irclog2html for #openjtag on 20050516

00:05.32*** join/#openjtag beewoolie (~beewoolie@florence.buici.com)
00:06.15*** join/#openjtag beewoolie (~beewoolie@florence.buici.com)
01:16.24*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
01:16.24*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
02:28.47*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
02:28.48*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
03:03.20*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
03:03.21*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
03:48.27*** join/#openjtag beewoolie-afk (~beewoolie@florence.buici.com)
10:36.18*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
10:36.18*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
10:58.02*** join/#openjtag ChanServ (ChanServ@services.)
10:58.02*** mode/#openjtag [+o ChanServ] by irc.freenode.net
11:22.39*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
11:22.39*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
11:32.43*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
11:32.43*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
12:08.19*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
12:08.19*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
13:20.40*** join/#openjtag prpplague (~billybob@72.22.134.184)
16:18.44*** join/#openjtag [g2] (~g2@g2.nslu2-linux)
16:47.58*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
16:47.58*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
18:11.53*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
18:11.53*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
18:57.02*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
18:57.02*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
19:00.29beewoolie-afk[g2]: Perhaps we should discuss the layout changes here?
19:01.03[g2]sure here is fine with me
19:01.15[g2]I'm sure at some point the openslug guys will be interested
19:01.22beewoolie-afkIndeed.
19:01.25[g2]but the crew is all here
19:01.32[g2]you, me, jacques and dyoung-zzzz
19:01.40beewoolie-afkI just wanted to get an idea of what you need without the cross-talk.
19:01.47[g2]and possibly ka6sox-away
19:02.04[g2]nod. It's much quieter in here
19:02.19[g2]actually I was going to change the layout of firmware image
19:02.43beewoolie-afkFlash layout is really a config issue and shouldn't be more than a custom config.
19:03.12[g2]exactly there are no changes to APEX, maybe just another custom config
19:03.21[g2]or maybe that is the default config
19:03.31beewoolie-afkBTW, did you figure out why your patch wasn't propagating?
19:03.51[g2]no I think that's an open issue
19:04.02[g2]I can test some more if you like
19:04.07[g2]it real easy
19:04.37[g2]the work-around is simple and quick enough, make clean && make
19:05.25[g2]beewoolie-afk, do you recommend having a config block for the APEX parameters ?
19:05.45beewoolie-afkwhat is a config block?
19:05.53[g2]a configuration block
19:05.57[g2]in flash
19:06.05beewoolie-afkAPEX has an environment.
19:06.11beewoolie-afkVAR=VALUE
19:06.49[g2]and the real question is making changes to that environment persistent across cold-boots
19:07.10beewoolie-afkRight.  It has an environment that it writes to flash.
19:07.38beewoolie-afkThere are default values compiled into the code.  These values can be overridden.
19:07.52[g2]Really ? does the writing ? But that's only if things are changed correct ?
19:08.02beewoolie-afkRight.
19:08.35beewoolie-afkI'm not using it at the moment on the slug because I don't want to clobber the second flash block.
19:08.38[g2]That'll need some testing and we''ll need to back up the last 144 bytes from the Redboot second block
19:08.55[g2]Ok.. that's good for now
19:09.32beewoolie-afkAt the moment, the environment is stored in the second flash block.  It could be put in the first block, after the loader.  Trouble is that there is no logic that allows for erasing the environment in a block without erasing the whole thing.
19:09.44beewoolie-afkThat was part of the issue.
19:10.08beewoolie-afkWe can certainly add some logic that pulls the MAC address from the second block.
19:10.28[g2]that's why we just have the user save off the last 144 bytes because the first 250,000 are all the same for everyone
19:11.22beewoolie-afkWe could write a piece of code that installs APEX on a runnin system s.t. it pulls the mac address automatically. and writes the first block with everything we need.
19:11.26[g2]it's a simple dd if=/dev/mdtblock0 of=redboot_save bs=1K skip=250 I think
19:11.48beewoolie-afkI'm suggesting something a little more elegant, but yeah.  I get that it's easy.
19:12.12beewoolie-afkI like having automated processes so that fumble fingers joe doesn't end up with a brick.
19:12.17[g2]Ok.. are you saying we reserve the last 144 bytes and and just uses the first 128K-144 bytes ?
19:12.41[g2]agreed on the deployablility issue
19:12.48[g2]automated is much better
19:13.20[g2]actually that could be part of a script
19:13.36[g2]an init script
19:13.38beewoolie-afkIt depends on whether or not you want to reorg flash.  If we plan to replace the bootloader, we can put everything APEX needs in one block and give the other block to the kernel...which appears to need it.
