Bernhard Walle [Sat, 2 May 2015 12:38:06 +0000 (14:38 +0200)]
mkfs.ubifs: Fix build with gcc 5.1
In gcc 5.1, the default C standard which is used to compile a C file,
has changed from gnu89 to gnu11. This changed the meaning of 'extern
inline'. See https://gcc.gnu.org/gcc-5/porting_to.html.
In mkfs.ubifs, this leads to multiple definitions of
hashtable_iterator_key and -hashtable_iterator_value. I think the most
pragmatic way to fix the issue is to replace 'extern inline' with
'static inline' here.
Signed-off-by: Bernhard Walle <bernhard@bwalle.de> Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Brian Norris [Mon, 15 Sep 2014 17:35:00 +0000 (10:35 -0700)]
libmtd: don't ignore "region index" parameter in mtd_regioninfo()
ioctl(MEMGETREGIONINFO) has one input parameter (regionindex) and three
output parameters (info about the erase region). There are two problems
in mtdinfo/libmtd here:
1. mtdinfo.c doesn't initialize its region_info_user struct, instead
passing uninitialized data to mtd_regioninfo()
2. mtd_regioninfo() fails to utilize the 'regidx' parameter to fill out
the regionindex parameter properly, so the garbage from mtdinfo.c is
propagated to the ioctl()
This means that mtdinfo will continuously probe the same (possibly
out-of-range) erase region, instead of looping over the valid regions.
Let's fix this in the mtd_regioninfo() helper, and at the same time,
let's zero out the mtdinfo.c buffer, as an additional precaution to keep
from using uninitialized data.
Initial error report from Yang, when running "mtdinfo /dev/mtd0" on a
Cavium 6100 board:
root@CN61XX:~# mtdinfo /dev/mtd0
mtd0
Name: phys_mapped_flash
Type: nor
Eraseblock size: 65536 bytes, 64.0 KiB
Amount of eraseblocks: 128 (8388608 bytes, 8.0 MiB)
Minimum input/output unit size: 1 byte
Sub-page size: 1 byte
Additional erase regions: 0
Character device major/minor: 90:0
Bad blocks are allowed: false
Device is writable: true
libmtd: error!: MEMGETREGIONINFO ioctl failed for erase region 0
error 22 (Invalid argument)
Eraseblock region 0: info is unavailable
libmtd: error!: MEMGETREGIONINFO ioctl failed for erase region 1
error 22 (Invalid argument)
Eraseblock region 1: info is unavailable
Reported-by: Yang Wei <Wei.Yang@windriver.com> Signed-off-by: Brian Norris <computersforpeace@gmail.com>
ubiformat: fix the subpage size hint on the error path
David Binderman <dcb314@hotmail.com> reports that the following piece of looks
wrong:
if (!args.subpage_size != mtd->min_io_size)
normsg("may be sub-page size is incorrect?");
I totally agree with him and I believe that we actually meant to have no
negation in fron to f 'args.subpage_size', so instead, the code should look
like this:
if (args.subpage_size != mtd->min_io_size)
normsg("may be sub-page size is incorrect?");
libmtd: fix mtd_dev_present return value on legacy systems
On legacy systems, if "/proc/mtd" doesn't exist or gives a read error,
mtd_dev_present returns -1 (since it calls legacy_dev_present), contrary
to what's specified in the header file.
This causes checks like
if (mtd_dev_present(n)) {
...
}
to give false positives. Fix this by comparing the return value to 1.
Signed-off-by: Guido Martínez <guido@vanguardiasur.com.ar> Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Guido Martínez [Thu, 14 Aug 2014 16:29:45 +0000 (13:29 -0300)]
nandtest: fix --reads argument
The --reads option specifies the argument as optional, but doesn't check
for a null optarg, which means that nandtest segfaults when run as
"nandtest --reads".
Fix this by making the argument required, and changing the help text to
not specify it as optional. Argument -r already specifies the argument
as required, so we fix this inconsistency too.
