From patchwork Tue Apr 26 19:15:22 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Adhemerval Zanella X-Patchwork-Id: 53244 Return-Path: X-Original-To: patchwork@sourceware.org Delivered-To: patchwork@sourceware.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 3ED273857C42 for ; Tue, 26 Apr 2022 19:17:20 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 3ED273857C42 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1651000640; bh=EM1ctNVSM6lio70GvBCs7uD7KunJWjbs6uQuNKDvYyM=; h=To:Subject:Date:In-Reply-To:References:List-Id:List-Unsubscribe: List-Archive:List-Post:List-Help:List-Subscribe:From:Reply-To: From; b=UD/muwOht2UoDPPY2xw2CF0lrHNY1YdeZOPh+wc8zkvGEAcD40xW8UzBCVu+j8vha Y3IdOQYuv1KtpqXF8jCHAdhDjvXhAVsiEBhIx1ri1SlExXbQRqlTpES70cwq8+s06E o5nn7weLuqmQDMG2GYlvxFDvi+EWS+hTZr5dJCd8= X-Original-To: libc-alpha@sourceware.org Delivered-To: libc-alpha@sourceware.org Received: from mail-oi1-x236.google.com (mail-oi1-x236.google.com [IPv6:2607:f8b0:4864:20::236]) by sourceware.org (Postfix) with ESMTPS id 8669A3857C41 for ; Tue, 26 Apr 2022 19:15:35 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 8669A3857C41 Received: by mail-oi1-x236.google.com with SMTP id y63so1343461oia.7 for ; Tue, 26 Apr 2022 12:15:35 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=EM1ctNVSM6lio70GvBCs7uD7KunJWjbs6uQuNKDvYyM=; b=kI8XmIQq1uWfMTtN4FNMs2DDBh9rSPs9s1zOj7OmjddCDPhxyYb+5n+xLXpUqnF5Aa TDPJ/MAMNhAWF/e7WUQKY2HqDLn2LsFo8xNVRwwiuD6KG30JUiiAYBTmfinGyRD6m1N3 SGQnhto1kd3ENjE8Gv7qF3Ju/vIie0+SB0WkRJi/J/aNmrC3+2MHKtzopsansljOBOTJ dZiy8Ig0G3mGK/VKKAUmCxnhy8UqEwrHgq7P7/3JeS/PCdgIbZImCdwBoGIe0t5Wl+ZF rZYw7c4CxQQ2t+maPGmG9oVBGcnRzWH4hc0ZGL4O76OtwuRcy7NreVz3595FhtzPuKKz tP4Q== X-Gm-Message-State: AOAM533oRV1mxXqeHf89XvmnVAwnfS+uGJQw6d4ZmzHLJ0SL6JfNJcl3 rKKiHt3626GvaTI2+Nx9ynpUy17mLcf/Fg== X-Google-Smtp-Source: ABdhPJzT6wSpGdS+xlNMC8SDLZjW66DJQAmq2dqd/gQzy/PRD6/8MLC1cHWvi5yOt96eIBWIHaz76g== X-Received: by 2002:a05:6808:144c:b0:325:64f3:d428 with SMTP id x12-20020a056808144c00b0032564f3d428mr1690782oiv.210.1651000533882; Tue, 26 Apr 2022 12:15:33 -0700 (PDT) Received: from birita.. ([2804:431:c7ca:4214:745b:d03c:b667:123b]) by smtp.gmail.com with ESMTPSA id 123-20020a4a0181000000b00324a1217e96sm5884129oor.35.2022.04.26.12.15.32 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Tue, 26 Apr 2022 12:15:33 -0700 (PDT) To: libc-alpha@sourceware.org, Florian Weimer Subject: [PATCH v2 3/4] Move libio lock single-thread optimization to generic libc-lock (BZ #27842) Date: Tue, 26 Apr 2022 16:15:22 -0300 Message-Id: <20220426191523.833171-4-adhemerval.zanella@linaro.org> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20220426191523.833171-1-adhemerval.zanella@linaro.org> References: <20220426191523.833171-1-adhemerval.zanella@linaro.org> MIME-Version: 1.0 X-Spam-Status: No, score=-12.5 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) 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-Patchwork-Original-From: Adhemerval Zanella via Libc-alpha From: Adhemerval Zanella Reply-To: Adhemerval Zanella Errors-To: libc-alpha-bounces+patchwork=sourceware.org@sourceware.org Sender: "Libc-alpha" This patch moves the single thread stdio optimization (d2e04918833d9) to the libc-lock. With generic support there is no need to add per-function code to handle the single-thread case, and it allows to removes the _IO_enable_locks (since once process goes multithread the locks will be always used). It also handles the memory streams requirement (de895ddcd7fc), since the SINGLE_THREAD_P already contains te information whether the process is multithread (so there is no need to disable the optimization because such stream are listed in _IO_list_all). Finally it also removed the flockfile uses a read-modify-write operation on _flags2 outside a lock region (BZ #27842). Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu. --- libio/Versions | 3 --- libio/feof.c | 2 -- libio/ferror.c | 2 -- libio/fputc.c | 2 -- libio/genops.c | 28 ---------------------------- libio/getc.c | 2 -- libio/getchar.c | 4 +--- libio/iofopncook.c | 2 -- libio/ioungetc.c | 2 -- libio/libio.h | 4 ---- libio/memstream.c | 3 --- libio/putc.c | 2 -- libio/wmemstream.c | 3 --- nptl/pthread_create.c | 3 --- stdio-common/flockfile.c | 1 - sysdeps/mach/libc-lock.h | 9 +++++---- sysdeps/nptl/libc-lock.h | 7 ++++--- 17 files changed, 10 insertions(+), 69 deletions(-) diff --git a/libio/Versions b/libio/Versions index b91a7bc914..b528fae670 100644 --- a/libio/Versions +++ b/libio/Versions @@ -159,9 +159,6 @@ libc { # Used by NPTL and librt __libc_fatal; - # Used by NPTL - _IO_enable_locks; - __fseeko64; __ftello64; } diff --git a/libio/feof.c b/libio/feof.c index 7f79b2795e..41e7beeca7 100644 --- a/libio/feof.c +++ b/libio/feof.c @@ -32,8 +32,6 @@ _IO_feof (FILE *fp) { int result; CHECK_FILE (fp, EOF); - if (!_IO_need_lock (fp)) - return _IO_feof_unlocked (fp); _IO_flockfile (fp); result = _IO_feof_unlocked (fp); _IO_funlockfile (fp); diff --git a/libio/ferror.c b/libio/ferror.c index d2489c54d3..a60efc3f0f 100644 --- a/libio/ferror.c +++ b/libio/ferror.c @@ -32,8 +32,6 @@ _IO_ferror (FILE *fp) { int result; CHECK_FILE (fp, EOF); - if (!_IO_need_lock (fp)) - return _IO_ferror_unlocked (fp); _IO_flockfile (fp); result = _IO_ferror_unlocked (fp); _IO_funlockfile (fp); diff --git a/libio/fputc.c b/libio/fputc.c index d29a9d0aee..c3d66500d1 100644 --- a/libio/fputc.c +++ b/libio/fputc.c @@ -32,8 +32,6 @@ fputc (int c, FILE *fp) { int result; CHECK_FILE (fp, EOF); - if (!_IO_need_lock (fp)) - return _IO_putc_unlocked (c, fp); _IO_acquire_lock (fp); result = _IO_putc_unlocked (c, fp); _IO_release_lock (fp); diff --git a/libio/genops.c b/libio/genops.c index ccfbcd149d..eca50bacbc 100644 --- a/libio/genops.c +++ b/libio/genops.c @@ -488,39 +488,11 @@ _IO_init (FILE *fp, int flags) _IO_init_internal (fp, flags); } -static int stdio_needs_locking; - -/* In a single-threaded process most stdio locks can be omitted. After - _IO_enable_locks is called, locks are not optimized away any more. - It must be first called while the process is still single-threaded. - - This lock optimization can be disabled on a per-file basis by setting - _IO_FLAGS2_NEED_LOCK, because a file can have user-defined callbacks - or can be locked with flockfile and then a thread may be created - between a lock and unlock, so omitting the lock is not valid. - - Here we have to make sure that the flag is set on all existing files - and files created later. */ -void -_IO_enable_locks (void) -{ - _IO_ITER i; - - if (stdio_needs_locking) - return; - stdio_needs_locking = 1; - for (i = _IO_iter_begin (); i != _IO_iter_end (); i = _IO_iter_next (i)) - _IO_iter_file (i)->_flags2 |= _IO_FLAGS2_NEED_LOCK; -} -libc_hidden_def (_IO_enable_locks) - void _IO_old_init (FILE *fp, int flags) { fp->_flags = _IO_MAGIC|flags; fp->_flags2 = 0; - if (stdio_needs_locking) - fp->_flags2 |= _IO_FLAGS2_NEED_LOCK; fp->_IO_buf_base = NULL; fp->_IO_buf_end = NULL; fp->_IO_read_base = NULL; diff --git a/libio/getc.c b/libio/getc.c index 8c79cf702f..18647aa3e1 100644 --- a/libio/getc.c +++ b/libio/getc.c @@ -34,8 +34,6 @@ _IO_getc (FILE *fp) { int result; CHECK_FILE (fp, EOF); - if (!