sample: openamp: add support for stm32h747i_disco board #38322

cameled · 2021-09-06T04:21:54Z

NOTE: This PR currently are not ready, still have some problem with M4 booting.

[EDIT] how to using test this sample on openamp sample

# Build M7 and M4
$ west build -p -b stm32h747i_disco_m7 samples/subsys/ipc/openamp

# flash M7
$ west flash

# flash M4 
$ west flash --build-dir build/openamp_remote-prefix/src/openamp_remote-build

cameled · 2021-09-06T04:37:16Z

@galak @hubertmis Hi, Did you have meet booting problem when working with open-amp and rpmsg_service samples? On stm32h747i_disco board, when run openamp and rpmsg_service sample, M4 are not boot.

stm32h747i_disco have a M7(master) and M4(remote). M4 not support openocd debug, I add led blinky to it's main function, but it's not blink.

I have check the M4 image on flash and zephyr.bin file, it's same.

carlocaione · 2021-09-06T07:45:04Z

@cameled just to let you know there is a discussion going on whether we want to deprecate rpmsg_service in favour of ipc_service.

erwango · 2021-09-06T07:45:24Z

@cameled Is that specific to those samples ?

cameled · 2021-09-06T07:50:42Z

@erwango Yes, I think this issue are not same as #37827
I test 8KB array on blinky sample base on this PR, M4 blink fine.

cameled · 2021-09-06T07:57:25Z

@carlescufi Thanks for you mentioning. open-amp raw interface are not friendly for end user, ~~If the ipc_service have merged~~, I will try to make stm32h747i_disco support it.

cameled · 2021-09-08T08:40:41Z

@erwango I get once time M4 boot success case with this PR, no matter power-down or reset, M4 will always boot success. After I re-flash the M4 board with the same image, boot failed happened again.

Looks there has some deeper issues that I can't find out.

cameled · 2021-09-09T07:13:23Z

@erwango I disabled the CM4 boot config "BCM4" on STM32Progrmmer->option bytes->User Configuration, then all the things return to normal, no open-amp or large size array(#37827) issues exist.

erwango · 2021-09-09T07:22:58Z

@erwango I disabled the CM4 boot config "BCM4" on STM32Progrmmer->option bytes->User Configuration, then all the things return to normal, no open-amp or large size array(#37827) issues exist.

@ABOSTM, can you check this is the expected configuration?
If yes, should we document this, or even implement a check to detect unsupported configurations?

cameled · 2021-09-09T08:42:10Z

Another issue, shared memory not work fine.

If write val1 to sram4(0x38000000) on CM7,

uint32_t *sp = (uint32_t *)0x38000000;
*sp  = val1;

then read val2 from sram4(0x38000000) on CM4.

uint32_t *sp = (uint32_t *)0x38000000;
val2 = *sp;

val1 and val2 are not equal ;-)

carlocaione · 2021-09-09T08:44:12Z

Another issue, shared memory not work fine.

If write val1 to sram4(0x38000000) on CM7,
uint32_t *sp = (uint32_t *)0x38000000;
*sp  = val1;
then read val2 from sram4(0x38000000) on CM4.
uint32_t *sp = (uint32_t *)0x38000000;
val2 = *sp;
val1 and val2 are not equal ;-)

Cache?

cameled · 2021-09-09T08:56:20Z

Another issue, shared memory not work fine.
If write val1 to sram4(0x38000000) on CM7,
uint32_t *sp = (uint32_t *)0x38000000;
*sp  = val1;
then read val2 from sram4(0x38000000) on CM4.
uint32_t *sp = (uint32_t *)0x38000000;
val2 = *sp;
val1 and val2 are not equal ;-)
Cache?

Yes, looks like it's a cache issue.
If I add CONFIG_NOCACHE_MEMORY=y to M7 Kconfig, then shared memory work fine.

cameled · 2021-09-09T09:10:27Z

@erwango Finally, open-amp and rpmsg_service samples work fine.

There have two issue block this PR.

M4 boot issue
~~Fix by disable the CM4 boot config "BCM4" on STM32Progrmmer->option bytes->User Configuration,~~
Fix by @ABOSTM soc: arm: stm32h7: rework STM32H7 dual core boot #38504
M7 memory cache issue
Fix by add CONFIG_NOCACHE_MEMORY=y to M7 Kconfig(thanks @carlocaione).

carlocaione · 2021-09-09T09:11:50Z

For (2) you want to take a look at OpenAMP/open-amp#293

ABOSTM · 2021-09-09T11:24:00Z

@erwango I disabled the CM4 boot config "BCM4" on STM32Progrmmer->option bytes->User Configuration, then all the things return to normal, no open-amp or large size array(#37827) issues exist.