The current nandtest performs a simple test which consists of:
1. erase block
2. write data
3. read and verify
In order to improve the nandtest strength, this commit adds a new parameter
to increase the number of "read and verify" iterations. In other words,
the test now consists of:
1. erase block
2. write data
3. read and verify (N times)
This seem to apply more pressure on a NAND driver's ECC engine and has been
used to discover stability problems with an old OMAP2.
include/common.h: fix build against recent 0.9.33 uClibc
An implementation of rpmatch() was backported to the 0.9.33 branch of uClibc.
So the uClibc version check introduced in commit 50c9e11f7e (include/common.h:
fix build against current uClibc) is not enough. Rename the local rpmatch()
implementation to avoid collision.
Signed-off-by: Baruch Siach <baruch@tkos.co.il> Acked-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Brian Norris <computersforpeace@gmail.com>
include/common.h: fix build against current uClibc
Commit dbe0fd17f2 (mtd-utils: new prompt() helper for talking to the user)
introduced a rpmatch() call. However, uClibc versions older than (not yet
released) 0.9.34 don't have rpmatch() implementation. Add one.
Artem Bityutskiy [Mon, 7 May 2012 08:44:38 +0000 (11:44 +0300)]
make_a_release.sh: suggest announcement e-mail
The 'make_a_release.sh' script appears to be extremely useful - I do not
forget things as I used to anymore (amending Makefile, signing, uploading
to the FTP server, etc). It is very useful that it suggest me exact commands
which I may just copy-past to my command line.
This patch improves the script and makes it suggest the e-mail announcement
which I may just copy-paste to my command line and the announcement will
be sent using 'git send-email' command. It will include all the interested
parties in CC.
Brian Norris [Mon, 31 Mar 2014 17:03:04 +0000 (10:03 -0700)]
ubiformat: correct "non-ubifs" warning message
UBI's raw flash scan actually scans for UBI data, not UBIFS data (there
*are* UBI users that are not UBIFS!), so correct the warning message.
This also matches the comment in libscan.h.
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Aaron Sierra [Fri, 27 Sep 2013 16:34:04 +0000 (11:34 -0500)]
mtd-utils: Makefile: add LDFLAGS_mkfs.ubifs
The build rule for mkfs.ubifs was missing an LDFLAGS_* variable like
mkfs.jffs2 had. This prevented mkfs.ubifs from being built against
explicit external libraries which is needed when cross-compiling.
This also adds UUIDCPPFLAGS and UUIDLDFLAGS variables to support the
mkfs.ubifs build.
Huang Shijie [Tue, 20 Aug 2013 05:58:37 +0000 (13:58 +0800)]
check the MLC nand type
In the current code, the MTD_NANDFLASH stands for both the SLC and MLC.
In the kernel, the MTD_NANDFLASH only stands for the SLC now,
so in order to keep the logic unchanged, we should also check the MLC
NAND by MTD_MLCNANDFLASH.
Signed-off-by: Huang Shijie <b32955@freescale.com> Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Mike Frysinger [Thu, 9 May 2013 17:59:24 +0000 (13:59 -0400)]
ubiupdatevol: add a --skip option
This already has a --size option for controlling how many bytes to read
from the input. Add a --skip option to control the offset into the input
too. This way people don't have to do `dd | ubiupdatevol`.
While we're here, I've fixed the types used with args.size and the read
loop so that they can hold the right sizes (like setting a 32bit+ size).
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Mike Frysinger [Thu, 9 May 2013 00:03:14 +0000 (20:03 -0400)]
nandwrite: add --input-{skip,size} options
If you have a file image and want to copy sub-portions out and into
NAND, there's no easy way to do that. You can use dd to extract it
to a temp file, or pipe it to nandwrite 1 page at a time. Both suck.
Add two new flags to explicitly set the size and offset of the input
file. Seeking stdin isn't currently supported as I'm not sure it's
necessary. It wouldn't be hard to add though...
