diff options
author | Tulio Magno Quites Machado Filho <tuliom@linux.vnet.ibm.com> | 2015-12-31 14:55:21 -0200 |
---|---|---|
committer | Tulio Magno Quites Machado Filho <tuliom@linux.vnet.ibm.com> | 2015-12-31 14:55:21 -0200 |
commit | eca182fcad77681d5ebe9ab49f91d33ed85d8289 (patch) | |
tree | 67b302abe291c2db58d1a45ee1e98d553b467539 | |
parent | 335863ea7cbc2c4c2a1947039565b781cf488a8f (diff) | |
parent | b057b4813c9f05c3cedff0c74b58c9c9d583f09f (diff) | |
download | glibc-eca182fcad77681d5ebe9ab49f91d33ed85d8289.tar glibc-eca182fcad77681d5ebe9ab49f91d33ed85d8289.tar.gz glibc-eca182fcad77681d5ebe9ab49f91d33ed85d8289.tar.bz2 glibc-eca182fcad77681d5ebe9ab49f91d33ed85d8289.zip |
Merge branch 'release/2.18/master' into ibm/2.18/master
Conflicts:
NEWS
-rw-r--r-- | ChangeLog | 46 | ||||
-rw-r--r-- | NEWS | 9 | ||||
-rw-r--r-- | elf/rtld.c | 15 | ||||
-rw-r--r-- | libio/wstrops.c | 8 | ||||
-rw-r--r-- | localedata/sort-test.sh | 6 | ||||
-rw-r--r-- | localedata/xfrm-test.c | 52 | ||||
-rw-r--r-- | posix/fnmatch_loop.c | 5 | ||||
-rw-r--r-- | posix/tst-fnmatch3.c | 32 | ||||
-rw-r--r-- | stdlib/cxa_thread_atexit_impl.c | 12 | ||||
-rw-r--r-- | string/strxfrm_l.c | 499 | ||||
-rw-r--r-- | sysdeps/generic/ldsodefs.h | 3 |
11 files changed, 562 insertions, 125 deletions
@@ -1,3 +1,49 @@ +2015-12-31 Leonhard Holz <leonhard.holz@web.de> + + [BZ #16009] + * string/strxfrm_l.c (STRXFRM): Allocate fixed size cache for + weights and rules. Use do_xfrm_cached if data fits in cache, + do_xfrm otherwise. Moved former main loop to... + * (do_xfrm_cached): New function. + * (do_xfrm): Non-caching version of do_xfrm_cached. Uses + find_idx, find_position and stack_push. + * (find_idx): New function. + * (find_position): Likewise. + * localedata/sort-test.sh: Added test run for do_xfrm. + * localedata/xfrm-test.c (main): Added command line option + -nocache to run the test with strings that are too large for + the STRXFRM cache. + +2015-12-30 Paul Pluzhnikov <ppluzhnikov@google.com> + + [BZ #17269] + * libio/wstrops.c (_IO_wstr_overflow): Guard against integer overflow + (enlarge_userbuf): Likewise. + +2015-12-30 Andreas Schwab <schwab@suse.de> + + [BZ #18032] + * posix/fnmatch_loop.c (FCT): Remove extra increment when skipping + over collating symbol inside a bracket expression. Minor cleanup. + * posix/tst-fnmatch3.c (do_test): Add test case. + +2015-12-30 Florian Weimer <fweimer@redhat.com> + + [BZ #19018] + * stdlib/cxa_thread_atexit_impl.c (__cxa_thread_atexit_impl): + Mangle function pointer before storing it. + (__call_tls_dtors): Demangle function pointer before calling it. + +2015-12-30 Florian Weimer <fweimer@redhat.com> + + [BZ #18928] + * sysdeps/generic/ldsodefs.h (struct rtld_global_ro): Remove + _dl_pointer_guard member. + * elf/rtld.c (_rtld_global_ro): Remove _dl_pointer_guard + initializer. + (security_init): Always set up pointer guard. + (process_envvars): Do not process LD_POINTER_GUARD. + 2015-05-11 Andreas Schwab <schwab@suse.de> [BZ #18007] @@ -11,9 +11,12 @@ Version 2.18.1 14143, 14155, 14547, 14699, 15073, 15128, 15532, 15427, 15522, 15680, 15723, 15734, 15735, 15797, 15892, 15895, 15909, 15915, 15917, 15946, - 15996, 16072, 16150, 16169, 16387, 