Missing memory mappings

Missing memory mappings

[Roger Phillips]

[-- Attachment #1: Type: text/plain, Size: 584 bytes --]

Greetings,

I'm doing post-mortem coredump analysis for multithreaded C++ applications on x64 Linux. Occasionally I get coredumps where some threads (not the crashing thread) has instruction pointers that look valid but cannot be matched to a memory region provided by info proc mappings. So my assumption is that these libs have been unloaded between the signal and the creation of the coredump and I wonder in which library load scenarios this could happen.

This might not be a gdb problem but a coredump problem but I don't know where else to find the expertise.

Regards.

Re: Missing memory mappings

[Martin Simmons]

>>>>> On Tue, 20 Feb 2024 12:27:57 +0000, Roger Phillips via Gdb said:
> 
> I'm doing post-mortem coredump analysis for multithreaded C++
> applications on x64 Linux. Occasionally I get coredumps where some
> threads (not the crashing thread) has instruction pointers that look
> valid but cannot be matched to a memory region provided by info proc
> mappings. So my assumption is that these libs have been unloaded
> between the signal and the creation of the coredump and I wonder in
> which library load scenarios this could happen.

Maybe the vdso section, which is not dumped?

> This might not be a gdb problem but a coredump problem but I don't
> know where else to find the expertise.

You could see what 'objdump -x' says about the coredump's sections.

__Martin