xc0000000
20:30:39Russ|werkbut then you aren't testing start.S
20:30:57erikmno, but usually you don't change that
20:31:13erikmand when you change it you want to test it for a bare system
20:31:18Russ|werkbut you might...
20:31:44erikmchanging memory timings, for example. so you would need to run from flash for that
20:32:22Russ|werkstart.S changes the memory timings
20:32:54erikmyes, but currently you would only like to test things like JFFS2 support, which is in blob-rest
20:34:02Russ|werkso having a start.S appended to the end would work for that
20:34:21Russ|werkjump to BLOB_RAM_BASE+the offset
20:34:34Russ|werkcopy it back, and jump to it
20:34:35erikmfor example, yes
20:34:51erikmI'll make it tomorrow
20:34:57Russ|werkI still think making start.S position independant would be better
20:35:05erikmRuss|werk: been there, done that/
20:35:10erikmRuss|werk: it just doesn't work
20:35:22Russ|werklike the memory test?
20:35:30Russ|werkoops, tested blob
20:35:34erikmwhat about the memory test?
20:35:44Russ|werkthe memory testing stuff would test blob
20:36:24erikmno, that's not what's wrong. you just can't get it to work. the manual says "use -fPIC" etc. but the point is that you need to have an ELF capable run-time linke
20:36:26erikmare
20:36:46Russ|werkI don't know arm code that well
20:36:47erikmand I refuse to put an ELF linker into a boot loader
20:37:02Russ|werkI could do it in m6800
20:37:19erikmit's not worth it for debugging code
20:37:37erikmI'll just make some blob-debug-start code
20:37:38Russ|werkskims through his arm quick reference sheet
20:38:31Russ|werkI can see doing blob-debug-start if only for the needed memory test special case
20:39:01Russ|werkother schemes invlove enabling the MMU and making it think it is at 0x00000000
20:39:15Russ|werkbut that leads to other problems
20:40:01erikmturning MMU is a big nono for me. it is a boot loader, not an OS
20:40:21erikmanyway, debug-start.S will do this:
20:40:30erikm- it turns on the LED
20:40:32Russ|werkI suppose if a user is hacking on start.S, they would have JTAG hardware
20:40:47erikm- it zeros the first 1MB
20:40:56erikm- it copies blob-rest to 0xc0000000
20:41:02erikm- it sets up a stack pointer
20:41:10erikm- it jumps to 0xc0000000
20:41:20erikmI think it even won't turn on the LED
20:41:27Russ|werkthe only question is where do you put it in blob
20:41:35Russ|werkturn off the LED before you jump to it
20:41:36erikmRuss|werk: we just don't
20:41:47Russ|werkthat way its easy to detect a bogus download
20:41:56erikmRuss|werk: it's a separate download target
20:42:04Russ|werkhmm....
20:42:33Russ|werkI dunno about that, I think it can be in the same target
20:42:35erikmmaybe even only enabled with an extra configure flag
20:42:46erikmRuss|werk: it's dead code for most situations
20:43:00erikmRuss|werk: it's only nice for developers, not for average users
20:43:06Russ|werkhell, it could possibly be c code, since it doesn't have to return
20:43:23Russ|werker, but c code can't quite setup a stack pointer
20:45:05Russ|werkdon't forget you can pass addresses to it in registers
20:45:28erikmRuss|werk: assembly is the easiest way.
20:45:40erikmRuss|werk: I'll work on it tomorrow morning
20:56:42erikmgoes home
20:56:46erikmsee you tomorrow

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with infobot logs, split per channel and by date, etc.