Compilation to arm fails

[ergo-services]

Compilation to arm fails

[halturin]

To be specific, I have tested it on aarch64 and all tests passed. I will try to find the same hardware. It's a bit difficult to resolve it with no access to the similar environment.

…

On Fri, Apr 22, 2022, 10:32 Szymon Wygnański ***@***.***> wrote: Thanks for your response. Indeed it compiles fine for arm64. However when I try to run it on Asus Tinker Board T it hangs and after a while the asus shuts off. To reproduce: cd ergo/examples/simple GOARCH=arm64 go build simple.go // scp to arm64 device on the arm64 device: ./simple The result: it hangs. I have to restart the device. Occasionally I get this error log: Message from ***@***.*** at Apr 22 08:13:22 ... kernel:[ 64.830251] Internal error: undefined instruction: 0 [#1] PREEMPT SMP Message from ***@***.*** at Apr 22 08:13:22 ... kernel:[ 64.964868] Process simple (pid: 4048, stack limit = 0xffff000013498000) Message from ***@***.*** at Apr 22 08:13:22 ... kernel:[ 65.199004] Code: d2800014 54000be1 b9404004 f9401c03 (23232323) However usually there is no log at all. I debugged it a little and found that it hangs on this line: https://github.com/ergo-services/ergo/blob/79bebaa/proto/dist/resolver.go#L225 However the value of the dsn variable seams to be correct: localhost:4369 On my local PC it works fine (but I have erlang installed) but on the arm machine it does not. Could you give me some clues, I would really appreciate it. — Reply to this email directly, view it on GitHub <#102 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA5ATFEMHDT77JI5ZYZS5LVGJP2ZANCNFSM5T7VMSSQ> . You are receiving this because you commented.Message ID: ***@***.***>

[finalclass]

Sure I will try it out. However I'm OOO currently and will be able to check it early next week.

[halturin]

Thanks for the report. I have tested it on arm64 only. Will take a look

[finalclass]

Thanks for your response. Indeed it compiles fine for arm64. However when I try to run it on Asus Tinker Edge T it hangs and after a while the asus shuts off.
To reproduce:

cd ergo/examples/simple
GOARCH=arm64 go build simple.go
// scp to arm64 device

on the arm64 device:

./simple

The result: it hangs. I have to restart the device.
Occasionally I get this error log:

Message from syslogd@vexing-eft at Apr 22 08:13:22 ...
 kernel:[   64.830251] Internal error: undefined instruction: 0 [#1] PREEMPT SMP

Message from syslogd@vexing-eft at Apr 22 08:13:22 ...
 kernel:[   64.964868] Process simple (pid: 4048, stack limit = 0xffff000013498000)

Message from syslogd@vexing-eft at Apr 22 08:13:22 ...
 kernel:[   65.199004] Code: d2800014 54000be1 b9404004 f9401c03 (23232323)

However usually there is no log at all.

I debugged it a little and found that it hangs on this line: https://github.com/ergo-services/ergo/blob/79bebaa/proto/dist/resolver.go#L225
However the value of the dsn variable seams to be correct: localhost:4369

On my local PC it works fine (but I have erlang installed) but on the arm machine it does not. Could you give me some clues, I would really appreciate it.

[halturin]

Sorry for the delay. I still don't have anything similar to your device.

May I ask you to try these fixes:
/ergo-services/[email protected]/lib/tools.go:166
replace:

  limit = 4294967000

by

  limit = math.MaxInt

it also requires to add "math" module to the import section.

And update ResourceUsage function in ergo-services/[email protected]/lib/osdep/linux.go
with this code

	var usage syscall.Rusage
	var utime, time int64
	if err := syscall.Getrusage(syscall.RUSAGE_SELF, &usage); err == nil {
		utime = int64(usage.Utime.Sec)*1000000000 + usage.Utime.Nano()
		stime = int64(usage.Stime.Sec)*1000000000 + usage.Stime.Nano()
	}
	return utime, stime

Or you can try this branch https://github.com/ergo-services/ergo/tree/fixarm

[finalclass]

Hi,
Unfortunately this does not seam to help. I've added go.mod to ergo/examples/simple:

module simpl.com/simple

go 1.18

replace github.com/ergo-services/ergo => ../../
replace github.com/ergo-services/ergo/etf => ../../etf
replace github.com/ergo-services/ergo/gen => ../../gen
replace github.com/ergo-services/ergo/node => ../../node
require github.com/ergo-services/ergo v1.999.211 // indirect

and then I've made the fixes you've mentioned. There was one mistake on line:

var utime, time int64

I assume it should be:

var utime, stime int64

I've built everything and ran the script again but it hanged the same way it did last time.

