Most boot sectors may contains marker for filesystem state. We can this
bit on every mount and warn user if some thing wrong, without checking
complete filesystem.
Signed-off-by: Oleksij Rempel <bug-track@fisher-privat.net>
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
Some OSos use reseved byte of boot sector to set state of the file
system. If first bit set, then filesystem is proably damaged - write
operation was not finished/cache not snycted/...
Signed-off-by: Oleksij Rempel <bug-track@fisher-privat.net>
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
mkdosfs will incorrectly calculate the number of sectors of a
given FAT partition if the number sectors are odd due to
count_blocks incorrectly handling the remainder of a division
operation. This miscalculation causes the OMAP4 bootloader to
fail to boot.
This bug can be observed by comparing the total sector size in
fdisk expert more to fsck.msdos; this discrepancy only shows up
when the number of sectors are odd.
See https://bugs.launchpad.net/ubuntu/+source/dosfstools/+bug/794043
for more information.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
char *".
These warnings appear when you compile the project with the option "-Wall", what
is done with the current default Makefile.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
https://bugzilla.redhat.com/show_bug.cgi?id=660154 for more information.
The problem is that alloc_rootdir_entry counts with 10000 files at max, but the
filename buffer is only 8 chars long. Due to pattern mask used it results to
only 10 files at max (FSCK0-9REC). If there is more than 10 files, it overflows
and hangs.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
This is helpful for SD cards or other removable media which have an enabled
write lock -- without the "Permission denied" message, the user has to strace
mkdosfs to find out what's going on.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
the resulting applications report strange errors on "large" (> 2 GiB)
partitions:
Seek to -2118967808:Invalid argument
Warning: Filesystem is FAT32 according to fat_length and fat32_length fields,
but has only 8613 clusters, less than the required minimum of 65525.
This may lead to problems on some systems.
This appears to be due to compilation with a 32-bit off_t and lseek() library
function.
Use lseek64 for positioning, and change some suspect uses of off_t to loff_t.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
Correct the code to align the root directory; it was broken before
since bs.dir_entries had already been set at the point of alignment.
This patch removes the dual use of bs.dir_entries and root_dir_entries
to carry the same information: the information is carried in
root_dir_entires exclusively, and then bs.dir_entries is set inside
setup_tables() at a late point.
The code to align the root directory is also wrapped in
if (align_structures); this avoids rounding the number of root
directory entries up to a whole sector when used with -a
(i.e. preserves the previous behavior.)
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
Align all data structures (reserved sectors, FATs, root directory for
FAT12/16) to an even multiple of the cluster size. This means that if
the partition is aligned, so will all clusters be. This adds
significant performance for anything where the physical sector size is
larger than the logical sector size, e.g. flash media or large-sector
hard disks.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
Firstly, packed attribute is added to the structure so that extension
is guarranteed to immediately follow name for the cross-name-extension
reads to succeed.
Secondly, writes into dir_entry->name that span through the extension as
well are split into two, so that FORTIFY_SOURCE's bound checking does
not abort dosfsck. There also was an off-by-one error in auto_rename()'s
sprintf().
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
If an orphan chain that linked to cluster 1 was reclaimed to a file,
deletion of the file would result in a filesystem panic.
Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>
