commit dc258ce62ae0bbb456c6a855dbb6b384ecf7e988 Author: Florian Weimer Date: Thu Aug 3 13:59:17 2017 +0200 getaddrinfo: Release resolver context on error in gethosts [BZ #21885] (cherry picked from commit 964263bb8d650f1681665c55704fb01a8e725621) diff --git a/ChangeLog b/ChangeLog index 8dbfc7eaff..28ce9c8479 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-03 Florian Weimer + + [BZ #21885] + * sysdeps/posix/getaddrinfo.c (gethosts): Release resolver context + on memory allocation failure. + 2017-08-02 Siddhesh Poyarekar * version.h (RELEASE): Set to "stable" diff --git a/NEWS b/NEWS index 8295f20c0a..9a64579658 100644 --- a/NEWS +++ b/NEWS @@ -5,6 +5,12 @@ See the end for copying conditions. Please send GNU C library bug reports via using `glibc' in the "product" field. +Version 2.26.1 + +The following bugs are resolved with this release: + + [21885] getaddrinfo: Release resolver context on error in gethosts + Version 2.26 Major new features: diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index efa7118498..699411cc92 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -255,6 +255,8 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, break; \ if (!scratch_buffer_grow (tmpbuf)) \ { \ + __resolv_context_enable_inet6 (res_ctx, res_enable_inet6); \ + __resolv_context_put (res_ctx); \ result = -EAI_MEMORY; \ goto free_and_return; \ } \ commit 665ce88d68fd13c5c4cbaf2808434c618745137c Author: Aurelien Jarno Date: Thu Aug 3 22:33:19 2017 +0000 i686/multiarch: Regenerate ulps This comes from running “make regen-ulps” on an AMD Opteron 2378 CPU. Changelog: * sysdeps/i386/i686/fpu/multiarch/libm-test-ulps: Regenerated. (cherry picked from commit 144bdab050bdd78c327d98a79b15c5b164e1d56f) diff --git a/ChangeLog b/ChangeLog index 28ce9c8479..3478699b10 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,7 @@ +2017-08-03 Aurelien Jarno + + * sysdeps/i386/i686/fpu/multiarch/libm-test-ulps: Regenerated. + 2017-08-03 Florian Weimer [BZ #21885] diff --git a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps index 81dd1a09ea..053f5ec972 100644 --- a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps +++ b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps @@ -58,7 +58,7 @@ double: 1 float128: 2 idouble: 1 ifloat128: 2 -ildouble: 4 +ildouble: 5 ldouble: 3 Function: "asin": @@ -1154,8 +1154,8 @@ float128: 4 idouble: 3 ifloat: 3 ifloat128: 4 -ildouble: 7 -ldouble: 7 +ildouble: 8 +ldouble: 8 Function: Imaginary part of "clog10_upward": double: 1 @@ -2013,8 +2013,8 @@ double: 3 float: 4 idouble: 3 ifloat: 4 -ildouble: 5 -ldouble: 5 +ildouble: 6 +ldouble: 6 Function: "hypot": double: 1 @@ -2205,8 +2205,8 @@ float128: 8 idouble: 3 ifloat: 4 ifloat128: 8 -ildouble: 5 -ldouble: 5 +ildouble: 6 +ldouble: 6 Function: "log": double: 1 commit a4e5aa1a443cfad09bc98f9bb527995371a53a88 Author: Aurelien Jarno Date: Thu Aug 3 22:35:48 2017 +0000 Fix the return type of the getentropy stub The return type of the getentropy stub is wrongly defined as ssize_t, while both the header and the Linux implementation define it as int. This patch fixes that. Changelog: * stdlib/getentropy.c (getentropy): Change return type to int. (cherry picked from commit 2b34e2716f1e84b2c3457ffc868c3dc775b55845) diff --git a/ChangeLog b/ChangeLog index 3478699b10..4357ad1eb8 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,7 @@ +2017-08-03 Aurelien Jarno + + * stdlib/getentropy.c (getentropy): Change return type to int. + 2017-08-03 Aurelien Jarno * sysdeps/i386/i686/fpu/multiarch/libm-test-ulps: Regenerated. diff --git a/stdlib/getentropy.c b/stdlib/getentropy.c index a71d4cd8f5..a88bbf8de3 100644 --- a/stdlib/getentropy.c +++ b/stdlib/getentropy.c @@ -21,7 +21,7 @@ /* Write LENGTH bytes of randomness starting at BUFFER. Return 0 on success and -1 on failure. */ -ssize_t +int getentropy (void *buffer, size_t length) { __set_errno (ENOSYS); commit 799859f6635d68487ea2472bd79d96a7639a1ab1 Author: H.J. Lu Date: Sun Aug 6 10:44:30 2017 -0700 x86-64: Use _dl_runtime_resolve_opt only with AVX512F [BZ #21871] On AVX machines with XGETBV (ECX == 1) like Skylake processors, (gdb) disass _dl_runtime_resolve_avx_opt Dump of assembler code for function _dl_runtime_resolve_avx_opt: 0x0000000000015890 <+0>: push %rax 0x0000000000015891 <+1>: push %rcx 0x0000000000015892 <+2>: push %rdx 0x0000000000015893 <+3>: mov $0x1,%ecx 0x0000000000015898 <+8>: xgetbv 0x000000000001589b <+11>: mov %eax,%r11d 0x000000000001589e <+14>: pop %rdx 0x000000000001589f <+15>: pop %rcx 0x00000000000158a0 <+16>: pop %rax 0x00000000000158a1 <+17>: and $0x4,%r11d 0x00000000000158a5 <+21>: bnd je 0x16200 <_dl_runtime_resolve_sse_vex> End of assembler dump. is slower than: (gdb) disass _dl_runtime_resolve_avx_slow Dump of assembler code for function _dl_runtime_resolve_avx_slow: 0x0000000000015850 <+0>: vorpd %ymm0,%ymm1,%ymm8 0x0000000000015854 <+4>: vorpd %ymm2,%ymm3,%ymm9 0x0000000000015858 <+8>: vorpd %ymm4,%ymm5,%ymm10 0x000000000001585c <+12>: vorpd %ymm6,%ymm7,%ymm11 0x0000000000015860 <+16>: vorpd %ymm8,%ymm9,%ymm9 0x0000000000015865 <+21>: vorpd %ymm10,%ymm11,%ymm10 0x000000000001586a <+26>: vpcmpeqd %xmm8,%xmm8,%xmm8 0x000000000001586f <+31>: vorpd %ymm9,%ymm10,%ymm10 0x0000000000015874 <+36>: vptest %ymm10,%ymm8 0x0000000000015879 <+41>: bnd jae 0x158b0 <_dl_runtime_resolve_avx> 0x000000000001587c <+44>: vzeroupper 0x000000000001587f <+47>: bnd jmpq 0x16200 <_dl_runtime_resolve_sse_vex> End of assembler dump. (gdb) since xgetbv takes much more cycles than single cycle operations like vpord/vvpcmpeq/ptest. _dl_runtime_resolve_opt should be used only with AVX512 where AVX512 instructions lead to lower CPU frequency on Skylake server. [BZ #21871] * sysdeps/x86/cpu-features.c (init_cpu_features): Set bit_arch_Use_dl_runtime_resolve_opt only with AVX512F. (cherry picked from commit d2cf37c0a2a375cf2fde69f1afbcc49e45368fc4) diff --git a/ChangeLog b/ChangeLog index 4357ad1eb8..764c827161 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-06 H.J. Lu + + [BZ #21871] + * sysdeps/x86/cpu-features.c (init_cpu_features): Set + bit_arch_Use_dl_runtime_resolve_opt only with AVX512F. + 2017-08-03 Aurelien Jarno * stdlib/getentropy.c (getentropy): Change return type to int. diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c index 1d087ea732..6f900840d4 100644 --- a/sysdeps/x86/cpu-features.c +++ b/sysdeps/x86/cpu-features.c @@ -244,10 +244,13 @@ init_cpu_features (struct cpu_features *cpu_features) |= bit_arch_Prefer_No_AVX512; /* To avoid SSE transition penalty, use _dl_runtime_resolve_slow. - If XGETBV suports ECX == 1, use _dl_runtime_resolve_opt. */ + If XGETBV suports ECX == 1, use _dl_runtime_resolve_opt. + Use _dl_runtime_resolve_opt only with AVX512F since it is + slower than _dl_runtime_resolve_slow with AVX. */ cpu_features->feature[index_arch_Use_dl_runtime_resolve_slow] |= bit_arch_Use_dl_runtime_resolve_slow; - if (cpu_features->max_cpuid >= 0xd) + if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable) + && cpu_features->max_cpuid >= 0xd) { unsigned int eax; commit 82efa1ffd43bed1494d20a4b86f6b15ac6bb5545 Author: Adhemerval Zanella Date: Mon Jul 17 10:59:59 2017 -0300 posix: Set p{read,write}v2 to return ENOTSUP (BZ#21780) Different than other architectures hppa-linux-gnu define different values for ENOTSUP and EOPNOTSUPP, where the later is a Linux specific one. This leads to tst-preadwritev{64}v2 tests failures: $ ./testrun.sh misc/tst-preadvwritev2 error: tst-preadvwritev2-common.c:35: preadv2 failure did not set errno to ENOTSUP (223) error: 1 test failures The straightforward fix is to return the POSIX defined ENOTSUP on all p{read,write}v{64}v2 implementations instead of Linux specific one. Checked on x86_64-linux-gnu and the tst-preadwritev{64}v2 on hppa-linux-gnu (although due the installed kernel on my testing system the pwritev{64}v2 with an invalid flag still fails due a known kernel issue [1]). [BZ #21780] * sysdeps/posix/preadv2.c (preadv2): Use ENOTSUP instead of EOPNOTSUPP. * sysdeps/posix/preadv64v2.c (preadv64v2): Likewise. * sysdeps/posix/pwritev2.c (pwritev2): Likewise. * sysdeps/posix/pwritev64v2.c (pwritev64v2): Likewise. * sysdeps/unix/sysv/linux/preadv2.c (preadv2): Likewise. * sysdeps/unix/sysv/linux/preadv64v2.c (preadv64v2): Likewise. * sysdeps/unix/sysv/linux/pwritev2.c (pwritev2): Likewise. * sysdeps/unix/sysv/linux/pwritev64v2.c (pwritev64v2): Likewise. [1] https://sourceware.org/ml/libc-alpha/2017-06/msg00726.html Cherry-pick of 852d63120783fae5bf85a067320dc4ba1ed59f11 diff --git a/ChangeLog b/ChangeLog index 764c827161..459062cc20 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,16 @@ +2017-08-09 Adhemerval Zanella + + [BZ #21780] + * sysdeps/posix/preadv2.c (preadv2): Use ENOTSUP instead of + EOPNOTSUPP. + * sysdeps/posix/preadv64v2.c (preadv64v2): Likewise. + * sysdeps/posix/pwritev2.c (pwritev2): Likewise. + * sysdeps/posix/pwritev64v2.c (pwritev64v2): Likewise. + * sysdeps/unix/sysv/linux/preadv2.c (preadv2): Likewise. + * sysdeps/unix/sysv/linux/preadv64v2.c (preadv64v2): Likewise. + * sysdeps/unix/sysv/linux/pwritev2.c (pwritev2): Likewise. + * sysdeps/unix/sysv/linux/pwritev64v2.c (pwritev64v2): Likewise. + 2017-08-06 H.J. Lu [BZ #21871] diff --git a/sysdeps/posix/preadv2.c b/sysdeps/posix/preadv2.c index 2a7cf11e27..d27f7028ed 100644 --- a/sysdeps/posix/preadv2.c +++ b/sysdeps/posix/preadv2.c @@ -28,7 +28,7 @@ preadv2 (int fd, const struct iovec *vector, int count, OFF_T offset, { if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } diff --git a/sysdeps/posix/preadv64v2.c b/sysdeps/posix/preadv64v2.c index e084f3f9e1..ce7cb40bf2 100644 --- a/sysdeps/posix/preadv64v2.c +++ b/sysdeps/posix/preadv64v2.c @@ -25,7 +25,7 @@ preadv64v2 (int fd, const struct iovec *vector, int count, OFF_T offset, { if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } diff --git a/sysdeps/posix/pwritev2.c b/sysdeps/posix/pwritev2.c index 5b7650c4fc..7ec8cbc407 100644 --- a/sysdeps/posix/pwritev2.c +++ b/sysdeps/posix/pwritev2.c @@ -28,7 +28,7 @@ pwritev2 (int fd, const struct iovec *vector, int count, OFF_T offset, { if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } diff --git a/sysdeps/posix/pwritev64v2.c b/sysdeps/posix/pwritev64v2.c index 0f2f9ef863..be98aeed9d 100644 --- a/sysdeps/posix/pwritev64v2.c +++ b/sysdeps/posix/pwritev64v2.c @@ -26,7 +26,7 @@ pwritev64v2 (int fd, const struct iovec *vector, int count, OFF_T offset, { if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } diff --git a/sysdeps/unix/sysv/linux/preadv2.c b/sysdeps/unix/sysv/linux/preadv2.c index 11fe85eaa8..137e2dd791 100644 --- a/sysdeps/unix/sysv/linux/preadv2.c +++ b/sysdeps/unix/sysv/linux/preadv2.c @@ -32,7 +32,7 @@ preadv2 (int fd, const struct iovec *vector, int count, off_t offset, # ifdef __NR_preadv2 ssize_t result = SYSCALL_CANCEL (preadv2, fd, vector, count, LO_HI_LONG (offset), flags); - if (result >= 0 || errno != ENOSYS) + if (result >= 0) return result; # endif /* Trying to emulate the preadv2 syscall flags is troublesome: @@ -46,7 +46,7 @@ preadv2 (int fd, const struct iovec *vector, int count, off_t offset, if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } return preadv (fd, vector, count, offset); diff --git a/sysdeps/unix/sysv/linux/preadv64v2.c b/sysdeps/unix/sysv/linux/preadv64v2.c index 9d7f8c9893..8f413253f4 100644 --- a/sysdeps/unix/sysv/linux/preadv64v2.c +++ b/sysdeps/unix/sysv/linux/preadv64v2.c @@ -30,7 +30,7 @@ preadv64v2 (int fd, const struct iovec *vector, int count, off64_t offset, #ifdef __NR_preadv64v2 ssize_t result = SYSCALL_CANCEL (preadv64v2, fd, vector, count, LO_HI_LONG (offset), flags); - if (result >= 0 || errno != ENOSYS) + if (result >= 0) return result; #endif /* Trying to emulate the preadv2 syscall flags is troublesome: @@ -44,7 +44,7 @@ preadv64v2 (int fd, const struct iovec *vector, int count, off64_t offset, if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } return preadv64 (fd, vector, count, offset); diff --git a/sysdeps/unix/sysv/linux/pwritev2.c b/sysdeps/unix/sysv/linux/pwritev2.c index 72f0471f96..8e5032fe2f 100644 --- a/sysdeps/unix/sysv/linux/pwritev2.c +++ b/sysdeps/unix/sysv/linux/pwritev2.c @@ -28,7 +28,7 @@ pwritev2 (int fd, const struct iovec *vector, int count, off_t offset, # ifdef __NR_pwritev2 ssize_t result = SYSCALL_CANCEL (pwritev2, fd, vector, count, LO_HI_LONG (offset), flags); - if (result >= 0 || errno != ENOSYS) + if (result >= 0) return result; # endif /* Trying to emulate the pwritev2 syscall flags is troublesome: @@ -42,7 +42,7 @@ pwritev2 (int fd, const struct iovec *vector, int count, off_t offset, if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } return pwritev (fd, vector, count, offset); diff --git a/sysdeps/unix/sysv/linux/pwritev64v2.c b/sysdeps/unix/sysv/linux/pwritev64v2.c index def9a0bc57..d2800c6657 100644 --- a/sysdeps/unix/sysv/linux/pwritev64v2.c +++ b/sysdeps/unix/sysv/linux/pwritev64v2.c @@ -30,7 +30,7 @@ pwritev64v2 (int fd, const struct iovec *vector, int count, off64_t offset, #ifdef __NR_pwritev64v2 ssize_t result = SYSCALL_CANCEL (pwritev64v2, fd, vector, count, LO_HI_LONG (offset), flags); - if (result >= 0 || errno != ENOSYS) + if (result >= 0) return result; #endif /* Trying to emulate the pwritev2 syscall flags is troublesome: @@ -44,7 +44,7 @@ pwritev64v2 (int fd, const struct iovec *vector, int count, off64_t offset, if (flags != 0) { - __set_errno (EOPNOTSUPP); + __set_errno (ENOTSUP); return -1; } return pwritev64 (fd, vector, count, offset); commit 302434688d925134065498b4a5574f6ee6bfb9fd Author: Florian Weimer Date: Thu Aug 10 09:10:36 2017 +0200 nss: Call __resolv_context_put before early return in get*_r [BZ #21932] This corrects an oversight introduced in commit 352f4ff9a268b81ef5d4b2413f582565806e4790 (resolv: Introduce struct resolv_context). (cherry picked from commit 3016149819268b14660f791b971910ccc2cc13e5) diff --git a/ChangeLog b/ChangeLog index 459062cc20..3127648626 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-09 Florian Weimer + + [BZ #21932] + * nss/getXXbyYY_r.c (REENTRANT_NAME): Call __resolv_context_put + before early return. + 2017-08-09 Adhemerval Zanella [BZ #21780] diff --git a/NEWS b/NEWS index 9a64579658..ec6cf34122 100644 --- a/NEWS +++ b/NEWS @@ -10,6 +10,7 @@ Version 2.26.1 The following bugs are resolved with this release: [21885] getaddrinfo: Release resolver context on error in gethosts + [21932] Unpaired __resolv_context_get in generic get*_r implementation Version 2.26 diff --git a/nss/getXXbyYY_r.c b/nss/getXXbyYY_r.c index 6c547ea1ca..bce80e05dd 100644 --- a/nss/getXXbyYY_r.c +++ b/nss/getXXbyYY_r.c @@ -234,6 +234,9 @@ INTERNAL (REENTRANT_NAME) (ADD_PARAMS, LOOKUP_TYPE *resbuf, char *buffer, H_ERRNO_VAR_P)) { case -1: +# ifdef NEED__RES + __resolv_context_put (res_ctx); +# endif return errno; case 1: #ifdef NEED_H_ERRNO @@ -253,7 +256,12 @@ INTERNAL (REENTRANT_NAME) (ADD_PARAMS, LOOKUP_TYPE *resbuf, char *buffer, nscd_status = NSCD_NAME (ADD_VARIABLES, resbuf, buffer, buflen, result