Łukasz Stelmach [Wed, 12 Nov 2014 11:44:21 +0000 (12:44 +0100)]
Revert "metadata: drop support for kernels < v3.15"
This reverts commit
53711b606ed83f9b8dd7dc1f3832a01bd6409d79.
Łukasz Stelmach [Wed, 12 Nov 2014 11:07:57 +0000 (12:07 +0100)]
Revert "metadata: follow 3.17 changes for current->start_time"
This reverts commit
47971994 to make kdbus work on 3.14 again.
Conflicts:
metadata.c
Change-Id: I6999069a4e4b0bb8632cdbe86e25bacdbcfc8742
Maciej Wereski [Thu, 8 May 2014 11:31:19 +0000 (13:31 +0200)]
Update to 0.
20140325.112233cet
Change-Id: I09d20fe82a92eb0b9cda81b882d622b496b9cc6b
Signed-off-by: Maciej Wereski <m.wereski@partner.samsung.com>
Maciej Wereski [Wed, 5 Mar 2014 08:51:02 +0000 (09:51 +0100)]
Update to 0.
20140224.202407cet
Change-Id: Ie211816178580e54b84af375e74f2bcb0e30b1d6
Signed-off-by: Maciej Wereski <m.wereski@partner.samsung.com>
Maciej Wereski [Tue, 25 Feb 2014 11:48:50 +0000 (12:48 +0100)]
Update to 0.
20140221.145534cet
Change-Id: I4be4430dcdbc7d2bba288906d377a912f14c11a5
Signed-off-by: Maciej Wereski <m.wereski@partner.samsung.com>
Karol Lewandowski [Mon, 27 Jan 2014 13:22:22 +0000 (14:22 +0100)]
packaging: Use date-based version string
kdbus doesn't have any release version we could use, fall back
to 0.<DATE>.<TIME> format for packaging purposes.
Karol Lewandowski [Fri, 24 Jan 2014 16:10:52 +0000 (17:10 +0100)]
packaging: Bump release
Karol Lewandowski [Fri, 24 Jan 2014 16:09:48 +0000 (17:09 +0100)]
packaging: Rename .spec to match real package name
Karol Lewandowski [Fri, 24 Jan 2014 16:09:22 +0000 (17:09 +0100)]
packaging: Fix various rpmlint errors
Karol Lewandowski [Fri, 24 Jan 2014 15:45:52 +0000 (16:45 +0100)]
packaging: Bump version
Karol Lewandowski [Mon, 28 Oct 2013 15:09:04 +0000 (16:09 +0100)]
Bring task_cgroup_path_from_hierarchy() back for pre-3.11 compatiblity
This commit partially reverts
ea2b8c798cd34149fa13010d24 ("message.c:
task_cgroup_path_from_hierarchy is now in 3.11-rc1, so remove it from
here").
We still require it for pre-3.11 kernels that Tizen is shipping.
Change-Id: I05568c5d8d473bb5257d3aa64f8a84b03e6da117
Signed-off-by: Karol Lewandowski <k.lewandowsk@samsung.com>
Karol Lewandowski [Wed, 22 Jan 2014 17:53:03 +0000 (18:53 +0100)]
Rebase to newest upstream and bump version
Rebased on top of
761aff75b ("connection: kill unneeded include").
Karol Lewandowski [Wed, 22 Jan 2014 16:49:11 +0000 (17:49 +0100)]
packaging: Provide -devel package with kdbus.h
Karol Lewandowski [Fri, 3 Jan 2014 17:28:24 +0000 (18:28 +0100)]
packaging: Fix spec to make kdbus work on ivi
Karol Lewandowski [Fri, 3 Jan 2014 14:25:40 +0000 (15:25 +0100)]
Add packaging to kdbus-bus package to be build from obs/gbs.
