PL080S has separate register to store transfer size in, allowing single
transfer to be much larger than in standard PL080.
This patch makes the amba-pl08x driver aware of this and removes writing
transfer size to reserved bits of CH_CONTROL register on PL080S, which
was not a problem witn transfer sizes fitting the original bitfield
of PL080, but now would overwrite other fields.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
PL080S is a modified version of PL080 that can be found on Samsung SoCs,
such as S3C6400 and S3C6410.
It has different offset of CONFIG register, separate CONTROL1 register
that holds transfer size and larger maximum transfer size.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This patch refactors debugging code that dumps LLI entries by moving it
into separate function, which is stubbed when VERBOSE_DEBUG is not
selected. This allows us to get rid of the ugly ifdef from the body of
pl08x_fill_llis_for_desc().
No functional change is introduced by this patch.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Currently memory allocated for LLIs is casted to an array of structs,
which is fragile and also limits the driver to a single, predefined LLI
layout, while there are some variants of PL08x, which have more fields
in LLI (namely PL080S with its extra CCTL2).
This patch makes LLIs a sequence of 32-bit words, which is just filled
with appropriate values in appropriate order and padded with required
amount of dummy words (currently zero, but PL080S will make better use
of this).
Suggested-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Some variants of PL08x (namely PL080S, found in Samsung S3C64xx SoCs)
have CONFIG register at different offset. This patch makes the driver
use offset from vendor data struct.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Further patch will introduce support for PL080S, which requires some
things to be done conditionally, thus increasing indentation level of
some functions even more.
This patch reduces indentation level of pl08x_getbytes_chan() function
by inverting several conditions and returning from function wherever
possible.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
The test here should be ">=" instead of ">". The cdd->chan_busy[] array
has "ALLOC_DECS_NUM" elements.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The pl330 DMA driver is broken in regard to handling a terminate all request
while it is processing the list of completed descriptors. This is most visible
when calling dmaengine_terminate_all() from within the descriptors callback for
cyclic transfers. In this case the TERMINATE_ALL transfer will clear the
work_list and stop the transfer. But after all callbacks for all completed
descriptors have been handled the descriptors will be re-enqueued into the (now
empty) work_list. So the next time dma_async_issue_pending() is called for the
channel these descriptors will be transferred again which will cause data
corruption. Similar issues can occur if dmaengine_terminate_all() is not called
from within the descriptor callback but runs on a different CPU at the same time
as the completed descriptor list is processed.
This patch introduces a new per channel list which will hold the completed
descriptors. While processing the list the channel's lock will be held to avoid
racing against dmaengine_terminate_all(). The lock will be released when calling
the descriptors callback though. Since the list of completed descriptors might
be modified (e.g. by calling dmaengine_terminate_all() from the callback) we can
not use the normal list iterator macros. Instead we'll need to check for each
loop iteration again if there are still items in the list. The drivers
TERMINATE_ALL implementation is updated to move descriptors from both the
work_list as well the new completed_list back to the descriptor pool. This makes
sure that none of the descripts finds its way back into the work list and also
that we do not call any futher complete callbacks after
dmaengine_terminate_all() has been called.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
With enabled pm_runtime in the kernel the device won't work because it
is not "on" during the probe function. This patch enables the device via
pm_runtime on probe so it remains activated.
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Before Randy figures out that this does not compile with CONFIG_BUG=n
here is a fix for it.
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
A bad merge resulted in a left-over free_irq() call. This patch removes it.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This configuration data will be used, when DMAC DT support is added to
r8a73a4.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
All shdma DMACs on ARM SoCs share certain register layout patterns, which
are currently defined in arch/arm/mach-shmobile/include/mach/dma-register.h.
That header is included by SoC-specific setup-*.c files to be used in DMAC
platform data. That header, however, cannot be directly used by the driver.
This patch copies those defines into a driver-local header to be used by
Device Tree configurations.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Macros, named like TEND or SAR lack a namespace and are too broadly named
for a global header. Besides, they aren't needed globally. Move them to
where they belong - into the driver. Some other macros aren't used at all,
remove them.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This facilitates DMAC DT support by eliminating the need in AUXDATA and
avoiding creating complex DT data. This also fits well with DMAC devices,
of which SoCs often have multiple identical copies and it is perfectly
valid to use a single configuration data set for all of them.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Platform data shouldn't be changed at run-time, so, pointers to it should
be const.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Switch shdma to using devm_* managed functions for allocation of memory,
requesting IRQs, mapping IO resources etc.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
On newer r-car SoCs the CHCLR register only contains one bit per channel,
to which a 1 has to be written to reset the channel. Older SoC versions had
one CHCLR register per channel, to which a 0 must be written to reset the
channel and clear its buffers. This patch adds support for the newer
layout.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This adds the ROM script addresses for i.MX25, i.MX5x and i.MX6 to the
SDMA driver needed for the driver to work without additional firmware.
