In some configurations, the BAM DMA controller is set up by a remote
processor and the local processor can simply start making use of it
without setting up the BAM. This is already supported using the
"qcom,controlled-remotely" property.
However, for some reason another possible configuration is that the
remote processor is responsible for powering up the BAM, but we are
still responsible for initializing it (e.g. resetting it etc).
This configuration is quite challenging to handle properly because
the power control is handled through separate channels
(e.g. device-specific SMSM interrupts / smem-states). Great care
must be taken to ensure the BAM registers are not accessed while
the BAM is powered off since this results in a bus stall.
Attempt to support this configuration with minimal device-specific
code in the bam_dma driver by tracking the number of requested
channels. Consumers of DMA channels are responsible to only request
DMA channels when the BAM was powered on by the remote processor,
and to release them before the BAM is powered off.
When the first channel is requested the BAM is initialized (reset)
and it is also put into reset when the last channel was released.
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Reviewed-by: Bhupesh Sharma <bhupesh.sharma@linaro.org>
Link: https://lore.kernel.org/r/20211018102421.19848-3-stephan@gerhold.net
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Without CONFIG_PM_SLEEP, the runtime suspend/resume functions
are unused, producing a warning:
../drivers/dma/sa11x0-dma.c:1042:12: error: 'sa11x0_dma_resume' defined but not used
../drivers/dma/sa11x0-dma.c:1004:12: error: 'sa11x0_dma_suspend' defined but not used
Signed-off-by: Cai Huoqing <caihuoqing@baidu.com>
Link: https://lore.kernel.org/r/20211026020508.550-1-caihuoqing@baidu.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Qing Wang <wangqing@vivo.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Link: https://lore.kernel.org/r/1633663733-47199-7-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Qing Wang <wangqing@vivo.com>
Link: https://lore.kernel.org/r/1633663733-47199-3-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Qing Wang <wangqing@vivo.com>
Link: https://lore.kernel.org/r/1633663733-47199-4-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Qing Wang <wangqing@vivo.com>
Link: https://lore.kernel.org/r/1633663733-47199-5-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Qing Wang <wangqing@vivo.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/1633663733-47199-6-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Wang Qing <wangqing@vivo.com>
Link: https://lore.kernel.org/r/1633663733-47199-2-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Don't populate the read-only array ds_lut on the stack but instead it
static. Also makes the object code smaller by 163 bytes:
Before:
text data bss dec hex filename
23508 4796 0 28304 6e90 ./drivers/dma/sh/rz-dmac.o
After:
text data bss dec hex filename
23281 4860 0 28141 6ded ./drivers/dma/sh/rz-dmac.o
(gcc version 11.2.0)
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210915112038.12407-1-colin.king@canonical.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The issue happens in an error handling path. If
of_dma_controller_register() fails, the function simply prints error
messages and returns error code, without decrementing the reference
count of pdev->device incremented earlier by
dma_async_device_register(), which may result in refcount leaks.
Fix it by invoking dma_async_device_unregister() before returning the
error code.
Signed-off-by: Xin Xiong <xiongx18@fudan.edu.cn>
Signed-off-by: Xiyu Yang <xiyuyang19@fudan.edu.cn>
Signed-off-by: Xin Tan <tanxin.ctf@gmail.com>
Link: https://lore.kernel.org/r/20210911070533.3114-1-xiongx18@fudan.edu.cn
Signed-off-by: Vinod Koul <vkoul@kernel.org>
As noted in the "Deprecated Interfaces, Language Features, Attributes,
and Conventions" documentation [1], size calculations (especially
multiplication) should not be performed in memory allocator (or similar)
function arguments due to the risk of them overflowing. This could lead
to values wrapping around and a smaller allocation being made than the
caller was expecting. Using those allocations could lead to linear
overflows of heap memory and other misbehaviors.
So, use the purpose specific kcalloc() function instead of the argument
size * count in the kzalloc() function.