H_ERRNO_VAR); if (nscd_status >= 0) - return nscd_status; + { +# ifdef NEED__RES + __resolv_context_put (res_ctx); +# endif + return nscd_status; + } } #endif commit c55ad6452e2d63ebf6fcaabb00bfd27aae02ffb6 Author: Florian Weimer Date: Thu Aug 10 15:58:28 2017 +0200 malloc: Avoid optimizer warning with GCC 7 and -O3 (cherry picked from commit eac43cbb8d808a40004aa0a4a286f5c5155beccb) diff --git a/ChangeLog b/ChangeLog index 3127648626..5ea9b8baa7 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-08-10 Florian Weimer + + * malloc/malloc.c (get_max_fast): Reimplement as an inline + function which calls __builtin_unreachable. + 2017-08-09 Florian Weimer [BZ #21932] diff --git a/malloc/malloc.c b/malloc/malloc.c index 54e406bcb6..e3ff778113 100644 --- a/malloc/malloc.c +++ b/malloc/malloc.c @@ -1658,6 +1658,9 @@ typedef struct malloc_chunk *mfastbinptr; #define arena_is_corrupt(A) (((A)->flags & ARENA_CORRUPTION_BIT)) #define set_arena_corrupt(A) ((A)->flags |= ARENA_CORRUPTION_BIT) +/* Maximum size of memory handled in fastbins. */ +static INTERNAL_SIZE_T global_max_fast; + /* Set value of max_fast. Use impossibly small value if 0. @@ -1668,8 +1671,20 @@ typedef struct malloc_chunk *mfastbinptr; #define set_max_fast(s) \ global_max_fast = (((s) == 0) \ ? SMALLBIN_WIDTH : ((s + SIZE_SZ) & ~MALLOC_ALIGN_MASK)) -#define get_max_fast() global_max_fast +static inline INTERNAL_SIZE_T +get_max_fast (void) +{ + /* Tell the GCC optimizers that global_max_fast is never larger + than MAX_FAST_SIZE. This avoids out-of-bounds array accesses in + _int_malloc after constant propagation of the size parameter. + (The code never executes because malloc preserves the + global_max_fast invariant, but the optimizers may not recognize + this.) */ + if (global_max_fast > MAX_FAST_SIZE) + __builtin_unreachable (); + return global_max_fast; +} /* ----------- Internal state representation and initialization ----------- @@ -1797,9 +1812,6 @@ static struct malloc_par mp_ = #endif }; -/* Maximum size of memory handled in fastbins. */ -static INTERNAL_SIZE_T global_max_fast; - /* Initialize a malloc_state struct. commit 2aa1a7a8f8b9b7879bc6eb1c34d1580f992c406d Author: Florian Weimer Date: Fri Aug 11 15:48:14 2017 +0200 assert: Suppress pedantic warning caused by statement expression [BZ #21242] (cherry picked from commit 8b2c63e4e2ad1fd161f80004ed30624f2a37b57b) diff --git a/ChangeLog b/ChangeLog index 5ea9b8baa7..0057c0902a 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,10 @@ +2017-08-11 Florian Weimer + + [BZ #21242] + * assert/assert.h [__GNUC__ && !__STRICT_ANSI__] (assert): + Suppress pedantic warning resulting from statement expression. + (__ASSERT_FUNCTION): Add missing __extension__. + 2017-08-10 Florian Weimer * malloc/malloc.c (get_max_fast): Reimplement as an inline diff --git a/NEWS b/NEWS index ec6cf34122..d57c4052cf 100644 --- a/NEWS +++ b/NEWS @@ -9,6 +9,7 @@ Version 2.26.1 The following bugs are resolved with this release: + [21242] assert: Suppress pedantic warning caused by statement expression [21885] getaddrinfo: Release resolver context on error in gethosts [21932] Unpaired __resolv_context_get in generic get*_r implementation diff --git a/assert/assert.h b/assert/assert.h index 22f019537c..6801cfeb10 100644 --- a/assert/assert.h +++ b/assert/assert.h @@ -91,13 +91,19 @@ __END_DECLS ? __ASSERT_VOID_CAST (0) \ : __assert_fail (#expr, __FILE__, __LINE__, __ASSERT_FUNCTION)) # else +/* The first occurrence of EXPR is not evaluated due to the sizeof, + but will trigger any pedantic warnings masked by the __extension__ + for the second occurrence. The explicit comparison against zero is + required to support function pointers and bit fields in this + context, and to suppress the evaluation of variable length + arrays. */ # define assert(expr) \ - ({ \ + ((void) sizeof ((expr) == 0), __extension__ ({ \ if (expr) \ ; /* empty */ \ else \ __assert_fail (#expr, __FILE__, __LINE__, __ASSERT_FUNCTION); \ - }) + })) # endif # ifdef __USE_GNU @@ -113,7 +119,7 @@ __END_DECLS C9x has a similar variable called __func__, but prefer the GCC one since it demangles C++ function names. */ # if defined __cplusplus ? __GNUC_PREREQ (2, 6) : __GNUC_PREREQ (2, 4) -# define __ASSERT_FUNCTION __PRETTY_FUNCTION__ +# define __ASSERT_FUNCTION __extension__ __PRETTY_FUNCTION__ # else # if defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L # define __ASSERT_FUNCTION __func__ commit 645b7635ba8fd58062245419e8bb668ab90cd3ec Author: Gabriel F. T. Gomes Date: Mon Aug 7 09:14:14 2017 -0300 powerpc: Restrict xssqrtqp operands to Vector Registers (bug 21941) POWER ISA 3.0 introduces the xssqrtqp instructions, which expects operands to be in Vector Registers (Altivec/VMX), even though this instruction belongs to the Vector-Scalar Instruction Set. In GCC's Extended Assembly for POWER, the 'wq' register constraint is provided for use with IEEE 754 128-bit floating-point values. However, this constraint does not limit the register allocation to Vector Registers (Altivec/VMX) and could assign a Vector-Scalar Register (VSX) to the operands of the instruction. This patch changes the register constraint used in sqrtf128 from 'wq' to 'v', in order to request a Vector Register (Altivec/VMX) for use with the xssqrtqp instruction. Tested for powerpc64le and --with-cpu=power9. [BZ #21941] * sysdeps/powerpc/fpu/math_private.h (__ieee754_sqrtf128): Since xssqrtqp requires operands to be in Vector Registers (Altivec/VMX), replace the register constraint 'wq' with 'v'. * sysdeps/powerpc/powerpc64le/power9/fpu/e_sqrtf128.c (__ieee754_sqrtf128): Likewise. (cherry picked from commit 4d98ace9de3183309cb394cd0110eda5ad2d2531) diff --git a/ChangeLog b/ChangeLog index 0057c0902a..6886cd9361 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,12 @@ +2017-08-10 Gabriel F. T. Gomes + + [BZ #21941] + * sysdeps/powerpc/fpu/math_private.h (__ieee754_sqrtf128): Since + xssqrtqp requires operands to be in Vector Registers + (Altivec/VMX), replace the register constraint 'wq' with 'v'. + * sysdeps/powerpc/powerpc64le/power9/fpu/e_sqrtf128.c + (__ieee754_sqrtf128): Likewise. + 2017-08-11 Florian Weimer [BZ #21242] diff --git a/sysdeps/powerpc/fpu/math_private.h b/sysdeps/powerpc/fpu/math_private.h index d8fd4923ac..396fd0562e 100644 --- a/sysdeps/powerpc/fpu/math_private.h +++ b/sysdeps/powerpc/fpu/math_private.h @@ -30,7 +30,7 @@ extern __always_inline _Float128 __ieee754_sqrtf128 (_Float128 __x) { _Float128 __z; - asm ("xssqrtqp %0,%1" : "=wq" (__z) : "wq" (__x)); + asm ("xssqrtqp %0,%1" : "=v" (__z) : "v" (__x)); return __z; } #endif diff --git a/sysdeps/powerpc/powerpc64le/power9/fpu/e_sqrtf128.c b/sysdeps/powerpc/powerpc64le/power9/fpu/e_sqrtf128.c index 769d3f8922..59fd8269f5 100644 --- a/sysdeps/powerpc/powerpc64le/power9/fpu/e_sqrtf128.c +++ b/sysdeps/powerpc/powerpc64le/power9/fpu/e_sqrtf128.c @@ -30,7 +30,7 @@ __float128 __ieee754_sqrtf128 (__float128 a) { __float128 z; - asm ("xssqrtqp %0,%1" : "=wq" (z) : "wq" (a)); + asm ("xssqrtqp %0,%1" : "=v" (z) : "v" (a)); return z; } strong_alias (__ieee754_sqrtf128, __sqrtf128_finite) commit c2921b17a37e887b8a5ca9d84b875b9ba702b79c Author: Gabriel F. T. Gomes Date: Mon Aug 21 14:23:27 2017 +0200 Do not use __builtin_types_compatible_p in C++ mode (bug 21930) The logic to define isinf for float128 depends on the availability of __builtin_types_compatible_p, which is only available in C mode, however, the conditionals do not check for C or C++ mode. This lead to an error in libstdc++ configure, as reported by bug 21930. This patch adds a conditional for C mode in the definition of isinf for float128. No definition is provided in C++ mode, since libstdc++ headers undefine isinf. Tested for powerpc64le (glibc test suite and libstdc++-v3 configure). [BZ #21930] * math/math.h (isinf): Check if in C or C++ mode before using __builtin_types_compatible_p, since this is a C mode feature. (cherry picked from commit 47a67213a9f51c5f8816d240500b10db605d8b77) diff --git a/ChangeLog b/ChangeLog index 6886cd9361..415fa3cc79 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-18 Gabriel F. T. Gomes + + [BZ #21930] + * math/math.h (isinf): Check if in C or C++ mode before using + __builtin_types_compatible_p, since this is a C mode feature. + 2017-08-10 Gabriel F. T. Gomes [BZ #21941] diff --git a/NEWS b/NEWS index d57c4052cf..75b82c899e 100644 --- a/NEWS +++ b/NEWS @@ -11,6 +11,7 @@ The following bugs are resolved with this release: [21242] assert: Suppress pedantic warning caused by statement expression [21885] getaddrinfo: Release resolver context on error in gethosts + [21930] Do not use __builtin_types_compatible_p in C++ mode [21932] Unpaired __resolv_context_get in generic get*_r implementation Version 2.26 diff --git a/math/math.h b/math/math.h index e21708045a..dea8dbe1ae 100644 --- a/math/math.h +++ b/math/math.h @@ -442,8 +442,12 @@ enum /* Return nonzero value if X is positive or negative infinity. */ # if __HAVE_DISTINCT_FLOAT128 && !__GNUC_PREREQ (7,0) \ - && !defined __SUPPORT_SNAN__ - /* __builtin_isinf_sign is broken for float128 only before GCC 7.0. */ + && !defined __SUPPORT_SNAN__ && !defined __cplusplus + /* Since __builtin_isinf_sign is broken for float128 before GCC 7.0, + use the helper function, __isinff128, with older compilers. This is + only provided for C mode, because in C++ mode, GCC has no support + for __builtin_types_compatible_p (and when in C++ mode, this macro is + not used anyway, because libstdc++ headers undefine it). */ # define isinf(x) \ (__builtin_types_compatible_p (__typeof (x), _Float128) \ ? __isinff128 (x) : __builtin_isinf_sign (x)) commit 5e989c36934d0f0cf13b7a53ef2fa440bce39210 Author: Gabriel F. T. Gomes Date: Mon Aug 14 17:51:51 2017 -0300 Do not use generic selection in C++ mode The logic to protect the use of generic selection (_Generic) does not check for C or C++ mode, however, generic selection is a C-only feature. Tested for powerpc64le. * misc/sys/cdefs.h (__HAVE_GENERIC_SELECTION): Define to 0, if in C++ mode. (cherry picked from commit 6913ad65e00bb32417ad39c41d292b976171e27e) diff --git a/ChangeLog b/ChangeLog index 415fa3cc79..23c00aed09 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-08-18 Gabriel F. T. Gomes + + * misc/sys/cdefs.h (__HAVE_GENERIC_SELECTION): Define to 0, if + in C++ mode. + 2017-08-18 Gabriel F. T. Gomes [BZ #21930] diff --git a/misc/sys/cdefs.h b/misc/sys/cdefs.h index 06523bfe9c..0c808216a4 100644 --- a/misc/sys/cdefs.h +++ b/misc/sys/cdefs.h @@ -464,17 +464,18 @@ # define __glibc_macro_warning(msg) #endif -/* Support for generic selection (ISO C11) is available in GCC since - version 4.9. Previous versions do not provide generic selection, - even though they might set __STDC_VERSION__ to 201112L, when in - -std=c11 mode. Thus, we must check for !defined __GNUC__ when - testing __STDC_VERSION__ for generic selection support. +/* Generic selection (ISO C11) is a C-only feature, available in GCC + since version 4.9. Previous versions do not provide generic + selection, even though they might set __STDC_VERSION__ to 201112L, + when in -std=c11 mode. Thus, we must check for !defined __GNUC__ + when testing __STDC_VERSION__ for generic selection support. On the other hand, Clang also defines __GNUC__, so a clang-specific check is required to enable the use of generic selection. */ -#if __GNUC_PREREQ (4, 9) \ - || __glibc_clang_has_extension (c_generic_selections) \ - || (!defined __GNUC__ && defined __STDC_VERSION__ \ - && __STDC_VERSION__ >= 201112L) +#if !defined __cplusplus \ + && (__GNUC_PREREQ (4, 9) \ + || __glibc_clang_has_extension (c_generic_selections) \ + || (!defined __GNUC__ && defined __STDC_VERSION__ \ + && __STDC_VERSION__ >= 201112L)) # define __HAVE_GENERIC_SELECTION 1 #else # define __HAVE_GENERIC_SELECTION 0 commit fb9a781e9d62c5d7a1f4196915cdfb7c6db59a0c Author: Florian Weimer Date: Mon Aug 21 16:13:49 2017 +0200 assert: Support types without operator== (int) [BZ #21972] (cherry picked from commit b5889d25e9bf944a89fdd7bcabf3b6c6f6bb6f7c) diff --git a/ChangeLog b/ChangeLog index 23c00aed09..8bc3ad9a46 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,14 @@ +2017-08-21 Florian Weimer + + [BZ #21972] + * assert/assert.h (assert): Use static_cast (bool) for C++. + Use the ternary operator in the warning branch for GNU C. + * assert/Makefile (tests): Add tst-assert-c++, tst-assert-g++. + (CFLAGS-tst-assert-c++.o): Compile in C++11 mode. + (CFLAGS-tst-assert-g++.o): Compile in GnU C++11 mode. + (LDLIBS-tst-assert-c++, LDLIBS-tst-assert-g++): Link with libstdc++. + * assert/tst-assert-c++.cc, assert/tst-assert-g++.cc: New files. + 2017-08-18 Gabriel F. T. Gomes * misc/sys/cdefs.h (__HAVE_GENERIC_SELECTION): Define to 0, if diff --git a/NEWS b/NEWS index 75b82c899e..1996e5fbef 100644 --- a/NEWS +++ b/NEWS @@ -13,6 +13,7 @@ The following bugs are resolved with this release: [21885] getaddrinfo: Release resolver context on error in gethosts [21930] Do not use __builtin_types_compatible_p in C++ mode [21932] Unpaired __resolv_context_get in generic get*_r implementation + [21972] assert macro requires operator== (int) for its argument type Version 2.26 diff --git a/assert/Makefile b/assert/Makefile index 1c3be9b01f..9ec1be81a9 100644 --- a/assert/Makefile +++ b/assert/Makefile @@ -25,6 +25,15 @@ include ../Makeconfig headers := assert.h routines := assert assert-perr __assert -tests := test-assert test-assert-perr +tests := test-assert test-assert-perr tst-assert-c++ tst-assert-g++ include ../Rules + +ifeq ($(have-cxx-thread_local),yes) +CFLAGS-tst-assert-c++.o = -std=c++11 +LDLIBS-tst-assert-c++ = -lstdc++ +CFLAGS-tst-assert-g++.o = -std=gnu++11 +LDLIBS-tst-assert-g++ = -lstdc++ +else +tests-unsupported += tst-assert-c++ tst-assert-g++ +endif diff --git a/assert/assert.h b/assert/assert.h index 6801cfeb10..640c95c063 100644 --- a/assert/assert.h +++ b/assert/assert.h @@ -85,7 +85,12 @@ __END_DECLS /* When possible, define assert so that it does not add extra parentheses around EXPR. Otherwise, those added parentheses would suppress warnings we'd expect to be detected by gcc's -Wparentheses. */ -# if !defined __GNUC__ || defined __STRICT_ANSI__ +# if defined __cplusplus +# define assert(expr) \ + (static_cast (expr) \ + ? void (0) \ + : __assert_fail (#expr, __FILE__, __LINE__, __ASSERT_FUNCTION)) +# elif !defined __GNUC__ || defined __STRICT_ANSI__ # define assert(expr) \ ((expr) \ ? __ASSERT_VOID_CAST (0) \ @@ -93,12 +98,11 @@ __END_DECLS # else /* The first occurrence of EXPR is not evaluated due to the sizeof, but will trigger any pedantic warnings masked by the __extension__ - for the second occurrence. The explicit comparison against zero is - required to support function pointers and bit fields in this - context, and to suppress the evaluation of variable length - arrays. */ + for the second occurrence. The ternary operator is required to + support function pointers and bit fields in this context, and to + suppress the evaluation of variable length arrays. */ # define assert(expr) \ - ((void) sizeof ((expr) == 0), __extension__ ({ \ + ((void) sizeof ((expr) ? 