We do not need slab allocations for ORB pointer write transactions
anymore in order to satisfy streaming DMA mapping constraints, thanks to
commit da28947e7e "firewire: ohci: avoid separate DMA mapping for
small AT payloads".
(Besides, the slab-allocated buffers that firewire-sbp2 used to provide
for 8-byte write requests were still not fully portable since they
shared a cacheline with unrelated CPU-accessed data.)
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
firewire-sbp2 already takes care for internal serialization where
required (ORB list accesses), and it does not use cmd->serial_number
internally. Hence it is safe to not grab the shost lock around
queuecommand.
While we are at housekeeping, drop a redundant struct member:
sbp2_command_orb.done is set once in a hot path and dereferenced once in
a hot path. We can as well dereference sbp2_command_orb.cmd->scsi_done
instead.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
firewire-core manages the following types of work items:
fw_card.br_work:
- resets the bus on a card and possibly sends a PHY packet before that
- does not sleep for long or not at all
- is scheduled via fw_schedule_bus_reset() by
- firewire-ohci's pci_probe method
- firewire-ohci's set_config_rom method, called by kernelspace
protocol drivers and userspace drivers which add/remove
Configuration ROM descriptors
- userspace drivers which use the bus reset ioctl
- itself if the last reset happened less than 2 seconds ago
fw_card.bm_work:
- performs bus management duties
- usually does not (but may in corner cases) sleep for long
- is scheduled via fw_schedule_bm_work() by
- firewire-ohci's self-ID-complete IRQ handler tasklet
- firewire-core's fw_device.work instances whenever the root node
device was (successfully or unsuccessfully) discovered,
refreshed, or rediscovered
- itself in case of resource allocation failures or in order to
obey the 125ms bus manager arbitration interval
fw_device.work:
- performs node probe, update, shutdown, revival, removal; including
kernel driver probe, update, shutdown and bus reset notification to
userspace drivers
- usually sleeps moderately long, in corner cases very long
- is scheduled by
- firewire-ohci's self-ID-complete IRQ handler tasklet via the
core's fw_node_event
- firewire-ohci's pci_remove method via core's fw_destroy_nodes/
fw_node_event
- itself during retries, e.g. while a node is powering up
iso_resource.work:
- accesses registers at the Isochronous Resource Manager node
- usually does not (but may in corner cases) sleep for long
- is scheduled via schedule_iso_resource() by
- the owning userspace driver at addition and removal of the
resource
- firewire-core's fw_device.work instances after bus reset
- itself in case of resource allocation if necessary to obey the
1000ms reallocation period after bus reset
fw_card.br_work instances should not, and instances of the others must
not, be executed in parallel by multiple CPUs -- but were not protected
against that. Hence allocate a non-reentrant workqueue for them.
fw_device.work may be used in the memory reclaim path in case of SBP-2
device updates. Hence we need a workqueue with rescuer and cannot use
system_nrt_wq.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Reviewed-by: Tejun Heo <tj@kernel.org>
When queueing iso packets, the run time is dominated by the two
MMIO accesses that set the DMA context's wake bit. Because most
drivers submit packets in batches, we can save much time by
removing all but the last wakeup.
The internal kernel API is changed to require a call to
fw_iso_context_queue_flush() after a batch of queued packets.
The user space API does not change, so one call to
FW_CDEV_IOC_QUEUE_ISO must specify multiple packets to take
advantage of this optimization.
In my measurements, this patch reduces the time needed to queue
fifty skip packets from userspace to one sixth on a 2.5 GHz CPU,
or to one third at 800 MHz.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
We do not need slab allocations anymore in order to satisfy
streaming DMA mapping constraints, thanks to commit da28947e7e
"firewire: ohci: avoid separate DMA mapping for small AT payloads".
(Besides, the slab-allocated buffers that firewire-core, firewire-sbp2,
and firedtv used to provide for 8-byte write and lock requests were
still not fully portable since they crossed cacheline boundaries or
shared a cacheline with unrelated CPU-accessed data. snd-firewire-lib
got this aspect right by using an extra kmalloc/ kfree just for the
8-byte transaction buffer.)
