Pull remoteproc updates from Bjorn Andersson:
"Qualcomm SM8650 audio, compute and modem remoteproc are added.
Qualcomm X1 Elite audio and compute remoteprocs are added, after
support for shutting down the bootloader-loaded firmware loaded into
the audio DSP..
A dozen drivers in the subsystem are transitioned to use devres
helpers for remoteproc and memory allocations - this makes it possible
to acquire in-kernel handle to individual remoteproc instances in a
cluster.
The release of DMA memory for remoteproc virtio is corrected to ensure
that restarting due to a watchdog bite doesn't attempt to allocate the
memory again without first freeing it.
Last, but not least, a couple of DeviceTree binding cleanups"
* tag 'rproc-v6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux: (30 commits)
remoteproc: qcom_q6v5_pas: Unload lite firmware on ADSP
remoteproc: qcom_q6v5_pas: Add support for X1E80100 ADSP/CDSP
dt-bindings: remoteproc: qcom,sm8550-pas: document the X1E80100 aDSP & cDSP
remoteproc: qcom_wcnss: Use devm_rproc_alloc() helper
remoteproc: qcom_q6v5_wcss: Use devm_rproc_alloc() helper
remoteproc: qcom_q6v5_pas: Use devm_rproc_alloc() helper
remoteproc: qcom_q6v5_mss: Use devm_rproc_alloc() helper
remoteproc: qcom_q6v5_adsp: Use devm_rproc_alloc() helper
dt-bindings: remoteproc: do not override firmware-name $ref
dt-bindings: remoteproc: qcom,glink-rpm-edge: drop redundant type from label
remoteproc: qcom: pas: correct data indentation
remoteproc: Make rproc_get_by_phandle() work for clusters
remoteproc: qcom: pas: Add SM8650 remoteproc support
remoteproc: qcom: pas: make region assign more generic
dt-bindings: remoteproc: qcom,sm8550-pas: document the SM8650 PAS
remoteproc: k3-dsp: Use devm_rproc_add() helper
remoteproc: k3-dsp: Use devm_ioremap_wc() helper
remoteproc: k3-dsp: Add devm action to release tsp
remoteproc: k3-dsp: Use devm_kzalloc() helper
remoteproc: k3-dsp: Use devm_ti_sci_get_by_phandle() helper
...
Multi-cluster remoteproc designs typically have the following DT
declaration:
remoteproc-cluster {
compatible = "soc,remoteproc-cluster";
core0: core0 {
compatible = "soc,remoteproc-core"
memory-region;
sram;
};
core1: core1 {
compatible = "soc,remoteproc-core"
memory-region;
sram;
}
};
A driver exists for the cluster rather than the individual cores
themselves so that operation mode and HW specific configurations
applicable to the cluster can be made.
Because the driver exists at the cluster level and not the individual
core level, function rproc_get_by_phandle() fails to return the
remoteproc associated with the phandled it is called for.
This patch enhances rproc_get_by_phandle() by looking for the cluster's
driver when the driver for the immediate remoteproc's parent is not
found.
Reported-by: Ben Levinsky <ben.levinsky@xilinx.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Co-developed-by: Tarak Reddy <tarak.reddy@amd.com>
Signed-off-by: Tarak Reddy <tarak.reddy@amd.com>
Co-developed-by: Tanmay Shah <tanmay.shah@amd.com>
Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
Link: https://lore.kernel.org/r/20240130154849.1018666-1-tanmay.shah@amd.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
The current memory region assign only supports a single
memory region.
But new platforms introduces more regions to make the
memory requirements more flexible for various use cases.
Those new platforms also shares the memory region between the
DSP and HLOS.
To handle this, make the region assign more generic in order
to support more than a single memory region and also permit
setting the regions permissions as shared.
Reviewed-by: Mukesh Ojha <quic_mojha@quicinc.com>
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20240123-topic-sm8650-upstream-remoteproc-v7-2-61283f50162f@linaro.org
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
The sparse tool complains about the remove of the _iomem attribute.
stm32_rproc.c:660:17: warning: cast removes address space '__iomem' of expression
Add '__force' to explicitly specify that the cast is intentional.
This conversion is necessary to cast to addresses pointer,
which are then managed by the remoteproc core as a pointer to a
resource_table structure.
Fixes: 8a471396d2 ("remoteproc: stm32: Move resource table setup to rproc_ops")
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
Link: https://lore.kernel.org/r/20240117135312.3381936-3-arnaud.pouliquen@foss.st.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
The sparse tool complains about the attribute conversion between
a _iomem void * and a void *:
stm32_rproc.c:122:12: sparse: sparse: incorrect type in assignment (different address spaces) @@ expected void *va @@ got void [noderef] __iomem * @@
stm32_rproc.c:122:12: sparse: expected void *va
stm32_rproc.c:122:12: sparse: got void [noderef] __iomem *
Add '__force' to explicitly specify that the cast is intentional.
