OESF Portables Forum
Model Specific Forums => Sharp Zaurus => Zaurus - pdaXrom => Topic started by: Ashley on April 15, 2004, 03:15:44 pm
-
Hi everybody!!
At last! And seems to work!...
I\'ve written a modification of pdaxrom installer which allows to easily resize nand to have JUST THE ROOT PARTITION, WITH NO HOME.
No fragmentation. Full space to install everything you what (100 MB free after install rom on a Cx60!!). Of course, /home now is on Root partition.
This is mainly thanks to ikm, who is the author of the trick.
It has been designed to be very careful before patching your Nand, but you must consider this as a beta version. Please backup your data before doing anything (maintenance menu backup is recommended). Also, for the moment, it was just tested with my 760. Take even more cares if you have another model.
You can find it in http://ciberia.ya.com/ashley_s/ (http://ciberia.ya.com/ashley_s/)
Contents of new.tar:
- updater.sh: to substitute pdaxrom updater.sh
- updater.decoded: decoded version of the last
- updater.decoded.cacko: decoded version of pdaxrom updater.sh
- mtdtweak & mtdparts: you must put these together with updater.sh and tools.tar and initrd.bin from pdaXrom
- encdesc: utility to encode and decode updater.sh files
Please try it with care and tell me bugs and succeeds in this topic :wink:
-
Sorry, can\'t download the file with links or wget.
I\'ve tried downloading the file with mozilla, but the file is corrupted. At least tar doesn\'t recognize it as valid tar archive...
-
I get 404
[779]> wget http://ciberia.ya.com/ashley_s/new.tar
--09:06:29-- http://ciberia.ya.com/ashley_s/new.tar
=> `new.tar\'
Connecting to ciberia.ya.com:80... connected!
HTTP request sent, awaiting response... 404 Not Found
09:06:30 ERROR 404: Not Found.
-
Okay, try downloading with referer.
wget --referer=http://ciberia.ya.com/ashley_s/index.html http://ciberia.ya.com/ashley_s/new.tar (http://ciberia.ya.com/ashley_s/new.tar)
This will work.
I hate referer rewrites........
-
Success - I think this is a more preferable setup! Maybe I should\'ve asked before trying it, but what\'s the worst that could happen if something went wrong?
-
I have a SL-C750 (64MB). The mtdtweaks/mtdparts address in mtd1 is different, it is 0x20AA4 on my device. If I reflash \"mtdtweaks\" into /dev/mtd1 using nandlogical, it works OK. When I reboot, /proc/mtd shows mtd2 as \"root\" with a size of 0x3900000, and no mtd3 and \"home\" partition, which is correct.
I can eraseall -j /dev/mtd2 just fine, however no matter what I do the system refuses to mount the partition. If I use a smaller partition size (35MB for example) then I can eraseall -j /dev/mtd2 and mount it just fine. Something refuses to allow mounting of such a large jffs2 partition (57MB) on the C750. I\'ve even manually created a jffs2 partition image using mkfs.jffs2 and dd\'ed it into /dev/mtd2. But mount simply refuses to work. Perhaps there is some other limit in the C750\'s kernel that won\'t allow this to work? Do you have any suggestions as to anything to try? I would love to get rid of the \"/home\" partition!
-
Hallo Ashley,
here correct fix for maximal size for root partition in SL-C kernel loader: http://www.pdaxrom.org/downloads/misc/slc-...artsize-fix.zip (http://www.pdaxrom.org/downloads/misc/slc-partsize-fix.zip)
unzip and copy updater.sh and mainte_fix.bin to CF and do flash update. It will rewrite kernel loader with new fixed version.
How it was hacked:
Bootloader was disassembled in ASM source source,
looking inside, we will see:
seg000:00020838 ADR R1, aConsoleTtys0Ro ; \"console=ttyS0 root=/dev/mtdblock2\"
seg000:0002083C MOV R2, #0x22 ; \'\"\'
seg000:00020840 BL sub_313AC
seg000:00020844 CMP R7, #0x4000000
seg000:00020848 BCS loc_30888
seg000:0002084C MOV R0, R7,LSR#14
seg000:00020850 MOV R7, R0,LSL#4
seg000:00020854 MOV R2, R7
seg000:00020858 ADD R0, SP, #0xF0
seg000:0002085C ADR R1, off_309F4
seg000:00020860 BL sub_30FFC
seg000:00020864 MOV R0, R6
seg000:00020868 ADR R1, aMtdpartsSharps ; \" mtdparts=sharpsl-nand:7168k@0k(smf),
\"
seg000:0002086C BL sub_30FC8
seg000:00020870 MOV R0, R6
seg000:00020874 ADD R1, SP, #0xF0
seg000:00020878 BL sub_30FC8
seg000:0002087C MOV R0, R6
seg000:00020880 ADR R1, aK7168kRootHome ; \"k@7168k(root),-(home)\"
seg000:00020884 BL sub_30FC8
seg000:00020888
seg000:00020888 loc_30888 ; CODE XREF: sub_30768+E0?j
also the same code from address 0x20b7c
seg000:00020B7C ADR R1, aConsoleTtys0_0 ; \"console=ttyS0 root=/dev/ram0\"
seg000:00020B80 MOV R2, #0x1D
seg000:00020B84 BL sub_313AC
seg000:00020B88 CMP R6, #0x4000000
seg000:00020B8C BCS loc_30BCC
checking for max root size there:
seg000:00020844 CMP R6, #0x4000000
and
seg000:00020B88 CMP R6, #0x4000000
as we already know, 0x4000000 is max size for 64MB, multiple it to 2 for 128,
it will:
seg000:00020844 CMP R6, #0x8000000
and
seg000:00020B88 CMP R6, #0x8000000
fix new value from hexedit in extracted mainte.bin, and write it back to Z.
Now run pdaXrom installer and increase root partition to 80 MB and reflash pdaXrom. Rebooting... login as root,
df -h /
root size now 80.0M ! :-)
-
Uuf!! I haven\'t been looking this for several hours and now it\'s plenty! (my first impression was nobody was interested!)
First thing: I\'m so sorry for the downloading problems! When I putted it seemed me it worked, but that nasty server dislikes any non-html file. Very sad. I just made a quick account just for this.
Then, a first solution: if you want i can send it by mail; for this PM me or write me (vicra dot 3111 at cajarural dot com). A better solution: Kirk to Enterprise, Kirk to Enterprise... Scotty, can you help me? :wink: No more experiments, as your site is the best resource for z stuff, it would honor me if you wish to make a little room for me
Success - I think this is a more preferable setup! Maybe I should\'ve asked before trying it, but what\'s the worst that could happen if something went wrong?
Thanks. The worse: installation fails always and then have to restore a maintenance menu backup.
I have a SL-C750 (64MB). The mtdtweaks/mtdparts address in mtd1 is different, it is 0x20AA4 on my device. If I reflash \"mtdtweaks\" into /dev/mtd1 using nandlogical, it works OK. When I reboot, /proc/mtd shows mtd2 as \"root\" with a size of 0x3900000, and no mtd3 and \"home\" partition, which is correct.
Nice. That sort of thing is what i expected when the model is diferent. Both ikm and me have a 760. Can any other 750 user confirm this? If so, we can do a simple mod to installer to do things right.