19:14.02beewoolie-afkSure.  The init script could be smart about the upgrade and only do it when necessary.
19:14.19beewoolie-afkThe presently isn't a MAGIC number in APEX, but we can add one easily enough.
19:14.52[g2]Since APEX is so good with space utilization, I'd prefer to use the second block
19:15.00[g2]as the environemnt
19:15.10beewoolie-afkFair enough.
19:15.20[g2]Then we still gain back 3 blocks
19:15.29[g2]1 for the sysconfig
19:15.31beewoolie-afkThere may come a time when APEX needs the extra space for NPE or whatever.
19:15.34beewoolie-afkRight.
19:15.41beewoolie-afkTwo others from???
19:15.44[g2]1 for the ramdisk start
19:15.49[g2]and 1 from the trailer
19:16.02beewoolie-afkThey waste a whole block for the ramdisk start marker?
19:16.22[g2]they only waste 16 bytes, but jffs2 needs to start on a block boundary :(
19:16.31[g2]which makes total sense to me
19:16.34beewoolie-afkThat 768K.  Kindofa lot.
19:16.43[g2]386K
19:16.49[g2]128 * 3
19:17.01beewoolie-afkOh.  Yeah.  I forgot that blocks are only 128k on this baby.
19:17.02[g2]382
19:17.11beewoolie-afkStill, that will help.
19:17.19[g2]absolutely
19:17.28beewoolie-afkNow, we're talking about Openslug, right?
19:17.32[g2]Right
19:17.44[g2]Do you want to support Unslung ?
19:18.06beewoolie-afkSo, this can be the upgrade plan.  Boot to existing firmware, run this script, reboot and you've got openslug.
19:18.21beewoolie-afkNo,  I don't care about unslung.  I just wanted to make sure we're talking about the same thing.
19:18.53[g2]Ok.. I need to think about OpenSlug compatibility for rwhitby
19:19.01[g2]but of the top of my head...
19:19.30[g2]The current default config runs for both Unslung and Openslug, but Unslung untested right now
19:19.47[g2]and we need to add some minor config changes like the 1 second autoboot
19:20.08[g2]well they'll both run after the minor changes...
19:20.42[g2]the second custom config is for the camp #5 guys that cut the cord with the linksyss compatibility
19:20.44beewoolie-afkI still don't get why you're so hot to lengthen the boot.
19:21.08[g2]it's the autoboot that's important
19:21.15[g2]at 1 second
19:21.35beewoolie-afkMaking it longer inconveniences everyone.  Leaving it shorter just means that people who need to break into the loader need to press the key when the power the unit on.
19:21.53[g2]1 second is too long ?
19:22.06beewoolie-afkI thought that you wanted it longer.
19:22.09[g2]no...
19:22.18[g2]I just want it *AUTO*
19:22.19beewoolie-afkOh.  The default config is already one second to autoboot.
19:22.26[g2]mine didn't
19:22.31beewoolie-afkYou must have overridden it in your config.
19:22.46beewoolie-afkThe default for nslu2 has autoboot enabled in 1.2.11.
19:22.48[g2]I'll double check but I don't think that's in 1.2.11
19:23.08beewoolie-afk#
19:23.08beewoolie-afk# Environment
19:23.08beewoolie-afk#
19:23.08beewoolie-afkCONFIG_NOR_BOOT=y
19:23.08beewoolie-afk# CONFIG_NAND_BOOT is not set
19:23.10beewoolie-afkCONFIG_AUTOBOOT=y
19:23.12beewoolie-afkCONFIG_AUTOBOOT_DELAY=10
19:23.14beewoolie-afkThat's the default.
19:23.29[g2]yes, but you must select the ixp4xx profile and that may change it
19:23.31beewoolie-afk10 as in 10 10ths.
19:23.37beewoolie-afkHuh?
19:23.45beewoolie-afk# make nslu2_config
19:23.50beewoolie-afkThat will set the defaults.
19:23.55[g2]Ah...
19:24.01beewoolie-afkJust like the kernel.
19:24.02[g2]here's the difference...
19:24.19[g2]I was going into make xconfig and just picking the ixp4xxx profile
19:24.48[g2]I like the make xxxx_config better
19:24.57beewoolie-afkBest to start with the nslu2_defaults and then make your changes.
19:25.22[g2]well I've got to change things right out of the box because my toolchain is in a different place
19:25.36beewoolie-afkah.
