IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
This is required for the shadow_copy2 module and "wide links = no". The file
system snapshots by nature are typically outside of share directory. So the
REALPATH result fails the wide links = no test.
This patch introduces
struct stat_ex {
dev_t st_ex_dev;
ino_t st_ex_ino;
mode_t st_ex_mode;
nlink_t st_ex_nlink;
uid_t st_ex_uid;
gid_t st_ex_gid;
dev_t st_ex_rdev;
off_t st_ex_size;
struct timespec st_ex_atime;
struct timespec st_ex_mtime;
struct timespec st_ex_ctime;
struct timespec st_ex_btime; /* birthtime */
blksize_t st_ex_blksize;
blkcnt_t st_ex_blocks;
};
typedef struct stat_ex SMB_STRUCT_STAT;
It is really large because due to the friendly libc headers playing macro
tricks with fields like st_ino, so I renamed them to st_ex_xxx.
Why this change? To support birthtime, we already have quite a few #ifdef's at
places where it does not really belong. With a stat struct that we control, we
can consolidate the nanosecond timestamps and the birthtime deep in the VFS
stat calls.
At this moment it is triggered by a request to support the birthtime field for
GPFS. GPFS does not extend the system level struct stat, but instead has a
separate call that gets us the additional information beyond posix. Without
being able to do that within the VFS stat calls, that support would have to be
scattered around the main smbd code.
It will very likely break all the onefs modules, but I think the changes will
be reasonably easy to do.
This is the first of a series of patches that change path based
operations to operate on a struct smb_filename instead of a char *.
This same concept already exists in source4.
My goals for this series of patches are to eventually:
1) Solve the stream vs. posix filename that contains a colon ambiguity
that currently exists.
2) Make unix_convert the only function that parses the stream name.
3) Clean up the unix_convert API.
4) Change all path based vfs operation to take a struct smb_filename.
5) Make is_ntfs_stream_name() a constant operation that can simply
check the state of struct smb_filename rather than re-parse the
filename.
6) Eliminate the need for split_ntfs_stream_name() to exist.
My strategy is to start from the inside at unix_convert() and work my
way out through the vfs layer, call by call. This first patch does
just that, by changing unix_convert and all of its callers to operate
on struct smb_filename. Since this is such a large change, I plan on
pushing the patches in phases, where each phase keeps full
compatibility and passes make test.
The API of unix_convert has been simplified from:
NTSTATUS unix_convert(TALLOC_CTX *ctx,
connection_struct *conn,
const char *orig_path,
bool allow_wcard_last_component,
char **pp_conv_path,
char **pp_saved_last_component,
SMB_STRUCT_STAT *pst)
to:
NTSTATUS unix_convert(TALLOC_CTX *ctx,
connection_struct *conn,
const char *orig_path,
struct smb_filename *smb_fname,
uint32_t ucf_flags)
Currently the smb_filename struct looks like:
struct smb_filename {
char *base_name;
char *stream_name;
char *original_lcomp;
SMB_STRUCT_STAT st;
};
One key point here is the decision to break up the base_name and
stream_name. I have introduced a helper function called
get_full_smb_filename() that takes an smb_filename struct and
allocates the full_name. I changed the callers of unix_convert() to
subsequently call get_full_smb_filename() for the time being, but I
plan to eventually eliminate get_full_smb_filename().
The aio_fork module does not need this, as it does not communicate via signals
but with pipes. Watching a strace log with those become_root() calls in aio.c
is absolutely awful, and it does affect performance.
Often times before creating a file, a client will first query to see
if it already exists. Since some systems have a case-insensitive stat
that is called from unix_convert, we can definitively return
STATUS_NO_SUCH_FILE to the client without scanning the whole
directory.
This code path is taken from trans2querypathinfo, but trans2findfirst
still does a full directory scan even though the get_real_filename
(the case-insensitive stat vfs call) can prevent this.
This patch adds the get_real_filename call to the trans2find* path,
and also changes the vfs_default behavior for
SMB_VFS_GET_REAL_FILENAME. Previously, in the absence of a
get_real_filename implementation, we would fallback to the full
directory scan. The default behavior now returns -1 and sets errno to
EOPNOTSUPP. This allows SMB_VFS_GET_REALFILENAME to be called from
trans2* and unix_convert.
This bug prompted several, fairly large changes to the of OneFS's
readdirplus() within Samba.
One fundamental problem is that we kept our cache cursor pointed at the
next entry to be returned from onefs_readdir(), while the resume cookie
needed to refill the cache such that our cursor would be on this entry,
was located in the previous cache entry. This meant that to correctly handle
seekdir() cases which could be found within the existing cache, and cases
where a cache reload was needed, required that the cache always hold
at least two entries: the entry we wished to return, and the previous entry
which held the resume cookie. Since the readdirplus() syscall gives us no
guarantee that it will always return these two direntries, there was a
fundamental problem with this design.
To fix this problem, I have rearchitected the onefs_readdir() path to keep
its pointer on the entry which contains the resume_cookie, not the entry
which will be returned next. Essentially, I changed onefs_readdir() from a
"return an entry then increment the cursor" model to "increment the cursor
then return an entry". By doing this, we only require that a single entry
be within the cache: the entry containing the resume cookie.
Second, there have been numerous off-by-one bugs in my implementation of
onefs_seekdir() which did a mapping between the 64-bit resume cookie
returned by readdirplus() and its own monotonically increasing "location"
offset. Furthermore, this design caused a somewhat frequent waste of
cycles, as in some cases we'd need to re-enumerate the entire directory to
recover the current "location" from an old resume cookie. As this code was
somewhat difficult to understand, prone to bugs, and innefficient in some
cases I decided it was better to wholesale replace it now, rather than later.
It is possible to algorithmically map the 64-bit resume cookies from
readdirplus() into 32-bit offset values which SMB requires. The onefs.so
module now calls into a system library to do this conversion. This greatly
simplifies both the seekdir() and telldir() paths and is more efficient.
Not being able to open the shadow copy directory is the same as having no
shadow copy support at all. The VFS module should in this case not log with
debug level 0 and set ENOSYS to indicate "no shadow copies used" to the higher
levels.