[halturin]

so there were a few issues

• compilation (it's been fixed, I assume).
• run built binary on arm hardware (it hangs on start, I suppose)

It seems I need to get the same HW (or a similar one) or a way to run it somehow in the VM (I have no clue how to do this so far).

[finalclass]

Yes the compilation issue has been fixed (actually I was compiling to the wrong target: it should be arm64, not just arm).
The running part appears to still be a problem. Let me know if you wish me to test anything else on the device.

[halturin]

I finally bought this device. Waiting for the shipment.

[halturin]

I've just tested simple example on my tinker board. No issues (master branch)

Could you please check on your board the same way? You may also want to add -ergo.trace to see extra debug info. Like this...

mendel@tinker:~$ ./simple -ergo.trace
2022/07/27 10:09:15 Start node with name "node@localhost" and cookie "cookies"
2022/07/27 10:09:15 Node listening range: 15000...65000
2022/07/27 10:09:15 Started embedded EMPD service and listen port: 4369
2022/07/27 10:09:15 EPMD accepted new connection from [::1]:46830
2022/07/27 10:09:15 Request from EPMD client: [0 22 120 58 152 77 0 0 6 0 5 0 4 110 111 100 101 0 4 17 59 1 0 0]
2022/07/27 10:09:15 [node@localhost] EPMD client: node registered
2022/07/27 10:09:15 [node@localhost] CORE registering behavior "erlang" in group "ergo:applications"
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1001>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1001> (registered name: "")
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1002>): net_kernel_sup
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1002>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1002> (registered name: "net_kernel_sup")
2022/07/27 10:09:15 [node@localhost] SUPERVISOR "net_kernel_sup" with restart strategy: one_for_one[permanent]
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1003>): net_kernel
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1003>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1003> (registered name: "net_kernel")
2022/07/27 10:09:15 NET_KERNEL: Init: []etf.Term(nil)
2022/07/27 10:09:15 [node@localhost] LINK process: <2C323E75.0.1002> => <2C323E75.0.1003>
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1004>): global_name_server
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1004>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1004> (registered name: "global_name_server")
2022/07/27 10:09:15 [node@localhost] LINK process: <2C323E75.0.1002> => <2C323E75.0.1004>
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1005>): rex
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1005>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1005> (registered name: "rex")
2022/07/27 10:09:15 REX: Init: []etf.Term(nil)
2022/07/27 10:09:15 [node@localhost] CORE registering behavior "erpc" in group "ergo:remote"
2022/07/27 10:09:15 [node@localhost] LINK process: <2C323E75.0.1002> => <2C323E75.0.1005>
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1006>): observer_backend
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1006>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1006> (registered name: "observer_backend")
2022/07/27 10:09:15 OBSERVER: Init: []etf.Term(nil)
2022/07/27 10:09:15 [node@localhost] RPC provide: proc_lib:translate_initial_call (gen.RPC)(0x1fc500)
2022/07/27 10:09:15 [node@localhost] RPC provide: appmon_info:start_link2 (gen.RPC)(0x1fc740)
2022/07/27 10:09:15 [node@localhost] LINK process: <2C323E75.0.1002> => <2C323E75.0.1006>
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1007>): erlang
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1007>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1007> (registered name: "erlang")
2022/07/27 10:09:15 ERLANG: Init: []etf.Term(nil)
2022/07/27 10:09:15 [node@localhost] LINK process: <2C323E75.0.1002> => <2C323E75.0.1007>
2022/07/27 10:09:15 [node@localhost] LINK process: <2C323E75.0.1001> => <2C323E75.0.1002>
2022/07/27 10:09:15 [node@localhost] CORE registering name (<2C323E75.0.1008>): gs1
2022/07/27 10:09:15 [node@localhost] CORE registering process: <2C323E75.0.1008>
2022/07/27 10:09:15 [node@localhost] CORE spawn a new process <2C323E75.0.1008> (registered name: "gs1")
2022/07/27 10:09:15 [node@localhost] CORE route message by pid (local) <2C323E75.0.1008>
2022/07/27 10:09:15 [node@localhost] GEN_SERVER <2C323E75.0.1008> got message from <2C323E75.0.1008>
2022/07/27 10:09:15 m: 100
HandleInfo: 100
2022/07/27 10:09:16 [node@localhost] CORE route message by pid (local) <2C323E75.0.1008>
2022/07/27 10:09:16 [node@localhost] GEN_SERVER <2C323E75.0.1008> got message from <2C323E75.0.1008>
2022/07/27 10:09:16 m: 101
HandleInfo: 101
2022/07/27 10:09:17 [node@localhost] CORE route message by pid (local) <2C323E75.0.1008>
2022/07/27 10:09:17 [node@localhost] GEN_SERVER <2C323E75.0.1008> got message from <2C323E75.0.1008>
2022/07/27 10:09:17 m: 102
HandleInfo: 102
2022/07/27 10:09:18 [node@localhost] CORE route message by pid (local) <2C323E75.0.1008>
2022/07/27 10:09:18 [node@localhost] GEN_SERVER <2C323E75.0.1008> got message from <2C323E75.0.1008>
2022/07/27 10:09:18 m: 103
HandleInfo: 103
2022/07/27 10:09:19 [node@localhost] CORE route message by pid (local) <2C323E75.0.1008>
2022/07/27 10:09:19 [node@localhost] GEN_SERVER <2C323E75.0.1008> got message from <2C323E75.0.1008>
2022/07/27 10:09:19 m: 104
HandleInfo: 104
2022/07/27 10:09:20 [node@localhost] CORE route message by pid (local) <2C323E75.0.1008>
2022/07/27 10:09:20 [node@localhost] GEN_SERVER <2C323E75.0.1008> got message from <2C323E75.0.1008>
2022/07/27 10:09:20 m: 105
HandleInfo: 105
2022/07/27 10:09:20 [node@localhost] CORE unregistering process: <2C323E75.0.1008>
2022/07/27 10:09:20 [node@localhost] CORE unregistering name (<2C323E75.0.1008>): gs1
2022/07/27 10:09:20 [node@localhost] MONITOR process terminated: <2C323E75.0.1008>
exited
2022/07/27 10:09:20 accept tcp 127.0.0.1:15000: use of closed network connection
mendel@tinker:~$

