From: LIU Hao <lh_mouse@126.com>
To: i.nixman@autistici.org
Cc: Wakely <jwakely@redhat.com>,
libstdc++@gcc.gnu.org, gcc-patches@gcc.gnu.org,
Eric Botcazou <ebotcazou@adacore.com>
Subject: Re: Adding a new thread model to GCC
Date: Mon, 24 Oct 2022 11:40:32 +0800 [thread overview]
Message-ID: <8a3105e7-c7aa-7f70-d8ae-2808ca126291@126.com> (raw)
In-Reply-To: <ea6d1c9603017c6c74b7042b07b5ed9f@autistici.org>
[-- Attachment #1.1: Type: text/plain, Size: 4199 bytes --]
在 2022/10/21 20:34, i.nixman@autistici.org 写道:
>
> got it...
> anyway it seems logical to me the way I proposed :)
>
>
Below is a message forwarded from mingw-w64-public, elaborating the necessity of a new thread model.
As there are objections from other mingw-w64 developers, I am putting those patches against
mingw-w64-crt on hold for now. Despite that, all threading facilities - mutexes, condition
variables, once flags, etc. - are still fully functional within the mcf thread model.
In addition, I will keep maintaining my personal builds (from GCC 12 release branch) with these
patches at https://gcc-mcf.lhmouse.com/.
-------- Forwarded Message --------
在 2022/10/23 18:06, Jacek Caban 写道:
>
> Please, let's not do that. It's possible to fix existing implementations, we don't need to make
> things more complicated than they are.
>
Okay okay, I think I have to compose a thorough list of problems that we are facing at the moment,
and had better have a permalink to the mailing list archive that I can reference elsewhere. I have
been tired of repeating the same grounds of arguments again and again:
1. In a DLL, destructors of static objects and callbacks that are registered
with `atexit()`, are executed by `LdrShutdownProcess()`, after all the other
thread have been terminated `ZwTerminateProcessO(NULL, status)`. This means
that, if another thread has been terminated while holding a mutex, the mutex
can never get unlocked. If a destructor attempts to lock the same mutex,
deadlocks will occur. Destructors of executables do not suffer from this
issue, because they are executed before `RtlExitUserProcess()`.
Standard behavior: Static destructors and exit callbacks should be executed
while other threads are running. If another thread attempts to access a
destroyed object, the behavior is undefined; the user is responsible to
prevent this from happening, by joining or suspending it.
2. Following 1, in a DLL, static destructors and exit callbacks are still
invoked when `_Exit()` or `quick_exit()` is called.
Standard behavior: `_Exit()` should not perform any cleanup; not even open
files are flushed. `quick_exit()` shall invoke all quick-exit callbacks in
reverse order, then call `_Exit()`.
3. There is a use-after-free bug [1] about thread-local destructors. I suspect
this is caused by emutls, because GCC uses `__cxa_thread_atexit()` to
register thread-local destructors, which could interleave with
`emutls_destroy()`.
Standard behavior: This is not allowed to happen. mcfgthread solves this
issue by running thread-local destructors and thread-specific key
destructors as two separate passes [3].
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80816
[2]
https://github.com/gcc-mirror/gcc/blob/f84e4fb44aa26b71fbc64e0532fd24d96e5caa3f/libgcc/emutls.c#L96
[3]
https://github.com/lhmouse/mcfgthread/blob/63e034d375caf585e2921cd3455f1048feb2172d/src/xglobals.c#L249
4. In the win32 thread model, thread-specific key destructors are called at
process exit [4], after static destructors.
Standard behavior: They shall be called only when a thread exits, and the
associated thread-specific values are not a null pointer. They shall not be
called when a program terminates; instead, users are responsible for
deallocating such resources before calling `exit()`. This requirement is
missing in POSIX, but formally specified by ISO/IEC 9899:2017, as the 4th
paragraph in '7.26.6.1 The tss_create function'.