The ROM script addresses are SoC specific and in some cases even tapeout
specific. This patch adds the ROM script addresses only for SoCs which
do not have a tapeout specific SDMA ROM, because currently it's unclear
how this case should be handled.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Use a struct type instead of an enum type for distinguishing between
different versions. This makes it simpler to handle multiple differences
without cluttering the code with comparisons for certain devtypes.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
As the driver now has its own xlate function and makes use of the
dma_get_slave_channel(), we need to manually set the DMA_PRIVATE flags.
Drivers which rely on of_dma_simple_xlate() do implicitly the same by
going through __dma_request_channel().
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Provide a callback to prepare cyclic DMA transfers.
This is for instance needed for audio channel transport.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
In order to fully support multiple transactions per channel, we need to
assure we get an interrupt for each completed transaction. That flags
bit is also our only way to tell at which descriptor a transaction ends.
So, remove the manual clearing of that bit, and then inline the only
remaining command that is left in append_pending_queue() for better
readability.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Currently, when an interrupt has occured for a channel, the tasklet
worker code will only look at the very last entry in the running list
and complete its cookie, and then dispose the entire running chain.
Hence, the first transaction's cookie will never complete.
In fact, the interrupt we should handle will be the one related to the
first descriptor in the chain with the ENDIRQEN bit set, so complete
the second transaction that is in fact still running.
As a result, the driver can't currently handle multiple transactions on
one chanel, and it's likely that no drivers exist that rely on this
feature.
Fix this by walking the running_chain and look for the first
descriptor that has the interrupt-enable bit set. Only queue
descriptors up to that point for completion handling, while leaving
the rest intact. Also, only make the channel idle if the list is
completely empty after such a cycle.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
In case of big endian CPU we have to convert either all fields in the structure
or leave this job to ACPICA. The second choice seems the best.
So, let's remove the ugly conversion that is not fully comprehensive anyway.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
If "num_disabled" is equal to STEDMA40_MAX_PHYS (32) then we would write
one space beyond the end of the pdata->disable_channels[] array.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
When CONFIG_ARM_LPAE=y the following build warning are generated:
drivers/dma/ste_dma40.c:3228:2: warning: format '%x' expects argument of type 'unsigned int', but argument 4 has type 'resource_size_t' [-Wformat]
drivers/dma/ste_dma40.c:3582:3: warning: format '%x' expects argument of type 'unsigned int', but argument 4 has type 'resource_size_t' [-Wformat]
drivers/dma/ste_dma40.c:3582:3: warning: format '%x' expects argument of type 'unsigned int', but argument 5 has type 'resource_size_t' [-Wformat]
drivers/dma/ste_dma40.c:3593:5: warning: format '%x' expects argument of type 'unsigned int', but argument 5 has type 'resource_size_t' [-Wformat]
According to Documentation/printk-formats.txt '%pa' can be used to properly
print 'resource_size_t'.
Also, for printing memory region the '%pr' is more convenient.
Reported-by: Kevin Hilman <khilman@linaro.org>
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Kevin Hilman <khilman@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Fix:
arch/arm/common/built-in.o: undefined reference to `edma_filter_fn'
seen with "make ARCH=arm allmodconfig"
Commit 6cba4355 (ARM: edma: Add DT and runtime PM support to the private EDMA
API) adds a dependency on edma_filter_fn() into arch/arm/common/edma.c. Since
this file is always built into the kernel, edma_filter_fn() must be built into
the kernel as well.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
In the shdma driver __iomem pointers are used to point to hardware
registers. Using typed pointers like "u32 __iomem *" in this case is
inconvenient, because then offsets, added to such pointers, have to be
devided by sizeof(u32) or similar. Switch the driver to use void
pointers, which avoids this clumsiness.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
struct sh_dmae_device::chan_reg is a pointer to u32, therefore when adding
offsets to it care should be taken to add offsets in sizeof(u32) units, not
in bytes. This patch corrects such a bug. While at it we also remove the
redundant parameter of the affected function.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Building dma_v3.o triggers a GCC warning:
drivers/dma/ioat/dma_v3.c: In function ‘__ioat3_prep_pq16_lock’:
drivers/dma/ioat/dma_v3.c:264:11: warning: array subscript is below array bounds [-Warray-bounds]
drivers/dma/ioat/dma_v3.c:264:11: warning: array subscript is below array bounds [-Warray-bounds]
This warning is caused by pq16_set_src(). It uses "int idx" as an index
to an eight element array. Changing "idx" to "unsigned" silences this
warning. Apparently GCC can then determine that "idx" will never be
negative.
Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Acked-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <djbw@fb.com>
dma_channel_rebalance() currently distributes channels by processor ID.
These IDs often change with the BIOS, and the order isn't related to
the DMA channel list (related to PCI bus ids).