I can eraseall -j /dev/mtd2 just fine, however no matter what I do the system refuses to mount the partition. If I use a smaller partition size (35MB for example) then I can eraseall -j /dev/mtd2 and mount it just fine. Something refuses to allow mounting of such a large jffs2 partition (57MB) on the C750. I\'ve even manually created a jffs2 partition image using mkfs.jffs2 and dd\'ed it into /dev/mtd2. But mount simply refuses to work. Perhaps there is some other limit in the C750\'s kernel that won\'t allow this to work? Do you have any suggestions as to anything to try? I would love to get rid of the \"/home\" partition!
I dislike very much hearing that. I think this tweak is even more needed by 700/750 users. You have tried the same things I would, so I can\'t give you a solution. I would like to see dmesg command line output and a nand dump of tweaked mtdparts, can you post them? May be this surpasses me, then sashz could be your only hope...
And sashz,
You found it!! You are a genius!! Boot loader disassebly!
It seems to be an old code from 700/750 that Sharp did not updated for 760. Things that happen. If we have had source code...
This is more useful for x60 users than having just 50 or 63 megs root partition, but the thing is that I prefer not to have any home partition, but mount /home to Root one. I like to install as many things as possible. I use my z for my daily work (I mainly use gnumeric) but I don\'t use /home to keep any work file, I store they on my CF card always (too many reflashes). Then, home space is useless for me. I think it maked much more sense in qtopia since root was ro. I have many linux boxes installed on my home computers, but no home partition, it is something i don\'t like for a desktop computer. And there is another drawback: as /home is used many times as a temp dir (installing, for exaple; you can change this, but sometimes you don\'t realize) you can\'t shorten too much it. The more you fragment, the more space you loose.
Well, enough.
Sashz, I hope you rewrite installer to automate your fix in next releases, but i think many people agree with me and would be more happy if the installer also gives the chance to have just one partition :wink:
I\'m afraid when i post this there will be dozens of new posts... let\'s see...
-
After using Ashley\'s modified update.sh, I went to reflash cacko 1.21. Problem - /dev/mtdblock3 no longer exists, hence no /home partition - but the installers are assuming it is there. I don\'t posess the skills to modify the partition table yet and was just about to do a nand restore when Sashz posted his partition size fix. Great timing for me!
-
After using Ashley\'s modified update.sh, I went to reflash cacko 1.21. Problem - /dev/mtdblock3 no longer exists, hence no /home partition - but the installers are assuming it is there. I don\'t posess the skills to modify the partition table yet and was just about to do a nand restore when Sashz posted his partition size fix. Great timing for me!
I wrote an option to restore original partition scheme in my installer. Isn\'t it working? After that restoration you should be able to install pdaxrom with its original installer or any other rom.
Ash
-
I\'m sending installer as an attachment. Tell me if it works.
-
I wrote an option to restore original partition scheme in my installer
You certainly did! I should pay more attention ;-) Thanks, it works well and is slightly cleaner than having a zero size mtdblock3 partition.
-
I dislike very much hearing that. I think this tweak is even more needed by 700/750 users. You have tried the same things I would, so I can\'t give you a solution. I would like to see dmesg command line output and a nand dump of tweaked mtdparts, can you post them? May be this surpasses me, then sashz could be your only hope...
I reverted my work all late last night but I will try to get a chance to try this again today. I do remember dmesg looked fine. An interesting thing is that if I formatted the 57MB partition as ext2 it would mount just fine - no errors and full 57MB partition available. So it appears to be just a problem with jffs2 (!?). I verified the eraseall worked, however, so the partition was getting formatted properly.
When I reverted everything back to original pdaXrom 1.0.5, I tried another test - simply making root partition 55MB (leaving the rest for the home partition.) Cacko installer properly did this, but the partition was no longer mountable then. I then tried 50MB and it also failed; when I tried 45MB root size it worked OK (?! i.e., 45MB root, 12MB home.)
I have two SL-C750\'s and will replace Qtopia on the other one and see if it behaves the same.
Thanks!
-
dvdrw, that sounds disconcerting. As your test show, the problem is not concerned with tweaking mtdparts. I think some people reported problems to this forum when tried to make big root partitions on 700/760 with early pdaxrom versions, just like happens to you. I didn\'t pay attention since it is a different problem from the x60 one, as sashz in his wisdom has showed. May be you should search that.
If it mounts ext2 but not jffs2, it sounds like a 750\'s kernel bug. But I understand you can\'t mount with either maintenance or pdaxrom kernel (am I wrong?). If so, don\'t know, maybe the bug is present in both kernels?! Or you just tried to mount with maintenance?
-
I have two SL-C750\'s and will replace Qtopia on the other one and see if it behaves the same.
I just tried a simple test on my second SL-C750, I tried installing normal unmodified pdaXrom 1.0.5 and resized NAND with 56MB for the root partition size (leaving 1MB for home), and the same problem occurs.
Here are the relevant entries from dmesg:
[span style=\'font-size:16pt;line-height:100%\']Linux version 2.4.18-rmk7-pxa3-embedix-021129 (zaurus@sharplinux) (gcc version 2.95.2 19991024 (release)) #1 Sat, 10 May 2003 14:51:36 +0900
CPU: Intel XScale-PXA255 revision 6
Machine: SHARP Shepherd
Converting old-style param struct to taglist
Memory clock: 99.53MHz (*27)
Run Mode clock: 199.07MHz (*2)
Turbo Mode clock: 398.13MHz (*2.0, active)
Security risk: creating user accessible mapping for 0x08000000 at 0xf1000000
On node 0 totalpages: 16384
zone(0): 16384 pages.
zone(1): 0 pages.
zone(2): 0 pages.
Kernel command line: console=ttyS0 root=/dev/ram0 mtdparts=sharpsl-nand:7168k@0k(smf),57344k@7168k(root),-(home) ZAURUS_UPDATE=/etc/update/update_sd
Console: colour dummy device 80x30
Calibrating delay loop... 397.31 BogoMIPS
Memory: 64MB = 64MB total
Memory: 53820KB available (1824K code, 520K data, 108K init)
...
Starting kswapd
JFFS2 version 2.1. © 2001, 2002 Red Hat, Inc., designed by Axis Communications AB.
reset resolution unkown =] 480x640
...
Corgi audio driver initialize
Sharp SL series flash device: 1000000 at 0
Using static partision definition
Creating 1 MTD partitions on \"sharpsl-flash\":
0x00120000-0x007f0000 : \"Filesystem\"
NAND device: Manufacture ID: 0x98, Chip ID: 0x76 (Toshiba TH58512FT)
Creating 3 MTD partitions on \"Toshiba TH58512FT\":
0x00000000-0x00700000 : \"smf\"
0x00700000-0x03f00000 : \"root\"
0x03f00000-0x04000000 : \"home\"
Linux Kernel Card Services 3.1.22
options: [pm]
...[/span]
And here\'s /proc/mtd:
[span style=\'font-size:16pt;line-height:100%\']# cat /proc/mtd
dev: size erasesize name
mtd0: 006d0000 00020000 \"Filesystem\"
mtd1: 00700000 00004000 \"smf\"
mtd2: 03800000 00004000 \"root\"
mtd3: 00100000 00004000 \"home\"
#[/span]
So mtd2 formats fine with \"eraseall -j\", but when I try mounting it I get this error:
[span style=\'font-size:16pt;line-height:100%\']# cd /tmp1
# mkdir foo
# mount -t jffs2 /dev/mtdblock2 /tmp1/foo
mount: Mounting /dev/mtdblock2 on /tmp1/foo failed: Invalid argument
#[/span]
If I reduce the size of mtd2 to 46MB then it mounts fine - at 47MB and above it fails. Here\'s /proc/mtd for the largest mtd2 I can create and still have it mount OK:
[span style=\'font-size:16pt;line-height:100%\']# cat /proc/mtd
dev: size erasesize name
mtd0: 006d0000 00020000 \"Filesystem\"
mtd1: 00700000 00004000 \"smf\"
mtd2: 02e00000 00004000 \"root\"
mtd3: 00b00000 00004000 \"home\"
#[/span]
Sashz, do you have any ideas as to what might be going on? This is certainly strange!