This change replaces kmalloc'ed lock transaction scratch buffers in
firewire-core, firedtv, and snd-firewire-lib by local stack allocations.
Perhaps the most notable result of the change is simpler locking because
there is no need to serialize usages of preallocated per-device buffers
anymore. Also, allocations and deallocations are simpler.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Acked-by: Clemens Ladisch <clemens@ladisch.de>
Current implementation of ohci_set_config_rom() uses a deferred
bus reset via fw_schedule_bus_reset(). If clients add multiple
unit descriptors to the config_rom in quick succession, the
deferred bus reset may not have fired before succeeding update
requests have come in. This can lead to an incorrect partial
update of the config_rom for both addition and removal of
config_rom descriptors, as the ohci_set_config_rom() routine
will return -EBUSY if a previous pending update has not been
completed yet; the requested update just gets dropped on the floor.
This patch recognizes that the "in-flight" update can be modified
until it has been processed by the bus-reset, and the locking
in the bus_reset_tasklet ensures that the update is done atomically
with respect to modifications made by ohci_set_config_rom(). The
-EBUSY error case is simply removed.
[Stefan R: The bug always existed at least theoretically. But it
became easy to trigger since 2.6.36 commit 02d37bed18 "firewire: core:
integrate software-forced bus resets with bus management" which
introduced long mandatory delays between janitorial bus resets.]
Signed-off-by: Benjamin Buchalter <bj@mhlabs.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (trivial style changes)
Cc: <stable@kernel.org> # 2.6.36.y and newer
When z==2, the condition "key == 2" is superfluous because it cannot
occur without "b == 3", as a descriptor with b!=3 and key==2 would be
an OUTPUT_MORE_IMMEDIATE descriptor which cannot be used alone.
Also remove magic numbers and needless computations on the b field.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
For AT packet payloads of up to eight bytes, we have enough unused space
in the DMA descriptors list so that we can put a copy of the payload
there and thus avoid having to create a separate streaming DMA mapping
for the payload buffer.
In a CPU-bound microbenchmark that just sends 8-byte packets, bandwidth
was measured to increase by 5.7 %, from 1009 KB/s to 1067 KB/s. In
practice, the only performance-sensitive usage of small asynchronous
packets is the SBP-2 driver's write to the ORB_POINTER register during
SCSI command submission.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
OHCI 1.1 5.7.3 not only forbids enabling or starting any DMA contexts
before the linkEnable bit is set, but also explicitly warns of undefined
behaviour if this order is violated.
Don't violate it then.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
firewire: core: ignore link-active bit of new nodes, fix device recognition
firewire: sbp2: revert obsolete 'fix stall with "Unsolicited response"'
firewire: core: increase default SPLIT_TIMEOUT value
firewire: ohci: Misleading kfree in ohci.c::pci_probe/remove
firewire: ohci: omit IntEvent.busReset check rom AT queueing
firewire: ohci: prevent starting of iso contexts with empty queue
firewire: ohci: prevent iso completion callbacks after context stop
firewire: core: rename some variables
firewire: nosy: should work on Power Mac G4 PCI too
firewire: core: fix card->reset_jiffies overflow
firewire: cdev: remove unneeded reference
firewire: cdev: always wait for outbound transactions to complete
firewire: cdev: remove unneeded idr_find() from complete_transaction()
firewire: ohci: log dead DMA contexts
Like the older ieee1394 core driver, firewire-core skipped scanning of
any new node whose PHY sent a self ID without "link active" bit. If a
device had this bit off mistakenly, it meant that it was inaccessible to
kernel drivers with the old IEEE 1394 driver stack but could still be
accessed by userspace drivers through the raw1394 interface.
But with firewire-core, userspace drivers don't get to see such buggy
devices anymore. This is effectively a driver regression since this
device bug is otherwise harmless.
We now attempt to scan all devices, even repeaters that don't have a
link or powered-down devices that have everything but their PHY shut
down when plugged in. This results in futile repeated scanning attempts
in case of such devices that really don't have an active link, but this
doesn't hurt since recent workqueue infrastructure lets us run more
concurrent scanning jobs than we can shake a stick at.
