From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [63.128.21.124]) by sourceware.org (Postfix) with ESMTP id DC2A1388A407 for ; Thu, 4 Mar 2021 14:14:36 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org DC2A1388A407 Received: from mimecast-mx01.redhat.com (mimecast-mx01.redhat.com [209.132.183.4]) (Using TLS) by relay.mimecast.com with ESMTP id us-mta-194-Gtx6NO37O2SR1ynr227y8g-1; Thu, 04 Mar 2021 09:14:34 -0500 X-MC-Unique: Gtx6NO37O2SR1ynr227y8g-1 Received: from smtp.corp.redhat.com (int-mx03.intmail.prod.int.phx2.redhat.com [10.5.11.13]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx01.redhat.com (Postfix) with ESMTPS id 58C08184214A for ; Thu, 4 Mar 2021 14:14:33 +0000 (UTC) Received: from tucnak.zalov.cz (ovpn-113-231.ams2.redhat.com [10.36.113.231]) by smtp.corp.redhat.com (Postfix) with ESMTPS id 02F3E60CCB; Thu, 4 Mar 2021 14:14:32 +0000 (UTC) Received: from tucnak.zalov.cz (localhost [127.0.0.1]) by tucnak.zalov.cz (8.16.1/8.16.1) with ESMTPS id 124EEVrj2798831 (version=TLSv1.3 cipher=TLS_AES_256_GCM_SHA384 bits=256 verify=NOT); Thu, 4 Mar 2021 15:14:31 +0100 Received: (from jakub@localhost) by tucnak.zalov.cz (8.16.1/8.16.1/Submit) id 124EEUg62798830; Thu, 4 Mar 2021 15:14:30 +0100 Date: Thu, 4 Mar 2021 15:14:30 +0100 From: Jakub Jelinek To: Florian Weimer Cc: libc-alpha@sourceware.org Subject: Re: [PATCH] pthread_once hangs when init routine throws an exception [BZ #18435] Message-ID: <20210304141430.GC745611@tucnak> Reply-To: Jakub Jelinek References: <20210303125227.GO3748@tucnak> <87blbznnmp.fsf@oldenburg.str.redhat.com> <20210304130003.GA745611@tucnak> <87czwfkq2f.fsf@oldenburg.str.redhat.com> MIME-Version: 1.0 In-Reply-To: <87czwfkq2f.fsf@oldenburg.str.redhat.com> X-Scanned-By: MIMEDefang 2.79 on 10.5.11.13 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-Spam-Status: No, score=-12.0 required=5.0 tests=BAYES_00, DKIMWL_WL_HIGH, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, RCVD_IN_DNSWL_LOW, RCVD_IN_MSPIKE_H4, RCVD_IN_MSPIKE_WL, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.2 X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 04 Mar 2021 14:14:39 -0000 On Thu, Mar 04, 2021 at 02:26:32PM +0100, Florian Weimer wrote: > I think the patch is okay, except for the nits I pointed out. Here is an updated patch. Changes from the first one are dropped __GNUC__ check, added tst-oncey*.c tests, avoiding indirect function call in the theoretical execute != 0 case and dropping __pthread_cleanup_cleanup because of that as it is now called only in case the unwind cleanup does the cleaning up which means neither unwind_stop nor __pthread_cleanup_pop did it. [PATCH] pthread_once hangs when init routine throws an exception [BZ #18435] This is another attempt at making pthread_once handle throwing exceptions from the init routine callback. As the new testcases show, just switching to the cleanup attribute based cleanup does fix the tst-once5 test, but breaks the new tst-oncey3 test. That is because when throwing exceptions, only the unwind info registered cleanups (i.e. C++ destructors or cleanup attribute), when cancelling threads and there has been unwind info from the cancellation point up to whatever needs cleanup both unwind info registered cleanups and THREAD_SETMEM (self, cleanup, ...) registered cleanups are invoked, but once we hit some frame with no unwind info, only the THREAD_SETMEM (self, cleanup, ...) registered cleanups are invoked. So, to stay fully backwards compatible (allow init routines without unwind info which encounter cancellation points) and handle exception throwing we actually need to register the pthread_once cleanups in both unwind info and in the THREAD_SETMEM (self, cleanup, ...) way. If an exception is thrown, only the former will happen and we in that case need to also unregister the THREAD_SETMEM (self, cleanup, ...) registered handler, because otherwise after catching the exception the user code could call deeper into the stack some cancellation point, get cancelled and then a stale cleanup handler would clobber stack and probably crash. If a thread calling init routine is cancelled and unwind info ends before the pthread_once frame, it will be cleaned up through self->cleanup as before. And if unwind info is present, unwind_stop first calls the self->cleanup registered handler for the frame, then it will call the unwind info registered handler but that will already see __do_it == 0 and do nothing. diff --git a/nptl/Makefile b/nptl/Makefile index 5f85dd7854..33766eaf7a 100644 --- a/nptl/Makefile +++ b/nptl/Makefile @@ -386,10 +386,6 @@ xtests += tst-eintr1 test-srcs = tst-oddstacklimit -# Test expected to fail on most targets (except x86_64) due to bug -# 18435 - pthread_once hangs when init routine throws an exception. -test-xfail-tst-once5 = yes - gen-as-const-headers = unwindbuf.sym \ pthread-pi-defines.sym diff --git a/nptl/pthreadP.h b/nptl/pthreadP.h index d2fd0826fe..93f3cef00f 100644 --- a/nptl/pthreadP.h +++ b/nptl/pthreadP.h @@ -604,6 +604,67 @@ extern void __pthread_cleanup_pop (struct _pthread_cleanup_buffer *buffer, # undef pthread_cleanup_pop # define pthread_cleanup_pop(execute) \ __pthread_cleanup_pop (&_buffer, (execute)); } + +# if defined __EXCEPTIONS && !defined __cplusplus +/* Structure to hold the cleanup handler information. */ +struct __pthread_cleanup_combined_frame +{ + void (*__cancel_routine) (void *); + void *__cancel_arg; + int __do_it; + struct _pthread_cleanup_buffer __buffer; +}; + +/* Special cleanup macros which register cleanup both using + __pthread_cleanup_{push,pop} and using cleanup attribute. This is needed + for pthread_once, so that it supports both throwing exceptions from the + pthread_once callback (only cleanup attribute works there) and cancellation + of the thread running the callback if the callback or some routines it + calls don't have unwind information. */ + +static __always_inline void +__pthread_cleanup_combined_routine (struct __pthread_cleanup_combined_frame + *__frame) +{ + if (__frame->__do_it) + { + __frame->__cancel_routine (__frame->__cancel_arg); + __frame->__do_it = 0; + __pthread_cleanup_pop (&__frame->__buffer, 0); + } +} + +static inline void +__pthread_cleanup_combined_routine_voidptr (void *__arg) +{ + struct __pthread_cleanup_combined_frame *__frame + = (struct __pthread_cleanup_combined_frame *) __arg; + if (__frame->__do_it) + { + __frame->__cancel_routine (__frame->__cancel_arg); + __frame->__do_it = 0; + } +} + +# define pthread_cleanup_combined_push(routine, arg) \ + do { \ + void (*__cancel_routine) (void *) = (routine); \ + struct __pthread_cleanup_combined_frame __clframe \ + __attribute__ ((__cleanup__ (__pthread_cleanup_combined_routine))) \ + = { .__cancel_routine = __cancel_routine, .__cancel_arg = (arg), \ + .__do_it = 1 }; \ + __pthread_cleanup_push (&__clframe.__buffer, \ + __pthread_cleanup_combined_routine_voidptr, \ + &__clframe); + +# define pthread_cleanup_combined_pop(execute) \ + __pthread_cleanup_pop (&__clframe.__buffer, 0); \ + __clframe.__do_it = 0; \ + if (execute) \ + __cancel_routine (__clframe.__cancel_arg); \ + } while (0) + +# endif #endif extern void __pthread_cleanup_push_defer (struct _pthread_cleanup_buffer *buffer, diff --git a/nptl/pthread_once.c b/nptl/pthread_once.c index 28d97097c7..7645da222a 100644 --- a/nptl/pthread_once.c +++ b/nptl/pthread_once.c @@ -111,11 +111,11 @@ __pthread_once_slow (pthread_once_t *once_control, void (*init_routine) (void)) /* This thread is the first here. Do the initialization. Register a cleanup handler so that in case the thread gets interrupted the initialization can be restarted. */ - pthread_cleanup_push (clear_once_control, once_control); + pthread_cleanup_combined_push (clear_once_control, once_control); init_routine (); - pthread_cleanup_pop (0); + pthread_cleanup_combined_pop (0); /* Mark *once_control as having finished the initialization. We need diff --git a/nptl/tst-once5.cc b/nptl/tst-once5.cc index b797ab3562..60fe1ef820 100644 --- a/nptl/tst-once5.cc +++ b/nptl/tst-once5.cc @@ -59,7 +59,7 @@ do_test (void) " throwing an exception", stderr); } catch (OnceException) { - if (1 < niter) + if (niter > 1) fputs ("pthread_once unexpectedly threw", stderr); result = 0; } @@ -75,7 +75,5 @@ do_test (void) return result; } -// The test currently hangs and is XFAILed. Reduce the timeout. -#define TIMEOUT 1 #define TEST_FUNCTION do_test () #include "../test-skeleton.c" diff --git a/sysdeps/pthread/Makefile b/sysdeps/pthread/Makefile index eeb64f9fb0..53b65ef349 100644 --- a/sysdeps/pthread/Makefile +++ b/sysdeps/pthread/Makefile @@ -231,7 +231,7 @@ generated += $(objpfx)tst-atfork2.mtrace \ tests-internal += tst-cancel25 tst-robust8 -tests += tst-oncex3 tst-oncex4 +tests += tst-oncex3 tst-oncex4 tst-oncey3 tst-oncey4 modules-names += tst-join7mod @@ -242,6 +242,8 @@ endif CFLAGS-tst-oncex3.c += -fexceptions CFLAGS-tst-oncex4.c += -fexceptions +CFLAGS-tst-oncey3.c += -fno-exceptions -fno-asynchronous-unwind-tables +CFLAGS-tst-oncey4.c += -fno-exceptions -fno-asynchronous-unwind-tables $(objpfx)tst-join7: $(libdl) $(shared-thread-library) $(objpfx)tst-join7.out: $(objpfx)tst-join7mod.so diff --git a/sysdeps/pthread/tst-oncey3.c b/sysdeps/pthread/tst-oncey3.c new file mode 100644 index 0000000000..08225b88dc --- /dev/null +++ b/sysdeps/pthread/tst-oncey3.c @@ -0,0 +1 @@ +#include "tst-once3.c" diff --git a/sysdeps/pthread/tst-oncey4.c b/sysdeps/pthread/tst-oncey4.c new file mode 100644 index 0000000000..9b4d98f3f1 --- /dev/null +++ b/sysdeps/pthread/tst-oncey4.c @@ -0,0 +1 @@ +#include "tst-once4.c" Jakub