Linux/AM5728: CMEM error

netrover

Part Number: AM5728

Tool/software: Linux

Hello,

I am getting an error trying to communicate between Linux and DSP1. Here are the messages running the program on the host:

-------------------------------------

root@am57xx-evm:~/test/messageq/hostex# ./hostex2 DSP1
--> main:
--> Main_main:
--> App_create:
CMEM_init success
MessageQ_Params_init
MessageQ_create
queue name DSP1:MsgQ:01
MessageQ_open
App_create: Host is ready
<-- App_create:
--> App_exec:
App_exec: sending message
user: 0xaa5a2000 phys: 0xa0000000
Data 0 in xor = 0xad[ 2211.926600] CMEMK Error: Failed to find a pool which fits 0x70

CMEM Error: getPool: Failed to get a pool fitting a size 0x70[ 2211.943117] CMEMK Error: ALLOC: invalid pool (-1) passed.

CMEM Error: allocPool: ioctl CMEM_IOCALLOC failed from pool -1: -1
CMEM_alloc() failed (returned NULL)
<-- App_exec: -1
<-- Main_main:
<-- main:
root@am57xx-evm:~/test/messageq/hostex#

-------------------------------------

I had checked the /sys/module/cmemk/parameters/phys_start and phys_end files via 'cat' and they showed 'null'. I then ran 'rmmod cmemk' and ran insmod this way,

insmod /lib/modules/4.4.19-rt25-gf572d285f0/extra/cmemk.ko phys_start=0xa0000000 phys_end=0xac000000 pools=4x50331648

The following messaged occured,

-------------------------------------

[ 6264.961831] CMEMK module: reference Linux version 4.4.19
[ 6264.969162] allocated heap buffer 0x40500000 of size 0x100000
[ 6264.975069] cmemk initialized
-------------------------------------

I checked phys_start and phys_end again, which now show 0xa0000000 and 0xac000000 respectively.

over 7 years ago

0 netrover over 7 years ago

Intellectual 850 points

More information. Using the ger_zel example with resource table for host and DSP1 only.

bitbucket.org/.../cmem.git

The error is happen in the second loop of the allocation. The first loop was find even for a PAYLOADSIZE of 0x700000. Here are the messages with a comment added in the loop followed by the loop code getting the error, (this occurs for both PAYLOADSIZE values used, 0x700000 and 0x70).
---------------------------------------------
root@am57xx-evm:~/test/messageq/hostex# ./hostex2 DSP1
--> main:
--> Main_main:
--> App_create:
CMEM_init success
MessageQ_Params_init
MessageQ_create
queue name DSP1:MsgQ:01
MessageQ_open
App_create: Host is ready
<-- App_create:
--> App_exec:
App_exec: sending message

At loop 0 of 4.

user: 0xaa5b5000 phys: 0xa0000000
Data 0 in xor = 0xc6

At loop 1 of 4.

[15557.259602] CMEMK Error: Failed to find a pool which fits 0x700000
CMEM Error: getPool: Failed to get a pool fitting a size 0x700000[15557.269706] CMEMK Error: ALLOC: invalid pool (-1) passed.

CMEM Error: allocPool: ioctl CMEM_IOCALLOC failed from pool -1: -1

CMEM_alloc() failed (returned NULL)
<-- App_exec: -1
<-- Main_main:
<-- main:
root@am57xx-evm:~/test/messageq/hostex#
---------------------------------------------

The code segment
---------------------------------------------
for (i = 0; i < 4; i++) {

printf("\n\tAt loop %d of 4.\n\n", i);
fflush(stdout);

msg->dataIn[i] = CMEM_allocPool(CMEM_getPool(PAYLOADSIZE), &cmemAttrs);
if (msg->dataIn[i] == NULL) {
printf("\nCMEM_alloc() failed (returned NULL)\n");
status = -1;
goto leave;
}
msg->dataInPhys[i] = CMEM_getPhys(msg->dataIn[i]);

printf("user: %p phys: 0x%x\n", msg->dataIn[i], msg->dataInPhys[i]);

uint8_t *dataIn = (uint8_t*)msg->dataIn[i];

msg->inXor[i] = 0;

for(j = 0; j < PAYLOADSIZE; j++) {
dataIn[j] = rand();
msg->inXor[i] ^= dataIn[j];
}

printf("Data %d in xor = 0x%x\n", i, msg->inXor[i]);
}
---------------------------------------------

0 Biser Gatchev-XID over 7 years ago in reply to netrover

TI__Guru**** 393215 points

The software team have been notified. They will respond here.

