async_syndrome_val check the P and Q blocks used for RAID6
calculations.
With DDF raid6, some of the data blocks might be NULL, so
this needs to be handled in the same way that async_gen_syndrome
handles it.
As async_syndrome_val calls async_xor, also enhance async_xor
to detect and skip NULL blocks in the list.
Signed-off-by: NeilBrown <neilb@suse.de>
md/raid6 passes a list of 'struct page *' to the async_tx routines,
which then either DMA map them for offload, or take the page_address
for CPU based calculations.
For RAID6 we sometime leave 'blanks' in the list of pages.
For CPU based calcs, we want to treat theses as a page of zeros.
For offloaded calculations, we simply don't pass a page to the
hardware.
Currently the 'blanks' are encoded as a pointer to
raid6_empty_zero_page. This is a 4096 byte memory region, not a
'struct page'. This is mostly handled correctly but is rather ugly.
So change the code to pass and expect a NULL pointer for the blanks.
When taking page_address of a page, we need to check for a NULL and
in that case use raid6_empty_zero_page.
Signed-off-by: NeilBrown <neilb@suse.de>
If we are unable to offload async_mult() or async_sum_product(), then
unmap the buffers before falling through to the synchronous path.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Testing on x86_64 with NDISKS=255 yields:
do_IRQ: modprobe near stack overflow (cur:ffff88007d19c000,sp:ffff88007d19c128)
...and eventually
general protection fault: 0000 [#1]
Moving the scribble buffers off the stack allows the test to complete
successfully.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Some engines have transfer size and address alignment restrictions. Add
a per-operation alignment property to struct dma_device that the async
routines and dmatest can use to check alignment capabilities.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Channel switching is problematic for some dmaengine drivers as the
architecture precludes separating the ->prep from ->submit. In these
cases the driver can select ASYNC_TX_DISABLE_CHANNEL_SWITCH to modify
the async_tx allocator to only return channels that support all of the
required asynchronous operations.
For example MD_RAID456=y selects support for asynchronous xor, xor
validate, pq, pq validate, and memcpy. When
ASYNC_TX_DISABLE_CHANNEL_SWITCH=y any channel with all these
capabilities is marked DMA_ASYNC_TX allowing async_tx_find_channel() to
quickly locate compatible channels with the guarantee that dependency
chains will remain on one channel. When
ASYNC_TX_DISABLE_CHANNEL_SWITCH=n async_tx_find_channel() may select
channels that lead to operation chains that need to cross channel
boundaries using the async_tx channel switch capability.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Some engines optimize operation by reading ahead in the descriptor chain
such that descriptor2 may start execution before descriptor1 completes.
If descriptor2 depends on the result from descriptor1 then a fence is
required (on descriptor2) to disable this optimization. The async_tx
api could implicitly identify dependencies via the 'depend_tx'
parameter, but that would constrain cases where the dependency chain
only specifies a completion order rather than a data dependency. So,
provide an ASYNC_TX_FENCE to explicitly identify data dependencies.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Port drivers/md/raid6test/test.c to use the async raid6 recovery
routines. This is meant as a unit test for raid6 acceleration drivers. In
addition to the 16-drive test case this implements tests for the 4-disk and
5-disk special cases (dma devices can not generically handle less than 2
sources), and adds a test for the D+Q case.
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
async_raid6_2data_recov() recovers two data disk failures
async_raid6_datap_recov() recovers a data disk and the P disk
These routines are a port of the synchronous versions found in
drivers/md/raid6recov.c. The primary difference is breaking out the xor
operations into separate calls to async_xor. Two helper routines are
introduced to perform scalar multiplication where needed.
async_sum_product() multiplies two sources by scalar coefficients and
then sums (xor) the result. async_mult() simply multiplies a single
source by a scalar.
This implemention also includes, in contrast to the original
synchronous-only code, special case handling for the 4-disk and 5-disk
array cases. In these situations the default N-disk algorithm will
present 0-source or 1-source operations to dma devices. To cover for
dma devices where the minimum source count is 2 we implement 4-disk and
5-disk handling in the recovery code.