This conversion is necessary to cast to virtual addresses pointer,used,
by the remoteproc core.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202312150052.HCiNKlqB-lkp@intel.com/
Fixes: 13140de09c ("remoteproc: stm32: add an ST stm32_rproc driver")
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
Link: https://lore.kernel.org/r/20240117135312.3381936-2-arnaud.pouliquen@foss.st.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Recovery remote processor failed when wdg irq received:
[ 0.842574] remoteproc remoteproc0: crash detected in cix-dsp-rproc: type watchdog
[ 0.842750] remoteproc remoteproc0: handling crash #1 in cix-dsp-rproc
[ 0.842824] remoteproc remoteproc0: recovering cix-dsp-rproc
[ 0.843342] remoteproc remoteproc0: stopped remote processor cix-dsp-rproc
[ 0.847901] rproc-virtio rproc-virtio.0.auto: Failed to associate buffer
[ 0.847979] remoteproc remoteproc0: failed to probe subdevices for cix-dsp-rproc: -16
The reason is that dma coherent mem would not be released when
recovering the remote processor, due to rproc_virtio_remove()
would not be called, where the mem released. It will fail when
it try to allocate and associate buffer again.
Releasing reserved memory from rproc_virtio_dev_release(), instead of
rproc_virtio_remove().
Fixes: 1d7b61c06d ("remoteproc: virtio: Create platform device for the remoteproc_virtio")
Signed-off-by: Joakim Zhang <joakim.zhang@cixtech.com>
Acked-by: Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231217053659.3245745-1-joakim.zhang@cixtech.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.
To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().
There is no change in behaviour as .remove() already returned zero
unconditionally.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20231123211657.518181-8-u.kleine-koenig@pengutronix.de
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Add the .find_loaded_rsc_table operation for i.MX DSP.
We need it for inter-process communication between DSP
and main core.
This callback is used to find the resource table (defined
in remote processor linker script) where the address of the
vrings along with the other allocated resources (carveouts etc)
are stored.
If this is not found, the vrings are not allocated and
the IPC between cores will not work.
Signed-off-by: Iuliana Prodan <iuliana.prodan@nxp.com>
Reviewed-by: Daniel Baluta <daniel.baluta@nxp.com>
Link: https://lore.kernel.org/r/20231013152731.23471-2-iuliana.prodan@oss.nxp.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Clang warns (or errors with CONFIG_WERROR=y):
drivers/remoteproc/st_remoteproc.c:357:6: error: variable 'ret' is used uninitialized whenever 'if' condition is true [-Werror,-Wsometimes-uninitialized]
357 | if (!ddata->config)
| ^~~~~~~~~~~~~~
drivers/remoteproc/st_remoteproc.c:442:9: note: uninitialized use occurs here
442 | return ret;
| ^~~
drivers/remoteproc/st_remoteproc.c:357:2: note: remove the 'if' if its condition is always false
357 | if (!ddata->config)
| ^~~~~~~~~~~~~~~~~~~
358 | goto free_rproc;
| ~~~~~~~~~~~~~~~
drivers/remoteproc/st_remoteproc.c:348:9: note: initialize the variable 'ret' to silence this warning
348 | int ret, i;
| ^
| = 0
1 error generated.
Set ret to -ENODEV, which seems to be a standard return code when
device_get_match_data() returns NULL.
Closes: https://github.com/ClangBuiltLinux/linux/issues/1944
Fixes: 5c77ebcd05 ("remoteproc: st: Use device_get_match_data()")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Link: https://lore.kernel.org/r/20231012-st_remoteproc-fix-sometimes-uninit-v1-1-f64d0f2d5b37@kernel.org
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
In lockstep mode following is TCM address map:
| *TCM* | *R5 View* | *Linux view* |
| R5_0 ATCM (128 KB) | 0x0000_0000 | 0xFFE0_0000 |
| R5_0 BTCM (128 KB) | 0x0002_0000 | 0xFFE2_0000 |
Current driver keeps single TCM carveout in lockstep mode
as ATCM and BTCM addresses form contiguous memory region.
Although the addresses are contiguous, it is not same type
of memory. ATCM typically holds interrupt or exception code
that must be accessed at high speed. BTCM typically holds
a block of data for intensive processing, such as audio or
video processing. As both are different types of memory,
they should be allocated as different carveout. This patch
is fixing TCM carveout allocation in lockstep mode.
Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
Link: https://lore.kernel.org/r/20230913024323.2768114-1-tanmay.shah@amd.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Because MT8195 SCP core 0 and core 1 both boot from head of SRAM and
have the same viewpoint of SRAM, SCP has a "core 1 SRAM offset"
configuration to control the access destination of SCP core 1 to boot
core 1 from different SRAM location.
The "core 1 SRAM offset" configuration is composed by a range
and an offset. It works like a simple memory mapped mechanism.
When SCP core 1 accesses a SRAM address located in the range,
the SCP bus adds the configured offset to the address to
shift the physical destination address on SCP SRAM. This shifting is
transparent to the software running on SCP core 1.
Signed-off-by: Tinghan Shen <tinghan.shen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20230901080935.14571-11-tinghan.shen@mediatek.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Previously, SCP core 0 controlled the power of L2TCM and dictated that
SCP core 1 could only boot after SCP core 0. To address this constraint,
extracted the power control flow of L2TCM and made it shared
between both cores, enabling support for arbitrary boot order.
The flow for controlling L2TCM power has been incorporated into the
mt8195_scp_before_load() and mt8195_scp_stop() APIs, which are
respectively invoked during the rproc->ops->start() and
rproc->ops->stop() operations. These APIs effectively serve the same
purpose as the rproc prepare()/unprepare() APIs."
Signed-off-by: Tinghan Shen <tinghan.shen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20230901080935.14571-10-tinghan.shen@mediatek.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
