Andrew Tridgell [Mon, 21 May 2007 23:35:14 +0000 (23:35 +0000)]
r23057: only call the async recv function for the first pending receive, not
all pending receives, when the transport dies. This is because the
async callback most commonly shuts down the connection (it's the only
reasonable thing to do when it's dead), and that frees the whole
context. That means that if we loop more than once, we'll end up using
freed memory.
r23044: - use uint32_t for handle id's
- include the session vuid in the SMB2 128-Bit wire handles
as SMB2 oplock breaks doesn't include a TID or VUID in the header
we need to make sure the handle is unique for the whole TCP connection
Andrew Tridgell [Sun, 20 May 2007 08:01:02 +0000 (08:01 +0000)]
r23017: fixed the warning we have been getting for a long time:
pvfs_close: failed to delete XXX
during the BASE-DELETE test. It was a real bug, and could result in a
delete on close triggering for a handle that had never fully opened.
r22993: - make it possible to load a dsdb_schema from ldif via the ejs bindings
- make it possible to set ntds_objectGUID and ntds_invocationId via the ejy bindings
r22992: add a function to attach the dsdb_schema to a ldb_context
loaded from ldif files. The first file should contain
the info's about the 'prefixMap' attribute in the first ldif chunk.
The 2nd file should contain all attributes and classes.
Andrew Tridgell [Thu, 17 May 2007 10:33:40 +0000 (10:33 +0000)]
r22972: added the basic ldif needed to support group policies in Samba4. WinXP
clients do correctly see our group policies, but the gpmc admin tool
doesn't yet work to allow you to edit the policies
Andrew Tridgell [Thu, 17 May 2007 08:47:04 +0000 (08:47 +0000)]
r22969: fix some more places where we could end up with more than one event
context. We now have an event context on the torture_context, and we
can also get one from the cli_credentials structure
Andrew Tridgell [Thu, 17 May 2007 07:52:33 +0000 (07:52 +0000)]
r22968: andrew bartlett pointed out that the parent process could add a fd to
the epoll context which would then appear in the children. To fix this
we need to check for pid changes in more places. Luckily on platforms
where we have epoll(), getpid() is very very cheap.
Andrew Tridgell [Thu, 17 May 2007 02:18:29 +0000 (02:18 +0000)]
r22959: cope with a rather interesting interaction between epoll() and
fork(). See
http://junkcode.samba.org/ftp/unpacked/junkcode/epoll_fork.c for why
this is needed
Ronnie Sahlberg [Thu, 17 May 2007 01:17:43 +0000 (01:17 +0000)]
r22958: For SRVSVC/NetFileClose only Administrator is allowed to close open files.
If a normal user tries to close a file that exists, even that users own files the server
responds with this error on w2k
if the file does not exist, the server instead responds with WERR_BADFILE
Michael Adam [Wed, 16 May 2007 11:15:16 +0000 (11:15 +0000)]
r22931: Fix logic in detection of the need to replace dlopen and friends.
Originally, dlfcn.o was only added to LIBREPLACEOBJ if dlopen
was found in libdl but header dlfcn.h was not appropriate.
Volker Lendecke [Tue, 15 May 2007 08:01:30 +0000 (08:01 +0000)]
r22886: Workaround to make RAW-NOTIFY a bit more reliable against Samba3. It
might take a bit for events to propagate from one to another connection,
in particular with a process per connection.
r22858: - let SMB2-LOCK-VALID-REQUEST pass against longhorn beta3
- add modify the SMB2-LOCK-BLOCK-WRITE test to also test reading
and name in SMB2-LOCK-RW-EXCLUSIV
- add SMB2-LOCK-NONE and SMB2-LOCK-SHARED
Andrew Bartlett [Mon, 14 May 2007 05:53:26 +0000 (05:53 +0000)]
r22838: Add in an explority test for what QFSINFO operations are valid on IPC$
It seems most opertaions are supported, but that the values are not
consistant. I think these are very much hand-hacked stubs, which is
what they will be on Samba4 too.
I'll need to add some more 'don't fail for...' hacks before this
passes against Win2k3.
Andrew Tridgell [Mon, 14 May 2007 01:04:01 +0000 (01:04 +0000)]
r22834: fixed a memory leak in the torture_open_connection() code, and removed
the duplicate handling of the unclist. This now exposes a function
torture_get_conn_index() which can be used by tests to work out which
host to talk to using the unclist. I will be using that as part of a
set of tests for clustered Samba that need to do auto-reconnect (to
allow testing of cluster node failure)
Andrew Tridgell [Mon, 14 May 2007 00:57:48 +0000 (00:57 +0000)]
r22830: merged the latest lib/events updates from ctdb to Samba4. This
includes a new EVENT_FD_AUTOCLOSE flag that prevents race conditions
where code using fd events might close a fd before releasing the
struct fd_event. That causes headaches for epoll.