AM6411: R5 remoteproc configuration

Hello Matthias,

Sorry for the delayed response since I was on leave last week.

This question mostly belongs to Linux, and currently, Linux experts are not available.

Now, we will follow the below steps; if they do not work out, then we will take help from Linux experts as well.

Please look at the below FAQ, and it clearly explains how to enable cores based on custom core selection for IPC in MCU+SDK and the next-generated Multicore app image.

The same things you have to do on the Linux side as well for customer core selection

https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1220573/faq-am6442-faq-am64x-generating-an-ipc-multicore-app-image-for-a-custom-selection-of-cores

Check below FAQ as well .

https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1245326/faq-rpmsg-how-to-add-multiple-endpoints-to-a-remote-core-running-rtos

Regards,

S.Anil.

Hello Matthias,

The information you are looking at is currently hiding in the next draft of the AM64x Academy, which will add a Multicore Academy alongside the AM64x Linux academy.

Here is the AM62x version of the page you are looking for:
https://dev.ti.com/tirex/explore/node?node=A__ARxKiGQnRRZIrn8bcmYQPw__linux_academy_am62x__XaWts8R__LATEST

Here is additional information that will be added to the AM64x version of that page:

Cores should be disabled if:

* The processor silicon variation does not have those remote cores. For example,
AM6422 & AM6421 have one R5F dual core subsystem instead of two subsystems,
and AM6412 & AM6411 have one R5F subsystem with a single R5F core.

* The application does not use the remote cores that are on the processor. Disabling
a core allows us to free up system resources (like memory) for usage by other cores.

...........................................

the warning about "device management subsystem R5F" does not apply to AM64x

.........................................

and then the section "Device tree changes to disable remote cores" changes like this:

For example, let's say we are using AM6411, and we are not using the M4F
core. In that case, we must also disable one of the R5F subsystems, and set
the other R5F subsystem to single-CPU mode.

and this is the AM64x patch file:

From 3af2225a8c3f8863c40c5423396c373d20538d5a Mon Sep 17 00:00:00 2001
From: Nick Saulnier <nsaulnier@ti.com>
Date: Mon, 2 Oct 2023 16:02:45 -0500
Subject: [PATCH] AM64x SDK 9.0: disable all cores but r5fss0_0

These cores are set to "disabled":
r5fss0_1, r5fss1_0, r5fss1_1, m4f

The mailbox node settings can be deleted, since the mailbox is no longer
being used for RPMsg communication between the disabled cores and Linux.

The disabled cores' memory allocations in DDR are released for use by other
parts of the processor system.

Since there is now only one MCU+ core running (r5fss0_0), the
rtos_ipc_memory_region is no longer used for IPC communication between MCU+
cores. This memory region can also be removed.

Signed-off-by: Nick Saulnier <nsaulnier@ti.com>
---
 arch/arm64/boot/dts/ti/k3-am642-sk.dts | 99 +++-----------------------
 1 file changed, 9 insertions(+), 90 deletions(-)

diff --git a/arch/arm64/boot/dts/ti/k3-am642-sk.dts b/arch/arm64/boot/dts/ti/k3-am642-sk.dts
index de0ff5af5e63..9cd48f35e2e7 100644
--- a/arch/arm64/boot/dts/ti/k3-am642-sk.dts
+++ b/arch/arm64/boot/dts/ti/k3-am642-sk.dts
@@ -50,60 +50,6 @@ main_r5fss0_core0_memory_region: r5f-memory@a0100000 {
 			reg = <0x00 0xa0100000 0x00 0xf00000>;
 			no-map;
 		};
-
-		main_r5fss0_core1_dma_memory_region: r5f-dma-memory@a1000000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa1000000 0x00 0x100000>;
-			no-map;
-		};
-
-		main_r5fss0_core1_memory_region: r5f-memory@a1100000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa1100000 0x00 0xf00000>;
-			no-map;
-		};
-
-		main_r5fss1_core0_dma_memory_region: r5f-dma-memory@a2000000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa2000000 0x00 0x100000>;
-			no-map;
-		};
-
-		main_r5fss1_core0_memory_region: r5f-memory@a2100000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa2100000 0x00 0xf00000>;
-			no-map;
-		};
-
-		main_r5fss1_core1_dma_memory_region: r5f-dma-memory@a3000000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa3000000 0x00 0x100000>;
-			no-map;
-		};
-
-		main_r5fss1_core1_memory_region: r5f-memory@a3100000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa3100000 0x00 0xf00000>;
-			no-map;
-		};
-
-		mcu_m4fss_dma_memory_region: m4f-dma-memory@a4000000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa4000000 0x00 0x100000>;
-			no-map;
-		};
-
-		mcu_m4fss_memory_region: m4f-memory@a4100000 {
-			compatible = "shared-dma-pool";
-			reg = <0x00 0xa4100000 0x00 0xf00000>;
-			no-map;
-		};
-
-		rtos_ipc_memory_region: ipc-memories@a5000000 {
-			reg = <0x00 0xa5000000 0x00 0x00800000>;
-			alignment = <0x1000>;
-			no-map;
-		};
 	};
 