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Mike Frysinger [Thu, 9 May 2013 00:03:13 +0000 (20:03 -0400)]
nandwrite: clean up length types
We use 'int' in many places to represent offsets/sizes. That obviously
does not play well with larger NAND devices on 32bit systems. Instead,
use the right type as needed:
- long long to represent the length of the image
- use fstat() rather than lseek();lseek(); to get the length of the image
- use size_t/ssize_t when working with read()
- tweak the printf formats as needed
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Mike Frysinger [Wed, 8 May 2013 16:27:26 +0000 (12:27 -0400)]
mkfs.ubifs: allow reformatting of devices
Sometimes I want to re-initialize an existing ubifs, but the tool
currently bails out if the volume is already formatted. Prompt the
user instead so they can decide.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Mike Frysinger [Wed, 8 May 2013 16:27:24 +0000 (12:27 -0400)]
move _GNU_SOURCE to the main makefile
A bunch of utils are relying on _GNU_SOURCE already. The new prompt code
uses getline() which is now part of POSIX, but in older versions of glibc,
it was behind _GNU_SOURCE as it was a GNU extension.
This change doesn't actually tie us to glibc. Only code that uses GNU
extensions does that. It just kills warning when using older versions of
glibc.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Mike Frysinger [Wed, 8 May 2013 16:26:13 +0000 (12:26 -0400)]
fix build errors w/newer kernel headers & glibc
Building with linux-headers-3.9 and glibc-2.17 fails like so:
In file included from summary.h:15:0,
from jffs2dump.c:37:
/usr/include/linux/uio.h:16:8: error: redefinition of 'struct iovec'
struct iovec
^
In file included from /usr/include/bits/fcntl-linux.h:38:0,
from /usr/include/bits/fcntl.h:61,
from /usr/include/fcntl.h:35,
from jffs2dump.c:25:
/usr/include/bits/uio.h:43:8: note: originally defined here
struct iovec
^
Elie De Brauwer [Fri, 1 Mar 2013 18:37:39 +0000 (19:37 +0100)]
integck.c: Fix buffer overflow in save_file
In the problem above I've spend several hours waiting for the issue to
appear, only to had the 'luck' that it was found in a file whose name was
256 bytes in length, resulting in the write to fail. Closer examination
showed that the buffer to store the path was 256 bytes in length, but this
buffer also includes /tmp and the read/write suffix and should be able to
contain a filename which is up to 255 bytes (NAME_MAX in linux/limits.h)
in size which is a bad fit. So that array is modified to FILENAME_MAX
(stdio_lim.h) and some checking is added to truncate the filename should
it cause an overflow.
The following log shows the first patch in action (see the correct seed),
and shows why this third patch is needed:
<quote>
integck: File Data:
integck: Offset: 0 Size: 1 Seed: 5008310 R.Off: 0
integck: 1 writes
integck: ============================================
integck: Write Info:
integck: Offset: 0 Size: 1 Seed: 5008310 R.Off: 0
integck: Offset: 0 Size: 1 Seed: 8246352 R.Off: 0
integck: Offset: 0 Size: 1 Seed: 5078796 R.Off: 0
integck: Offset: 0 Size: 1 Seed: 2267087 R.Off: 0
integck: Offset: 0 Size: 1 Seed: 3602680 R.Off: 0
integck: 5 writes or truncations
integck: ============================================
integck: Saving /tmp/yqcnfygfitaatyeyvffrguegcdttamcnyhowhgieljfuxfipiljsjcbluaeaghwyinkggommsbwnmvekihgnwgiibccpbwfrpxuxwkmnyghnutrudienngxwgorudbskedaaekiuiyqksfazrwzfwbfhzjjqoiulebtlpbfiuffmsnguqkjzqjqizimsmhbqqagaebjdhqwmzdxghiavtcxubegawlgtvstuqurkurpnrckjfkgostdtpg.integ.sav.readn
integck: error!: condition 'w_fd != -1' failed in save_file() at integck.c:1445
integck: error 36 (File name too long)
</quote>
Signed-off-by: Elie De Brauwer <eliedebrauwer@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Elie De Brauwer [Fri, 1 Mar 2013 18:37:38 +0000 (19:37 +0100)]
integck.c: rework file_check_data to immediately dump the buffer containing the errors
See my problem description int the previous commit, the point is that integck
in file_check_data reads a buffer, and then checks if the data is correct, it
will do a seek(0), and reread from the same fd. The point is that in the
scenario I observed integck failed (due to a buffer mismatch) but the it saved
(and what was in flash) was actually correct. So I modified this function to
dump the buffers to stderr at the moment an error is found.
