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 reverts commit bd30c9c128.
While this does indeed slightly simplify code, it simplifies too much: Soon we
will need filename_convert_dirfsp_nosymlink raw without looking at
UCF_LCOMP_LNK_OK. So in hindsight this went too far.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
OPEN_REPARSE_POINT will trigger symlinks not being followed but
returned, even if we have "follow symlinks = yes". Prepare for setting
UCF_LCOMP_LNK_OK for this case in a central place.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
fdos_mode(fsp) unconditionally dereferences fsp, so we can get rid of
checking for fsp!=NULL.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15754
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Mon Nov 18 12:43:28 UTC 2024 on atb-devel-224
Better look at the final code, not at the patch. The idea is to call
filename_convert_dirfsp() from fd_openat() and just have one place to
follow symlinks.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Tue Nov 12 19:21:11 UTC 2024 on atb-devel-224
Looks more complex, but we don't have to go through all of
fd_openat()/non_widelink_open() for opening streams. SMB_VFS_OPENAT
knows how to deal with this.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Soon we will call filename_convert_dirfsp() on these, which can't deal
with paths that are invalid by containing . and .. as path components.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Looks more complex, but this avoids calling openat_pathref_fsp, which
eventually calls into non_widelink_open(). We need to open the pretty paranoid
SMB_ASSERT in openat_pathref_fsp_lcomp() a little bit.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Get a pathref handle on the file system root. This will serve as
"basedir" for filename_convert_dirfsp_rel() to turn an absolute path
into one relative to the handle created here.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Soon we'll have a caller that needs the last component as a relative
file name. Make sure it does not have to call get_lcomp or so.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Unused so far, but soon we'll call this routine with a basedir that's
somewhere below the share root.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
This avoids a full path traversal if /proc/self/fd is available.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Use the /proc/self/fd trick to make get_real_filename_fullscan_at look
a bit nicer and faster in strace. Direct SMB_VFS_OPENAT also is
cheaper in user space, we don't need the full fd_openat and
non_widelink_open magic here.
Also avoid opening ".", which can fail where a full path open would
succeed: If the directory in question does not give "x" perms to the
user, we get a handle on the dir as such but can't cd into it.
I haven't seen real-world cases of this, but one of our tests creates
such a scenario. I have further refactoring in my local tree that make
this patch necessary.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
So far fd_openat needed to be called with cwd_fsp and "." if you
wanted to open the share root, this enables using the absolute share
path. The next patch will remove sending the "." with cwd_fsp as a
simplification. Enable that with this little change.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Make it clear that non_widelink_open being called for absolute paths
is not done during regular operations, i.e. when we have passed our
pathnames through filename_convert_dirfsp and we work on a real
dirfsp.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
We can't deal with snapdir paths in non_widelink_open fully with
snapdirseverywhere active: There is no way for
shadow_copy2_parent_pathname() to work when a snapshot directory is
below the directory that we want to calculate the parent for. What is
the parent directory supposed to point at? I don't know.
For me the only way out is to accept that we should ignore what
happens behind shadow_copy2's path manipulation in core
smbd. This *might* open symlink races, but the whole point of
snapshots is that they are r/o copies of the real active file system
and as such they should be immune to those races.
Found while trying to refactor code around fd_openat()
P.S: This code will go away pretty soon
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
If we do the S_ISLNK check in the lower level, the if-condition is
simpler and we get the close_file_free() call for free.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
A valid tcon will have changed fsp_get_pathref_fd() to AT_FDCWD, -100
on Linux.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
The next patch will add another excluding condition, this change
keeps the if-condition that is changed here simple.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
vfs_GetWd depends upon a current tcon in fake_acls, otherwise it will
call openat with an invalid dirfd on a relative pathname.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
We have to undo the fsp allocation and open we've done. Not noticed,
it's probably highly unlikely OpenDir_fsp() fails.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Tue Nov 12 13:26:10 UTC 2024 on atb-devel-224
Use reopen_from_fsp(), this will use the /proc/self/fd/<fd> trick if
available, no need to go through non_widelink_open
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
By default enabled on non-clustered Samba, disabled on clustered Samba, the
reason being the expected additional load caused by forcing strict rename to be
enabled.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
"deferred" will be used in the function by a later commit...
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
This matches the notify_fname() argument name and the next commit is going to
add an "action" variable.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Checks for Directory Lease breaks on the parent directory of smb_fname. Gets a
sharemode lock on the locking.tdb record of the directory, hence it mustn't be
called if the caller still has another sharmode lock.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
notify_fname() for NOTIFY_ACTION_DIRLEASE_BREAK will soon need the lease of the
current open and to implement "MS-FSA 2.1.4.12 Algorithm to Check for an Oplock
Break" with flags=PARENT_OBJECT.
