| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Shuttle-reviewed by jra :)
Guenther
|
|
Guenther
|
|
Parts of the Samba RPC client and server code misinterpret authenticated
packets.
DCE authenticated packets actually look like this :
+--------------------------+
|header |
| ... frag_len (packet len)|
| ... auth_len |
+--------------------------+
| |
| Data payload |
... ....
| |
+--------------------------+
| |
| auth_pad_len bytes |
+--------------------------+
| |
| Auth footer |
| auth_pad_len value |
+--------------------------+
| |
| Auth payload |
| (auth_len bytes long) |
+--------------------------+
That's right. The pad bytes come *before* the footer specifying how many pad
bytes there are. In order to read this you must seek to the end of the packet
and subtract the auth_len (in the packet header) and the auth footer length (a
known value).
The client and server code gets this right (mostly) in 3.0.x -> 3.4.x so long
as the pad alignment is on an 8 byte boundary (there are some special cases in
the code for this).
Tridge discovered there are some (DRS replication) cases where on 64-bit
machines where the pad alignment is on a 16-byte boundary. This breaks the
existing S3 hand-optimized rpc code.
This patch removes all the special cases in client and server code, and allows
the pad alignment for generated packets to be specified by changing a constant
in include/local.h (this doesn't affect received packets, the new code always
handles them correctly whatever pad alignment is used).
This patch also works correctly with rpcclient using sign+seal from
the 3.4.x and 3.3.x builds (testing with 3.0.x and 3.2.x to follow)
so even as a server it should still work with older libsmbclient and
winbindd code.
Jeremy
|
|
Re-arrange the operations order so SMB_VFS_CONNECT is done
first as root (to allow modules to correctly initialize themselves).
Reviewed modules to check if they needed CONNECT invoked as
a user (which we previously did) and it turns out any of them
that cared needed root permissions anyway.
Jeremy.
|
|
At the formerly used process_result statement we have alone one
NT_STATUS_IS_OK() which never could be hit in our case as we only go here
if NT_STATUS_EQUAL is not ok.
|
|
|
|
To provide the user with the same SID when doing Kerberos logins, attempt to do
a make_server_info_sam instead of a make_server_info_pw.
|
|
|
|
Detected while showing this code to obnox :-)
|
|
There's no need to still hold the g_lock tdb-level lock while telling the
waiters to retry
|
|
In g_lock_unlock we have a little race between the process_exists and
messaging_send call: We only send to 5 waiters now, they all might have died
between us checking their existence and sending the message. This change makes
g_lock_lock retry at least once every minute.
|
|
Only notify the first 5 pending lock waiters. This avoids a thundering herd
problem that is really nasty in a cluster. It also makes acquiring a lock a bit
more FIFO, lock waiters are added to the end of the array.
|
|
Only check the existence of the lock owner in g_lock_parse, check the rest of
the records only when we got the lock successfully. This reduces the load on
process_exists which can involve a network roundtrip in the clustered case.
|
|
g_lock_parse might have thrown away entries from the locks array because the
processes were not around anymore. Don't store the orphaned entries.
|
|
|
|
We still dont get the marshalling right, disable and XP will just fall back to
level 6.
Guenther
|
|
Update the manpage accordingly.
|
|
A return code of 1 from initgroups() is OK since apparently it means
the gid has already been set. The man page doesn't mention this.
|
|
convert smbcacls, sharesec and web/
|
|
|
|
|
|
|
|
|
|
This uses another char* cast hack. Left alone for now.
|
|
|
|
the sort_query_replies() in nmblib.c is a TODO. It uses a hack that
treats a char* as a structure. I've left that one alone for now.
|
|
This one was a bit trickier. I'd appreciate it if someone else can
look over this.
|
|
|
|
|
|
We have already returned if (argc < 1) above
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
and 0 in the places where it does.
Jeremy
|
|
behavior.
Cause all exit paths to go through one place, where all cleanup is
done. change_to_root_user() for pathname operations that should succeed if
the path exists, even if the connecting user has no access.