Signed-off-by: Elie De Brauwer <eliedebrauwer@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Elie De Brauwer [Fri, 1 Mar 2013 18:37:37 +0000 (19:37 +0100)]
integck.c: Only verify the operation after all datastructures have been updated
<quote>
integck: File Data:
integck: Offset: 0 Size: 196 Seed: 5999877 R.Off: 0
integck: Offset: 196 Size: 33 Seed: 4160795 R.Off: 0
integck: Offset: 229 Size: 1252 Seed: 8070052 R.Off: 0
integck: Offset: 1481 Size: 612 Seed: 4160795 R.Off: 1285
integck: Offset: 2093 Size: 6 Seed: 6946586 R.Off: 0
integck: Offset: 2099 Size: 536 Seed: 4160795 R.Off: 1903
integck: Offset: 2635 Size: 1562 Seed: 9845455 R.Off: 0
integck: Offset: 4197 Size: 80 Seed: 702818 R.Off: 0
integck: Offset: 4277 Size: 115 Seed: 9845455 R.Off: 1642
integck: 9 writes
integck: ============================================
integck: Write Info:
integck: Offset: 826 Size: 357 Seed: 5908448 R.Off: 0
integck: Offset: 4197 Size: 80 Seed: 702818 R.Off: 0
...
</quote>
And I would expect the file data listing to include at offset 826 something
with a size of 357 and a seed of 5908448. Clearly it is not there (which
is already extremely confusing). The point is that file_write_info first
updates the raw_write, then verifies the data (passing the new write)
and only after that updates the write structure. But in file_check_data
only the newly written data is verified (passed as an argument) whilst
the save_file() function to dump the file uses the raw_writes to recreate
the written data (while raw_writes is only updated after after this check
would have succeeded). Several lines to say that in this patch the verify
only gets called _after_ the datastructures are updated.
Signed-off-by: Elie De Brauwer <eliedebrauwer@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Uwe Kleine-König [Thu, 28 Feb 2013 09:42:26 +0000 (10:42 +0100)]
flash_otp_write: fix a buffer overflow on NAND with write size > 2048
I'm not aware of any chip having a write size bigger than 2048 today.
Still checking for that instead of a sleeping problem to bite us maybe
in a few years is easy.
flash_otp_write might see only a single byte when reading from stdin for
the first tim. In this case (and without this patch) it pads to
$writesize with '\xff's and writes that out. In the next iteration it
reads the 2nd byte, pads and writes again. So the 2nd byte is written to
offset $writesize instead of 1.
Brian Norris [Sun, 25 Nov 2012 07:26:09 +0000 (23:26 -0800)]
ubi-tests: fix pthreads linking
Technically, '-l' linker options should be included only after the
objects which must link to the library. So when we include '-lpthread'
in the LDFLAGS variable, it gets placed too early (i.e., before the
io_paral.o object), and so the pthread linkage never occurs. The
following error probably only shows up with linkers that don't link
pthreads by default.
$ make tests V=1
...
gcc -I../../ubi-utils//include -I ../../include -lpthread -Wall -Wextra -Wwrite-strings -Wno-sign-compare -ffunction-sections -fdata-sections -Wl,--gc-sections -g -o /home/norris/git/mtd-utils/tests/ubi-tests/io_paral /home/norris/git/mtd-utils/tests/ubi-tests/io_paral.o /home/norris/git/mtd-utils/tests/ubi-tests/helpers.o libubi.a
/home/norris/git/mtd-utils/tests/ubi-tests/io_paral.o: In function `main':
/home/norris/git/mtd-utils/tests/ubi-tests/io_paral.c:287: undefined reference to `pthread_create'
/home/norris/git/mtd-utils/tests/ubi-tests/io_paral.c:295: undefined reference to `pthread_create'
/home/norris/git/mtd-utils/tests/ubi-tests/io_paral.c:303: undefined reference to `pthread_join'
collect2: ld returned 1 exit status
make[2]: *** [/home/norris/git/mtd-utils/tests/ubi-tests/io_paral] Error 1
...
$ ld --version
GNU ld (GNU Binutils for Ubuntu) 2.22
...
$ gcc --version
gcc (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3
...