19:25.56[g2]can I make nslu2_default CROSS_COMPILE=.....linux-
19:25.57beewoolie-afkI think that you can use an environment variable for that.
19:26.07beewoolie-afkAt least, it is suppose to work that way.
19:26.16[g2]or CROSS_COMILE=....linux- make nslu2_default
19:26.25[g2]and it gets picked up ?
19:26.34beewoolie-afkHmm.  No, I think it needs to be defined for the build.
19:26.42beewoolie-afk#CROSS_COMPILE=... make
19:27.14[g2]well it's easier to just sed the nslu2_defaults file and make nslu2_defaults
19:28.07beewoolie-afkalternatively, we can make it possible to put the compiler in the user's path.
19:28.24beewoolie-afkcross_compile=arm-linux-  ...
19:28.30beewoolie-afkPerhaps that should be the default.
19:28.49[g2]we're splitting hairs here I think
19:29.04[g2]we just need to make it a little clearer on the setup
19:29.09[g2]I was doing it wrong
19:29.09beewoolie-afkSort of, except the cleanliness of the build is the first thing that most people see.
19:29.18[g2]sure...
19:29.18beewoolie-afkLike you.
19:29.43beewoolie-afkI'll add some notes about that in the readme.
19:36.41[g2]beewoolie-afk, I've got an env.c patch and a new default config for openslug
19:36.55[g2]then we'll have nslu2_config and openslug_config
19:36.56beewoolie-afkemail?
19:37.05beewoolie-afkThat sounds good.
19:37.07[g2]sure I'll e-mail in a couple minutes
19:37.15[g2]I was going to test them first :)
19:37.35beewoolie-afktest schmest
19:37.45[g2]Right not the only change in the Opeslug is the path for me :)
19:38.20[g2]Does the memuconfig know which default profile was used ?
19:39.13beewoolie-afkI don't know what that would mean.
19:39.19beewoolie-afkWhat is a profile?
19:39.34[g2]sorry which default configuration
19:40.15[g2]you are correct about the 1 second autoboot being there when one makes the nslu2/openslug_config
19:40.39beewoolie-afkIt works the same way as the kernel.
19:40.44beewoolie-afk# make foo_config
19:40.51beewoolie-afkWill set the configuration from a template.
19:43.49*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
19:43.50*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
19:43.52beewoolie-afkWell, again it's like the kernel.  I used to symlink configs so that changes would be saved to the original file.  I don't do that anymore.
19:43.58[g2]what would be "nice" is that the IX{42x and it should really be IXP4xx would say "Linksys NSLU2", "Avila", "OpenSlug"
19:44.29beewoolie-afkif you make foo_config, then edit the config, build, and do make bar_config, the foo changes are lost unless you copy .config to another file.
19:44.47beewoolie-afkHuh?
19:44.47[g2]exactly and that's totally fine
19:45.49beewoolie-afkAre you saying that you want the implementations to list openslug/unslung?
19:46.04[g2]What I'm saying is just like in the kernel there are processor selections in the processor family having a similar concept or a different heading could reminded the APEX config users which profile is currently being used
19:46.13beewoolie-afkThe avilla should be added to the implementation list, just in case there are some hardware changes.
19:46.34beewoolie-afkWhen you use xconfig, you can see the implementation name
19:46.42[g2]I'll do the avila port and submit the patches and files to you
19:46.53beewoolie-afkMenucofig shows it, too.
19:47.14beewoolie-afkI suppose that the default memory config should change for the avilla.
19:47.25beewoolie-afkI don't know if there is anything else, though.
19:47.50[g2]Ok... are you saying the MACH_NSLU2 ?
19:47.54beewoolie-afkBTW, do you use quilt?  It makes it really easy to develop patches.
19:48.08[g2]beewoolie-afk, not yet on the quilt thing
19:48.22beewoolie-afkthere should be a MACH_AVILLA, I suppose.
19:48.33[g2]but it would be nice to say I go quilting with my friends :)
19:48.39beewoolie-afkThe MACH's are implementations of the CPU or CPU family.
19:48.56beewoolie-afkquilting on-line with your pals.
19:49.57beewoolie-afkNow that I think of it, it is important to have an AVILLA target because there is some NSLU2 specific GPIO code in APEX already.
19:50.40[g2]right
19:51.01[g2]althought they don't effect the avila platform right now
19:51.12[g2]it's GPIO 8 right ?
