check-ipc-flood.sh does not work with IPC4

[marc-hb]

Example in https://sof-ci.01.org/softestpr/PR1057/build454/devicetest/index.html (or in Intel daily test run 28116)

2023-06-15 17:55:26 UTC _ /home/ubuntu/sof-test/test-case/check-ipc-flood.sh need debug version firmware
2023-06-15 17:55:26 UTC _ Starting func_exit_handler(), exit status=2, FUNCNAME stack:
2023-06-15 17:55:26 UTC _  main()  @  /home/ubuntu/sof-test/test-case/check-ipc-flood.sh

cc:

• file system detect for setup default debugfs for ipc flood test #191
• Auxiliary devices in SOF break check-ipc-flood.sh (IPC3) #467
• [FEATURE] Test IPC large message IO whilst playback and capture are active #847

[ranj063]

I'm not sure if flood test will ever work with IPC4. This requires the addition of a new kind of IPC into IPC4 and I dont think there are any plans for it. Can we simply omit these from testing for IPC4?

[marc-hb]

Can we simply omit these from testing for IPC4?

We sure can if everyone agrees. Where can we bring that question up?

BEFORE we remove that test from the corresponding test plans we need to fix the misleading error message. We can probably just add something like if ipc4; then skip_test "only IPC3 is supported"

[ranj063]

Can we simply omit these from testing for IPC4?

We sure can if everyone agrees. Where can we bring that question up?

BEFORE we remove that test from the corresponding test plans we need to fix the misleading error message. We can probably just add something like if ipc4; then skip_test "only IPC3 is supported"

yes, lets just do that for now

[marc-hb]

wontfix

[marc-hb]

I'll see if I can quickly fix the error message.

[marc-hb]

So to decide whether it should test or SKIP, the short (and pretty poorly written) check-ipc-flood.sh currently looks for this in the kernel logs:

Aug 13 18:08:19 jf-cml-sku0983-sdw-1 kernel: sof-audio-pci-intel-cnl 0000:00:1f.3: Firmware info: version 2:2:10-e05ca
Aug 13 18:08:19 jf-cml-sku0983-sdw-1 kernel: sof-audio-pci-intel-cnl 0000:00:1f.3: Firmware: ABI 3:23:0 Kernel ABI 3:23:0
Aug 13 18:08:19 jf-cml-sku0983-sdw-1 kernel: sof-audio-pci-intel-cnl 0000:00:1f.3: Firmware debug build 65535 on dtermin.-extman - options:
                                              GDB: disabled
                                              lock debug: disabled
                                              lock vdebug: disabled

Instead of this:

Aug 13 11:17:35 up2 kernel: sof-audio-pci-intel-apl 0000:00:0e.0: Firmware info: version 2:2:0-57864
Aug 13 11:17:35 up2 kernel: sof-audio-pci-intel-apl 0000:00:0e.0: Firmware: ABI 3:22:1 Kernel ABI 3:23:0

In other words, if Firmware debug is found in the kernel log then it runs, otherwise it skips. (It does not even look for sof-audio... or ... build, sigh...)

This is quite ugly.

Is there some more reliable and hopefully generic (IPC-independent) way to check whether this IPC is supported?

sound/soc/sof/ipc3.c

[         943.437] (         142.500) c0 ipc                  src/ipc/ipc3/handler.c:1605 INFO ipc: new cmd 0xb0010000
[         951.771] (           8.333) c0 ipc                  src/ipc/ipc3/handler.c:1650 ERROR ipc: unknown command type 2952790016

EDIT: this is based on this kernel code, so it's not as "ugly" as I thought it was:

int sof_ipc3_validate_fw_version(struct snd_sof_dev *sdev)
{
        dev_info(sdev->dev,
                 "Firmware info: version %d:%d:%d-%s\n",  v->major, v->minor,
                 v->micro, v->tag);
        dev_info(sdev->dev,
                 "Firmware: ABI %d:%d:%d Kernel ABI %d:%d:%d\n",

 ....

        if (ready->flags & SOF_IPC_INFO_BUILD) {
                dev_info(sdev->dev,
                         "Firmware debug build %d on %s-%s - options:\n"
                         " GDB: %s\n"

[marc-hb]

How about this:

root@up2:~# echo 1 > /sys/kernel/debug/sof/ipc_flood_count ; echo $?
-bash: echo: write error: Invalid argument
1

In other words, a "flood" of just 1 just to probe the feature. If that works, then flood "for real" with either the 10,000 default or the command line argument. Otherwise SKIP.

The risk is wrongly SKIPping because the entire flooding feature is broken (when it shouldn't be) and it fails even with 1.

cc: @ujfalusi

[marc-hb]

Here's a much smaller and simpler SKIP message fix, please review:

• check-ipc-flood.sh: stop asking Zephyr users to use CONFIG_DEBUG #1227

[marc-hb]

Just noticed check-ipc-flood.sh does not care about firmware logs...

[ujfalusi]

echo 1 > /sys/kernel/debug/sof/snd_sof.ipc_flood.0/ipc_flood_count 
-bash: echo: write error: Invalid argument

and in kernel:

[ 1579.215851] snd_sof:ipc3_log_header: sof-audio-pci-intel-tgl 0000:00:1f.3: ipc tx: 0xb0010000: GLB_TEST_MSG: IPC_FLOOD
[ 1579.215969] sof-audio-pci-intel-tgl 0000:00:1f.3: ipc tx error for 0xb0010000 (msg/reply size: 8/0): -22
[ 1579.215977] snd_sof_ipc_flood_test snd_sof.ipc_flood.0: ipc flood test failed at 0 iterations

so, the firmware have the IPC flood test disabled, because the firmware is not built with CONFIG_DEBUG, there is not much the kernel do here, it cannot know this...

[marc-hb]

Thanks for the additional information @ujfalusi.

Better error message just merged with a reference to this issue. Closing as wontfix for now.