From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qv1-xf2f.google.com (mail-qv1-xf2f.google.com [IPv6:2607:f8b0:4864:20::f2f]) by sourceware.org (Postfix) with ESMTPS id 0A64338708EF for ; Mon, 28 Dec 2020 13:59:56 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org 0A64338708EF Received: by mail-qv1-xf2f.google.com with SMTP id et9so4940460qvb.10 for ; Mon, 28 Dec 2020 05:59:56 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=CmjILUq3e+tzu0+DoNa/P85jUVwSWxrmFynE6BdtRKc=; b=nWfcAqfSN9s1ywKjefyOJn4d8DHA8mFFjEmy94mTuhHuVi0YEb9XsVe0XheC5h1B8t QUihy8u2RKhIbjyeOyLNQ1C+MT75k/jtcYcDAXhkULdeJcRu3pSs9VZbJkj6ai1vPuZn HyPM3+cytF68tzSdrSlYlEEv1Ho8nF3AKTEAbCxEpwU8XFzyFfRnPzIATl0MkSbUtUO4 ohB82gDuMjMcCVlYkOVGJzF4fVUwwF24JECWo4G1dRdd6NXQ+y0RNKwT2gNo7Ao+f3Mf Kwp5lm+vVqzWZXVAK3XZyJkUf8lk8PKu4jZjtZZfBt+zpWv+NvweOrW+glPRjnUVUgoJ Pxwg== X-Gm-Message-State: AOAM532HqMDIN8tkBNDdjIajZu/QCaD/d6QVg+wjDiFCAEXQjQQ53Akd yi/Xelj1Gk4ga8vD1Q82KsE5wDiqtJPIlw== X-Google-Smtp-Source: ABdhPJzNUsoF0rYz8a+PbSM7y7/ebUxBvYtq/gFeDmy7OwI42DJJKTrVlK6iFfwPQUTDIZ4paXHZIw== X-Received: by 2002:a05:6214:1485:: with SMTP id bn5mr47614846qvb.17.1609163994888; Mon, 28 Dec 2020 05:59:54 -0800 (PST) Received: from localhost.localdomain ([177.194.48.209]) by smtp.googlemail.com with ESMTPSA id i17sm24243744qtg.77.2020.12.28.05.59.53 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 28 Dec 2020 05:59:54 -0800 (PST) From: Adhemerval Zanella To: libc-alpha@sourceware.org, Paul Eggert Cc: bug-gnulib@gnu.org Subject: [PATCH v2 4/6] stdlib: Sync canonicalize with gnulib [BZ #10635] [BZ #26592] [BZ #26341] [BZ #24970] Date: Mon, 28 Dec 2020 10:59:42 -0300 Message-Id: <20201228135944.1018303-4-adhemerval.zanella@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20201228135944.1018303-1-adhemerval.zanella@linaro.org> References: <20201228135944.1018303-1-adhemerval.zanella@linaro.org> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Spam-Status: No, score=-13.7 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, RCVD_IN_DNSWL_NONE, 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: Mon, 28 Dec 2020 13:59:58 -0000 It sync with gnulib version bbaba6ce5 with the following difference require fix a glibc build: --- stdlib/canonicalize.c +++ lib/canonicalize-lgpl.c @@ -52,7 +52,6 @@ # endif # define GCC_LINT 1 # define _GL_ATTRIBUTE_PURE __attribute__ ((__pure__)) -# define FUNC_REALPATH_WORKS 1 #else # define __canonicalize_file_name canonicalize_file_name # define __realpath realpath (it is due the glibc is built with -Wundef -Werror). It fixes multiple realpath issues: - Portability fixes for errno clobbering on free (BZ#10635). The function does not call free directly anymore, although it might be done through scratch_buffer_free. The free errno clobbering is being tracked by BZ#17924. - Pointer arithmetic overflows in realpath (BZ#26592). - Realpath cyclically call __alloca(path_max) to consume too much stack space (BZ#26341). - Realpath mishandles EOVERFLOW; stat not needed anyway (BZ#24970). The check is done through faccessat now. Checked on x86_64-linux-gnu. --- stdlib/canonicalize.c | 564 ++++++++++++++++++++-------- sysdeps/unix/sysv/linux/faccessat.c | 3 +- 2 files changed, 403 insertions(+), 164 deletions(-) diff --git a/stdlib/canonicalize.c b/stdlib/canonicalize.c index 3fcb399a5d..24a027b83f 100644 --- a/stdlib/canonicalize.c +++ b/stdlib/canonicalize.c @@ -16,43 +16,212 @@ License along with the GNU C Library; if not, see . */ -#include +#ifndef _LIBC +/* Don't use __attribute__ __nonnull__ in this compilation unit. Otherwise gcc + optimizes away the name == NULL test below. */ +# define _GL_ARG_NONNULL(params) + +# define _GL_USE_STDLIB_ALLOC 1 +# include +#endif + +/* Specification. */ #include -#include -#include -#include -#include + #include +#include +#include +#include #include +#include +#include +#include #include -#include +#include +#include +#include + +#ifdef _LIBC +# include +# include +# ifdef __ASSUME_FACCESSAT2 +# define FACCESSAT_NEVER_EOVERFLOWS __ASSUME_FACCESSAT2 +# else +# define FACCESSAT_NEVER_EOVERFLOWS true +# endif +# define GCC_LINT 1 +# define _GL_ATTRIBUTE_PURE __attribute__ ((__pure__)) +# define FUNC_REALPATH_WORKS 1 +#else +# define __canonicalize_file_name canonicalize_file_name +# define __realpath realpath +# include "pathmax.h" +# define __faccessat faccessat +# if defined _WIN32 && !defined __CYGWIN__ +# define __getcwd _getcwd +# elif HAVE_GETCWD +# if IN_RELOCWRAPPER + /* When building the relocatable program wrapper, use the system's getcwd + function, not the gnulib override, otherwise we would get a link error. + */ +# undef getcwd +# endif +# if defined VMS && !defined getcwd + /* We want the directory in Unix syntax, not in VMS syntax. + The gnulib override of 'getcwd' takes 2 arguments; the original VMS + 'getcwd' takes 3 arguments. */ +# define __getcwd(buf, max) getcwd (buf, max, 0) +# else +# define __getcwd getcwd +# endif +# else +# define __getcwd(buf, max) getwd (buf) +# endif +# define __mempcpy mempcpy +# define __pathconf pathconf +# define __rawmemchr rawmemchr +# define __readlink readlink +# define __stat stat +#endif -/* Return the canonical absolute name of file NAME. A canonical name - does not contain any `.', `..' components nor any repeated path - separators ('/') or symlinks. All path components must exist. If - RESOLVED is null, the result is malloc'd; otherwise, if the - canonical name is PATH_MAX chars or more, returns null with `errno' - set to ENAMETOOLONG; if the name fits in fewer than PATH_MAX chars, - returns the name in RESOLVED. If the name cannot be resolved and - RESOLVED is non-NULL, it contains the path of the first component - that cannot be resolved. If the path can be resolved, RESOLVED - holds the same value as the value returned. */ +/* Suppress bogus GCC -Wmaybe-uninitialized warnings. */ +#if defined GCC_LINT || defined lint +# define IF_LINT(Code) Code +#else +# define IF_LINT(Code) /* empty */ +#endif -char * -__realpath (const char *name, char *resolved) +#ifndef DOUBLE_SLASH_IS_DISTINCT_ROOT +# define DOUBLE_SLASH_IS_DISTINCT_ROOT false +#endif +#ifndef FACCESSAT_NEVER_EOVERFLOWS +# define FACCESSAT_NEVER_EOVERFLOWS false +#endif + +#if !FUNC_REALPATH_WORKS || defined _LIBC + +/* Return true if FILE's existence can be shown, false (setting errno) + otherwise. Follow symbolic links. */ +static bool +file_accessible (char const *file) +{ +# if defined _LIBC || HAVE_FACCESSAT + int r = __faccessat (AT_FDCWD, file, F_OK, AT_EACCESS); +# else + struct stat st; + int r = __stat (file, &st); +# endif + + return ((!FACCESSAT_NEVER_EOVERFLOWS && r < 0 && errno == EOVERFLOW) + || r == 0); +} + +/* True if concatenating END as a suffix to a file name means that the + code needs to check that the file