Linux/AM5728: AM572x - Access IPC between A15 and M4 using docker

Hi, Christine,

What ProcSDK release version are you based on? We have not tried running Ubuntu on top of AM5728 EVM. Does it run without the Ubuntu?
Is this application TI IPC example? If not, could you try to see if it works with TI IPC ex02_messageq example. The prebuilt images are in

root@am57xx-evm:/usr/bin/ipc/examples/ex02_messageq/debug# ls
app_host server_dsp2.xe66 server_ipu2.xem4
server_dsp1.xe66 server_ipu1.xem4

Rex

Hi Rex,

We are running on a Linux built by Yocto using the 5.0.0.15 SDK with meta-virtualization layer into for docker.

Yes the application does run natively outside of a docker container.

The application is based on the TI IPC example.  Basically I took the ex02_messageq example and modified it for our purposes.

Our guess is permissions issues on  /tmp/LAD from the docker container.  If you have other ideas, please let us know.

Thanks,

Christine

Hi, Christine,

Let me discuss internally to see what the ubuntu docker container would possibly have anything to do with it.

Rex

Hi,

We have added the meta-virtualization layer to our Yocto image to support docker. We are running the LAD daemon natively in Linux and the IPC test application is running in Docker and can connect to the LAD daemon as shown here:

Initializing LAD... [0.602339] NameServer_setup: entered, refCount=0
[0.602365] NameServer_setup: creating listener thread
[0.602480] NameServer_setup: exiting, refCount=1
[0.602574] listener_cb: Entered Listener thread.
[0.602590] NameServer: waiting for unblockFd: 0, and socks: maxfd: 0
[0.602596] NameServer_create(): 'GateMP'
[0.602648]
opening FIFO: /tmp/LAD/LADCMDS
[143.359481] Retrieving command...
[143.359844]
LAD_CONNECT:
[143.359861] client FIFO name = /tmp/LAD/106
[143.359874] client PID = 106
[143.359886] assigned client handle = 0
[143.359932] FIFO /tmp/LAD/106 created
[143.360074] FIFO /tmp/LAD/106 opened for writing
[143.360124] sent response
[143.360135] DONE
[143.360144] Retrieving command...
[143.360244]
DETECTED CONNECTED CLIENT #0 HAS DEPARTED![143.360257]
Doing DISCONNECT on behalf of client...[143.360265]
closing FIFO /tmp/LAD/106 (filePtr=0x32808)
[143.360286] done, unlinking /tmp/LAD/106

however, the IPC test application encounters a HW Spin Lock failure:

root@am57xx-evm:/home# ./app_host ipu2
--> main:
Ipc_start: GateHWSpinlock_start failed: -10
Ipc_start failed: status = -1
<-- main:

The -10 error code looks like the same error as we looked at yesterday can't open /dev/mem
or mapping error.

source code:
/* Fall back to /dev/mem if hwspinlock_user driver is not supported */
Mod->fd = open("/dev/hwspinlock", O_RDWR);
if (Mod->fd < 0) {
Mod->fd = open ("/dev/mem", O_RDWR | O_SYNC);
}
else {
Mod->useHwlockDrv = true;
}

if (Mod->fd < 0){
LOG0("GateHWSpinlock_start: failed to open the spinlock device");
status = GateHWSpinlock_E_OSFAILURE;
}

/* map the hardware lock registers into the local address space */
if (!Mod->useHwlockDrv && status == GateHWSpinlock_S_SUCCESS) {
dst = (UInt32)mmap(NULL, _GateHWSpinlock_cfgParams.size,
(PROT_READ | PROT_WRITE),
(MAP_SHARED), Mod->fd,
(off_t)_GateHWSpinlock_cfgParams.baseAddr);

if (dst == (UInt32)MAP_FAILED) {
LOG0("GateHWSpinlock_start: Memory map failed")
status = GateHWSpinlock_E_OSFAILURE;
close(Mod->fd);
Mod->fd = -1;
}
else {
Mod->baseAddr = (UInt32 *)(dst + _GateHWSpinlock_cfgParams.offset);
status = GateHWSpinlock_S_SUCCESS;
}
}


Any tips would be appreciated.

Thanks,
Christine

We are sharing the ti ipc libraries into the docker image and the libraries are being found and used, but encounter the HW Spin Lock failure.