Thanks!!
-
dvdrw, that sounds disconcerting. As your test show, the problem is not concerned with tweaking mtdparts. I think some people reported problems to this forum when tried to make big root partitions on 700/760 with early pdaxrom versions, just like happens to you. I didn\'t pay attention since it is a different problem from the x60 one, as sashz in his wisdom has showed. May be you should search that.
Interesting, I didn\'t know this! I will search now.
If it mounts ext2 but not jffs2, it sounds like a 750\'s kernel bug. But I understand you can\'t mount with either maintenance or pdaxrom kernel (am I wrong?). If so, don\'t know, maybe the bug is present in both kernels?! Or you just tried to mount with maintenance?
This is exactly what I am thinking...I hope I can figure out a workaround for this. Note that I have only been trying the mount with the maintenance kernel. Could you tell me how I could boot with the pdaXrom kernel and try this?
-
So mtd2 formats fine with \"eraseall -j\", but when I try mounting it I get this error:
[span style=\'font-size:16pt;line-height:100%\']# cd /tmp1
# mkdir foo
# mount -t jffs2 /dev/mtdblock2 /tmp1/foo
mount: Mounting /dev/mtdblock2 on /tmp1/foo failed: Invalid argument
#[/span]
Is anything additional showing up in \'dmesg\' after you issued a failing mount command? How long does it take for the mount command to fail?
From my experience, there were two problems with jffs2 I stumbled upon. First, when the flash is dirty, you have to have at least one valid inode for the mount to succeed. Second, when the partition size is quite big, jffs2 fails to allocate enough memory, because it uses kmalloc, and the required allocation size is bigger than 128k (the maximum that kmalloc can allocate). The latter problem is fixed by changing the code to use vmalloc instead of kalloc, but anyway, it shows up only with bigger partitions (say, 256M or 512M with the appropriate erasesize). It doesn\'t show up even on 128M with the 16k erasesize, as in SL-C760. So I do not really know what to suggest at the moment, further investigation is needed. Wells, inserting tons of printk\'s to the kernel always helps, but there may be a more simple solution
-
Is anything additional showing up in \'dmesg\' after you issued a failing mount command? How long does it take for the mount command to fail?
It fails instantly and there are no entries added to dmesg after issuing the mount command.
So I do not really know what to suggest at the moment, further investigation is needed. Wells, inserting tons of printk\'s to the kernel always helps, but there may be a more simple solution
Are the sources to the Zaurus maintenance kernel available somewhere? Do you know if anyone has done an alternative, RAM-based distro that I could boot from and work w/the mtd partitions?
-
This is exactly what I am thinking...I hope I can figure out a workaround for this. Note that I have only been trying the mount with the maintenance kernel. Could you tell me how I could boot with the pdaXrom kernel and try this?
You can try to make a big jffs2 file system on a sd or cf with a running pdaxrom. You must have one, since you are reflashing :wink:
If this works, you may try the only solution I find: update maintenance kernel. This, of course, is very risky and you can brick your z :? , but you seem to know what you do very well :wink:
I\'ll update to a maintenance 760 kernel. I can mail you a nandlogical READ of mine.
With some cares, you could even restore your mainte kernel with a running pdaxrom, if it goes wrong:
- Before doing any thing, leave pdaxrom in a useful state.
- Make a nandlogical READ of your maintenance kernel. Put it on sd or cf.
- Copy nandlogical tool from maintenance to sd or cf.
- nandlogical WRITE 760 kernel image. Reboot.
- Try maintenance. If everything works, fine, if not:
- Reboot and run pdaxrom (it should, no matter how mainte kernel is).
- Use nandlogical to WRITE back your kernel.
I think this is doable, but it\'s your chance.
Tell me if you dare :wink:
-
Ok, now I know what\'s happening with SL-C750. Just look at Sharp\'s kernel source:
$ cd linux/fs/jffs2/
(Note that SL-C760 is codenamed \'husky\')
$ grep -r HUSKY *
build.c:#ifdef CONFIG_ARCH_PXA_HUSKY
build.c:#ifdef CONFIG_ARCH_PXA_HUSKY
fs.c:#ifdef CONFIG_ARCH_PXA_HUSKY
nodelist.h:#ifdef CONFIG_ARCH_PXA_HUSKY
super.c:#ifdef CONFIG_ARCH_PXA_HUSKY
super-v24.c:#ifdef CONFIG_ARCH_PXA_HUSKY
All these #ifdefs (except one) say that for C760 we should use consistent_alloc, for all other models kmalloc should be used instead. Now it becomes apparent: only C760 has two large (64Mb) partitions, so large that kmalloc can\'t live with them.
The remaining one #ifdef says to increase reserved bad blocks from 24 to 80 for husky. The larger partitions, the bigger possible amount of bad blocks. Or maybe C760 flashes are worse than on other models
The outcome: you can\'t use standard maintenance kernel on C750 to create one big root partition. You will have to reflash it.
How to fix the kernel: remove the aforementioned conditional statements there, leaving only the C760 code. This way it will most probably work
-
ikm, you are THE root!
The outcome: you can\'t use standard maintenance kernel on C750 to create one big root partition. You will have to reflash it.
Ok, now we are sure: it is the only way.
How to fix the kernel: remove the aforementioned conditional statements there, leaving only the C760 code. This way it will most probably work
Or just try to flash a unflashed 760 kernel, as I told. Seems to be a quicker solution :wink:
Ash
-
Are the sources to the Zaurus maintenance kernel available somewhere?
I think this is just a copy of the same kernel with a slightly modified CMDLINE and a ramdisk. The OK-Menu will probably save where to look for the updater.sh in a dedicated area on flash (RAM is cleared during reboot) and then reboot.
-
... The outcome: you can\'t use standard maintenance kernel on C750 to create one big root partition. You will have to reflash it.
How to fix the kernel: remove the aforementioned conditional statements there, leaving only the C760 code. This way it will most probably work
yes this makes sense. I\'ve rebuilt the \"shepherd\" kernel but have defined the HUSKY config define for the jffs2 code. everything built fine, but I am not sure where to flash zImage in the updater. I tried flashing over the second kernel image location at 0x00220000 but that killed the device (even the \"d+m\" reset appears dead.) I\'m trying a \"c+d\" CF recover now to un-brick it...
-
success at last!!!1 I now have pdaXrom running with a single 57MB root partition on my C750! many thanks to ikm and ashley who made this possible for this happy user
-
Congratulations dvdrw!
many thanks to ikm and ashley who made this possible for this happy user
You are welcome, but you are the main cause of your succeed.
Now, I want some feedback, in order to update installer to work with 750s :wink:
I\'ve rebuilt the \"shepherd\" kernel but have defined the HUSKY config define for the jffs2 code. everything built fine, but I am not sure where to flash zImage in the updater.
Can you mail me zImage? The one you built works, and not sure if i can compile it correctly. I can\'t ever test it.
