Merge remote-tracking branch 'origin/time_upgrade' into 64_and_32_bit_time_tests

Added a README to basic-tests

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Renamed timothy_tests to stress_tests

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Added a makefile to run all the unit tests.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Moved unit tests into their own folder.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Added test target to makefile.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Added a test target to makefile

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Removed obsolete test from linux_tests.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Removed obsolete timothy_test/yaffs_and_linux_mirror_tests

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Updated timothy_tests/stress_tester

Updated the mount point's path.
Broke out some commonly used shared functions and defines into a
shared file.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Removed old test from timothy_tests/quick_tests

Linux now allows directories to be opened.
When quick_tests were written, open was supposed to
return the error code EISDIR.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Updated is_yaffs_working_test_fix

Added pthread library to the makefile and changed
the mountpoint to the new "yflash2".

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Fixed segfault in yaffs_guts

Remounting a device would cause a segfault because the
device's temp_buffer would not be reinitialised. Resulting
in a null pointer being called as a function, creating the segfault.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Enhanced printf statements in is_yaffs_working_tests

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Fixed mountpoint path for is_yaffs_working_test

The old path "/yaffs2/" seem to have been removed.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Added -lpthread to is_yaffs_working_tests's makefile

Also removed dead code.

Signed-off-by: Timothy Manning <codedraftsman@gmail.com>

Added endianness support for YTIME_T

Added a inline function that handles swapping the Endianness for YTIME_T.

Signed-off-by: Timothy Manning <codedraughtsman@users.noreply.github.com>

Added comments to yaffs_DoUtimre

Signed-off-by: Timothy Manning <codedraughtsman@users.noreply.github.com>

Changed some tabs to spaces.

Signed-off-by: Timothy Manning <codedraughtsman@users.noreply.github.com>

Created a new test to check the size of YTIME_T

Added support for 64 bit time to yaffs direct.

These changes still need to be applied to the kernel and rtems.

Added some comments to dtest.c

Also expanded some of the printf description text.

Fixed timothy_tests/quick_tests

One of the string buffers was to small and would not compile.
The string buffer's size was increased and it all compliles.

Note: the tests still fail due to a unrelated segfault, currently
fixing this.

Index the block state string list because the order was wrong.

Thanks to Cory Wiltshire for pointing out the problem and proposing
a fix.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

kernel Makefile won't work after moving cache handling into a separated file.

Signed-off-by: Luca (Wei) Chen <wei@cvluca.com>

yaffs_guts.c : Fix missing temporary buffers when doing yaffs format on inband tags.

The low level init was not setting up temporary buffers. This was causing yaffs_format
to crash in the case of formating a device with inband tags.

Solution is to move the temporary buffer creation into low level init.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

New yaffs2_image_maker tool to work with 4kB pages - still under test

This appears to be working but more testing required.

The tool now takes two new parameters for modifying the bytes per page
and pages per block.

For instance to create an inband tags image using 4kB virtual pages (ie. 32
virtual pages per block) try:

./yaffs2_image_maker -i test-dir2 -o outimage -w working_file -I -c 4096  -B 32

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs_image_maker: Add ability to handle oob tags and no tags ecc

This still has some limits - it assumes 2k pages.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add a README for the image maker

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add endian feature to image maker and do some tidy up.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add an image maker tool that generates image files for 2k page * 64 devices

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Some clean up and more modular cache management

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add CONFIG_YAFFS_NO_YAFFS1 flag to remove yaffs1 code when not needed

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Move cache code to own file

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Move cache handling into its own file

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add CONFIG_YAFFS_SMALL_RAM to reduce RAM footprint

Reduce number of handles and search contexts.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add CONFIG_YAFFS_NO_TRACE to turn off all tracing

This allows a foootprint reduction.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs-vfs-multi.c: Handle search context properly to prevent leaking

Fix potential memory leak due to search context not being closed properly.

