Prevent SEGV on corrupted msdos directories (Veo Zhang)

PR/50908: David Binderman: Optimize memset's

Avoid mixing cluster numbers and sector numbers. Makes code more readable.

From: http://marc.info/?t=140304310700005&r=1&w=2
When truncating cluster chains fix the length of the cluster head.

branches: 1.25.10; 1.25.20;
undo part of the previous patch, which I suspect is bad.

PR/44529: Martin Danielsson: fsck_msdos crashes when verifying corrupt file
system. Avoid using the long name index when it is 0. Refactor the code to
avoid duplication.

branches: 1.23.2;
fix sign-compare issues

branches: 1.22.4;
Move to 2 clause license, approved by Wolfgang Solfrank.

branches: 1.21.12; 1.21.14; 1.21.16;
Remove more duplicate #includes,
from Slava Semushin <slava.semushin@gmail.com>

- don't leak memory on allocation failure.
- s/perror/perr and print how much we were trying to allocate.

the root directory of non fat32 filesystems is stored in a special area.
a couple of corner cases can cause it to fail to write out that area
after it performs repairs.

Coverity CID 1078: Add assertion check for NULL pointer.

Kill __P(), ANSIfy, remove main() prototype.

use bounded string op

Recognize "dirty" signature on FAT as set by Windows 95 OSR2+ and don't report
it just as "odd byte sequence". Handle the dirty flag similar way
it's treated on e.g. FFS - e.g. clear it separately from rest of FAT.
Changing of FAT is now postponed until after all checks are done,
check.c:checkfilesys() was cleaned considerably and several messages
has been adjusted to be more accurate.

If file system was modified and no error has occured, just return 0
instead of 4 - there is no need for forcing a reboot in this case.

This addresses bin/5638 by Christos Zoulas.

from Erik Bertelsen <erik@mediator.uni-c.dk>
{ put } { in } { lots } { of } { these } { to } { shut } { up } { egcs }

__AUDIT__ cleanup.

Fix some bugs & annoyances when reparing defect disks

Add support for FAT32
Support dropping lost cluster chains (on request by Christoph Badura)
While being here, fix some minor bugs, mostly with format strings

* cleanup for WARNS=1
* getopt returns -1 not EOF
* cleanup manpage

Fix a few bugs, especially when repairing disks:
Print correct pathname in error messages
Fix unterminating loop when trying to correct a bad fat
Require fat media byte to be the same as the one in the bpb
Fix unterminating loop when looking for free directory slot in LOST.DIR
(bad disk image provided by Christoph Badura)

util -> fsutil

Fix delete() call in dir.c differently than Christos' way.
Require at least one filesystem as program argument.
Update man page to reflect the latter.

- use error printing code from fsck(8)
- remove fstab walking code
- add missing prototypes
- add missing argument in dir.c near "Truncate", where delete() was being
called with fewer args. I think the missing argument should be a zero,
but Wolfgang should check...

Fix null pointer derefernce
Fixes PR#2765. Thanks David.

Fix gcc warnings.

Don't allocate data structures for non-directories
Free any data structures when finishing a check (at the latest)

Check return values of malloc
Generate full pathnames only on request
Minor cleanup, RCS Ids in .h files

branches: 1.1.4;
Add fsck utility for msdosfs

PR/50908: David Binderman: Optimize memset's

Avoid mixing cluster numbers and sector numbers. Makes code more readable.

From: http://marc.info/?t=140304310700005&r=1&w=2
When truncating cluster chains fix the length of the cluster head.

branches: 1.25.10; 1.25.20;
undo part of the previous patch, which I suspect is bad.

PR/44529: Martin Danielsson: fsck_msdos crashes when verifying corrupt file
system. Avoid using the long name index when it is 0. Refactor the code to
avoid duplication.

branches: 1.23.2;
fix sign-compare issues

branches: 1.22.4;
Move to 2 clause license, approved by Wolfgang Solfrank.

branches: 1.21.12; 1.21.14; 1.21.16;
Remove more duplicate #includes,
from Slava Semushin <slava.semushin@gmail.com>

- don't leak memory on allocation failure.
- s/perror/perr and print how much we were trying to allocate.

the root directory of non fat32 filesystems is stored in a special area.
a couple of corner cases can cause it to fail to write out that area
after it performs repairs.

Coverity CID 1078: Add assertion check for NULL pointer.

Kill __P(), ANSIfy, remove main() prototype.

use bounded string op

Recognize "dirty" signature on FAT as set by Windows 95 OSR2+ and don't report
it just as "odd byte sequence". Handle the dirty flag similar way
it's treated on e.g. FFS - e.g. clear it separately from rest of FAT.
Changing of FAT is now postponed until after all checks are done,
check.c:checkfilesys() was cleaned considerably and several messages
has been adjusted to be more accurate.

If file system was modified and no error has occured, just return 0
instead of 4 - there is no need for forcing a reboot in this case.

This addresses bin/5638 by Christos Zoulas.

from Erik Bertelsen <erik@mediator.uni-c.dk>
{ put } { in } { lots } { of } { these } { to } { shut } { up } { egcs }

__AUDIT__ cleanup.

Fix some bugs & annoyances when reparing defect disks

Add support for FAT32
Support dropping lost cluster chains (on request by Christoph Badura)
While being here, fix some minor bugs, mostly with format strings

* cleanup for WARNS=1
* getopt returns -1 not EOF
* cleanup manpage

Fix a few bugs, especially when repairing disks:
Print correct pathname in error messages
Fix unterminating loop when trying to correct a bad fat
Require fat media byte to be the same as the one in the bpb
Fix unterminating loop when looking for free directory slot in LOST.DIR
(bad disk image provided by Christoph Badura)

util -> fsutil

Fix delete() call in dir.c differently than Christos' way.
Require at least one filesystem as program argument.
Update man page to reflect the latter.

- use error printing code from fsck(8)
- remove fstab walking code
- add missing prototypes
- add missing argument in dir.c near "Truncate", where delete() was being
called with fewer args. I think the missing argument should be a zero,
but Wolfgang should check...

Fix null pointer derefernce
Fixes PR#2765. Thanks David.

Fix gcc warnings.

Don't allocate data structures for non-directories
Free any data structures when finishing a check (at the latest)

Check return values of malloc
Generate full pathnames only on request
Minor cleanup, RCS Ids in .h files

branches: 1.1.4;
Add fsck utility for msdosfs