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Marcus Prebble [Tue, 16 Oct 2012 11:51:01 +0000 (13:51 +0200)]
mkfs.ubifs: Improve error handling of is_contained()
The is_contained() function returns -1 if an error occurs when
canonicalizing the output file path/root directory. This resulted in the
confusing error message 'Error: The output file cannot be in the UBIFS
root' when specifying a non-existent directory for the output.
This patch changes the error handling to display a different error
message for the case when is_contained() returns -1.
Additionally it frees all memory allocated by is_contained().
Signed-off-by: Marcus Prebble <marcus.prebble@axis.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Artem Bityutskiy [Wed, 10 Oct 2012 11:23:18 +0000 (14:23 +0300)]
mkfs.ubifs: rewrite path checking
We use the 'in_path()' function to check whether the output image is
withing the mkfs.ubifs root directory or not. However, this function
is not correct and it fails for the following situation, as
Marcus Prebble <marcus.prebble@axis.com> reports:
1. We have our root file-system mounted at / and want to build an image
out of it.
2. We have tmpfs mounted at /tmp
3. We mount the root file-system under /tmp/newroot
4. We run mkfs.ubifs with -r /tmp/newroot -o /tmp/image
And this fails. It fails because 'in_path()' misses this use-case.
This patch re-implements the check completely. Now we use 'realpath()'
to find canonical paths and just check that the output file is not
under the root mkfs.ubifs directory.
Reported-by: Marcus Prebble <marcus.prebble@axis.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Tested-by: Marcus Prebble <marcus.prebble@axis.com>
Richard Genoud [Wed, 12 Sep 2012 14:38:35 +0000 (16:38 +0200)]
mkfs.jffs2: correct some warnings using PRIdoff_t
When compiled with WITHOUT_LARGEFILE, there was warnings like that:
warning: format '%9llu' expects type 'long long unsigned int',
but argument 3 has type '__off_t'
Using PRIdoff_t corrects that.
Signed-off-by: Richard Genoud <richard.genoud@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Richard Genoud [Wed, 12 Sep 2012 14:37:19 +0000 (16:37 +0200)]
ubiformat: fix failure on big partitions (>4Gio)
The offset (which is 64bits when mtd-utils are not compile with
WITHOUT_LARGEFILE) is calculated like that:
offset = nb * size;
But nb and size are int, so on 32bits platforms, there's a possible
overflow.
So, it should be replace with:
offset = (off_t)nb * size;
If WITHOUT_LARGEFILE is defined, there still be an overflow, but it's
what we want, right ?
Cheney Chen tested an ubiformat on a NAND (5.9 GiB mtd part).
Richard Genoud [Wed, 22 Aug 2012 16:04:37 +0000 (18:04 +0200)]
ubiattach: fail if kernel ignores max_beb_per1024
If the kernel doesn't know the max_beb_per1024 parameter in the attach
ioctl, but the call still succeeded ubi_attach and ubi_attach_mtd will
return 1 instead of 0.
In this case, the ubiattach command will detach the device and fail with
an error message.
Signed-off-by: Richard Genoud <richard.genoud@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Richard Genoud [Wed, 22 Aug 2012 16:04:36 +0000 (18:04 +0200)]
ubiattach: introduce max_beb_per1024 in UBI_IOCATT
The ioctl UBI_IOCATT has been extended with max_beb_per1024 parameter.
This parameter is used for adjusting the "maximum expected number of
bad blocks per 1024 blocks" for each mtd device.
The number of physical erase blocks (PEB) that UBI will reserve for bad
block handling is now:
whole_flash_chipset__PEB_number * max_beb_per1024 / 1024
This means that for a 4096 PEB NAND device with 3 MTD partitions:
mtd0: 512 PEB
mtd1: 1536 PEB
mtd2: 2048 PEB
the commands:
ubiattach -m 0 -d 0 -b 20 /dev/ubi_ctrl
ubiattach -m 1 -d 1 -b 20 /dev/ubi_ctrl
ubiattach -m 2 -d 2 -b 20 /dev/ubi_ctrl
will attach mtdx to UBIx and reserve:
80 PEB for bad block handling on UBI0
80 PEB for bad block handling on UBI1
80 PEB for bad block handling on UBI2
=> for the whole device, 240 PEB will be reserved for bad block
handling.