0 Rex Chang over 7 years ago in reply to Biser Gatchev-XID

TI__Guru 50170 points

Were you trying to run TI IPC ex02 messageQ example?

0 netrover over 7 years ago in reply to Rex Chang

Intellectual 850 points

I used the host/DSP1 example posted here,

bitbucket.org/.../cmem.git

0 netrover over 7 years ago in reply to netrover

Intellectual 850 points

Also, I am wondering if the resource table included in the example, rsc_table_dsp.h, has some conflicts within all of the define statements.
Thanks

0 Rex Chang over 7 years ago in reply to netrover

TI__Guru 50170 points

Shouldn't, but last time I ran this example was 1 or 2 releases ago. I'll retry it with the same release as yours (ProcSDK 3.2, kernel v4.4) and see what's going on.

0 netrover over 7 years ago in reply to Rex Chang

Intellectual 850 points

Hello,

I noticed that with I removed and then insmod cmemk again with the following command,

insmod /lib/modules/4.4.19-rt25-gf572d285f0/extra/cmemk.ko phys_start=0xa0000000 phys_end=0xac000000 pools=4x50331648

And then cat /proc/cmem, I still get the default 1 pool that occurs during the boot process,
------------------------------------------------------------------------------
Block 0: Pool 0: 1 bufs size 0xc000000 (0xc000000 requested)

Pool 0 busy bufs:

Pool 0 free bufs:
id 0: phys addr 0xa0000000
-----------------------------------------------------------------------------
I was looking at the cmemk/parameters file, which shows what I used with insmod,

cat /sys/module/cmemk/parameters/pools --- shows ----> 4x50331648

Could this be related to getting only one pool? I though a pool could be used more than once if enough memory is available?

Thanks,
Dan

0 netrover over 7 years ago in reply to netrover

Intellectual 850 points

Ok, I changed the cmem dtsi file and created 8 pools of equal size using all of the available memory.
Creating four pools now works. Now the the second DSP2 will not load using the same resource table that points to the DSP2 base address of 0x9f000000.

This is the output allocating buffers in four different pools:
-----------------------------------------------------------------------------
root@am57xx-evm:~/test/messageq/hostex# ./hostex2 DSP1
--> main:
--> Main_main:
--> App_create:
CMEM_init success
MessageQ_Params_init
MessageQ_create
queue name DSP1:MsgQ:01
MessageQ_open
App_create: Host is ready
<-- App_create:
--> App_exec:
App_exec: sending message

At loop 0 of 4.

user: 0xb4e41000 phys: 0xaa800000

inXor: 0
msg->inXor[i]: 0
inXor: 16
msg->inXor[i]: 0
At loop 1 of 4.

user: 0xb3641000 phys: 0xa9000000

inXor: 0
msg->inXor[i]: 0
inXor: 16
msg->inXor[i]: 0
At loop 2 of 4.

user: 0xb1e41000 phys: 0xa7800000

inXor: 0
msg->inXor[i]: 0
inXor: 16
msg->inXor[i]: 0
At loop 3 of 4.

user: 0xb0641000 phys: 0xa6000000

inXor: 0
msg->inXor[i]: 0
inXor: 16
msg->inXor[i]: 0
inXor: 16
msg->inXor[0]: 0
Message put done.