1 : 0), __extension__ ({ \ if (expr) \ ; /* empty */ \ else \ diff --git a/assert/tst-assert-c++.cc b/assert/tst-assert-c++.cc new file mode 100644 index 0000000000..12a5e690cb --- /dev/null +++ b/assert/tst-assert-c++.cc @@ -0,0 +1,78 @@ +/* Tests for interactions between C++ and assert. + Copyright (C) 2017 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 + . */ + +#include + +/* The C++ standard requires that if the assert argument is a constant + subexpression, then the assert itself is one, too. */ +constexpr int +check_constexpr () +{ + return (assert (true), 1); +} + +/* Objects of this class can be contextually converted to bool, but + cannot be compared to int. */ +struct no_int +{ + no_int () = default; + no_int (const no_int &) = delete; + + explicit operator bool () const + { + return true; + } + + bool operator! () const; /* No definition. */ + template bool operator== (T) const; /* No definition. */ + template bool operator!= (T) const; /* No definition. */ +}; + +/* This class tests that operator== is not used by assert. */ +struct bool_and_int +{ + bool_and_int () = default; + bool_and_int (const no_int &) = delete; + + explicit operator bool () const + { + return true; + } + + bool operator! () const; /* No definition. */ + template bool operator== (T) const; /* No definition. */ + template bool operator!= (T) const; /* No definition. */ +}; + +static int +do_test () +{ + { + no_int value; + assert (value); + } + + { + bool_and_int value; + assert (value); + } + + return 0; +} + +#include diff --git a/assert/tst-assert-g++.cc b/assert/tst-assert-g++.cc new file mode 100644 index 0000000000..8c06402825 --- /dev/null +++ b/assert/tst-assert-g++.cc @@ -0,0 +1,19 @@ +/* Tests for interactions between C++ and assert. GNU C++11 version. + Copyright (C) 2017 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 + . */ + +#include commit 3aeab55ee17ca527e4597bc8397c0434c3f8b34e Author: Gabriel F. T. Gomes Date: Mon Aug 21 15:45:57 2017 -0300 Add missing bug fixes to NEWS diff --git a/NEWS b/NEWS index 1996e5fbef..0534c5296e 100644 --- a/NEWS +++ b/NEWS @@ -10,9 +10,12 @@ Version 2.26.1 The following bugs are resolved with this release: [21242] assert: Suppress pedantic warning caused by statement expression + [21780] posix: Set p{read,write}v2 to return ENOTSUP + [21871] x86-64: Use _dl_runtime_resolve_opt only with AVX512F [21885] getaddrinfo: Release resolver context on error in gethosts [21930] Do not use __builtin_types_compatible_p in C++ mode [21932] Unpaired __resolv_context_get in generic get*_r implementation + [21941] powerpc: Restrict xssqrtqp operands to Vector Registers [21972] assert macro requires operator== (int) for its argument type Version 2.26 commit 77db8772bd3f6f2bbad697dcf46861ce310f5b95 Author: Florian Weimer Date: Thu Aug 10 16:06:52 2017 +0200 __inet6_scopeid_pton: Remove attribute_hidden, internal_function The hidden attribute was overridden by libc_hidden_proto on GNU/Linux. It is incorrect because the function is used from nscd. internal_function is not supposed to be used across DSO boundaries, so this commit removes it (again, due to the use in nscd). (cherry picked from commit f87cc2bfba9b844da48a63441c6099342b1551c7) diff --git a/ChangeLog b/ChangeLog index 8bc3ad9a46..ad05da8ade 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,10 @@ +2017-08-10 Florian Weimer + + * inet/net-internal.h (__inet6_scopeid_pton): Remove + attribute_hidden, internal_function. + * inet/inet6_scopeid_pton.c (__inet6_scopeid_pton): Remove + internal_function. + 2017-08-21 Florian Weimer [BZ #21972] diff --git a/inet/inet6_scopeid_pton.c b/inet/inet6_scopeid_pton.c index e09b1cb34d..cc8803fa10 100644 --- a/inet/inet6_scopeid_pton.c +++ b/inet/inet6_scopeid_pton.c @@ -28,7 +28,7 @@ /* Parse SOURCE as a scope ID for ADDRESS. Return 0 on success and -1 on error. */ -internal_function int +int __inet6_scopeid_pton (const struct in6_addr *address, const char *scope, uint32_t *result) { diff --git a/inet/net-internal.h b/inet/net-internal.h index 2b2632c7ba..b2135893e8 100644 --- a/inet/net-internal.h +++ b/inet/net-internal.h @@ -25,8 +25,7 @@ #include int __inet6_scopeid_pton (const struct in6_addr *address, - const char *scope, uint32_t *result) - internal_function attribute_hidden; + const char *scope, uint32_t *result); libc_hidden_proto (__inet6_scopeid_pton) commit 6043d77a47de297b62084c1c261cdada082bf09c Author: Andreas Schwab Date: Mon Aug 28 19:49:18 2017 +0200 ldd: never run file directly (cherry picked from commit eedca9772e99c72ab4c3c34e43cc764250aa3e3c) diff --git a/ChangeLog b/ChangeLog index ad05da8ade..fa27c6f66f 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-16 Andreas Schwab + + [BZ #16750] + CVE-2009-5064 + * elf/ldd.bash.in: Never run file directly. + 2017-08-10 Florian Weimer * inet/net-internal.h (__inet6_scopeid_pton): Remove diff --git a/NEWS b/NEWS index 0534c5296e..756e849643 100644 --- a/NEWS +++ b/NEWS @@ -7,8 +7,17 @@ using `glibc' in the "product" field. Version 2.26.1 +Security related changes: + + CVE-2009-5064: The ldd script would sometimes run the program under + examination directly, without preventing code execution through the + dynamic linker. (The glibc project disputes that this is a security + vulnerability; only trusted binaries must be examined using the ldd + script.) + The following bugs are resolved with this release: + [16750] ldd: Never run file directly. [21242] assert: Suppress pedantic warning caused by statement expression [21780] posix: Set p{read,write}v2 to return ENOTSUP [21871] x86-64: Use _dl_runtime_resolve_opt only with AVX512F diff --git a/elf/ldd.bash.in b/elf/ldd.bash.in index 7dd1fccf24..686785e235 100644 --- a/elf/ldd.bash.in +++ b/elf/ldd.bash.in @@ -164,18 +164,6 @@ warning: you do not have execution permission for" "\`$file'" >&2 fi done case $ret in - 0) - # If the program exits with exit code 5, it means the process has been - # invoked with __libc_enable_secure. Fall back to running it through - # the dynamic linker. - try_trace "$file" - rc=$? - if [ $rc = 5 ]; then - try_trace "$RTLD" "$file" - rc=$? - fi - [ $rc = 0 ] || result=1 - ;; 1) # This can be a non-ELF binary or no binary at all. nonelf "$file" || { @@ -183,7 +171,7 @@ warning: you do not have execution permission for" "\`$file'" >&2 result=1 } ;; - 2) + 0|2) try_trace "$RTLD" "$file" || result=1 ;; *) commit ef8566d72af5e03c1b82cf02efb794268a347f8c Author: Gabriel F. T. Gomes Date: Mon Aug 14 13:46:15 2017 -0300 Provide a C++ version of issignaling that does not use __MATH_TG The macro __MATH_TG contains the logic to select between long double and _Float128, when these types are ABI-distinct. This logic relies on __builtin_types_compatible_p, which is not available in C++ mode. On the other hand, C++ function overloading provides the means to distinguish between the floating-point types. The overloading resolution will match the correct parameter regardless of type qualifiers, i.e.: const and volatile. Tested for powerpc64le, s390x, and x86_64. * math/math.h [defined __cplusplus] (issignaling): Provide a C++ definition for issignaling that does not rely on __MATH_TG, since __MATH_TG uses __builtin_types_compatible_p, which is only available in C mode. (CFLAGS-test-math-issignaling.cc): New variable. * math/Makefile [CXX] (tests): Add test-math-issignaling. * math/test-math-issignaling.cc: New test for C++ implementation of type-generic issignaling. * sysdeps/powerpc/powerpc64le/Makefile [subdir == math] (CXXFLAGS-test-math-issignaling.cc): Add -mfloat128 to the build options of test-math-issignaling on powerpc64le. (cherry picked from commit a16e8bc08edca84d507715c66d6cddbbc7ed3b62) diff --git a/ChangeLog b/ChangeLog index fa27c6f66f..527026ba5d 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,17 @@ +2017-08-22 Gabriel F. T. Gomes + + * math/math.h [defined __cplusplus] (issignaling): Provide a C++ + definition for issignaling that does not rely on __MATH_TG, + since __MATH_TG uses __builtin_types_compatible_p, which is only + available in C mode. + (CFLAGS-test-math-issignaling.cc): New variable. + * math/Makefile [CXX] (tests): Add test-math-issignaling. + * math/test-math-issignaling.cc: New test for C++ implementation + of type-generic issignaling. + * sysdeps/powerpc/powerpc64le/Makefile [subdir == math] + (CXXFLAGS-test-math-issignaling.cc): Add -mfloat128 to the build + options of test-math-issignaling on powerpc64le. + 2017-08-16 Andreas Schwab [BZ #16750] diff --git a/math/Makefile b/math/Makefile index e09b0c0545..0130fcf38b 100644 --- a/math/Makefile +++ b/math/Makefile @@ -203,7 +203,7 @@ tests-static = test-fpucw-static test-fpucw-ieee-static \ test-signgam-ullong-static test-signgam-ullong-init-static ifneq (,$(CXX)) -tests += test-math-isinff test-math-iszero +tests += test-math-isinff test-math-iszero test-math-issignaling endif ifneq (no,$(PERL)) @@ -350,6 +350,7 @@ CFLAGS-test-signgam-ullong-init-static.c = -std=c99 CFLAGS-test-math-isinff.cc = -std=gnu++11 CFLAGS-test-math-iszero.cc = -std=gnu++11 +CFLAGS-test-math-issignaling.cc = -std=gnu++11 CFLAGS-test-iszero-excess-precision.c = -fexcess-precision=standard CFLAGS-test-iseqsig-excess-precision.c = -fexcess-precision=standard diff --git a/math/math.h b/math/math.h index dea8dbe1ae..add86af724 100644 --- a/math/math.h +++ b/math/math.h @@ -474,7 +474,24 @@ enum # include /* Return nonzero value if X is a signaling NaN. */ -# define issignaling(x) __MATH_TG ((x), __issignaling, (x)) +# ifndef __cplusplus +# define issignaling(x) __MATH_TG ((x), __issignaling, (x)) +# else + /* In C++ mode, __MATH_TG cannot be used, because it relies on + __builtin_types_compatible_p, which is a C-only builtin. On the + other hand, overloading provides the means to distinguish between + the floating-point types. The overloading resolution will match + the correct parameter (regardless of type qualifiers (i.e.: const + and volatile). */ +extern "C++" { +inline int issignaling (float __val) { return __issignalingf (__val); } +inline int issignaling (double __val) { return __issignaling (__val); } +inline int issignaling (long double __val) { return __issignalingl (__val); } +# if __HAVE_DISTINCT_FLOAT128 +inline int issignaling (_Float128 __val) { return __issignalingf128 (__val); } +# endif +} /* extern C++ */ +# endif /* Return nonzero value if X is subnormal. */ # define issubnormal(x) (fpclassify (x) == FP_SUBNORMAL) diff --git a/math/test-math-issignaling.cc b/math/test-math-issignaling.cc new file mode 100644 index 0000000000..22ae9e1bca --- /dev/null +++ b/math/test-math-issignaling.cc @@ -0,0 +1,113 @@ +/* Test for the C++ implementation of issignaling. + Copyright (C) 2017 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 + . */ + +#define _GNU_SOURCE 1 +#include +#include + +#include + +/* There is no signaling_NaN for _Float128 in std::numeric_limits. + Include ieee754_float128.h and use the bitfields in the union + ieee854_float128.ieee_nan to build a signaling NaN. */ +#if __HAVE_DISTINCT_FLOAT128 +# include +#endif + +static bool errors; + +static void +check (int actual, int expected, const char *actual_expr, int line) +{ + if (actual != expected) + { + errors = true; + printf ("%s:%d: error: %s\n", __FILE__, line, actual_expr); + printf ("%s:%d: expected: %d\n", __FILE__, line, expected); + printf ("%s:%d: actual: %d\n", __FILE__, line, actual); + } +} + +#define CHECK(actual, expected) \ + check ((actual), (expected), #actual, __LINE__) + +template +static void +check_type () +{ + typedef std::numeric_limits limits; + CHECK (issignaling (T{0}), 0); + if (limits::has_infinity) + { + CHECK (issignaling (limits::infinity ()), 0); + CHECK (issignaling (-limits::infinity ()), 0); + } + if (limits::has_quiet_NaN) + CHECK (issignaling (limits::quiet_NaN ()), 0); + if (limits::has_signaling_NaN) + CHECK (issignaling (limits::signaling_NaN ()), 1); +} + +#if __HAVE_DISTINCT_FLOAT128 +static void +check_float128 () +{ + ieee854_float128 q; + + q.d = 0; + CHECK (issignaling (q.d), 0); + + /* Infinity. */ + q.ieee.negative = 0; + q.ieee.exponent = 0x7FFF; + q.ieee.mantissa0 = 0x0000; + q.ieee.mantissa1 = 0x00000000; + q.ieee.mantissa2 = 0x00000000; + q.ieee.mantissa3 = 0x00000000; + CHECK (issignaling (q.d), 0); + + /* Quiet NaN. */ + q.ieee_nan.quiet_nan = 1; + q.ieee_nan.mantissa0 = 0x0000; + CHECK (issignaling (q.d), 0); + + /* Still a quiet NaN. */ + q.ieee_nan.quiet_nan = 1; + q.ieee_nan.mantissa0 = 0x4000; + CHECK (issignaling (q.d), 0); + + /* Signaling NaN. */ + q.ieee_nan.quiet_nan = 0; + q.ieee_nan.mantissa0 = 0x4000; + CHECK (issignaling (q.d), 1); +} +#endif + +static int +do_test (void) +{ + check_type (); + check_type (); + check_type (); +#if __HAVE_DISTINCT_FLOAT128 + check_float128 (); +#endif + return errors; +} + +#include diff --git a/sysdeps/powerpc/powerpc64le/Makefile b/sysdeps/powerpc/powerpc64le/Makefile index 77617b670a..19adbfa1c1 100644 --- a/sysdeps/powerpc/powerpc64le/Makefile +++ b/sysdeps/powerpc/powerpc64le/Makefile @@ -16,6 +16,7 @@ $(foreach suf,$(all-object-suffixes),%f128_r$(suf)): CFLAGS += -mfloat128 $(foreach suf,$(all-object-suffixes),$(objpfx)test-float128%$(suf)): CFLAGS += -mfloat128 $(foreach suf,$(all-object-suffixes),$(objpfx)test-ifloat128%$(suf)): CFLAGS += -mfloat128 CFLAGS-libm-test-support-float128.c += -mfloat128 +CFLAGS-test-math-issignaling.cc += -mfloat128 $(objpfx)test-float128% $(objpfx)test-ifloat128%: \ gnulib-tests += $(f128-loader-link) endif commit 35dded99a89db873b06270ca7f21245a0faf712a Author: Gabriel F. T. Gomes Date: Wed Aug 23 10:16:54 2017 -0300 Fix the C++ version of issignaling when __NO_LONG_DOUBLE_MATH is defined When __NO_LONG_DOUBLE_MATH is defined, __issignalingl is not available, thus issignaling with long double argument should call __issignaling, instead. Tested for powerpc64le. * math/math.h [defined __cplusplus] (issignaling): In the long double case, call __issignalingl only if __NO_LONG_DOUBLE_MATH is not defined. Call __issignaling, otherwise. (cherry picked from commit 3d7b66f66cb223e899a7ebc0f4c20f13e711c9e0) diff --git a/ChangeLog b/ChangeLog index 527026ba5d..04b56b555c 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-24 Gabriel F. T. Gomes + + * math/math.h [defined __cplusplus] (issignaling): In the long + double case, call __issignalingl only if __NO_LONG_DOUBLE_MATH + is not defined. Call __issignaling, otherwise. + 2017-08-22 Gabriel F. T. Gomes * math/math.h [defined __cplusplus] (issignaling): Provide a C++ diff --git a/math/math.h b/math/math.h index add86af724..60dfa31592 100644 --- a/math/math.h +++ b/math/math.h @@ -486,7 +486,15 @@ enum extern "C++" { inline int issignaling (float __val) { return __issignalingf (__val); } inline int issignaling (double __val) { return __issignaling (__val); } -inline int issignaling (long double __val) { return __issignalingl (__val); } +inline int +issignaling (long double __val) +{ +# ifdef __NO_LONG_DOUBLE_MATH + return __issignaling (__val); +# else + return __issignalingl (__val); +# endif +} # if __HAVE_DISTINCT_FLOAT128 inline int issignaling (_Float128 __val) { return __issignalingf128 (__val); } # endif commit 58270c0049404ef2f878fdd45df55f17f0b8c1f7 Author: Gabriel F. T. Gomes Date: Tue Aug 22 16:34:42 2017 -0300 Provide a C++ version of iszero that does not use __MATH_TG (bug 21930) When signaling nans are enabled (with -fsignaling-nans), the C++ version of iszero uses the fpclassify macro, which is defined with __MATH_TG. However, when support for float128 is available, __MATH_TG uses the builtin __builtin_types_compatible_p, which is only available in C mode. This patch refactors the C++ version of iszero so that it uses function overloading to select between the floating-point types, instead of relying on fpclassify and __MATH_TG. Tested for powerpc64le, s390x, x86_64, and with build-many-glibcs.py. [BZ #21930] * math/math.h [defined __cplusplus && defined __SUPPORT_SNAN__] (iszero): New C++ implementation that does not use fpclassify/__MATH_TG/__builtin_types_compatible_p, when signaling nans are enabled, since __builtin_types_compatible_p is a C-only feature. * math/test-math-iszero.cc: When __HAVE_DISTINCT_FLOAT128 is defined, include ieee754_float128.h for access to the union and member ieee854_float128.ieee. [__HAVE_DISTINCT_FLOAT128] (do_test): Call check_float128. [__HAVE_DISTINCT_FLOAT128] (check_float128): New function. * sysdeps/powerpc/powerpc64le/Makefile [subdir == math] (CXXFLAGS-test-math-iszero.cc): Add -mfloat128 to the build options of test-math-zero on powerpc64le. (cherry picked from commit 42496114ec0eb7d6d039d05d4262e109951c600c) diff --git a/ChangeLog b/ChangeLog index 04b56b555c..23ded7f03d 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,20 @@ +2017-08-28 Gabriel F. T. Gomes + + [BZ #21930] + * math/math.h [defined __cplusplus && defined __SUPPORT_SNAN__] + (iszero): New C++ implementation that does not use + fpclassify/__MATH_TG/__builtin_types_compatible_p, when + signaling nans are enabled, since __builtin_types_compatible_p + is a C-only feature. + * math/test-math-iszero.cc: When __HAVE_DISTINCT_FLOAT128 is + defined, include ieee754_float128.h for access to the union and + member ieee854_float128.ieee. + [__HAVE_DISTINCT_FLOAT128] (do_test): Call check_float128. + [__HAVE_DISTINCT_FLOAT128] (check_float128): New function. + * sysdeps/powerpc/powerpc64le/Makefile [subdir == math] + (CXXFLAGS-test-math-iszero.cc): Add -mfloat128 to the build + options of test-math-zero on powerpc64le. + 2017-08-24 Gabriel F. T. Gomes * math/math.h [defined __cplusplus] (issignaling): In the long diff --git a/math/math.h b/math/math.h index 60dfa31592..7c0fc6dbb3 100644 --- a/math/math.h +++ b/math/math.h @@ -513,15 +513,40 @@ inline int issignaling (_Float128 __val) { return __issignalingf128 (__val); } # endif # else /* __cplusplus */ extern "C++" { +# ifdef __SUPPORT_SNAN__ +inline int +iszero (float __val) +{ + return __fpclassifyf (__val) == FP_ZERO; +} +inline int +iszero (double __val) +{ + return __fpclassify (__val) == FP_ZERO; +} +inline int +iszero (long double __val) +{ +# ifdef __NO_LONG_DOUBLE_MATH + return __fpclassify (__val) == FP_ZERO; +# else + return __fpclassifyl (__val) == FP_ZERO; +# endif +} +# if __HAVE_DISTINCT_FLOAT128 +inline int +iszero (_Float128 __val) +{ + return __fpclassifyf128 (__val) == FP_ZERO; +} +# endif +# else template inline bool iszero (__T __val) { -# ifdef __SUPPORT_SNAN__ - return fpclassify (__val) == FP_ZERO; -# else return __val == 0; -# endif } +# endif } /* extern C++ */ # endif /* __cplusplus */ #endif /* Use IEC_60559_BFP_EXT. */ diff --git a/math/test-math-iszero.cc b/math/test-math-iszero.cc index 027e972654..5c07261626 100644 --- a/math/test-math-iszero.cc +++ b/math/test-math-iszero.cc @@ -22,6 +22,13 @@ #include +/* Support for _Float128 in std::numeric_limits is limited. + Include ieee754_float128.h and use the bitfields in the union + ieee854_float128.ieee_nan to build corner-case inputs. */ +#if __HAVE_DISTINCT_FLOAT128 +# include +#endif + static bool errors; static void @@ -72,12 +79,84 @@ check_type () std::numeric_limits::has_denorm == std::denorm_absent); } +#if __HAVE_DISTINCT_FLOAT128 +static void +check_float128 () +{ + ieee854_float128 q; + + q.d = 0.0Q; + CHECK (iszero (q.d), 1); + q.d = -0.0Q; + CHECK (iszero (q.d), 1); + q.d = 1.0Q; + CHECK (iszero (q.d), 0); + q.d = -1.0Q; + CHECK (iszero (q.d), 0); + + /* Normal min. */ + q.ieee.negative = 0; + q.ieee.exponent = 0x0001; + q.ieee.mantissa0 = 0x0000; + q.ieee.mantissa1 = 0x00000000; + q.ieee.mantissa2 = 0x00000000; + q.ieee.mantissa3 = 0x00000000; + CHECK (iszero (q.d), 0); + q.ieee.negative = 1; + CHECK (iszero (q.d), 0); + + /* Normal max. */ + q.ieee.negative = 0; + q.ieee.exponent = 0x7FFE; + q.ieee.mantissa0 = 0xFFFF; + q.ieee.mantissa1 = 0xFFFFFFFF; + q.ieee.mantissa2 = 0xFFFFFFFF; + q.ieee.mantissa3 = 0xFFFFFFFF; + CHECK (iszero (q.d), 0); + q.ieee.negative = 1; + CHECK (iszero (q.d), 0); + + /* Infinity. */ + q.ieee.negative = 0; + q.ieee.exponent = 0x7FFF; + q.ieee.mantissa0 = 0x0000; + q.ieee.mantissa1 = 0x00000000; + q.ieee.mantissa2 = 0x00000000; + q.ieee.mantissa3 = 0x00000000; + CHECK (iszero (q.d), 0); + + /* Quiet NaN. */ + q.ieee_nan.quiet_nan = 1; + q.ieee_nan.mantissa0 = 0x0000; + CHECK (iszero (q.d), 0); + + /* Signaling NaN. */ + q.ieee_nan.quiet_nan = 0; + q.ieee_nan.mantissa0 = 0x4000; + CHECK (iszero (q.d), 0); + + /* Denormal min. */ + q.ieee.negative = 0; + q.ieee.exponent = 0x0000; + q.ieee.mantissa0 = 0x0000; + q.ieee.mantissa1 = 0x00000000; + q.ieee.mantissa2 = 0x00000000; + q.ieee.mantissa3 = 0x00000001; + CHECK (iszero (q.d), 0); + q.ieee.negative = 1; + CHECK (iszero (q.d), 0); +} +#endif + static int do_test (void) { check_type (); check_type (); check_type (); +#if __HAVE_DISTINCT_FLOAT128 + check_float128 (); +#endif return errors; } diff --git a/sysdeps/powerpc/powerpc64le/Makefile b/sysdeps/powerpc/powerpc64le/Makefile index 19adbfa1c1..dea2290736 100644 --- a/sysdeps/powerpc/powerpc64le/Makefile +++ b/sysdeps/powerpc/powerpc64le/Makefile @@ -17,7 +17,8 @@ $(foreach suf,$(all-object-suffixes),$(objpfx)test-float128%$(suf)): CFLAGS += - $(foreach suf,$(all-object-suffixes),$(objpfx)test-ifloat128%$(suf)): CFLAGS += -mfloat128 CFLAGS-libm-test-support-float128.c += -mfloat128 CFLAGS-test-math-issignaling.cc += -mfloat128 -$(objpfx)test-float128% $(objpfx)test-ifloat128%: \ +CFLAGS-test-math-iszero.cc += -mfloat128 +$(objpfx)test-float128% $(objpfx)test-ifloat128% $(objpfx)test-math-iszero: \ gnulib-tests += $(f128-loader-link) endif commit 947e2e0a9410c18eb44144456c7fe8e7c0e2a999 Author: Joseph Myers Date: Tue Aug 22 00:30:51 2017 +0000 Fix position of tests-unsupported definition in assert/Makefile. tests-unsupported has to be defined before the inclusion of Rules in a subdirectory Makefile; otherwise it is ineffective. This patch fixes the ordering in assert/Makefile, where a recent test addition put tests-unsupported too late (resulting in build failures when the C++ compiler was missing or broken, and thereby showing up the unrelated bug 21987). Incidentally, I don't see why these tests depend on $(have-cxx-thread_local) rather than just a working C++ compiler. Tested in such a configuration (broken compiler/libstdc++) with build-many-glibcs.py. * assert/Makefile [$(have-cxx-thread_local)]: Move conditional variable definitions above inclusion of ../Rules. (cherry picked from commit 75dfe623df945db7dd3c12a206d743c45c16b5ed) diff --git a/ChangeLog b/ChangeLog index 23ded7f03d..87fa54c57b 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-08-22 Joseph Myers + + * assert/Makefile [$(have-cxx-thread_local)]: Move conditional + variable definitions above inclusion of ../Rules. + 2017-08-28 Gabriel F. T. Gomes [BZ #21930] diff --git a/assert/Makefile b/assert/Makefile index 9ec1be81a9..222ab516f0 100644 --- a/assert/Makefile +++ b/assert/Makefile @@ -27,8 +27,6 @@ headers := assert.h routines := assert assert-perr __assert tests := test-assert test-assert-perr tst-assert-c++ tst-assert-g++ -include ../Rules - ifeq ($(have-cxx-thread_local),yes) CFLAGS-tst-assert-c++.o = -std=c++11 LDLIBS-tst-assert-c++ = -lstdc++ @@ -37,3 +35,5 @@ LDLIBS-tst-assert-g++ = -lstdc++ else tests-unsupported += tst-assert-c++ tst-assert-g++ endif + +include ../Rules commit 4fdd75e4463801b9d8f329769df1a26145560656 Author: Florian Weimer Date: Tue Aug 8 18:48:05 2017 +0200 getaddrinfo: Remove unreachable return statement from gaih_inet (cherry picked from commit 0df595b23a829c9169ec418a19eef9006b4ae801) diff --git a/ChangeLog b/ChangeLog index 87fa54c57b..591c753fcc 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-08-08 Florian Weimer + + * sysdeps/posix/getaddrinfo.c (gaih_inet): Remove unreachable + return statement. + 2017-08-22 Joseph Myers * assert/Makefile [$(have-cxx-thread_local)]: Move conditional diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index 699411cc92..09f85fc472 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -420,13 +420,9 @@ gaih_inet (const char *name, const struct gaih_service *service, alloca_account (sizeof (struct gaih_servtuple), alloca_used); - if ((rc = gaih_inet_serv (service->name, - tp, req, newp, tmpbuf))) - { - if (rc) - continue; - return rc; - } + if (gaih_inet_serv (service->name, + tp, req, newp, tmpbuf) != 0) + continue; *pst = newp; pst = &(newp->next); commit 5253749232749adb535d9b2bf7d43173b191ebef Author: Florian Weimer Date: Mon Sep 4 11:44:10 2017 +0200 Synchronize support/ infrastructure with master This commit updates the support/ subdirectory to commit 65329bd233db9d1b8b94e90734a564705b619260 on the master branch. diff --git a/support/namespace.h b/support/namespace.h index 859c2fda3f..9eddb1a0e9 100644 --- a/support/namespace.h +++ b/support/namespace.h @@ -66,7 +66,9 @@ struct support_chroot_configuration { /* File contents. The files are not created if the field is NULL. */ - const char *resolv_conf; + const char *resolv_conf; /* /etc/resolv.conf. */ + const char *hosts; /* /etc/hosts. */ + const char *host_conf; /* /etc/host.conf. */ }; /* The result of the creation of a chroot. */ @@ -78,8 +80,11 @@ struct support_chroot /* Path to the chroot directory. */ char *path_chroot; - /* Path to the /etc/resolv.conf file. */ - char *path_resolv_conf; + /* Paths to files in the chroot. These are absolute and outside of + the chroot. */ + char *path_resolv_conf; /* /etc/resolv.conf. */ + char *path_hosts; /* /etc/hosts. */ + char *path_host_conf; /* /etc/host.conf. */ }; /* Create a chroot environment. The returned data should be freed diff --git a/support/support_chroot.c b/support/support_chroot.c index c0807b313a..f3ef551b05 100644 --- a/support/support_chroot.c +++ b/support/support_chroot.c @@ -24,6 +24,23 @@ #include #include +/* If CONTENTS is not NULL, write it to the file at DIRECTORY/RELPATH, + and store the name in *ABSPATH. If CONTENTS is NULL, store NULL in + *ABSPATH. */ +static void +write_file (const char *directory, const char *relpath, const char *contents, + char **abspath) +{ + if (contents != NULL) + { + *abspath = xasprintf ("%s/%s", directory, relpath); + add_temp_file (*abspath); + support_write_file_string (*abspath, contents); + } + else + *abspath = NULL; +} + struct support_chroot * support_chroot_create (struct support_chroot_configuration conf) { @@ -39,15 +56,10 @@ support_chroot_create (struct support_chroot_configuration conf) xmkdir (path_etc, 0777); add_temp_file (path_etc); - if (conf.resolv_conf != NULL) - { - /* Create an empty resolv.conf file. */ - chroot->path_resolv_conf = xasprintf ("%s/resolv.conf", path_etc); - add_temp_file (chroot->path_resolv_conf); - support_write_file_string (chroot->path_resolv_conf, conf.resolv_conf); - } - else - chroot->path_resolv_conf = NULL; + write_file (path_etc, "resolv.conf", conf.resolv_conf, + &chroot->path_resolv_conf); + write_file (path_etc, "hosts", conf.hosts, &chroot->path_hosts); + write_file (path_etc, "host.conf", conf.host_conf, &chroot->path_host_conf); free (path_etc); @@ -67,5 +79,7 @@ support_chroot_free (struct support_chroot *chroot) { free (chroot->path_chroot); free (chroot->path_resolv_conf); + free (chroot->path_hosts); + free (chroot->path_host_conf); free (chroot); } commit bdd8422cfb1fe04cb20617495156fb232b00d23c Author: Florian Weimer Date: Fri Sep 1 08:56:46 2017 +0200 getaddrinfo: Use &errno has the errno pointer Similar code in nss/getXXbyYY_r.c is already using &errno as the argument. (cherry picked from commit 924b121c5978689001ae28cf1c8497371dad4f71) diff --git a/ChangeLog b/ChangeLog index 591c753fcc..04304bb929 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-09-01 Florian Weimer + + * sysdeps/posix/getaddrinfo.c (gethosts): Use errno directly. + (getcanonname): Likewise. + (gaih_inet): Likewise. + 2017-08-08 Florian Weimer * sysdeps/posix/getaddrinfo.c (gaih_inet): Remove unreachable diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index 09f85fc472..1a16820b7e 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -247,11 +247,10 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, char *localcanon = NULL; \ no_data = 0; \ while (1) { \ - rc = 0; \ status = DL_CALL_FCT (fct, (name, _family, &th, \ tmpbuf->data, tmpbuf->length, \ - &rc, &herrno, NULL, &localcanon)); \ - if (rc != ERANGE || herrno != NETDB_INTERNAL) \ + &errno, &herrno, NULL, &localcanon)); \ + if (errno != ERANGE || herrno != NETDB_INTERNAL) \ break; \ if (!scratch_buffer_grow (tmpbuf)) \ { \ @@ -261,11 +260,11 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, goto free_and_return; \ } \ } \ - if (status == NSS_STATUS_SUCCESS && rc == 0) \ + if (status == NSS_STATUS_SUCCESS && errno == 0) \ h = &th; \ else \ h = NULL; \ - if (rc != 0) \ + if (errno != 0) \ { \ if (herrno == NETDB_INTERNAL) \ { \ @@ -335,9 +334,8 @@ getcanonname (service_user *nip, struct gaih_addrtuple *at, const char *name) { char buf[256]; int herrno; - int rc; if (DL_CALL_FCT (cfct, (at->name ?: name, buf, sizeof (buf), - &s, &rc, &herrno)) != NSS_STATUS_SUCCESS) + &s, &errno, &herrno)) != NSS_STATUS_SUCCESS) /* If the canonical name cannot be determined, use the passed string. */ s = (char *) name; @@ -353,7 +351,6 @@ gaih_inet (const char *name, const struct gaih_service *service, const struct gaih_typeproto *tp = gaih_inet_typeproto; struct gaih_servtuple *st = (struct gaih_servtuple *) &nullserv; struct gaih_addrtuple *at = NULL; - int rc; bool got_ipv6 = false; const char *canon = NULL; const char *orig_name = name; @@ -395,7 +392,8 @@ gaih_inet (const char *name, const struct gaih_service *service, st = (struct gaih_servtuple *) alloca_account (sizeof (struct gaih_servtuple), alloca_used); - if ((rc = gaih_inet_serv (service->name, tp, req, st, tmpbuf))) + int rc = gaih_inet_serv (service->name, tp, req, st, tmpbuf); + if (__glibc_unlikely (rc != 0)) return rc; } else @@ -495,7 +493,7 @@ gaih_inet (const char *name, const struct gaih_service *service, idn_flags |= IDNA_USE_STD3_ASCII_RULES; char *p = NULL; - rc = __idna_to_ascii_lz (name, &p, idn_flags); + int rc = __idna_to_ascii_lz (name, &p, idn_flags); if (rc != IDNA_SUCCESS) { /* No need to jump to free_and_return here. */ @@ -793,15 +791,14 @@ gaih_inet (const char *name, const struct gaih_service *service, while (1) { - rc = 0; status = DL_CALL_FCT (fct4, (name, pat, tmpbuf->data, tmpbuf->length, - &rc, &herrno, + &errno, &herrno, NULL)); if (status == NSS_STATUS_SUCCESS) break; if (status != NSS_STATUS_TRYAGAIN - || rc != ERANGE || herrno != NETDB_INTERNAL) + || errno != ERANGE || herrno != NETDB_INTERNAL) { if (herrno == TRY_AGAIN) no_data = EAI_AGAIN; commit 701f7873da013fb19d9120317322cd78333e63c2 Author: Florian Weimer Date: Fri Sep 1 08:57:07 2017 +0200 getaddrinfo: Use &h_errno has the h_errno pointer This simplifies the code because it is not necessary to propagate the temporary h_errno value to the thread-local variable. It also increases compatibility with NSS modules which update only one of the two places. (cherry picked from commit 53250a21b81474ef4e78090a4a9a63d8471e1091) diff --git a/ChangeLog b/ChangeLog index 04304bb929..d53c8fe083 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-09-01 Florian Weimer + + * sysdeps/posix/getaddrinfo.c (gethosts): Use h_errno directly. + (getcanonname): Likewise. + (gaih_inet): Likewise. + 2017-09-01 Florian Weimer * sysdeps/posix/getaddrinfo.c (gethosts): Use errno directly. diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index 1a16820b7e..076e1fa62b 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -241,7 +241,6 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, #define gethosts(_family, _type) \ { \ - int herrno; \ struct hostent th; \ struct hostent *h; \ char *localcanon = NULL; \ @@ -249,8 +248,8 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, while (1) { \ status = DL_CALL_FCT (fct, (name, _family, &th, \ tmpbuf->data, tmpbuf->length, \ - &errno, &herrno, NULL, &localcanon)); \ - if (errno != ERANGE || herrno != NETDB_INTERNAL) \ + &errno, &h_errno, NULL, &localcanon)); \ + if (errno != ERANGE || h_errno != NETDB_INTERNAL) \ break; \ if (!scratch_buffer_grow (tmpbuf)) \ { \ @@ -266,18 +265,17 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, h = NULL; \ if (errno != 0) \ { \ - if (herrno == NETDB_INTERNAL) \ + if (h_errno == NETDB_INTERNAL) \ { \ - __set_h_errno (herrno); \ __resolv_context_enable_inet6 (res_ctx, res_enable_inet6); \ __resolv_context_put (res_ctx); \ result = -EAI_SYSTEM; \ goto free_and_return; \ } \ - if (herrno == TRY_AGAIN) \ + if (h_errno == TRY_AGAIN) \ no_data = EAI_AGAIN; \ else \ - no_data = herrno == NO_DATA; \ + no_data = h_errno == NO_DATA; \ } \ else if (h != NULL) \ { \ @@ -333,9 +331,8 @@ getcanonname (service_user *nip, struct gaih_addrtuple *at, const char *name) if (cfct != NULL) { char buf[256]; - int herrno; if (DL_CALL_FCT (cfct, (at->name ?: name, buf, sizeof (buf), - &s, &errno, &herrno)) != NSS_STATUS_SUCCESS) + &s, &errno, &h_errno)) != NSS_STATUS_SUCCESS) /* If the canonical name cannot be determined, use the passed string. */ s = (char *) name; @@ -594,14 +591,13 @@ gaih_inet (const char *name, const struct gaih_service *service, int rc; struct hostent th; struct hostent *h; - int herrno; while (1) { rc = __gethostbyname2_r (name, AF_INET, &th, tmpbuf->data, tmpbuf->length, - &h, &herrno); - if (rc != ERANGE || herrno != NETDB_INTERNAL) + &h, &h_errno); + if (rc != ERANGE || h_errno != NETDB_INTERNAL) break; if (!scratch_buffer_grow (tmpbuf)) { @@ -626,12 +622,9 @@ gaih_inet (const char *name, const struct gaih_service *service, } else { - if (herrno == NETDB_INTERNAL) - { - __set_h_errno (herrno); - result = -EAI_SYSTEM; - } - else if (herrno == TRY_AGAIN) + if (h_errno == NETDB_INTERNAL) + result = -EAI_SYSTEM; + else if (h_errno == TRY_AGAIN) result = -EAI_AGAIN; else /* We made requests but they turned out no data. @@ -654,8 +647,7 @@ gaih_inet (const char *name, const struct gaih_service *service, { /* Try to use nscd. */ struct nscd_ai_result *air = NULL; - int herrno; - int err = __nscd_getai (name, &air, &herrno); + int err = __nscd_getai (name, &air, &h_errno); if (air != NULL) { /* Transform into gaih_addrtuple list. */ @@ -746,9 +738,9 @@ gaih_inet (const char *name, const struct gaih_service *service, goto free_and_return; else if (__nss_not_use_nscd_hosts == 0) { - if (herrno == NETDB_INTERNAL && errno == ENOMEM) + if (h_errno == NETDB_INTERNAL && errno == ENOMEM) result = -EAI_MEMORY; - else if (herrno == TRY_AGAIN) + else if (h_errno == TRY_AGAIN) result = -EAI_AGAIN; else result = -EAI_SYSTEM; @@ -787,23 +779,21 @@ gaih_inet (const char *name, const struct gaih_service *service, if (fct4 != NULL) { - int herrno; - while (1) { status = DL_CALL_FCT (fct4, (name, pat, tmpbuf->data, tmpbuf->length, - &errno, &herrno, + &errno, &h_errno, NULL)); if (status == NSS_STATUS_SUCCESS) break; if (status != NSS_STATUS_TRYAGAIN - || errno != ERANGE || herrno != NETDB_INTERNAL) + || errno != ERANGE || h_errno != NETDB_INTERNAL) { - if (herrno == TRY_AGAIN) + if (h_errno == TRY_AGAIN) no_data = EAI_AGAIN; else - no_data = herrno == NO_DATA; + no_data = h_errno == NO_DATA; break; } commit 8f46c6052408a23a77ecf46aa378120c1a4afe37 Author: Florian Weimer Date: Fri Sep 1 08:57:28 2017 +0200 getaddrinfo: Properly set errno for NSS function lookup failure (cherry picked from commit ad816a5e00ce891a2cea8187638fa0e00f83aaf6) diff --git a/ChangeLog b/ChangeLog index d53c8fe083..2f959d4c36 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-09-01 Florian Weimer + + * sysdeps/posix/getaddrinfo.c (gaih_inet): Make reporting of NSS + function lookup failures more reliable. + 2017-09-01 Florian Weimer * sysdeps/posix/getaddrinfo.c (gethosts): Use h_errno directly. diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index 076e1fa62b..eaf8bafcf4 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -923,13 +923,17 @@ gaih_inet (const char *name, const struct gaih_service *service, } else { + /* Could not locate any of the lookup functions. + The NSS lookup code does not consistently set + errno, so we need to supply our own error + code here. The root cause could either be a + resource allocation failure, or a missing + service function in the DSO (so it should not + be listed in /etc/nsswitch.conf). Assume the + former, and return EBUSY. */ status = NSS_STATUS_UNAVAIL; - /* Could not load any of the lookup functions. Indicate - an internal error if the failure was due to a system - error other than the file not being found. We use the - errno from the last failed callback. */ - if (errno != 0 && errno != ENOENT) - __set_h_errno (NETDB_INTERNAL); + __set_h_errno (NETDB_INTERNAL); + __set_errno (EBUSY); } } commit 7ab87bccb657b02cac5a3360b11c67aff901de2e Author: Florian Weimer Date: Fri Sep 1 08:57:52 2017 +0200 getaddrinfo: In gaih_inet, use h_errno for certain status values only h_errno is not set for NSS_STATUS_SUCCESS, so its value might not be accurate at this point. (cherry picked from commit a2881ef01450295782b065f2f850f340d5c12c14) diff --git a/ChangeLog b/ChangeLog index 2f959d4c36..c58cd5ccd0 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-09-01 Florian Weimer + + * sysdeps/posix/getaddrinfo.c (gaih_inet): Only use h_errno if + status indicates it is set. + 2017-09-01 Florian Weimer * sysdeps/posix/getaddrinfo.c (gaih_inet): Make reporting of NSS diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index eaf8bafcf4..9d9e7e2bf2 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -949,7 +949,10 @@ gaih_inet (const char *name, const struct gaih_service *service, __resolv_context_enable_inet6 (res_ctx, res_enable_inet6); __resolv_context_put (res_ctx); - if (h_errno == NETDB_INTERNAL) + /* If we have a failure which sets errno, report it using + EAI_SYSTEM. */ + if ((status == NSS_STATUS_TRYAGAIN || status == NSS_STATUS_UNAVAIL) + && h_errno == NETDB_INTERNAL) { result = -EAI_SYSTEM; goto free_and_return; commit 7966331555df43bb7e2a55ce5a17a330e57f487f Author: Florian Weimer Date: Mon Sep 4 11:25:34 2017 +0200 getaddrinfo: Return EAI_NODATA if gethostbyname2_r reports NO_DATA [BZ #21922] (cherry picked from commit 5f8340f583fe3d4f5734bd2371c5a45ecff2db0d) diff --git a/ChangeLog b/ChangeLog index c58cd5ccd0..f46bbb7c0d 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-09-01 Florian Weimer + + [BZ #21922] + * sysdeps/posix/getaddrinfo.c (gaih_inet): Report EAI_NODATA error + coming from gethostbyname2_r. + 2017-09-01 Florian Weimer * sysdeps/posix/getaddrinfo.c (gaih_inet): Only use h_errno if diff --git a/NEWS b/NEWS index 756e849643..97eb21e868 100644 --- a/NEWS +++ b/NEWS @@ -22,6 +22,7 @@ The following bugs are resolved with this release: [21780] posix: Set p{read,write}v2 to return ENOTSUP [21871] x86-64: Use _dl_runtime_resolve_opt only with AVX512F [21885] getaddrinfo: Release resolver context on error in gethosts + [21922] getaddrinfo with AF_INET(6) returns EAI_NONAME, not EAI_NODATA [21930] Do not use __builtin_types_compatible_p in C++ mode [21932] Unpaired __resolv_context_get in generic get*_r implementation [21941] powerpc: Restrict xssqrtqp operands to Vector Registers diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index 9d9e7e2bf2..0cf87c224d 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -619,6 +619,14 @@ gaih_inet (const char *name, const struct gaih_service *service, } *pat = addrmem; } + else + { + if (h_errno == NO_DATA) + result = -EAI_NODATA; + else + result = -EAI_NONAME; + goto free_and_return; + } } else { commit a71a3374cd8cf53776c33994f69ec184c26f2129 Author: Florian Weimer Date: Mon Sep 4 11:27:24 2017 +0200 getaddrinfo: Fix error handling in gethosts [BZ #21915] [BZ #21922] The old code uses errno as the primary indicator for success or failure. This is wrong because errno is only set for specific combinations of the status return value and the h_errno variable. (cherry picked from commit f4a6be2582b8dfe8adfa68da3dd8decf566b3983) diff --git a/ChangeLog b/ChangeLog index f46bbb7c0d..3e32d14dbf 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,17 @@ +2017-09-01 Florian Weimer + + [BZ #21915] + [BZ #21922] + * sysdeps/posix/getaddrinfo.c (gethosts): Look at NSS function + result to determine success or failure, not the errno value. + * nss/Makefile (tests): Add tst-nss-files-hosts-erange. + (tst-nss-files-hosts-erange): Link with -ldl. + * nss/tst-nss-files-hosts-erange.c: New file. + * nss/tst-resolv-basic.c (response): Handle nodata.example. + (do_test): Add NO_DATA tests. + * resolv/tst-resolv-basic.c (test_nodata_nxdomain): New function. + (do_test): Call it. + 2017-09-01 Florian Weimer [BZ #21922] diff --git a/NEWS b/NEWS index 97eb21e868..8fbf4241d1 100644 --- a/NEWS +++ b/NEWS @@ -22,6 +22,7 @@ The following bugs are resolved with this release: [21780] posix: Set p{read,write}v2 to return ENOTSUP [21871] x86-64: Use _dl_runtime_resolve_opt only with AVX512F [21885] getaddrinfo: Release resolver context on error in gethosts + [21915] getaddrinfo: incorrect result handling for NSS service modules [21922] getaddrinfo with AF_INET(6) returns EAI_NONAME, not EAI_NODATA [21930] Do not use __builtin_types_compatible_p in C++ mode [21932] Unpaired __resolv_context_get in generic get*_r implementation diff --git a/nss/Makefile b/nss/Makefile index d9f6d41181..91b1c21567 100644 --- a/nss/Makefile +++ b/nss/Makefile @@ -58,6 +58,11 @@ tests = test-netdb test-digits-dots tst-nss-getpwent bug17079 \ tst-nss-test5 xtests = bug-erange +# Tests which need libdl +ifeq (yes,$(build-shared)) +tests += tst-nss-files-hosts-erange +endif + # If we have a thread library then we can test cancellation against # some routines like getpwuid_r. ifeq (yes,$(have-thread-library)) @@ -154,3 +159,5 @@ $(patsubst %,$(objpfx)%.out,$(tests)) : \ ifeq (yes,$(have-thread-library)) $(objpfx)tst-cancel-getpwuid_r: $(shared-thread-library) endif + +$(objpfx)tst-nss-files-hosts-erange: $(libdl) diff --git a/nss/tst-nss-files-hosts-erange.c b/nss/tst-nss-files-hosts-erange.c new file mode 100644 index 0000000000..beb7aa9fa0 --- /dev/null +++ b/nss/tst-nss-files-hosts-erange.c @@ -0,0 +1,109 @@ +/* Parse /etc/hosts in multi mode with a trailing long line (bug 21915). + Copyright (C) 2017 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 + . */ + + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +struct support_chroot *chroot_env; + +#define X10 "XXXXXXXXXX" +#define X100 X10 X10 X10 X10 X10 X10 X10 X10 X10 X10 +#define X1000 X100 X100 X100 X100 X100 X100 X100 X100 X100 X100 + +static void +prepare (int argc, char **argv) +{ + chroot_env = support_chroot_create + ((struct support_chroot_configuration) + { + .resolv_conf = "", + .hosts = + "127.0.0.1 localhost localhost.localdomain\n" + "::1 localhost localhost.localdomain\n" + "192.0.2.1 example.com\n" + "#" X1000 X100 "\n", + .host_conf = "multi on\n", + }); +} + +static int +do_test (void) +{ + support_become_root (); + if (!support_can_chroot ()) + return EXIT_UNSUPPORTED; + + __nss_configure_lookup ("hosts", "files"); + if (dlopen (LIBNSS_FILES_SO, RTLD_LAZY) == NULL) + FAIL_EXIT1 ("could not load " LIBNSS_DNS_SO ": %s", dlerror ()); + + xchroot (chroot_env->path_chroot); + + errno = ERANGE; + h_errno = NETDB_INTERNAL; + check_hostent ("gethostbyname example.com", + gethostbyname ("example.com"), + "name: example.com\n" + "address: 192.0.2.1\n"); + errno = ERANGE; + h_errno = NETDB_INTERNAL; + check_hostent ("gethostbyname2 AF_INET example.com", + gethostbyname2 ("example.com", AF_INET), + "name: example.com\n" + "address: 192.0.2.1\n"); + { + struct addrinfo hints = + { + .ai_family = AF_UNSPEC, + .ai_socktype = SOCK_STREAM, + .ai_protocol = IPPROTO_TCP, + }; + errno = ERANGE; + h_errno = NETDB_INTERNAL; + struct addrinfo *ai; + int ret = getaddrinfo ("example.com", "80", &hints, &ai); + check_addrinfo ("example.com AF_UNSPEC", ai, ret, + "address: STREAM/TCP 192.0.2.1 80\n"); + if (ret == 0) + freeaddrinfo (ai); + + hints.ai_family = AF_INET; + errno = ERANGE; + h_errno = NETDB_INTERNAL; + ret = getaddrinfo ("example.com", "80", &hints, &ai); + check_addrinfo ("example.com AF_INET", ai, ret, + "address: STREAM/TCP 192.0.2.1 80\n"); + if (ret == 0) + freeaddrinfo (ai); + } + + support_chroot_free (chroot_env); + return 0; +} + +#define PREPARE prepare +#include diff --git a/resolv/tst-resolv-basic.c b/resolv/tst-resolv-basic.c index 64eedbbd81..66a0e8a165 100644 --- a/resolv/tst-resolv-basic.c +++ b/resolv/tst-resolv-basic.c @@ -50,7 +50,7 @@ response (const struct resolv_response_context *ctx, qname_compare = qname + 2; else qname_compare = qname; - enum {www, alias, nxdomain, long_name} requested_qname; + enum {www, alias, nxdomain, long_name, nodata} requested_qname; if (strcmp (qname_compare, "www.example") == 0) requested_qname = www; else if (strcmp (qname_compare, "alias.example") == 