I tried flashing over the second kernel image location at 0x00220000 but that killed the device (even the \"d+m\" reset appears dead.)
What extrange. Did you reflash to a different location later? Or what did you do to make it work?
I\'m trying a \"c+d\" CF recover now to un-brick it...
Did you succeed doing this too? Fabulous! Maybe you can post a howto for Foxdie to this topic: http://www.zaurususergroup.com/index.php?n...iewtopic&t=2897 (https://www.oesf.org/forums/index.php?showtopic=2897) he seems desperate.
And finally, i\'ll need that you confirm the offset of mtdparts in a 750. I think it\'s all what i need to make a 2nd version of installer suitable for that model. Now i need a 700 voluntarious user :wink:
Oh man, i look like your boss!
-
Now, I want some feedback, in order to update installer to work with 750s :wink:
OK, so I found afterwards that I actually didn\'t need to fix and rebuild the Zaurus kernel. I originally hit a problem with actually booting the final installed pdaXrom kernel after doing the install, so Cacko\'s C750 kernel has the same issue that the stock Zaurus C750 kernel does (probably because it is based on the \"Sheperd\" device config). So rather than try to rebuild the pdaXrom kernel, I simply tried flashing the pdaXrom C760 kernel onto the C750 and voila, it worked!
So as a test, I tried flashing the pdaXrom *C760* kernel into my SL-C750 in the maintenance kernel location (0x220000), and it worked - I was able to create and mount a full 57MB jffs2 partition.
So, to get a SL-C750 to work with a single root partition, an extra step needs to be done - the maintenance kernel needs to be upgraded with the pdaXrom kernel *first* - then do a reboot, then do the mainte.bin patch to repartition the MTDs for single root, then reboot, then install the pdaXrom *C760* kernel (NOT the C750 kernel), and finally nandcp the filesystem to mtd2 and final reboot.
The offset for the mtdtweak/mtdparts patch in the C750 is 0x20aa4, so if you change the MTDCMD= assignment in your script to to \"MTDCMD=0x20aa4\" then the mtdparms= patch will work.
It is important to note that I\'m not sure if there are any issues with using the pdaXrom C760 kernel for both maintenance and main kernels on the SL-C750 - however the device appears to work fine with it.
If you update your version of the updater.sh script to do all of this, and then post it then I can test it for you!
I\'m trying a \"c+d\" CF recover now to un-brick it...
Did you succeed doing this too? Fabulous! Maybe you can post a howto for Foxdie to this topic: http://www.zaurususergroup.com/index.php?n...iewtopic&t=2897 (https://www.oesf.org/forums/index.php?showtopic=2897) he seems desperate.
Actually everything you need to do the C+D trick is at http://pocketworkstation.org/files/recover/ (http://pocketworkstation.org/files/recover/). It is very handy, you can completely recover your device with this process.
-
Hi everybody!
I\'ve just rebuit my installer to work properly with 750s, thanks to dvdrw\'s headaches. :wink:
To use it, just substitute installer.sh from pdaxrom, and add mtdparts & mtdtweak to your flash card, like prior version.
Zaurus sl-c750 need to update their maintenance kernel, as we have discussed in this topic, in order to be able to mount *big* jffs2 partitions. So, the first option you must select from the installer menu is \"4 Update SL-C750 Maintenance Kernel\". Don\'t worry, it makes a backup copy of you original maintenance kernel into your flash card (named mkern750.bak); if you want to get back, option \"5 Restore SL-C750 Maintenance Kernel\" restores it when ever you want. The rest steps are as prior version: resize to only root and install, which can also be reversed as you now. I try to write the most careful code I can, \'cause we are playing with fire :wink:
Reply for bugs, congratulations, insults...
The new installer is only available in the following attachment.
PS: Thank you again ikm and dvdrw, this won\'t be possible without you
-
Reply for bugs, congratulations, insults...
Hi Ashley,
Thank you very much for putting this together! I haven\'t had a chance to try it yet, but in checking the update script source I think there is a little problem - on the 750 you have to run the \"main\" kernel with the Cacko 760 kernel as well (!), as their 750 kernel has the same jffs2 allocation problem as the default Zaurus 750 kernel.
So at around line 437 in your current updater.decoded, you should change this line:
for TARGETFILE in /tmp1/tools/kernel-$MODEL/zImage.bin initrd.bin INITRD.BIN
to perhaps something like this sequence:
if [ "$MODEL" == "SL-C750" ]; then
KERNEL_IMAGE=kernel-SL-C760
else
KERNEL_IMAGE=kernel-$MODEL
fi
for TARGETFILE in /tmp1/tools/$KERNEL_IMAGE/zImage.bin initrd.bin INITRD.BIN
Again I haven\'t had a chance to try this yet, but will do this tonight.
BTW, it would be nice if Cacko could fix their 750 kernel to include the jffs2 fix directly. Is there any way to let them know about this?
Thanks again!
-
hi Ashley,
I\'ve made some necessary changes in your new installer for the C750 single-root install (as well as a few small cosmetic changes). I have attached the updated version. I have tested this version on my C750 by NAND restoring back to the factory Zaurus ROM, then doing a clean install using this updater with the Cacko 1.0.5 initrd.bin and tools.tar. I\'ve also tested your maintenance ROM backup/restore operations and they work great. so many many thanks again for your work on this!!!!
p.s. as there is no longer any mtd3 it is possible to eliminate the bootup warnings about mtdblock and mtdblock3 not found by commenting out the relevant statements from /etc/rc.d/rc.rofilesys (e.g., RW_MTD_LINE references, etc)
-
Hi dvdrw!!
Sorry for the delay in posting, now my job keeps me much more busy :?
Thank you very very much for testing and making necessary mods for me!!
It\'s true, I forgot to make the changes about non-maintenance 750 kernel
p.s. as there is no longer any mtd3 it is possible to eliminate the bootup warnings about mtdblock and mtdblock3 not found by commenting out the relevant statements from /etc/rc.d/rc.rofilesys (e.g., RW_MTD_LINE references, etc)
Also, you should remove mtdblock3 line from /etc/fstab :wink:
I\'m very happy it works, and that you tested and corrected it!
And, people, DARE trying it! There could be still bugs (surely for 700, not yet tested), but for 860/760/750s seems to work perfectly. So don\'t be afraid. The installer is very careful before anything is done. And the beneficts for me are obvious: I have many many apps installed just in my nand and have still a lot of free space!
Maybe, I should remember that after installing pdaXrom with this installer you\'ll get 100MB free nand space on a x60, which can be much more thanks to jffs2 zlib transparent compression. As an example: in my z, df is saying I have 60MB used, 60MB free (!!) and \"du /\" says I have 105MB used !!!!! (no cards in) Should I say anything else...?
-
I\'ve just realized that attachments are NOT shown if you are not logged as a user! So, to be able to see attachments and download them you must log in!
-
BTW, it would be nice if Cacko could fix their 750 kernel to include the jffs2 fix directly. Is there any way to let them know about this?
I will.
-
Ashley, just tried your pdaXrom patch and it worked perfectly on my SL-760! This is what the ROM was missing... since I prefer to install the apps internally and not on the slower SD / CF cards which are better suited for data storage. now I finally do not have to limit what applications I install. Thanks again for your great contribution to the pdaXrom project!
