- 13 Sep, 2021 12 commits
-
-
Kari Argillander authored
We do not have any reason to keep old linear search in. Before this was used for error path or if table was so big that it cannot be allocated. Current binary search implementation won't need error path. Remove old references to linear entry search. Signed-off-by:
Kari Argillander <kari.argillander@gmail.com> Signed-off-by:
Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
-
Kari Argillander authored
We could try to optimize algorithm to first fill just small table and after that use bigger table all the way up to ARRAY_SIZE(offs). This way we can use bigger search array, but not lose benefits with entry count smaller < ARRAY_SIZE(offs). Signed-off-by:
-
Kari Argillander authored
Current binary search allocates memory for table and fill whole table before we start actual binary search. This is quite inefficient because table fill will always be O(n). Also if table is huge we need to reallocate memory which is costly. This implementation use just stack memory and always when table is full we will check if last element is <= and if not start table fill again. The idea was that it would be same cost as table reallocation. Signed-off-by:
-
Kari Argillander authored
We have lot of unnecessary headers in these files. Remove them so that we help compiler a little bit. Signed-off-by:
-
Kari Argillander authored
There is lot of headers which we do not need in this file. Delete them and add what we really need. Here is list which identify why we need this header. <linux/kernel.h> // min() <linux/slab.h> // kzalloc() <linux/stddef.h> // offsetof() <linux/string.h> // memcpy(), memset() <linux/types.h> // u8, size_t, etc. "debug.h" // PtrOffset() Signed-off-by:
-
Kari Argillander authored
There is no headers. They will be included through ntfs_fs.c, but that is not right thing to do. Let's include headers what this file need straight away. types.h is needed for __le16, u8 etc. kernel.h is needed for le16_to_cpu() Signed-off-by:
-
Kari Argillander authored
We only need linux/types.h for types like u8 etc. So we can remove rest and help compiler a little bit. Signed-off-by:
-
Kari Argillander authored
size_t needs header. Add missing header guards so that compiler will only include these ones. Signed-off-by:
-
Kari Argillander authored
We do not have headers at all in this file. We should have them so that not every .c file needs to include all of the stuff which this file need for building. This way we can remove some headers from other files and get better picture what is needed. This can save some compilation time. And this can help if we sometimes want to separate this one big header. Also use forward declarations for structs and enums when it not included straight with include and it is used in function declarations input. This will prevent possible compiler warning: xxx declared inside parameter list will not be visible outside of this definition or declaration Here is list which I made when parsing this. There is not necessarily all example from this header file, but this just proofs we need it. <linux/blkdev.h> SECTOR_SHIFT <linux/buffer_head.h> sb_bread(), put_bh <linux/cleancache.h> put_page() <linux/fs.h> struct inode (Just struct ntfs_inode need it) <linux/highmem.h> kunmap(), kmap() <linux/kernel.h> cpu_to_leXX() ALIGN <linux/mm.h> kvfree() <linux/mutex.h> struct mutex, mutex_(un/try)lock() <linux/page-flags.h> PageError() <linux/pagemap.h> read_mapping_page() <linux/rbtree.h> struct rb_root <linux/rwsem.h> struct rw_semaphore <linux/slab.h> krfree(), kzalloc() <linux/string.h> memset() <linux/time64.h> struct timespec64 <linux/types.h> uXX, __leXX <linux/uidgid.h> kuid_t, kgid_t <asm/div64.h> do_div() <asm/page.h> PAGE_SIZE "debug.h" ntfs_err() (Just one entry. Maybe we can drop this) "ntfs.h" Do you even ask? Signed-off-by:
-
Kari Argillander authored