_IO_need_lock (fp)) - return _IO_getc_unlocked (fp); _IO_acquire_lock (fp); result = _IO_getc_unlocked (fp); _IO_release_lock (fp); diff --git a/libio/getchar.c b/libio/getchar.c index e0f4f471b3..4ab1e7b065 100644 --- a/libio/getchar.c +++ b/libio/getchar.c @@ -33,10 +33,8 @@ int getchar (void) { int result; - if (!_IO_need_lock (stdin)) - return _IO_getc_unlocked (stdin); _IO_acquire_lock (stdin); result = _IO_getc_unlocked (stdin); _IO_release_lock (stdin); return result; -} \ No newline at end of file +} diff --git a/libio/iofopncook.c b/libio/iofopncook.c index 16bcc7f4a3..5ecca41717 100644 --- a/libio/iofopncook.c +++ b/libio/iofopncook.c @@ -162,8 +162,6 @@ _IO_cookie_init (struct _IO_cookie_file *cfile, int read_write, _IO_mask_flags (&cfile->__fp.file, read_write, _IO_NO_READS+_IO_NO_WRITES+_IO_IS_APPENDING); - cfile->__fp.file._flags2 |= _IO_FLAGS2_NEED_LOCK; - /* We use a negative number different from -1 for _fileno to mark that this special stream is not associated with a real file, but still has to be treated as such. */ diff --git a/libio/ioungetc.c b/libio/ioungetc.c index 92d5572e21..547953dd16 100644 --- a/libio/ioungetc.c +++ b/libio/ioungetc.c @@ -33,8 +33,6 @@ ungetc (int c, FILE *fp) CHECK_FILE (fp, EOF); if (c == EOF) return EOF; - if (!_IO_need_lock (fp)) - return _IO_sputbackc (fp, (unsigned char) c); _IO_acquire_lock (fp); result = _IO_sputbackc (fp, (unsigned char) c); _IO_release_lock (fp); diff --git a/libio/libio.h b/libio/libio.h index 42ff6143af..e1e8ffdc54 100644 --- a/libio/libio.h +++ b/libio/libio.h @@ -88,7 +88,6 @@ typedef struct #define _IO_FLAGS2_USER_WBUF 8 #define _IO_FLAGS2_NOCLOSE 32 #define _IO_FLAGS2_CLOEXEC 64 -#define _IO_FLAGS2_NEED_LOCK 128 /* _IO_pos_BAD is an off64_t value indicating error, unknown, or EOF. */ #define _IO_pos_BAD ((off64_t) -1) @@ -212,9 +211,6 @@ extern int _IO_ftrylockfile (FILE *) __THROW; #define _IO_cleanup_region_end(_Doit) /**/ #endif -#define _IO_need_lock(_fp) \ - (((_fp)->_flags2 & _IO_FLAGS2_NEED_LOCK) != 0) - extern int _IO_vfscanf (FILE * __restrict, const char * __restrict, __gnuc_va_list, int *__restrict); extern __ssize_t _IO_padn (FILE *, int, __ssize_t); diff --git a/libio/memstream.c b/libio/memstream.c index 7ab61876ba..15022b72ee 100644 --- a/libio/memstream.c +++ b/libio/memstream.c @@ -92,9 +92,6 @@ __open_memstream (char **bufloc, size_t *sizeloc) new_f->fp.bufloc = bufloc; new_f->fp.sizeloc = sizeloc; - /* Disable single thread optimization. BZ 21735. */ - new_f->fp._sf._sbf._f._flags2 |= _IO_FLAGS2_NEED_LOCK; - return (FILE *) &new_f->fp._sf._sbf; } libc_hidden_def (__open_memstream) diff --git a/libio/putc.c b/libio/putc.c index 44e30649c9..f4cc2024b6 100644 --- a/libio/putc.c +++ b/libio/putc.c @@ -25,8 +25,6 @@ _IO_putc (int c, FILE *fp) { int result; CHECK_FILE (fp, EOF); - if (!