Linux/AM5728: CMEM error

I should also note that dmesg reports CMA pools being created for different cores but /proc/iomem doesn't show any CMEM allocation:

Thanks for your help,

Dermot

Hi, Dermot,

I tried running big data using SDK 4.3 prebuilt images and big data binaries but not seeing any issue. My modified cmem dts file not for running big data example, but for other application which requires adding an extra cmem block at 0xb000 0000. which you can see in my log when dump the /proc/cmem file. Hope this change does not confuse you from running the big data example.

Did you use the prebuilt big data binaries or you rebuilt it? Did you download the big data DSP image by changing the soft link to big data DSP image? Below is my logs

am57xx-evm login: root
root@am57xx-evm:~# [ 19.054186] omap_hwmod: mmu1_dsp2: _wait_target_disable failed
[ 19.066693] omap_hwmod: mmu0_dsp2: _wait_target_disable failed

root@am57xx-evm:~# uname -a
Linux am57xx-evm 4.9.69-g9ce43c71ae #2 SMP PREEMPT Tue May 22 11:09:56 EDT 2018 armv7l GNU/Linux
root@am57xx-evm:~#
root@am57xx-evm:~#
root@am57xx-evm:~#
root@am57xx-evm:~# cat /proc/cmem

Block 0: Pool 0: 1 bufs size 0xc000000 (0xc000000 requested)

Pool 0 busy bufs:

Pool 0 free bufs:
id 0: phys addr 0xa0000000

Block 1: Pool 0: 1 bufs size 0xc000000 (0xc000000 requested)

Pool 0 busy bufs:

Pool 0 free bufs:
id 0: phys addr 0xb0000000
root@am57xx-evm:~# lsmod | grep cmem
cmemk 35419 0
root@am57xx-evm:~# dmesg | grep cmem
[ 3.984692] cmemk: loading out-of-tree module taints kernel.
[ 4.003672] cmemk initialized
root@am57xx-evm:~# ls -l /lib/firmware/dra7-dsp1-fw.xe66
lrwxrwxrwx 1 root root 54 Mar 26 15:34 /lib/firmware/dra7-dsp1-fw.xe66 -> /usr/bin/simple_buffer_example/release/server_dsp.xe66
root@am57xx-evm:~# /usr/bin/simple_buffer_example/release/app_host DSP1
--> main:
[ 101.303934] omap_hwmod: mmu0_dsp2: _wait_target_disable failed
[ 101.309829] omap-iommu 41501000.mmu: 41501000.mmu: version 3.0
[ 101.317843] omap-iommu 41502000.mmu: 41502000.mmu: version 3.0
--> Main_main:
--> App_create:
App_create: Host is ready
<-- App_create:
--> App_exec:
CMEM_init success
CMEM_getPool success
CMEM_allocPool success: Allocated buffer 0xaa550000
SharedRegion_setup success
App_taskFxn: SR_1, base 0xaa550000, len=1000000
HeapMem_setup success
HeapMem_create success
App_taskFxn: SR_1 heap, totalSize=16777216,totalFreeSize=16777216,largestFreeSize=16777216
App_taskFxn: SR_1 heap, buf=0x0xaa550080,size=16777216
App_exec: sending message 1
Shared memory phys Addr ffffffffa0000000
App_exec: sending message 2
App_exec: sending message 3
App_exec: message received 1
App_exec: Preparing message 4
App_exec: Sending message 4
App_exec: message received 2
App_exec: Preparing message 5
App_exec: Sending message 5
App_exec: message received 3
App_exec: Preparing message 6
App_exec: Sending message 6
App_exec: message received 4
App_exec: Preparing message 7
App_exec: Sending message 7
App_exec: message received 5
App_exec: Preparing message 8
App_exec: Sending message 8
App_exec: message received 6
App_exec: Preparing message 9
App_exec: Sending message 9
App_exec: message received 7
App_exec: Preparing message 10
App_exec: Sending message 10
App_exec: message received 8
App_exec: Preparing message 11
App_exec: Sending message 11
App_exec: message received 9
App_exec: Preparing message 12
App_exec: Sending message 12
App_exec: message received 10
App_exec: Preparing message 13
App_exec: Sending message 13
App_exec: message received 11
App_exec: Preparing message 14
App_exec: Sending message 14
App_exec: message received 12
App_exec: Preparing message 15
App_exec: Sending message 15
App_exec: message received 13
App_exec: Preparing message 16
App_exec: Sending message 16
App_exec: message received: 14
App_exec: message received: 15
App_exec: message received: 16
App_exec: Data check clean
<-- App_exec: 0
--> App_delete:
<-- App_delete:
<-- Main_main:

Host: Test Passed
<-- main:
root@am57xx-evm:~# [ 112.487408] omap_hwmod: mmu1_dsp2: _wait_target_disable failed
[ 112.501275] omap_hwmod: mmu0_dsp2: _wait_target_disable failed

root@am57xx-evm:~#

Hi Rex,

Thanks for your reply. I'm using the SDK 4.3 pre-built images only. I create an SD card using the create-sdcard.sh script with two partitions, one for boot (containing MLO, uEnv.txt and u-boot.img) and one for root (containing the rootfs).

Even without attempting to run the Big Data IPC demo, I think my problem is the empty /proc/cmem. Does this get populated by the cmemk module when it loads? I tried unloading and reloading with insmod but I just used the sample parameters provided at processors.wiki.ti.com/.../CMEM_Overview however phys_start and phys_end get reported as unknown parameters. Where might I find the parameters used to load it from the rootfs? I presume it gets loaded somewhere but I'm having trouble finding where.