This should fix accessibility of Focusrite Saffire PRO 26 I/O:
http://sourceforge.net/mailarchive/forum.php?thread_name=20110314215622.5c751bb0%40stein&forum_name=ffado-user
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Now that firewire-core sets the local node's SPLIT_TIMEOUT to 2 seconds
per default, commit a481e97d3c is no
longer required.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The SPLIT_TIMEOUT mechanism is intended to detect requests that somehow
got lost. However, when the timeout value is too low, transactions that
could have been completed successfully will be cancelled. Furthermore,
there are chips whose firmwares ignore the configured split timeout and
send late split response; known examples are the DM1x00 (BeBoB), TCD22x0
(DICE), and some OXUF936QSE firmwares.
This patch changes the default timeout to two seconds, which happens to
be the default on other OSes, too.
Actual lost requests are extremely rare, so there should be no practical
downside to increasing the split timeout even on devices that work
correctly.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Add a driver for two playback-only FireWire devices based on the OXFW970
chip.
v2: better AMDTP API abstraction; fix fw_unit leak; small fixes
v3: cache the iPCR value
v4: FireWave constraints; fix fw_device reference counting;
fix PCR caching; small changes and fixes
v5: volume/mute support; fix crashing due to pcm stop races
v6: fix build; one-channel volume for LaCie
v7: use signed values to make volume (range checks) work; fix function
block IDs for volume/mute; always use channel 0 for LaCie volume
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Acked-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Tested-by: Jay Fenlason <fenlason@redhat.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
It seems drivers/firewire/ohci.c is making some optimistic assumptions
about struct fw_ohci and that member "card" will always remain the first
member of the struct.
Plus it's probably going to confuse a lot of static code analyzers too.
So I wonder if there is a good reason not to free the ohci struct just
like it was allocated instead of the tricky &ohci->card way?
Signed-off-by: Oleg Drokin <green@linuxhacker.ru>
It is perhaps just a rudiment from before mainline submission of the
driver.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Since commit 82b662dc41 "flush AT contexts after bus reset for OHCI 1.2",
the driver takes care of any AT packets that were enqueued during a bus
reset phase. The check from commit 76f73ca1b2 is therefore no longer
necessary and the MMIO read can be avoided.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
If a misguided program tried to start an isochronous context before it
has queued any packets, the call would appear to succeed, but the
context would not actually go into the running state, and the OHCI
controller would then raise an unrecoverableError interrupt because the
first Z value is zero and thus invalid. The driver logs such errors,
but there is no mechanism to report this back to the program.
Add an explicit check so that this error can be returned synchronously.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
To prevent the iso packet callback from being called after
fw_iso_context_stop() has returned, make sure that the
context's tasklet has finished executing before that.
This fixes access-after-free bugs that have so far been
observed only in the upcoming snd-firewire-speakers driver,
but can theoretically also happen in the firedtv driver.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
In manage_channel(), rename the variables "c" and "i" to the more
expressive "bit" and "channel".
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The first board generation of Power Mac G4 ("Yikes!", those with PCI
graphics) still had a PCILynx controller like their G3 predecessors,
but not the later AGP models. (Jonathan Woithe recalls to have heard
of it, and some web sources reinforce it.)
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
On a 32-bit machine with, e.g., HZ=1000, jiffies will overflow after
about 50 days, so if there are between 25 and 50 days between bus
resets, the card->reset_jiffies comparisons can get wrong results.