-
Thanks, OpKnight. I thought so, it was a pity to waste the nand due to partitions and limits. I thought it even before installing the very first version of pdaxrom (2003 cacko x11 roms mounted root as a compressed loop, just like qtopia, and it was needed, but not now). And I agree with you, nand is muuuuch faster than flash cards...
-
is it possible to that for the SL5600 for the default rom? or am I missing something?
-
is it possible to that for the SL5600 for the default rom? or am I missing something?
That should be possible, the internals of 5600 are very similar to C devices.
BTW, how\'s your ROM port project? Did you try to flash your 5600?
-
euh... I haven\'t flash it, I\'ve been playing around with the qt3 library, I\'ll try to fry my Z this afternoon.
Sorry for the delay maslovsky... ;-)
-
Has anyone else ever tried this on a C860? For me the new installer stopped with the message:
Flash pattern NOT recognized. Cannot patch flash.
I browsed a little through updater.decoded to figure out what went wrong and eventually found that if I change MTDCMD from 0X209F4 to 0X20A0C it would work, and in fact it did so without bricking my precious. I did this from a shell prompt. Is there any documentation about things like the scrambled updater.sh format or the boot process in general?
I assume that an almost full partition is bad for wear levelling, all writes can only be distributed among the few free blocks, so that\'s another good reason to have only one partition.
-
Has anyone else ever tried this on a C860?
Not yet, I presume. You seem to be the first...
For me the new installer stopped with the message:
Flash pattern NOT recognized. Cannot patch flash.
Fortunately, I wrote a pre-patch check routine to ensure that can be done without risk :wink:
I browsed a little through updater.decoded to figure out what went wrong and eventually found that if I change MTDCMD from 0X209F4 to 0X20A0C it would work, and in fact it did so without bricking my precious.
As there are so few differences between 760 and 860, I didn\'t thought this to be one, but actually it is. Thank you for testing, investigating and posting the actual 860 mtdparts offset! I\'m going to update this on the installer, very easy.
Is there any documentation about things like the scrambled updater.sh format or the boot process in general?
I posted source from endecsh, the encoder-decoder app, in previuos downloads. I don\'t know why Sharp made this silly encoding to installer.sh. The boot process is somehow described in several topics, this one has some info we know as well, but I don\'t know any complete reference (wiki, howto...)
I assume that an almost full partition is bad for wear levelling, all writes can only be distributed among the few free blocks, so that\'s another good reason to have only one partition.
You must be right , that is something I never thought :wink:
-
At last I found a few minutes of free time...
Here is the pdaxrom modified installer fixed for 860s (just added 1 line :roll: )
Ash
-
Hello Ashley,
I have just tried the new_installer-0.3.tar.gz but it says \"flash pattern not recognized. Cannot patch flash.\" I used the pdaxrom 1.05, copied the content of your first patch over it and updated that with the content of new_installer-0.3.tar.gz.
Did I miss something or is there something wrong with the new_installer-0.3.tar.gz?
Thanks for your effort.
Regards,
Fabian
-
In the updater after the error message was issued, ecape to the shell and compare the contents of mtdnand and mtdtweak. The error means that the installer got the MTDCMD offset wrong, and from the position of a common string in mtdnand and mtdtweak you should be able to calculate the correct offset. If you are adventurous (brick risk), you can then edit the offset in updater.decoded and run the updated script directly from the shell prompt. Post the new offset and your exact model specs. Good luck.
-
Hi!
Fabian, what model do you have? As far as I know nobody has tested this installer with a 700. In the rest japanese models it works.
Have you tried what Zazz says? If so, can you post and attach mtdnand contents or your actual offset?
-
Hello Ashley,
I forgot to mention that I have a 860 (that\'s why I tried your third patch). As far as I understand the other postings it should work. I haven\'t tried what Zazz said since I don\'t want to take the risk of bricking my precious unit :-)
Greetings,
Fabian
-
Strange.
Zazz, can you test the installer 0.3 in your unit? (Edit it to just check if you want, without installing ) I just added a line wich changes MTDCMD to the offset you told me. Maybe there are two 860 versions?
Fabian, can you post mtdnand anyway, without trying to patch your z? (you can dump it even with a running system)
-
Zazz, can you test the installer 0.3 in your unit? (Edit it to just check if you want, without installing ) I just added a line wich changes MTDCMD to the offset you told me. Maybe there are two 860 versions?
I installed with the new installer (0.3) on my C860 and it worked without
problems.
Himchen
-
I installed with the new installer (0.3) on my C860 and it worked without problems.
Thanks for posting this, so I am not alone with my C860 having the MTDCMD offset that it has.
Maybe we should package some minmal and unobtrusive test script together with /sbin/nandlogical that does a pure read-only test and that can be run directly without booting into maintenance mode to have more people confirm their offsets without risk.
-
I installed with the new installer (0.3) on my C860 and it worked without
problems.
Good news. Congratulations!
Maybe we should package some minmal and unobtrusive test script together with /sbin/nandlogical that does a pure read-only test and that can be run directly without booting into maintenance mode to have more people confirm their offsets without risk.
Good idea! I agree. Maybe this weekend I got some time :wink:
-
Hello Ashley,
how do I get the mtdnand values? I executed mtdtweak, the result is sharpsl-nand:7168k@0k(smf), is this the number you are looking for?
Regards,
Fabian
-
great work Ashley!
thanks.
work as a charm in my 760.
piotr
-
Hi everybody!
At last I\'ve found some time to write and test a small utility to test if the kernel command line is recognized before trying modified installer. This tool works in any (I supose) running rom, not in maintenance mode, and does not modify anything on z. So, you can safely test it
The testing routines are the same I wrote for modified installer, so if they succeed, you should not have any problem to use modified installer.
To use it, just untar to your z and run mtdtest.sh.
It will create a dump file (80 bytes) of your mtdparts kernel command line called mtdnand. If testing fails, plase report attaching it to your message, this will help me a lot :wink:
Ash
-
OK, I tried it on my 860 w/o success. The file doesn\'t want to stay attached, the contents are below:
nd:7168k@0k(smf),k@7168k(root),-(home) ZAURUS_RESET=HARD jffs2_orphaned_
-
OK, I tried it on my 860 w/o success. The file doesn\'t want to stay attached, the contents are below:
nd:7168k@0k(smf),k@7168k(root),-(home) ZAURUS_RESET=HARD jffs2_orphaned_
The problem is that if you paste it as text, I can\'t see nulls. However, it seems your offset is different from other 860s!! It is 24 bytes less than 0X20A0C. Can you try to change offset to 0x209F4? (You just have to edit mtdtest.sh, line 23)
-
The 760 setting worked, good call. Any idea why?
-
Sharp Zaurus SL-C760 detected.
Testing mtdparts kernel command line from your nand...
Dumping data starting at 0x000209f4 and ending at 0x00020a44...
Dumping 209f4 - 20a44
mtdparts recognized!
You can use modified installer to install pdaXrom in a single root partition!
-
Hi,
I am using Your modification from few days and it work great!
I discover that after installing ipk or at boot time I am getting such a message:
ldconfig: /usr/local/lib/libXp.so.6 is not a symbolic link
ldconfig: /usr/local/lib/libXp.so.2 is not a symbolic link
Everything works ok but I just want you to ask it is somehow connected to your modification?
It is a way to fix it or it is ok as it is?? :-)
Piotr
ps. What does your latest file?? If Ive installed pdaxrom on one partition should I check something with that script??
thanks
-
The 760 setting worked, good call. Any idea why?
