It's easier to read code with fewer levels of indentation, remove a level
of indentation in at_xdmac_advance_work()
if (!foo() & !bar()) {
}
was replaced by:
if (foo() || bar())
return;
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-9-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Since tx_submit can be called from a hard IRQ, xfers_list must be
protected with a lock to avoid concurency on the list's elements.
Since at_xdmac_handle_cyclic() is called from a tasklet, spin_lock_irq
is enough to protect from a hard IRQ.
Fixes: e1f7c9eee7 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-8-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Move the free desc to the tail of the list, so that the sequence of
descriptors is more track-able in case of debug. One would know which
descriptor should come next and could easier catch concurrency over
descriptors for example. virt-dma uses list_splice_tail_init() as well,
follow the core driver.
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-7-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The transfer descriptors were wrongly moved to the free descriptors list
before calling the tx desc callback. As the DMA engine drivers drop any
locks before calling the callback function, txd could be taken again,
resulting in its callback called prematurely. Fix the race for the tx desc
callback by moving the xfer desc into the free desc list after the
callback is invoked.
Fixes: e1f7c9eee7 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-6-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Caller of dma_cookie_complete is expected to hold a lock to prevent
concurrency over the channel's completed cookie marker. Call
dma_cookie_complete() with the lock held.
Fixes: e1f7c9eee7 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-5-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
It is desirable to do the prints without the lock held if possible, so
move the print after the lock is released.
Fixes: e1f7c9eee7 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-4-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Cyclic channels must too call issue_pending in order to start a transfer.
Start the transfer in issue_pending regardless of the type of channel.
This wrongly worked before, because in the past the transfer was started
at tx_submit level when only a desc in the transfer list.
Fixes: e1f7c9eee7 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-3-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
tx_submit is supposed to push the current transaction descriptor to a
pending queue, waiting for issue_pending() to be called. issue_pending()
must start the transfer, not tx_submit(), thus remove
at_xdmac_start_xfer() from at_xdmac_tx_submit(). Clients of at_xdmac that
assume that tx_submit() starts the transfer must be updated and call
dma_async_issue_pending() if they miss to call it (one example is
atmel_serial).
As the at_xdmac_start_xfer() is now called only from
at_xdmac_advance_work() when !at_xdmac_chan_is_enabled(), the
at_xdmac_chan_is_enabled() check is no longer needed in
at_xdmac_start_xfer(), thus remove it.
Fixes: e1f7c9eee7 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211215110115.191749-2-tudor.ambarus@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The following sysfs attributes will be obsolete due to the name change of
tokens to read buffers:
max_tokens
token_limit
group/tokens_allowed
group/tokens_reserved
group/use_token_limit
Create new entries and have old entry print warning of deprecation.
New attributes to replace the token ones:
max_read_buffers
read_buffer_limit
group/read_buffers_allowed
group/read_buffers_reserved
group/use_read_buffer_limit
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163951339488.2988321.2424012059911316373.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
DSA spec v1.2 has changed the term of "bandwidth tokens" to "read buffers"
in order to make the concept clearer. Deprecate bandwidth token
naming in the driver and convert to read buffers in order to match with
the spec and reduce confusion when reading the spec.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163951338932.2988321.6162640806935567317.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
By the spec, wq size and group association is not changeable unless device
is disabled. Exclude clearing the shadow copy on wq disable/reset. This
allows wq type to be changed after disable to be re-enabled.
Move the size and group association to its own cleanup and only call it
during device disable.
Fixes: 0dcfe41e9a ("dmanegine: idxd: cleanup all device related bits after disabling device")
Reported-by: Lucas Van <lucas.van@intel.com>
Tested-by: Lucas Van <lucas.van@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163951291732.2987775.13576571320501115257.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Change the driver where WQ interrupt is requested only when wq is being
enabled. This new scheme set things up so that request_threaded_irq() is
only called when a kernel wq type is being enabled. This also sets up for
future interrupt request where different interrupt handler such as wq
occupancy interrupt can be setup instead of the wq completion interrupt.
Not calling request_irq() until the WQ actually needs an irq also prevents
wasting of CPU irq vectors on x86 systems, which is a limited resource.
idxd_flush_pending_descs() is moved to device.c since descriptor flushing
is now part of wq disable rather than shutdown().
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163942149487.2412839.6691222855803875848.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The descriptor flushing for shutdown is not holding the irq_entry list
lock. If there's ongoing interrupt completion handling, this can corrupt
the list. Add locking to protect list walking. Also refactor the code so
it's more compact.
