ALSA IEC 61883-1/6 packet streaming engine uses pre-computed parameters
ideal for nominal sampling transfer frequency (STF) to transfer packets
to device since it was added 2011. As a result of user experience for a
decade, it is clear that the sequence is not suitable to some actual
devices. It takes the devices to generate noise, and causes any type of
discontinuity in the series of packet transferred from the device. It's
required for the engine to transfer packets according to effective STF.
The effective STF is given by media clock recovered by the sequence of
packet transferred from the target device. In the previous commit, the
sequence is already cached. The media clock recovery can be achieved by
analyzing the sequence.
In technological world, many ideas are proposed for media clock recovery.
However, the small part of them could be actually adopted in our case
since floating point arithmetic is not mostly available in Linux kernel
land.
This commit adopts the simple way from them; sequence replay, which means
that the sequence of parameters from incoming packet is used as is to
transfer outgoing packets. The media clock is not computed internally,
but the sequence of outgoing packet superficially looks to be generated by
the media clock.
The association between source and destination is decided when starting
AMDTP domain. When the target device supports a pair of isochronous packet
streams, the tx stream is source and the rx stream is destination. When it
supports two pair of streams, each of tx stream is associated to
corresponding rx stream in its order. When it supports less number of tx
streams than rx streams, the fist tx stream is selected for all of rx
streams. When it supports more tx streams than rx streams, the first tx
packet is associated to the rx stream.
As I noted in previous commit, the sequence of parameters from incoming
packet is different between devices, time to time. It is worse idea to
replay the sequence of parameters from a device for the sequence of
packet to the other devices even if they are in the same category of
device.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210527122611.173711-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In design of audio and music unit in IEEE 1394 bus, feedback of
effective sampling transfer frequency (STF) is delivered by packets
transferred from device. The devices supported by ALSA firewire stack
are categorized to three groups regarding to it.
* Group 1:
* Echo Audio Fireworks board module
* Oxford Semiconductor OXFW971 ASIC
* Digidesign Digi00x family
* Tascam FireWire series
* RME Fireface series
* Group 2:
* BridgeCo. DM1000/DM1100/DM1500 ASICs for BeBoB solution
* TC Applied Technologies DICE ASICs
* Group 3:
* Mark of the Unicord FireWire series
In group 1, the effective STF is determined by the sequence of the number
of events per packet. In group 2, the sequence of presentation timestamp
expressed in syt field of CIP header is interpreted as well. In group 3,
the presentation timestamp is expressed in source packet header (SPH) of
each data block.
I note that some models doesn't take care of effective STF with large
internal buffer. It's reasonable to name it as group 0:
* Group 0
* Oxford Semiconductor OXFW970 ASIC
The effective STF is known to be slightly different from nominal STF for
all of devices, and to be different between the devices. Furthermore, the
effective STF is known to be shifted for long-period transmission. This
makes it hard for software to satisfy the effective STF when processing
packets to the device.
The effective STF is deterministic as a result of analyzing the batch of
packet transferred from the device. For the analysis, caching the sequence
of parameter in the packet is required.
This commit adds an option so that AMDTP domain structure takes AMDTP
stream structure to cache the sequence of parameters in packet transferred
from the device. The parameters are offset ticks of syt field against the
cycle to receive the packet and the number of data blocks per packet.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210527122611.173711-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The member of callbacked in AMDTP stream structure is not used anymore.
Instead, ready_processing member is used to wake up yielding task of user
process.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210524031346.50539-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The devices based on BeBoB ASICs or the devices in Tascam FireWire
series transfer a batch of NODATA packet or empty packet in the beginning
of packet streaming. To avoid processing them, current implementation uses
an option to skip processing content of tx packet during some initial
cycles. However, the hard-coded number is not enough useful.
This commit drops content of packets till the packet includes any event
firstly. The function of option is to skip processing content of tx packet
with any event after dropping.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210524031346.50539-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In later commit, transfer delay is used in both IR and IT contexts. This
commit refactors regardless of transfer delay.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210522013303.49596-7-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Current implementation pools the sequence in AMDTP domain. This is convenient
regarding to memory usage and computation time, however inconvenient for the
devices such that several rx streams are expected to transfer timing
information independently.
This commit refactors to pool the sequence per rx packet stream.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210522013303.49596-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Many devices are unaware of syt field in rx CIP for playback timing.
This commit adds a flag to cancel processing syt field. Actually,
syt calculation is required to decide the number of events per rx packet.
The flag put 0xffff to CIP header of rx packet. On the other hand,
The value of syt field in CIP header of tx packet is unavailable. The
sequence of packet descriptor for tx packet includes 0 for the offset
of syt field to avoid computation.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210522013303.49596-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The check of error is just done for the case that CIP header is available.
This commit moves auto variable into the branch to process CIP header.