[1] https://www.kernel.org/doc/html/v5.14/process/deprecated.html#open-coded-arithmetic-in-allocator-arguments
Signed-off-by: Len Baker <len.baker@gmx.com>
Link: https://lore.kernel.org/r/20210904145813.5161-1-len.baker@gmx.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Modify the prototype from xilinx_dma_tx_descriptor to
xilinx_dma_alloc_tx_descriptor and xilinx_dma_channel_set_config
to xilinx_vdma_channel_set_config in API description to
fix below linux kernel-doc warnings.
drivers/dma/xilinx/xilinx_dma.c:800: warning: expecting
prototype for xilinx_dma_tx_descriptor(). Prototype was
for xilinx_dma_alloc_tx_descriptor() instead.
drivers/dma/xilinx/xilinx_dma.c:2471: warning: expecting
prototype for xilinx_dma_channel_set_config(). Prototype
was for xilinx_vdma_channel_set_config() instead.
Signed-off-by: Shravya Kumbham <shravya.kumbham@xilinx.com>
Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Link: https://lore.kernel.org/r/1631525316-2323-1-git-send-email-radhey.shyam.pandey@xilinx.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Use the helper macro SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() instead of the
verbose operators ".suspend_noirq /.resume_noirq/.freeze_noirq/
.thaw_noirq/.poweroff_noirq/.restore_noirq", because the
SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() is a nice helper macro that could
be brought in to make code a little clearer, a little more concise.
Signed-off-by: Cai Huoqing <caihuoqing@baidu.com>
Link: https://lore.kernel.org/r/20210828090117.1814-1-caihuoqing@baidu.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Device reset clears the MSIXPERM table and the device registers. Re-program
the MSIXPERM table and re-enable the error interrupts post reset.
Fixes: 745e92a6d8 ("dmaengine: idxd: idxd: move remove() bits for idxd 'struct device' to device.c")
Reported-by: Sanjay Kumar <sanjay.k.kumar@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/163054188513.2853562.12077053294595278181.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Fix following panic on system halt:
Requesting system halt
[ 10.600000] spi spi0.1: spi_device 0.1 cleanup
[ 10.630000] fsl_edma_chan_mux() fsl_chan->edma->n_chans 64 dmamux_nr 0
[ 10.630000] *** ZERO DIVIDE *** FORMAT=4
[ 10.630000] Current process id is 38
[ 10.630000] BAD KERNEL TRAP: 00000000
[ 10.630000] PC: [<402f09ba>] fsl_edma_chan_mux+0x7c/0x12e
...
Some architecture as mcf5441x (ColdFire) may not have
a dmamux, so dmamux_nr is set to 0. This patch considers this case.
Signed-off-by: Angelo Dureghello <angelo.dureghello@timesys.com>
Link: https://lore.kernel.org/r/20210901211610.662077-1-angelo.dureghello@timesys.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Using list_move_tail() instead of list_del() + list_add_tail()
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Bixuan Cui <cuibixuan@huawei.com>
Acked-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20210908092826.67765-1-cuibixuan@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The ptdma driver has added debugfs support, but this fails to build
when debugfs is disabled:
drivers/dma/ptdma/ptdma-debugfs.c: In function 'ptdma_debugfs_setup':
drivers/dma/ptdma/ptdma-debugfs.c:93:54: error: 'struct dma_device' has no member named 'dbg_dev_root'
93 | debugfs_create_file("info", 0400, pt->dma_dev.dbg_dev_root, pt,
| ^
drivers/dma/ptdma/ptdma-debugfs.c:96:55: error: 'struct dma_device' has no member named 'dbg_dev_root'
96 | debugfs_create_file("stats", 0400, pt->dma_dev.dbg_dev_root, pt,
| ^
drivers/dma/ptdma/ptdma-debugfs.c:102:52: error: 'struct dma_device' has no member named 'dbg_dev_root'
102 | debugfs_create_dir("q", pt->dma_dev.dbg_dev_root);
| ^
Remove the #ifdef in the header, as this only saves a few bytes,
but would require ugly #ifdefs in each driver using it.
Simplify the other user while we're at it.
Fixes: e2fb2e2a33 ("dmaengine: ptdma: Add debugfs entries for PTDMA")
Fixes: 26cf132de6 ("dmaengine: Create debug directories for DMA devices")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://lore.kernel.org/r/20210920122017.205975-1-arnd@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Since pm_runtime_put is done when tegra_adma_probe is successful, we
cannot do pm_runtime_put_sync again in tegra_adma_remove.
Fix this by removing the pm_runtime_put_sync in tegra_adma_remove.