No change in behaviour for now, all callers pass lease=NULL;
Also change path arg to struct smb_filename.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
This will be a common pattern in the upcoming code dealing with Directory Lease
breaks: when checking for Directory Lease breaks on the parent directory, we
take the sharemode lock on the parent and hence by then must have dropped the
sharemode lock on the object that performs the Directory Lease break check. That
functionality will be later added to notify_fname().
This means the notification will be sent a wee bit later, but that's something
the change notification protocol has to live with anyway.
For Directory Leases the sequence to check for Directory Lease breaks is to call
MS-FSA 2.1.4.12 "Algorithm to Check for an Oplock Break" with
flags=PARENT_OBJECT at the end of each relevant operation, eg creating a file
asf, when processing of the operation that will call 2.1.4.12 is already
completed.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Check for open files recursively when renaming a directory and wait for handle
lease breaks.
As delay_for_handle_lease_break_send() does the same check as
have_file_open_below(), remove have_file_open_below() from can_rename() so it is
not called twice for SMB2 renames, and add calls to have_file_open_below() to the
SMB1 entry rename entry points.
This is a bit ugly, but I don't see any other good way of doing this.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Let have_file_open_below() be the single function to check if
"have_file_open_below" and let it check internally whether to just search the
fsp list in the process or traversing locking.tdb based on the setting of
"strict rename".
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
In the future not all callers will have a share_mode_lock around.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
All existing callers use share_mode_forall_read(), so no change in behaviour.
Note: doing the indirection via the function pointers "ro_fn" and "rw_fn" in a
single state "struct share_mode_forall_state" avoids duplicating
share_mode_forall_dump_fn() and share_mode_forall_fn() and has the benefit of
code sharing of these functions for both read-only and read-write cases.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
This ensures common cleanup code via
smbXsrv_connection_shutdown_send() ->
-> smbXsrv_session_disconnect_xconn()
-> smbXsrv_session_remove_channel()
-> smb2srv_session_shutdown_send()
is used if the last (only) connection goes away as well. In the future this
should be implemented for the
xconn->has_cluster_movable_ip
case.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Add a version of delay_rename_for_lease_break() that is usable in other places
where we have to check for handle lease breaks. No change in behaviour.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
in_input_buffer just points into the smbd_smb2_request iovecs data which is
guarenteed to have the same lifetime as the deferred rename processing, no need
to make a copy.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15608
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
We must also break leases on other opens if the open of the rename doesn't have
a lease itself. The existing test test_lease_v2_rename() that was added
alongside the deferred rename server code didn't cover this case.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15697
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Thu Oct 31 12:47:24 UTC 2024 on atb-devel-224
For a compound related request chain of eg CREATE+NOTIFY+GETINFO, the NOTIFY
will typically go async. When this is noted in smbd_smb2_request_pending_queue()
the pending async tevent_req is cancelled which means we return
NT_STATUS_CANCELLED to the client while Windows returns
NT_STATUS_INTERNAL_ERROR.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15697
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
This was missing from commit 6140c3177a and causes
all opens of directories to be handled as stat opens, bypassing the sharemode
check.
Not adding a test at this time, as my (hopefully) soon to be merged Directory
Leases branch has a test which actually detected this problem.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15732
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Tue Oct 29 12:44:49 UTC 2024 on atb-devel-224
Essentially FSP_POSIX_FLAGS_OPEN implies FSP_POSIX_FLAGS_PATHNAMES, so we can
just remove FSP_POSIX_FLAGS_PATHNAMES and simplify things.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
This option of the vfs_fruit VFS module that could be used to enable POSIX
directory rename behaviour for OS X clients has been removed as it could result
in severe problems for Windows clients.
As a possible workaround it is possible to prevent creation of .DS_Store files
(a Finder thingy to store directory view settings) on network mounts by running
$ defaults write com.apple.desktopservices DSDontWriteNetworkStores true
on the Mac.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15732
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Mon Oct 14 12:23:04 UTC 2024 on atb-devel-224
So far, the create context is used as it is sent by the client.
Now we first check whether posix extensions are negotiated.
Pair-Programmed-With: Ralph Boehme <slow@samba.org>
Signed-off-by: Jule Anger <janger@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
stat opens should not cause a oplock/lease downgrade if
they don't have a lease attached to itself.
Note that opens broken to NONE still count if they are
non-stat opens...
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15649
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15651
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Thu Oct 10 13:59:18 UTC 2024 on atb-devel-224
We create a new helper function to retrieve the fs capabilties via
STATVFS call. Additionally set other capabilities based on specific
parametric options. This is 99.9% taken from SMB_VFS_FS_CAPABILITIES
implementation of vfs_default in preparation to be called from any
required vfs module.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15716
Signed-off-by: Anoop C S <anoopcs@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
This fixes a O(n²) performance regression in notifyd. The problem was
that we had a watcher per notify instance. This changes the code to have
a watcher per notify db entry.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14430
Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Tue Oct 1 14:22:43 UTC 2024 on atb-devel-224
