08-17-2012, 09:13 AM | #61 |
Carpe diem, c'est la vie.
Posts: 6,433
Karma: 10773668
Join Date: Nov 2011
Location: Multiverse 6627A
Device: K1 to PW3
|
The amazon GPL source code contains patch files to be used if you replace any source files with unpatched versions. Perhaps some of those patches may be needed when using newer source code too (with suitable modification).
Before downgrading to an older version, it may be worth seeing if one of amazon's patches makes the newer version work. |
08-17-2012, 01:23 PM | #62 |
BLAM!
Posts: 13,492
Karma: 26047188
Join Date: Jun 2010
Location: Paris, France
Device: Kindle 2i, 3g, 4, 5w, PW, PW2, PW5; Kobo H2O, Forma, Elipsa, Sage, C2E
|
@knc1: And to expand on that, GCC's auto-vectorization isn't enabled at anything lower than -O3, IIRC .
|
Advert | |
|
08-17-2012, 07:31 PM | #63 |
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
smbfs.ko for the 3 from the lab126 2.6.26 kernel, via the 2006q3 CS TC
Works with the 2006q3 CS TC as knc1 said. I feel this is probably the best route. I'm having trouble with the patches applying as the build directory in BR is "./built/linux-custom"
I have also amended it to do the: Spoiler:
in arch/arm/configs/imx35_luigi_defconfig but I'm still fumbling my way through using patch and BR currently requires a typed in 'fix' to reference the correct file. I'll RTFM - again. Anyways: the modules look much happier now: [root@kindle root]# insmod /mnt/us/modules/smbfs.ko [root@kindle root]# lsmod Module Size Used by smbfs 64800 0 Narry a moan [root@kindle root]# rmmod smbfs [root@kindle root]# strace -o /mnt/us/opt/insmod.trace insmod /mnt/us/modules/smbfs.ko Spoiler:
[smbfs.ko for the 3 from the lab126 2.6.26 kernel, via the 2006q3 CS TC] So I will be using that TC for my testing on the arm of 'USING THE RIGHT HAMMER FOR THE JOB' The link for the 2006q3 TC is available on the KEK project tools repository. That's where I got mine. I'll update the BR - "doing it like an Amazon" - guide when I have it slicker and solid. but in essence it still works. I will update the Best Practice choice of tools now. Hopefully these Best Practices will help to form an accessible reference for the future digital hitch-hiker. Insert requests for custom modules here. Thanks. I'll remove the binfmt_misc.ko in the OP just in case... Last edited by twobob; 08-17-2012 at 07:40 PM. |
08-17-2012, 07:51 PM | #64 |
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
cifs.ko for the 3 from the lab126 2.6.26 kernel, via the 2006q3 CS TC
[root@kindle root]# modinfo /mnt/us/modules/cifs.ko
Code:
filename: /mnt/us/modules/cifs.ko author: Steve French <sfrench@us.ibm.com> license: GPL description: VFS to access servers complying with the SNIA CIFS Specification e.g. Samba and Windows version: 1.53 vermagic: 2.6.26-rt-lab126 mod_unload modversions ARMv6 depends: srcversion: 55B503E37F4DFB20AA9BE56 parm: cifs_max_pending:Simultaneous requests to server. Default: 50 Range: 2 to 256 (int) parm: cifs_min_small:Small network buffers in pool. Default: 30 Range: 2 to 256 (int) parm: cifs_min_rcv:Network buffers in pool. Default: 4 Range: 1 to 64 (int) parm: CIFSMaxBufSize:Network buffer size (not including header). Default: 16384 Range: 8192 to 130048 (int) Code:
[root@kindle root]# lsmod Module Size Used by cifs 238340 0 binfmt_misc 8552 0 smbfs 64800 0 g_ether 21096 0 |
08-17-2012, 07:55 PM | #65 |
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
A quick eyeball comparison of what modinfo thinks about an internal module and a recreated one using the BR - lab126 method
[root@kindle root]# modinfo fuse Code:
filename: /lib/modules/2.6.26-rt-lab126/kernel/fs/fuse/fuse.ko alias: char-major-10-229 license: GPL description: Filesystem in Userspace author: Miklos Szeredi <miklos@szeredi.hu> depends: vermagic: 2.6.26-rt-lab126 mod_unload modversions ARMv6 Code:
filename: /mnt/us/modules/fuse.ko alias: char-major-10-229 author: Miklos Szeredi <miklos@szeredi.hu> description: Filesystem in Userspace license: GPL vermagic: 2.6.26-rt-lab126 mod_unload modversions ARMv6 depends: |