19:51.40[g2]The default IXP425 platform used 8-11 for INTA-D iirc
19:52.05[g2]The slug uses 8-10, and avila uses 8-11 but it's split between two pci buses
19:52.07beewoolie-afkI think so.  The NSLU2 doesn't.  It uses 8 for power-down, IIRC.
19:52.22[g2]ok... maybe I'm off by 1
19:52.27[g2]maybe it's 9-12
19:52.33beewoolie-afk14 and 15 are clocks
19:52.37[g2]and 9-11 on the slug
19:52.38beewoolie-afk13 is PCI reset
19:52.47beewoolie-afk12, 11, 10, 9 are ints, IIRC.
19:52.53beewoolie-afksomething like that.
19:52.56[g2]nod.
19:53.17[g2]Oh... so 13 is a hard PCI reset ?
19:53.27beewoolie-afkShould this be AVILA_XXXX or just generic AVILA
19:53.51[g2]I haven't looked closely at the boards and the does only the 2347
19:53.53beewoolie-afkWell, there is a pin that performs PCI reset.  I think, though, that it's a PCI peripheral reset that goes to the USB controller.
19:54.17[g2]so that resets all the target devices
19:54.28[g2]PCI target devices that is
19:55.29beewoolie-afkIs there an ixp425 on that board?
19:55.49[g2]it's a 422 but they do 420, 422 and 425
19:55.56[g2]I've got two 425s here
19:56.02beewoolie-afkOn the same model number?
19:56.04[g2]a BB and BD
19:56.14[g2]the RV042 and RV082
19:56.26[g2]266 @ 425 and 533 @ 425
19:56.53[g2]Those are the devices that are getting pimped out tomorrow or the next day
19:57.01beewoolie-afkThere are a couple of ways to handle this.  I can add a sub-menu for avila so that all of the options are selectable.
19:57.12beewoolie-afkOr I can add top-level entries for each.
19:57.22beewoolie-afkIt depends on how many choices there are
19:57.25[g2]Let's leave the avila stuff alone for now
19:57.41beewoolie-afkI just added one.
19:58.02[g2]At the top level ?
19:58.09beewoolie-afkYeah.  Below nslu2
19:58.35beewoolie-afkconfig MACH_AVILA2347
19:58.35beewoolie-afk<PROTECTED>
19:58.35beewoolie-afk<PROTECTED>
19:58.35beewoolie-afk<PROTECTED>
19:58.37beewoolie-afk<PROTECTED>
19:58.41beewoolie-afk<PROTECTED>
19:58.43beewoolie-afk<PROTECTED>
19:59.31[g2]Ok, I'll send you the default config file and any other patches required
19:59.35beewoolie-afkko
20:00.31[g2]So let me test the OpenSlug changes first and we'll have that fixed and finished, then we can have the default template for the avila stuff
20:01.07[g2]I might float a message on the avila board and let them know it all works for 2.6 and I could possibly test their default config too
20:41.32*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
20:41.32*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
20:54.43[g2]jacques I've got an APEX question for you ?
20:54.55jacquesfor me?
20:55.00[g2]yeah for you
20:55.16[g2]you're interested in APEX right ?
20:55.31jacquesyes
20:55.57[g2]Ok.. after I give beewoolie-afk some minor tweaks I think APEX is ready to go on the slug
20:56.48[g2]basically, you'll download the latest, tweak for your compile path, make openslug_config; make and have an apex.bin for flashing
20:57.27[g2]jacques, your running OpenSlug right ?
20:57.31[g2]you are
20:57.37jacquesyes
20:57.50[g2]ok... so the question then becomes......
20:58.16[g2]That solves the normal (current Redboot configuration).
20:58.40[g2]With a different APEX config file we can have a custom flash layout
20:58.54[g2]where we lose all the legacy stuff
20:59.03[g2]and gain back 3 128K blocks
20:59.24[g2]plus a APEX config area for persistent changes
21:03.16*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
21:03.17*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
21:03.42[g2]3) Reflash the bootloard, "dd if=apex_1.2.11_openslug.bin of=/dev/mtdblock0"
21:04.04[g2]4) make sure /initrd isn't mounted (umount /initrd)
21:04.12[g2]5) reboot "shutdown -r now"
21:04.24[g2]6) do the happy dance as it all comes up
21:04.54jacquesok, now I want to flash a new openslug image, how do I do it?