_IO_need_lock (fp)) - return _IO_putc_unlocked (c, fp); _IO_acquire_lock (fp); result = _IO_putc_unlocked (c, fp); _IO_release_lock (fp); diff --git a/libio/wmemstream.c b/libio/wmemstream.c index 9366ef4aad..abaf421069 100644 --- a/libio/wmemstream.c +++ b/libio/wmemstream.c @@ -94,9 +94,6 @@ open_wmemstream (wchar_t **bufloc, size_t *sizeloc) new_f->fp.bufloc = bufloc; new_f->fp.sizeloc = sizeloc; - /* Disable single thread optimization. BZ 21735. */ - new_f->fp._sf._sbf._f._flags2 |= _IO_FLAGS2_NEED_LOCK; - return (FILE *) &new_f->fp._sf._sbf; } diff --git a/nptl/pthread_create.c b/nptl/pthread_create.c index e7a099acb7..4f45ea36bc 100644 --- a/nptl/pthread_create.c +++ b/nptl/pthread_create.c @@ -740,9 +740,6 @@ __pthread_create_2_1 (pthread_t *newthread, const pthread_attr_t *attr, collect_default_sched (pd); } - if (__glibc_unlikely (__nptl_nthreads == 1)) - _IO_enable_locks (); - /* Pass the descriptor to the caller. */ *newthread = (pthread_t) pd; diff --git a/stdio-common/flockfile.c b/stdio-common/flockfile.c index a5decb450f..49f72c69ab 100644 --- a/stdio-common/flockfile.c +++ b/stdio-common/flockfile.c @@ -22,7 +22,6 @@ void __flockfile (FILE *stream) { - stream->_flags2 |= _IO_FLAGS2_NEED_LOCK; _IO_lock_lock (*stream->_lock); } weak_alias (__flockfile, flockfile); diff --git a/sysdeps/mach/libc-lock.h b/sysdeps/mach/libc-lock.h index 225eb67f5a..ee38948d1e 100644 --- a/sysdeps/mach/libc-lock.h +++ b/sysdeps/mach/libc-lock.h @@ -106,9 +106,9 @@ typedef struct __libc_lock_recursive_opaque__ __libc_lock_recursive_t; __libc_lock_recursive_t *const __lock = &(NAME); \ void *__self = __libc_lock_owner_self (); \ int __r = 0; \ - if (__self == __lock->owner) \ + if (!SINGLE_THREAD_P && __self == __lock->owner) \ ++__lock->cnt; \ - else if ((__r = lll_trylock (__lock->lock)) == 0) \ + else if (!SINGLE_THREAD_P && (__r = lll_trylock (__lock->lock)) == 0) \ __lock->owner = __self, __lock->cnt = 1; \ __r; \ }) @@ -117,7 +117,7 @@ typedef struct __libc_lock_recursive_opaque__ __libc_lock_recursive_t; ({ \ __libc_lock_recursive_t *const __lock = &(NAME); \ void *__self = __libc_lock_owner_self (); \ - if (__self != __lock->owner) \ + if (!SINGLE_THREAD_P && __self != __lock->owner) \ { \ lll_lock (__lock->lock, 0); \ __lock->owner = __self; \ @@ -132,7 +132,8 @@ typedef struct __libc_lock_recursive_opaque__ __libc_lock_recursive_t; if (--__lock->cnt == 0) \ { \ __lock->owner = 0; \ - lll_unlock (__lock->lock, 0); \ + if (!SINGLE_THREAD_P) \ + lll_unlock (__lock->lock, 0); \ } \ }) diff --git a/sysdeps/nptl/libc-lock.h b/sysdeps/nptl/libc-lock.h index 6c2d6acfd1..abd84e71b4 100644 --- a/sysdeps/nptl/libc-lock.h +++ b/sysdeps/nptl/libc-lock.h @@ -86,7 +86,7 @@ typedef struct { int lock; int cnt; void *owner; } __libc_lock_recursive_t; # define __libc_lock_lock_recursive(NAME) \ do { \ void *self = THREAD_SELF; \ - if ((NAME).owner != self) \ + if (!SINGLE_THREAD_P && (NAME).owner != self) \ { \ lll_lock ((NAME).lock, LLL_PRIVATE); \ (NAME).owner = self; \ @@ -104,7 +104,7 @@ typedef struct { int lock; int cnt; void *owner; } __libc_lock_recursive_t; ({ \ int result = 0; \ void *self = THREAD_SELF; \ - if ((NAME).owner != self) \ + if (!SINGLE_THREAD_P && (NAME).owner != self) \ { \ if (lll_trylock ((NAME).lock) == 0) \ { \ @@ -131,7 +131,8 @@ typedef struct { int lock; int cnt; void *owner; } __libc_lock_recursive_t; if (--(NAME).cnt == 0) \ { \ (NAME).owner = NULL; \ - lll_unlock ((NAME).lock, LLL_PRIVATE); \ + if (!SINGLE_THREAD_P) \ + lll_unlock ((NAME).lock, LLL_PRIVATE); \ } \ } while (0) #else