16414, 16430, 16431, 16510, 16617, - 16618, 16885, 16916, 16943, 16958, 17048, 17137, 17187, 17325, 17625, - 17630, 18007, 18104, 18287. + 15996, 16009, 16072, 16150, 16169, 16387, 16414, 16430, 16431, 16510, + 16617, 16618, 16885, 16916, 16943, 16958, 17048, 17137, 17187, 17269, + 17325, 17625, 17630, 18007, 18032, 18104, 18287, 18928, 19018. + +* The LD_POINTER_GUARD environment variable can no longer be used to + disable the pointer guard feature. It is always enabled. * CVE-2014-8121 The NSS backends shared internal state between the getXXent and getXXbyYY NSS calls for the same database, causing a denial-of-service diff --git a/elf/rtld.c b/elf/rtld.c index 91da88cdd1..b2fce266fb 100644 --- a/elf/rtld.c +++ b/elf/rtld.c @@ -162,7 +162,6 @@ struct rtld_global_ro _rtld_global_ro attribute_relro = ._dl_hwcap_mask = HWCAP_IMPORTANT, ._dl_lazy = 1, ._dl_fpu_control = _FPU_DEFAULT, - ._dl_pointer_guard = 1, ._dl_pagesize = EXEC_PAGESIZE, ._dl_inhibit_cache = 0, @@ -857,15 +856,12 @@ security_init (void) #endif /* Set up the pointer guard as well, if necessary. */ - if (GLRO(dl_pointer_guard)) - { - uintptr_t pointer_chk_guard = _dl_setup_pointer_guard (_dl_random, - stack_chk_guard); + uintptr_t pointer_chk_guard + = _dl_setup_pointer_guard (_dl_random, stack_chk_guard); #ifdef THREAD_SET_POINTER_GUARD - THREAD_SET_POINTER_GUARD (pointer_chk_guard); + THREAD_SET_POINTER_GUARD (pointer_chk_guard); #endif - __pointer_chk_guard_local = pointer_chk_guard; - } + __pointer_chk_guard_local = pointer_chk_guard; /* We do not need the _dl_random value anymore. The less information we leave behind, the better, so clear the @@ -2606,9 +2602,6 @@ process_envvars (enum mode *modep) GLRO(dl_use_load_bias) = envline[14] == '1' ? -1 : 0; break; } - - if (memcmp (envline, "POINTER_GUARD", 13) == 0) - GLRO(dl_pointer_guard) = envline[14] != '0'; break; case 14: diff --git a/libio/wstrops.c b/libio/wstrops.c index 828393081e..19193e4826 100644 --- a/libio/wstrops.c +++ b/libio/wstrops.c @@ -95,8 +95,11 @@ _IO_wstr_overflow (fp, c) wchar_t *old_buf = fp->_wide_data->_IO_buf_base; size_t old_wblen = _IO_wblen (fp); _IO_size_t new_size = 2 * old_wblen + 100; - if (new_size < old_wblen) + + if (__glibc_unlikely (new_size < old_wblen) + || __glibc_unlikely (new_size > SIZE_MAX / sizeof (wchar_t))) return EOF; + new_buf = (wchar_t *) (*((_IO_strfile *) fp)->_s._allocate_buffer) (new_size * sizeof (wchar_t)); @@ -186,6 +189,9 @@ enlarge_userbuf (_IO_FILE *fp, _IO_off64_t offset, int reading) return 1; _IO_size_t newsize = offset + 100; + if (__glibc_unlikely (newsize > SIZE_MAX / sizeof (wchar_t))) + return 1; + wchar_t *oldbuf = wd->_IO_buf_base; wchar_t *newbuf = (wchar_t *) (*((_IO_strfile *) fp)->_s._allocate_buffer) (newsize diff --git a/localedata/sort-test.sh b/localedata/sort-test.sh index b74ae9d17a..6001103ac3 100644 --- a/localedata/sort-test.sh +++ b/localedata/sort-test.sh @@ -49,11 +49,17 @@ for l in $lang; do ${common_objpfx}localedata/xfrm-test $id < $cns.in \ > ${common_objpfx}localedata/$cns.xout || here=1 cmp -s $cns.in ${common_objpfx}localedata/$cns.xout || here=1 + LOCPATH=${common_objpfx}localedata GCONV_PATH=${common_objpfx}/iconvdata \ + LC_ALL=$l ${test_program_prefix} \ + ${common_objpfx}localedata/xfrm-test $id -nocache < $cns.in \ + > ${common_objpfx}localedata/$cns.nocache.xout || here=1 + cmp -s $cns.in ${common_objpfx}localedata/$cns.nocache.xout || here=1 if test $here -eq 0; then echo "$l xfrm-test OK" else echo "$l xfrm-test FAIL" diff -u $cns.in ${common_objpfx}localedata/$cns.xout | sed 's/^/ /' + diff -u $cns.in ${common_objpfx}localedata/$cns.nocache.xout | sed 's/^/ /' status=1 fi done diff --git a/localedata/xfrm-test.c b/localedata/xfrm-test.c index 80fc1fd846..5fc6b0a801 100644 --- a/localedata/xfrm-test.c +++ b/localedata/xfrm-test.c @@ -23,7 +23,10 @@ #include <stdio.h> #include <stdlib.h> #include <string.h> +#include <stdbool.h> +/* Keep in sync with string/strxfrm_l.c. */ +#define SMALL_STR_SIZE 4095 struct lines { @@ -37,6 +40,7 @@ int main (int argc, char *argv[]) { int result = 0; + bool nocache = false; size_t nstrings, nstrings_max; struct lines *strings; char *line = NULL; @@ -44,7 +48,18 @@ main (int argc, char *argv[]) size_t n; if (argc < 2) - error (1, 0, "usage: %s <random seed>", argv[0]); + error (1, 0, "usage: %s <random seed> [-nocache]", argv[0]); + + if (argc == 3) + { + if (strcmp (argv[2], "-nocache") == 0) + nocache = true; + else + { + printf ("Unknown option %s!\n", argv[2]); + exit (1); + } + } setlocale (LC_ALL, ""); @@ -59,9 +74,9 @@ main (int argc, char *argv[]) while (1) { - char saved, *newp; - int needed; - int l; + char saved, *word, *newp; + size_t l, line_len, needed; + if (getline (&line, &len, stdin) < 0) break; @@ -83,10 +98,35 @@ main (int argc, char *argv[]) saved = line[l]; line[l] = '\0'; - needed = strxfrm (NULL, line, 0); + + if (nocache) + { + line_len = strlen (line); + word = malloc (line_len + SMALL_STR_SIZE + 1); + if (word == NULL) + { + printf ("malloc failed: %m\n"); + exit (1); + } + memset (word, ' ', SMALL_STR_SIZE); + memcpy (word + SMALL_STR_SIZE, line, line_len); + word[line_len + SMALL_STR_SIZE] = '\0'; + } + else + word = line; + + needed = strxfrm (NULL, word, 0); newp = malloc (needed + 1); - strxfrm (newp, line, needed + 1); + if (newp == NULL) + { + printf ("malloc failed: %m\n"); + exit (1); + } + strxfrm (newp, word, needed + 1); strings[nstrings].xfrm = newp; + + if (nocache) + free (word); line[l] = saved; ++nstrings; } diff --git a/posix/fnmatch_loop.c b/posix/fnmatch_loop.c index 078b98242e..5002ccbce1 100644 --- a/posix/fnmatch_loop.c +++ b/posix/fnmatch_loop.c @@ -951,14 +951,13 @@ FCT (pattern, string, string_end, no_leading_period, flags, ends, alloca_used) } else if (c == L('[') && *p == L('.')) { - ++p; while (1) { c = *++p; - if (c == '\0') + if (c == L('\0')) return FNM_NOMATCH; - if (*p == L('.') && p[1] == L(']')) + if (c == L('.') && p[1] == L(']')) break; } p += 2; diff --git a/posix/tst-fnmatch3.c b/posix/tst-fnmatch3.c new file mode 100644 index 0000000000..75bc00a2c5 --- /dev/null +++ b/posix/tst-fnmatch3.c @@ -0,0 +1,32 @@ +/* Test for fnmatch not reading past the end of the pattern. + Copyright (C) 2014-2015 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#include <fnmatch.h> + +int +do_test (void) +{ + if (fnmatch ("[[:alpha:]'[:alpha:]\0]", "a", 0) != FNM_NOMATCH) + return 1; + if (fnmatch ("[a[.