Fixes: 8f47d1a5e5 ("dmaengine: idxd: connect idxd to dmaengine subsystem")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163942148935.2412839.18282664745572777280.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
With irq_entry already being associated with the wq in a 1:1 relationship,
embed the irq_entry in the idxd_wq struct and remove back pointers for
idxe_wq and idxd_device. In the process of this work, clean up the interrupt
handle assignment so that there's no decision to be made during submit
call on where interrupt handle value comes from. Set the interrupt handle
during irq request initialization time.
irq_entry 0 is designated as special and is tied to the device itself.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163942148362.2412839.12055447853311267866.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
There are currently 2 ways to create a set of sysfs files for a
kobj_type, through the default_attrs field, and the default_groups
field. Move the ioatdma sysfs code to use default_groups field which has
been the preferred way since aa30f47cf6 ("kobject: Add support for
default attribute groups to kobj_type") so that we can soon get rid of
the obsolete default_attrs field.
Cc: Vinod Koul <vkoul@kernel.org>
Cc: dmaengine@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20220104163330.1338824-1-gregkh@linuxfoundation.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The variables src_addr and dst_addr handle DMA addresses, so these should
be declared as dma_addr_t.
Fixes: 667b925144 ("dmaengine: uniphier-xdmac: Add UniPhier external DMA controller driver")
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Link: https://lore.kernel.org/r/1639456963-10232-1-git-send-email-hayashi.kunihiko@socionext.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add support for R-Car S4-8. We can reuse R-Car V3U code so that
renames variable names as "gen4".
Note that some registers of R-Car V3U do not exist on R-Car S4-8,
but none of them are used by the driver for now.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Ulrich Hecht <uli+renesas@fpond.eu>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20211222114507.1252947-3-yoshihiro.shimoda.uh@renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Tag for dmaengine slave_id removal topic branch which should be merged
into v5.17
-----BEGIN PGP SIGNATURE-----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=Mpf8
-----END PGP SIGNATURE-----
Merge tag 'dmaengine_topic_slave_id_removal_5.17' into next
Merge the tag dmaengine_topic_slave_id_removal_5.17 into next. This
brings in the slave_id removal topic changes
'shdma_slave_used' is a bitmap. So use 'bitmap_zalloc()' to simplify code,
improve the semantic and avoid some open-coded arithmetic in allocator
arguments.
Also change the corresponding 'kfree()' into 'bitmap_free()' to keep
consistency.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/3efaf2784424ae3d7411dc47f8b6b03e7bb8c059.1637702701.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The pointer hwdesc is being initialized with a value that is never
read, it is being updated later in a for-loop. The assignment is
redundant and can be removed.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Reviewed-by: Amelie Delaunay <amelie.delaunay@foss.st.com>
Link: https://lore.kernel.org/r/20211204140032.548066-1-colin.i.king@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add a sysfs knob to allow tuning of retries for the kernel ENQCMDS
descriptor submission. While on host, it is not as likely that ENQCMDS
return busy during normal operations due to the driver controlling the
number of descriptors allocated for submission. However, when the driver is
operating as a guest driver, the chance of retry goes up significantly due
to sharing a wq with multiple VMs. A default value is provided with the
system admin being able to tune the value on a per WQ basis.
Suggested-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163820629464.2702134.7577370098568297574.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
For some devices with only half-duplex capabilities, it doesn't make
much sense to use one DMA channel per direction, as both channels will
never be active at the same time.
Add support for bidirectional I/O on DMA channels. The client drivers
can then request a "tx-rx" DMA channel which will be used for both
directions.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Link: https://lore.kernel.org/r/20211206174259.68133-7-paul@crapouillou.net
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The JZ4760 and JZ4760B SoCs have two regular DMA controllers with 6
channels each. They also have an extra DMA controller named MDMA
with only 2 channels, that only supports memcpy operations, and one
named BDMA with only 3 channels, that is mostly used for transfers
between memories and the BCH controller.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Link: https://lore.kernel.org/r/20211206174259.68133-5-paul@crapouillou.net
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The JZ4760 SoC has a hardware problem with chan0 not enabling properly
if it's enabled before chan1, after a reset (works fine afterwards).
This is worked around in the probe function by just enabling then
disabling chan1.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Link: https://lore.kernel.org/r/20211206174259.68133-4-paul@crapouillou.net
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add SYSFW defined rchan_oes_offset number for J721S2 SoC in soc data.
Signed-off-by: Aswath Govindraju <a-govindraju@ti.com>
Acked-by: Peter Ujfalusi <peter.ujfalusi@gmail.com>
Link: https://lore.kernel.org/r/20211119132315.15901-2-a-govindraju@ti.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Make use of the struct_size() helper instead of an open-coded version, in
order to avoid any potential type mistakes or integer overflows that, in
the worst scenario, could lead to heap overflows.
Link: https://github.com/KSPP/linux/issues/160
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20211208001013.GA62330@embeddedor
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Handle errors when trying to map the IRQ for the DMA channels.
