Only access non-locals after they're initialised#40
Open
aszlig wants to merge 1 commit into
Open
Conversation
This fixes a long-standing issue on aarch64, which is caused by mimalloc in Python 3 calling close() too early: #0 0x0000007ff75952ac in __pthread_kill_implementation () from .../lib/libc.so.6 #1 0x0000007ff753c5b8 in raise () from .../lib/libc.so.6 #2 0x0000007ff7525d38 in abort () from .../lib/libc.so.6 #3 0x0000007ff7299818 in __gnu_cxx::__verbose_terminate_handler() () from .../lib/libstdc++.so.6 #4 0x0000007ff729704c in __cxxabiv1::__terminate(void (*)()) () from .../lib/libstdc++.so.6 #5 0x0000007ff728e49c in std::terminate() () from .../lib/libstdc++.so.6 #6 0x0000007ff72973dc in __cxa_throw () from .../lib/libstdc++.so.6 #7 0x0000007ff728ed7c in std::__throw_bad_array_new_length() () from .../lib/libstdc++.so.6 #8 0x0000007ff7f10c54 in std::__new_allocator<std::__detail::_Hash_node_base*>::allocate (this=0x7fffffe890, __n=18446744073709551557) at .../include/c++/15.2.0/bits/new_allocator.h:139 #9 0x0000007ff7f6e898 in std::allocator_traits<std::allocator<std::__detail::_Hash_node_base*> >::allocate (__a=..., __n=18446744073709551557) at .../include/c++/15.2.0/bits/alloc_traits.h:614 #10 std::__detail::_Hashtable_alloc<std::allocator<std::__detail::_Hash_node<int, false> > >::_M_allocate_buckets (this=0x7ff7fb02a0 <all_fds>, __bkt_count=18446744073709551557) at .../include/c++/15.2.0/bits/hashtable_policy.h:1605 #11 0x0000007ff7f6e500 in std::_Hashtable<int, int, std::allocator<int>, std::__detail::_Identity, std::equal_to<int>, std::hash<int>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, true, true> >::_M_allocate_buckets (this=0x7ff7fb02a0 <all_fds>, __bkt_count=18446744073709551557) at .../include/c++/15.2.0/bits/hashtable.h:413 #12 0x0000007ff7f6d944 in std::_Hashtable<int, int, std::allocator<int>, std::__detail::_Identity, std::equal_to<int>, std::hash<int>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, true, true> >::_M_rehash (this=0x7ff7fb02a0 <all_fds>, __bkt_count=18446744073709551557) at .../include/c++/15.2.0/bits/hashtable.h:2754 #13 0x0000007ff7f6c0b8 in std::_Hashtable<int, int, std::allocator<int>, std::__detail::_Identity, std::equal_to<int>, std::hash<int>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, true, true> >::_M_insert_unique_node (this=0x7ff7fb02a0 <all_fds>, __bkt=3, __code=3, __node=0x5555583580, __n_elt=1) at .../include/c++/15.2.0/bits/hashtable.h:2479 #14 0x0000007ff7f6a034 in std::_Hashtable<int, int, std::allocator<int>, std::__detail::_Identity, std::equal_to<int>, std::hash<int>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, true, true> >::_M_emplace_uniq<int const&> (this=0x7ff7fb02a0 <all_fds>) at .../include/c++/15.2.0/bits/hashtable.h:2365 #15 0x0000007ff7f685a8 in std::_Hashtable<int, int, std::allocator<int>, std::__detail::_Identity, std::equal_to<int>, std::hash<int>, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, true, true> >::insert (this=0x7ff7fb02a0 <all_fds>, __v=@0x7fffffebb8: 3) at .../include/c++/15.2.0/bits/hashtable.h:1049 #16 0x0000007ff7f66b60 in std::unordered_set<int, std::hash<int>, std::equal_to<int>, std::allocator<int> >::insert (this=0x7ff7fb02a0 <all_fds>, __x=@0x7fffffebb8: 3) at .../include/c++/15.2.0/bits/unordered_set.h:475 #17 0x0000007ff7f64c80 in Systemd::init (rules=...) at ../src/systemd.cc:213 #18 0x0000007ff7f14bd0 in init_rules () at ../src/preload.cc:77 #19 0x0000007ff7f19edc in ip2unix_wrap_close (fd=4) at ../src/preload.cc:677 #20 0x0000007ff78cf430 in mi_process_init.part () from .../lib/libpython3.13.so.1.0 #21 0x0000007ff7894ee8 in _mi_process_init () from .../lib/libpython3.13.so.1.0 #22 0x0000007ff7fc32c4 in call_init () from .../lib/ld-linux-aarch64.so.1 #23 0x0000007ff7fc341c in _dl_init () from .../lib/ld-linux-aarch64.so.1 #24 0x0000007ff7fdbea0 in _start () from .../lib/ld-linux-aarch64.so.1 Here the non-local all_fds variable is an unordered_set which hasn't yet been initialised but is accessed because _mi_process_init has a lower priority and thus is initialised *before* all_fds. Since mi_process_init calls close(), an insert on the all_fds variable is triggered by the systemd rule initialisation, which then fails because all_fds hasn't been initialised yet. The all_fds variable isn't the only one that's affected, so in order to avoid wrapping all non-local variables in lazy initalisation wrappers, we now short-circuit all the library calls that trigger the ruleset initialisation machinery. While at it, I also fixed a potential data race that could emerge in multithreaded programs. While in practice most networking code is initialised before actually spawning threads, it's still something that could happen, so let's use std::call_once on rule initialisation. Signed-off-by: aszlig <aszlig@nix.build>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This fixes a long-standing issue on aarch64, which is caused by mimalloc in Python 3 calling
closetoo early:Here the non-local
all_fdsvariable is anunordered_setwhich hasn't yet been initialised but is accessed because _mi_process_init has a lower priority and thus is initialised beforeall_fds.Since mi_process_init calls
close, an insert on theall_fdsvariable is triggered by the systemd rule initialisation, which then fails becauseall_fdshasn't been initialised yet.The
all_fdsvariable isn't the only one that's affected, so in order to avoid wrapping all non-local variables in lazy initalisation wrappers, we now short-circuit all the library calls that trigger the ruleset initialisation machinery.