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 because we need a new pointer for LDB after the fork,
and with LMDB we can not longer rely on tdb_reopen_all() to do that
for us.
This can not be done in reinit_after_fork() due to the dependency loop
this would create.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Fix a bug when a child dies when a request is pending in the child. If the
signal handler fires before epoll finds out the other end of the parent-child
socket is closed, we close the socket on our side without taking care of the
pending request. This causes two problems: First, that one pending request
never is replied to properly, and secondly, we might end up with EPOLL_DEL on a
wrong file descriptor. This causes all sorts of trouble if we hit an active
one.
The fix for this problem is not to close the socket in winbind_child_died().
This however stops an idle child that dies hard from being properly cleaned up.
The fix for that is to add the child->monitor_fde that is set pending only when
no child request is active. This way we can remove the close(sock) in the
signal handler.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13309
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
This removes the special handling for idmap_child() after the "This is
a little tricky" comment. I believe this was not required at all, the
idmap_child is part of the winbindd_children list.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13309
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Note that we only walk the domain children, which all have
child->domain != NULL. So we don't need that check anymore.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13309
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Look at setup_domain_child(): There we always set child->domain. The only other
two children are the idmap and locator children, which don't have a domain set.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13309
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Step 0 in removing winbindd_children as a variable: We have access to
all children via our domain list and the two explicit children. There's
no need to separately maintain a list of winbind children. Maintaining
child->pid != 0 is sufficient to make sure we only walk active children.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13309
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
If we have multiple domain children, it's important
that the first idle child takes over the next waiting request.
Before we had the problem that a request could get stuck in the
queue of a busy child, while later requests could get served fine by
other children.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13292
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Fri Feb 23 09:04:23 CET 2018 on sn-devel-144
Commit ed3bc614cc got the logic wrong while
trying to implement the logic we had in init_child_connection(),
which was removed by commit d61f3626b7.
Instead of doing a WINBINDD_GETDCNAME request (which would caused an error
because the implementation was removed in commit
958fdaf5c3), we sent the callers request
and interpreted the result as WINBINDD_GETDCNAME response, which
led to an empty dcname variable. As result the domain child
opened a connection to the primary domain in order to lookup
a dc.
If we want to connect the primary domain from the parent via
a domain child of the primary domain.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13295
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
This will reduce the diff for the following changes.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13295
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
If the (winbind) client gave up we call TALLOC_FREE(state->mem_ctx)
in remove_client(). This triggers a recursive talloc_free() for all
in flight requests.
In order to maintain the winbindd parent-child protocol, we need
to keep the orphaned wb_simple_trans request until the parent
got the response from the child.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13290
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
We need a way to keep the child->queue blocked without relying on
the current 'req' (wb_child_request_state).
The next commit will make use of this.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13290
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
messaging already provides the sender id
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Fri Feb 16 00:56:36 CET 2018 on sn-devel-144
This improves the situation when a client request blocks a winbind
child. This might be a slow samlogon or lookupnames to a domain that's
far away. With random selection of the child for new request coming in
we could end up with a long queue when other, non-blocked children
could serve those new requests. Choose the shortest queue.
This is an immediate and simple fix. Step two will be to have a
per-domain and not a per-child queue. Right now we're pre-selecting
the check-out queue at Fry's randomly without looking at the queue
length. With this change we're picking the shortest queue. The better
change will be what Fry's really does: One central queue and red/green
lights on the busy/free checkout counters.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Mon Feb 12 19:51:35 CET 2018 on sn-devel-144
At lot of callers require a valid schannel connection.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=13259
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Don't use winbind_messaging_context anymore.
This fixes a bug analysed by Peter Somogyi <PSOMOGYI@hu.ibm.com>: If a
parent winbind forks, it only called reinit_after_fork on
winbind_messaging_context. On the other hand, deep in dbwrap_open we use
server_messaging_context(). This is not reinitialized by
winbind_reinit_after fork, so the parent and child share a ctdb
connection. This is invalid, because replies from ctdb end up in the
wrong process.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13150
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
There's no point in having two global event contexts
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13150
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
We have too many variables called _creds :-)
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Wed Aug 30 22:44:45 CEST 2017 on sn-devel-144
domain->methods is always set to cache_methods; call those functions
directly instead of going through the indirection.
Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
If needed we can add this into actual tevent.
Preparing to remove source3/lib/events.c
Bug: https://bugzilla.samba.org/show_bug.cgi?id=12283
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
When cancelling a request at the main winbindd process, that is currently
being served by a child winbindd process, just freeing all objects related
to the request is not enough, as the next bytes to come through the pipe
from the child process are the response to the cancelled request, and the
object reading those bytes will be the next request. This breaks the protocol.
This change, upon canceling a request that is being served, closes the
connection to the child process, causing the next request to be served
by a new child process (and the detached child to die eventually).
BUG: https://bugzilla.samba.org/show_bug.cgi?id=11358
Signed-off-by: Uri Simchoni <urisimchoni@gmail.com>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Mon Jun 29 14:00:24 CEST 2015 on sn-devel-104
Signed-off-by: Richard Sharpe <rsharpe@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed May 6 07:03:27 CEST 2015 on sn-devel-104
For consistency with request read side.
Signed-off-by: David Disseldorp <ddiss@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Mon Jan 20 14:23:10 CET 2014 on sn-devel-104
The socket and request are the only arguments required, the entire
winbind child state structure is not needed.
This allows for the separation of the request and response structures,
which is useful for asynchronous conversion.
Signed-off-by: David Disseldorp <ddiss@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