Addresses-Coverity: ("Uninitialized scalar variable")
Fixes: c09010eeb3 ("ALSA: firewire-lib: handle the case that empty isochronous packet payload for CIP")
Suggested-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520130409.GA170303@workstation
Signed-off-by: Takashi Iwai <tiwai@suse.de>
When starting AMDTP domain, tasks in process context yields running CPU
till all of isochronous context get callback, with an assumption that
it's OK to process content of packet.
However several isochronous cycles are skipped to transfer rx packets, or
the content of rx packets are dropped, to manage the timing to start
processing the packets.
This commit changes the timing for tasks in process context to wake up
when processing content of packet is actually ready.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520040154.80450-9-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
DICE ASICs support several pairs of isochronous packet streaming and
expect software to queue packets with the same timing information into
the same isochronous cycle.
This commit adds structure member to manage the cycle to start
processing packet in several IT contexts. The cycle is decided when
batch of isochronous cycle is skipped in callback to isochronous context
for IRQ target.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520040154.80450-8-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
It's several hundred cycles from starting isochronous contexts and the
actual cycle to start processing content of packet. It's useless to
start the context for IRQ target apart from the other contexts.
This commit refactors helper function to start AMDTP domain in the point.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520040154.80450-7-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Current implementation of ALSA IEC 61883-1/6 packet streaming engine
allows drivers to decide isochronous cycle to start IR context. This
option is mainly used to avoid processing the sequence of packet with
some quirks; e.g. discontinuity of counter. However, it's inconvenient
to fail to continue packet processing when the target device doesn't
start transmission of packet till the decided cycle.
This commit changes the behaviour. As an alternative to the start cycle
for IR context, the cycle count to drop content of packet in the beginning
of IR context.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520040154.80450-6-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
DICE ASICs support several pairs of isochronous packet streaming. It's
convenient for drivers to process content of the packet in the same cycle
timing.
This commit adds structure member to manage the cycle to start
processing packet in several IR contexts. The cycle is decided in the
first callback of the IR contexts. The content of packet is dropped
till the cycle.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520040154.80450-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
It's convenient to calculate the size of context payload apart from
the size of isochronous packet payload.
This commit adds a helper function for it.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210520040154.80450-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This commit fulfils sequence descriptors for skipped cycle when
it's one cycle. This is preparation for future integration.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210518130048.146596-9-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Within devices supported by drivers in ALSA firewire stack, OXFW-based
devices and Fireface devices are known to skip isochronous cycle for
packet transmission. The former is due to the jumbo payload quirk. The
latter is due to vendor protocol in which empty packet is not
transferred in blocking mode.
Although nothing to do just for handling events of the packet, packet
continuity is necessarily for media clock recovery. This commit checks
whether any cycle is continue or not.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210518130048.146596-8-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Some macros and functions are renamed so that they compute isochronous
cycle within maximum count of second in isochronous context of 1394
OHCI.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210518130048.146596-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Two quadlets are at least included in isochronous packet payload for
Common Isochronous Packet (CIP) format in IEC 61883-1. However, it's
better to equip ALSA IEC 61883-1/6 packet streaming engine for contrary
packet.
This commit handles isochronous cycle to process such packet so that the
cycle is skipped.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210518130048.146596-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
It brings some inconvenience in practice to use enumerated type for
variable to which bitwise OR with enumerator constant is assigned.
This commit replaces declarations of enumerated type with int type.
Suggested-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210518084557.102681-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The snd_firewire_lib:amdtp_packet tracepoints event includes index of
packet processed in a context handling. However in IR context, it is not
calculated as expected.
Cc: <stable@vger.kernel.org>
Fixes: 753e717986 ("ALSA: firewire-lib: use packet descriptor for IR context")
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210513125652.110249-6-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The quadlets for CIP header is handled as a part of IR context header,
thus it doesn't join in IR context payload. However current calculation
includes the quadlets in IR context payload.
Cc: <stable@vger.kernel.org>
Fixes: f11453c7cc ("ALSA: firewire-lib: use 16 bytes IR context header to separate CIP header")
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210513125652.110249-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The check for size of isochronous packet payload just cares of the size of
IR context payload without the size of CIP header.
Cc: <stable@vger.kernel.org>
Fixes: f11453c7cc ("ALSA: firewire-lib: use 16 bytes IR context header to separate CIP header")
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210513125652.110249-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The tasklet is an old API that should be deprecated, usually can be
converted to another decent API. In FireWire driver, a tasklet is
still used for offloading the AMDTP PCM stream handling. It can be
achieved gracefully with a work queued, too.
This patch replaces the tasklet usage in firewire-lib driver with a
simple work. The conversion is fairly straightforward but for the
in_interrupt() checks that are replaced with the check using the
current_work().
Note that in_interrupt() in amdtp_packet tracepoint is still kept as
is. This is the place that is probed by both softirq of 1394 OHCI and
a user task of a PCM application, and the work handling is already
filtered in amdtp_domain_stream_pcm_pointer().
Tested-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200909163659.21708-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.