\0.]]", "a", 0) != FNM_NOMATCH) + return 1; + return 0; +} + +#define TEST_FUNCTION do_test () +#include "../test-skeleton.c" diff --git a/stdlib/cxa_thread_atexit_impl.c b/stdlib/cxa_thread_atexit_impl.c index dfd4c7e694..6b7455de25 100644 --- a/stdlib/cxa_thread_atexit_impl.c +++ b/stdlib/cxa_thread_atexit_impl.c @@ -42,6 +42,10 @@ static __thread struct link_map *lm_cache; int __cxa_thread_atexit_impl (dtor_func func, void *obj, void *dso_symbol) { +#ifdef PTR_MANGLE + PTR_MANGLE (func); +#endif + /* Prepend. */ struct dtor_list *new = calloc (1, sizeof (struct dtor_list)); new->func = func; @@ -83,9 +87,13 @@ __call_tls_dtors (void) while (tls_dtor_list) { struct dtor_list *cur = tls_dtor_list; - tls_dtor_list = tls_dtor_list->next; + dtor_func func = cur->func; +#ifdef PTR_DEMANGLE + PTR_DEMANGLE (func); +#endif - cur->func (cur->obj); + tls_dtor_list = tls_dtor_list->next; + func (cur->obj); __rtld_lock_lock_recursive (GL(dl_load_lock)); diff --git a/string/strxfrm_l.c b/string/strxfrm_l.c index 3812ed6096..2e72bbb6ec 100644 --- a/string/strxfrm_l.c +++ b/string/strxfrm_l.c @@ -40,8 +40,23 @@ #define CONCAT(a,b) CONCAT1(a,b) #define CONCAT1(a,b) a##b +/* Maximum string size that is calculated with cached indices. Right now this + is an arbitrary value open to optimizations. SMALL_STR_SIZE * 4 has to be + lower than __MAX_ALLOCA_CUTOFF. Keep localedata/xfrm-test.c in sync. */ +#define SMALL_STR_SIZE 4095 + #include "../locale/localeinfo.h" +/* Group locale data for shorter parameter lists. */ +typedef struct +{ + uint_fast32_t nrules; + unsigned char *rulesets; + USTRING_TYPE *weights; + int32_t *table; + USTRING_TYPE *extra; + int32_t *indirect; +} locale_data_t; #ifndef WIDE_CHAR_VERSION @@ -80,115 +95,330 @@ utf8_encode (char *buf, int val) } #endif - -size_t -STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n, __locale_t l) +/* Find next weight and rule index. Inlined since called for every char. */ +static __always_inline size_t +find_idx (const USTRING_TYPE **us, int32_t *weight_idx, + unsigned char *rule_idx, const locale_data_t *l_data, const int pass) { - struct __locale_data *current = l->__locales[LC_COLLATE]; - uint_fast32_t nrules = current->values[_NL_ITEM_INDEX (_NL_COLLATE_NRULES)].word; - /* We don't assign the following values right away since it might be - unnecessary in case there are no rules. */ - const unsigned char *rulesets; - const int32_t *table; - const USTRING_TYPE *weights; - const USTRING_TYPE *extra; - const int32_t *indirect; - uint_fast32_t pass; - size_t needed; - size_t last_needed; - const USTRING_TYPE *usrc; - size_t srclen = STRLEN (src); - int32_t *idxarr; - unsigned char *rulearr; - size_t idxmax; - size_t idxcnt; - int use_malloc; + /* Prepare variables required by findidx(). */ + int32_t *table = l_data->table; + int32_t *indirect = l_data->indirect; + USTRING_TYPE *extra = l_data->extra; #include WEIGHT_H + int32_t tmp = findidx (us, -1); + *rule_idx = tmp >> 24; + int32_t idx = tmp & 0xffffff; + size_t len = l_data->weights[idx++]; - if (nrules == 0) + /* Skip over indices of previous levels. */ + for (int i = 0; i < pass; i++) { - if (n != 0) - STPNCPY (dest, src, MIN (srclen + 