David Disseldorp [Thu, 20 Jul 2017 09:10:57 +0000 (11:10 +0200)]
build: fix ceph_statx check when configured with libcephfs_dir
When configured with a custom libcephfs_dir, the ceph_statx check fails
to link. This is due to the location of the ceph-common dependency,
which is installed under a ceph subdirectory.
Signed-off-by: David Disseldorp <ddiss@suse.de> Reviewed-by: Jeff Layton <jlayton@samba.org>
(cherry picked from commit ff7df3d3f5259362a6bb6780d6b532e57e89681d)
Autobuild-User(v4-6-test): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(v4-6-test): Tue Feb 13 13:36:13 CET 2018 on sn-devel-144
Ralph Boehme [Wed, 6 Dec 2017 21:09:52 +0000 (22:09 +0100)]
vfs_fruit: set delete-on-close for empty finderinfo
We previously removed the stream from the underlying filesystem stream
backing store when the client zeroes out FinderInfo in the AFP_AfpInfo
stream, but this causes certain operations to fail (eg stat) when trying
to access the stream over any file-handle open on that stream.
So instead of deleting, set delete-on-close on the stream. The previous
commit already implemented not to list list streams with delete-on-close
set which is necessary to implemenent correct macOS semantics for this
particular stream.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Tue Jan 9 17:09:12 CET 2018 on sn-devel-144
Ralph Boehme [Thu, 7 Dec 2017 16:32:35 +0000 (17:32 +0100)]
vfs_fruit: filter out AFP_AfpInfo streams with pending delete-on-close
This is in preperation of fixing the implementation of removing the
AFP_AfpInfo stream by zeroing the FinderInfo out.
We currently remove the stream blob from the underyling filesystem
backing store, but that results in certain operations to fail on any
still open file-handle.
The fix comes in the next commit which will convert to backing store
delete operation to a set delete-on-close on the stream.
This commit adds filtering on streams that have the delete-on-close
set. It is only needed for the fruit:metadata=stream case, as with
fruit:metadata=netatalk the filtering is already done in
fruit_streaminfo_meta_netatalk().
Ralph Boehme [Thu, 7 Dec 2017 12:43:02 +0000 (13:43 +0100)]
s4/torture/fruit: enhance zero AFP_AfpInfo stream test
This test more operations in the zeroed out FinderInfo test, ensuring
after zeroing out FinderInfo, operations on the filehandle still work
and that enumerating streams doesn't return the stream anymore.
David Disseldorp [Wed, 10 Jan 2018 13:03:09 +0000 (14:03 +0100)]
vfs_default: use VFS statvfs macro in fs_capabilities
Currently the vfs_default fs_capabilities handler calls statvfs
directly, rather than calling the vfs macro. This behaviour may cause
issues for VFS modules that delegate fs_capabilities handling to
vfs_default but offer their own statvfs hook.
Signed-off-by: David Disseldorp <ddiss@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
(cherry picked from commit 4b25c9f4a4d336a16894452862ea059701b025de)
Autobuild-User(v4-6-test): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(v4-6-test): Mon Jan 22 13:14:03 CET 2018 on sn-devel-144
David Disseldorp [Wed, 10 Jan 2018 00:37:14 +0000 (01:37 +0100)]
vfs_ceph: add fs_capabilities hook to avoid local statvfs
Adding the fs_capabilities() hook to the CephFS VFS module avoids
fallback to the vfs_default code-path, which calls statvfs() against the
share path on the *local* filesystem.
Signed-off-by: David Disseldorp <ddiss@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
(cherry picked from commit 2724e0cac29cd1632ea28075a740fcc888affb36)
Jeremy Allison [Wed, 3 Jan 2018 17:52:33 +0000 (09:52 -0800)]
s3: smbd: Use identical logic to test for kernel oplocks on a share.
Due to inconsistent use of lp_kernel_oplocks() we could miss kernel
oplocks being on/off in some of our oplock handling code, and thus
use the wrong logic.
Ensure all logic around koplocks and lp_kernel_oplocks() is consistent.
