Xlnx add trace buf and update address space by tnmysh · Pull Request #98 · OpenAMP/openamp-system-reference

tnmysh · 2026-03-13T04:00:16Z

Add trace buffer support
Update current DDR addresses to match latest system device-tree standards
fix trivial compiler warning during building matrix multiply demo

arnopo · 2026-03-31T07:37:34Z

 collect (APP_INC_DIRS "${CMAKE_CURRENT_SOURCE_DIR}")

 if (EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/${PROJECT_SYSTEM}/gic_init.c")
  collect (APP_COMMON_SOURCES ${PROJECT_SYSTEM}/gic_init.c)


s/exampples/examples/ in commit message

That said you can remove examples: as done for other patches

arnopo · 2026-03-31T07:39:45Z

+	int32_t len;
+	va_list args;
+
+


useless double blank line

arnopo · 2026-03-31T07:42:22Z


-#define NUM_TABLE_ENTRIES           1
+#define NUM_TABLE_ENTRIES           2
+#define RSC_TRACE_SZ		    (4*1024)


nitpicking. The trace contain char, why not just just

Suggested change

#define RSC_TRACE_SZ (4*1024)

#define RSC_TRACE_SZ 4096

tnmysh · 2026-04-07T15:39:44Z

Hi @arnopo,

I found bug in this patch series when using internal tools, please do not merge it until next update.

Thanks,
Tanmay

tnmysh · 2026-05-18T19:12:58Z

@arnopo, @edmooring this change is ready for review.

platform_info.h was changed to platform_info_common.h, replace it accordingly for echo demo. Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

fix following compiler warning: warning: passing argument 1 of 'rpmsg_matrix_app' fr om incompatible pointer type [-Wincompatible-pointer-types] 47 | rpmsg_matrix_app(rpdev, platform); | ^~~~~ Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

As per latest SDT standard AMD-xilinx EDF yocto is released with fixed memory regions for each core to use. Update the default address for cluster-A core0 to use as per this standard. Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

trace buffer is a circular buffer to store logs. Linux remoteproc subsystem creates a debugfs file that can be used to print these logs on linux side. Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

edmooring

This comes down to two questions:

Which is the correct size for the TCM lockstep region, 0x40000 or 0x 60000?
Why is the log level escalated for levels above "debug"?

edmooring · 2026-05-22T07:36:04Z

 MEMORY
 {
-   psu_ddr_S_AXI_BASEADDR : ORIGIN = 0x3ED00000, LENGTH = 0x00040000
+   psu_ddr_S_AXI_BASEADDR : ORIGIN = 0x9800000, LENGTH = 0x00060000


This now conflicts with the comment in line 18 above, Size has changed from 0x40000 to 0x60000 without changing the comment. Which is correct?

@edmooring, actually the memory modified here is not TCM, but DDR. So the firmware code can be split and some sections can be loaded into TCM, and some are loaded in the DDR.

The TCM memory is represented here with: psu_r5_tcm_ram_0_S_AXI_BASEADDR

Which I think is 64KB. I agree that the comment should be modified, as the TCM is 64-KB chunks. I will update comment.

edmooring · 2026-05-22T07:38:10Z

 MEMORY
 {
-   psu_ddr_S_AXI_BASEADDR : ORIGIN = 0x3ED00000, LENGTH = 0x00040000
+   psu_ddr_S_AXI_BASEADDR : ORIGIN = 0x9800000, LENGTH = 0x00060000


Same as above. Either the comment in line 18 is wrong or the code in line 21 is wrong.

Same explanation as above.

edmooring · 2026-05-22T07:42:32Z

 #ifndef SHARED_MEM_PA
 #if XPAR_CPU_ID == 0
-#define SHARED_MEM_PA  0x3ED40000UL
+#define SHARED_MEM_PA  0x9860000UL


Again, why the size change from 0x40000 to 0x60000? This also applies to line 106 below.

This SHARED_MEM_PA is different region than mentioned in the linker script.

This is where vring buffers are located. This address is moved in the new spec.

edmooring · 2026-05-22T07:58:56Z

+} circ;
+
+static void rsc_trace_putchar(char c)
+{


A very small nit:
The circular buffer will be left in a logically inconsistent state after c_len calls to rsc_trace_puchar(), where c_pos actually points past the end of the buffer. This might cause problems for readers of the buffer. Switching the order so the put to the buffer is done first, and then the check for the end of the buffer would keep the state of the buffer logically consistent between calls to rsc_trace_putchar().

edmooring · 2026-05-22T08:07:31Z

+	int32_t len;
+	va_list args;
+
+	if (level > METAL_LOG_DEBUG)


This needs an explanatory comment. Why are we escalating the warning level?

tnmysh requested review from arnopo and edmooring March 13, 2026 04:00

tnmysh force-pushed the xlnx_add_trace_buf_and_update_address_space branch from f6a6178 to 833a6b4 Compare March 13, 2026 21:26

arnopo requested changes Mar 18, 2026

View reviewed changes

tnmysh force-pushed the xlnx_add_trace_buf_and_update_address_space branch from 833a6b4 to b59ecf4 Compare March 27, 2026 03:45

tnmysh requested a review from arnopo March 27, 2026 03:48

arnopo reviewed Mar 31, 2026

View reviewed changes

tnmysh force-pushed the xlnx_add_trace_buf_and_update_address_space branch from b59ecf4 to 1e20941 Compare April 3, 2026 20:28

tnmysh force-pushed the xlnx_add_trace_buf_and_update_address_space branch from 1e20941 to 6796ba0 Compare May 18, 2026 15:19

tnmysh added 4 commits May 20, 2026 08:01

legacy_apps: echo: include new header

b18d275

platform_info.h was changed to platform_info_common.h, replace it accordingly for echo demo. Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

legacy_apps: xlnx: update reserved mem regions

a62c136

As per latest SDT standard AMD-xilinx EDF yocto is released with fixed memory regions for each core to use. Update the default address for cluster-A core0 to use as per this standard. Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

legacy_apps: zynqmp_r5: add trace buffer support

9456f3e

trace buffer is a circular buffer to store logs. Linux remoteproc subsystem creates a debugfs file that can be used to print these logs on linux side. Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>

tnmysh force-pushed the xlnx_add_trace_buf_and_update_address_space branch from 6796ba0 to 9456f3e Compare May 20, 2026 15:02

tnmysh requested a review from arnopo May 20, 2026 15:12

arnopo added this to the Release V2026.04 milestone May 20, 2026

arnopo approved these changes May 21, 2026

View reviewed changes

edmooring requested changes May 22, 2026

View reviewed changes

Conversation

tnmysh commented Mar 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tnmysh commented Apr 7, 2026

Uh oh!

tnmysh commented May 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

edmooring left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

tnmysh commented Mar 13, 2026 •

edited

Loading

tnmysh commented May 18, 2026 •

edited

Loading