The differences between 760 and 860 are very very few. Maybe early 860s are just 760s with a different kernel (for USB storage) and latest (another release, i presume) include more changes.
Then, it\'s easy to use modified installer. You can edit last version to change MTDCMD offset (and encode it to make usable as installer) or use version 0.2, which as a matter of fact was changed to suit (some) 860s.
If this happens to more people i can modify installer again to automate this thing.
Ash.
-
piotr, I don\'t know why this happens to you. I\'ve not seen that on my z, and don\'t think it is related to have just the root partition. Also, I don\'t know how to fix it :wink:
ps. What does your latest file?? If Ive installed pdaxrom on one partition should I check something with that script??
thanks
It is just a testing tool for people who are afraid to run installer directly :wink: It runs in normal mode, not maintenance. If it succeeds, you can *safely* use modified installer.
Ash.
-
Thanks!
:-)
-
Just reflashed my C860 with pdaXrom and a single root partition, using updater v0.3. It worked like a charm. :-D
Would be awesome if this will become part of the standard pdaXrom.
btw, when is the new release due? Looking forward to it *jumps up and down*
-
I downloaded the version from pdaxrom.org... does anyone know what I am supposed to do with the .bin file that came in the zip?
Thanks!
-B
-
Sorry - a careful re-read of the forum posts answered my question... just put it in with the other s on the cf....
Thanks!
-
Welp that worked well - I booted as to flash with the small updater and .bin file in there - let it do its thing - rebooted - flashed to 121mb root and then reinstalled the pdaxrom - I have also remounted root on the sd card. Had 100mb free after the os - nice... now everything fits!
have a good weekend folks!
-Brian
-
Uh, thats remonted HOME on the SD card... left the old /home in case the sd card is out - and just mounted the sd card as /home right over it...
-B
-
gandalf -- nm = new mexico?
Scott
-
yup NM as in new mexico...
-B
-
Hi
Sorry, stupid newbie question. I just can't see the attachments, although I am logged in. Already tried it in safari (macos x), mozilla, ie, firefox, no go. Where are they supposed to be downloadable?
(thnx ashley et al, dying to try this)
Eric
-
I have same question. Is anybody possible to make installer's last version avalaible?
-
Same question here...
-
Is there anyone who has the current installer (v 0.3)? Please put it online...
-
Is there anyone who has the current installer (v 0.3)? Please put it online...
I still have a copy. Use at your own risk.
Edit: Hmm, the stupid board does not seem to allow the upload of files that are not images ("You cannot upload this type of file"). I've put a copy here (http://zaurus.zapto.org/zazz/ashley/new_installer-0.3.tar.gz).
-
Big thanks, zazz. I hope the risk is a usual one caused by bad testing but not a result of troyans you've put. 8)
-
Thanx! now I have 121 MB root!
-
Wow!!
I did NOT expected so much traffic here! Yes, I'm a bit offline lately...
Sorry all who tried to use the vanished modified installer. I can swear it was there before zug changed. Thank you Zazz to put it online again! I got the same problem trying to attach it again
Congratulations to all who have run it successfully!
-
This installer rules! It's so much easier installing apps now... Thanks Ashley!
-
yup NM as in new mexico...
-B
where in NM? I lived in Santa Fe for 10 years... just moved to NYC in 2001.
Scott
-
thnx zazz and ashley
-
Has any one tried this on a C700?
I would try it myself, but do not want to take a chance of bricking my Z.
-
Hello all, I am new to Linux and have a CL-750 about 6 months old. I have read this entire post and I am still confused. I want to get the latest ROM updates and want to start to learn Linux, but am not sure where to start. In the d/l section there are several ROM's available. How do I know which one to use? My current ROM is 1.10 JP. Hopefully since I am new I get 30 days of free stupid questions?????
I also posted in the newbie section.
-
Has any one tried this on a C700?
Nobody reported it
I would try it myself, but do not want to take a chance of bricking my Z.
I think the installer is mature. None has claimed about it . And it takes very care before doing anything. Yourself. Also, I can mail you the testing tool, useful to check if everything is sane.
I want to get the latest ROM updates and want to start to learn Linux, but am not sure where to start.
I suggest you to try other apps & roms to get practice. pdaXrom isn't hard, nor is my simple modified installer, but they are *not* targeted to Linux newbies
-
Hello all, I am new to Linux and have a CL-750 about 6 months old. I have read this entire post and I am still confused. I want to get the latest ROM updates and want to start to learn Linux, but am not sure where to start. In the d/l section there are several ROM's available. How do I know which one to use? My current ROM is 1.10 JP. Hopefully since I am new I get 30 days of free stupid questions?????
I also posted in the newbie section.
Well, in short, to get you started: there are a variety of roms, and they include:
1) the original SHARP rom, several versions available (just use the latest). That's a rom that comes with the device preinstalled, and it is the only official rom for the Zaurus. It is based on the Qtopia environment, which is made for the mobile devices and is not used on the regular desktop linux installations. There are many third-party programs available for it. Unfortunately, as the most mainstream linux software was not made for the qtopia, it is not feasible to run that software on this rom.
2) several flavours of the modified SHARP rom. They differ in the number of packages preinstalled, fonts and other stuff. All software for the original rom will work on these ones.
3) the OpenZaurus rom, based on the opie environment. Opie environment is based on the qtopia codebase, but has its own paths of evolution. To put it short and stupid, the OpenZaurus rom is an effort to create something that barely resembles the original SHARP rom, but with the open source license. This openness means that all the source codes are freely available for distribution and modifications.
4) this pdaXrom project. Unlike all other roms, the idea here is to turn the Zaurus to the linux subnotebook. It runs X server for graphics (hence the name), and all graphical applications are the regular X clients. This is the most important thing about it, as all regular linux desktops work the same way. This has great merits, as it makes possible to run the majority of all desktop applications on the Zaurus effortlessly, taking account of the device limitations and subtleties, of course (3.7" 640x480 screen, touchscreen instead of mouse etc). The only thing that is needed in order to get a program running is a mere recompilation, maybe with some tweaks.
5) Debian on Zaurus projects, there are several. Also Gentoo for Zaurus. The idea is to run these regular linux distributions on the Zaurus. As these distributions are not tailored for this task, the whole experience is quite extreme to say the least, suitable for the geekish soft of people only
So you have a variety of options to choose from. If you want to learn linux, pdaxrom seems to be quite adequate. But it certainly is not for neebies. On the other hand, in case you have been using the device for 6 months already, maybe it is a time to give it a try? It is you who decide
-
can we post a "summa" of the "just root" setup?
files needed, instructions, for c860 and other zaurii.
I'm a bit confused
-
Is it possible to get the mtdtest.sh?
John
-
Quick post
I've put vanished stuff here (http://es.geocities.com/vrsgazquez/).
First of all, you can use testing tool (mtdparts_test.tar.gz). Just decompress into your running rom and exec ./test/mtdtest.sh. Report any problem or success.
If testing is ok, you should be able to use modified installer. Follow the instructions from pdaxrom site to prepare your flash card. Before rebooting your z, put files from new installer to your root dir of your flash card (overwrite updater.sh). Reboot in maintenance mode, as told by pdaxrom. Patch mtdparts with 2n option "NAND Flash repartition: only Root". If you have a 750 you should also update maintenance kernel. Reboot in maintenance mode. Install new rom with option 1. Reboot. Voilà .