21:05.22jacquessince AFAIK NIC isn't working under APEX
21:05.27jacquesso no tftp
21:05.59[g2]jacques, The NIC and USB are 100% with the lastest APEX
21:06.19[g2]I'm booting via APEX to SDA1
21:06.22jacquesyou mean once linux is booted
21:06.34[g2]and the IXP automatically comes up properly
21:06.41[g2]sure
21:06.49jacquesI mean, what is the equivalent of load -r -v -b 0x01000000 /nslu2/rootfs.jffs2
21:07.01[g2]well for now...
21:07.13jacquesxmodem? :-\
21:07.16[g2]boot into linux and dd if=reboot.bin of=/dev/mtdblock0
21:07.28[g2]and restore redboot or xmodem stuff down.
21:07.38dyoung-zzzzipkg upgrade openslug-kernel
21:07.46dyoung-zzzzshould work.
21:07.55[g2]dyoung-zzzz, welcome.
21:08.00[g2]there are minor issues with that
21:08.14jacqueslet's say I can't boot into linux
21:08.17dyoung-zzzzyou guys made quite the backlog.
21:08.21[g2]:)
21:08.25dyoung-zzzzstill catching it up.
21:08.27jacquesI have to JTAG redboot back on ?
21:08.37[g2]ok...
21:08.45[g2]You've got serial
21:08.52[g2]and the kernel is still good right ?
21:09.04jacquessounds like I need a slug specifically for testing APEX since I can't do dual duty as an openslug tester with it
21:09.13[g2]I think we can just change the kernel cmd line from APEX and NFS boot
21:09.14jacquesno, let's assume the kernel is non-bootable
21:09.38jacquesonly thing that workd is APEX
21:10.09[g2]Ok... You X or ZMODEM the kernel in (untested) then NFS boot
21:10.30[g2]well'll have to add a small initrd for the modules networking modules to be loaded from the ramdisk
21:10.31jacquesOK so APEX *does* have [xz]modem that's good, untested is bad
21:11.14[g2]Ummmm... would you like me to wipe your butt next time you go to the bathroom too ?
21:11.16[g2]:)
21:11.33jacquesyour point?
21:12.15[g2]My point I thought you were interested in APEX.
21:12.24jacquesI think I need at least one more slug, then the JTAG'ed one can be specifically for testing APEX
21:12.36dyoung-zzzzGood Plan.
21:12.58dyoung-zzzzJust as I should serial-enable my openslug test bed so I can use the jtag one for testing funner stuff.
21:13.00[g2]very good plan
21:13.02jacquessure, I'm just trying to see where it is feature-wise
21:13.21[g2]that makes great sense
21:13.38jacquesyeah, I have two serial kits left - one usb and one not
21:14.04jacquesI'll probably use the USB one - wish I could have found more
21:14.21[g2]So we'll test the serial and make sure it works
21:14.43[g2]My real question to you and dyoung-zzzz is about the flash layout
21:15.09jacqueswhere will the partition table be stored?  passed in on the kernel command line only?
21:15.15[g2]would you stay with the current layout or want to gain the 3 flash block back and move to a non-standard layout
21:16.03[g2]well APEX doesn't need the partition table right now
21:16.12jacquesonly thing I can think of that's necessary in the current layout is partition table, and if we can somehow store that in the APEX config area, we can regain all that space
21:16.37[g2]Right I'd move it to the APEX environement area
21:16.45jacqueshow are the partitions communicated to the kernel?
21:17.03dyoung-zzzzI'm in favor or swiping the mac address from the end of redboot and putting it someplace intelligent.
21:17.39[g2]dyoung-zzzz, nod. we are going to backup the last 144 bytes of Redboot which are the most of the unique ones
21:18.33[g2]although if you're running OpenSlug the mac addr is already in /etc/network/interfaces
21:18.49dyoung-zzzzyeah, but it would need a way to get there in the first place right?
21:19.11[g2]right you have to boot OpenSlug first right ?
21:19.19dyoung-zzzzyou do?
21:19.56jacquesseems on first boot openslug extracts MAC from redboot and puts it in /etc/network/interfaces
21:20.04dyoung-zzzzin that case, its critical to have that information someplace smart because when you reflash then turnup the thing, the turnup wont know what its mac is anymore
21:20.39[g2]sure turnup copies /etc/network/interfaces after its been built
21:20.45dyoung-zzzzanyways, thats one of my peeves about redboot.
21:20.54beewoolie-afkjacques: BTW, xmodem receive in APEX *is* tested.  I use it frequently.
21:21.18[g2]beewoolie-afk, I was just speaking from personal experience
21:21.25beewoolie-afkdyoung-zzzz: I believe we can teach APEX to program the MAC address.