Change-Id: I3b630841956115d20e3d12a2d7dad45f21a2d75e
Signed-off-by: Jacek Pielaszkiewicz <j.pielaszkie@samsung.com>
[ Rebased to newest upstream ]
Signed-off-by: Karol Lewandowski <k.lewandowsk@samsung.com>
Daniel Mack [Thu, 11 Sep 2014 16:02:07 +0000 (18:02 +0200)]
kdbus.txt: document wildcard policy entries
Daniel Mack [Thu, 11 Sep 2014 15:34:52 +0000 (17:34 +0200)]
match: make kdbus_match_db_remove() void
Daniel Mack [Thu, 11 Sep 2014 15:34:52 +0000 (17:34 +0200)]
connection: allow wildcard matches for policy holders
Daniel Mack [Thu, 11 Sep 2014 15:32:42 +0000 (17:32 +0200)]
connection: reformat
Daniel Mack [Thu, 11 Sep 2014 15:32:11 +0000 (17:32 +0200)]
endpoint: cosmetics
Daniel Mack [Thu, 11 Sep 2014 15:10:29 +0000 (17:10 +0200)]
connection: check return value of kdbus_meta_append()
Daniel Mack [Thu, 11 Sep 2014 14:38:35 +0000 (16:38 +0200)]
connection: return -EOPNOTSUPP if the connection has already died
Daniel Mack [Wed, 10 Sep 2014 12:45:21 +0000 (14:45 +0200)]
kdbus.txt: add more documentation
Daniel Mack [Wed, 10 Sep 2014 16:03:52 +0000 (18:03 +0200)]
match: don't fail on invalid item types
Daniel Mack [Wed, 10 Sep 2014 12:26:19 +0000 (14:26 +0200)]
notify: explictly set payload_type = KDBUS_PAYLOAD_KERNEL
Daniel Mack [Tue, 9 Sep 2014 12:38:02 +0000 (14:38 +0200)]
kdbus.txt: kill reference to sd-memfd.h
Djalal Harouni [Mon, 8 Sep 2014 13:18:26 +0000 (14:18 +0100)]
policy: make policy checks work across user namespaces
Use the global kuid_t and kgid_t to store uid/gid values, this way we
our policy checks will work across user namespaces.
Note that currently we ignore that the user is privileged in its own
namespaces and the policy access kuid_t and kgid_t were mapped into that
namespace. If this is requested we can add it later a la:
fs/inode.c:inode_owner_or_capable()
Add kdbus_policy_make_access() to convert the user provided info to the
current user namespace. Userspace struct is not changed, only the kernel
one.
This patch fixes test #8 of test-kdbus-policy
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
[daniel: group kdbus_policy_db_entry_access->{uid,gid} in a union]
Daniel Mack [Mon, 8 Sep 2014 15:14:46 +0000 (17:14 +0200)]
connection: factor out queue management code to queue.c
Move code that deals with the message queue and its items into queue.c.
This way, we can work towards a cleaner structure and cut down connection.c
which has grown too big.
The sepearation is not really strict at this point, and might see some
more cleanups in the future.
Daniel Mack [Mon, 8 Sep 2014 13:54:38 +0000 (15:54 +0200)]
connection: rename 'entry' variable in kdbus_conn_kmsg_send()
Djalal Harouni [Mon, 8 Sep 2014 13:18:25 +0000 (14:18 +0100)]
test: update policy tests to handle user namespaces
Upstream kernels allow unprivileged users to create user namespaces
and change their uid/gid.
These patches update kdbus policy logic to handle this case and
improve our current checks across user namespaces.
So this patch adds:
* kdbus_test_waitpid() to get exit code of childs.
* kdbus_clone_userns_test() that performs the test inside a new
user namespace.
* Converts all the other tests to return CHECK_OK, CHECK_SKIP or
CHECK_ERR so we are consistent.
Currently we fail at kdbus_clone_userns_test() test #8. The next patch
will fix this issue.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Daniel Mack [Fri, 5 Sep 2014 16:44:07 +0000 (18:44 +0200)]
policy: indent
Daniel Mack [Fri, 5 Sep 2014 16:39:57 +0000 (18:39 +0200)]
connection: check policy for reply-awaiting messages as well
The current logic in bogus. We need to call kdbus_conn_check_access()
for messages that require a message as well. Also, teach
kdbus_conn_check_access() a mode in which not to look up expected
replies of a connection.
Daniel Mack [Sun, 24 Aug 2014 16:17:46 +0000 (18:17 +0200)]
connection: drop unused argument of kdbus_conn_payload_add()
Daniel Mack [Wed, 20 Aug 2014 20:53:48 +0000 (22:53 +0200)]
metadata: kdbus_meta_append_auxgroups(): actually set the group ids
As we need metadata in other context than queue items, we don't want
to rely on the queue item logic to clean up after us and fix up the
values. For now, add back the code to actually fill the payload. We
might add code to make the queue patching conditional later.