The only known issue is with some 860s, wich are intermediate versions between 760s and later 860s. If it's your case (testing and new_installer-0.3.tar.gz will safely warn) you should use new-installer-fixed.tar.gz.
I think it's all
-
nice feed Ash
anyway, you can give those files to pdaXrom crew to let them integrate your procedures with the installation they will release. In this way the new pdaXrom user will immediatly see the possibility of a "just root" setup, without the need of reflashing after your thread/feed/files discovery
-
hum for some reason it keeps comming up as pdaxrom 1.0.0 what am i doing wrong?
-
hum for some reason it keeps comming up as pdaxrom 1.0.0 what am i doing wrong?
You're likely doing everything right, as the pdaXrom 1.0.5 still advertises itself as 1.0.0 (the version number was forgotten to be increased). Check if the on-the-fly rotation works -- if it does, you've got 1.0.5, cause the autorotation was introduced in 1.0.5
-
ah indeed it is true.
Thanks for writing that, its nice having all that space to play with while i get used to my 860 and figure out exactly what apps i want installed.
-
Ok, guys, i need your help...
I have messed up my c750 so i have to reflash it, and i will soon leave home for my hollidays, so i'm not sure i'll be able to wait until next pdaxrom release... Yet i want to use this wonderfull one partition script but i am not sure about this step :
If you have a 750 you should also update maintenance kernel. Reboot in maintenance mode.
Can someone help me ? I would really need a step by step help as i don't understand anything to the japanese menu...
Thanks in advance.
-
ok, sorry, i'm just stuuuuuuuuuuuupid...
nothing had to be done in the japanese menu except the normal procedure for reflashing...
The all updating and repartitioning thing comes in english with the rom installation procedure, so it is really not big a deal.
Ashley : your tool is WONDERFUL !!!!!!!!!! No install pb, no disfunctioning, no bug... i have a C750 and i flashed pdaxrom 1.0.5 that has now 58 368 ko to run... Thank you !!!!!
-
You're welcomed!
-
First of all, you can use testing tool (mtdparts_test.tar.gz). Just decompress into your running rom and exec ./test/mtdtest.sh. Report any problem or success.
If testing is ok, you should be able to use modified installer. Follow the instructions from pdaxrom site to prepare your flash card. Before rebooting your z, put files from new installer to your root dir of your flash card (overwrite updater.sh). Reboot in maintenance mode, as told by pdaxrom. Patch mtdparts with 2n option "NAND Flash repartition: only Root". If you have a 750 you should also update maintenance kernel. Reboot in maintenance mode. Install new rom with option 1. Reboot. Voilà .
I finally tried to install this on my C700. The testing tool reported "success." When I tried to "NAND Flash repartition: only Root," I got an error message "Flash pattern NOT recognized. Cannot patch flash."
What should I try next?
-
Actually, I am not sure that the testing tool ran correctly. If I type "exec ./test/mtdtest.sh" at the console command line, the first line I get is :
:bad interpreter: No such file or directory
the next line is:
bash: /mnt/cf/test/mtdtest.sh: Success
Does anyone have any suggestions?
-
Hello,
got the same problem on my c 700
do i have to edit my updater.sh
or will it not work under c 700 ?
thanks,
Denis
-
Hi!, good summer!
I don't know anything about this error message. Never happened with a non-700. All this stuff remains untested on 700, you seem to be the first ones . Maybe bcut is not working properly. It is just used to compare (into a loop) mtdnand with mtdparts and mtdtweak files. You can compare them by any other app or even by hand (just 80 Bytes).
I don't know if c700 kernel command line is just offseted from other z or is not compatible with this tweak. I will be able to investigate if you send me your mtdnand file
Ash.
-
Ashley, the other problem is that the C700 kernel probably needs to be patched and recompiled to work with 64Mb large jffs2 partitions. That could be easily done, the only problem is that I don't really know what exact kernel patches are applied by the pdaXrom developers to their kernels.
-
Woah, I'm sure this has been thought of but I just want to query this and introduce a word of caution.
If you start the image installer process (using the OK button from a cold start) option 3 lets you format the 'User Partition'.
An (already installed) traditional ROM then has the job of re-populating this to a usable state - this is the traditional way of resetting one of these units to a virgin state.
Does anyone know what happens if you select this having repartitioned the NAND ? and what happens if this option can't find the User partition ?
It could possibly brick at worst case.
BTW: While we are at it I'm assuming that the Service Menus are in ROM and not FLASH but is this Japanese utility menu the same or does this class as a flash based bootloader ? If so then it would need patching to avoid the possibility of this issue.
You could argue that if you have gone to the effort of partitioning the flash like this then you are in an 'informed position' and wouldn't try this option. There again what happens if something 'breaks' that would be possible to fix in a service depot and you send it to some poor hapless dude whose first procedure is to wipe the user partition.
Hmm.. you may get back a brick and be charged for wasting time.
Wouldn't a safer option be to just make the initial root partition as small as possible, carry only the kernel, some minimal scripts and busybox that on an erase would set up the user partition as the real root. On a normal boot the root would pivot the root mount onto the user partition and remount itself in perhaps /boot ?.
This seems safer and keeps the partitioning scheme cool with the loader menus etc. It also means that you are still free to install almost everything including base stuff from IPK if your distro allows it.
-
Woah, I'm sure this has been thought of but I just want to query this and introduce a word of caution.
If you start the image installer process (using the OK button from a cold start) option 3 lets you format the 'User Partition'.
An (already installed) traditional ROM then has the job of re-populating this to a usable state - this is the traditional way of resetting one of these units to a virgin state.
Does anyone know what happens if you select this having repartitioned the NAND ? and what happens if this option can't find the User partition ?
It could possibly brick at worst case.
I don't really think it would brick the device. It would possibly screw up while trying to perform the thing, but I highly doubt it would trash a reserved service/utility part of the flash.
BTW: While we are at it I'm assuming that the Service Menus are in ROM and not FLASH but is this Japanese utility menu the same or does this class as a flash based bootloader ? If so then it would need patching to avoid the possibility of this issue.
As far as I understand it, both service menus and the Japanese utility are in flash. You can erase them, you can patch them, you can do whatever you want with them. But there is always a possibility to bootstrap the device, as the very basic first-stage bootloader resides in ROM and has a built-in possibility to restore the flash. There is an instruction for doing that, the link on it was posted on this board quite a while ago, but I think it's possible to find it. Or maybe someone could post it here if he has it saved.
You could argue that if you have gone to the effort of partitioning the flash like this then you are in an 'informed position' and wouldn't try this option. There again what happens if something 'breaks' that would be possible to fix in a service depot and you send it to some poor hapless dude whose first procedure is to wipe the user partition.
Hmm.. you may get back a brick and be charged for wasting time.
Well, in the worst case scenario, a real brick can drop from the rooftops and brick you on a way to service ;) My point is that thinking about that is a bit over the top. And well, I guess someone with the experience of the lowest-level bootstraping could just try that out :)
Wouldn't a safer option be to just make the initial root partition as small as possible, carry only the kernel, some minimal scripts and busybox that on an erase would set up the user partition as the real root. On a normal boot the root would pivot the root mount onto the user partition and remount itself in perhaps /boot ?.
This seems safer and keeps the partitioning scheme cool with the loader menus etc. It also means that you are still free to install almost everything including base stuff from IPK if your distro allows it.