Thanks,
Dermot

Hi, Dermot,

Using command line to configure CMEM was from much older release (MCSDK or may be Proc SDK 1.x or 2.x) though it should still work in the current release. The configuration has benn moved to using device tree and is in am57xx-evm-cmem.dtsi file. It's done during kernel boot using prebuilt images without any changes. Let me scratch my SD card and retry it so we are on the same page.

Rex

Hi Rex,

Yes - I just realised that it seems to get loaded without parameters from looking at the /etc/modules-load.d/cmemk.conf file.

The cmem configuration in my am57xx-evm-cmem.dtsi file is:

cmem {
compatible = "ti,cmem";
#address-cells = <1>;
#size-cells = <0>;

#pool-size-cells = <2>;

status = "okay";

cmem_block_0: cmem_block@0 {
reg = <0>;
memory-region = <&cmem_block_mem_0>;
cmem-buf-pools = <1 0x0 0x0c000000>;
};

cmem_block_1: cmem_block@1 {
reg = <1>;
memory-region = <&cmem_block_mem_1_ocmc3>;
};
};


Is there anyway that I can confirm that this dts include has made it into my actual booted system? I'm guessing because it's .dtsi and not .dts, there's not point in my looking for a corresponding .dtb file?

Thanks,
Dermot

Hi, Dermot,

The dtsi file is included by the dts file. Check your u-boot logs to find out what board type it is and find the corresponding dts file.

CPU : DRA752-GP ES2.0
Model: TI AM572x EVM Rev A3
Board: AM572x EVM REV A.30

In my case, it is am57xx-evm-reva3.dts

You can also find them in /proc/device-tree/reserved-memory.

Rex

Hi Rex,

Yes I think I can confirm that the dts file is am57xx-beagle-x15-revb1.dts.  It has a corresponding .dtb file in /boot on my device.

It contains a model string : model = "TI AM5728 BeagleBoard-X15 rev B1";

which matches what gets posted during boot (presumably by u-boot):

Presumably this matching string is how the correct .dtb file gets selected during the boot process?

The last line of this dts file includes this cmem dtsi file:

#include "am57xx-evm-cmem.dtsi"

I'm still a newbie when it comes to u-boot and device tree but I think this looks OK?

Thanks & I really appreciate your help!

Dermot

Hey Rex,

I'll try building out an image again from the pre-built components. I had been stopping (and disabling) the ti-mct-daemon.service as per the "Running IPC Examples" link.

Could you do me a favour and check that if you reboot your system, you can see content in /proc/cmem regardless of the state of the mct daemon?

Thanks,
Dermot

Hi, Dermot,

Just noticed that my post this morning isn't shown in the thread for some reason. Let me recollect them.

You don't need to rebuild the images. All binaries are included in the filesystem under /usr/bin/simple_buffer_example. Check my logs to see exact where they are located.

Yes, I see content in /proc/cmem without killing ti-mctd daemon.

Rex

Hi Rex,

Yeah - same thing happened to my post yesterday - after spending some time writing detail, collecting screenshots, etc. the post seemed to get submitted OK but when I refreshed the page it was nowhere to be seen.  So, I ended up posting a more concise version  2nd time!

Anyway, I decided to rebuild the kernel, modules and device tree binary just to see if that made a difference.

Unfortunately, the "sudo make ARCH=arm INSTALL_MOD_PATH=/media/rootfs modules_install" command from the link below doesn't actually install the modules in the INSTALL_MOD_PATH.

http://software-dl.ti.com/processor-sdk-linux/esd/docs/latest/linux/Foundational_Components.html#kernel 

In fact, even to run with sudo, I had to change the command to avoid gcc being used instead of the arm cross compiler:

sudo bash -c ' export CROSS_COMPILE=../../linux-devkit/sysroots/x86_64-arago-linux/usr/bin/arm-linux-gnueabihf- ; make ARCH=arm INSTALL_MOD_PATH=/home/dmurphy/my_build/modules/ modules_install'

Regardless of whether I use sudo or not to make modules_install, nothing ends up in the INSTALL_MOD_PATH directory.  The Make just ends with the following output:

...

CC sound/soc/generic/snd-soc-simple-card-utils.mod.o
LD [M] sound/soc/generic/snd-soc-simple-card-utils.ko
CC sound/soc/generic/snd-soc-simple-card.mod.o
LD [M] sound/soc/generic/snd-soc-simple-card.ko
CC sound/soc/omap/snd-soc-omap-hdmi-audio.mod.o
LD [M] sound/soc/omap/snd-soc-omap-hdmi-audio.ko
CC sound/usb/snd-usb-audio.mod.o
LD [M] sound/usb/snd-usb-audio.ko
CC sound/usb/snd-usbmidi-lib.mod.o
LD [M] sound/usb/snd-usbmidi-lib.ko
dmurphy@tibuilder:~/ti-processor-sdk-linux-am57xx-evm-04.03.00.05/board-support/linux-4.9.69+gitAUTOINC+9ce43c71ae-g9ce43c71ae$

I may have to raise this issue under another E2E support request.

 

Dermot,

If gcc is called instead of the cross-compiler, it means your PATH isn't set to point to the cross-compiler. Be sure it points to the directory containing arm-linux-gnueabihf-gcc.

I didn't have issue when I did the installation. Do you see contents in /media/rootfs or where the mount point is? Be sure the mount point is correct.

Rex