Signed-off-by: Dongliang Mu <mudongliangabcd@gmail.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Link: https://lore.kernel.org/r/20211021031432.3466261-1-mudongliangabcd@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The previous commit 059e969c2a ("dmaengine: tegra210-adma: Using
pm_runtime_resume_and_get to replace open coding") forgets to replace
the pm_runtime_get_sync in the tegra_adma_probe, but removes the
pm_runtime_put_noidle.
Fix this by continuing to replace pm_runtime_get_sync with
pm_runtime_resume_and_get in tegra_adma_probe.
Fixes: 059e969c2a ("dmaengine: tegra210-adma: Using pm_runtime_resume_and_get to replace open coding")
Signed-off-by: Dongliang Mu <mudongliangabcd@gmail.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Link: https://lore.kernel.org/r/20211021030538.3465287-1-mudongliangabcd@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
In rcar_dmac_probe, if pm_runtime_resume_and_get fails, it forgets to
disable runtime PM. And of_dma_controller_free should only be invoked
after the success of of_dma_controller_register.
Fix this by refactoring the error handling code.
Signed-off-by: Dongliang Mu <mudongliangabcd@gmail.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20211020143546.3436205-1-mudongliangabcd@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Before the `callback_result` callback was introduced drivers coded their
invocation to the callback in a similar way to:
if (cb->callback) {
spin_unlock(&dma->lock);
cb->callback(cb->callback_param);
spin_lock(&dma->lock);
}
With the introduction of `callback_result` two helpers where introduced to
transparently handle both types of callbacks. And drivers where updated to
look like this:
if (dmaengine_desc_callback_valid(cb)) {
spin_unlock(&dma->lock);
dmaengine_desc_callback_invoke(cb, ...);
spin_lock(&dma->lock);
}
dmaengine_desc_callback_invoke() correctly handles both `callback_result`
and `callback`. But we forgot to update the dmaengine_desc_callback_valid()
function to check for `callback_result`. As a result DMA descriptors that
use the `callback_result` rather than `callback` don't have their callback
invoked by drivers that follow the pattern above.
Fix this by checking for both `callback` and `callback_result` in
dmaengine_desc_callback_valid().
Fixes: f067025bc6 ("dmaengine: add support to provide error result from a DMA transation")
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Acked-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20211023134101.28042-1-lars@metafoo.de
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add support for setting dma coherent mask, dma mask is set to 64 bit
Signed-off-by: Pandith N <pandith.n@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20211001140812.24977-4-pandith.n@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Added hardware handshake selection in channel config,
for mem2per and per2mem case.
The peripheral specific handshake interface needs to be
programmed in src_per, dst_per bits of CHx_CFG register.
Signed-off-by: Pandith N <pandith.n@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20211001140812.24977-3-pandith.n@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Added support for DMA controller with more than 8 channels.
DMAC register map changes based on number of channels.
Enabling DMAC channel:
DMAC_CHENREG has to be used when number of channels <= 8
DMAC_CHENREG2 has to be used when number of channels > 8
Configuring DMA channel:
CHx_CFG has to be used when number of channels <= 8
CHx_CFG2 has to be used when number of channels > 8
Suspending and resuming channel:
DMAC_CHENREG has to be used when number of channels <= 8 DMAC_CHSUSPREG
has to be used for suspending a channel > 8
Signed-off-by: Pandith N <pandith.n@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20211001140812.24977-2-pandith.n@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Theorically, address pointers used by STM32 DMA must be chosen so as to
ensure that all transfers within a burst block are aligned on the address
boundary equal to the size of the transfer.
If this is always the case for peripheral addresses on STM32, it is not for
memory addresses if the user doesn't respect this alignment constraint.
To avoid a weird behavior of the DMA controller in this case (no error
triggered but data are not transferred as expected), force no burst.
Signed-off-by: Amelie Delaunay <amelie.delaunay@foss.st.com>
Link: https://lore.kernel.org/r/20211011094259.315023-4-amelie.delaunay@foss.st.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
buf_addr parameter of stm32_dma_set_xfer_param function is a dma_addr_t.
We only need to check the remainder of buf_addr/max_width, so, no need to
use do_div and extra u64 addr. Use '%' instead.