Christof Schmitt [Wed, 13 Dec 2017 18:34:23 +0000 (11:34 -0700)]
smbd: Fix coredump on failing chdir during logoff
server_exit does an internal tree disconnect which requires a chdir to
the share directory. In case the file system encountered a problem and
the chdir call returns an error, this triggers a SERVER_EXIT_ABNORMAL
which in turn results in a panic and a coredump. As the log already
indicates the problem (chdir returned an error), avoid the
SERVER_EXIT_ABNORMAL in this case and not trigger a coredump.
Signed-off-by: Christof Schmitt <cs@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Sat Dec 16 01:56:06 CET 2017 on sn-devel-144
This module allow injecting errors in vfs calls. It only implements one
case (return ESTALE from chdir), but the idea is to extend this to more
vfs functions and more errors when needed.
If PULL_DB control times out but the remote node is still sending the
data, then the tevent_req for pull_database_send will be freed without
removing the message handler. So when the data is received, srvid
handler will be called and it will try to access tevent_req which will
result in use-after-free and abort.
Signed-off-by: Amitay Isaacs <amitay@gmail.com> Reviewed-by: Martin Schwenke <martin@meltin.net>
Uri Simchoni [Tue, 5 Dec 2017 18:56:49 +0000 (20:56 +0200)]
sysacls: change datatypes to 32 bits
The SMB_ACL_PERMSET_T and SMB_ACL_PERM_T were defined as
mode_t, which is 16-bits on some (non-Linux) systems. However,
pidl *always* encodes mode_t as uint32_t. That created a bug on
big-endian systems as sys_acl_get_permset() returns a SMB_ACL_PERMSET_T
pointer to an internal a_perm structure member defined in IDL as a mode_t,
which pidl turns into a uin32_t in the emitted header file.
Uri Simchoni [Tue, 5 Dec 2017 18:49:03 +0000 (20:49 +0200)]
pysmbd: fix use of sysacl API
Fix pysmbd to use the sysacl (POSIX ACL support) as intended, and
not assume too much about the inner structure and implementation
of the permissions in the sysacl API.
This will allow the inner structure to change in a following commit.
Reviewed-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Andrew Bartlett <abartlet@samba.org>
(cherry picked from heimdal commit 19f9fdbcea11013cf13ac72c416f161ee55dee2b)
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Mon Aug 28 15:10:54 CEST 2017 on sn-devel-144
Reviewed-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Andrew Bartlett <abartlet@samba.org>
(cherry picked from heimdal commit e8317b955f5a390c4f296871ba6987ad05478c95)
g_lock: fix cleanup of stale entries in g_lock_trylock()
g_lock_trylock() always incremented the counter 'i', even after cleaning a stale
entry at position 'i', which means it skipped checking for a conflict against
the new entry at position 'i'.
As result a process could get a write lock, while there're still
some read lock holders. Once we get into that problem, also more than
one write lock are possible.
Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Wed Dec 20 20:31:48 CET 2017 on sn-devel-144
(similar to commit 576fb4fb5dc506bf55e5cf87973999dca444149b)
Autobuild-User(v4-6-test): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(v4-6-test): Fri Dec 22 22:11:00 CET 2017 on sn-devel-144
Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Wed Dec 6 23:16:54 CET 2017 on sn-devel-144
Volker Lendecke [Thu, 30 Nov 2017 20:06:53 +0000 (21:06 +0100)]
messaging: Always register the unique id
The winbind child does not call serverid_register, so the unique id is not
registered. ctdbd_process_exists now calls CTDB_CONTROL_CHECK_PID_SRVID, which
then fails.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=13180 Signed-off-by: Volker Lendecke <vl@samba.org>
Autobuild-User(v4-7-test): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(v4-7-test): Fri Dec 15 15:35:25 CET 2017 on sn-devel-144
Volker Lendecke [Wed, 29 Nov 2017 15:45:40 +0000 (16:45 +0100)]
pthreadpool: Fix starvation after fork
After the race is before the race:
1) Create an idle thread
2) Add a job: This won't create a thread anymore
3) Immediately fork
The idle thread will be woken twice before it's actually woken up: Both
pthreadpool_add_job and pthreadpool_prepare_pool call cond_signal, for
different reasons. We must look at pool->prefork_cond first because otherwise
we will end up in a blocking job deep within a fork call, the helper thread
must take its fingers off the condvar as quickly as possible. This means that
after the fork there's no idle thread around anymore that would pick up the job
submitted in 2). So we must keep the idle threads around across the fork.