[ Impact: asynchronous raid6 recovery routines for 2data and datap cases ]
Cc: Yuri Tikhonov <yur@emcraft.com>
Cc: Ilya Yanok <yanok@emcraft.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: David Woodhouse <David.Woodhouse@intel.com>
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
[ Based on an original patch by Yuri Tikhonov ]
This adds support for doing asynchronous GF multiplication by adding
two additional functions to the async_tx API:
async_gen_syndrome() does simultaneous XOR and Galois field
multiplication of sources.
async_syndrome_val() validates the given source buffers against known P
and Q values.
When a request is made to run async_pq against more than the hardware
maximum number of supported sources we need to reuse the previous
generated P and Q values as sources into the next operation. Care must
be taken to remove Q from P' and P from Q'. For example to perform a 5
source pq op with hardware that only supports 4 sources at a time the
following approach is taken:
p, q = PQ(src0, src1, src2, src3, COEF({01}, {02}, {04}, {08}))
p', q' = PQ(p, q, q, src4, COEF({00}, {01}, {00}, {10}))
p' = p + q + q + src4 = p + src4
q' = {00}*p + {01}*q + {00}*q + {10}*src4 = q + {10}*src4
Note: 4 is the minimum acceptable maxpq otherwise we punt to
synchronous-software path.
The DMA_PREP_CONTINUE flag indicates to the driver to reuse p and q as
sources (in the above manner) and fill the remaining slots up to maxpq
with the new sources/coefficients.
Note1: Some devices have native support for P+Q continuation and can skip
this extra work. Devices with this capability can advertise it with
dma_set_maxpq. It is up to each driver how to handle the
DMA_PREP_CONTINUE flag.
Note2: The api supports disabling the generation of P when generating Q,
this is ignored by the synchronous path but is implemented by some dma
devices to save unnecessary writes. In this case the continuation
algorithm is simplified to only reuse Q as a source.
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Ilya Yanok <yanok@emcraft.com>
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
We currently walk the parent chain when waiting for a given tx to
complete however this walk may race with the driver cleanup routine.
The routines in async_raid6_recov.c may fall back to the synchronous
path at any point so we need to be prepared to call async_tx_quiesce()
(which calls dma_wait_for_async_tx). To remove the ->parent walk we
guarantee that every time a dependency is attached ->issue_pending() is
invoked, then we can simply poll the initial descriptor until
completion.
This also allows for a lighter weight 'issue pending' implementation as
there is no longer a requirement to iterate through all the channels'
->issue_pending() routines as long as operations have been submitted in
an ordered chain. async_tx_issue_pending() is added for this case.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
If module_init and module_exit are nops then neither need to be defined.
[ Impact: pure cleanup ]
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Replace the flat zero_sum_result with a collection of flags to contain
the P (xor) zero-sum result, and the soon to be utilized Q (raid6 reed
solomon syndrome) zero-sum result. Use the SUM_CHECK_ namespace instead
of DMA_ since these flags will be used on non-dma-zero-sum enabled
platforms.
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
async_xor() needs space to perform dma and page address conversions. In
most cases the code can simply reuse the struct page * array because the
size of the native pointer matches the size of a dma/page address. In
order to support archs where sizeof(dma_addr_t) is larger than
sizeof(struct page *), or to preserve the input parameters, we utilize a
memory region passed in by the caller.
Since the code is now prepared to handle the case where it cannot
perform address conversions on the stack, we no longer need the
!HIGHMEM64G dependency in drivers/dma/Kconfig.
[ Impact: don't clobber input buffers for address conversions ]
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Prepare the api for the arrival of a new parameter, 'scribble'. This
will allow callers to identify scratchpad memory for dma address or page
address conversions. As this adds yet another parameter, take this
opportunity to convert the common submission parameters (flags,
dependency, callback, and callback argument) into an object that is
passed by reference.