Fixes: e0ebdbdcb4 ("dmaengine: stm32-dma: take address into account when computing max width")
Signed-off-by: Amelie Delaunay <amelie.delaunay@foss.st.com>
Link: https://lore.kernel.org/r/20211011094259.315023-3-amelie.delaunay@foss.st.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
To prevent accidental repeated completion, mark pending descriptor
complete in terminate_all. It can be the case when terminate_all is called
while no end of transfer interrupt occurs.
Signed-off-by: Amelie Delaunay <amelie.delaunay@foss.st.com>
Link: https://lore.kernel.org/r/20211011094259.315023-2-amelie.delaunay@foss.st.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Total amount of SG list entries executed in a single burst is limited by
the number of available DMA descriptors.
This information is useful for device drivers utilizing this DMA engine.
Signed-off-by: Artur Rojek <contact@artur-rojek.eu>
Acked-by: Paul Cercueil <paul@crapouillou.net>
Link: https://lore.kernel.org/r/20210829195805.148964-1-contact@artur-rojek.eu
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Currently, DMA clocks are turned on by the bootloader.
This patch adds support for DMA clock handling so that
the driver manages the DMA clocks.
Fixes: 5000d37042 ("dmaengine: sh: Add DMAC driver for RZ/G2L SoC")
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20210923102451.11403-1-biju.das.jz@bp.renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
memset() and memcpy() on an MMIO region like here results in a
lockup at startup on mpc5200 platform (since this first happens
during probing of the ATA and Ethernet drivers). Use memset_io()
and memcpy_toio() instead.
Fixes: 2f9ea1bde0 ("bestcomm: core bestcomm support for Freescale MPC5200")
Cc: stable@vger.kernel.org # v5.14+
Signed-off-by: Anatolij Gustschin <agust@denx.de>
Link: https://lore.kernel.org/r/20211014094012.21286-1-agust@denx.de
Signed-off-by: Vinod Koul <vkoul@kernel.org>
AT_XDMAC_CC_PERID() should be used to setup bits 24..30 of XDMAC_CC
register. Using it without parenthesis around 0x7f & (i) will lead to
setting all the time zero for bits 24..30 of XDMAC_CC as the << operator
has higher precedence over bitwise &. Thus, add paranthesis around
0x7f & (i).
Fixes: 15a03850ab ("dmaengine: at_xdmac: fix macro typo")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211007111230.2331837-3-claudiu.beznea@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
at_xdmac could be used on SoCs which supports backup mode (where most
of the SoC power, including power to DMA controller, is closed at suspend
time). Thus, on resume, the settings which were previously done need to be
restored. Do the same for axi configuration.
Fixes: f40566f220 ("dmaengine: at_xdmac: add AXI priority support and recommended settings")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Link: https://lore.kernel.org/r/20211007111230.2331837-2-claudiu.beznea@microchip.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The variable err is being initialized with a value that is never read, it
is being updated later on. The assignment is redundant and can be removed
and move the declaration into the local scope.
Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20211015123447.27560-1-colin.king@canonical.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
ADMAIF FIFO uses a ring buffer and it is divided amongst the available
channels. The default FIFO size (in multiples of 16 words) of ADMAIF TX/RX
channels is as below:
* On Tegra210,
channel 1 to 2 : size = 3
channel 3 to 10: size = 2
* On Tegra186 and later,
channel 1 to 4 : size = 3
channel 5 to 20: size = 2
As per recommendation from HW, FIFO size of ADMA channel should be same as
the corresponding ADMAIF channel it maps to. FIFO corruption is observed if
the sizes do not match. We are using the default FIFO sizes for ADMAIF and
there is no plan to support any custom values.
Thus at runtime, override the ADMA channel FIFO size value depending on the
corresponding ADMAIF channel.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Link: https://lore.kernel.org/r/1631722025-19873-4-git-send-email-spujar@nvidia.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Trivial change to add description for 'adma_get_burst_config' in chip
data structure.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Link: https://lore.kernel.org/r/1631722025-19873-3-git-send-email-spujar@nvidia.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The 'has_outstanding_reqs' member description order in structure
'tegra_adma_chip_data' does not match with the corresponding member
declaration. The same is true for member assignment in chip data
structures declared for Tegra210 and Tegra186.
This is a trivial fix to re-order the mentioned member for a better
readability.
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Link: https://lore.kernel.org/r/1631722025-19873-2-git-send-email-spujar@nvidia.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