The quick solution to re-create one helper thread in pthreadpool_parent has a
fatal flaw: What do we do if that pthread_create call fails? We're deep in an
application calling fork(), and doing fancy signalling from there is really
something we must avoid.
This has one potential performance issue: If we have hundreds of idle threads
(do we ever have that) during the fork, the call to pthread_mutex_lock on the
fork_mutex from pthreadpool_server (the helper thread) will probably cause a
thundering herd when the _parent call unlocks the fork_mutex. The solution for
this to just keep one idle thread around. But this adds code that is not
strictly required functionally for now.
More detailed explanation from Jeremy:
First, understanding the problem the test reproduces:
add a job (num_jobs = 1) -> creates thread to run it.
job finishes, thread sticks around (num_idle = 1).
num_jobs is now zero (initial job finished).
a) Idle thread is now waiting on pool->condvar inside
pthreadpool_server() in pthread_cond_timedwait().
Now, add another job ->
pthreadpool_add_job()
-> pthreadpool_put_job()
This adds the job to the queue.
Oh, there is an idle thread so don't
create one, do:
pthread_cond_signal(&pool->condvar);
and return.
Now call fork *before* idle thread in (a) wakes from
the signaling of pool->condvar.
In the parent (child is irrelevent):
Go into: pthreadpool_prepare() ->
pthreadpool_prepare_pool()
Set the variable to tell idle threads to exit:
pool->prefork_cond = &prefork_cond;
then wake them up with:
pthread_cond_signal(&pool->condvar);
This does nothing as the idle thread
is already awoken.
b) Idle thread wakes up and does:
Reduce idle thread count (num_idle = 0)
pool->num_idle -= 1;
Check if we're in the middle of a fork.
if (pool->prefork_cond != NULL) {
Yes we are, tell pthreadpool_prepare()
we are exiting.
pthread_cond_signal(pool->prefork_cond);
And exit.
pthreadpool_server_exit(pool);
return NULL;
}
So we come back from the fork in the parent with num_jobs = 1,
a job on the queue but no idle threads - and the code that
creates a new thread on job submission was skipped because
an idle thread existed at point (a).
OK, assuming that the previous explaination is correct, the
fix is to create a new pthreadpool context mutex:
pool->fork_mutex
and in pthreadpool_server(), when an idle thread wakes up and
notices we're in the prepare fork state, it puts itself to
sleep by waiting on the new pool->fork_mutex.
And in pthreadpool_prepare_pool(), instead of waiting for
the idle threads to exit, hold the pool->fork_mutex and
signal each idle thread in turn, and wait for the pool->num_idle
to go to zero - which means they're all blocked waiting on
pool->fork_mutex.
When the parent continues, pthreadpool_parent()
unlocks the pool->fork_mutex and all the previously
'idle' threads wake up (and you mention the thundering
herd problem, which is as you say vanishingly small :-)
and pick up any remaining job.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=13179 Signed-off-by: Volker Lendecke <vl@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
(cherry picked from commit f6858505aec9f1004aeaffa83f21e58868749d65)
Ralph Boehme [Mon, 9 Oct 2017 11:29:05 +0000 (13:29 +0200)]
winbindd: idmap_rid: error code for failing id-to-sid mapping request
NT_STATUS_NO_SUCH_DOMAIN triggers complete request failure in the parent
winbindd. By returning NT_STATUS_NONE_MAPPED winbindd lets the individual
mapping fail but keeps processing any remaining mapping requests.
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): Tue Oct 10 19:57:37 CEST 2017 on sn-devel-144
Ralph Boehme [Mon, 25 Sep 2017 13:42:08 +0000 (15:42 +0200)]
winbindd: idmap_rid: don't rely on the static domain list
The domain list in the idmap child is inherited from the parent winbindd
process and may not contain all domains in case enumerating trusted
domains didn't finish before the first winbind request that triggers the
idmap child fork comes along.
The previous commits added the domain SID as an additional argument to
the wbint_UnixIDs2Sids request, storing the domain SID in struct
idmap_domain.