Also, take this opportunity to fix up the kerneldoc and add notes about
the relevant ASYNC_TX_* flags for each routine.
[ Impact: moves api pass-by-value parameters to a pass-by-reference struct ]
Signed-off-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In support of inter-channel chaining async_tx utilizes an ack flag to
gate whether a dependent operation can be chained to another. While the
flag is not set the chain can be considered open for appending. Setting
the ack flag closes the chain and flags the descriptor for garbage
collection. The ASYNC_TX_DEP_ACK flag essentially means "close the
chain after adding this dependency". Since each operation can only have
one child the api now implicitly sets the ack flag at dependency
submission time. This removes an unnecessary management burden from
clients of the api.
[ Impact: clean up and enforce one dependency per operation ]
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
'zero_sum' does not properly describe the operation of generating parity
and checking that it validates against an existing buffer. Change the
name of the operation to 'val' (for 'validate'). This is in
anticipation of the p+q case where it is a requirement to identify the
target parity buffers separately from the source buffers, because the
target parity buffers will not have corresponding pq coefficients.
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
To allow an async_tx routine to be compiled away on HAS_DMA=n arch it
needs to be declared __always_inline otherwise the compiler may emit
code and cause a link error.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Now that clients no longer need to be notified of channel arrival
dma_async_client_register can simply increment the dmaengine_ref_count.
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
async_tx and net_dma each have open-coded versions of issue_pending_all,
so provide a common routine in dmaengine.
The implementation needs to walk the global device list, so implement
rcu to allow dma_issue_pending_all to run lockless. Clients protect
themselves from channel removal events by holding a dmaengine reference.
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Allowing multiple clients to each define their own channel allocation
scheme quickly leads to a pathological situation. For memory-to-memory
offload all clients can share a central allocator.
This simply moves the existing async_tx allocator to dmaengine with
minimal fixups:
* async_tx.c:get_chan_ref_by_cap --> dmaengine.c:nth_chan
* async_tx.c:async_tx_rebalance --> dmaengine.c:dma_channel_rebalance
* split out common code from async_tx.c:__async_tx_find_channel -->
dma_find_channel
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Simply, if a client wants any dmaengine channel then prevent all dmaengine
modules from being removed. Once the clients are done re-enable module
removal.
Why?, beyond reducing complication:
1/ Tracking reference counts per-transaction in an efficient manner, as
is currently done, requires a complicated scheme to avoid cache-line
bouncing effects.
2/ Per-transaction ref-counting gives the false impression that a
dma-driver can be gracefully removed ahead of its user (net, md, or
dma-slave)
3/ None of the in-tree dma-drivers talk to hot pluggable hardware, but
if such an engine were built one day we still would not need to notify
clients of remove events. The driver can simply return NULL to a
->prep() request, something that is much easier for a client to handle.
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
async_tx.ko is a consumer of dma channels. A circular dependency arises
if modules in drivers/dma rely on common code in async_tx.ko. It
prevents either module from being unloaded.
Move dma_wait_for_async_tx and async_tx_run_dependencies to dmaeninge.o
where they should have been from the beginning.
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Mapping the destination multiple times is a misuse of the dma-api.
Since the destination may be reused as a source, ensure that it is only
mapped once and that it is mapped bidirectionally. This appears to add
ugliness on the unmap side in that it always reads back the destination
address from the descriptor, but gcc can determine that dma_unmap is a
nop and not emit the code that calculates its arguments.
Cc: <stable@kernel.org>
Cc: Saeed Bishara <saeed@marvell.com>
Acked-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
* Rename 'next' to 'dep'
* Move the channel switch check inside the loop to simplify
termination
Acked-by: Ilya Yanok <yanok@emcraft.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Should clear the next pointer of the TX if we are sure that the
next TX (say NXT) will be submitted to the channel too. Overwise,
we break the chain of descriptors, because we lose the information
about the next descriptor to run. So next time, when invoke
async_tx_run_dependencies() with TX, it's TX->next will be NULL, and
NXT will be never submitted.