Message get done.

inXor: 16
msg->inXor[0]: 16
<-- App_exec: 0

--> App_delete:
<-- App_delete:
<-- Main_main:
<-- main:
-----------------------------------------------------------------------------
Thanks

0 netrover over 7 years ago in reply to netrover

Intellectual 850 points

Also, a clarification may be needed in configuration the cmem pools.
If cmem is configured in a dtsi file that also means the 'insmod cmemk ...' command cannot update the pools configuration?
Thanks

0 Rex Chang over 6 years ago in reply to netrover

TI__Guru 50170 points

I don't see any issue running the ex02_messageq example. The relevant logs from my run are shown below:

[ 9.160260] remoteproc3: powering up 41000000.dsp
[ 9.192047] remoteproc3: Booting fw image dra7-dsp2-fw.xe66, size 4528816
[ 9.222290] omap_hwmod: mmu0_dsp2: _wait_target_disable failed
[ 9.228186] omap-iommu 41501000.mmu: 41501000.mmu: version 3.0
[ 9.234131] omap-iommu 41502000.mmu: 41502000.mmu: version 3.0
[ 9.297451] remoteproc3: remote processor 41000000.dsp is now up
[ 9.313289] virtio_rpmsg_bus virtio2: rpmsg host is online
[ 9.313672] virtio_rpmsg_bus virtio2: creating channel rpmsg-proto addr 0x3d
[ 9.333941] remoteproc2: powering up 40800000.dsp
[ 9.356598] remoteproc2: Booting fw image dra7-dsp1-fw.xe66, size 4681692
[ 9.378185] omap_hwmod: mmu0_dsp1: _wait_target_disable failed
[ 9.384076] omap-iommu 40d01000.mmu: 40d01000.mmu: version 3.0
[ 9.390011] omap-iommu 40d02000.mmu: 40d02000.mmu: version 3.0
[ 9.440844] remoteproc2: remote processor 40800000.dsp is now up
[ 9.458507] virtio_rpmsg_bus virtio3: rpmsg host is online

root@am57xx-evm:/lib/firmware# ls -l dra7*.xe66
lrwxrwxrwx 1 root root 25 Apr 26 2017 dra7-dsp1-fw.xe66 -> ex02_server_dsp1_320.xe66
lrwxrwxrwx 1 root root 25 Apr 26 2017 dra7-dsp2-fw.xe66 -> ex02_server_dsp2_320.xe66

root@am57xx-evm:~# uname -a
Linux am57xx-evm 4.4.32-gadde2ca9f8 #8 SMP PREEMPT Fri Apr 7 14:16:56 EDT 2017 armv7l GNU/Linux

root@am57xx-evm:~# ./app_host_ex02_320 DSP1
--> main:
--> Main_main:
--> App_create:
App_create: Host is ready
<-- App_create:
--> App_exec:
App_exec: sending message 1
App_exec: sending message 2
App_exec: sending message 3
App_exec: message received, sending message 4
App_exec: message received, sending message 5
App_exec: message received, sending message 6
App_exec: message received, sending message 7
App_exec: message received, sending message 8
App_exec: message received, sending message 9
App_exec: message received, sending message 10
App_exec: message received, sending message 11
App_exec: message received, sending message 12
App_exec: message received, sending message 13
App_exec: message received, sending message 14
App_exec: message received, sending message 15
App_exec: message received
App_exec: message received
App_exec: message received
<-- App_exec: 0
--> App_delete:
<-- App_delete:
<-- Main_main:
<-- main:
root@am57xx-evm:~#

root@am57xx-evm:~# lsmod | grep -ir cmem
cmemk 31462 0

You can try to follow the instructions in IPC Quick Start Guide at processors.wiki.ti.com/.../Processor_SDK_IPC_Quick_Start_Guide

You do need to build the IPC library on ARM Linux side before building the IPC examples on DSP RTOS side. If the RTOS SDK and CCS are not installed at the default folders, then TI_RTOS_PATH and TI_CCS_PATH need to be exported.

The ex02_messageq example is included in RTOS SDK and under RTOS_SDK/IPC_xx_xx_xx/examples/DRA7_linux_elf/ex02_messageq

I am using ProcSDK 3.2 release which has kernel 4.4.32. Last time I ran the example was using ProcSDK 3.1 which is the same kernel version as yours, v4.4.19.

Rex

Processors

Processors forum

Linux/AM5728: CMEM error