Ralph Boehme [Mon, 25 Sep 2017 13:39:39 +0000 (15:39 +0200)]
winbindd: pass domain SID to wbint_UnixIDs2Sids
This makes the domain SID available to the idmap child for
wbint_UnixIDs2Sids mapping request. It's not used yet anywhere, this
comes in the next commit.
Ralph Boehme [Mon, 25 Sep 2017 11:25:57 +0000 (13:25 +0200)]
winbindd: add domain SID to idmap mapping domains
Fetch the domain SID for every domain in the idmap-domain map. This is
in preperation of passing the domain SID as an additional argument to
xid2sid requests to the idmap child.
Jeremy Allison [Wed, 29 Nov 2017 21:16:43 +0000 (13:16 -0800)]
s3: libsmb: Fix reversing of oldname/newname paths when creating a reparse point symlink on Windows from smbclient.
This happened as smbd doesn't support reparse points so we couldn't test.
This was the reverse of the (tested) symlink parameters in the unix extensions
symlink command.
Rename parameters to link_target instead of oldname so this is clearer.
Christof Schmitt [Tue, 28 Nov 2017 17:59:06 +0000 (10:59 -0700)]
pthreadpool: Undo put_job when returning error
When an error is returned to the caller of pthreadpool_add_job, the job
should not be kept in the internal job array. Otherwise the caller might
free the data structure and a later worker thread would still reference
it.
When it is not possible to create a single worker thread, the system
might be out of resources or hitting a configured limit. In this case
fall back to calling the job function synchronously instead of raising
the error to the caller and possibly back to the SMB client.
STARTUP control is primarily used to synchronise tcp tickles from running
nodes to a node which has just started up. Earlier STARTUP control was
sent (using BROADCAST_ALL) after setup event. Once the other nodes in
the cluster connected to this node, the queued up messages would be sent
and the tcp tickles would get synchronised.
Recent fix to drop messages to disconnected or not-yet-connected nodes,
the STARTUP control was never sent to the remote nodes and the tcp
tickles did not get synchronised.
To fix this problem send the STARTUP control (using BROADCAST_CONNECTED)
after startup event. By this time all the running nodes in the cluster
are connected.
Signed-off-by: Amitay Isaacs <amitay@gmail.com> Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Thu Nov 30 15:29:48 CET 2017 on sn-devel-144
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Thu Nov 2 03:16:11 CET 2017 on sn-devel-144
Signed-off-by: Jeremy Allison <jra@samba.org> Reviewed-by: Ralph Böhme <slow@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Thu Nov 30 05:47:12 CET 2017 on sn-devel-144
Noel Power [Wed, 29 Nov 2017 12:52:32 +0000 (13:52 +0100)]
testprogs: Fix a typo in the net ads test
Signed-off-by: Noel Power <noel.power@suse.com> Reviewed-by: Andreas Schneider <asn@samba.org> Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Thu Nov 30 01:47:24 CET 2017 on sn-devel-144
Signed-off-by: Noel Power <noel.power@suse.com> Reviewed-by: Andreas Schneider <asn@samba.org> Reviewed-by: Ralph Boehme <slow@samba.org>
(cherry picked from commit 4be05c835e9d8b8f13856d592aaf42b40ce397c2)
Signed-off-by: Noel Power <noel.power@suse.com> Reviewed-by: Andreas Schneider <asn@samba.org> Reviewed-by: Ralph Boehme <slow@samba.org>
(cherry picked from commit 3048ae318fc8b4d1b7663826972306372430a463)
Ralph Boehme [Wed, 15 Nov 2017 15:52:48 +0000 (16:52 +0100)]
vfs_fruit: proper VFS-stackable conversion of FinderInfo
This fixes the problem that conversion failed with
fruit:metadata=stream. Before we were calling ad_set() which stores the
metadata in the Netatalk compatible format.
Rewrite to fully go through the VFS by calling SMB_VFS_CREATE_FILE() and
SMB_VFS_PWRITE().
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed Nov 29 08:38:06 CET 2017 on sn-devel-144
Ralph Boehme [Fri, 17 Nov 2017 11:53:42 +0000 (12:53 +0100)]
s4/torture: fruit: remove use of localdir from test_adouble_conversion test
The previous use of localdir and torture_setup_local_file() was
motivated by the fact that by default vfs_fruit rejects access to files
with a "._" prefix.
