Frank Haverkamp [Sat, 24 Nov 2007 10:04:51 +0000 (11:04 +0100)]
ubi-utils: nand2bin had ECC calculation problems
Fixed a problem when ECC was checked. The correction was not properly
done by subpage. Added more output for the moment to be able to figure
out more potential problems.
Added testcase: bin2nand2bin.sh and biterror inject program inject_biterror.pl
Interface
o ECC correction disabled by default. Switch to turn it
explicitly on. The user must specify what he wants to be done.
Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
Frank Haverkamp [Sat, 24 Nov 2007 10:04:51 +0000 (11:04 +0100)]
ubi-utils: various fixes in unubi
The extraction of data from blocks used for dynamic volumes was
totally broken. The data size was calculated wrong. This fix is not
perfect, the alignment is still ignored.
The parameter "header-size" is very misleading. It does not reflect
the vid hdr offset properly. I assume therefor that it only works for
the layout I am using where the vid hdr is at the _end_ of the 1st
NAND page (2048).
I added the generation of a textfile with information about the blocks
which are going into the internal graph representation.
Instead of a graph I think that a simple array will simplify the code
very much. The array must than be sorted properly to cope with older
and newer block-copies but that should not be a problem.
discussed the tool with my coleage Andreas Arnez and we found that it
might be a good idea to replace it even with a perl program for the
same purpose since that would offer the flexibility to change it on
the fly when needed. The tool is mainly used for crash analysis so it
could be an advantage to change it without needing a C-compiler.
Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
Frank Haverkamp [Sat, 24 Nov 2007 10:04:50 +0000 (11:04 +0100)]
ubi-utils: Update and bugfix unubi
Drake Dawsett had done some changes to our unubi which where not yet
published. I hereby send his latest code and integrated the parameter
handling which was changed in our version. When reviewing this very
huge patch we need to ensure that possible changes of others are not
reversed.
Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
Frank Haverkamp [Wed, 26 Sep 2007 11:40:10 +0000 (13:40 +0200)]
ubi-utils: Nand2bin add more information when bad blocks occur
When bad blocks occur more debug information is needed.
To identify the bad blocks the blocksize is required.
The split block mode was added to generate files per
block. In split mode the bad block markers are analyzed
and a summary is printed at the end.
Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
Artem Bityutskiy [Wed, 29 Aug 2007 15:51:07 +0000 (18:51 +0300)]
ubi-utils: get rid of compile warnings
Just silly hacks. Also remove udevsettle() invocation from UBI
library. If it must be called, it is not library's business
anyway. Wa added it to make scripts which use ubimkvol utility
run, so it is the only caller that really needs this. So
just move the cruft to the utility.
David Woodhouse [Tue, 14 Aug 2007 14:54:49 +0000 (22:54 +0800)]
Switch multicast distribution system to round-robin mode
Send one packet from each eraseblock in turn, rather than all packets
for one eraseblock together. This means that bursts of loss are evenly
spread between blocks. It also makes the client side a bit more complex
if you can't assume that there's anywhere except the flash to store its
intermediate data.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
David Woodhouse [Fri, 10 Aug 2007 06:40:01 +0000 (14:40 +0800)]
Fine-tine the TX timing for serve_image.
It just busy-waits until it's time to send the next packet now. Which
sucks, but gives us good timing for debugging. Needs work (and actually
if we have hrtimers we probably _can_ sleep and still have proper
timing).
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Josh Boyer [Fri, 3 Aug 2007 13:55:51 +0000 (08:55 -0500)]
Increment mkfs.jffs2 version number
Several change have gone into mkfs.jffs2 since we switch from CVS, but the
version number never got incremented. Bump the version number up so people
can distinguish between older and newer versions.
Josh Boyer [Fri, 3 Aug 2007 13:49:10 +0000 (08:49 -0500)]
Remove $Id:$ tags
The $Id:$ tags are left over from the old CVS repository. Several files have
since been changed, and they generally have little value in a git repo so they
should be removed.
Josh Boyer [Fri, 3 Aug 2007 13:34:38 +0000 (08:34 -0500)]
Disable LZO compression by default
When the recent LZO compression support went into mtd-utils recently,
it was enabled by default. This means that mkfs.jffs2 will default to
creating images with lzo compression as the default.