21:21.57[g2]I see near zero need to go there
21:22.31beewoolie-afkYou're fussing about MAC addresses.  I'm saying there are options.
21:22.32jacquesbeewoolie-afk, great :-)
21:22.49[g2]right there are lots of options
21:22.53beewoolie-afkThere isn't really a need to worry about losing the MAC address as it's printed on the box.
21:23.34[g2]I'm just trying figure out if you guys want to regain the 3 blocks and lose the legacy stuff
21:23.54beewoolie-afkjacques: FYI, the only important feature missing from APEX which Redboot has is NPE support.  It's kinda hairy, so I've elected to avoid it for now.
21:24.48jacquesbeewoolie-afk, yeah, I can understand how that would be a difficult painful feature to add (NPE)
21:24.49[g2]beewoolie-afk, you saw the note about moving the partition table to the APEX environment area right ?
21:24.52dyoung-zzzzdont we really only gain back 1 block?
21:25.31[g2]dyoung-zzzz, sysconf area 1, Ramdisk fragmentation 1, Trailer 1, 1+1+1=3
21:25.36dyoung-zzzzor are you counting the last block too, in which case thats 2
21:25.56[g2]Plus the sysconfig block
21:26.01dyoung-zzzzOk, so the suggestion is that apex will always fit in 1 block?
21:26.24[g2]1 block APEX code, 1 Block APEX environment
21:26.36[g2]same size as Redboot which is 2 blocks right ?
21:27.00dyoung-zzzzright, I saw talk of "future exansion" above someplace with the suggestion of reserving 2blocks for apex code.
21:28.55[g2]Ok so the 144 bytes (or less) from the end of Redboot and the partition table from the trailer get moved to the APEX environment block
21:29.07[g2]say the last 1 K or 2K
21:29.54jacquesI think max partition table size is 4kB (not positive)
21:30.11jacques256 bytes x 16 entries
21:30.26[g2]Ok... it's a 128K block with hardly anything in it
21:30.38[g2]I'm sure we've got plenty of room
21:31.03jacquesyes, I don't think that will be an issue
21:32.28[g2]so for discussion sake... Let's gzip the sysconf info, put the last 144 bytes, and have room for a 4K partition table at the end of the environment block
21:32.36[g2]that's the last 144 bytes of Redboot
21:32.51[g2]and fill the partition table with the proper values
21:33.37[g2]OK... This makes total sense to me.... All these things can be done at runtime with APEX
21:34.10[g2]So by default, APEX is like it currently is, meaning it doesn't initialize the environment block
21:34.38[g2]The procedure is flash OpenSlug
21:34.50[g2]and one can ipkg install APEX
21:37.15[g2]which saves off a copy of Redboot and reflashes APEX as the bootloader
21:38.37[g2]then there's another routine that can enable the environment block and move stuff around on the jffs2 partition to setup the extended functionality
21:39.40beewoolie-afkCan I get a clarification here?  Is the 'partition table' one of those redboot partition table thingys?
21:39.55[g2]it's a kernel partition table
21:39.57beewoolie-afkI ask, because there isn't a standard for it.
21:40.09jacquesthere is
21:40.13beewoolie-afkFlash partitions are configured with the command line.
21:40.30[g2]actually there and ordered dependency
21:40.36[g2]the commandline is just one option
21:40.36jacqueswe used existing kernel code to read the partition table
21:40.45jacquesand kernel by default looks at start of last block for it
21:41.02beewoolie-afkI think you're talking about a redboot partition table.
21:41.21beewoolie-afkstatic const char *part_probes[] __initdata = {"cmdlinepart", "RedBoot", NULL};
21:41.34beewoolie-afkThat's how the physmap driver handles partition table parsing.
21:41.45beewoolie-afkIt looks to the command line and then looks for a RedBoot partition table.
21:42.25beewoolie-afkIOW, we don't really need the redboot legacy stuff.
21:42.38beewoolie-afkIt's better IMHO to use the command line.  It's more clear and easier to change.
21:43.25jacquesI don't see a problem with that
21:43.33[g2]nslu2_2.6.11.patch:+static const char *probes[] = { "cmdlinepart", "RedBoot", "NSLU2", NULL };
21:43.51[g2]that's the patch applied by OE
21:43.56jacquesI was just saying before that the partition info had to be stored somewhere and passed to the kernel somehow
21:43.58beewoolie-afkAh.  The it's a non-standard format.
21:44.04beewoolie-afkAll the more reason to go to the command line.
21:44.18jacquesnon-standard how?
