[PATCH] eal: support including mapped memory in core dump

[Lewis Donzis]

----- On Oct 23, 2024, at 6:18 PM, Dmitry Kozlyuk dmitry.kozliuk at gmail.com wrote:
> Lewis, testing on FreeBSD would be appreciated.

Well, unfortunately, it's not working very well...  The contigmem memory was not included in the core dump.

I added logging just before the madvise() call to prove that it's calling madvise() properly, and it looks good.
I then tried forcing all madvise() calls to use EAL_DODUMP.
And finally, I tried removing the call to madvise() altogether.

None of those things changed the size of the core dump from what it was, and my test of trying to access an mbuf in the debugger still says the memory is inaccessible.

What's even worse is, it appears that I had already tried the above work-arounds back on May 28th and mentioned that in an e-mail to you.  Apparently I totally forgot about those experiments!  At the time, I wondered if it was just something fundamental about FreeBSD and contigmem.

It may take some digging through the FreeBSD source code to see how the dumper works.

So I apologize for not remembering that this wasn't entirely a DPDK issue.  That said, it's nice that this is available on Linux and maybe we'll find a way to make it work on FreeBSD.

Thanks again for the efforts,
lew