host.Init forked processes persist and maintain high CPU load

[BrockMatthews]

Describe the bug Processes spawned by host.Init() when using ftdi maintain a high CPU load on multiple threads that persists.

To Reproduce Steps to reproduce the behavior:

1. Run program

   
   package main

import ( "log" "runtime" "time"

"periph.io/x/host/v3"
"periph.io/x/host/v3/ftdi"

)

func main() { host.Init() log.Println("Init done.")

ftdi.All()

buffer := make([]byte, 1024)
runtime.Stack(buffer, true)

log.Println("Running goroutines:")
log.Println(string(buffer))

time.Sleep(1 * time.Hour)

}

2. Run it.

2023/10/04 00:05:43 Init done. 2023/10/04 00:05:43 Running goroutines: 2023/10/04 00:05:43 goroutine 1 [running]: main.main() /home/name/go/src/periphio_test/test.go:19 +0x78

3. Check process

top -H -p

Threads: 12 total, 0 running, 12 sleeping, 0 stopped, 0 zombie %Cpu(s): 4.7 us, 19.3 sy, 0.0 ni, 75.7 id, 0.0 wa, 0.0 hi, 0.3 si, 0.0 st KiB Mem : 3074384 total, 2296660 free, 562180 used, 215544 buff/cache KiB Swap: 0 total, 0 free, 0 used. 2406060 avail Mem

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
5264 name 20 0 1327928 14104 3280 S 13.0 0.5 0:08.36 test_aa
5278 name 20 0 1327928 14104 3280 S 12.6 0.5 0:08.72 test_aa
5277 name 20 0 1327928 14104 3280 S 10.0 0.5 0:06.38 test_aa
5262 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.01 test_aa
5263 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa
5265 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa
5266 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.01 test_aa
5267 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa
5268 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa
5269 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa
5270 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa
5279 name 20 0 1327928 14104 3280 S 0.0 0.5 0:00.00 test_aa

**Expected behavior**
No processes to persist and produce high CPU load after host.Init() completion. 

**Platform (please complete the following information):**
 - OS: NXP i.MX Release Distro 4.14-sumo (sumo)
 - Board: custom

**Additional context**
Device: FTDI FT232H

Device is completely usable after host.Init()

ftdi_sio driver unbound from used FT232H, but module not removed as it is required by other FTDI devices on the system.

[maruel]

This is surprising. Can you use https://pkg.go.dev/runtime/pprof to try to determine what is spinning? Keep in mind the ftdi driver is known to not be at the quality bar that the project generally aims to achieve.

[BrockMatthews]

Thanks. The profiling output isn't awfully specific but here it is

Duration: 10.21s, Total samples = 5.66s (55.44%)
Entering interactive mode (type "help" for commands, "o" for options)
(pprof) top
Showing nodes accounting for 5660ms, 100% of 5660ms total
      flat  flat%   sum%        cum   cum%
    2390ms 42.23% 42.23%     2390ms 42.23%  [libpthread-2.27.so]
    2330ms 41.17% 83.39%     2330ms 41.17%  runtime._LostExternalCode
     840ms 14.84% 98.23%      840ms 14.84%  [libc-2.27.so]
     100ms  1.77%   100%      100ms  1.77%  [PeriphioTest_aarch64]
         0     0%   100%     5660ms   100%  runtime._ExternalCode

Please let me know of anything else I can provide.

[maruel]

Sorry I won't be able to diagnose with this data. You will likely need to investigate more deeply yourself.

[BrockMatthews]

On deeper inspection the issue seems to be caused by the ftd2xx driver with the call to FT_Open in d2xx_posix.go open(). This happens specifically on the device that is used and has ftdi_sio unbound, and on none of the other FTDI devices iterated through.