21:44.32beewoolie-afkNSLU2 is not a partition 'standard'.  No other platform uses it.
21:44.32jacquesit's a standard RedBoot partition table
21:44.43beewoolie-afkThen why call it something different?
21:44.55jacquesoh, I dunno what that is, I was referring to the "RedBoot" part above
21:45.20[g2]well we could easily change the patch to be "cmdlinepart", "RedBoot", NULL
21:45.33jacqueswhen I wrote the code in slugimage to build tha partition table, I referred to the kernel code for RedBoot partition table
21:45.42beewoolie-afkIn the end,I think it's better to go to the command line.  We don't even have to change the kernel.
21:45.50jacquesso I dunno what NSLU2 format means
21:46.17jacquesyes, we are ascillating between two different things here
21:46.21jacquesoscillating
21:46.51jacquesI have already agreed command line is fine, but we are also arguing about the current partition table format name
21:47.00beewoolie-afkI'm trying to pull it back.  The more that we do that is simple, command-line or ATAG based, the more conforming our solution will be.
21:47.03beewoolie-afkRight.
21:47.33[g2]beewoolie-afk, nod.
21:47.49[g2]jacques, As far as I'm aware it's the RedBoot format
21:48.03jacques[g2], me too
21:48.30jacquesand I wrote the slugimage code by looking at the kernel redboot format code so I think I know what I'm talking about
21:48.37dyoung-zzzzNSLU2 from that kernel patch is the stock factory non-standard static partition scheme that linksys uses.  We're pretty much ignoring it for everything made post-slugimage.
21:48.37jacques(could be wrong tho)
21:48.42beewoolie-afkjacques: Just to clarify one other point.  APEX is really small.  It will remain so unless we decide we have to have NPE.  I've been plannng to add a feature where it is possible to put the environment in the same block as the loader.  It means we could reduce the memory usage by another block.
21:49.46jacquesbeewoolie-afk, I think that would be for specialized application tho - because I would like to be able to change the env data without worrying about corrupting bootloader
21:50.00jacquespersonally for slug I don't think it's worth it for that one block
21:50.02beewoolie-afkThat isn't really the problem.
21:50.58beewoolie-afkThe environment code is smart about writing data.  The trouble is that the allocated space can fill with old settings.  Erasing the whole block requires erasing the same block that the loader is in and rewriting it.
21:51.10beewoolie-afkIt just needs a special command to do it.
21:51.11jacquesah nice
21:51.36jacquesso you have a section of 0xFF and just write new config data until it's full
21:51.37beewoolie-afkAs I said, it's just a small clarification.  
21:51.47jacquesclever
21:51.49[g2]well you are risking the boot block when you do that and it isn't worth it for that 1 block
21:52.25beewoolie-afkyeah.  to both of you.  That's why it hasn't come up.  it is more important on systems with larger erase block sizes.
21:52.37[g2]agreeed
21:53.28[g2]so do you want the partition table copied or not ?
21:53.52jacquesmaybe not copied so much as translated into a command line format
21:54.08[g2]-dinnerthat's totally possible too
21:54.20jacquesin practice, we never got close to the 16 entry limit
21:54.33jacquesseems a waste of space 4kB just for partition table
21:54.53[g2]-dinnerwell it's empty anyway
21:54.58[g2]-dinneroff to dinner
21:55.22dyoung-zzzzYou guys are the experts.
22:02.50jacquesdang I wish slugs would go on sale somewhere
22:11.08dyoung-zzzzIs there a clearing house of tivo hacking info someplace?  Googling returns too many hits to be helpful.
22:11.52dyoung-zzzzBasically I inheireted a Series2 which I just slapped a 120G into.  its software version was updated to 7.1; but I want to be able to telnet to it and screw around wiht it, etc...
22:12.09dyoung-zzzzdo I need to downgrade to 4.0.1 or something to do that?
22:15.19jacquestivo underground
22:16.07jacqueshttp://www.tivocommunity.com/tivo-vb/forumdisplay.php?f=24
22:19.00[g2]-dinnerjacques, last I bought they were $79US at newegg
22:19.20jacques[g2], yeah, I know, I would prefer to pay less
22:19.44jacquesI check newegg almost every day and they are holding steady
22:20.24jacquesdyoung-zzzz, if you need info about things which aren't allowed on tivocommunity.com, there are other forums
22:20.48dyoung-zzzzOk
22:21.06dyoung-zzzzI just wanna be able to telnet to the thinig for now.
22:21.15dyoung-zzzzI'll get to "those other topics" a little later.