Cc: <stable@kernel.org> [2.6.26]
Signed-off-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Ilya Yanok <yanok@emcraft.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx: (24 commits)
I/OAT: I/OAT version 3.0 support
I/OAT: tcp_dma_copybreak default value dependent on I/OAT version
I/OAT: Add watchdog/reset functionality to ioatdma
iop_adma: cleanup iop_chan_xor_slot_count
iop_adma: document how to calculate the minimum descriptor pool size
iop_adma: directly reclaim descriptors on allocation failure
async_tx: make async_tx_test_ack a boolean routine
async_tx: remove depend_tx from async_tx_sync_epilog
async_tx: export async_tx_quiesce
async_tx: fix handling of the "out of descriptor" condition in async_xor
async_tx: ensure the xor destination buffer remains dma-mapped
async_tx: list_for_each_entry_rcu() cleanup
dmaengine: Driver for the Synopsys DesignWare DMA controller
dmaengine: Add slave DMA interface
dmaengine: add DMA_COMPL_SKIP_{SRC,DEST}_UNMAP flags to control dma unmap
dmaengine: Add dma_client parameter to device_alloc_chan_resources
dmatest: Simple DMA memcpy test client
dmaengine: DMA engine driver for Marvell XOR engine
iop-adma: fix platform driver hotplug/coldplug
dmaengine: track the number of clients using a channel
...
Fixed up conflict in drivers/dca/dca-sysfs.c manually
All callers of async_tx_sync_epilog have called async_tx_quiesce on the
depend_tx, so async_tx_sync_epilog need only call the callback to
complete the operation.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Ensure forward progress is made when a dmaengine driver is unable to
allocate an xor descriptor by breaking the dependency chain with
async_tx_quisce() and issue any pending descriptors.
Tested with iop-adma by setting device->max_xor = 2 to force multiple
calls to device_prep_dma_xor for each call to async_xor and limiting the
descriptor slot pool to 5. Discovered that the minimum descriptor pool
size for iop-adma is 2 * iop_chan_xor_slot_cnt(device->max_xor) + 1.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
When the number of source buffers for an xor operation exceeds the hardware
channel maximum async_xor creates a chain of dependent operations. The result
of one operation is reused as an input to the next to continue the xor
calculation. The destination buffer should remain mapped for the duration of
the entire chain. To provide this guarantee the code must no longer be allowed
to fallback to the synchronous path as this will preclude the buffer from being
unmapped, i.e. the dma-driver will potentially miss the descriptor with
!DMA_COMPL_SKIP_DEST_UNMAP.
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In the rcu update side, don't use list_for_each_entry_rcu().
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
commit 636bdeaa 'dmaengine: ack to flags: make use of the unused bits in
the 'ack' field' missed an ->ack conversion in
crypto/async_tx/async_memset.c
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Move rcu-protected lists from list.h into a new header file rculist.h.
This is done because list are a very used primitive structure all over the
kernel and it's currently impossible to include other header files in this
list.h without creating some circular dependencies.
For example, list.h implements rcu-protected list and uses rcu_dereference()
without including rcupdate.h. It actually compiles because users of
rcu_dereference() are macros. Others RCU functions could be used too but
aren't probably because of this.
Therefore this patch creates rculist.h which includes rcupdates without to
many changes/troubles.
Signed-off-by: Franck Bui-Huu <fbuihuu@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Josh Triplett <josh@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
'ack' is currently a simple integer that flags whether or not a client is done
touching fields in the given descriptor. It is effectively just a single bit
of information. Converting this to a flags parameter allows the other bits to
be put to use to control completion actions, like dma-unmap, and capture
results, like xor-zero-sum == 0.
Changes are one of:
1/ convert all open-coded ->ack manipulations to use async_tx_ack
and async_tx_test_ack.