@ABOSTM, can you check this is the expected configuration?
If yes, should we document this, or even implement a check to detect unsupported configurations?

It is not the expected behavior, it should work with BCM4 enabled (especially as it is the default configuration out of the box), and it used to work when I worked on this dualcore. So it sounds like a regression

cameled · 2021-09-14T05:06:59Z

@erwango I disabled the CM4 boot config "BCM4" on STM32Progrmmer->option bytes->User Configuration, then all the things return to normal, no open-amp or large size array(#37827) issues exist.

@ABOSTM, can you check this is the expected configuration?
If yes, should we document this, or even implement a check to detect unsupported configurations?

It is not the expected behavior, it should work with BCM4 enabled (especially as it is the default configuration out of the box), and it used to work when I worked on this dualcore. So it sounds like a regression

There have some unknown areas, why open-amp not running code(POST_KERNEL and main) can affect the CM4 booting process.

lenghonglin · 2021-10-31T12:36:22Z

the pull request is not counting ? @cameled

arnopo

minor comments else LGTM

subsys/ipc/rpmsg_service/rpmsg_backend.c

samples/subsys/ipc/openamp/boards/stm32h747i_disco_m7.overlay

samples/subsys/ipc/rpmsg_service/boards/stm32h747i_disco_m7.overlay

erwango · 2021-11-05T09:23:53Z

@cameled I've had a try on my board and unfortunately I can't get it working.
Here is what I can get on m7 side:

*** Booting Zephyr OS build v2.7.99-1108-g4eec821986ba  ***
Starting application thread!

OpenAMP[master] demo started

On M4:

*** Booting Zephyr OS build v2.7.99-1108-g4eec821986ba  ***
Starting application thread!

OpenAMP[remote] demo started

Trying to debug (m7 side), it seems to be doing right, waiting in arch_cpu_idle.
Any advice on what I can do to see what is going on ?

cameled · 2021-11-05T09:44:17Z

@cameled I've had a try on my board and unfortunately I can't get it working. Here is what I can get on m7 side:
*** Booting Zephyr OS build v2.7.99-1108-g4eec821986ba  ***
Starting application thread!

OpenAMP[master] demo started
On M4:
*** Booting Zephyr OS build v2.7.99-1108-g4eec821986ba  ***
Starting application thread!

OpenAMP[remote] demo started
Trying to debug (m7 side), it seems to be doing right, waiting in arch_cpu_idle. Any advice on what I can do to see what is going on ?

Quick fix:
add CONFIG_NOCACHE_MEMORY=y to CM7, there have a cache issue on M7 Core.

erwango · 2021-11-05T09:52:13Z

Quick fix:
add CONFIG_NOCACHE_MEMORY=y to CM7, there have a cache issue on M7 Core.

\o/ working fine now! Thanks
Is there a plan to fix this in another PR ? Otherwise, what about adding this in samples/subsys/ipc/openamp/boards/stm32h747i_disco_m7.conf
This would avoid others to face the same issue.

cameled · 2021-11-05T10:02:14Z

Quick fix:
add CONFIG_NOCACHE_MEMORY=y to CM7, there have a cache issue on M7 Core.

\o/ working fine now! Thanks Is there a plan to fix this in another PR ? Otherwise, what about adding this in samples/subsys/ipc/openamp/boards/stm32h747i_disco_m7.conf This would avoid others to face the same issue.

@carlocaione have merged the openamp cache helpers OpenAMP/open-amp#293 to Zephyr. But there have an interface missing issue report by @lenghonglin #38322 (comment).

erwango · 2021-11-05T10:31:59Z

@carlocaione have merged the openamp cache helpers OpenAMP/open-amp#293 to Zephyr. But there have an interface missing issue report by @lenghonglin #38322 (comment).

What about adding CONFIG_NOCACHE_MEMORY with a note specifying this is a temp workaround waiting the issue to be fixed ? Otherwise I'd prefer we avoid to merge current PR if not working as is.

carlocaione

What about adding CONFIG_NOCACHE_MEMORY with a note specifying this is a temp workaround waiting the issue to be fixed ? Otherwise I'd prefer we avoid to merge current PR if not working as is.

Agreed

arnopo · 2021-11-05T15:21:35Z

What about adding CONFIG_NOCACHE_MEMORY with a note specifying this is a temp workaround waiting the issue to be fixed ? Otherwise I'd prefer we avoid to merge current PR if not working as is.