1, n)); - - return srclen; + idx += len; + len = l_data->weights[idx++]; } - rulesets = (const unsigned char *) - current->values[_NL_ITEM_INDEX (_NL_COLLATE_RULESETS)].string; - table = (const int32_t *) - current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_TABLE,SUFFIX))].string; - weights = (const USTRING_TYPE *) - current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_WEIGHT,SUFFIX))].string; - extra = (const USTRING_TYPE *) - current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_EXTRA,SUFFIX))].string; - indirect = (const int32_t *) - current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_INDIRECT,SUFFIX))].string; - use_malloc = 0; + *weight_idx = idx; + return len; +} - assert (((uintptr_t) table) % __alignof__ (table[0]) == 0); - assert (((uintptr_t) weights) % __alignof__ (weights[0]) == 0); - assert (((uintptr_t) extra) % __alignof__ (extra[0]) == 0); - assert (((uintptr_t) indirect) % __alignof__ (indirect[0]) == 0); +static int +find_position (const USTRING_TYPE *us, const locale_data_t *l_data, + const int pass) +{ + int32_t weight_idx; + unsigned char rule_idx; + const USTRING_TYPE *usrc = us; - /* Handle an empty string as a special case. */ - if (srclen == 0) - { - if (n != 0) - *dest = L('\0'); - return 0; - } + find_idx (&usrc, &weight_idx, &rule_idx, l_data, pass); + return l_data->rulesets[rule_idx * l_data->nrules + pass] & sort_position; +} - /* We need the elements of the string as unsigned values since they - are used as indeces. */ - usrc = (const USTRING_TYPE *) src; - - /* Perform the first pass over the string and while doing this find - and store the weights for each character. Since we want this to - be as fast as possible we are using `alloca' to store the temporary - values. But since there is no limit on the length of the string - we have to use `malloc' if the string is too long. We should be - very conservative here. */ - if (! __libc_use_alloca ((srclen + 1) * (sizeof (int32_t) + 1))) - { - idxarr = (int32_t *) malloc ((srclen + 1) * (sizeof (int32_t) + 1)); - rulearr = (unsigned char *) &idxarr[srclen]; - - if (idxarr == NULL) - /* No memory. Well, go with the stack then. - - XXX Once this implementation is stable we will handle this - differently. Instead of precomputing the indeces we will - do this in time. This means, though, that this happens for - every pass again. */ - goto try_stack; - use_malloc = 1; - } - else +/* Do the transformation. */ +static size_t +do_xfrm (const USTRING_TYPE *usrc, STRING_TYPE *dest, size_t n, + const locale_data_t *l_data) +{ + int32_t weight_idx; + unsigned char rule_idx; + uint_fast32_t pass; + size_t needed = 0; + size_t last_needed; + + /* Now the passes over the weights. */ + for (pass = 0; pass < l_data->nrules; ++pass) { - try_stack: - idxarr = (int32_t *) alloca (srclen * sizeof (int32_t)); - rulearr = (unsigned char *) alloca (srclen + 1); + size_t backw_len = 0; + last_needed = needed; + const USTRING_TYPE *cur = usrc; + const USTRING_TYPE *backw_start = NULL; + + /* We assume that if a rule has defined `position' in one section + this is true for all of them. */ + int position = find_position (cur, l_data, pass); + + if (position == 0) + { + while (*cur != L('\0')) + { + const USTRING_TYPE *pos = cur; + size_t len = find_idx (&cur, &weight_idx, &rule_idx, l_data, + pass); + int rule = l_data->rulesets[rule_idx * l_data->nrules + pass]; + + if ((rule & sort_forward) != 0) + { + /* Handle the pushed backward sequence. */ + if (backw_start != NULL) + { + for (size_t i = backw_len; i > 0; ) + { + int32_t weight_idx; + unsigned char rule_idx; + size_t len = find_idx (&backw_start, &weight_idx, + &rule_idx, l_data, pass); + if (needed + i < n) + for (size_t j = len; j > 0; j--) + dest[needed + i - j] = + l_data->weights[weight_idx++]; + + i -= len; + } + + needed += backw_len; + backw_start = NULL; + backw_len = 0; + } + + /* Now handle the forward element. */ + if (needed + len < n) + while (len-- > 0) + dest[needed++] = l_data->weights[weight_idx++]; + else + /* No more characters fit into the buffer. */ + needed += len; + } + else + { + /* Remember start of the backward sequence & track length. */ + if (backw_start == NULL) + backw_start = pos; + backw_len += len; + } + } + + + /* Handle the pushed backward sequence. */ + if (backw_start != NULL) + { + for (size_t i = backw_len; i > 0; ) + { + size_t len = find_idx (&backw_start, &weight_idx, &rule_idx, + l_data, pass); + if (needed + i < n) + for (size_t j = len; j > 0; j--) + dest[needed + i - j] = + l_data->weights[weight_idx++]; + + i -= len; + } + + needed += backw_len; + } + } + else + { + int val = 1; +#ifndef WIDE_CHAR_VERSION + char buf[7]; + size_t buflen; +#endif + size_t i; + + while (*cur != L('\0')) + { + const USTRING_TYPE *pos = cur; + size_t len = find_idx (&cur, &weight_idx, &rule_idx, l_data, + pass); + int rule = l_data->rulesets[rule_idx * l_data->nrules + pass]; + + if ((rule & sort_forward) != 0) + { + /* Handle the pushed backward sequence. */ + if (backw_start != NULL) + { + for (size_t p = backw_len; p > 0; p--) + { + size_t len; + int32_t weight_idx; + unsigned char rule_idx; + const USTRING_TYPE *backw_cur = backw_start; + + /* To prevent a warning init the used vars. */ + len = find_idx (&backw_cur, &weight_idx, + &rule_idx, l_data, pass); + + for (i = 1; i < p; i++) + len = find_idx (&backw_cur, &weight_idx, + &rule_idx, l_data, pass); + + if (len != 0) + { +#ifdef WIDE_CHAR_VERSION + if (needed + 1 + len < n) + { + dest[needed] = val; + for (i = 0; i < len; ++i) + dest[needed + 1 + i] = + l_data->weights[weight_idx + i]; + } + needed += 1 + len; +#else + buflen = utf8_encode (buf, val); + if (needed + buflen + len < n) + { + for (i = 0; i < buflen; ++i) + dest[needed + i] = buf[i]; + for (i = 0; i < len; ++i) + dest[needed + buflen + i] = + l_data->weights[weight_idx + i]; + } + needed += buflen + len; +#endif + val = 1; + } + else + ++val; + } + + backw_start = NULL; + backw_len = 0; + } + + /* Now handle the forward element. */ + if (len != 0) + { +#ifdef WIDE_CHAR_VERSION + if (needed + 1 + len < n) + { + dest[needed] = val; + for (i = 0; i < len; ++i) + dest[needed + 1 + i] = + l_data->weights[weight_idx + i]; + } + needed += 1 + len; +#else + buflen = utf8_encode (buf, val); + if (needed + buflen + len < n) + { + for (i = 0; i < buflen; ++i) + dest[needed + i] = buf[i]; + for (i = 0; i < len; ++i) + dest[needed + buflen + i] = + l_data->weights[weight_idx + i]; + } + needed += buflen + len; +#endif + val = 1; + } + else + ++val; + } + else + { + /* Remember start of the backward sequence & track length. */ + if (backw_start == NULL) + backw_start = pos; + backw_len++; + } + } + + /* Handle the pushed backward sequence. */ + if (backw_start != NULL) + { + for (size_t p = backw_len; p > 0; p--) + { + size_t len; + int32_t weight_idx; + unsigned char rule_idx; + const USTRING_TYPE *backw_cur = backw_start; + + /* To prevent a warning init the used vars. */ + len = find_idx (&backw_cur, &weight_idx, + &rule_idx, l_data, pass); + + for (i = 1; i < p; i++) + len = find_idx (&backw_cur, &weight_idx, + &rule_idx, l_data, pass); + + if (len != 0) + { +#ifdef WIDE_CHAR_VERSION + if (needed + 1 + len < n) + { + dest[needed] = val; + for (i = 0; i < len; ++i) + dest[needed + 1 + i] = + l_data->weights[weight_idx + i]; + } + needed += 1 + len; +#else + buflen = utf8_encode (buf, val); + if (needed + buflen + len < n) + { + for (i = 0; i < buflen; ++i) + dest[needed + i] = buf[i]; + for (i = 0; i < len; ++i) + dest[needed + buflen + i] = + l_data->weights[weight_idx + i]; + } + needed += buflen + len; +#endif + val = 1; + } + else + ++val; + } + } + } + + /* Finally store the byte to separate the passes or terminate + the string. */ + if (needed < n) + dest[needed] = pass + 1 < l_data->nrules ? L('\1') : L('\0'); + ++needed; } - idxmax = 0; - do + /* This is a little optimization: many collation specifications have + a `position' rule at the end and if no non-ignored character + is found the last \1 byte is immediately followed by a \0 byte + signalling this. We can avoid the \1 byte(s). */ + if (needed > 2 && needed == last_needed + 1) { - int32_t tmp = findidx (&usrc, -1); - rulearr[idxmax] = tmp >> 24; - idxarr[idxmax] = tmp & 0xffffff; - - ++idxmax; + /* Remove the \1 byte. */ + if (--needed <= n) + dest[needed - 1] = L('\0'); } - while (*usrc != L('\0')); - /* This element is only read, the value never used but to determine - another value which then is ignored. */ - rulearr[idxmax] = '\0'; + /* Return the number of bytes/words we need, but don't count the NUL + byte/word at the end. */ + return needed - 1; +} + +/* Do the transformation using weight-index and rule cache. */ +static size_t +do_xfrm_cached (STRING_TYPE *dest, size_t n, const locale_data_t *l_data, + size_t idxmax, int32_t *idxarr, const unsigned char *rulearr) +{ + uint_fast32_t nrules = l_data->nrules; + unsigned char *rulesets = l_data->rulesets; + USTRING_TYPE *weights = l_data->weights; + uint_fast32_t pass; + size_t needed = 0; + size_t last_needed; + size_t idxcnt; - /* Now the passes over the weights. We now use the indeces we found - before. */ - needed = 0; + /* Now the passes over the weights. */ for (pass = 0; pass < nrules; ++pass) { size_t backw_stop = ~0ul; @@ -434,14 +664,91 @@ STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n, __locale_t l) dest[needed - 1] = L('\0'); } - /* Free the memory if needed. */ - if (use_malloc) - free (idxarr); - /* Return the number of bytes/words we need, but don't count the NUL byte/word at the end. */ return needed - 1; } + +size_t +STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n, __locale_t l) +{ + locale_data_t l_data; + struct __locale_data *current = l->__locales[LC_COLLATE]; + l_data.nrules = current->values[_NL_ITEM_INDEX (_NL_COLLATE_NRULES)].word; + + /* Handle byte comparison case. */ + if (l_data.nrules == 0) + { + size_t srclen = STRLEN (src); + + if (n != 0) + STPNCPY (dest, src, MIN (srclen + 1, n)); + + return srclen; + } + + /* Handle an empty string, code hereafter relies on strlen (src) > 0. */ + if (*src == L('\0')) + { + if (n != 0) + *dest = L('\0'); + return 0; + } + + /* Get the locale data. */ + l_data.rulesets = (unsigned char *) + current->values[_NL_ITEM_INDEX (_NL_COLLATE_RULESETS)].string; + l_data.table = (int32_t *) + current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_TABLE,SUFFIX))].string; + l_data.weights = (USTRING_TYPE *) + current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_WEIGHT,SUFFIX))].string; + l_data.extra = (USTRING_TYPE *) + current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_EXTRA,SUFFIX))].string; + l_data.indirect = (int32_t *) + current->values[_NL_ITEM_INDEX (CONCAT(_NL_COLLATE_INDIRECT,SUFFIX))].string; + + assert (((uintptr_t) l_data.table) % __alignof__ (l_data.table[0]) == 0); + assert (((uintptr_t) l_data.weights) % __alignof__ (l_data.weights[0]) == 0); + assert (((uintptr_t) l_data.extra) % __alignof__ (l_data.extra[0]) == 0); + assert (((uintptr_t) l_data.indirect) % __alignof__ (l_data.indirect[0]) == 0); + + /* We need the elements of the string as unsigned values since they + are used as indeces. */ + const USTRING_TYPE *usrc = (const USTRING_TYPE *) src; + + /* Allocate cache for small strings on the stack and fill it with weight and + rule indices. If the cache size is not sufficient, continue with the + uncached xfrm version. */ + size_t idxmax = 0; + const USTRING_TYPE *cur = usrc; + int32_t *idxarr = alloca (SMALL_STR_SIZE * sizeof (int32_t)); + unsigned char *rulearr = alloca (SMALL_STR_SIZE + 1); + /* Prepare variables required by findidx(). */ + int32_t *table = l_data.table; + int32_t *indirect = l_data.indirect; + USTRING_TYPE *extra = l_data.extra; +#include WEIGHT_H + + do + { + int32_t tmp = findidx (&cur, -1); + rulearr[idxmax] = tmp >> 24; + idxarr[idxmax] = tmp & 0xffffff; + + ++idxmax; + } + while (*cur != L('\0') && idxmax < SMALL_STR_SIZE); + + /* This element is only read, the value never used but to determine + another value which then is ignored. */ + rulearr[idxmax] = '\0'; + + /* Do the transformation. */ + if (*cur == L('\0')) + return do_xfrm_cached (dest, n, &l_data, idxmax, idxarr, rulearr); + else + return do_xfrm (usrc, dest, n, &l_data); +} libc_hidden_def (STRXFRM) #ifndef WIDE_CHAR_VERSION diff --git a/sysdeps/generic/ldsodefs.h b/sysdeps/generic/ldsodefs.h index e7b0516aaf..6ccbd7138e 100644 --- a/sysdeps/generic/ldsodefs.h +++ b/sysdeps/generic/ldsodefs.h @@ -589,9 +589,6 @@ struct rtld_global_ro /* List of auditing interfaces. */ struct audit_ifaces *_dl_audit; unsigned int _dl_naudit; - - /* 0 if internal pointer values should not be guarded, 1 if they should. */ - EXTERN int _dl_pointer_guard; }; # define __rtld_global_attribute__ # ifdef IS_IN_rtld |