Djalal Harouni [Wed, 20 Aug 2014 16:41:03 +0000 (17:41 +0100)]
test: use audit.loginuid and audit.sessionid when dumping the audit item
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 20 Aug 2014 15:58:57 +0000 (16:58 +0100)]
kdbus: do the audit loginuid translation as late as possible
Do the namespace translation just before pushing into the receiver's
slice, so we map to the receiver's user namespace.
Patch tested.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Daniel Mack [Tue, 19 Aug 2014 17:04:42 +0000 (19:04 +0200)]
metadata: add back code to fill creds data
It was overlooked that metadata is also used for connection info calls.
Until we find a nicer way, just do the creds assigment in this item twice:
once at metadata creation time, and again when we deliver the message.
Daniel Mack [Tue, 19 Aug 2014 14:22:53 +0000 (16:22 +0200)]
Revert "test/Makefile: only look at $(KERNELDIR)/usr/include for files with quotes"
This reverts commit
baeb2d8bbf2360524d4104ea851407a14085cef9.
Turns out this confuses cpp even more.
For now, people have to to 'make headers_install' in the kernel dir before
building kdbus.
Daniel Mack [Tue, 19 Aug 2014 13:19:40 +0000 (15:19 +0200)]
test/Makefile: only look at $(KERNELDIR)/usr/include for files with quotes
This is a hack to make cpp look at system include files first, and then
consult $(KERNELDIR)/usr/include. For this to work, include directives
referring to files in uapi have to be addressed with quotes.
Daniel Mack [Tue, 19 Aug 2014 12:50:06 +0000 (14:50 +0200)]
Makefile: deal with realtive paths in test/
Invoke realpath to convert a possible relative path into an absolute one.
Daniel Mack [Tue, 19 Aug 2014 11:42:53 +0000 (13:42 +0200)]
Makefile: export KERNELDIR for tests
We include linux/memfd.h now, make sure it's searched for in $KERNELDIR.
Djalal Harouni [Mon, 18 Aug 2014 16:59:20 +0000 (17:59 +0100)]
test: add the metadata tests inside new namespaces
Currently this only handles user namespace. Use it to inspect the
dumped metadata/credentials.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Mon, 18 Aug 2014 16:13:10 +0000 (17:13 +0100)]
test: add conn_recv() that use poll() and msg_recv()
Will be used by tests.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Mon, 18 Aug 2014 10:59:59 +0000 (11:59 +0100)]
test: move shared definitions to kdbus-util.h to use them in other tests
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Mon, 18 Aug 2014 14:38:03 +0000 (15:38 +0100)]
test: add helpers for user namespace mapping
Will be used by other tests.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Tue, 19 Aug 2014 01:43:57 +0000 (02:43 +0100)]
connection: move the install creds into the slice to its own function
Move the install creds into the receiver's slice to its own function
kdbus_conn_creds_install().
Use from_kuid_munged(), so the uid mapping never fails.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Tue, 19 Aug 2014 01:43:56 +0000 (02:43 +0100)]
metadata: document creds_item_off and auxgrps_item_off fields
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Daniel Mack [Mon, 18 Aug 2014 15:07:49 +0000 (17:07 +0200)]
handle: return POLLOUT | POLLWRNORM from kdbus_handle_poll()
Also return POLLOUT | POLLWRNORM from kdbus_handle_poll() when a message
arrived. IOW: in all non-error conditions.
Daniel Mack [Tue, 12 Aug 2014 10:35:50 +0000 (12:35 +0200)]
metadata: drop support for kernels < v3.15
Daniel Mack [Mon, 11 Aug 2014 09:31:25 +0000 (11:31 +0200)]
Kconfig: depend on TMPFS
Daniel Mack [Mon, 12 May 2014 22:28:19 +0000 (00:28 +0200)]
test: adopt new memfd interface
Daniel Mack [Mon, 12 May 2014 12:39:59 +0000 (14:39 +0200)]
switch to upstream kernel memfd implementation (ABI+API break)
Now that a generic memfd solution is merged upstream, drop our own
implementation and switch over.
Daniel Mack [Sat, 9 Aug 2014 08:19:35 +0000 (10:19 +0200)]
metadata: follow 3.17 changes for current->start_time
As of Linux kernel commit
57e0be041d9e21a73 ("sched: Make
task->real_start_time nanoseconds based"), current->start_time is
stored in nanoseconds, so we don't need to do convert anymore.