name is that of a searchable + directory, since the canonicalize_filename_mode_stk code won't + check this later anyway when it checks an ordinary file name + component within END. END must either be empty, or start with a + slash. */ + +static bool _GL_ATTRIBUTE_PURE +suffix_requires_dir_check (char const *end) +{ + /* If END does not start with a slash, the suffix is OK. */ + while (ISSLASH (*end)) + { + /* Two or more slashes act like a single slash. */ + do + end++; + while (ISSLASH (*end)); + + switch (*end++) + { + default: return false; /* An ordinary file name component is OK. */ + case '\0': return true; /* Trailing "/" is trouble. */ + case '.': break; /* Possibly "." or "..". */ + } + /* Trailing "/.", or "/.." even if not trailing, is trouble. */ + if (!*end || (*end == '.' && (!end[1] || ISSLASH (end[1])))) + return true; + } + + return false; +} + +/* Append this to a file name to test whether it is a searchable directory. + On POSIX platforms "/" suffices, but "/./" is sometimes needed on + macOS 10.13 , and should also work on + platforms like AIX 7.2 that need at least "/.". */ + +#if defined _LIBC || defined LSTAT_FOLLOWS_SLASHED_SYMLINK +static char const dir_suffix[] = "/"; +#else +static char const dir_suffix[] = "/./"; +#endif + +/* Return true if DIR is a searchable dir, false (setting errno) otherwise. + DIREND points to the NUL byte at the end of the DIR string. + Store garbage into DIREND[0 .. strlen (dir_suffix)]. */ + +static bool +dir_check (char *dir, char *dirend) +{ + strcpy (dirend, dir_suffix); + return file_accessible (dir); +} + +static idx_t +get_path_max (void) +{ +# ifdef PATH_MAX + long int path_max = PATH_MAX; +# else + /* The caller invoked realpath with a null RESOLVED, even though + PATH_MAX is not defined as a constant. The glibc manual says + programs should not do this, and POSIX says the behavior is undefined. + Historically, glibc here used the result of pathconf, or 1024 if that + failed; stay consistent with this (dubious) historical practice. */ + int err = errno; + long int path_max = __pathconf ("/", _PC_PATH_MAX); + __set_errno (err); +# endif + return path_max < 0 ? 1024 : path_max <= IDX_MAX ? path_max : IDX_MAX; +} + +/* Act like __realpath (see below), with an additional argument + rname_buf that can be used as temporary storage. + + If GCC_LINT is defined, do not inline this function with GCC 10.1 + and later, to avoid creating a pointer to the stack that GCC + -Wreturn-local-addr incorrectly complains about. See: + https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93644 + Although the noinline attribute can hurt performance a bit, no better way + to pacify GCC is known; even an explicit #pragma does not pacify GCC. + When the GCC bug is fixed this workaround should be limited to the + broken GCC versions. */ +#if __GNUC_PREREQ (10, 1) +# if defined GCC_LINT || defined lint +__attribute__ ((__noinline__)) +# elif __OPTIMIZE__ && !