We do not have header files at all in this file. Add following headers and there is also explanation which for it was added. Note that explanation might not be complete, but it just proofs it is needed. <linux/blkdev.h> // SECTOR_SHIFT <linux/build_bug.h> // static_assert() <linux/kernel.h> // cpu_to_le64, cpu_to_le32, ALIGN <linux/stddef.h> // offsetof() <linux/string.h> // memcmp() <linux/types.h> //__le32, __le16 "debug.h" // PtrOffset(), Add2Ptr() Signed-off-by:
-
Kari Argillander authored
Add forward declarations for structs so that we can include this file without warnings even without linux/fs.h Signed-off-by:
-
Colin Ian King authored
The variable err is being initialized with a value that is never read, it is being updated later on. The assignment is redundant and can be removed. Addresses-Coverity: ("Unused value") Signed-off-by:Colin Ian King <colin.king@canonical.com> Reviewed-by:
-
- 09 Sep, 2021 9 commits
-
-
Kari Argillander authored
Show options should show option according documentation when some value is not default or when ever coder wants. Uid/gid are problematic because it is hard to know which are defaults. In file system there is many different implementation for this problem. Some file systems show uid/gid when they are different than root, some when user has set them and some show them always. There is also problem that what if root uid/gid change. This code just choose to show them always. This way we do not need to think this any more. Signed-off-by:
-
Kari Argillander authored
Rename mount option no_acs_rules to (no)acsrules. This allow us to use possibility to mount with options noaclrules or aclrules. Acked-by:
Christian Brauner <christian.brauner@ubuntu.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
-
Kari Argillander authored
Other fs drivers are using iocharset= mount option for specifying charset. So add it also for ntfs3 and mark old nls= mount option as deprecated. Reviewed-by:
Pali Rohár <pali@kernel.org> Signed-off-by:
-
Kari Argillander authored
If we call Opt_nohidden with just keyword hidden, then we can use hidden/nohidden when mounting. We already use this method for almoust all other parameters so it is just logical that this will use same method. Acked-by:
-
Kari Argillander authored
init_fs_context() is meant to initialize s_fs_info (spi). Move spi initializing code there which we can initialize before fill_super(). Signed-off-by:
-
Kari Argillander authored
We have now new mount api as described in Documentation/filesystems. We should use it as it gives us some benefits which are desribed here lore.kernel.org/linux-fsdevel/159646178122.1784947.11705396571718464082.stgit@warthog.procyon.org.uk/ Nls loading is changed a to load with string. This did make code also little cleaner. Also try to use fsparam_flag_no as much as possible. This is just nice little touch and is not mandatory but it should not make any harm. It is just convenient that we can use example acl/noacl mount options. Signed-off-by:
-
Kari Argillander authored
Use pointer to mount options. We want to do this because we will use new mount api which will benefit that we have spi and mount options in different allocations. When we remount we do not have to make whole new spi it is enough that we will allocate just mount options. Please note that we can do example remount lot cleaner but things will change in next patch so this should be just functional. Signed-off-by:
-
Kari Argillander authored
Remove unnecesarry remount flag handling. This does not do anything for this driver. We have already set SB_NODIRATIME when we fill super. Also noatime should be set from mount option. Now for some reson we try to set it when remounting. Lazytime part looks like it is copied from f2fs and there is own mount parameter for it. That is why they use it. We do not set lazytime anywhere in our code. So basically this just blocks lazytime when remounting. Acked-by:
-
Kari Argillander authored
Remove unnecesarry mount option noatime because this will be handled by VFS. Our option parser will never get opt like this. Acked-by:
-
- 02 Sep, 2021 4 commits
-
-
Kari Argillander authored