22:36.52*** join/#openjtag ka6sox-office (~tking@ka6sox.nslu2-linux)
22:36.57*** join/#openjtag jbowler (~jbowler@jbowler.nslu2-linux)
22:37.03[g2]jbowler, welcome
22:37.12[g2]beewoolie-afk, around ?
22:37.40[g2]jbowler, beewoolie is the author of APEX
22:37.52*** join/#openjtag ByronT (byrontodd@ByronT.nslu2-linux)
22:38.01jbowlerAh, ok.  Let me state the problem...
22:38.05[g2]I've been working with him to get stuff fully running on the slug and other IXP OEM boards
22:38.29jbowlerWith RedBoot, RedBoot SysConf partition->openslug jffs2 /etc/default/sysconf (say) and this supplies the eth0 MAC address
22:38.33[g2]the problems is if we move sysconf you lose where you are getting your MAC frome
22:38.46jbowlerWith APEX something needs to read and generate the same file with equivalent information.
22:39.16jbowlerThat's easy - we just need some 1130 bytes somewhere to store the useful part of the original RedBoot SysConf
22:39.26[g2]144
22:39.57[g2]for the Redboot, and we've been talking about saving the whole sysconf gzipped or something like that
22:40.00jbowlerIt's not just 144.  The user may have run LinkSys config and stored useful data there.
22:40.20[g2]144 for Redboot, more for sysconf
22:40.41jbowlerThere's no need to gzip it - actually that's bad.  Putting gunzip into /linuxrc is bad ;-)
22:41.12[g2]that makes a bunch of sense if you are pulling it from there
22:41.35[g2]actually, we can add something to /proc like macaddr
22:41.44dyoung-zzzz[g2] right.
22:41.51dyoung-zzzzthat was the latest plan, but I slacked on it.
22:41.56[g2]and it can be pulled from flash or passed on the cmd line
22:41.59dyoung-zzzzand the patch never got written.
22:42.14jbowler[g2]: if it can be added to /proc then it can be added to the driver.  Much better, no change for user land mistakes.
22:42.18[g2]well APEX passes cmdline options fine
22:42.23jbowlers/change/chance/
22:42.53[g2]yeah.. except for those licensing issues it makes total sense
22:43.07jbowlerMy feeling is that the MAC should not need to be set by ifconfig - it should default to the allocated MAC if at all possible.
22:43.11[g2]that's in the 425 section which is GPL right ?
22:43.21jbowler[g2]: correct
22:43.24[g2]jbowler, totally agreed
22:43.58jbowlerAnyway, so what, if the NSLU2 kernel just has the two numbers stored somewhere that's fine (I assume there are two, but maybe LinkSys only bought one...)
22:44.10[g2]OK.. SO.. that begs the question of what do we do with the current OpenSlug for setting the MAC addr ?
22:44.18*** join/#openjtag prpplague (~billybob@72.22.134.184)
22:44.35ka6sox-officeisn't that going to be bootloader specific?
22:44.53jbowlerWhy do we need to do anything?  If I can't read hw_addr from SysConf I won't set it ;-)
22:45.52[g2]jbowler, how do you know it's sysconf ? are you checking the partition name, are you checking the lenght at the beginning are you parsing the tag ?
22:46.07jbowler[g2]: yes, to all of the above
22:46.28[g2]Ok. so if there's no sysconf partition, you could look elsewhere
22:46.30jbowlerIf I can find something called "SysConf" in /proc/mtd I can find it.
22:47.21jbowler[g2]: yes, and if I can't find SysConf in /proc/mtd I know to look elsewhere.
23:11.03*** join/#openjtag purl (ibot@apt.bot.TimRiker.active.supporter.pdpc)
23:11.04*** topic/#openjtag is This is the place to discuss development of Open JTAG tools and Bootloaders
23:11.43[g2]beewoolie-afk, currently you are not using the partition tables at all you are just using the address/offset/lenght from APEX environment vars
23:12.24[g2]when we move to the driver model are you planning to use those ? or would you rather not support that ?
23:14.17beewoolie-afkPresently, regions are specified literally.
23:14.38beewoolie-afkOnce I add some of the features I've talked about for chaining driver, we can use aliases to refer to flash partitions.
23:14.56beewoolie-afkalias part0  nor:0+128k
23:15.42beewoolie-afk[g2]: BTW, I got distracted.  you don't need to specify the length of the receive region for an xmodem receive.
23:16.03beewoolie-afkapex> xr nor:0           # download to start of flash

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