There are cases where cacheline-unaligned raid operations can hang the
dma channel. Simply disable these operations by increasing the
alignment constraints published to async_tx. The raid456 driver always
issues page aligned requests, so the only in-kernel user of the ioatdma
driver that is affected by this change is dmatest.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Use separate locks for the descriptor prep (producer) and descriptor
cleanup (consumer) paths. Allows the producer path to run concurrently
with the cleanup path. Inspired by Documentation/circular-buffer.txt.
Cc: David Howells <dhowells@redhat.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Constify struct sysfs_ops.
This is part of the ops structure constification
effort started by Arjan van de Ven et al.
Benefits of this constification:
* prevents modification of data that is shared
(referenced) by many other structure instances
at runtime
* detects/prevents accidental (but not intentional)
modification attempts on archs that enforce
read-only kernel data at runtime
* potentially better optimized code as the compiler
can assume that the const data cannot be changed
* the compiler/linker move const data into .rodata
and therefore exclude them from false sharing
Signed-off-by: Emese Revfy <re.emese@gmail.com>
Acked-by: David Teigland <teigland@redhat.com>
Acked-by: Matt Domsch <Matt_Domsch@dell.com>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Acked-by: Hans J. Koch <hjk@linutronix.de>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Rename for_each_bit to for_each_set_bit in the kernel source tree. To
permit for_each_clear_bit(), should that ever be added.
The patch includes a macro to map the old for_each_bit() onto the new
for_each_set_bit(). This is a (very) temporary thing to ease the migration.
[akpm@linux-foundation.org: add temporary for_each_bit()]
Suggested-by: Alexey Dobriyan <adobriyan@gmail.com>
Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Artem Bityutskiy <dedekind@infradead.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If the calling convention of ->timer_fn() and ->cleanup_fn() are unified
across hardware versions we can drop parameters to ioat_init_channel() and
unify ioat_is_dma_complete() implementations.
Both ->timer_fn() and ->cleanup_fn() are modified to expect a struct
dma_chan pointer.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The hardware automatically disables further interrupts after each event
until rearmed. This allows a delay to be injected between the occurence
of the interrupt and the running of the cleanup routine. The delay is
scaled by the descriptor backlog and then written to the INTRDELAY
register which specifies the number of microseconds to hold off
interrupt delivery after an interrupt event occurs. According to
powertop this reduces the interrupt rate from ~5000 intr/s to ~150
intr/s per without affecting throughput (simple dd to a raid6 array).
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Since ioat_cleanup_preamble() and the update of the last completed
descriptor are not synchronized there is a chance that two cleanup threads
can see descriptors to clean. If the first cleans up all pending
descriptors then the second will trigger the BUG_ON.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The pending == 2 case no longer exists in the driver so, we can use
ioat2_ring_pending() outside the lock to determine if there might be any
descriptors in the ring that the hardware has not seen.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
We already disallow raid operations while DCA is globally enabled, so
having it locally enabled is a nop and confusing when reading the code.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Fix typo in ioat2_quiesce. check 'tmo' is zero, not 'end'. Also applies
to 2.6.32.3
Cc: <stable@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Put the ioat2 and ioat3 state machines in the halted state with all
errors cleared.
The ioat1 init path is not disturbed for stability, there are no
reported ioat1 initiaization issues.
Cc: <stable@kernel.org>
Reported-by: Roland Dreier <rdreier@cisco.com>
Tested-by: Roland Dreier <rdreier@cisco.com>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
When continuing a pq calculation the driver needs 3 extra sources. The
driver can perform a 3 source calculation with a single descriptor, but
needs an extended descriptor to process up to 8 sources in one
operation. However, in the p-disabled case only one extra source is
needed. When continuing a p-disabled operation there are occasions
(i.e. 0 < src_cnt % 8 < 3) where the tail operation does not need an
extended descriptor. Properly account for this fact otherwise invalid
'dmacount' values will be written to hardware usually causing the
channel to halt with 'invalid descriptor' errors.