Daniel Mack [Sun, 17 Aug 2014 19:43:25 +0000 (21:43 +0200)]
metadata: set item->size in kdbus_meta_append_item()
This saves us some logic on the call sites.
Daniel Mack [Sun, 17 Aug 2014 19:36:22 +0000 (21:36 +0200)]
metadata: remove kdbus_meta_find_item()
Remove kdbus_meta_find_item(), and store offsets of items inside the
metadata in the metadata itself. This is nicer than walking the list again
in order to find a specific element.
Daniel Mack [Sat, 16 Aug 2014 09:58:28 +0000 (11:58 +0200)]
connection: drop unused paramenter to kdbus_conn_*_install
The connection parameter is not used, so drop it from the function's
parameters.
Daniel Mack [Sat, 16 Aug 2014 08:54:08 +0000 (10:54 +0200)]
connection: address possible auxgrp count mismatch
The number of auxgrps could theoretically change between the preparation of
the the meta data item and the scan in kdbus_conn_queue_alloc().
Solve this by clamping the number of group elements we look at.
Daniel Mack [Sat, 16 Aug 2014 08:53:28 +0000 (10:53 +0200)]
util: introduce KDBUS_ITEM_PAYLOAD_SIZE
Add a simple macro to determine an item's payload size, and use it in
message.c.
Daniel Mack [Sat, 16 Aug 2014 08:41:09 +0000 (10:41 +0200)]
metadata: rename kdbus_meta_offset_of() to kdbus_meta_find_item()
Return a pointer to the item itself, so we can as well use its size.
Marcel Holtmann [Fri, 15 Aug 2014 19:43:37 +0000 (21:43 +0200)]
handle: Return POLLOUT | POLLWRNORM mask when no messages are pending
To facility the feature of doing an asynchronous sending of messages
when the bus is idle, make sure to return POLLOUT | POLLWRNORM from
kdbus_handle_poll.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Daniel Mack [Fri, 15 Aug 2014 20:08:08 +0000 (22:08 +0200)]
connection: fix error rewind exit path in kdbus_conn_msg_install()
Daniel Mack [Fri, 15 Aug 2014 19:45:10 +0000 (21:45 +0200)]
connection, metadata: convert auxgrp information as late as possible
Same as with the cred block - do the from_kgid() conversion shortly before
installing the message in the receiver's task.
Daniel Mack [Fri, 15 Aug 2014 17:46:10 +0000 (19:46 +0200)]
metadata: allow NULL pointers in kdbus_meta_append_data()
Allow NULL pointers to be passed as data argument to
kdbus_meta_append_data(), so users can use the function to only make space
in the metadata without actually copying any payload.
Daniel Mack [Fri, 15 Aug 2014 17:22:43 +0000 (19:22 +0200)]
connection, metadata: do namespace translation as late as possible
We have to do the namespace translation just before a message is
installed into the receiver's task, so we actually know about the
currently used pid and user namespaces.
To achive this, store the byte offset pointer in the message in
kdbus_conn_queue_alloc(), and patch the values over from
kdbus_conn_msg_install().
Thanks to Djalal Harouni for pointing this out.
Daniel Mack [Fri, 15 Aug 2014 17:20:51 +0000 (19:20 +0200)]
metadata: add kdbus_meta_offset_of()
kdbus_meta_offset_of() will iterate over the given metadata and return
the offset of the item of type @type, if it exists.
Daniel Mack [Thu, 14 Aug 2014 20:20:54 +0000 (22:20 +0200)]
policy, connection: rename policy cache related functions
Just so they reflect a little better what they're doing.
Daniel Mack [Thu, 14 Aug 2014 20:14:49 +0000 (22:14 +0200)]
policy: rename send_access_hash to talk_access_hash
We've already renamed the SEND policy to TALK for more clarity,
so let's do the same with the access hash.
Daniel Mack [Thu, 14 Aug 2014 14:36:40 +0000 (16:36 +0200)]
names: flush policy TALK cache when a connection loses a name
When a connection loses a name, make sure cache entries that allow
TALK access to it are properly removed. Upon the next TALK query,
the connection's names will be iterated and checked again.
Daniel Mack [Thu, 14 Aug 2014 14:14:33 +0000 (16:14 +0200)]
test: test-kdbus: fix monitor and byebye tests
Since commit
7015a1e6 ("prevent special-purpose connections from
owning names or exchanging messages"), ioctl() calls return
-EOPNOTSUPP they are issued to a handle of a inapproprate type.