Usually in file system init() messages are only displayed in info level. Change level from notice to info, but keep CONFIG_NTFS3_64BIT_CLUSTER in notice level. Also this need even more attention so let's put big warning here so that nobody will not try accidentally use it. There is also no good reason to display internal stuff like binary tree search. This is always on option which can only disabled for debugging purposes by developer. Also this message does not even check if developer has disabled it or not so it is useless info. Signed-off-by:
-
Kari Argillander authored
Files already have SDPX identifier so no reason to keep boilerplates in these files anymore. Signed-off-by:
Eric Biggers <ebiggers@kernel.org> Signed-off-by:
-
Kari Argillander authored
This check will be also performed in generic_file_read_iter() so we do not want to check this two times in a row. This was founded with Smatch fs/ntfs3/file.c:803 ntfs_file_read_iter() warn: unused return: count = iov_iter_count() Signed-off-by:
-
Kari Argillander authored
Use fiemap_prep() to check valid flags. It also shrink request scope (@len) to what the fs can actually handle. This address following Smatch static checker warning: fs/ntfs3/frecord.c:1894 ni_fiemap() warn: potential integer overflow from user 'vbo + len' Because fiemap_prep() shrinks @len this cannot happened anymore. Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Link: lore.kernel.org/ntfs3/20210825080440.GA17407@kili/ Fixes: 4342306f ("fs/ntfs3: Add file operations and implementation") Signed-off-by:
-
- 31 Aug, 2021 3 commits
-
-
Konstantin Komarov authored
Signed-off-by: -
Konstantin Komarov authored
Rename now works "Add new name and remove old name". "Remove old name and add new name" may result in bad inode if we can't add new name and then can't restore (add) old name. Signed-off-by: -
Kari Argillander authored
For some reason we have FAT ioctl calls. Even old ntfs driver did not use these. We should not use these because it his hard to get things out of kernel when they are upstream. That's why we remove these for now. More discussion is needed what ioctl should be implemented and what is important. Signed-off-by:
-
- 30 Aug, 2021 1 commit
-
-
Kari Argillander authored
Capitalize comments and end with period for better reading. Also function comments are now little more kernel-doc style. This way we can easily convert them to kernel-doc style if we want. Note that these are not yet complete with this style. Example function comments start with /* and in kernel-doc style they start /**. Use imperative mood in function descriptions. Change words like ntfs -> NTFS, linux -> Linux. Use "we" not "I" when commenting code. Signed-off-by:
-
- 27 Aug, 2021 11 commits
-
-
Dan Carpenter authored
There are three bugs in this code: 1) If indx_get_root() fails, then return -EINVAL instead of success. 2) On the "/* make root external */" -EOPNOTSUPP; error path it should free "re" but it has a memory leak. 3) If indx_new() fails then it will lead to an error pointer dereference when we call put_indx_node(). I've re-written the error handling to be more clear. Fixes: 82cae269 ("fs/ntfs3: Add initialization of super block") Signed-off-by:
-
Dan Carpenter authored
The "e" pointer is dereferenced before it has been checked for NULL. Move the dereference after the NULL check to prevent an Oops. Fixes: 82cae269 ("fs/ntfs3: Add initialization of super block") Signed-off-by:
-
Dan Carpenter authored
Return -EINVAL if ni_find_attr() fails. Don't return success. Fixes: 82cae269 ("fs/ntfs3: Add initialization of super block") Signed-off-by:
-
Dan Carpenter authored
The ntfs_get_ea() function returns negative error codes or on success it returns the length. In the original code a zero length return was treated as -ENODATA and results in a NULL return. But it should be treated as an invalid length and result in an PTR_ERR(-EINVAL) return. Fixes: be71b5cb ("fs/ntfs3: Add attrib operations") Signed-off-by:
-
Dan Carpenter authored