Thank you to George Pee for reporting this issue.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add malloc monitoring to get a sense of RAM use

Note this should only be monitoring the malloc usage by Yaffs, not the
malloc usage in things like flash simulators which are not relevant in
real systems.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Begign bug: Fix incorrect return value

The return value was not being used, but it was wrong.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Merge branch 'sizeofsizeof'

Fix two bugs:

sizeof(sizeof(hdr)) used instaed of sizeof(hr)

remounting a read-only fs in rw did not clean up defered operations
properly.

Add some RTEMS notes

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Handle cleanup when remounting rw after mounting read-only

Deferend clean ups were not happening properly.

Thanks to andriy.grytsenko@globallogic.com for providing a
patch that forms the babsis of these changes.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Make buffers larger to stop compiler complaining

The buffers were not overflowing, but the check was complaining.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

fix wrong use of sizeof in yaffs_checkpt_close

There is a distinct bug that sizeof was used incorrectly,
causing the judgment to be ineffective and always flushing checkpt buffer.

Signed-off-by: liaohua <liaohua4@huawei.com>

Add test to dereference symlink

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add more basic tests and some clean up.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add stats to flash simulator

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

rtems yests: FSX

Add readme and some tweaks.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add fsx file test and split up basic tests.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

rtems support: Rudimentary test harness now working.

Lots more test cases to add.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Makefile.rtems: Add a few missing files

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs_tagsmarshall.c: Fix prototype dependency

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

rtems_yaffs: Remove unnecessary flushing of symlink

Flushing only has meaning on regular files which have cached data.
Symlinks do not have any cached data, therefore flushing has no purpose. Remove.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

rtems-yaffs: Checkpoint the filesystem at unmount time

Adding a checkpoint greatly reduces the amount of time required to mount
the filesystem on the next boot.  After a clean unmount, remount time
is only a second or few.  After an unclean reboot, remount time requires
a scan of the filesystem, which can take a few minutes.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

rtems_yaffs: Add symlink support

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

rtems-yaffs: Modes to freshen to latest Yaffs and build in yaffs2/rtems

Makefile uses symlinks to files outside yaffs2/rtems aso that rtems can be built without
modifying the directory structure and deleting files.

Various tweaks needed to support changes in function prototypes.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Import rtems wrapper from rtems git

https://git.rtems.org/sebh/rtems-yaffs2.git/
Commit:  23be2ac372173ecf701a0a

Most of this work was authored by

Sebastian Huber <sebastian.huber@embedded-brains.de>
Sebastien Bourdeauducq <sebastien@milkymist.org>

Note that the RTEMS licencing applies to all the rtems wrapper code.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

fix memory leak when /proc/yaffs is read

Thanks to Jisheng Zhang <Jisheng.Zhang@synaptics.com> for supplying this patch

There is a kernel memory leak observed when the proc file /proc/yaffs
is read. This reason is that in yaffs_proc_open, single_open is called
and the respective release function is not called during release.

Fix with correct release function - single_release().

Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>

Fix race yaffs_flush_inodes against iput

needs inode->i_sb->s_inode_list_lock protects
  inode->i_sb->s_inodes, inode->i_sb_list

Signed-off-by: chengchao <chengchao@kedacom.com>

Akiko: Fix Yaffs Linux compilation for 4.9 to 4.19

Thanks Akiko for supplying this patch

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Only load name if read succeeds

Thanks Akiko for supplying this patch.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Merge branch 'master' of ssh://aleph1.co.uk/home/aleph1/git/yaffs2

Clean up some version dependencies

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Fix hang caused by returning with lock held

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Fix building for mkyaffsimage and mkyaffs2image

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Fix copyright

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Clean up some compilation warnings for VxWorks

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add Adre's patch for Linux 4.9

Kudos to Andre Renaud <andre@ignavus.net>

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add struct definition to yaffs_nameval header

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs direct: Fix opening of a directory.

If opening a directory using yaffs_open(), Yaffs Direct was checking
that the file was being opened with O_RDONLY.

In fact those flags should actually be ignored and it should be treated
as if opened O_RDONLY.