For example, a share can now be defined with a path of /root/only/access
(where /root/only/access is a directory path with all components only
accessible to root e.g. root owned, permissions 700 on every component).
Non-root users will now correctly connect, but get ACCESS_DENIED on
all activities (which matches Windows behavior). Previously, non-root
users would get NT_STATUS_BAD_NETWORK_NAME on doing a TConX to this
share, even though it's a perfectly valid share path (just not accessible
to them).
This change was inspired by the research I did for bug #7126, which
was reported by bepi@adria.it.
As this is a change in a core function, I'm proposing to leave
this only in master for 3.6.0, not back-port to any existing releases.
This should give us enough time to decide if this is the way we want this to
behave (as Windows) or if we prefer the previous behavior.
Jeremy.
|
|
Jeremy.
|
|
The "lock spin time" parameter mimics the following Windows
setting which by default is 250ms in Windows and 200ms in Samba.
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\LanmanServer\Parameters\LockViolationDelay
When a client sends repeated, non-blocking, contending BRL requests
to a Windows server, after the first Windows starts treating these
requests as timed blocking locks with the above timeout.
As an efficiency, I've changed the behavior when this setting is 0,
to skip this logic and treat all requests as non-blocking locks.
This gives the smbd server behavior similar to the 3.0 release with
the do_spin_lock() implementation.
I've also changed the blocking lock parameter in the call from
push_blocking_lock_request() to true as all requests made in this
path are blocking by definition.
|
|
This hasn't been turned on or been capable of doing so for
many years now. Makes this jumbo function smaller...
Jeremy.
|
|
NT_STATUS_OBJECT_PATH_INVALID error"
This reverts commit 2fdd8b10c6abadd27c579e772c0482214d2363a5.
This fix is incorrect. The original code works as desired,
I made a mistake here.
Jeremy.
|
|
NT_STATUS_OBJECT_PATH_INVALID error
As tridge's comment says, we should be ignoring ACCESS_DENIED
on the share path in a TconX call, instead allowing the mount
and having individual SMB calls fail (as Windows does). The
original code erroneously caught SMB_VFS_STAT != 0 and errored
out on that.
Jeremy.
|
|
Called, from key_exists, scan_sorted_subkeys re-creates the sorted
subkeys record of the given key and then searches through it.
The race is that between creation and parsing of the sorted subkey
record, another process that stores some other subkey of the same
parent key will delete the sorted subkey record, resulting in an
WERR_BADFILE of an operation that should actually succeed.
This patch fixes the issue by wrapping the creation and parsing
into a transaction.
Michael
|
|
Michael
|
|
This made smbd crash in g_lock_lock() when trying to start a
transaction on a db with an already started transaction,
e.g. in a tcon_and_X where the share_info.tdb was not yet
initialized but share_info.tdb was already locked by another
process or writing acces to the winreg rpc pipe where the
registry tdb was already locked by another process.
What we really _want_ to do here by design is to react to
MSG_DBWRAP_G_LOCK_RETRY messages that are either sent
by a client doing g_lock_unlock or by ourselves when
we receive a CTDB_SRVID_SAMBA_NOTIFY or
CTDB_SRVID_RECONFIGURE message from ctdbd, i.e. when
either a client holding a lock or a complete node
has died.
Doing this properly involves calling tevent_loop_once(),
but doing this here with the main ctdbd messaging context
creates a nested event loop when g_lock_lock() is called
from the main event loop.
So as a quick fix, we act a little corasely here: we do
a select on the ctdb connection fd and when it is readable
or we get EINTR, then we retry without actually parsing
any ctdb packages or dispatching messages. This means that
we retry more often than necessary and intended by design,
but this does not harm and it is unobtrusive. When we have
finished, the main loop will pick up all the messages and
ctdb packets. The only extra twist is that we cannot use
timed events here but have to handcode a timeout for select.
Michael
|
|
Michael
|