Advert | |
|
08-17-2012, 11:22 PM | #66 |
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
The new trimmed down /usr/bin built with BR (The ln -s type binaries are elided)
Spoiler:
|
08-17-2012, 11:36 PM | #67 |
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
[root@kindle sbin]# ./usb_modeswitch --help
Code:
* usb-modeswitch: handle USB devices with multiple modes * Version 1.1.2alpha (C) Josua Dietze 2010 * Based on libusb0 (0.1.12 and above) ! PLEASE REPORT NEW CONFIGURATIONS ! Usage: usb-modeswitch [-hvpVPmMrdHn] [-c filename] -h, --help this help -e, --version print version number and exit -v, --default-vendor NUM vendor ID to look for (mandatory) -p, --default-product NUM product ID to look for (mandatory) -V, --target-vendor NUM target vendor (optional, for success check) -P, --target-product NUM target model (optional, for success check) -C, --target-class NUM target device class -m, --message-endpoint NUM where to direct the message (optional) -M, --message-content <msg> command to send (hex number as string) -2 <msg>, -3 <msg> additional commands to send -n, --need-response read a response to the message transfer -r, --response-endpoint NUM where from read the response (optional) -d, --detach-only just detach the storage driver -H, --huawei-mode apply a special procedure -S, --sierra-mode apply a special procedure -O, --sony-mode apply a special procedure -G, --gct-mode apply a special procedure -R, --reset-usb reset the device in the end -c, --config <filename> load different config file -Q, --quiet don't show progress or error messages -W, --verbose print all settings before running -D, --sysmode specific result and syslog message -s, --success NUM check switching result after NUM secs -I, --no-inquire do not get device details (default on) -i, --interface NUM select initial USB interface (default 0) -u, --configuration NUM select USB configuration -a, --altsetting NUM select alternative USB interface setting |
08-18-2012, 01:30 AM | #68 |
Going Viral
Posts: 17,212
Karma: 18210809
Join Date: Feb 2012
Location: Central Texas
Device: No K1, PW2, KV, KOA
|
Here is one, well, one pair, of programs to add to the collection:
http://www.daemonology.net/bsdiff/ Should be easy to add-in (and push upstream) to Buildroot. The general use would be to build bsdiff for the host and bspatch for the target but the makefile isn't organized that way. So just build the pair for both host and target with two calls to GENTARGETS ( ... ) If you haven't added a package file to Buildroot before, for this one just copy one of the simplest ones you can find - text edit the various names. The GENTARGETS macro will write the actual makefile glue between Buildroot and the package's makefile. This package depends on one external package, libbz2 but I suspect that package is already defined in Buildroot. Upstream **might** want a patch included to change that CFLAGS line (which would over-ride the optimization level selected in Buildroot and also contains a LDFLAGS item. - - - - Oh, as I am sure you have learned by now - you can build any individual package in Buildroot with the correct "make ... ... ..." Voodo command. |
08-18-2012, 01:37 AM | #69 |
Going Viral
Posts: 17,212
Karma: 18210809
Join Date: Feb 2012
Location: Central Texas
Device: No K1, PW2, KV, KOA
|
Now that we have them (Samba and CIFS), I guess it is time to google up the correct config file, its contents, decide where to put it, decide what to export to M$-Windows (Linux, MacOSx), figure out a nice little patch to the firewall rules.
Maybe keep it safe and simple, with a read-only export of some part of the filesystem. Last edited by knc1; 08-18-2012 at 10:02 AM. |
08-18-2012, 01:45 AM | #70 |
Going Viral
Posts: 17,212
Karma: 18210809
Join Date: Feb 2012
Location: Central Texas
Device: No K1, PW2, KV, KOA
|
Another little chore . . . .