 	vusb_main: fixed-regulator-vusb-main5v0 {
@@ -556,44 +502,25 @@ mbox_main_r5fss0_core0: mbox-main-r5fss0-core0 {
 		ti,mbox-rx = <0 0 2>;
 		ti,mbox-tx = <1 0 2>;
 	};
-
-	mbox_main_r5fss0_core1: mbox-main-r5fss0-core1 {
-		ti,mbox-rx = <2 0 2>;
-		ti,mbox-tx = <3 0 2>;
-	};
 };
 
 &mailbox0_cluster3 {
 	status = "disabled";
 };
 
-&mailbox0_cluster4 {
-	mbox_main_r5fss1_core0: mbox-main-r5fss1-core0 {
-		ti,mbox-rx = <0 0 2>;
-		ti,mbox-tx = <1 0 2>;
-	};
-
-	mbox_main_r5fss1_core1: mbox-main-r5fss1-core1 {
-		ti,mbox-rx = <2 0 2>;
-		ti,mbox-tx = <3 0 2>;
-	};
-};
-
 &mailbox0_cluster5 {
 	status = "disabled";
 };
 
-&mailbox0_cluster6 {
-	mbox_m4_0: mbox-m4-0 {
-		ti,mbox-rx = <0 0 2>;
-		ti,mbox-tx = <1 0 2>;
-	};
-};
-
 &mailbox0_cluster7 {
 	status = "disabled";
 };
 
+/* set R5F subsystem to single-CPU mode */
+&main_r5fss0 {
+	ti,cluster-mode = <2>;
+};
+
 &main_r5fss0_core0 {
 	mboxes = <&mailbox0_cluster2 &mbox_main_r5fss0_core0>;
 	memory-region = <&main_r5fss0_core0_dma_memory_region>,
@@ -601,27 +528,19 @@ &main_r5fss0_core0 {
 };
 
 &main_r5fss0_core1 {
-	mboxes = <&mailbox0_cluster2 &mbox_main_r5fss0_core1>;
-	memory-region = <&main_r5fss0_core1_dma_memory_region>,
-			<&main_r5fss0_core1_memory_region>;
+	status = "disabled";
 };
 
 &main_r5fss1_core0 {
-	mboxes = <&mailbox0_cluster4 &mbox_main_r5fss1_core0>;
-	memory-region = <&main_r5fss1_core0_dma_memory_region>,
-			<&main_r5fss1_core0_memory_region>;
+	status = "disabled";
 };
 
 &main_r5fss1_core1 {
-	mboxes = <&mailbox0_cluster4 &mbox_main_r5fss1_core1>;
-	memory-region = <&main_r5fss1_core1_dma_memory_region>,
-			<&main_r5fss1_core1_memory_region>;
+	status = "disabled";
 };
 
 &mcu_m4fss {
-	mboxes = <&mailbox0_cluster6 &mbox_m4_0>;
-	memory-region = <&mcu_m4fss_dma_memory_region>,
-			<&mcu_m4fss_memory_region>;
+	status = "disabled";
 };
 
 &ecap0 {
-- 
2.34.1

Regards,

Nick

Feel free to poke around the other pages added in the Multicore academy, there is other information that might be interesting for you.

e.g., if the firmware is being loaded by Linux remoteproc driver, you should check out section
"how to create remote core firmware that can be loaded by Linux:"
https://dev.ti.com/tirex/explore/node?node=A__AVP4Q-S9hfvfv8aDiSbBJw__linux_academy_am62x__XaWts8R__LATEST

Regards,

Nick

Hi Nick,

thank you for your reply - I finally got around to try it out. Your patch is fairly similar to the changes I attempted to make myself before I opened this thread. Unfortunately, the changes are not sufficient to make the Remoteproc driver work (on current ti-linux-kernel 6.1):

• The ti_k3_r5_remoteproc has a consistency check that will fail with "MCU cluster requires both R5F cores to be enabled but num_cores is set to = 1" when I modify the Device Tree as suggested
• When I comment out that check, the whole system will hang after printing "platform 78000000.r5f: configured R5F for remoteproc mode"

Regards,
Matthias

Hello Matthias,

Apologies for the delayed response, I lost your thread over Thanksgiving. I will try to replicate your results on Friday. Please ping the thread if I have not responded by Monday.

Regards,

Nick

Hi Nick,

thanks for the new patch, the AM6411 R5 is working nicely now - I was able to boot the IPC example and run the rpmsg_client_sample test with the changes applied.

Regards,
Matthias

Thanks for confirming Mattias! Please feel free to create a new thread if you have any other questions or feedback.

Regards,

Nick