__NO_INLINE__ +# define GCC_BOGUS_WRETURN_LOCAL_ADDR +# endif +#endif +static char * +realpath_stk (const char *name, char *resolved, + struct scratch_buffer *rname_buf) { - char *rpath, *dest, *extra_buf = NULL; - const char *start, *end, *rpath_limit; - long int path_max; + char *dest; + char const *start; + char const *end; int num_links = 0; if (name == NULL) { /* As per Single Unix Specification V2 we must return an error if - either parameter is a null pointer. We extend this to allow - the RESOLVED parameter to be NULL in case the we are expected to - allocate the room for the return value. */ + either parameter is a null pointer. We extend this to allow + the RESOLVED parameter to be NULL in case the we are expected to + allocate the room for the return value. */ __set_errno (EINVAL); return NULL; } @@ -60,166 +229,235 @@ __realpath (const char *name, char *resolved) if (name[0] == '\0') { /* As per Single Unix Specification V2 we must return an error if - the name argument points to an empty string. */ + the name argument points to an empty string. */ __set_errno (ENOENT); return NULL; } -#ifdef PATH_MAX - path_max = PATH_MAX; -#else - path_max = __pathconf (name, _PC_PATH_MAX); - if (path_max <= 0) - path_max = 1024; -#endif + struct scratch_buffer extra_buffer, link_buffer; + scratch_buffer_init (&extra_buffer); + scratch_buffer_init (&link_buffer); + scratch_buffer_init (rname_buf); + char *rname_on_stack = rname_buf->data; + char *rname = rname_on_stack; + bool end_in_extra_buffer = false; + bool failed = true; - if (resolved == NULL) - { - rpath = malloc (path_max); - if (rpath == NULL) - return NULL; - } - else - rpath = resolved; - rpath_limit = rpath + path_max; + /* This is always zero for Posix hosts, but can be 2 for MS-Windows + and MS-DOS X:/foo/bar file names. */ + idx_t prefix_len = FILE_SYSTEM_PREFIX_LEN (name); - if (name[0] != '/') + if (!IS_ABSOLUTE_FILE_NAME (name)) { - if (!__getcwd (rpath, path_max)) - { - rpath[0] = '\0'; - goto error; - } - dest = __rawmemchr (rpath, '\0'); + while (!__getcwd (rname, rname_buf->length)) + { + if (errno != ERANGE) + { + dest = rname; + goto error; + } + if (!scratch_buffer_grow (rname_buf)) + goto error_nomem; + rname = rname_buf->data; + } + dest = __rawmemchr (rname, '\0'); + start = name; + prefix_len = FILE_SYSTEM_PREFIX_LEN (rname); } else { - rpath[0] = '/'; - dest = rpath + 1; + dest = __mempcpy (rname, name, prefix_len); + *dest++ = '/'; + if (DOUBLE_SLASH_IS_DISTINCT_ROOT) + { + if (prefix_len == 0 /* implies ISSLASH (name[0]) */ + && ISSLASH (name[1]) && !ISSLASH (name[2])) + *dest++ = '/'; + *dest = '\0'; + } + start = name + prefix_len; } - for (start = end = name; *start; start = end) + for ( ; *start; start = end) { - struct stat64 st; - int n; - - /* Skip sequence of multiple path-separators. */ - while (*start == '/') - ++start; - - /* Find end of path component. */ - for (end = start; *end && *end != '/'; ++end) - /* Nothing. */; - - if (end - start == 0) - break; - else if (end - start == 1 && start[0] == '.') - /* nothing */; - else if (end - start == 2 && start[0] == '.' && start[1] == '.') - { - /* Back up to previous component, ignore if at root already. */ - if (dest > rpath + 1) - while ((--dest)[-1] != '/'); - } + /* Skip sequence of multiple file name separators. */ + while (ISSLASH (*start)) + ++start; + + /* Find end of component. */ + for (end = start; *end && !ISSLASH (*end); ++end) + /* Nothing. */; + + /* Length of this file name component; it can be zero if a file + name ends in '/'. */ + idx_t startlen = end - start; + + if (startlen == 0) + break; + else if (startlen == 1 && start[0] == '.') + /* nothing */; + else if (startlen == 2 && start[0] == '.' && start[1] == '.') + { + /* Back up to previous component, ignore if at root already. */ + if (dest > rname + prefix_len + 