Date
1 - 8 of 8
Using the Cortex M4 on IMX7D
|
Michael Graichen
Hey,
I have just started working on a IMX7D (https://www.kontron.de/products/boards-and-standard-form-factors/smarc/smarc-samx7.html) together with Zephyr I noticed that the Colibri iMX7 Board has support for the Cortex M4. So I've build an example by using: cmake -GNinja -DBOARD=colibri_imx7d_m4 ../samples/basic/blinky and tried to load that on the Cortex M4 echo -n zephyr.elf > /sys/class/remoteproc/remoteproc0/firmware and echo -n start > /sys/class/remoteproc/remoteproc0/state Unfortunatly the start fails because the elf-file has no resource table. I can select TOP ---> Additional libraries ---> [*] coprocessor resource table within menuconfig but the resulting elf-file is still missing the resource table. What am i missing? How can a add a resource table to the elf-file? best regards Michael |
|
|
|
Kumar Gala
On Feb 26, 2021, at 3:43 AM, Michael Graichen <michael.graichen@...> wrote: I don’t believe the remoteproc loading was ever supported for the imx family with zephyr. This is something you’d need to add support for. I think it might be supported on the STM32 MP1 as an example. (I added Arnaud on CC as he added the STM32 MP1 support). For the i.mx I think you can load via u-boot and there’s a section in the zephyr board docs here about it: https://docs.zephyrproject.org/latest/boards/arm/colibri_imx7d_m4/doc/index.html - k |
|
|
|
Arnaud POULIQUEN
Hi,
toggle quoted message
Show quoted text
-----Original Message-----The Linux remoteproc framework allows you to load a firmware without a resource table. The resource table is useful only if you want to use the RPMsg protocol or a trace buffer. So this should work, except if the IMX platform driver doesn't allow this.... Right, I think today only the stm32mp1 support the resource table in zephyr, but this should be enough generic to be supported by another platform. Anyway if you don't want to use RPMsg, probably better as a first step to understand why you can not load a firmware without resource table. Regards Arnaud
|
|
|
|
Marco Cavallini
Hi,
I tested Zephyr with a Colibri i.MX7 and it works our of the box. west build -p auto -b colibri_imx7d_m4 samples/philosophers afterwards I loaded and run the binary using u-boot -- Marco Cavallini - KOAN - Italy |
|
|
|
Michael Graichen
Am 26.02.21 um 17:27 schrieb Kumar Gala:
Thanks for your reply.On Feb 26, 2021, at 3:43 AM, Michael Graichen <michael.graichen@...> wrote:I don’t believe the remoteproc loading was ever supported for the imx family with zephyr. This is something you’d need to add support for. I think it might be supported on the STM32 MP1 as an example. (I added Arnaud on CC as he added the STM32 MP1 support). This link was the reason why I thought to give it a try. But I'm using barebox as a bootloader right now, so I thought to test "the Linux way" first. I guess I will test u-boot next week. I have looked over the boot instruction from the link. u-boot is using the plain .bin file for loading the M4 while remoteproc is using the .elf file. Is it right that all those environment variables, especially 'setenv m4addr 0x007f800' that are set in u-boot before booting the M4 should come out of the resource table when using remoteproc? Do you have more information about how a resource table for the IMX7-M4 should look like? Have a great weekend. Michael |
|
|
|
Michael Graichen
Am 27.02.21 um 09:32 schrieb Marco Cavallini:
Hi, Thanks you you reply. guess I will test that next week. What Memory addresses did you use when you tested the "out of the Box" example? Have a great weekend Michael |
|
|
|
Michael Graichen
Am 26.02.21 um 18:06 schrieb Arnaud POULIQUEN:
Hi,-----Original Message-----The Linux remoteproc framework allows you to load a firmware without a resource table. Hey Arnaud, Thanks for your reply. I've seen that stm32_rproc driver has additionally functions within static const struct rproc_ops st_rproc_ops https://elixir.bootlin.com/linux/v5.11.2/source/drivers/remoteproc/stm32_rproc.c#L544 especially .parse_fw = stm32_rproc_parse_fw, within the plain rproc driver rproc->ops->parse_fw = rproc_elf_load_rsc_table is registerd within "static int rproc_alloc_ops()" https://elixir.bootlin.com/linux/v5.11.2/source/drivers/remoteproc/remoteproc_core.c#L2202 and is called by doing the "echo start" command. https://elixir.bootlin.com/linux/v5.11.2/source/drivers/remoteproc/remoteproc_elf_loader.c#L359 but fails after the find_table() function. so if i just skip that for a test in https://elixir.bootlin.com/linux/v5.11.2/source/drivers/remoteproc/remoteproc_core.c#L1464, does that take me over the point? I can not see the code where rproc actually "loads" the firmware to the M4? best regards and a great weekend Michael |
|
|
|
Arnaud POULIQUEN
Hi,
This is the generic elf parser functions that are available here: https://elixir.bootlin.com/linux/latest/source/drivers/remoteproc/remoteproc_elf_loader.c Have you specified the firmware name? Here is the sequence to load a firmware from Linux user application: cp $1 /lib/firmware/ echo $1> /sys/class/remoteproc/remoteproc0/firmware echo start >/sys/class/remoteproc/remoteproc0/state with $1 is the firmware to load This function return NULL or the address of the resource table, but not return an error Do you have a kernel error message ?
This is mainly done in https://elixir.bootlin.com/linux/latest/source/drivers/remoteproc/remoteproc_elf_loader.c Regards, Arnaud
|
|
|