2/ set the ack bit at prep time where possible
3/ make drivers store the flags at prep time
4/ add flags to the device_prep_dma_interrupt prototype
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Shrink struct dma_async_tx_descriptor and introduce
async_tx_channel_switch to properly inject a channel switch interrupt in
the descriptor stream. This simplifies the locking model as drivers no
longer need to handle dma_async_tx_descriptor.lock.
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
If the channel cannot perform the operation in one call to
->device_prep_dma_zero_sum, then fallback to the xor+page_is_zero path.
This only affects users with arrays larger than 16 devices on iop13xx or
32 devices on iop3xx.
Cc: <stable@kernel.org>
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The source and destination addresses are included to allow channel
selection based on address alignment.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Pass a full set of flags to drivers' per-operation 'prep' routines.
Currently the only flag passed is DMA_PREP_INTERRUPT. The expectation is
that arch-specific async_tx_find_channel() implementations can exploit this
capability to find the best channel for an operation.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Reviewed-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
The tx_set_src and tx_set_dest methods were originally implemented to allow
an array of addresses to be passed down from async_xor to the dmaengine
driver while minimizing stack overhead. Removing these methods allows
drivers to have all transaction parameters available at 'prep' time, saves
two function pointers in struct dma_async_tx_descriptor, and reduces the
number of indirect branches..
A consequence of moving this data to the 'prep' routine is that
multi-source routines like async_xor need temporary storage to convert an
array of linear addresses into an array of dma addresses. In order to keep
the same stack footprint of the previous implementation the input array is
reused as storage for the dma addresses. This requires that
sizeof(dma_addr_t) be less than or equal to sizeof(void *). As a
consequence CONFIG_DMADEVICES now depends on !CONFIG_HIGHMEM64G. It also
requires that drivers be able to make descriptor resources available when
the 'prep' routine is polled.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Remove the unused ASYNC_TX_ASSUME_COHERENT flag. Async_tx is
meant to hide the difference between asynchronous hardware and synchronous
software operations, this flag requires clients to understand cache
coherency consequences of the async path.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
single list_head variable initialized with LIST_HEAD_INIT could almost
always can be replaced with LIST_HEAD declaration, this shrinks the code
and looks better.
Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
do_async_xor must be compiled away on !HAS_DMA archs.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Fix dma_wait_for_async_tx to not loop forever in the case where a
dependency chain is longer than two entries. This condition will not
happen with current in-kernel drivers, but fix it for future drivers.
Found-by: Saeed Bishara <saeed.bishara@gmail.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Andrew Morton:
[async_memcpy] is very wrong if both ASYNC_TX_KMAP_DST and
ASYNC_TX_KMAP_SRC can ever be set. We'll end up using the same kmap
slot for both src add dest and we get either corrupted data or a BUG.
Evgeniy Polyakov:
Btw, shouldn't it always be kmap_atomic() even if flag is not set.
That pages are usual one returned by alloc_page().
So fix the usage of kmap_atomic and kill the ASYNC_TX_KMAP_DST and
ASYNC_TX_KMAP_SRC flags.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The async_tx api provides methods for describing a chain of asynchronous
bulk memory transfers/transforms with support for inter-transactional
dependencies. It is implemented as a dmaengine client that smooths over
the details of different hardware offload engine implementations. Code
that is written to the api can optimize for asynchronous operation and the
api will fit the chain of operations to the available offload resources.
I imagine that any piece of ADMA hardware would register with the
'async_*' subsystem, and a call to async_X would be routed as
appropriate, or be run in-line. - Neil Brown
async_tx exploits the capabilities of struct dma_async_tx_descriptor to
provide an api of the following general format:
struct dma_async_tx_descriptor *
async_<operation>(..., struct dma_async_tx_descriptor *depend_tx,
dma_async_tx_callback cb_fn, void *cb_param)
{
struct dma_chan *chan = async_tx_find_channel(depend_tx, <operation>);
struct dma_device *device = chan ? chan->device : NULL;
int int_en = cb_fn ? 1 : 0;
struct dma_async_tx_descriptor *tx = device ?
device->device_prep_dma_<operation>(chan, len, int_en) : NULL;
if (tx) { /* run <operation> asynchronously */
...
tx->tx_set_dest(addr, tx, index);
...
tx->tx_set_src(addr, tx, index);
...
async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
} else { /* run <operation> synchronously */
...