* On my SuperMicro dual E5 machine, first socket has processor IDs [0-7]
(and [16-23] for hyperthreads), second socket has [8-15]+[24-31]
=> channels are properly allocated to local CPUs.
* On Dells R720 with same processors, first socket has even processor IDs,
second socket has odd numbers
=> half the processors get channels on the remote socket, causing
cross-NUMA traffic and lower DMA performance.
Change nth_chan() to return the channel with min table_count and in the
NUMA node of the given CPU, if any. If none, the (non-local) channel with
min table_count is returned. nth_chan() is therefore renamed into min_chan()
since we don't iterate until the nth channel anymore. In practice, the
behavior is the same because first channels are taken first and are then
ignored because they got an additional reference.
The new code has a slightly higher complexity since we always scan the
entire list of channels for finding the minimal table_count (instead
of stopping after N chans), and because we check whether the CPU is in the
DMA device locality mask. Overall we still have time complexity =
number of chans x number of processors. This rebalance is rarely used,
so this won't hurt.
On the above SuperMicro machine, channels are still allocated the same.
On the Dells, there are no locality issue anymore (MEMCPY channel X goes
to processor X and to its hyperthread sibling).
Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
Signed-off-by: Dan Williams <djbw@fb.com>
Disable RAID on non-Atom platform and remove related fixups such as the
64-byte alignement restriction on legacy DMA operations (introduced in
commit f26df1a1 as a workaround for silicon errata).
Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
Acked-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Jon Mason <jon.mason@intel.com>
Signed-off-by: Dan Williams <djbw@fb.com>
The mv_xor driver had never been used in a big-endian context, and
therefore was not using the hardware features to support such an
execution environment. The hardware provides a "descriptor swap" bit
that automatically swaps the bytes of the DMA descriptors, within
blocks of 8 bytes. This requires a different DMA descriptor layout on
big-endian systems, as well as enabling this "descriptor swap" bit.
This mechanism is exactly identical to the one already used in the
mv643xx_eth network driver and the mvneta network driver.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Dan Williams <djbw@fb.com>
In order to support big-endian execution, the mv_xor driver is changed
to use the readl_relaxed() and writel_relaxed() accessors that
properly convert from the CPU endianess to the device endianess (which
in the case of Marvell XOR hardware is always little-endian).
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Dan Williams <djbw@fb.com>
Let's move the behaviour of printing no error message back to the pre v3.10
times. It means we will use debug level in the described case, and a warning
level otherwise.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Dan Williams <djbw@fb.com>
There is a really little chance when we are able to create a directory and are
not able to create nodes under it. So, this patch just removes those checks.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Dan Williams <djbw@fb.com>
The debugfs interface brought a copy of the test case parameters. This makes
different set of values under /sys/module/dmatest/parameters/ and
/sys/kernel/debug/dmatest/. The user might be confused by the divergence of
values.
The proposed solution in this patch is to make module parameters writable and
remove them from the debugfs. Though we're still using debugfs to control test
runner and getting results.
Documentation part is updated accordingly.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Suggested-by: Dan Williams <djbw@fb.com>
Signed-off-by: Dan Williams <djbw@fb.com>
In Rob's recent pull request the patch
ARM: highbank: select ARCH_DMA_ADDR_T_64BIT for LPAE
promotes dma_addr_t to 64bit, so printk generates a warning about
an incorrect type. Fix this by casting it to u64 and using %llx.
Fixing long lines on the way.
Signed-off-by: Andre Przywara <andre.przywara@linaro.org>
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
When dma_addr_t is 64 bits long, compilation of the AMBA PL08x DMA
driver breaks due to a missing 64bit%8bit modulo operation.
Looking more closely the divisor in these operations can only be
1, 2 or 4, so the full featured '%' modulo operation is overkill and
can be replaced with simple bit masking.
Change from v1:
Replace open-coded function with existing IS_ALIGNED macro and use a
macro around that to avoid a line becoming too long.
Signed-off-by: Andre Przywara <andre.przywara@linaro.org>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
With all mxs-dma clients moved to use generic DMA helper, the code
left from generic DMA binding conversion can be removed now.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
After the patch: "2ccaef0 dma: imx-sdma: make channel0 operations atomic",
the "done" completion is not used any more.
Just remove it.
Signed-off-by: Huang Shijie <b32955@freescale.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Remove unneeded error handling on the result of a call to
platform_get_resource when the value is passed to devm_ioremap_resource.
A simplified version of the semantic patch that makes this change is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@@
expression pdev,res,n,e,e1;
expression ret != 0;
identifier l;
@@
- res = platform_get_resource(pdev, IORESOURCE_MEM, n);
... when != res
- if (res == NULL) { ... \(goto l;\|return ret;\) }
... when != res
+ res = platform_get_resource(pdev, IORESOURCE_MEM, n);
e = devm_ioremap_resource(e1, res);
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