Add a check for when the kzalloc() in init_rsttbl() fails. Some of the callers checked for NULL and some did not. I went down the call tree and added NULL checks where ever they were missing. Fixes: b46acd6a ("fs/ntfs3: Add NTFS journal") Signed-off-by:
-
Kari Argillander authored
Use kcalloc/kmalloc_array over kzalloc/kmalloc when we allocate array. Checkpatch found these after we did not use our own defined allocation wrappers. Reviewed-by:
-
Kari Argillander authored
Problem with these wrapper is that we cannot take off example GFP_NOFS flag. It is not recomended use those in all places. Also if we change one driver specific wrapper to kernel wrapper then it would look really weird. People should be most familiar with kernel wrappers so let's just use those ones. Driver specific alloc wrapper also confuse some static analyzing tools, good example is example kernels checkpatch tool. After we converter these to kernel specific then warnings is showed. Following Coccinelle script was used to automate changing. virtual patch @alloc depends on patch@ expression x; expression y; @@ ( - ntfs_malloc(x) + kmalloc(x, GFP_NOFS) | - ntfs_zalloc(x) + kzalloc(x, GFP_NOFS) | - ntfs_vmalloc(x) + kvmalloc(x, GFP_NOFS) | - ntfs_free(x) + kfree(x) | - ntfs_vfree(x) + kvfree(x) | - ntfs_memdup(x, y) + kmemdup(x, y, GFP_NOFS) ) Reviewed-by:
-
Kari Argillander authored
The static checkers (Smatch) were complaining because QuadAlign() was buggy. If you try to align something higher than UINT_MAX it got truncated to a u32. Smatch warning was: fs/ntfs3/attrib.c:383 attr_set_size_res() warn: was expecting a 64 bit value instead of '~7' So that this will not happen again we will change all these macros to kernel made ones. This can also help some other static analyzing tools to give us better warnings. Patch was generated with Coccinelle script and after that some style issue was hand fixed. Coccinelle script: virtual patch @alloc depends on patch@ expression x; @@ ( - #define QuadAlign(n) (((n) + 7u) & (~7u)) | - QuadAlign(x) + ALIGN(x, 8) | - #define IsQuadAligned(n) (!((size_t)(n)&7u)) | - IsQuadAligned(x) + IS_ALIGNED(x, 8) | - #define Quad2Align(n) (((n) + 15u) & (~15u)) | - Quad2Align(x) + ALIGN(x, 16) | - #define IsQuad2Aligned(n) (!((size_t)(n)&15u)) | - IsQuad2Aligned(x) + IS_ALIGNED(x, 16) | - #define Quad4Align(n) (((n) + 31u) & (~31u)) | - Quad4Align(x) + ALIGN(x, 32) | - #define IsSizeTAligned(n) (!((size_t)(n) & (sizeof(size_t) - 1))) | - IsSizeTAligned(x) + IS_ALIGNED(x, sizeof(size_t)) | - #define DwordAlign(n) (((n) + 3u) & (~3u)) | - DwordAlign(x) + ALIGN(x, 4) | - #define IsDwordAligned(n) (!((size_t)(n)&3u)) | - IsDwordAligned(x) + IS_ALIGNED(x, 4) | - #define WordAlign(n) (((n) + 1u) & (~1u)) | - WordAlign(x) + ALIGN(x, 2) | - #define IsWordAligned(n) (!((size_t)(n)&1u)) | - IsWordAligned(x) + IS_ALIGNED(x, 2) | ) Reported-by:
-
Kari Argillander authored
First of this fix one none utf8 char in this comment block. Maybe this happened because error in filesystem ;) Also this block was hard to read because long lines so make it max 80 long. And while we doing this stuff make little better grammer. Signed-off-by:
-
Jiapeng Chong authored
Eliminate the follow versioncheck warning: ./fs/ntfs3/inode.c: 16 linux/version.h not needed. Reported-by:
Abaci Robot <abaci@linux.alibaba.com> Fixes: 82cae269 ("fs/ntfs3: Add initialization of super block") Signed-off-by:
Jiapeng Chong <jiapeng.chong@linux.alibaba.com> Reviewed-by:
-
Gustavo A. R. Silva authored
Fix the following fallthrough warnings: fs/ntfs3/inode.c:1792:2: warning: unannotated fall-through between switch labels [-Wimplicit-fallthrough] fs/ntfs3/index.c:178:2: warning: unannotated fall-through between switch labels [-Wimplicit-fallthrough] This helps with the ongoing efforts to globally enable -Wimplicit-fallthrough for Clang. Link: https://github.com/KSPP/linux/issues/115Signed-off-by:
Gustavo A. R. Silva <gustavoars@kernel.org> Reviewed-by:
Nathan Chancellor <nathan@kernel.org> Signed-off-by:
-