This may seems a waste of space, but as far as the bad blocks can appear
every where on a flash device, in the worst case scenario they can
all appear in one MTD partition.
So the maximum number of expected erase blocks given by the NAND
manufacturer should be reserve on each MTD partition.
Signed-off-by: Richard Genoud <richard.genoud@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Artem Bityutskiy [Fri, 3 Aug 2012 07:01:18 +0000 (10:01 +0300)]
load_nandsim.sh: remove useless function
We do not really need a separate complex function to check if nandsim
is already loaded or not. Besides, it is safer to check /proc/mtd
in case nandsim is compiled-in.
Tomer Barletz [Tue, 26 Jun 2012 21:46:41 +0000 (14:46 -0700)]
mtd-utils: Check mtdoffset is not larger than mtd.size in case of a bad block.
mtdoffset is being tested against mtd.size in the outer two loops, but
the third nested one does not test against it.
In case of a bad block we'll try to access an out of bounds offset in
the next MEMGETBADBLOCK ioctl, which will fail with EINVAL.
In case mtdoffset is indeed larger than the partition size, we need to
bail, since there are not enough "good" blocks to complete the write.
Brian Norris [Wed, 28 Mar 2012 23:59:06 +0000 (16:59 -0700)]
Makefile: fixup previous 'make clean' fix
Apparently, Makefile comments need to be made without indentation. Otherwise,
they are printed out as shell commands. This fix prevents seeing this in your
shell during 'make clean':
$ make clean
...
# findutils v4.1.x (RHEL 4) do not have '+' syntax
...
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Acked-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Brian Norris [Fri, 9 Mar 2012 17:49:09 +0000 (09:49 -0800)]
Makefile: fix "make clean" for old GNU find
findutils v4.1.x does not have the `-exec CMD {} +' syntax. We can just as
easily use the `-exec CMD {} \;' syntax. However, it will launch a lot more
`rm' processes, so we only use it if the first form fails with an error.
This isn't a perfect solution (`find -exec +' can fail for other reasons)
but it works well enough.
This problem manifests itself in RHEL 4, findutils 4.1.20:
$ make clean
rm -f /XXX/mtd-utils/*.o /XXX/mtd-utils/ftl_format ...
find: missing argument to `-exec'
make: *** [clean] Error 1
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Acked-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Artem Bityutskiy [Wed, 7 Mar 2012 14:08:30 +0000 (16:08 +0200)]
mkfs.ubifs: do not ignore --max-leb-cnt when formatting an UBI volume
When the output file is an UBI volume - mkfs.ubifs just sets --max-leb-cnt
to the volume size and ignores the user-supplied --max-leb-cnt value, which
is wrong. Let's set it to the volume size only if the user did not supply
--max-leb-cnt.
Brian Norris [Wed, 8 Feb 2012 21:26:20 +0000 (13:26 -0800)]
libmtd: perform device checking first
If we don't check for the MTD before calling `legacy_get_dev_info1', we may
get errors like:
libmtd: MTD subsystem is old and does not support sysfs, so MTD character device nodes have to exist
libmtd: error!: "/dev/mtd2" is not a character device
mtdinfo: error!: libmtd failed get MTD device 2 information
error 22 (Invalid argument)
So reverse the order of these two checks.
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Brian Norris [Sat, 11 Feb 2012 04:35:05 +0000 (20:35 -0800)]
libmtd: fix mtd_write() issues for large data-only writes
ioctl(MEMWRITE) is implemented with memdup_user(), and so it allocates
kernel memory in contiguous regions. This limits its usefulness for large
amounts of data, since contiguous kernel memory can become scarce. I have
experienced "out of memory" problems with ubiformat, for instance, which
writes in eraseblock-sized regions:
This error can be mitigated for now by only using ioctl(MEMWRITE) when we
need to write OOB data, since we can only do this in small transactions
anyway. Then, data-only transactions (like those originating from
ubiformat) can be carried out with write() calls.
This issue can also be solved within the kernel ioctl(), but either way,
this patch is still useful, since write() is more straightforward (and
efficient?) than ioctl() for data-only writes.
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
projects / mtd-utils.git / log