BUG: segfault when getting memory info on M1

[jqnatividad]

This code segfaults on sysinfo 0.25.3. It worked without issues previously on earlier versions of sysinfo:

    const MEMORY_LIMITED_BUFFER: u64 = 100 * 1_000_000; // 100 MB

    let mem_limited_buffer = if System::IS_SUPPORTED {
         let mut sys = System::new_all();
         sys.refresh_memory();
         (sys.total_memory() * 1000) / 10 // 10 percent of total memory
     } else {
         MEMORY_LIMITED_BUFFER
     };

[GuillaumeGomez]

Thanks for the report! Since I don't have a mac M1 myself, I'll need someone else with one to either try to fix it or provide a lot more information. Luckily, there weren't many PRs:

• Add support for get temperature on m1 #792
• Greatly improve source code to make it simpler to handle both x86 and arch targets #814
• Add macos M1 check on CI #813
• Fix leak in macOS disk retrieval #812

Can you provide information from a memory tool so we can see which part of the code is causing this new bug?

[jqnatividad]

Hi @GuillaumeGomez , would love to provide the information you need.

I know you said you don't have access to a mac M1, but do you have a specific memory tool in mind?

Alternatively, is there a way I can build my project to get the information you need?

FYI, I use sysinfo to check for available memory with the qsv extsort command here, and to do a hardware survey here.

[GuillaumeGomez]

I don't know which tools exist on mac. The goal is to get the stacktrace and possibly which buffer was wrongly accessed so by default I'd say valgrind but any tool that can get such information is just as good.

[jqnatividad]

Hmmm... it's not segfaulting anymore @GuillaumeGomez . I added tests in my CI to exercise sysinfo and its now working as expected.
When I filed the report, I can reproduce it reliably just by using 0.25.3 and checking for memory....
But now, it's working OK

[jqnatividad]

Oooppsss.... I take that back.
Ran the test without doing cargo update. It's still segfaulting... continuing to investigate...

[schell]

I just git bisect'd a segfault on my M1 down to bumping sysinfo to 0.25.3. This is the relevant crashed thread in my crash report:

Thread 14 Crashed::  Dispatch queue: com.apple.root.default-qos.overcommit
0   libobjc.A.dylib               	       0x1b39989c0 objc_release + 16
1   CoreFoundation                	       0x1b3b86370 cow_cleanup + 172
2   CoreFoundation                	       0x1b3b86274 -[__NSDictionaryM dealloc] + 148
3   IOKit                         	       0x1b65d4208 __IOHIDEventSystemClientIntRelease + 220
4   IOKit                         	       0x1b65d0f3c _IOHIDObjectRetainCount + 416
5   libdispatch.dylib             	       0x1b394c1b4 _dispatch_client_callout + 20
6   libdispatch.dylib             	       0x1b395da04 _dispatch_root_queue_drain + 680
7   libdispatch.dylib             	       0x1b395e104 _dispatch_worker_thread2 + 164
8   libsystem_pthread.dylib       	       0x1b3b0c324 _pthread_wqthread + 228
9   libsystem_pthread.dylib       	       0x1b3b0b080 start_wqthread + 8

[jqnatividad]

Just wanted to confirm that this is still a problem with sysinfo v 0.26.0...

[GuillaumeGomez]

Yes, no changes on my end to fix this. @lqd nicely proposed to take a look whenever he has time for it.