Well, in my opinion, while I think it would be a small tiny bit safer, it would not pay off a single megabyte of the precise flash space wasted. Just relax, there are much more pressing issues on the agenda to mediate about :)
-
Ashley, the other problem is that the C700 kernel probably needs to be patched and recompiled to work with 64Mb large jffs2 partitions. That could be easily done, the only problem is that I don't really know what exact kernel patches are applied by the pdaXrom developers to their kernels.
You are right. And it may not be as simple as the trick used with 750s (using the 760 kernel). Recompilation may be needed. The worse is I don't own a 700, so I can't play this game. Any volunteer?
As far as I understand it, both service menus and the Japanese utility are in flash. You can erase them, you can patch them, you can do whatever you want with them. But there is always a possibility to bootstrap the device, as the very basic first-stage bootloader resides in ROM and has a built-in possibility to restore the flash. There is an instruction for doing that, the link on it was posted on this board quite a while ago, but I think it's possible to find it. Or maybe someone could post it here if he has it saved.
http://pocketworkstation.org/files/recover/ (http://pocketworkstation.org/files/recover/)
Of course I agree with you, ikm. Nice post , as usual
iamasmith, I wrote a simple partition restoration to the origianl root+user scheme in the installer, so you always have the choice to go back by rebooting again into the installer. So don't worry so much. Installer was written months ago, and I don't know how many people has tried it, but no complaint received, only congratulations... and fears
-
OK, paranoia kicking in I guess.
I wouldn't design a system that you couldn't restore even if all the flash was gone so I guess why should Sharp.
I have, however, worked in 'the industry' a long time and have learned to be cautious - this helps to prevent those 'cold sweat' moments that people end up with.
Good info on the reflash stuff, thanks. Not so encouraging that the 860 archive is marked 'probably_not_working' though guess which unit I've got...
Anyway, thanks for the response. It sets my mind at rest a little on this one.
-
Ashley, the other problem is that the C700 kernel probably needs to be patched and recompiled to work with 64Mb large jffs2 partitions. That could be easily done, the only problem is that I don't really know what exact kernel patches are applied by the pdaXrom developers to their kernels.
You are right. And it may not be as simple as the trick used with 750s (using the 760 kernel). Recompilation may be needed. The worse is I don't own a 700, so I can't play this game. Any volunteer?
I could handle the fix and recompilation, but I would need the precise pdaXrom kernel patch they apply, and their .config file. The result would most probably work without any testing at all, but only if both the patch and the .config are right.
As far as I understand it, both service menus and the Japanese utility are in flash. You can erase them, you can patch them, you can do whatever you want with them. But there is always a possibility to bootstrap the device, as the very basic first-stage bootloader resides in ROM and has a built-in possibility to restore the flash. There is an instruction for doing that, the link on it was posted on this board quite a while ago, but I think it's possible to find it. Or maybe someone could post it here if he has it saved.
http://pocketworkstation.org/files/recover/ (http://pocketworkstation.org/files/recover/)
Of course I agree with you, ikm. Nice post , as usual
Thanks
-
Well, I'm getting some interesting results trying to modify my 750 to have a singe root partition. I've got RC2 successfully flashed with 45 mg on mtdblock2 and 12mg on mtdblock3 (/home).
When I try and reflash with Ashley's "new-installer-fix" I got this error flash pattern not recognized I got this error before and after using the update maintance kernel (I restored it after it didn't resolve error)
Next I rebooted back into pdaXrom and tried the mtdtest script and got the error open flash:No such file or directory and the rest of the script fails.
So in looking at the script I see that LOGOCAL_MTD=/dev/mtd1 However in looking at /dev/ there is no mtd1, only mtd. So I modified script to LOGOCAL_MTD=/dev/mtd and then ran it with the following error GETMEMINFO: Inappropriate ioctl for device
Since the script fails there is no mtdnand file to examine.
Any ideas Ashley?
Thanks,
Jerry
-
So in looking at the script I see that LOGOCAL_MTD=/dev/mtd1 However in looking at /dev/ there is no mtd1, only mtd. So I modified script to LOGOCAL_MTD=/dev/mtd and then ran it with the following error GETMEMINFO: Inappropriate ioctl for device
Try using not just /dev/mtd, but rather /dev/mtd/1.
-
Thanks ikm, that worked. It created the mtdnand file and of course it didn't recognized the pattern. Reading the mtdnand hex is beyond my little brain.
Cheers,
Jerry
-
Jerry, that mtdnand looks jerky. It makes no sense to me. How did you attached it?
BTW, I haven't tried RC2 yet...
-
Ashley:
I ran mtdtest from my Cf card and just put my Cf in my card reader and uploaded directly from there.
Thanks,
Jerry
-
Ashley:
Is the output of mtdtest supposed to be hex or ascii? If it is ascii I must have really borked my internal memory somehow or perhaps the binaries from your mtdtest got corrupted somehow in the download or extraction process. I'll try redownloading again and run again. If the output is the same, would a NAND restore from a SYSC750.BKD possibly fix the problem?
Thanks for your help.
Jerry
-
Ok, retried everything and this file looks different in hex than the previous file (but it's still Greek to me)
Thanks,
Jerry
-
with a SL-5600:
Dumping data starting at 0x00020aa4 and ending at 0x00020af4...
Dumping 20aa4 - 20af4
mtdparts flash pattern NOT recognized. You cannot use modified installer :-(
Please report. Attaching mtdnand file will help a lot
-
Hi There, just tried that installer and it really worked. But
now i got a damn problem:
The Install of pdaXrom with the new installer worked, i made the
backup to the 760-kernel, but forgot to set the partiton-size from
45MB to 57MB. And now...
...i cannot flash anymore. I´m able to reach the Flash-Menu, i can
choose SD, but after that he just starts pdaXrom. pdaXrom itself runs
smoothly, but since i forgot to set the partition on 57MB i´m out of space,
and that´s why i want to reflash.
Even so i cannot go into the Servicemenu, cause the "D"+"M"-Method
don´t work anymore!!!
Need really really help!
Nico
P.S.: Tried that with my SD-Card (pdaXrom-install-files on it), and with my
CF-Card (Cacko1.22-install-files on it), both didn´t work.
-
...sure you have ac attached, haven't you? and kept out the battery for at least 5 sec? If not, that could be a point; anyway, it's too simple to be the case - just starting help by the basis
-
good points at last, never underestimate anything
But tried it again carefully now, with eyes, ears and mind open
...still nothin...
[EDIT] But if it´s a help, i can describe the screen i get from pdaXrom.
It doesn´t boot completely, i´ve got the blue pdaXrom-screen at top,
under that a few pages text going through so fast that i can´t say what
it was. But it halted now, and the last line is:
INIT: no more processes left in this runlevel
and that´s it. I think he has all modules loaded necessary for pdaXrom
to start, i read something about "usb", "cramfs", "net_fd", "vendorID" and
so on...
[EDIT2] and what the hell happened that i can´t access the servicemenu???
-
Hi,
where to find all information about nandlogical and boot-process?
nandlogical:
1) the second argument always is READ or WRITE?
2)if I copy from device mtd1 at address 0x70000- what I get?
(or maybe full map of /dev/mtdX ???)
3)which mtd's are logical devices and which aren't?
4)Is always secure (for zaurus) to test with nandlogical when reading?
(I have never used nandlogical yet)
boot-process:
where to find full description?
where are located kernels (Sharp and this one from Cacko)-
maybe it is possible to mount /dev/mtdx and copy Sharp's kernel to cf card?
(I am using Cacko 1.22 on C860)
Thanks in advance
Janusz