To fix this, ensure that this timestamp always uses 64 bits.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: "Stefan Richter" <stefanr@s5r6.in-berlin.de>
For outbound transactions, the IDR's and the callback's references now
have exactly the same lifetime, so we do not need both of them.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: "Stefan Richter" <stefanr@s5r6.in-berlin.de>
We must not use fw_cancel_transaction() because it cannot correctly
abort still-active transactions. The only place in core-cdev where this
matters is when the file is released. Instead of trying to abort the
transactions, we wait for them to complete normally, i.e., until all
outbound transaction resources have been removed from the IDR tree.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: "Stefan Richter" <stefanr@s5r6.in-berlin.de>
Outbound transactions are never aborted with release_client_resource(),
so it is not necessary for complete_transaction() to check whether the
resource is still registered. Only shutdown_resource() can abort such
an transaction, and this is already handled with the in_shutdown check.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: "Stefan Richter" <stefanr@s5r6.in-berlin.de>
When a DMA context goes into the dead state (and the controller thus
stops working correctly), logging this error and the controller's error
code might be helpful for debugging.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
firewire: core: fix unstable I/O with Canon camcorder
* 'fwnet' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
firewire: net: is not experimental anymore
firewire: net: invalidate ARP entries of removed nodes
thanks to Clemens' and Maxim's fixes to firewire-ohci and -net in the
last two kernel releases.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This makes it possible to resume communication with a node that dropped
off the bus for a brief period. Otherwise communication will only be
possible after ARP cache entry timeouts.
Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (rebased)
Regression since commit 1038953674, "firewire: core: check for 1394a
compliant IRM, fix inaccessibility of Sony camcorder":
The camcorder Canon MV5i generates lots of bus resets when asynchronous
requests are sent to it (e.g. Config ROM read requests or FCP Command
write requests) if the camcorder is not root node. This causes drop-
outs in videos or makes the camcorder entirely inaccessible.
https://bugzilla.redhat.com/show_bug.cgi?id=633260
Fix this by allowing any Canon device, even if it is a pre-1394a IRM
like MV5i are, to remain root node (if it is at least Cycle Master
capable). With the FireWire controller cards that I tested, MV5i always
becomes root node when plugged in and left to its own devices.
Reported-by: Ralf Lange
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Cc: <stable@kernel.org> # 2.6.32.y and newer
PAGE_KERNEL_RO is not available on all architectures, so its use
in the new AR code broke compilation on sparc64.
Because the read-only mapping was just a debugging aid, just use
PAGE_KERNEL instead.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
James Bottomley wrote:
> On Thu, 2011-01-13 at 08:27 +0100, Clemens Ladisch wrote:
>> firewire: ohci: fix compilation on arches without PAGE_KERNEL_RO, e.g. sparc
>>
>> PAGE_KERNEL_RO is not available on all architectures, so its use in the
>> new AR code broke compilation on sparc64.
>>
>> Because the R/O mapping is only used to catch drivers that try to write
>> to the reception buffer and not actually required for correct operation,
>> we can just use a normal PAGE_KERNEL mapping where _RO is not available.
[...]
>> +/*
>> + * For archs where PAGE_KERNEL_RO is not supported;
>> + * mapping the AR buffers readonly for the CPU is just a debugging aid.
>> + */
>> +#ifndef PAGE_KERNEL_RO
>> +#define PAGE_KERNEL_RO PAGE_KERNEL
>> +#endif
>
> This might cause interesting issues on sparc64 if it ever acquired a
> PAGE_KERNEL_RO. Sparc64 has extern pgprot_t for it's PAGE_KERNEL types
> rather than #defines, so the #ifdef check wouldn't see this.
>
> I think either PAGE_PROT_RO becomes part of our arch API (so all
> architectures are forced to add it), or, if it's not part of the API,
> ohci isn't entitled to use it. The latter seems simplest since you have
> no real use for write protection anyway.
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Instead of starting the split transaction timeout timer when any request
is submitted, start it only when the destination's ACK_PENDING has been
received. This prevents us from using a timeout that is too short, and,
if the controller's AT queue is emptying very slowly, from cancelling
a packet that has not yet been sent.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
"firewire: ohci: restart iso DMA contexts on resume from low power mode"
added the flag struct context.active and "firewire: ohci: cache the
context run bit" added struct context.running.
These flags contain the same information; combine them.
Also, normalize whitespace in pci_resume().