FAT-32 does not support symbolic links (required to get the symantics of *nix library sonames correct). There are (at least) two ways to deal with that: FAT{,16,32} does partially support hard-links (at least until you run dosfsck on it) ; The target library of the links could be copied to each of the less specific names (makes a large increase in disk usage, but we have 3Gbytes) ; or some combination of the above. It would probably be appropriate to do that in a custom, post-build, script. Decisions, decisions, decisions. . . . Last edited by knc1; 08-18-2012 at 02:08 AM. |
08-18-2012, 04:50 AM | #71 | ||
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
Quote:
this tells me that doubtless some underlying process is being invoked that supports this. possibly tar - tar -h, --dereference follow symlinks; archive and dump the files they point to --hard-dereference follow hard links; archive and dump the files they refer to EDIT: I looked around this issue some more and have experience created junction points / hard links in MS server application but AIUI FAT would see this as a file error (IIRC). It's Not impossible by any means: I found example code to create symbolic link support for FAT drives but it WOULD REQUIRE at the very minimum a windows related process on the filesytem (invoking wine might do it), there are also various MS utilities throughout the ages that could do this. BUT I don't think this is a good solution given the PITA and probably fschk complaints. There would need to be a good reason and IMHO I don't see one yet. In this case the 'rolled out' version of the file would probably be far less painful to manage. I am using these 'Rolled Out' copies successfully now. Quote:
cp the rootfs.tar to a temp. cp rootfs.tar rootfs.temp.tar me@dev ~/BLDS/kindle $ tar --delete -vf ./rootfs.temp.tar ./dev/ ./home/ ./media/ ./mnt/ ./opt/ ./proc/ ./root/ ./run/ ./sys/ ./tmp/ ./var/ ./linuxrc copy that onto your device in the root of mnt/us (or somewhere on the fat partition anyways) and run tar with --extract --dereference on it. I'll go check the exact local tar command as I don't use it (I extract via sftp) But it will be something like [root@kindle]# tar -xvf ./rootfs.temp.tar --dereference err no. busybox tar doesn't support that. I'll have another think. That process will get us the good stuff easily splattered into a nice tar. Last edited by twobob; 08-18-2012 at 07:38 AM. Reason: added thoughts (now I am out of the bath hehehe) |
||
08-18-2012, 07:08 AM | #72 |
Going Viral
Posts: 17,212
Karma: 18210809
Join Date: Feb 2012
Location: Central Texas
Device: No K1, PW2, KV, KOA
|
The way I had in mind, once discovered by dos-fsck, would result in identical copies under different names being created to "fix" what isn't broken.
So might as well just start out with multiple copies. Then one only must figure out a way to enforce: ldconfig -X ... Otherwise it would try to replace the copies with the symbolic links. That sort of thing is usually done by replacing /sbin/ldconfig with a shell script that calls the re-named ldconfig with option -X plus an edited version of the original arguments. Hmm... Now we are talking about a change to the filesystem that would break any OTA update when (if) /sbin/ldconfig was ever tested. Hmm... Unless we learn to always patch the md5sum manifest file when we change a system file. Yup, I forsee the need for that bsdiff/bspatch pair of routines mentioned above. |
08-18-2012, 07:37 AM | #73 |
Going Viral
Posts: 17,212
Karma: 18210809
Join Date: Feb 2012
Location: Central Texas
Device: No K1, PW2, KV, KOA
|
I don't think that busybox tar understands: --dereference
Check for that on your K3 - (mine is still torn apart). |
08-18-2012, 07:39 AM | #74 |
( ͡° ͜ʖ ͡°){ʇlnɐɟ ƃǝs}Tır
Posts: 6,586
Karma: 6299991
Join Date: Jun 2012
Location: uti gratia usura (Yao ying da ying; Mo ying da yieng)
Device: PW-WIFI|K5-3G+WIFI| K4|K3-3G|DXG|K2| Rooted Nook Touch
|
|
08-18-2012, 07:54 AM | #75 |
Going Viral
Posts: 17,212
Karma: 18210809
Join Date: Feb 2012
Location: Central Texas
Device: No K1, PW2, KV, KOA
|
There was a patch posted on the BR M.L. yesterday to fix the QT build.
In the top level Makefile, in the list of unexported symbols, add: unexport QMAKESPEC The list should be somewhere near line 209. |
Tags |
buildroot, compile, howto, kernal, kindle 3.3 |
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Linux | Titus | enTourage Archive | 6 | 09-07-2010 11:23 AM |
So as a Linux..... | tajreed | General Discussions | 13 | 04-05-2010 12:51 PM |
Linux? | Ron46 | Ectaco jetBook | 8 | 12-24-2009 08:07 AM |
Linux | coredump | Sony Reader | 12 | 11-17-2009 05:07 AM |
Linux Help | clewley | Bookeen | 3 | 04-04-2008 08:40 AM |