Signed-off-by: Romain Perier <romain.perier@gmail.com>
Signed-off-by: Allen Pais <allen.lkml@gmail.com>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200902040221.354941-3-allen.lkml@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In previous commit, the sequence of syt offset and the number of data
blocks per packet is calculated for pool in AMDTP domain structure in
advance of processing outgoing packets.
This commit uses the sequence for outgoing packet processing to obsolete
per-stream processing of the sequence.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-11-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In current implementation, sequence of syt offset and the number of data
blocks is generated when packets for outgoing stream are going to be
queued.
This commit generates and pools the sequence independently of the
processing of outgoing packets for future extension.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-10-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
For future extension, storage is required to store packet sequence in
incoming AMDTP stream to recover media clock for outgoing AMDTP stream.
This commit adds the storage to AMDTP domain for this purpose. The
packet sequence is represented by 'struct seq_desc' which has two
members; syt_offset and the number of data blocks. The size of storage
is decided according to the size of packet queue.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-9-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
When calculating the number of data blocks per packet, some states are
stored in AMDTP stream structure. This is inconvenient when reuse the
calculation from non-stream structure.
This commit applies refactoring to helper function for the calculation
so that the function doesn't touch AMDTP stream structure.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-8-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
When calculating syt offset, some states are stored in AMDTP stream
structure. This is inconvenient when reuse the calculation from
non-stream structure.
This commit applies refactoring to helper function for the calculation
so that the function doesn't touch AMDTP stream structure.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-7-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In current implementation for outgoing AMDTP packet, the value of syt
field in CIP header is computed when calculating syt offset. For
future extension, it's convenient to split the computation and
calculation.
This commit splits them.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-6-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Although the parameter for packet queue and IRQ timing is calculated when
AMDTP stream starts, the calculated parameters are the same between
streams in AMDTP domain.
This commit moves the calculation and decide the parameters when AMDTP
domain starts.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In current implementation, AMDTP domain structure and AMDTP stream
structure has one way of reference from the former to the latter. For
future extension, bidirectional reference is needed.
This commit adds a member into stream structure to refer to domain
structure to which the stream belongs.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In descriptor of isochronous context in 1394 OHCI, the field of second
has 3 bit, thus the maximum value is 8. The value is used for correct
cycle calculation.
This commit replaces hard-coded value with macro to obsolete magic
number.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200508043635.349339-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Some devices have a quirk to postpone transmission of isoc packet for
several dozen or hundred isoc cycles since configured to transmit.
Furthermore, some devices have a quirk to transmit isoc packet with
discontinued data of its header.
In 1394 OHCI specification, software allows to start isoc context with
certain isoc cycle. Linux firewire subsystem has kernel API to use it
as well.
This commit uses the functionality of 1394 OHCI controller to handle
the quirks. At present, this feature is convenient to ALSA bebob and
fireface driver. As a result, some devices can be safely handled, as
long as I know:
- MAudio FireWire solo
- MAudio ProFire Lightbridge
- MAudio FireWire 410
- Roland FA-66
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191018061911.24909-7-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This commit changes AMDTP domain to run on an IT context of 1394 OHCI as
IRQ target. No hardware interrupt is scheduled for the other isoc
contexts. All of the isoc context are processed in a callback for an isoc
context of IRQ target.
The IRQ target is automatically selected from a list of AMDTP streams,
thus users of AMDTP domain should add an AMDTP stream for IT context
at least.
The reason to select IT context as IRQ target is that the IT context
runs on local 1394 OHCI controller and it can be used as reliable,
constant IRQ generator. On the other hand, IR context can include skip
cycle according to isoc packet transferred by device.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191018061911.24909-6-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The aim of AMDTP domain is to process several isoc context in the same
time. However, current implementation is against this idea because it
flushes each isoc context in the end of processing context callback.
This commit cancels it.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191018061911.24909-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
An isoc context for AMDTP stream is flushed to queue packet
by a call of pcm.ack. This commit extends this for AMDTP
domain.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191018061911.24909-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
An isoc context for AMDTP stream is flushed to queue packet
by a call of pcm.pointer. This commit extends this for AMDTP
domain.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191018061911.24909-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In IEC 61883-1/6, one isoc packet can transfer events up to the value of
syt interval. This comes from the interval of isoc cycle. As 1394 OHCI
controller can generate hardware IRQ per isoc packet, the interval is
calculated as 125 usec.
In IEC 61883-1/6, two ways of transmission is described; blocking and
non-blocking methods. In blocking method, the sequence of packet includes
'empty' or 'NODATA' packets which include no events. In non-blocking
method, the number of events per packet is variable up to the syt
interval.
This commit uses double of the value of syt interval as minimum available
size of PCM period due to the above protocol design.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191017155424.885-13-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Linux driver for 1394 OHCI controller voluntarily flushes isoc context
when total size of accumulated context header reached PAGE_SIZE. This
kicks tasklet for the isoc context. This is inconvenient to process
runtime of PCM substream.
This commit adds a restriction of the maximum size of PCM period to
avoid this behaviour.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20191017155424.885-12-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