Since a previous commit allowed SMB access to ._ files, rewrite the
test_adouble_conversion() test to create the ._ AppleDouble file over
SMB.
This also renders torture_setup_local_file() obsolete.
Ralph Boehme [Thu, 16 Nov 2017 06:58:34 +0000 (07:58 +0100)]
s4/torture: rework stream names tests usage of local xattr call
Previously this test, that tests for correct conversion of ':' in stream
names, only worked with streams_xattr with "fruit:metadata" set to
"netatalk".
In order to have test coverage for fruit shares with other configs,
split the test into two:
one test creates the stream over SMB and run against all shares, the
other one is the unmodified existing test and is only run against the
share with streams_xattr and fruit:metadata=netatalk.
Ralph Boehme [Wed, 15 Nov 2017 17:38:41 +0000 (18:38 +0100)]
selftest: add localdir option to fruit subtests
A subsequent commits modifies an existing tests that needs $localdir to
also run against "vfs_fruit_metadata_stream" and
"vfs_fruit_stream_depot". This reveals test failures, those will be
fixed in a subsequent commit.
Ralph Boehme [Tue, 21 Nov 2017 13:34:28 +0000 (14:34 +0100)]
s3/loadparm: don't mark IPC$ as autoloaded
A related problem that affects configuration for the hidden IPC$
share. This share is marked a "autoloaded" and such shares are not
reloaded when requested. That resulted in the tcon to IPC$ still using
encrpytion after running the following sequence of changes:
In 6a the client simply encrypted packets on the IPC$ tcon. In 6b the
client got a tcon failure with NT_STATUS_ACCESS_DENIED, but silently
ignore the error.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Nov 28 02:02:37 CET 2017 on sn-devel-144
Ralph Boehme [Tue, 21 Nov 2017 13:28:48 +0000 (14:28 +0100)]
s3/loadparm: ensure default service options are not changed
Rename sDefault to _sDefault and make it const. sDefault is make a copy
of _sDefault in in the initialisation function lp_load_ex().
As we may end up in setup_lp_context() without going through
lp_load_ex(), sDefault may still be uninitialized at that point, so I'm
initializing lp_ctx->sDefault from _sDefault.
Ralph Boehme [Wed, 22 Nov 2017 10:49:57 +0000 (11:49 +0100)]
s3/loadparm: allocate a fresh sDefault object per lp_ctx
This is in preperation of preventing direct access to sDefault in all
places that currently modify it.
As currently s3/loadparm is afaict not accessing lp_ctx->sDefault, but
changes sDefault indirectly through lp_parm_ptr() this change is just a
safety measure to prevent future breakage.
Signed-off-by: Timur I. Bakeyev <timur@iXsystems.com> Reviewed-by: Volker Lendecke <vl@samba.org> Reviewed-by: Uri Simchoni <uri@samba.org>
(cherry picked from commit 8034b88d4e771663c4b7b581fb6b1992c33d5d96)
Signed-off-by: Timur I. Bakeyev <timur@iXsystems.com> Reviewed-by: Volker Lendecke <vl@samba.org> Reviewed-by: Uri Simchoni <uri@samba.org>
(cherry picked from commit e9e4cd4d2b6ae376cba90fdb56b9c9684f2dc492)
Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
(cherry picked from commit 9466796c87cc4ca8d32da553421cd8ecef1bb8e4)
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): Sat Nov 18 04:07:24 CET 2017 on sn-devel-144
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.
Signed-off-by: Jeremy Allison <jra@samba.org> Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Wed Nov 15 19:50:54 CET 2017 on sn-devel-144
Signed-off-by: Jeremy Allison <jra@samba.org> Reviewed-by: Andreas Schneider <asn@samba.org>
(cherry picked from commit aaa52ab7b5ae711b80e3967ab1ecc91888c346f6)
Signed-off-by: Jeremy Allison <jra@samba.org> Reviewed-by: Andreas Schneider <asn@samba.org>
(cherry picked from commit eefc7a27155b70d027b1193187dd435267d863ea)
Autobuild-User(v4-6-test): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(v4-6-test): Fri Nov 17 13:59:02 CET 2017 on sn-devel-144
projects / thirdparty / samba.git / log