Cc: <stable@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The completion of a pq operation is notified with a null descriptor
appended to the end of the chain. This descriptor needs to be visible
to dma clients otherwise the client is precluded from ensuring all
operations are quiesced before freeing channel resources, i.e. due to
descriptor polling it may get the completion notification ahead of the
interrupt delivered by the null descriptor.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
ioat3.2 does not support asynchronous error notifications which makes
the driver experience latencies when non-zero pq validate results are
expected. Provide a mechanism for turning off async_xor_val and
async_syndrome_val via Kconfig. This approach is generally useful for
any driver that specifies ASYNC_TX_DISABLE_CHANNEL_SWITCH and would like
to force the async_tx api to fall back to the synchronous path for
certain operations.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Error interrupts and error completions may cause channel hangs, so
poll the channel status register after a timeout.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
RAID operations cause a system hang on platforms with DCA
(Direct-Cache-Access) enabled. So turn off RAID capabilities in this
case.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/dma/ioat/dma_v3.c: In function 'ioat3_prep_memset_lock':
drivers/dma/ioat/dma_v3.c:439: warning: 'fill' may be used uninitialized in this function
drivers/dma/ioat/dma_v3.c:437: warning: 'desc' may be used uninitialized in this function
drivers/dma/ioat/dma_v3.c: In function '__ioat3_prep_xor_lock':
drivers/dma/ioat/dma_v3.c:489: warning: 'xor' may be used uninitialized in this function
drivers/dma/ioat/dma_v3.c:486: warning: 'desc' may be used uninitialized in this function
drivers/dma/ioat/dma_v3.c: In function '__ioat3_prep_pq_lock':
drivers/dma/ioat/dma_v3.c:631: warning: 'pq' may be used uninitialized in this function
drivers/dma/ioat/dma_v3.c:628: warning: 'desc' may be used uninitialized in this function
gcc-4.0, unlike gcc-4.3, does not see that these variables are
initialized before use. Convert the descriptor loops to do-while make
this initialization apparent.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/dma/ioat/dma_v2.c: In function 'ioat2_dma_prep_memcpy_lock':
drivers/dma/ioat/dma_v2.c:680: warning: 'hw' may be used uninitialized in this function
drivers/dma/ioat/dma_v2.c:681: warning: 'desc' may be used uninitialized in this function
Cc: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
With the addition of ioat_max_alloc_order it is not clear what the
maximum allocation order is, so document that in the modinfo. Also take
an opportunity to kill a stray semicolon.
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
A new ring implementation and the addition of raid functionality
constitutes a bump in the driver major version number.
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
This patch enables DCA support on multiple-IOH/multiple-IIO architectures.
It modifies dca module by replacing single dca_providers list
with dca_domains list, each domain containing separate list of providers.
This approach lets dca driver manage multiple domains, i.e. sets of providers
and requesters mapped back to the same PCI root complex device.
The driver takes care to register each requester to a provider
from the same domain.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
All the necessary fields for handling an ioat2,3 ring entry can fit into
one cacheline. Move ->len prior to ->txd in struct ioat_ring_ent, and
move allocation of these entries to a hw-cache-aligned kmem cache to
reduce the number of cachelines dirtied for descriptor management.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Drop ioatdma's use of tx_list from struct dma_async_tx_descriptor in
preparation for removal of this field.
Cc: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Trivial cleanup to make the PCI ID table easier to read.
[dan.j.williams@intel.com: extended to v3.2 devices]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The ioatdma module is missing aliases for the PCI devices it supports,
so it is not autoloaded on boot. Add a MODULE_DEVICE_TABLE() to get
these aliases.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The cleanup routine for the raid cases imposes extra checks for handling
raid descriptors and extended descriptors. If the channel does not
support raid it can avoid this extra overhead by using the ioat2 cleanup
path.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Jasper Forest introduces raid offload support via ioat3.2 support. When
raid offload is enabled two (out of 8 channels) will report raid5/raid6
offload capabilities. The remaining channels will only report ioat3.0
capabilities (memcpy).