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The DMA context run control bit is entirely controlled by software, so
it is safe to cache it. This allows the driver to avoid doing an
additional MMIO read when queueing an AT packet.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Add comments
- on why bus_reset_tasklet flushes AT queues,
- that commit 76f73ca1b2 can possibly be reverted now.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Acked-by: Jarod Wilson <jarod@redhat.com>
The OHCI 1.2 (draft) specification, clause 7.2.3.3, allows and
recommends that, after a bus reset, the controller does not flush all
the packets in the AT queues. Therefore, the driver has to do this
itself.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
At ifup, carrier status would be shown on even if it actually was off.
Also add an include for ethtool_ops rather than to rely on the one from
netdevice.h.
Note, we can alas not use fwnet_device_mutex to serialize access to
dev->peer_count (as I originally wanted). This would cause a lock
inversion:
- fwnet_probe | takes fwnet_device_mutex
+ register_netdev | takes rtnl_mutex
- devinet_ioctl | takes rtnl_mutex
+ fwnet_open | ...must not take fwnet_device_mutex
Hence use the dev->lock spinlock for serialization.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
To make userland, e.g. NetworkManager work with firewire, we need to
detect whether cable is plugged or not. Simple and correct way of doing
that is just counting number of peers. No peers - no link and vice
versa.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Unfortunately its easy to trigger such error messages by removing the
cable while sending streams of data over the link.
Such errors are normal, and therefore this patch stops firewire-net from
flooding the kernel log with these errors, by combining series of same
errors together.
Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
(Stefan R:) Eventually we should remove this logging when firewire-net
and related firewire-ohci facilities have been stabilized.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Restore iso channels DMA so that iso channels could continue to work
after resume from RAM/disk.
Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Some lousy BIOSes, e.g. my Aspire 5720 BIOS forget to restore the GUID
register on resume from RAM.
Fix that by setting it to the last value that was read from it.
Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The buffers used for the selfIDs packets and the AR request and response
descriptors end up using three pages because dma_alloc_coherent()
allocates at least one page per call. However, these data structures
would all fit into 4 KB, so we can save space by using a common buffer
for them.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
When the isochRx/isochTx bit is clear, we do not need to read the
corresponding iso interrupt event register.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Change the header of PHY packets to be sent to include a pseudo
transaction code. This makes the header consistent with that of
received PHY packets, and allows at_context_queue_packet() and
log_ar_at_event() to see the packet type directly instead of having
to deduce it from the header length or even from the header contents.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
To remove the error information from the controller's queue and to allow
more posted writes, the driver has to read the failed posted write
address before clearing the postedWriteErr interrupt bit.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
(Stefan R:) The spec is somewhat fuzzy about the actual requirements.
To err on the safe side, let's do these two read accesses.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Make sure that interrupt event clear bit writes are executed before the
interrupt handler returns.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Add proper error handling for the context_init() calls.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Move the AR DMA descriptors out of the buffer pages, and map the buffer
pages linearly into the kernel's address space. This allows the driver
to ignore any page boundaries in the DMA data and thus to avoid any
copying around of packet payloads.
This fixes the bug where S800 packets that are so big (> 4080 bytes)
that they can be split over three pages were not handled correctly.
Due to the changed algorithm, we can now use arbitrarily many buffer
pages, which improves performance because the controller can more easily
unload its DMA FIFO.
Furthermore, using streaming DMA mappings should improve perfomance on
architectures where coherent DMA mappings are not cacheable. Even on
other architectures, the caching behaviour should be improved slightly
because the CPU no longer writes to the buffer pages.
v2: Detect the last filled buffer page by searching the descriptor's
residual count value fields in order (like in the old code), instead
of going backwards through the transfer status fields; it looks as
if some controllers do not set the latter correctly.
v3: Fix an old resume bug that would now make the handler run into
a BUG_ON, and replace that check with more useful error handling.
Increase the buffer size for better performance with non-TI chips.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Maxim Levitsky writes:
Works almost perfectly. I can still see RCODE_BUSY errors
sometimes, not very often though. 64K here eliminates these errors
completely. This is most likely due to nouveau drivers and lowest
perf level I use to lower card temperature. That increases
latencies too much I think. Besides that the IO is just perfect.
Tested-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>