As reported by valrgrind
==30002== Conditional jump or move depends on uninitialised value(s)
==30002== at 0x5CB883C: pa_cmsg_ancil_data_close_fds (pstream.c:193)
==30002== by 0x5CBB161: do_write (pstream.c:759)
==30002== by 0x5CB8B51: do_pstream_read_write (pstream.c:233)
==30002== by 0x5CB8EE8: io_callback (pstream.c:279)
...
The pa_cmsg_ancil_data structure has two main guards:
'creds_valid', which implies that it holds credentials
information, and 'nfd', which implies it holds file descriptors.
When code paths create a credentials ancillary data structure,
they just set the 'nfd' guard to zero. Typically, the rest of
pa_cmsg_ancil_data fields related to fds are _all_ left
_uninitialized_.
pa_cmsg_ancil_data_close_fds() has broken the above contract:
it accesses the new 'close_fds_on_cleanup' flag, which is related
to file descriptors, without checking the 'nfd == 0' guard first.
Fix this inconsistency.
Reported-by: Alexander E. Patrakov <patrakov@gmail.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
Signed-off-by: Arun Raghavan <arun@arunraghavan.net>
Now that we have the necessary infrastructure to memexport and
mempimport a memfd memblock, extend that support higher up in the
chain with pstreams.
A PA endpoint can now _transparently_ send a memfd memblock to the
other end by simply calling pa_pstream_send_memblock() – provided
the block's memfd pool was earlier registered with the pstream.
If the pipe does not support memfd transfers, we fall back to
sending the block's full data instead of just its reference.
** Further details:
A single pstream connection usually transfers blocks from multiple
pools including the server's srbchannel mempool, the client's
audio data mempool, and the server's global core mempool.
If these mempools are memfd-backed, we now require registering
them with the pstream before sending any blocks they cover. This
is done to minimize fd passing overhead and avoid fd leaks.
Moreover, to support all these pools without hard-coding their
number or nature in the Pulse communication protocol itself, a new
REGISTER_MEMFD_SHMID command is introduced. That command can be
sent _anytime_ during the pstream's lifetime and is used for
creating on demand SHM ID to memfd mappings.
Suggested-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
To transfer memfd-backed blocks without passing their fd every time,
thus minimizing overhead and avoiding fd leaks, a command is sent
with the memfd fd as ancil data very early on.
This command has an ID that uniquely identifies the memfd region.
Further memfd block references are then exclusively done using this
ID.
This commit implements the details of such 'permanent' mappings on
the receiving end, using memimport segments.
Suggested-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
In future commits, server-wide SHMs will be replaced with per-client
ones that will be dynamically created and freed according to clients
connections open and close.
Meanwhile, current PA design does not guarantee that the per-client
mempool blocks are referenced only by client-specific objects.
Thus reference count the pools and let each memblock inside the pool
itself, or just attached to it, increment the pool's refcount upon
allocation. This way, per-client mempools will only be freed when no
further component in the system holds any references to its blocks.
DiscussionLink: https://goo.gl/qesVMV
Suggested-by: Tanu Kaskinen <tanuk@iki.fi>
Suggested-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
We need to guard the pstream with an extra ref to ensure
it is not destroyed at the time we check whether or not the
srbchannel is destroyed.
Reported-by: Takashi Iwai <tiwai@suse.de>
BugLink: http://bugzilla.opensuse.org/show_bug.cgi?id=950487
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
In case SHM is full or disabled, audio data is sent through the
io/srbchannel. When this channel in turn gets full, memblocks
could previously be split up. This could lead to crashes in case
the split was on non-frame boundaries (in combination with full
memblock queues).
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=88452
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
FSF addresses used in PA sources are no longer valid and rpmlint
generates numerous warnings during packaging because of this.
This patch changes all FSF addresses to FSF web page according to
the GPL how-to: https://www.gnu.org/licenses/gpl-howto.en.html
Done automatically by sed-ing through sources.
For writing, we prefer writing through the srbchannel if one is available,
and we have no ancil data to send.
For reading, we support reading from both in parallel. This meant replicating
a struct used for reading, so a lot of this patch is just a search/replace in
do_read to use the appropriate channel for reading.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
The shared ringbuffer memblock must be writable by both sides.
This makes it possible to send such a memblock over a pstream without
the "both sides writable" information getting lost.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
This patch adds support to iochannel, pstream and pstream-util
to send file descriptors over a unix pipe.
Currently we don't support writing both creds and fds in the same
packet, it's either one or the other (or neither).
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
The file descriptors are read from the iochannel just like the creds are.
So instead of passing just creds (and creds_valid), we now pass the
entire pa_ancil struct.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Now that we don't *always* get a callback after having written
something, make sure we can continue writing as long as it fully
succeeds.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
During a stream, most packets sent are either memblocks (with SHM info),
or requests for more data. These are only slightly bigger than the
header.
This patch makes it possible to write these packages in one write
instead of two: a memcpy of just a few bytes is worth saving extra
syscalls for write and poll.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Instead of using PA_SCACHE_ENTRY_SIZE_MAX, the size for FRAME_SIZE_MAX_ALLOW is
set directly to the same value. This removes the need for the core-scache.h
include, which caused an unwanted dependency of libpulsecommon on libpulsecore.
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=41539
The check whether POSIX socket.h or WIN32 winsock2.h must be included can be
made centrally. The downside is that some functionality of e.g. arpa/inet.h is
also implemented in winsock.h, so that some files that don't use socket
functions, but do use inet.h functions, must also include pulsecore/socket.h.
(as well as arpa/inet.h)
When logging a suppression message do so on the same log level as the
suppressed messages.
Cherry picked by Colin Guthrie from ec5a785712
with a couple of additional changes due to extra limiting in master
that was not present in stable-queue.
pa_memblock is now an opaque structure. Access to its fields is now done
through various accessor functions in a thread-safe manner.
pa_memblock_acquire() and pa_memblock_release() are now used to access the
attached audio data. Why? To allow safe manipulation of the memory pointer
maintained by the memory block. Internally _acquire() and _release() maintain a
reference counter. Please do not confuse this reference counter whith the one
maintained by pa_memblock_ref()/_unref()!
As a side effect this patch removes all direct usages of AO_t and replaces it
with pa_atomic_xxx based code.
This stuff needs some serious testing love. Especially if threads are actively
used.
git-svn-id: file:///home/lennart/svn/public/pulseaudio/trunk@1404 fefdeb5f-60dc-0310-8127-8f9354f1896f
pa_logXXX(__FILE__":
and replace them by
pa_logXXX("
git-svn-id: file:///home/lennart/svn/public/pulseaudio/trunk@1272 fefdeb5f-60dc-0310-8127-8f9354f1896f
is to allocate all audio memory blocks from a per-process memory pool which is
available as read-only SHM segment to other local processes. Then, instead of
writing the actual audio data to the socket just write references to this
shared memory pool.
To work optimally all memory blocks should now be of type PA_MEMBLOCK_POOL or
PA_MEMBLOCK_POOL_EXTERNAL. The function pa_memblock_new() now generates memory
blocks of this type by default.
git-svn-id: file:///home/lennart/svn/public/pulseaudio/trunk@1266 fefdeb5f-60dc-0310-8127-8f9354f1896f