The main motivation here is to be able to handle probe deferral. E.g. when
using DT overlays it is possible that the DMA controller is probed before
interrupt controller, depending on the order in the DT.
In order to support this switch from irq_of_parse_and_map() to
of_irq_get(), which internally does the same, but it will return
EPROBE_DEFER when the interrupt controller is not yet available.
As a result other errors, such as an invalid IRQ specification, or missing
IRQ are also properly handled.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Link: https://lore.kernel.org/r/20211208114212.234130-1-lars@metafoo.de
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The display driver wants to pass a custom flag to the DMA engine driver,
which it started doing by using the slave_id field that was traditionally
used for a different purpose.
As there is no longer a correct use for the slave_id field, it should
really be removed, and the remaining users changed over to something
different.
The new mechanism for passing nonstandard settings is using the
.peripheral_config field, so use that to pass a newly defined structure
here, making it clear that this will not work in portable drivers.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211122222203.4103644-10-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The slave_id was previously used to pick one DMA slave instead of another,
but this is now done through the DMA descriptors in device tree.
For the qcom_adm driver, the configuration is documented in the DT
binding to contain a tuple of device identifier and a "crci" field,
but the implementation ends up using only a single cell for identifying
the slave, with the crci getting passed in nonstandard properties of
the device, and passed through the dma driver using the old slave_id
field. Part of the problem apparently is that the nand driver ends up
using only a single DMA request ID, but requires distinct values for
"crci" depending on the type of transfer.
Change both the dmaengine driver and the two slave drivers to allow
the documented binding to work in addition to the ad-hoc passing
of crci values. In order to no longer abuse the slave_id field, pass
the data using the "peripheral_config" mechanism instead.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211122222203.4103644-9-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
It appears that the code that reads the slave_id from the channel config
was copied incorrectly from other drivers. Nothing ever sets this field
on platforms that use this driver, so remove the reference.
Reviewed-by: Baolin Wang <baolin.wang7@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211122222203.4103644-8-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The last driver referencing the slave_id on Marvell PXA and MMP platforms
was the SPI driver, but this stopped doing so a long time ago, so the
TODO from the earlier patch can no be removed.
Fixes: b729bf3453 ("spi/pxa2xx: Don't use slave_id of dma_slave_config")
Fixes: 13b3006b8e ("dma: mmp_pdma: add filter function")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211122222203.4103644-7-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The slave device is picked through either devicetree or a filter
function, and any remaining out-of-tree drivers would have warned
about this usage since 2015.
Stop interpreting the field finally so it can be removed from
the interface.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211122222203.4103644-6-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Nothing sets the slave_id field any more, so stop accessing
it to allow the removal of this field.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211122222203.4103644-11-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Ming reported that with the abort path of the descriptor submission, there
can be a window where a completed descriptor can be missed to be completed
by the irq completion thread:
CPU A CPU B
Submit (successful)
Submit (fail)
irq_process_work_list() // empty
llist_abort_desc()
// remove all descs from pending list
irq_process_pending_llist() // empty
exit idxd_wq_thread() with no processing
Add opportunistic descriptor completion in the abort path in order to
remove the missed completion.
Fixes: 6b4b87f2c3 ("dmaengine: idxd: fix submission race window")
Reported-by: Ming Li <ming4.li@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163898288714.443911.16084982766671976640.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Smatch reports below warnings [1] wrt dereferencing rm_res when it can
potentially be ERR_PTR(). This is possible when entire range is
allocated to Linux
Fix this case by making sure, there is no deference of rm_res when its
ERR_PTR().
[1]:
drivers/dma/ti/k3-udma.c:4524 udma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
drivers/dma/ti/k3-udma.c:4537 udma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
drivers/dma/ti/k3-udma.c:4681 bcdma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
drivers/dma/ti/k3-udma.c:4696 bcdma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
drivers/dma/ti/k3-udma.c:4711 bcdma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
drivers/dma/ti/k3-udma.c:4848 pktdma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
drivers/dma/ti/k3-udma.c:4861 pktdma_setup_resources() error: 'rm_res' dereferencing possible ERR_PTR()
Reported-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Acked-by: Peter Ujfalusi <peter.ujfalusi@gmail.com>
Link: https://lore.kernel.org/r/20211209180957.29036-1-vigneshr@ti.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The variable used for returning status in
`ppc440spe_adma_dma2rxor_prep_src' function is never changed
and this function just need to return 0. Thus, the `rval' can
be removed and return 0 from `ppc440spe_adma_dma2rxor_prep_src'.
Signed-off-by: Jason Wang <wangborong@cdjrlc.com>
Link: https://lore.kernel.org/r/20211114060856.239314-1-wangborong@cdjrlc.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Dan reports that smatch has found idxd_wq_quiesce() is being called inside
the idxd->dev_lock. idxd_wq_quiesce() calls wait_for_completion() and
therefore it can sleep. Move the call outside of the spinlock as it does
not need device lock.