Agreed

How CONFIG_NOCACHE_MEMORY is working?
Look like an associated no cacheable section can be defined in the linker...
In this case Defining the shared memory in a specific section would make sense, not only has a temporary work around.
As the rpmsg no copy interface is not used, defining the shared memory in a non cacheable area would probably be more efficient.

cameled · 2021-11-12T01:50:23Z

How CONFIG_NOCACHE_MEMORY is working? Look like an associated no cacheable section can be defined in the linker... In this case Defining the shared memory in a specific section would make sense, not only has a temporary work around. As the rpmsg no copy interface is not used, defining the shared memory in a non cacheable area would probably be more efficient.

From ARM Cortex-M7 Reference Manual, section "4.1.3 Initializing and enabling the L1 cache", L1 cache are disabled by default.

stm32h7 enable this at soc_m7.c file with CONFIG_NOCACHE_MEMORY check.

#ifndef CONFIG_NOCACHE_MEMORY
	if (!(SCB->CCR & SCB_CCR_DC_Msk)) {
		SCB_EnableDCache();
	}
#endif /* CONFIG_NOCACHE_MEMORY */

cameled · 2021-11-12T05:53:09Z

@arnopo no cacheable section like #30403? I prefer to implement the cache control for the Cortex-M, which will be a more complete solution. Also found libmetal metal_io_region have mem_flags.

erwango · 2021-11-15T08:34:46Z

stm32h7 enable this at soc_m7.c file with CONFIG_NOCACHE_MEMORY check.

This part is indeed uncorrect and I hope #40146 should help providing a better solution.
You can use it for now, but we should keep in mind this is temporary.
Ideally, we should remove usage of NOCACHE_MEMORY which is not proper here and replace with a H7 cache deactivation specific flag.

cameled · 2021-12-22T05:45:31Z

Will update at this weekend.

cameled · 2022-01-06T07:16:25Z

Will update at this weekend.

Sorry for the delay. After check the ipc_service sample, I find there have an ipm channel number requirement.
ipm_stm32_hsem mailbox only have one channel for Tx/Rx.

carlocaione · 2022-01-06T08:26:19Z

Sorry for the delay. After check the ipc_service sample, I find there have an ipm channel number requirement.
ipm_stm32_hsem mailbox only have one channel for Tx/Rx.

That's fine. You can use just one single channel for tx/rx.

This add openamp sample run on stm32h747i_disco board. The CM7 on stm32h747 support memory cache. If the cache enabled on CM7, CM4 will receive incorrect data. This just disable the memory cache on CM7, future user may want to use more flexible way to fix it. Signed-off-by: HaiLong Yang <cameledyang@pm.me>

cameled · 2022-01-10T08:51:54Z

Remove rpmsg_service sample support.

Not support ipc_service, it based on mbox api, stm32h747 current not support it.

This just disable the memory cache on CM7, future user may want to use more flexible way to fix it.

github-actions · 2022-04-23T00:30:25Z

This pull request has been marked as stale because it has been open (more than) 60 days with no activity. Remove the stale label or add a comment saying that you would like to have the label removed otherwise this pull request will automatically be closed in 14 days. Note, that you can always re-open a closed pull request at any time.

cameled · 2022-05-05T09:31:57Z

Close this PR.

cameled requested review from ioannisg and nashif as code owners September 6, 2021 04:21

github-actions bot added area: Documentation area: Samples Samples labels Sep 6, 2021

cameled force-pushed the rpmsg_serv_747 branch 2 times, most recently from 0e02b86 to 6b969e9 Compare September 6, 2021 04:57

erwango requested review from arnopo and erwango September 6, 2021 07:42

erwango self-assigned this Sep 6, 2021

erwango added this to the v3.0.0 milestone Sep 6, 2021

erwango added the platform: STM32 ST Micro STM32 label Sep 7, 2021

cameled mentioned this pull request Sep 11, 2021

stm32h747i_disco M4 not working, if use large size(>1KB) global array #37827

Closed

ABOSTM mentioned this pull request Sep 13, 2021

soc: arm: stm32h7: rework STM32H7 dual core boot #38504

Merged

cameled force-pushed the rpmsg_serv_747 branch from b9eb86b to a653bf8 Compare September 14, 2021 06:41

arnopo previously approved these changes Nov 4, 2021

View reviewed changes

subsys/ipc/rpmsg_service/rpmsg_backend.c Outdated Show resolved Hide resolved

samples/subsys/ipc/openamp/boards/stm32h747i_disco_m7.overlay Outdated Show resolved Hide resolved

samples/subsys/ipc/rpmsg_service/boards/stm32h747i_disco_m7.overlay Outdated Show resolved Hide resolved

cameled force-pushed the rpmsg_serv_747 branch from 4b6bb57 to 4eec821 Compare November 5, 2021 06:00