Follow this change in the test as well.
Alban Crequy [Thu, 14 Aug 2014 11:13:46 +0000 (12:13 +0100)]
bus.h: add missing include to linux/kref.h
Symptoms:
> kdbus/bus.h:56:14: error: field ‘kref’ has incomplete type
> struct kref kref;
Signed-off-by: Alban Crequy <alban.crequy@collabora.co.uk>
Daniel Mack [Tue, 12 Aug 2014 13:02:31 +0000 (15:02 +0200)]
pool: remove unneeded semicolon
Daniel Mack [Sat, 9 Aug 2014 13:00:20 +0000 (15:00 +0200)]
metadata: return -EFAULT if get_task_mm(current) fails
Daniel Mack [Sat, 9 Aug 2014 12:13:57 +0000 (14:13 +0200)]
test: add test code for KDBUS_ITEM_AUXGROUPS
Daniel Mack [Sat, 9 Aug 2014 12:12:51 +0000 (14:12 +0200)]
metadata: add new item type KDBUS_ITEM_AUXGROUPS (ABI break)
Add new data type to dump the auxiliary groups the sending task is a member
of.
Djalal Harouni [Tue, 5 Aug 2014 01:46:08 +0000 (02:46 +0100)]
test: split conn_update() into update attach-flags and update policy
Since ordinary connections are only interested in the attach-flags and
policy holders in policies, split conn_update() into:
1) conn_update_attach_flags()
2) conn_update_policy()
This way we use the conn_update_policy() function in test-kdbus-policy
with a policy-holding connection and we pass all the tests. This
prevents messing up with the attach-flags.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Tue, 5 Aug 2014 01:46:07 +0000 (02:46 +0100)]
connection: improve kdbus_cmd_conn_update() connection type checks
Do another round of connection type checks inside the KDBUS_ITEM
iterator.
We need this since we do not want to allow ordinary connections to
update policy entries that belong to another policy holder connection.
We also do it for the attach flags since only ordinary connections are
interessted in it.
And update a kdbus_policy_set() call to only pass a one name per
policy-holding connection
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Tue, 5 Aug 2014 01:46:06 +0000 (02:46 +0100)]
handle: allow KDBUS_CMD_CONN_UPDATE ioctl for policy holders
Allow KDBUS_CMD_CONN_UPDATE for KDBUS_CONN_POLICY_HOLDER connections.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Tue, 5 Aug 2014 01:09:09 +0000 (02:09 +0100)]
handle: return -EOPNOTSUPP instead of -EPERM if an operation is not supported
If userspace calls in with the wrong connection type, just return
-EOPNOTSUPP instead of -EPERM.
This will not confuse unprivileged and privileged processes, and permits
to identify legitimate -EPERM errors.
This just converts errors introduced in commit
7015a1e6746
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Tue, 5 Aug 2014 01:09:08 +0000 (02:09 +0100)]
test: sync the policy tests with the recent activators and policy holders changes
Recent commit
7015a1e6746e0c2 prevents special-purpose connections from
owning names, so update the test-kdbus-policy tests to follow and test
these changes.
Create a new policy holder connection which will register the policy for
an X name, and make the first conn_db[0] connection own that name.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Kay Sievers [Mon, 4 Aug 2014 14:18:08 +0000 (16:18 +0200)]
use conn->type instead of conn->flags where possible
Kay Sievers [Mon, 4 Aug 2014 14:05:39 +0000 (16:05 +0200)]
prevent special-purpose connections from owning names or exchanging messages
Kay Sievers [Thu, 31 Jul 2014 21:02:54 +0000 (23:02 +0200)]
kdbus.h: clarify policy holder docs
Kay Sievers [Thu, 31 Jul 2014 15:15:13 +0000 (17:15 +0200)]
do not use __ prefix for ordinary functions
Kay Sievers [Thu, 31 Jul 2014 14:59:27 +0000 (16:59 +0200)]
domain: add BUG_ON()
Kay Sievers [Thu, 31 Jul 2014 14:52:50 +0000 (16:52 +0200)]
do not use __ prefix for ordinary functions
Kay Sievers [Thu, 31 Jul 2014 14:51:08 +0000 (16:51 +0200)]
test: do not use __ prefix for ordinary functions
Djalal Harouni [Wed, 30 Jul 2014 20:11:58 +0000 (21:11 +0100)]
domain: remove dead kdbus_domain_user_find_or_new()
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 30 Jul 2014 20:11:57 +0000 (21:11 +0100)]
connection: fix user quota accounting corruption
First use kzalloc to allocate the users array, so we do not reference
unintialized values.