Fixes: 5b0c68c473 ("dmaengine: idxd: support reporting of halt interrupt")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163716858508.1721911.15051495873516709923.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The commit in the Fixes: tag has changed the logic of the code and now it
is likely that the probe will return an early success (0), even if not
completely executed.
This should lead to a crash or similar issue later on when the code
accesses to some never allocated resources.
Change the '!err' into a 'err' when checking if
'dma_set_mask_and_coherent()' has failed or not.
While at it, simplify the code and remove the "can't success code" related
to 32 DMA mask.
As stated in [1], 'dma_set_mask_and_coherent(DMA_BIT_MASK(64))' can't fail
if 'dev->dma_mask' is non-NULL. And if it is NULL, it would fail for the
same reason when tried with DMA_BIT_MASK(32).
[1]: https://lkml.org/lkml/2021/6/7/398
Fixes: ecb8c88bd3 ("dmaengine: dw-edma-pcie: switch from 'pci_' to 'dma_' API")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/935fbb40ae930c5fe87482a41dcb73abf2257973.1636492127.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Vinod Koul <vkoul@kernel.org>
This new CDMA binding for device_prep_dma_memcpy_sg was partially borrowed
from xlnx kernel tree, an expanded with extended address space support when
linking descriptor segments and checking for incorrect zero transfer size.
Signed-off-by: Adrian Larumbe <adrianml@alumnos.upm.es>
Link: https://lore.kernel.org/r/20211101180825.241048-4-adrianml@alumnos.upm.es
Signed-off-by: Vinod Koul <vkoul@kernel.org>
This is the old DMA_SG interface that was removed in commit
c678fa6634 ("dmaengine: remove DMA_SG as it is dead code in kernel"). It
has been renamed to DMA_MEMCPY_SG to better match the MEMSET and MEMSET_SG
naming convention.
It should only be used for mem2mem copies, either main system memory or
CPU-addressable device memory (like video memory on a PCI graphics card).
Bringing back this interface was prompted by the need to use the Xilinx
CDMA device for mem2mem SG transfers.
Signed-off-by: Adrian Larumbe <adrianml@alumnos.upm.es>
Link: https://lore.kernel.org/r/20211101180825.241048-3-adrianml@alumnos.upm.es
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Coverity complains of an uninitialized variable:
5. uninit_use_in_call: Using uninitialized value config.dst_per when calling axi_chan_config_write. [show details]
6. uninit_use_in_call: Using uninitialized value config.hs_sel_src when calling axi_chan_config_write. [show details]
CID 121164 (#1-3 of 3): Uninitialized scalar variable (UNINIT)
7. uninit_use_in_call: Using uninitialized value config.src_per when calling axi_chan_config_write. [show details]
418 axi_chan_config_write(chan, &config);
Fix this by initializing the structure to 0 which should at least be benign in axi_chan_config_write(). Also fix
what looks like a cut-n-paste error when initializing config.hs_sel_dst.
Fixes: 824351668a ("dmaengine: dw-axi-dmac: support DMAX_NUM_CHANNELS > 8")
Cc: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Cc: Vinod Koul <vkoul@kernel.org>
Cc: dmaengine@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
Link: https://lore.kernel.org/r/20211025181656.31658-1-tim.gardner@canonical.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
"Interrupt handle revoked" is an event that happens when the driver is
running on a guest kernel and the VM is migrated to a new machine.
The device will trigger an interrupt that signals to the guest driver
that the interrupt handles need to be replaced.
The misc irq thread function calls a helper function to handle the
event. The function uses the WQ percpu_ref to quiesce the kernel
submissions. It then replaces the interrupt handles by requesting
interrupt handle command for each I/O MSIX vector. Once the handle is
updated, the driver will unblock the submission path to allow new
submissions.
The submitter will attempt to acquire a percpu_ref before submission. When
the request fails, it will wait on the wq_resurrect 'completion'.
The driver does anticipate the possibility of descriptors being submitted
before the WQ percpu_ref is killed. If a descriptor has already been
submitted, it will return with incorrect interrupt handle status. The
descriptor will be re-submitted with the new interrupt handle on the
completion path. For descriptors with incorrect interrupt handles,
completion interrupt won't be triggered.
At the completion of the interrupt handle refresh, the handling function
will call idxd_int_handle_refresh_drain() to issue drain descriptors to
each of the wq with associated interrupt handle. The drain descriptor will have
interrupt request set but without completion record. This will ensure all
descriptors with incorrect interrupt completion handle get drained and
a completion interrupt is triggered for the guest driver to process them.
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Co-Developed-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163528420189.3925689.18212568593220415551.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
