This is a portability fix and reduces stack usage.

The DMA mapping API cannot map on-stack addresses, as explained in
Documentation/DMA-mapping.txt.  Convert the two cases of on-stack packet
payload buffers in firedtv (payload of write requests in avc_write and
of lock requests in cmp_lock) to slab-allocated memory.

We use the 512 bytes sized FCP frame buffer in struct firedtv for this
purpose.  Previously it held only incoming FCP responses, now it holds
pending FCP requests and is then overwriten by an FCP response from the
tuner subunit.  Ditto for CMP lock requests and responses.  Accesses to
the payload buffer are serialized by fdtv->avc_mutex.

As a welcome side effect, stack usage of the AV/C transaction functions
is reduced by 512 bytes.

Alas, avc_register_remote_control() is a special case:  It previously
did not wait for a response.  To fit better in with the other FCP
transactions, let it wait for an interim response.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Control of more than one AV/C device at once --- e.g. camcorders, tape
decks, audio devices, TV tuners --- failed or worked only unreliably,
depending on driver implementation.  This affected kernelspace and
userspace drivers alike and was caused by firewire-core's inability to
accept multiple registrations of FCP listeners.

The fix allows multiple address handlers to be registered for the FCP
command and response registers.  When a request for these registers is
received, all handlers are invoked, and the Firewire response is
generated by the core and not by any handler.

The cdev API does not change, i.e., userspace is still expected to send
a response for FCP requests; this response is silently ignored.
Signed-off-by: NClemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (changelog, rebased, whitespace)

All read transactions initiated by firedtv are only quadlet-sized, hence
the backend->read call can be simplified a little.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Packet DMA buffers are queued either initially all at once (then, a
queueing failure will cause firedtv to release the DMA context as a
whole) or subsequently one by one as they recycled after use (then a
failure is extremely unlikely).  Therefore we can be a little less
cautious when counting at which packet buffer to set the interrupt flag.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

Cache only addresses of whole pages, not of each buffer chunk.  Besides,
page addresses can be obtained by page_address() instead of kmap() since
they were allocated in lowmem.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>

The firedtv DVB driver will now work not only on top of the old ieee1394
driver stack but also on the new firewire driver stack.

Alongside to the firedtv-1394.c backend for driver binding and I/O, the
firedtv-fw.c backend is added.  Depending on which of the two 1394
stacks is configured, one or the other or both backends will be built
into the firedtv driver.

This has been tested with a DVB-T and a DVB-C box on x86-64 and x86-32
together with a few different controllers (Agere FW323, a NEC chip, TI
TSB82AA2, TSB43AB22/A, VIA VT6306).
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NMauro Carvalho Chehab <mchehab@redhat.com>