1) + for (--dest; dest > rname && !ISSLASH (dest[-1]); --dest) + continue; + if (DOUBLE_SLASH_IS_DISTINCT_ROOT + && dest == rname + 1 && !prefix_len + && ISSLASH (*dest) && !ISSLASH (dest[1])) + dest++; + } else - { - size_t new_size; - - if (dest[-1] != '/') - *dest++ = '/'; - - if (dest + (end - start) >= rpath_limit) - { - ptrdiff_t dest_offset = dest - rpath; - char *new_rpath; - - if (resolved) - { - __set_errno (ENAMETOOLONG); - if (dest > rpath + 1) - dest--; - *dest = '\0'; - goto error; - } - new_size = rpath_limit - rpath; - if (end - start + 1 > path_max) - new_size += end - start + 1; - else - new_size += path_max; - new_rpath = (char *) realloc (rpath, new_size); - if (new_rpath == NULL) - goto error; - rpath = new_rpath; - rpath_limit = rpath + new_size; - - dest = rpath + dest_offset; - } - - dest = __mempcpy (dest, start, end - start); - *dest = '\0'; - - if (__lstat64 (rpath, &st) < 0) - goto error; - - if (S_ISLNK (st.st_mode)) - { - char *buf = __alloca (path_max); - size_t len; - - if (++num_links > __eloop_threshold ()) - { - __set_errno (ELOOP); - goto error; - } - - n = __readlink (rpath, buf, path_max - 1); - if (n < 0) - goto error; - buf[n] = '\0'; - - if (!extra_buf) - extra_buf = __alloca (path_max); - - len = strlen (end); - if (path_max - n <= len) - { - __set_errno (ENAMETOOLONG); - goto error; - } - - /* Careful here, end may be a pointer into extra_buf... */ - memmove (&extra_buf[n], end, len + 1); - name = end = memcpy (extra_buf, buf, n); - - if (buf[0] == '/') - dest = rpath + 1; /* It's an absolute symlink */ - else - /* Back up to previous component, ignore if at root already: */ - if (dest > rpath + 1) - while ((--dest)[-1] != '/'); - } - else if (!S_ISDIR (st.st_mode) && *end != '\0') - { - __set_errno (ENOTDIR); - goto error; - } - } + { + if (!ISSLASH (dest[-1])) + *dest++ = '/'; + + while (rname + rname_buf->length - dest + < startlen + sizeof dir_suffix) + { + idx_t dest_offset = dest - rname; + if (!scratch_buffer_grow_preserve (rname_buf)) + goto error_nomem; + rname = rname_buf->data; + dest = rname + dest_offset; + } + + dest = __mempcpy (dest, start, startlen); + *dest = '\0'; + + char *buf; + ssize_t n; + while (true) + { + buf = link_buffer.data; + idx_t bufsize = link_buffer.length; + n = __readlink (rname, buf, bufsize - 1); + if (n < bufsize - 1) + break; + if (!scratch_buffer_grow (&link_buffer)) + goto error_nomem; + } + if (0 <= n) + { + if (++num_links > __eloop_threshold ()) + { + __set_errno (ELOOP); + goto error; + } + + buf[n] = '\0'; + + char *extra_buf = extra_buffer.data; + idx_t end_idx IF_LINT (= 0); + if (end_in_extra_buffer) + end_idx = end - extra_buf; + idx_t len = strlen (end); + while (extra_buffer.length <= len + n) + { + if (!scratch_buffer_grow_preserve (&extra_buffer)) + goto error_nomem; + extra_buf = extra_buffer.data; + } + if (end_in_extra_buffer) + end = extra_buf + end_idx; + + /* Careful here, end may be a pointer into extra_buf... */ + memmove (&extra_buf[n], end, len + 1); + name = end = memcpy (extra_buf, buf, n); + end_in_extra_buffer = true; + + if (IS_ABSOLUTE_FILE_NAME (buf)) + { + idx_t pfxlen = FILE_SYSTEM_PREFIX_LEN (buf); + + dest = __mempcpy (rname, buf, pfxlen); + *dest++ = '/'; /* It's an absolute symlink */ + if (DOUBLE_SLASH_IS_DISTINCT_ROOT) + { + if (ISSLASH (buf[1]) && !ISSLASH (buf[2]) && !pfxlen) + *dest++ = '/'; + *dest = '\0'; + } + /* Install the new prefix to be in effect hereafter. */ + prefix_len = pfxlen; + } + else + { + /* Back up to previous component, ignore if at root + already: */ + if (dest > rname + prefix_len + 1) + for (--dest; dest > rname && !ISSLASH (dest[-1]); --dest) + continue; + if (DOUBLE_SLASH_IS_DISTINCT_ROOT && dest == rname + 1 + && ISSLASH (*dest) && !ISSLASH (dest[1]) && !prefix_len) + dest++; + } + } + else if (! (suffix_requires_dir_check (end) + ? dir_check (rname, dest) + : errno == EINVAL)) + goto error; + } } - if (dest > rpath + 1 && dest[-1] == '/') + if (dest > rname + prefix_len + 1 && ISSLASH (dest[-1])) --dest; - *dest = '\0'; - - assert (resolved == NULL || resolved == rpath); - return rpath; + if (DOUBLE_SLASH_IS_DISTINCT_ROOT && dest == rname + 1 && !prefix_len + && ISSLASH (*dest) && !ISSLASH (dest[1])) + dest++; + failed = false; error: - assert (resolved == NULL || resolved == rpath); - if (resolved == NULL) - free (rpath); - return NULL; + *dest++ = '\0'; + if (resolved != NULL && dest - rname <= get_path_max ()) + rname = strcpy (resolved, rname); + +error_nomem: + scratch_buffer_free (&extra_buffer); + scratch_buffer_free (&link_buffer); + if (failed || rname == resolved) + scratch_buffer_free (rname_buf); + + if (failed) + return NULL; + + if (rname == resolved) + return rname; + idx_t rname_size = dest - rname; + if (rname == rname_on_stack) + { + rname = malloc (rname_size); + if (rname == NULL) + return NULL; + return memcpy (rname, rname_on_stack, rname_size); + } + char *result = realloc (rname, rname_size); + return result != NULL ? result : rname; +} + +/* Return the canonical absolute name of file NAME. A canonical name + does not contain any ".", ".." components nor any repeated file name + separators ('/') or symlinks. All file name components must exist. If + RESOLVED is null, the result is malloc'd; otherwise, if the + canonical name is PATH_MAX chars or more, returns null with 'errno' + set to ENAMETOOLONG; if the name fits in fewer than PATH_MAX chars, + returns the name in RESOLVED. If the name cannot be resolved and + RESOLVED is non-NULL, it contains the name of the first component + that cannot be resolved. If the name can be resolved, RESOLVED + holds the same value as the value returned. */ + +char * +__realpath (const char *name, char *resolved) +{ + #ifdef GCC_BOGUS_WRETURN_LOCAL_ADDR + #warning "GCC might issue a bogus -Wreturn-local-addr warning here." + #warning "See ." + #endif + struct scratch_buffer rname_buffer; + return realpath_stk (name, resolved, &rname_buffer); } libc_hidden_def (__realpath) versioned_symbol (libc, __realpath, realpath, GLIBC_2_3); +#endif /* !FUNC_REALPATH_WORKS || defined _LIBC */ #if SHLIB_COMPAT(libc, GLIBC_2_0, GLIBC_2_3) diff --git a/sysdeps/unix/sysv/linux/faccessat.c b/sysdeps/unix/sysv/linux/faccessat.c index 5d078371b5..5bb1051c06 100644 --- a/sysdeps/unix/sysv/linux/faccessat.c +++ b/sysdeps/unix/sysv/linux/faccessat.c @@ -24,7 +24,7 @@ int -faccessat (int fd, const char *file, int mode, int flag) +__faccessat (int fd, const char *file, int mode, int flag) { int ret = INLINE_SYSCALL_CALL (faccessat2, fd, file, mode, flag); #if __ASSUME_FACCESSAT2 @@ -73,3 +73,4 @@ faccessat (int fd, const char *file, int mode, int flag) return INLINE_SYSCALL_ERROR_RETURN_VALUE (EACCES); #endif /* !__ASSUME_FACCESSAT2 */ } +weak_alias (__faccessat, faccessat) -- 2.25.1