Older kernels cannot decompress such images, so lzo should be disabled
by default. The patch below fixes this. To enable lzo compression
from now on, simply pass '-X lzo' as an option to mkfs.jffs2.
Richard Purdie [Tue, 10 Jul 2007 15:03:47 +0000 (16:03 +0100)]
Add favourlzo compression mode
Add a favourlzo compression mode to mtd-utils
This allows lzo compression to be used in the cases where the
compression ratio isn't quite as good zlib. This can make sense in
certain use cases because LZO decompression is much faster than zlib.
Signed-off-by: Richard Purdie <rpurdie@openedhand.com> Signed-off-by: Josh Boyer <jwboyer@gmail.com>
Richard Purdie [Tue, 10 Jul 2007 15:03:31 +0000 (16:03 +0100)]
Add lzo support to mtd-utils
Add LZO support to mtd-utils to generate LZO compressed jffs2 images
Unlike the kernel version, the standard lzo userspace library is used
along with lzo1x_999_compress rather than the lzo1x_1_compress version
since better compression ratios can be obtained (at no significant cost
to decompression time).
Signed-off-by: Richard Purdie <rpurdie@openedhand.com> Signed-off-by: Josh Boyer <jwboyer@gmail.com>
Adrian Hunter [Wed, 18 Jul 2007 08:50:53 +0000 (11:50 +0300)]
Add more information to integrity test error message
When the integrity test encounters a file that does not contain
the expected data, it lists the data that it expected to find
in terms of writes to the file.
Now the test also displays a list of "raw" writes that includes
writes that have been truncated away, or completely overwritten
by other writes.
The test also now displays the pid because it is used as the
initial random seed.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
Alexander Schmidt [Tue, 10 Jul 2007 15:53:54 +0000 (17:53 +0200)]
UBI-utils: add compare feature
This is a new feature for pfiflash, called "--compare". It allows the user
to simulate a pfiflash session without actually changing the flash
content. If the flash content is equal to the data in the pfif file,
pfiflash returns zero. A positive value is returned when the flash
content differs from the pfi file, which indicates that an update is
necessary. This feature is useful when a controller mounts an NFS share
during boot and has to determine if a pfi file stored on this share
contains a code update. Modified PDD values are also registered by the
compare feature.
Signed-off-by: Alexander Schmidt <alexs@linux.vnet.ibm.com>
Frank Haverkamp [Wed, 14 Mar 2007 13:17:07 +0000 (14:17 +0100)]
MTD-Utils: fix handling of ioctl return value in nand-utils
Older kernel do not implement the MTDFILEMODE ioctl. In this case
nandwrite and nanddump should have used MEMGETOOBSEL in combination
with MEMSETOOBSEL. Unfortunately the return value of the unsucessfull
ioctl is not -ENOTTY, but -1 and errno contains ENOTTY. This change
fixes this issue. I have not tested all cornercases. Would be
good if someone could do more careful testing than I did, or maybe
reviewing is sufficient in this case.
Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com> Signed-off-by: Josh Boyer <jwboyer@gmail.com>
Frank Haverkamp [Wed, 21 Feb 2007 09:40:13 +0000 (10:40 +0100)]
UBI-Utils: Increment version of pfiflash
Incrementing the version number was forgotten in the previous patch. Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com> Signed-off-by: Josh Boyer <jwboyer@gmail.com>
Frank Haverkamp [Mon, 11 Dec 2006 13:34:23 +0000 (14:34 +0100)]
[MTD] UBI Utils: Tools should have individual version numbers
The tools had a mixture of different version numbers. This is changed now.
The internal change to move to remove glibc dependencies should be reflected
by an increase of the version number, so that we can react if trouble is
seen with the new code.
Singed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
Drake Dowsett [Mon, 6 Nov 2006 15:54:10 +0000 (16:54 +0100)]
[MTD] UBI: rework of off-line UBI analysis tool
The new version can create a gnuplot graph of the erase count statistics.
It can also extract UBI volumes and single blocks with a preanalysis of
the EC as well as the VID header. It has a manual page too ;-).
Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
projects / mtd-utils.git / log