Hacker News Reader

Comment by ajross | Hacker News Reader

Did gcc/libstdc++ have the same problem? The report you link is for clang, it looks like.

Yes, gcc/libstdc++ have the same problem. That's where I saw it first and then I tried llvm/libunwind and saw the same thing.

What's the actual lock on? The unwinding shouldn't involve shared mutable state (as someone who's been deep into the DWARF unwinding VM bytecode).

In libgcc it's in Unwind_Find_FDE - we think it's a lock around walking the loaded dynamic libraries. I haven't personally dug much deeper into it but my folks here and the llvm engineers seem to be pretty certain that's the problem (this: https://github.com/gcc-mirror/gcc/blob/master/libgcc/unwind-...). Right now we are rearranging our compiler so we throw fewer exceptions because you don't have to optimize things that you don't do :-).

Looks to me like it only tries to protect the building of the shared / sorted "seen_objects". You don't want two threads rebuilding it at the same time. Although there must be a way to work around this. Maybe something like optimistically walk through the seen list, then grab the lock to update and walk again without a lock? You should be able to safely walk a linked list forwards even with another thread inserting into it, right?

There is some work libc size to make the lock optional and enabling it only at the first dlopen.

Edit: last time I investigated the issue ended up here: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71744

I think that this is a lock around dlopen(). dlopen changes the list of mapped objects (and therefore, the mapping from instruction pointer to unwind information).

Reply on news.ycombinator.com