[finalclass]

Sorry for the late answer.
That's quite strange becuase for me it still does not work. I don't have any customizations on the device, just a bare tinker os installation.

$ uname -a
Linux vexing-eft 4.14.98-imx #1 SMP PREEMPT Wed Jun 9 15:32:53 UTC 2021 aarch64 GNU/Linux

It seams that we even have the same kernel versions (only the compilation time is different).

[halturin]

what version of golang are you using? OS, distro?

PS: forgot to mention. I've updated simple.go by adding flag.Parse() in order to use -ergo.trace
PPS: you may contact me via telegram halturin for instant messaging.

[halturin]

just to make sure I've tested one more time (recently updated my OS environment)

[finalclass]

I'm compiling it on:

$ go version
go version go1.19 linux/amd64
$ uname -a
Linux rog 5.15.59-1-MANJARO #1 SMP PREEMPT Wed Aug 3 11:20:04 UTC 2022 x86_64 GNU/Linux

[halturin]

May I ask you to build it using golang 1.18?

[finalclass]

Unfortunately it's the same.

[halturin]

could you please add flag.Parse() into the main function of simple.go and run it with -ergo.trace on this board?

[halturin]

btw, is this hostname (or localhost) present in the /ets/hosts?

[finalclass]

[halturin]

thanks for the quick reply. I'll take a look at why the node couldn't connect to itself on the 4369.
PS: do you have any firewall settings on this board?

[halturin]

and the last thing I would like to ask you... could you please try the v220 branch? It's got some changes in the network module.

[finalclass]

This got me thinking. If it hangs on trying to connect to itself then a simple http server should also be broken:

So I guess you can close this issue because it's clearly not related to the ergo library.

[finalclass]

There is one difference: with the http server it does not hang (the server). I can close it with Ctrl+C. When it comes to curl it hangs.

[halturin]

I guess there must be a firewall setting that drops any incoming connections. You may check this out with iptables -L and iptables -F to flush them out. Otherwise, I have no clue what it could be. Looks weird.

[finalclass]

It seams that it's not a firewall

When I find some time I will try to reinstall the system.

[halturin]

Since this bug is not related to ergo I'm closing it.