[4]
https://github.com/mingw-w64/mingw-w64/blob/d0a034a04d312434b842c4869a8a900568d8db98/mingw-w64-crt/crt/tlsthrd.c#L134
5. Wait operations, of timed mutexes and condition variables, should take
absolute time points as `struct timespec`.
Standard behavior: Both POSIX and ISO C specifies them as such, while all
Windows APIs take relative durations as a 32-bit integer of milliseconds,
which can also easily get overflown.
--
Best regards,
LIU Hao
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]
next prev parent reply other threads:[~2022-10-24 3:40 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-01 18:34 LIU Hao
2022-10-01 20:02 ` Bernhard Reutner-Fischer
2022-10-02 12:54 ` LIU Hao
2022-10-03 5:03 ` Bernhard Reutner-Fischer
2022-10-04 8:06 ` LIU Hao
2022-10-04 19:45 ` Bernhard Reutner-Fischer
2022-10-04 12:44 ` LIU Hao
2022-10-04 13:13 ` Xi Ruoyao
2022-10-04 13:45 ` LIU Hao
2022-10-05 1:23 ` Xi Ruoyao
2022-10-10 15:56 ` LIU Hao
2022-10-11 13:22 ` LIU Hao
2022-10-14 9:39 ` Jonathan Yong
2022-10-19 13:55 ` Jonathan Yong
2022-10-19 19:53 ` Bernhard Reutner-Fischer
2022-10-20 1:25 ` LIU Hao
2022-10-21 9:58 ` Jonathan Wakely
2022-10-21 10:09 ` i.nixman
2022-10-21 10:48 ` Jonathan Wakely
2022-10-21 10:55 ` i.nixman
2022-10-21 11:36 ` LIU Hao
2022-10-21 11:54 ` i.nixman
2022-10-21 12:19 ` LIU Hao
2022-10-21 12:34 ` i.nixman
2022-10-24 3:40 ` LIU Hao [this message]
2022-10-24 20:50 ` Jacek Caban
2022-10-21 12:13 ` Jacek Caban
2022-10-21 12:29 ` LIU Hao
2022-10-21 12:40 ` Jacek Caban
2022-10-21 11:44 ` Eric Botcazou
2022-10-21 11:55 ` i.nixman
2022-10-21 12:30 ` Jacek Caban
2022-10-23 0:36 ` NightStrike
2022-10-24 6:53 ` i.nixman
2022-10-24 8:15 ` Eric Botcazou
2022-10-24 8:20 ` i.nixman
2022-10-31 9:18 ` Eric Botcazou
2022-10-31 15:22 ` i.nixman
2022-12-18 11:14 ` Jonathan Yong
2022-11-01 5:22 ` i.nixman
2022-11-01 9:09 ` Eric Botcazou
2022-11-02 12:05 ` i.nixman
2022-11-02 21:27 ` Eric Botcazou
2022-11-02 21:40 ` i.nixman
2022-12-16 17:18 ` Jonathan Wakely
2022-12-16 19:20 ` Eric Botcazou
2022-12-22 12:21 ` Jonathan Yong
2022-12-22 12:28 ` i.nixman
2022-12-23 23:59 ` Jonathan Yong
2022-12-24 5:58 ` NightStrike
2022-12-24 6:27 ` i.nixman
2022-12-24 13:50 ` i.nixman
2022-12-24 15:42 ` i.nixman
2022-12-24 15:57 ` i.nixman
2022-12-24 21:22 ` i.nixman
2022-12-25 1:10 ` Jonathan Yong
2023-01-09 21:56 ` Eric Botcazou
2022-12-24 6:22 ` i.nixman
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8a3105e7-c7aa-7f70-d8ae-2808ca126291@126.com \
--to=lh_mouse@126.com \
--cc=ebotcazou@adacore.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=i.nixman@autistici.org \
--cc=jwakely@redhat.com \
--cc=libstdc++@gcc.gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).