I have an idea what the bug can be

[ram-nat]

Awesome work with your patch!

So, I ran into the same crash on my Ubuntu server and after digging through the code, I have an idea what the issue is - it is not due to re-entrancy issues, but a use after free bug in the code.

I think the code is incorrectly doing HASH_CLEAR after it has free'd the nodes (this is only a theory at this point, but one that I'm more than 50% confident in).

So here - https://github.com/crass/mhddfs/blob/c62547d50a574ed9e3b04a8533be1c350b08c2ea/src/main.c#L245-L251

Replace with:

    // free memory
    HASH_ITER(hh, items_ht, item, tmp) {
                HASH_DEL(items_ht, item);
        free(item->name);
        free(item->st);
        free(item);
    }

I haven't tried this out yet, but plan to do so when I get some time (I've downgraded to 1.38 in the meantime). The only advantage over your patch is that it won't have a new dependency and can continue to use ut_hash.