<operation>
...
async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
}
return tx;
}
async_tx_find_channel() returns a capable channel from its pool. The
channel pool is organized as a per-cpu array of channel pointers. The
async_tx_rebalance() routine is tasked with managing these arrays. In the
uniprocessor case async_tx_rebalance() tries to spread responsibility
evenly over channels of similar capabilities. For example if there are two
copy+xor channels, one will handle copy operations and the other will
handle xor. In the SMP case async_tx_rebalance() attempts to spread the
operations evenly over the cpus, e.g. cpu0 gets copy channel0 and xor
channel0 while cpu1 gets copy channel 1 and xor channel 1. When a
dependency is specified async_tx_find_channel defaults to keeping the
operation on the same channel. A xor->copy->xor chain will stay on one
channel if it supports both operation types, otherwise the transaction will
transition between a copy and a xor resource.
Currently the raid5 implementation in the MD raid456 driver has been
converted to the async_tx api. A driver for the offload engines on the
Intel Xscale series of I/O processors, iop-adma, is provided in a later
commit. With the iop-adma driver and async_tx, raid456 is able to offload
copy, xor, and xor-zero-sum operations to hardware engines.
On iop342 tiobench showed higher throughput for sequential writes (20 - 30%
improvement) and sequential reads to a degraded array (40 - 55%
improvement). For the other cases performance was roughly equal, +/- a few
percentage points. On a x86-smp platform the performance of the async_tx
implementation (in synchronous mode) was also +/- a few percentage points
of the original implementation. According to 'top' on iop342 CPU
utilization drops from ~50% to ~15% during a 'resync' while the speed
according to /proc/mdstat doubles from ~25 MB/s to ~50 MB/s.
The tiobench command line used for testing was: tiobench --size 2048
--block 4096 --block 131072 --dir /mnt/raid --numruns 5
* iop342 had 1GB of memory available
Details:
* if CONFIG_DMA_ENGINE=n the asynchronous path is compiled away by making
async_tx_find_channel a static inline routine that always returns NULL
* when a callback is specified for a given transaction an interrupt will
fire at operation completion time and the callback will occur in a
tasklet. if the the channel does not support interrupts then a live
polling wait will be performed
* the api is written as a dmaengine client that requests all available
channels
* In support of dependencies the api implicitly schedules channel-switch
interrupts. The interrupt triggers the cleanup tasklet which causes
pending operations to be scheduled on the next channel
* Xor engines treat an xor destination address differently than a software
xor routine. To the software routine the destination address is an implied
source, whereas engines treat it as a write-only destination. This patch
modifies the xor_blocks routine to take a an explicit destination address
to mirror the hardware.
Changelog:
* fixed a leftover debug print
* don't allow callbacks in async_interrupt_cond
* fixed xor_block changes
* fixed usage of ASYNC_TX_XOR_DROP_DEST
* drop dma mapping methods, suggested by Chris Leech
* printk warning fixups from Andrew Morton
* don't use inline in C files, Adrian Bunk
* select the API when MD is enabled
* BUG_ON xor source counts <= 1
* implicitly handle hardware concerns like channel switching and
interrupts, Neil Brown
* remove the per operation type list, and distribute operation capabilities
evenly amongst the available channels
* simplify async_tx_find_channel to optimize the fast path
* introduce the channel_table_initialized flag to prevent early calls to
the api
* reorganize the code to mimic crypto
* include mm.h as not all archs include it in dma-mapping.h
* make the Kconfig options non-user visible, Adrian Bunk
* move async_tx under crypto since it is meant as 'core' functionality, and
the two may share algorithms in the future
* move large inline functions into c files
* checkpatch.pl fixes
* gpl v2 only correction
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-By: NeilBrown <neilb@suse.de>