carlocaione approved these changes Nov 5, 2021

View reviewed changes

carlocaione requested changes Nov 5, 2021

View reviewed changes

cameled force-pushed the rpmsg_serv_747 branch from 4eec821 to d74fe90 Compare November 12, 2021 03:53

arnopo mentioned this pull request Nov 17, 2021

Introduce DT-based OpenAMP IPC service backend #40365

Merged

cameled force-pushed the rpmsg_serv_747 branch from d74fe90 to 8e968ba Compare January 6, 2022 07:24

cameled changed the title ~~ipc: rpmsg_service: add stm32h747i_disco board support~~ sample: openamp: add support for stm32h747i_disco board Jan 10, 2022

cameled force-pushed the rpmsg_serv_747 branch from 8e968ba to 928b82f Compare January 10, 2022 08:43

dleach02 dismissed arnopo’s stale review via 928b82f February 11, 2022 16:03

dkalowsk modified the milestones: v3.0.0, v3.1.0 Feb 21, 2022

github-actions bot added the Stale label Apr 23, 2022

cameled closed this May 5, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

sample: openamp: add support for stm32h747i_disco board #38322

sample: openamp: add support for stm32h747i_disco board #38322

cameled commented Sep 6, 2021 •

edited

cameled commented Sep 6, 2021

carlocaione commented Sep 6, 2021

erwango commented Sep 6, 2021

cameled commented Sep 6, 2021

cameled commented Sep 6, 2021 •

edited

cameled commented Sep 8, 2021 •

edited

cameled commented Sep 9, 2021

erwango commented Sep 9, 2021

cameled commented Sep 9, 2021

carlocaione commented Sep 9, 2021

cameled commented Sep 9, 2021 •

edited

cameled commented Sep 9, 2021 •

edited

carlocaione commented Sep 9, 2021

ABOSTM commented Sep 9, 2021

cameled commented Sep 14, 2021

lenghonglin commented Oct 31, 2021

arnopo left a comment

erwango commented Nov 5, 2021

cameled commented Nov 5, 2021

erwango commented Nov 5, 2021

cameled commented Nov 5, 2021 •

edited

erwango commented Nov 5, 2021

carlocaione left a comment

arnopo commented Nov 5, 2021 •

edited

cameled commented Nov 12, 2021 •

edited

cameled commented Nov 12, 2021

erwango commented Nov 15, 2021

cameled commented Dec 22, 2021

cameled commented Jan 6, 2022

carlocaione commented Jan 6, 2022

cameled commented Jan 10, 2022

github-actions bot commented Apr 23, 2022

cameled commented May 5, 2022

sample: openamp: add support for stm32h747i_disco board #38322

sample: openamp: add support for stm32h747i_disco board #38322

Conversation

cameled commented Sep 6, 2021 • edited

cameled commented Sep 6, 2021

carlocaione commented Sep 6, 2021

erwango commented Sep 6, 2021

cameled commented Sep 6, 2021

cameled commented Sep 6, 2021 • edited

cameled commented Sep 8, 2021 • edited

cameled commented Sep 9, 2021

erwango commented Sep 9, 2021

cameled commented Sep 9, 2021

carlocaione commented Sep 9, 2021

cameled commented Sep 9, 2021 • edited

cameled commented Sep 9, 2021 • edited

carlocaione commented Sep 9, 2021

ABOSTM commented Sep 9, 2021

cameled commented Sep 14, 2021

lenghonglin commented Oct 31, 2021

arnopo left a comment

Choose a reason for hiding this comment

erwango commented Nov 5, 2021

cameled commented Nov 5, 2021

erwango commented Nov 5, 2021

cameled commented Nov 5, 2021 • edited

erwango commented Nov 5, 2021

carlocaione left a comment

Choose a reason for hiding this comment

arnopo commented Nov 5, 2021 • edited

cameled commented Nov 12, 2021 • edited

cameled commented Nov 12, 2021

erwango commented Nov 15, 2021

cameled commented Dec 22, 2021

cameled commented Jan 6, 2022

carlocaione commented Jan 6, 2022

cameled commented Jan 10, 2022

github-actions bot commented Apr 23, 2022

cameled commented May 5, 2022

cameled commented Sep 6, 2021 •

edited

cameled commented Sep 6, 2021 •

edited

cameled commented Sep 8, 2021 •

edited

cameled commented Sep 9, 2021 •

edited

cameled commented Sep 9, 2021 •

edited

cameled commented Nov 5, 2021 •

edited

arnopo commented Nov 5, 2021 •

edited

cameled commented Nov 12, 2021 •

edited