Therefore override flags if opening a directory.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Changes to test scripts

These just clean up a few things.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Fix yaffs_endian.h for Linux

Yaffs loff_t is kept as loff_t and converted into Y_LOFF_T when
when exported for Yaffs Direct.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Changes to enable Linux 4.8 compilation

timothy-tests: Fix paths for wrongly failing tests

All pass now.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

timothy-tests: Change mountpoint names to match test framework

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Get timothy quick tests building cleanly.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

More compilation clean up

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Compilation clean up

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Chean up nand test script

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs2: Yaffs endian support

Allow the forcing of endians so that Yaffs can be used on mixed endian
machines, or a flash image can be extracted and used elsewhere.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Change checkpoint object to using a bitfield

This is done to support changing to support endian
manipulation.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add ability to hide lost_n_found directory

Some people don't want it showing up in the directory structure.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Split qsort from the main Yaffs Direct base

Use the system qsort if available. If not, then add in the
optional qsort.c

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Write object headers with stored file extents

Previously file object headers were being written with file extents
rather than stored extents.

This could lead to situations where the data in the cache was not yet
written to flash but had been accounted for in the object header file
size.

Under power fail this could cause zero-filled blobs at the end of a file.

This change tracks the stored file size and only write object headers
reflecting the extents of the file that has actually been stored to flash,
thus avoiding the holes.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs_vfs_multi.c: Some cleanups

Remove some build time strings so that the binary deos not change
every time the code is compiled.

Add some support for kernel changes

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Change enums to u32 for stored fields

enum sizes are not fixed, so it is better to specify an actual sized
storage element.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add usr mode flags

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Further extensions for VxWorks wrapper

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Some changes to get VxWorks wrapper working

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: Remove unnecessary variable from checkpooint reading

No point in tracking ok if we're also going to break.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: Keep checkpt rd/wr buffers

Continual malloc/free on some systemcs can cause fragmentation. Better
to just allocate the buffers once and hold unto them.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: Remove unused variable

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs direct: Fix wrong condition for O_CREAT | O_EXCL

Thanks Chris for pointing this out.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: Clean up defeated conditional.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: Add handling for . and .. at end of path

yaffs_opendir("/nand/x/./") or yaffs_opendir("/nand/x/../") were
not working, but yaffs_opendir("/nand/x/.") and yaffs_opendir("/nand/x/..")
were.

Make them all work.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Add some more basic tests

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: Fix yaffs_readdir when dereferencing a symlink

If the directory is a symlink then follow the symlink.

eg
yaffs_mount("/nand/");
yaffs_symlink("x", "/nand/sym");
yaffs_mkdir("/nand/x",0666);

After that:

yaffs_opendir("/nand/sym/") works, but opendir("/nand/sym") fails because the
symlink itself is found. Fix by dereferencing the symlink in yaffs_opendir().

Thanks to Trent Lillehaugen for finding that.

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

Fix directory look up.

Thanks to Trent Lillehaugen for new algorithm:

Let's say I have two yaffs partitions: "/" and "/foo".
If I call, yaffs_mkdir("/foobar", 0), yaffsfs_FindDevice will match the path
with the "/foo" partition.  I would expect it to match "/" instead.
It will also match "/fo/o/bar" with "/foo", again I think it should match
with "/".

Signed-off-by: Charles Manning <cdhmanning@gmail.com>

yaffs: fix missing checkpooint on yaffs remount

For yaffs file system, the mode of reading or writing is restricted at
four pointer where are mnt->mnt_flags,mnt->mnt_sb->s_flags,mtd->flags
and dev->read_only,the first three is used handle file and file
system(eg,remount) operation, and last one(dev->read_only) almost is used
handle checkpoint of yaffs2. However, in current code, the dev->read_only
only can be changed at first time when the yaffs2 file system is mounted,
later it can't be changed again(eg,mount -o remount), the result is that
the checkpoint's saving operation always can't succeed if you set readonly
mode for yaffs2 file system when it is mounted at the first time.

To fix this issue, we implement yaffs_remount_fs() which allows the
rootfs to be remounted as r/w.

Also fix one missing argument in yaffs_flush_whole_cache() call.

Signed-off-by: Wenlin Kang <wenlin.kang@windriver.com>