Signed-off-by: Tom Picard <tom.s.picard@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The async_tx api uses the DMA_INTERRUPT operation type to terminate a
chain of issued operations with a callback routine.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
If a platform advertises pq capabilities, but not xor, then use
ioat3_prep_pqxor and ioat3_prep_pqxor_val to simulate xor support.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
ioat3.2 adds support for raid6 syndrome generation (xor sum of galois
field multiplication products) using up to 8 sources. It can also
perform an pq-zero-sum operation to validate whether the syndrome for a
given set of sources matches a previously computed syndrome.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
This adds a hardware specific self test to be called from ioat_probe.
In the ioat3 case we will have tests for all the different raid
operations, while ioat1 and ioat2 will continue to just test memcpy.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
ioat3.2 adds xor offload support for up to 8 sources. It can also
perform an xor-zero-sum operation to validate whether all given sources
sum to zero, without writing to a destination. Xor descriptors differ
from memcpy in that one operation may require multiple descriptors
depending on the number of sources. When the number of sources exceeds
5 an extended descriptor is needed. These descriptors need to be
accounted for when updating the DMA_COUNT register.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Tag completion writes for direct cache access to reduce the latency of
checking for descriptor completions.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Export driver attributes for diagnostic purposes:
'ring_size': total number of descriptors available to the engine
'ring_active': number of descriptors in-flight
'capabilities': supported operation types for this channel
'version': Intel(R) QuickData specfication revision
This also allows some chattiness to be removed from the driver startup
as this information is now available via sysfs.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Up until this point the driver for Intel(R) QuickData Technology
engines, specification versions 2 and 3, were mostly identical save for
a few quirks. Version 3.2 hardware adds many new capabilities (like
raid offload support) requiring some infrastructure that is not relevant
for v2. For better code organization of the new funcionality move v3
and v3.2 support to its own file dma_v3.c, and export some routines from
the base files (dma.c and dma_v2.c) that can be reused directly.
The first new capability included in this code reorganization is support
for v3.2 memset operations.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
ioat3.2 adds raid5 and raid6 offload capabilities.
Signed-off-by: Tom Picard <tom.s.picard@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In preparation for adding more operation types to the ioat3 path the
driver needs to honor the DMA_PREP_FENCE flag. For example the async_tx api
will hand xor->memcpy->xor chains to the driver with the 'fence' flag set on
the first xor and the memcpy operation. This flag in turn sets the 'fence'
flag in the descriptor control field telling the hardware that future
descriptors in the chain depend on the result of the current descriptor, so
wait for all writes to complete before starting the next operation.
Note that ioat1 does not prefetch the descriptor chain, so does not
require/support fenced operations.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Increment the allocation order of the descriptor ring every time we run
out of descriptors up to a maximum of allocation order specified by the
module parameter 'ioat_max_alloc_order'. After each idle period
decrement the allocation order to a minimum order of
'ioat_ring_alloc_order' (i.e. the default ring size, tunable as a module
parameter).
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In order to support dynamic resizing of the descriptor ring or polling
for a descriptor in the presence of a hung channel the reset handler
needs to make progress while in a non-preemptible context. The current
workqueue implementation precludes polling channel reset completion
under spin_lock().
This conversion also allows us to return to opportunistic cleanup in the
ioat2 case as the timer implementation guarantees at least one cleanup
after every descriptor is submitted. This means the worst case
completion latency becomes the timer frequency (for exceptional
circumstances), but with the benefit of avoiding busy waiting when the
lock is contended.
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Save 4 bytes per software descriptor by transmitting tx_cnt in an unused
portion of the hardware descriptor.
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Mark all single use initialization routines with __devinit.
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The register write in ioat_dma_cleanup_tasklet is unfortunate in two
ways:
1/ It clears the extra 'enable' bits that we set at alloc_chan_resources time
2/ It gives the impression that it disables interrupts when it is in
fact re-arming interrupts
[ Impact: fix, persist the value of the chanctrl register when re-arming ]
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