And free the old conn->msg_users array not the newly allocated 'users'
one.
Patch tested, and users will hit the KDBUS_CONN_MAX_MSGS_PER_USER limit
and fail with -ENOBUFS
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 30 Jul 2014 20:11:56 +0000 (21:11 +0100)]
bus: improve user quota accounting and domain locking
Currently kdbus_bus_new() execution path might take the domain lock
three times on success, four times on failure.
kdbus_bus_new():
=> kdbus_domain_user_find_or_new(): takes it 2 times (and it is racy)
+
kdbus_bus_new(): take it an extra time to account the user and link
the bus into the domain bus_list.
And as discussed in the previous patch kdbus_domain_user_find_or_new()
is racy, so convert to __kdbus_domain_user_account()
This allows also to improve the execution path to take the domain lock
only one time in case of success. On failure the domain lock will be
taken two times.
kdbus_bus_new():
=> take domain lock
=> check if domain is still active/connected
=> __kdbus_domain_user_account()
...
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 30 Jul 2014 20:11:55 +0000 (21:11 +0100)]
kdbus: improve user quota accounting by using kdbus_domain_user_account()
Currently kdbus_domain_user_find_or_new() is used to find a user
domain or create a new one and link it into the domain.
kdbus_domain_user_find_or_new() may fail due to memory allocation
errors or if the domain was shutdown, but since callers will
receive only a NULL pointer on failure, they assume -ENOMEM and
ignore -ESHUTDOWN. Fix this in kdbus_domain_user_account() by returning
the appropriate error code.
There are also some races with kdbus_domain_user_find_or_new(), if it is
called with the same parameters and if we do not find a previously
linked domain user, then both threads will race to assign a different ID
for the same uid, thus, invalidating the users array. We fix this in the
new kdbus_domain_user_account() and __kdbus_domain_user_account() by
taking the domain lock only one time.
Replace some kdbus_domain_user_find_or_new() calls with
kdbus_domain_user_account(). The last one in bus.c is updated in the
next patch.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 30 Jul 2014 20:11:54 +0000 (21:11 +0100)]
domain: add kdbus_domain_user_account()
Add kdbus_domain_user_account() to account and link users into a
domain.
This function will take the domain lock, and it will be used as a
replacement for kdbus_domain_user_find_or_new().
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 30 Jul 2014 20:11:53 +0000 (21:11 +0100)]
domain: add __kdbus_domain_user_account() to account domain users
Add __kdbus_domain_user_account() to account and link users into a
domain.
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 30 Jul 2014 20:11:52 +0000 (21:11 +0100)]
domain: add kdbus_domain_user_assign_id() to assign IDs to domain users
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Djalal Harouni [Wed, 23 Jul 2014 16:34:32 +0000 (17:34 +0100)]
test: correctly set the 'ret' variable
Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Greg Kroah-Hartman [Wed, 2 Jul 2014 05:46:39 +0000 (22:46 -0700)]
Merge pull request #21 from torstehu/fix-warning
Fix annoying compile warning due to missing cast
Torstein Husebø [Sun, 29 Jun 2014 15:41:31 +0000 (17:41 +0200)]
Fix annoying compile warning due to missing cast
Daniel Mack [Tue, 24 Jun 2014 13:41:47 +0000 (15:41 +0200)]
connection: don't append to uninitialized metadata
kmsg->meta is only valid if conn_src != NULL.
Daniel Mack [Tue, 24 Jun 2014 09:24:49 +0000 (11:24 +0200)]
connection: clean up conditional code path
For broadcasts, we return anway, so there's no point in an 'else' branch.
Daniel Mack [Tue, 24 Jun 2014 09:15:48 +0000 (11:15 +0200)]
connection: rename kdbus_check_send_perm() to kdbus_conn_check_access()
Keep the function namespace clean.
Daniel Mack [Mon, 23 Jun 2014 17:08:27 +0000 (19:08 +0200)]
connection: factor out code into kdbus_conn_add_expected_reply()
This helps cutting down kdbus_conn_kmsg_send() to make it more readable.