0) @@ -59,6 +59,8 @@ response (const struct resolv_response_context *ctx, requested_qname = nxdomain; else if (strcmp (qname_compare, LONG_NAME) == 0) requested_qname = long_name; + else if (strcmp (qname_compare, "nodata.example") == 0) + requested_qname = nodata; else { support_record_failure (); @@ -87,6 +89,8 @@ response (const struct resolv_response_context *ctx, resolv_response_close_record (b); resolv_response_open_record (b, "www.example", qclass, qtype, 0); break; + case nodata: + return; case nxdomain: FAIL_EXIT1 ("unreachable"); } @@ -267,6 +271,55 @@ test_bug_21295 (void) } } +/* Run tests which do not expect any data. */ +static void +test_nodata_nxdomain (void) +{ + /* Iterate through different address families. */ + int families[] = { AF_UNSPEC, AF_INET, AF_INET6, -1 }; + for (int i = 0; families[i] >= 0; ++i) + /* If do_tcp, prepend "t." to the name to trigger TCP + fallback. */ + for (int do_tcp = 0; do_tcp < 2; ++do_tcp) + /* If do_nxdomain, trigger an NXDOMAIN error (DNS failure), + otherwise use a NODATA response (empty but successful + answer). */ + for (int do_nxdomain = 0; do_nxdomain < 2; ++do_nxdomain) + { + int family = families[i]; + char *name = xasprintf ("%s%s.example", + do_tcp ? "t." : "", + do_nxdomain ? "nxdomain" : "nodata"); + + if (family != AF_UNSPEC) + { + if (do_nxdomain) + check_h (name, family, "error: HOST_NOT_FOUND\n"); + else + check_h (name, family, "error: NO_ADDRESS\n"); + } + + const char *expected; + if (do_nxdomain) + expected = "error: Name or service not known\n"; + else + expected = "error: No address associated with hostname\n"; + + check_ai (name, "80", family, expected); + + struct addrinfo hints = + { + .ai_family = family, + .ai_flags = AI_V4MAPPED | AI_ALL, + }; + check_ai_hints (name, "80", hints, expected); + hints.ai_flags |= AI_CANONNAME; + check_ai_hints (name, "80", hints, expected); + + free (name); + } +} + static int do_test (void) { @@ -439,29 +492,8 @@ do_test (void) "address: DGRAM/UDP 2001:db8::4 80\n" "address: RAW/IP 2001:db8::4 80\n"); - check_h ("nxdomain.example", AF_INET, - "error: HOST_NOT_FOUND\n"); - check_h ("nxdomain.example", AF_INET6, - "error: HOST_NOT_FOUND\n"); - check_ai ("nxdomain.example", "80", AF_UNSPEC, - "error: Name or service not known\n"); - check_ai ("nxdomain.example", "80", AF_INET, - "error: Name or service not known\n"); - check_ai ("nxdomain.example", "80", AF_INET6, - "error: Name or service not known\n"); - - check_h ("t.nxdomain.example", AF_INET, - "error: HOST_NOT_FOUND\n"); - check_h ("t.nxdomain.example", AF_INET6, - "error: HOST_NOT_FOUND\n"); - check_ai ("t.nxdomain.example", "80", AF_UNSPEC, - "error: Name or service not known\n"); - check_ai ("t.nxdomain.example", "80", AF_INET, - "error: Name or service not known\n"); - check_ai ("t.nxdomain.example", "80", AF_INET6, - "error: Name or service not known\n"); - test_bug_21295 (); + test_nodata_nxdomain (); resolv_test_end (aux); diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c index 0cf87c224d..2c4b6d6793 100644 --- a/sysdeps/posix/getaddrinfo.c +++ b/sysdeps/posix/getaddrinfo.c @@ -242,28 +242,26 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, #define gethosts(_family, _type) \ { \ struct hostent th; \ - struct hostent *h; \ char *localcanon = NULL; \ no_data = 0; \ - while (1) { \ - status = DL_CALL_FCT (fct, (name, _family, &th, \ - tmpbuf->data, tmpbuf->length, \ - &errno, &h_errno, NULL, &localcanon)); \ - if (errno != ERANGE || h_errno != NETDB_INTERNAL) \ - break; \ - if (!scratch_buffer_grow (tmpbuf)) \ - { \ - __resolv_context_enable_inet6 (res_ctx, res_enable_inet6); \ - __resolv_context_put (res_ctx); \ - result = -EAI_MEMORY; \ - goto free_and_return; \ - } \ - } \ - if (status == NSS_STATUS_SUCCESS && errno == 0) \ - h = &th; \ - else \ - h = NULL; \ - if (errno != 0) \ + while (1) \ + { \ + status = DL_CALL_FCT (fct, (name, _family, &th, \ + tmpbuf->data, tmpbuf->length, \ + &errno, &h_errno, NULL, &localcanon)); \ + if (status != NSS_STATUS_TRYAGAIN || h_errno != NETDB_INTERNAL \ + || errno != ERANGE) \ + break; \ + if (!scratch_buffer_grow (tmpbuf)) \ + { \ + __resolv_context_enable_inet6 (res_ctx, res_enable_inet6); \ + __resolv_context_put (res_ctx); \ + result = -EAI_MEMORY; \ + goto free_and_return; \ + } \ + } \ + if (status == NSS_STATUS_NOTFOUND \ + || status == NSS_STATUS_TRYAGAIN || status == NSS_STATUS_UNAVAIL) \ { \ if (h_errno == NETDB_INTERNAL) \ { \ @@ -277,9 +275,9 @@ convert_hostent_to_gaih_addrtuple (const struct addrinfo *req, else \ no_data = h_errno == NO_DATA; \ } \ - else if (h != NULL) \ + else if (status == NSS_STATUS_SUCCESS) \ { \ - if (!convert_hostent_to_gaih_addrtuple (req, _family,h, &addrmem)) \ + if (!convert_hostent_to_gaih_addrtuple (req, _family, &th, &addrmem)) \ { \ __resolv_context_enable_inet6 (res_ctx, res_enable_inet6); \ __resolv_context_put (res_ctx); \ commit 85cfe508568530eed2d9cfd34110c21721d1f99e Author: Florian Weimer Date: Wed Sep 6 13:43:01 2017 +0200 tst-res_use_inet6: Enhance test to cover IPv4-to-IPv6 address mapping This requires more control over the response data, so it is now determined by flags embedded in the query name. (cherry picked from commit 5e9c4d17feb9910f489ad2915d0b6e00597a0f11) diff --git a/ChangeLog b/ChangeLog index 3e32d14dbf..dd71f6c427 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,15 @@ +2017-09-06 Florian Weimer + + Enhance tst-res_use_inet6 to test IPv4-to-IPv6 address mapping. + * resolv/tst-res_use_inet6.c (response): Process flags embedded in + the QNAME. + (test_gai): Adjust query names. Add additional tests. + (test_get2_any, test_get2_no_inet6, test_get2_inet6): Split from + test_get2. Adjust query names. Add additional tests. + (test_no_inet6): New function, extracted from threadfunc. + (threadfunc): Call test_get2_any, test_get2_inet6, test_no_inet6. + Add additional tests. + 2017-09-01 Florian Weimer [BZ #21915] diff --git a/resolv/tst-res_use_inet6.c b/resolv/tst-res_use_inet6.c index 6f3db08892..1522d5c5f5 100644 --- a/resolv/tst-res_use_inet6.c +++ b/resolv/tst-res_use_inet6.c @@ -19,18 +19,44 @@ #include #include #include +#include #include #include #include +/* Produce a response based on QNAME: Certain characters in the first + label of QNAME trigger the inclusion of resource records: + + 'a' A record (IPv4 address) + 'q' AAAA record (quad A record, IPv6 address) + 'm' record type must match QTYPE (no additional records) + + QTYPE is ignored for record type selection if 'm' is not + specified. */ static void response (const struct resolv_response_context *ctx, struct resolv_response_builder *b, const char *qname, uint16_t qclass, uint16_t qtype) { - bool include_both = strcmp (qname, "both.example") == 0; - bool include_a = qtype == T_A || include_both; - bool include_aaaa = qtype == T_AAAA || include_both; + bool include_a = false; + bool include_aaaa = false; + bool include_match = false; + for (const char *p = qname; *p != '.' && *p != '\0'; ++p) + { + if (*p == 'a') + include_a = true; + else if (*p == 'q') + include_aaaa = true; + else if (*p == 'm') + include_match = true; + } + if (include_match) + { + if (qtype == T_A) + include_aaaa = false; + else if (qtype == T_AAAA) + include_a = false; + } resolv_response_init (b, (struct resolv_response_flags) {}); resolv_response_add_question (b, qname, qclass, qtype); @@ -64,16 +90,21 @@ test_gai (void) .ai_protocol = IPPROTO_TCP, }; struct addrinfo *ai; - int ret = getaddrinfo ("www1.example", "80", &hints, &ai); - check_addrinfo ("getaddrinfo AF_UNSPEC www1.example", ai, ret, + int ret = getaddrinfo ("qam.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_UNSPEC qam.example", ai, ret, "address: STREAM/TCP 192.0.2.17 80\n" "address: STREAM/TCP 2001:db8::1 80\n"); if (ret == 0) freeaddrinfo (ai); - ret = getaddrinfo ("both.example", "80", &hints, &ai); + ret = getaddrinfo ("am.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_UNSPEC am.example", ai, ret, + "address: STREAM/TCP 192.0.2.17 80\n"); + if (ret == 0) + freeaddrinfo (ai); + ret = getaddrinfo ("qa.example", "80", &hints, &ai); /* Combined A/AAAA responses currently result in address duplication. */ - check_addrinfo ("getaddrinfo AF_UNSPEC both.example", ai, ret, + check_addrinfo ("getaddrinfo AF_UNSPEC qa.example", ai, ret, "address: STREAM/TCP 192.0.2.17 80\n" "address: STREAM/TCP 192.0.2.17 80\n" "address: STREAM/TCP 2001:db8::1 80\n" @@ -89,13 +120,18 @@ test_gai (void) .ai_protocol = IPPROTO_TCP, }; struct addrinfo *ai; - int ret = getaddrinfo ("www1.example", "80", &hints, &ai); - check_addrinfo ("getaddrinfo AF_INET www1.example", ai, ret, + int ret = getaddrinfo ("qam.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_INET qam.example", ai, ret, + "address: STREAM/TCP 192.0.2.17 80\n"); + if (ret == 0) + freeaddrinfo (ai); + ret = getaddrinfo ("am.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_INET am.example", ai, ret, "address: STREAM/TCP 192.0.2.17 80\n"); if (ret == 0) freeaddrinfo (ai); - ret = getaddrinfo ("both.example", "80", &hints, &ai); - check_addrinfo ("getaddrinfo AF_INET both.example", ai, ret, + ret = getaddrinfo ("qa.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_INET qa.example", ai, ret, "address: STREAM/TCP 192.0.2.17 80\n"); if (ret == 0) freeaddrinfo (ai); @@ -108,40 +144,131 @@ test_gai (void) .ai_protocol = IPPROTO_TCP, }; struct addrinfo *ai; - int ret = getaddrinfo ("www1.example", "80", &hints, &ai); + int ret = getaddrinfo ("qa.example", "80", &hints, &ai); check_addrinfo ("getaddrinfo (AF_INET6)", ai, ret, "address: STREAM/TCP 2001:db8::1 80\n"); if (ret == 0) freeaddrinfo (ai); - ret = getaddrinfo ("both.example", "80", &hints, &ai); - check_addrinfo ("getaddrinfo AF_INET6 both.example", ai, ret, + ret = getaddrinfo ("am.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_INET6 am.example", ai, ret, + "error: No address associated with hostname\n"); + if (ret == 0) + freeaddrinfo (ai); + ret = getaddrinfo ("qam.example", "80", &hints, &ai); + check_addrinfo ("getaddrinfo AF_INET6 qam.example", ai, ret, "address: STREAM/TCP 2001:db8::1 80\n"); if (ret == 0) freeaddrinfo (ai); } } -/* Test that gethostbyname2 is not influenced by RES_USE_INET6. */ +/* Test that gethostbyname2 is mostly not influenced by + RES_USE_INET6. */ static void -test_get2 (void) +test_get2_any (void) { - check_hostent ("gethostbyname2 AF_INET www1.example", - gethostbyname2 ("www1.example", AF_INET), - "name: www1.example\n" + check_hostent ("gethostbyname2 AF_INET am.example", + gethostbyname2 ("am.example", AF_INET), + "name: am.example\n" "address: 192.0.2.17\n"); - check_hostent ("gethostbyname2 AF_INET both.example", - gethostbyname2 ("both.example", AF_INET), - "name: both.example\n" + check_hostent ("gethostbyname2 AF_INET a.example", + gethostbyname2 ("a.example", AF_INET), + "name: a.example\n" + "address: 192.0.2.17\n"); + check_hostent ("gethostbyname2 AF_INET qm.example", + gethostbyname2 ("qm.example", AF_INET), + "error: NO_ADDRESS\n"); + check_hostent ("gethostbyname2 AF_INET q.example", + gethostbyname2 ("q.example", AF_INET), + "error: NO_RECOVERY\n"); + check_hostent ("gethostbyname2 AF_INET qam.example", + gethostbyname2 ("qam.example", AF_INET), + "name: qam.example\n" + "address: 192.0.2.17\n"); + check_hostent ("gethostbyname2 AF_INET qa.example", + gethostbyname2 ("qa.example", AF_INET), + "name: qa.example\n" "address: 192.0.2.17\n"); - check_hostent ("gethostbyname2 AF_INET6 www1.example", - gethostbyname2 ("www1.example", AF_INET6), - "name: www1.example\n" + check_hostent ("gethostbyname2 AF_INET6 qm.example", + gethostbyname2 ("qm.example", AF_INET6), + "name: qm.example\n" + "address: 2001:db8::1\n"); + check_hostent ("gethostbyname2 AF_INET6 q.example", + gethostbyname2 ("q.example", AF_INET6), + "name: q.example\n" "address: 2001:db8::1\n"); - check_hostent ("gethostbyname2 AF_INET6 both.example", - gethostbyname2 ("both.example", AF_INET6), - "name: both.example\n" + check_hostent ("gethostbyname2 AF_INET6 qam.example", + gethostbyname2 ("qam.example", AF_INET6), + "name: qam.example\n" "address: 2001:db8::1\n"); + check_hostent ("gethostbyname2 AF_INET6 qa.example", + gethostbyname2 ("qa.example", AF_INET6), + "name: qa.example\n" + "address: 2001:db8::1\n"); + /* Additional AF_INET6 tests depend on RES_USE_INET6; see below. */ +} + +/* gethostbyname2 tests with RES_USE_INET6 disabled. */ +static void +test_get2_no_inet6 (void) +{ + test_get2_any (); + + check_hostent ("gethostbyname2 AF_INET6 am.example", + gethostbyname2 ("am.example", AF_INET6), + "error: NO_ADDRESS\n"); + check_hostent ("gethostbyname2 AF_INET6 a.example", + gethostbyname2 ("a.example", AF_INET6), + "error: NO_RECOVERY\n"); +} + +/* gethostbyname2 tests with RES_USE_INET6 enabled. */ +static void +test_get2_inet6 (void) +{ + test_get2_any (); + + check_hostent ("gethostbyname2 AF_INET6 am.example", + gethostbyname2 ("am.example", AF_INET6), + "name: am.example\n" + "address: ::ffff:192.0.2.17\n"); + check_hostent ("gethostbyname2 AF_INET6 a.example", + gethostbyname2 ("a.example", AF_INET6), + "error: NO_RECOVERY\n"); +} + +/* Collection of tests which assume no RES_USE_INET6 flag. */ +static void +test_no_inet6 (void) +{ + check_hostent ("gethostbyname (\"a.example\")", + gethostbyname ("a.example"), + "name: a.example\n" + "address: 192.0.2.17\n"); + check_hostent ("gethostbyname (\"qa.example\")", + gethostbyname ("qa.example"), + "name: qa.example\n" + "address: 192.0.2.17\n"); + check_hostent ("gethostbyname (\"am.example\")", + gethostbyname ("am.example"), + "name: am.example\n" + "address: 192.0.2.17\n"); + check_hostent ("gethostbyname (\"qam.example\")", + gethostbyname ("qam.example"), + "name: qam.example\n" + "address: 192.0.2.17\n"); + check_hostent ("gethostbyname (\"q.example\")", + gethostbyname ("q.example"), + "error: NO_RECOVERY\n"); + check_hostent ("gethostbyname (\"qm.example\")", + gethostbyname ("qm.example"), + "error: NO_ADDRESS\n"); + test_get2_no_inet6 (); + test_get2_no_inet6 (); + test_gai (); + test_get2_no_inet6 (); + test_get2_no_inet6 (); } static void * @@ -153,28 +280,42 @@ threadfunc (void *ignored) .response_callback = response }); - check_hostent ("gethostbyname (\"www1.example\")", - gethostbyname ("www1.example"), - "name: www1.example\n" - "address: 192.0.2.17\n"); - check_hostent ("gethostbyname (\"both.example\")", - gethostbyname ("both.example"), - "name: both.example\n" - "address: 192.0.2.17\n"); - test_get2 (); - test_gai (); + TEST_VERIFY ((_res.options & RES_USE_INET6) == 0); + test_no_inet6 (); _res.options |= RES_USE_INET6; - check_hostent ("gethostbyname (\"www1.example\")", - gethostbyname ("www1.example"), - "name: www1.example\n" + check_hostent ("gethostbyname (\"a.inet6.example\")", + gethostbyname ("a.inet6.example"), + "error: NO_RECOVERY\n"); + check_hostent ("gethostbyname (\"am.inet6.example\")", + gethostbyname ("am.inet6.example"), + "name: am.inet6.example\n" + "address: ::ffff:192.0.2.17\n"); + check_hostent ("gethostbyname (\"qa.inet6.example\")", + gethostbyname ("qa.inet6.example"), + "name: qa.inet6.example\n" + "address: 2001:db8::1\n"); + check_hostent ("gethostbyname (\"qam.inet6.example\")", + gethostbyname ("qam.inet6.example"), + "name: qam.inet6.example\n" "address: 2001:db8::1\n"); - check_hostent ("gethostbyname (\"both.example\")", - gethostbyname ("both.example"), - "name: both.example\n" + check_hostent ("gethostbyname (\"q.inet6.example\")", + gethostbyname ("q.inet6.example"), + "name: q.inet6.example\n" "address: 2001:db8::1\n"); - test_get2 (); + check_hostent ("gethostbyname (\"qm.inet6.example\")", + gethostbyname ("qm.inet6.example"), + "name: qm.inet6.example\n" + "address: 2001:db8::1\n"); + test_get2_inet6 (); + test_get2_inet6 (); test_gai (); + test_get2_inet6 (); + test_get2_inet6 (); + + TEST_VERIFY (_res.options & RES_USE_INET6); + _res.options &= ~RES_USE_INET6; + test_no_inet6 (); resolv_test_end (obj); commit 3005466abe8fb80ad4ff51865f1e28dd81c43347 Author: Florian Weimer Date: Wed Sep 6 15:11:44 2017 +0200 nss_dns: Remove dead PTR IPv4-to-IPv6 mapping code (cherry picked from commit c77eb96925b719001237ca7c9e3cef40d795d66b) diff --git a/ChangeLog b/ChangeLog index dd71f6c427..fa215c2729 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,15 @@ +2017-09-06 Florian Weimer + + Remove dead PTR IPv4-to-IPv6 mapping code from nss_dns. + * resolv/nss_dns/dns-host.c (getanswer_r): Remove dead code. + * resolv/tst-res_use_inet6.c (response_ptr_v4, response_ptr_v6): + New functions. + (response): Call them. Add 'p', '6' flag processing. + (test_reverse): New function. + (test_get2_any): Call it. + (test_no_inet6): Add 'p' test. + (test_inet6): Likewise. + 2017-09-06 Florian Weimer Enhance tst-res_use_inet6 to test IPv4-to-IPv6 address mapping. diff --git a/resolv/nss_dns/dns-host.c b/resolv/nss_dns/dns-host.c index 7cd54ab504..1e85e4f08f 100644 --- a/resolv/nss_dns/dns-host.c +++ b/resolv/nss_dns/dns-host.c @@ -889,19 +889,6 @@ getanswer_r (struct resolv_context *ctx, /* bind would put multiple PTR records as aliases, but we don't do that. */ result->h_name = bp; - if (have_to_map) - { - n = strlen (bp) + 1; /* for the \0 */ - if (__glibc_unlikely (n >= MAXHOSTNAMELEN)) - { - ++had_error; - break; - } - bp += n; - linebuflen -= n; - if (map_v4v6_hostent (result, &bp, &linebuflen)) - goto too_small; - } *h_errnop = NETDB_SUCCESS; return NSS_STATUS_SUCCESS; case T_A: diff --git a/resolv/tst-res_use_inet6.c b/resolv/tst-res_use_inet6.c index 1522d5c5f5..d819f921d6 100644 --- a/resolv/tst-res_use_inet6.c +++ b/resolv/tst-res_use_inet6.c @@ -16,31 +16,101 @@ License along with the GNU C Library; if not, see . */ +#include #include #include +#include +#include #include #include #include #include +#include #include +/* Handle IPv4 reverse lookup responses. Product a PTR record + A-B-C-D.v4.example. */ +static void +response_ptr_v4 (const struct resolv_response_context *ctx, + struct resolv_response_builder *b, + const char *qname, uint16_t qclass, uint16_t qtype) +{ + int bytes[4]; + int offset = -1; + TEST_VERIFY (sscanf (qname, "%d.%d.%d.%d.in-addr.arpa%n", + bytes + 0, bytes + 1, bytes + 2, bytes + 3, + &offset) == 4); + TEST_VERIFY (offset == strlen (qname)); + resolv_response_init (b, (struct resolv_response_flags) {}); + resolv_response_add_question (b, qname, qclass, qtype); + resolv_response_section (b, ns_s_an); + resolv_response_open_record (b, qname, qclass, T_PTR, 0); + char *name = xasprintf ("%d-%d-%d-%d.v4.example", + bytes[3], bytes[2], bytes[1], bytes[0]); + resolv_response_add_name (b, name); + free (name); + resolv_response_close_record (b); +} + +/* Handle IPv6 reverse lookup responses. Produce a PTR record + <32 hex digits>.v6.example. */ +static void +response_ptr_v6 (const struct resolv_response_context *ctx, + struct resolv_response_builder *b, + const char *qname, uint16_t qclass, uint16_t qtype) +{ + + TEST_VERIFY_EXIT (strlen (qname) > 64); + + char bytes[33]; + for (int i = 0; i < 64; ++i) + if ((i % 2) == 0) + { + TEST_VERIFY (isxdigit ((unsigned char) qname[i])); + bytes[31 - i / 2] = qname[i]; + } + else + TEST_VERIFY_EXIT (qname[i] == '.'); + bytes[32] = '\0'; + + resolv_response_init (b, (struct resolv_response_flags) {}); + resolv_response_add_question (b, qname, qclass, qtype); + resolv_response_section (b, ns_s_an); + resolv_response_open_record (b, qname, qclass, T_PTR, 0); + char *name = xasprintf ("%s.v6.example", bytes); + resolv_response_add_name (b, name); + free (name); + resolv_response_close_record (b); +} + /* Produce a response based on QNAME: Certain characters in the first label of QNAME trigger the inclusion of resource records: 'a' A record (IPv4 address) 'q' AAAA record (quad A record, IPv6 address) + 'p' PTR record 'm' record type must match QTYPE (no additional records) + '6' stop flag processing if QTYPE == AAAA + + For 'a' and 'q', QTYPE is ignored for record type selection if 'm' + is not specified. - QTYPE is ignored for record type selection if 'm' is not - specified. */ + in-addr.arpa and ip6.arpa queries are handled separately in + response_ptr_v4 and response_ptr_v6. */ static void response (const struct resolv_response_context *ctx, struct resolv_response_builder *b, const char *qname, uint16_t qclass, uint16_t qtype) { + if (strstr (qname, ".in-addr.arpa") != NULL) + return response_ptr_v4 (ctx, b, qname, qclass, qtype); + else if (strstr (qname, ".ip6.arpa") != NULL) + return response_ptr_v6 (ctx, b, qname, qclass, qtype); + bool include_a = false; bool include_aaaa = false; bool include_match = false; + bool include_ptr = false; for (const char *p = qname; *p != '.' && *p != '\0'; ++p) { if (*p == 'a') @@ -49,6 +119,10 @@ response (const struct resolv_response_context *ctx, include_aaaa = true; else if (*p == 'm') include_match = true; + else if (*p == 'p') + include_ptr = true; + else if (*p == '6' && qtype == T_AAAA) + break; } if (include_match) { @@ -70,11 +144,17 @@ response (const struct resolv_response_context *ctx, } if (include_aaaa) { - char ipv6[16] - = {0x20, 0x01, 0xd, 0xb8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; - resolv_response_open_record (b, qname, qclass, T_AAAA, 0); - resolv_response_add_data (b, &ipv6, sizeof (ipv6)); - resolv_response_close_record (b); + char ipv6[16] + = {0x20, 0x01, 0xd, 0xb8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + resolv_response_open_record (b, qname, qclass, T_AAAA, 0); + resolv_response_add_data (b, &ipv6, sizeof (ipv6)); + resolv_response_close_record (b); + } + if (include_ptr) + { + resolv_response_open_record (b, qname, qclass, T_PTR, 0); + resolv_response_add_name (b, "ptr-target.example"); + resolv_response_close_record (b); } } @@ -162,6 +242,65 @@ test_gai (void) } } +/* Test gethostbyaddr and getnameinfo. The results are independent of + RES_USE_INET6. */ +static void +test_reverse (void) +{ + { + char ipv4[4] = { 192, 0, 2, 17 }; + check_hostent ("gethostbyaddr AF_INET", + gethostbyaddr (ipv4, sizeof (ipv4), AF_INET), + "name: 192-0-2-17.v4.example\n" + "address: 192.0.2.17\n"); + } + { + char ipv6[16] + = {0x20, 0x01, 0xd, 0xb8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + check_hostent ("gethostbyaddr AF_INET", + gethostbyaddr (ipv6, sizeof (ipv6), AF_INET6), + "name: 20010db8000000000000000000000001.v6.example\n" + "address: 2001:db8::1\n"); + } + + { + struct sockaddr_in addr = + { + .sin_family = AF_INET, + .sin_addr = { .s_addr = htonl (0xc0000211) }, + .sin_port = htons (80) + }; + char host[NI_MAXHOST]; + char service[NI_MAXSERV]; + int ret = getnameinfo ((struct sockaddr *) &addr, sizeof (addr), + host, sizeof (host), service, sizeof (service), + NI_NUMERICSERV); + TEST_VERIFY (ret == 0); + TEST_VERIFY (strcmp (host, "192-0-2-17.v4.example") == 0); + TEST_VERIFY (strcmp (service, "80") == 0); + } + { + char ipv6[16] + = {0x20, 0x01, 0xd, 0xb8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + struct sockaddr_in6 addr = + { + .sin6_family = AF_INET6, + .sin6_port = htons (80), + }; + TEST_VERIFY (sizeof (ipv6) == sizeof (addr.sin6_addr)); + memcpy (&addr.sin6_addr, ipv6, sizeof (addr.sin6_addr)); + char host[NI_MAXHOST]; + char service[NI_MAXSERV]; + int ret = getnameinfo ((struct sockaddr *) &addr, sizeof (addr), + host, sizeof (host), service, sizeof (service), + NI_NUMERICSERV); + TEST_VERIFY (ret == 0); + TEST_VERIFY + (strcmp (host, "20010db8000000000000000000000001.v6.example") == 0); + TEST_VERIFY (strcmp (service, "80") == 0); + } +} + /* Test that gethostbyname2 is mostly not influenced by RES_USE_INET6. */ static void @@ -207,6 +346,8 @@ test_get2_any (void) "name: qa.example\n" "address: 2001:db8::1\n"); /* Additional AF_INET6 tests depend on RES_USE_INET6; see below. */ + + test_reverse (); } /* gethostbyname2 tests with RES_USE_INET6 disabled. */ @@ -254,6 +395,10 @@ test_no_inet6 (void) gethostbyname ("am.example"), "name: am.example\n" "address: 192.0.2.17\n"); + check_hostent ("gethostbyname (\"amp.example\")", + gethostbyname ("amp.example"), + "name: amp.example\n" + "address: 192.0.2.17\n"); check_hostent ("gethostbyname (\"qam.example\")", gethostbyname ("qam.example"), "name: qam.example\n" @@ -307,6 +452,28 @@ threadfunc (void *ignored) gethostbyname ("qm.inet6.example"), "name: qm.inet6.example\n" "address: 2001:db8::1\n"); + check_hostent ("gethostbyname (\"amp.inet6.example\")", + gethostbyname ("amp.inet6.example"), + "error: NO_RECOVERY\n"); + check_hostent ("gethostbyname (\"qmp.inet6.example\")", + gethostbyname ("qmp.inet6.example"), + "name: qmp.inet6.example\n" + "address: 2001:db8::1\n"); + check_hostent ("gethostbyname (\"ap.inet6.example\")", + gethostbyname ("ap.inet6.example"), + "error: NO_RECOVERY\n"); + check_hostent ("gethostbyname (\"6ap.inet6.example\")", + gethostbyname ("6ap.inet6.example"), + "name: 6ap.inet6.example\n" + "address: ::ffff:192.0.2.17\n"); + check_hostent ("gethostbyname (\"am6p.inet6.example\")", + gethostbyname ("am6p.inet6.example"), + "name: am6p.inet6.example\n" + "address: ::ffff:192.0.2.17\n"); + check_hostent ("gethostbyname (\"qp.inet6.example\")", + gethostbyname ("qp.inet6.example"), + "name: qp.inet6.example\n" + "address: 2001:db8::1\n"); test_get2_inet6 (); test_get2_inet6 (); test_gai (); commit 905a6129147e7ee80e8918e23efe212433b8cce7 Author: Florian Weimer Date: Wed Sep 6 15:46:54 2017 +0200 resolv: Fix memory leak with OOM during resolv.conf parsing [BZ #22095] (cherry picked from commit 5670c4ab256114e869b1df4b05653aa5f909182c) diff --git a/ChangeLog b/ChangeLog index fa215c2729..ef2041b9e5 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-09-06 Florian Weimer + + [BZ #22095] + * resolv/res_init.c (res_vinit_1): Avoid memory leak in case of + dynarray allocation failure. + 2017-09-06 Florian Weimer Remove dead PTR IPv4-to-IPv6 mapping code from nss_dns. diff --git a/NEWS b/NEWS index 8fbf4241d1..62959274b3 100644 --- a/NEWS +++ b/NEWS @@ -28,6 +28,7 @@ The following bugs are resolved with this release: [21932] Unpaired __resolv_context_get in generic get*_r implementation [21941] powerpc: Restrict xssqrtqp operands to Vector Registers [21972] assert macro requires operator== (int) for its argument type + [22095] resolv: Fix memory leak with OOM during resolv.conf parsing Version 2.26 diff --git a/resolv/res_init.c b/resolv/res_init.c index fa46ce7813..4e1f9fe8de 100644 --- a/resolv/res_init.c +++ b/resolv/res_init.c @@ -446,6 +446,11 @@ res_vinit_1 (FILE *fp, struct resolv_conf_parser *parser) (&parser->nameserver_list); if (p != NULL) *p = sa; + else + { + free (sa); + return false; + } } continue; } commit 27233446a62ca35ce0b54566279a99a6774d4210 Author: Florian Weimer Date: Wed Sep 6 15:47:27 2017 +0200 resolv: __resolv_conf_attach must not free passed conf object [BZ #22096] (cherry picked from commit a83047308196e3e54716a39dd85c0a08b198d6bd) diff --git a/ChangeLog b/ChangeLog index ef2041b9e5..e98a4bbf3c 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-09-06 Florian Weimer + + [BZ #22096] + * resolv/resolv_conf.c (__resolv_conf_attach): Do not free conf in + case of failure to obtain the global conf object. + 2017-09-06 Florian Weimer [BZ #22095] diff --git a/NEWS b/NEWS index 62959274b3..9bcb176171 100644 --- a/NEWS +++ b/NEWS @@ -29,6 +29,7 @@ The following bugs are resolved with this release: [21941] powerpc: Restrict xssqrtqp operands to Vector Registers [21972] assert macro requires operator== (int) for its argument type [22095] resolv: Fix memory leak with OOM during resolv.conf parsing + [22096] resolv: __resolv_conf_attach must not free passed conf object Version 2.26 diff --git a/resolv/resolv_conf.c b/resolv/resolv_conf.c index f391d30c27..e0f296d02e 100644 --- a/resolv/resolv_conf.c +++ b/resolv/resolv_conf.c @@ -600,10 +600,7 @@ __resolv_conf_attach (struct __res_state *resp, struct resolv_conf *conf) struct resolv_conf_global *global_copy = get_locked_global (); if (global_copy == NULL) - { - free (conf); - return false; - } + return false; /* Try to find an unused index in the array. */ size_t index; commit d265b6129184dd94da600187b67cef9125bc58c7 Author: Florian Weimer Date: Wed Sep 6 11:25:14 2017 +0200 __libc_dynarray_emplace_enlarge: Add missing else Before, arrays of small elements received a starting allocation size of 8, not 16. diff --git a/ChangeLog b/ChangeLog index e98a4bbf3c..f82fd1f97d 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-09-06 Florian Weimer + + * malloc/dynarray_emplace_enlarge.c + (__libc_dynarray_emplace_enlarge): Add missing else. + 2017-09-06 Florian Weimer [BZ #22096] diff --git a/malloc/dynarray_emplace_enlarge.c b/malloc/dynarray_emplace_enlarge.c index dfc70017ce..09cd09268b 100644 --- a/malloc/dynarray_emplace_enlarge.c +++ b/malloc/dynarray_emplace_enlarge.c @@ -32,7 +32,7 @@ __libc_dynarray_emplace_enlarge (struct dynarray_header *list, size. */ if (element_size < 4) new_allocated = 16; - if (element_size < 8) + else if (element_size < 8) new_allocated = 8; else new_allocated = 4; commit 8a1adb593969e099604537804f594efe01e04f6f Author: Florian Weimer Date: Wed Aug 30 20:10:56 2017 +0200 dynarray: Set errno on overflow-induced allocation failure This allows the caller to return directly on such an error, with an appropriate errno value. (cherry picked from commit 5898f4548efdcd7c0fd437a74eeb80facc51a117) diff --git a/ChangeLog b/ChangeLog index f82fd1f97d..87cc2f4865 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,11 @@ +2017-08-30 Florian Weimer + + * malloc/dynarray_emplace_enlarge.c + (__libc_dynarray_emplace_enlarge): Set errno on overflow. + * malloc/dynarray_resize.c (__libc_dynarray_resize): Likewise. + * malloc/tst-dynarray.c (test_long_overflow): New function. + (do_test): Call it. + 2017-09-06 Florian Weimer * malloc/dynarray_emplace_enlarge.c diff --git a/malloc/dynarray_emplace_enlarge.c b/malloc/dynarray_emplace_enlarge.c index 09cd09268b..a15245f4cb 100644 --- a/malloc/dynarray_emplace_enlarge.c +++ b/malloc/dynarray_emplace_enlarge.c @@ -17,6 +17,7 @@ . */ #include +#include #include #include #include @@ -43,8 +44,11 @@ __libc_dynarray_emplace_enlarge (struct dynarray_header *list, { new_allocated = list->allocated + list->allocated / 2 + 1; if (new_allocated <= list->allocated) - /* Overflow. */ - return false; + { + /* Overflow. */ + __set_errno (ENOMEM); + return false; + } } size_t new_size; diff --git a/malloc/dynarray_resize.c b/malloc/dynarray_resize.c index e6dc9fbc68..63c981bf61 100644 --- a/malloc/dynarray_resize.c +++ b/malloc/dynarray_resize.c @@ -17,6 +17,7 @@ . */ #include +#include #include #include #include @@ -38,7 +39,11 @@ __libc_dynarray_resize (struct dynarray_header *list, size_t size, size_t new_size_bytes; if (check_mul_overflow_size_t (size, element_size, &new_size_bytes)) - return false; + { + /* Overflow. */ + __set_errno (ENOMEM); + return false; + } void *new_array; if (list->array == scratch) { diff --git a/malloc/tst-dynarray.c b/malloc/tst-dynarray.c index 2206d75e31..d11f7bb8a3 100644 --- a/malloc/tst-dynarray.c +++ b/malloc/tst-dynarray.c @@ -18,6 +18,9 @@ #include "tst-dynarray-shared.h" +#include +#include + #define DYNARRAY_STRUCT dynarray_long #define DYNARRAY_ELEMENT long #define DYNARRAY_PREFIX dynarray_long_ @@ -463,6 +466,31 @@ test_long_init (void) } } +/* Test overflow in resize. */ +static void +test_long_overflow (void) +{ + { + struct dynarray_long dyn; + dynarray_long_init (&dyn); + errno = EINVAL; + TEST_VERIFY (!dynarray_long_resize + (&dyn, (SIZE_MAX / sizeof (long)) + 1)); + TEST_VERIFY (errno == ENOMEM); + TEST_VERIFY (dynarray_long_has_failed (&dyn)); + } + + { + struct dynarray_long_noscratch dyn; + dynarray_long_noscratch_init (&dyn); + errno = EINVAL; + TEST_VERIFY (!dynarray_long_noscratch_resize + (&dyn, (SIZE_MAX / sizeof (long)) + 1)); + TEST_VERIFY (errno == ENOMEM); + TEST_VERIFY (dynarray_long_noscratch_has_failed (&dyn)); + } +} + /* Test NUL-terminated string construction with the add function and the simple finalize function. */ static void @@ -538,6 +566,7 @@ do_test (void) test_int (); test_str (); test_long_init (); + test_long_overflow (); test_zstr (); return 0; } commit dca8b177f6bb521638f4d8a845b4cb10348137c9 Author: H.J. Lu Date: Thu Aug 31 06:28:31 2017 -0700 Place $(elf-objpfx)sofini.os last [BZ #22051] Since sofini.os terminates .eh_frame section, it should be placed last. [BZ #22051] * Makerules (build-module-helper-objlist): Filter out $(elf-objpfx)sofini.os. (build-shlib-objlist): Append $(elf-objpfx)sofini.os if it is needed. (cherry picked from commit ecd0747df388f3925d4839740b0abcf43d0a9fb2) diff --git a/ChangeLog b/ChangeLog index 87cc2f4865..f9e423dbb3 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,11 @@ +2017-08-31 H.J. Lu + + [BZ #22051] + * Makerules (build-module-helper-objlist): Filter out + $(elf-objpfx)sofini.os. + (build-shlib-objlist): Append $(elf-objpfx)sofini.os if it is + needed. + 2017-08-30 Florian Weimer * malloc/dynarray_emplace_enlarge.c diff --git a/Makerules b/Makerules index 9bb707c168..828a445f24 100644 --- a/Makerules +++ b/Makerules @@ -686,14 +686,17 @@ $(build-module-helper) -o $@ $(shlib-lds-flags) \ $(call after-link,$@) endef +# sofini.os must be placed last since it terminates .eh_frame section. build-module-helper-objlist = \ $(patsubst %_pic.a,$(whole-archive) %_pic.a $(no-whole-archive),\ $(filter-out %.lds $(map-file) $(+preinit) $(+postinit) \ + $(elf-objpfx)sofini.os \ $(link-libc-deps),$^)) build-module-objlist = $(build-module-helper-objlist) $(LDLIBS-$(@F:%.so=%).so) build-shlib-objlist = $(build-module-helper-objlist) \ - $(LDLIBS-$(@F:lib%.so=%).so) + $(LDLIBS-$(@F:lib%.so=%).so) \ + $(filter $(elf-objpfx)sofini.os,$^) # Don't try to use -lc when making libc.so itself. # Also omits crti.o and crtn.o, which we do not want commit 86553be84d071aee8464b2c2fa03858d12118852 Author: H.J. Lu Date: Thu Sep 7 13:56:58 2017 -0700 Use "static const char domain[] =" * resolv/tst-resolv-qtypes.c (domain): Changed to "const char domain[] =". (cherry picked from commit 78bfa877b36e8f33c99cbe9a16eb73f5a2adc0c8) diff --git a/ChangeLog b/ChangeLog index f9e423dbb3..96c461c441 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-09-07 H.J. Lu + + * resolv/tst-resolv-qtypes.c (domain): Changed to + "const char domain[] =". + 2017-08-31 H.J. Lu [BZ #22051] diff --git a/resolv/tst-resolv-qtypes.c b/resolv/tst-resolv-qtypes.c index 06ea3dbd14..da3325f80c 100644 --- a/resolv/tst-resolv-qtypes.c +++ b/resolv/tst-resolv-qtypes.c @@ -50,7 +50,7 @@ response (const struct resolv_response_context *ctx, resolv_response_close_record (b); } -static const char * const domain = "www.example.com"; +static const char domain[] = "www.example.com"; static int wrap_res_query (int type, unsigned char *answer, int answer_length) commit 5f5532caf820f47f714d09766082f1e070c12cfb Author: Markus Trippelsdorf Date: Fri Sep 8 19:57:12 2017 +0000 Update x86_64 ulps for AMD Ryzen. * sysdeps/x86_64/fpu/libm-test-ulps: Update for AMD Ryzen. (cherry picked from commit 4c03a696800e3bb4b433626a65c4fef465dcc98b) diff --git a/ChangeLog b/ChangeLog index 96c461c441..c180f8dbca 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,7 @@ +2017-09-08 Markus Trippelsdorf + + * sysdeps/x86_64/fpu/libm-test-ulps: Update for AMD Ryzen. + 2017-09-07 H.J. Lu * resolv/tst-resolv-qtypes.c (domain): Changed to diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps index c347642044..1e28aaa9f1 100644 --- a/sysdeps/x86_64/fpu/libm-test-ulps +++ b/sysdeps/x86_64/fpu/libm-test-ulps @@ -1733,10 +1733,10 @@ ldouble: 3 Function: Imaginary part of "ctan_upward": double: 2 -float: 1 +float: 2 float128: 5 idouble: 2 -ifloat: 1 +ifloat: 2 ifloat128: 5 ildouble: 3 ldouble: 3 commit 56ce01906ecb0a3e04411f8ceb60b27f4877f070 Author: H.J. Lu Date: Wed Aug 23 08:22:52 2017 -0700 string/stratcliff.c: Replace int with size_t [BZ #21982] Fix GCC 7 errors when string/stratcliff.c is compiled with -O3: stratcliff.c: In function ‘do_test’: cc1: error: assuming signed overflow does not occur when assuming that (X - c) <= X is always true [-Werror=strict-overflow] [BZ #21982] * string/stratcliff.c (do_test): Declare size, nchars, inner, middle and outer with size_t instead of int. Repleace %d and %Zd with %zu in printf. Update "MAX (0, nchars - 128)" and "MAX (outer, nchars - 64)" to support unsigned outer and nchars. Also exit loop when outer == 0. (cherry picked from commit 376b40a27a6783ea7f73ae577be320df66aeb36d) diff --git a/ChangeLog b/ChangeLog index c180f8dbca..84721f6265 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,12 @@ +2017-09-11 H.J. Lu + + [BZ #21982] + * string/stratcliff.c (do_test): Declare size, nchars, inner, + middle and outer with size_t instead of int. Repleace %d and + %Zd with %zu in printf. Update "MAX (0, nchars - 128)" and + "MAX (outer, nchars - 64)" to support unsigned outer and + nchars. Also exit loop when outer == 0. + 2017-09-08 Markus Trippelsdorf * sysdeps/x86_64/fpu/libm-test-ulps: Update for AMD Ryzen. diff --git a/string/stratcliff.c b/string/stratcliff.c index e28b0c5058..4320336c9a 100644 --- a/string/stratcliff.c +++ b/string/stratcliff.c @@ -58,8 +58,8 @@ int do_test (void) { - int size = sysconf (_SC_PAGESIZE); - int nchars = size / sizeof (CHAR); + size_t size = sysconf (_SC_PAGESIZE); + size_t nchars = size / sizeof (CHAR); CHAR *adr; CHAR *dest; int result = 0; @@ -80,7 +80,17 @@ do_test (void) } else { - int inner, middle, outer; + size_t inner, middle, outer, nchars64, max128; + + if (nchars > 64) + nchars64 = nchars - 64; + else + nchars64 = 0; + + if (nchars > 128) + max128 = nchars - 128; + else + max128 = 0; mprotect (adr, size, PROT_NONE); mprotect (adr + 2 * nchars, size, PROT_NONE); @@ -93,59 +103,65 @@ do_test (void) MEMSET (adr, L('T'), nchars); /* strlen/wcslen test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (inner = MAX (outer, nchars - 64); inner < nchars; ++inner) + for (inner = MAX (outer, nchars64); inner < nchars; ++inner) { adr[inner] = L('\0'); if (STRLEN (&adr[outer]) != (size_t) (inner - outer)) { - printf ("%s flunked for outer = %d, inner = %d\n", + printf ("%s flunked for outer = %zu, inner = %zu\n", STRINGIFY (STRLEN), outer, inner); result = 1; } adr[inner] = L('T'); } + if (outer == 0) + break; } /* strnlen/wcsnlen test */ - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars; outer >= max128; --outer) { - for (inner = MAX (outer, nchars - 64); inner < nchars; ++inner) + for (inner = MAX (outer, nchars64); inner < nchars; ++inner) { adr[inner] = L('\0'); if (STRNLEN (&adr[outer], inner - outer + 1) != (size_t) (inner - outer)) { - printf ("%s flunked for outer = %d, inner = %d\n", + printf ("%s flunked for outer = %zu, inner = %zu\n", STRINGIFY (STRNLEN), outer, inner); result = 1; } adr[inner] = L('T'); } + if (outer == 0) + break; } - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars; outer >= max128; --outer) { - for (inner = MAX (outer, nchars - 64); inner <= nchars; ++inner) + for (inner = MAX (outer, nchars64); inner <= nchars; ++inner) { if (STRNLEN (&adr[outer], inner - outer) != (size_t) (inner - outer)) { - printf ("%s flunked bounded for outer = %d, inner = %d\n", + printf ("%s flunked bounded for outer = %zu, inner = %zu\n", STRINGIFY (STRNLEN), outer, inner); result = 1; } } + if (outer == 0) + break; } /* strchr/wcschr test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (middle = MAX (outer, nchars - 64); middle < nchars; ++middle) + for (middle = MAX (outer, nchars64); middle < nchars; ++middle) { for (inner = middle; inner < nchars; ++inner) { @@ -158,8 +174,8 @@ do_test (void) || (inner != middle && (cp - &adr[outer]) != middle - outer)) { - printf ("%s flunked for outer = %d, middle = %d, " - "inner = %d\n", + printf ("%s flunked for outer = %zu, middle = %zu, " + "inner = %zu\n", STRINGIFY (STRCHR), outer, middle, inner); result = 1; } @@ -168,6 +184,8 @@ do_test (void) adr[middle] = L('T'); } } + if (outer == 0) + break; } /* Special test. */ @@ -180,9 +198,9 @@ do_test (void) } /* strrchr/wcsrchr test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (middle = MAX (outer, nchars - 64); middle < nchars; ++middle) + for (middle = MAX (outer, nchars64); middle < nchars; ++middle) { for (inner = middle; inner < nchars; ++inner) { @@ -195,8 +213,8 @@ do_test (void) || (inner != middle && (cp - &adr[outer]) != middle - outer)) { - printf ("%s flunked for outer = %d, middle = %d, " - "inner = %d\n", + printf ("%s flunked for outer = %zu, middle = %zu, " + "inner = %zu\n", STRINGIFY (STRRCHR), outer, middle, inner); result = 1; } @@ -205,12 +223,14 @@ do_test (void) adr[middle] = L('T'); } } + if (outer == 0) + break; } /* memchr test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (middle = MAX (outer, nchars - 64); middle < nchars; ++middle) + for (middle = MAX (outer, nchars64); middle < nchars; ++middle) { adr[middle] = L('V'); @@ -218,32 +238,36 @@ do_test (void) if (cp - &adr[outer] != middle - outer) { - printf ("%s flunked for outer = %d, middle = %d\n", + printf ("%s flunked for outer = %zu, middle = %zu\n", STRINGIFY (MEMCHR), outer, middle); result = 1; } adr[middle] = L('T'); } + if (outer == 0) + break; } - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars; outer >= max128; --outer) { CHAR *cp = MEMCHR (&adr[outer], L('V'), nchars - outer); if (cp != NULL) { - printf ("%s flunked for outer = %d\n", + printf ("%s flunked for outer = %zu\n", STRINGIFY (MEMCHR), outer); result = 1; } + if (outer == 0) + break; } /* These functions only exist for single-byte characters. */ #ifndef WCSTEST /* rawmemchr test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (middle = MAX (outer, nchars - 64); middle < nchars; ++middle) + for (middle = MAX (outer, nchars64); middle < nchars; ++middle) { adr[middle] = L('V'); @@ -251,19 +275,21 @@ do_test (void) if (cp - &adr[outer] != middle - outer) { - printf ("%s flunked for outer = %d, middle = %d\n", + printf ("%s flunked for outer = %zu, middle = %zu\n", STRINGIFY (rawmemchr), outer, middle); result = 1; } adr[middle] = L('T'); } + if (outer == 0) + break; } /* memrchr test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (middle = MAX (outer, nchars - 64); middle < nchars; ++middle) + for (middle = MAX (outer, nchars64); middle < nchars; ++middle) { adr[middle] = L('V'); @@ -271,44 +297,50 @@ do_test (void) if (cp - &adr[outer] != middle - outer) { - printf ("%s flunked for outer = %d, middle = %d\n", + printf ("%s flunked for outer = %zu, middle = %zu\n", STRINGIFY (memrchr), outer, middle); result = 1; } adr[middle] = L('T'); } + if (outer == 0) + break; } - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars; outer >= max128; --outer) { CHAR *cp = memrchr (&adr[outer], L('V'), nchars - outer); if (cp != NULL) { - printf ("%s flunked for outer = %d\n", + printf ("%s flunked for outer = %zu\n", STRINGIFY (memrchr), outer); result = 1; } + if (outer == 0) + break; } #endif /* strcpy/wcscpy test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (inner = MAX (outer, nchars - 64); inner < nchars; ++inner) + for (inner = MAX (outer, nchars64); inner < nchars; ++inner) { adr[inner] = L('\0'); if (STRCPY (dest, &adr[outer]) != dest || STRLEN (dest) != (size_t) (inner - outer)) { - printf ("%s flunked for outer = %d, inner = %d\n", + printf ("%s flunked for outer = %zu, inner = %zu\n", STRINGIFY (STRCPY), outer, inner); result = 1; } adr[inner] = L('T'); } + if (outer == 0) + break; } /* strcmp/wcscmp tests */ @@ -322,14 +354,14 @@ do_test (void) if (STRCMP (adr + middle, dest + nchars - outer) <= 0) { - printf ("%s 1 flunked for outer = %d, middle = %d\n", + printf ("%s 1 flunked for outer = %zu, middle = %zu\n", STRINGIFY (STRCMP), outer, middle); result = 1; } if (STRCMP (dest + nchars - outer, adr + middle) >= 0) { - printf ("%s 2 flunked for outer = %d, middle = %d\n", + printf ("%s 2 flunked for outer = %zu, middle = %zu\n", STRINGIFY (STRCMP), outer, middle); result = 1; } @@ -348,16 +380,16 @@ do_test (void) { if (STRNCMP (adr + middle, dest + nchars - outer, inner) != 0) { - printf ("%s 1 flunked for outer = %d, middle = %d, " - "inner = %d\n", + printf ("%s 1 flunked for outer = %zu, middle = %zu, " + "inner = %zu\n", STRINGIFY (STRNCMP), outer, middle, inner); result = 1; } if (STRNCMP (dest + nchars - outer, adr + middle, inner) != 0) { - printf ("%s 2 flunked for outer = %d, middle = %d, " - "inner = %d\n", + printf ("%s 2 flunked for outer = %zu, middle = %zu, " + "inner = %zu\n", STRINGIFY (STRNCMP), outer, middle, inner); result = 1; } @@ -365,14 +397,14 @@ do_test (void) if (STRNCMP (adr + middle, dest + nchars - outer, outer) >= 0) { - printf ("%s 1 flunked for outer = %d, middle = %d, full\n", + printf ("%s 1 flunked for outer = %zu, middle = %zu, full\n", STRINGIFY (STRNCMP), outer, middle); result = 1; } if (STRNCMP (dest + nchars - outer, adr + middle, outer) <= 0) { - printf ("%s 2 flunked for outer = %d, middle = %d, full\n", + printf ("%s 2 flunked for outer = %zu, middle = %zu, full\n", STRINGIFY (STRNCMP), outer, middle); result = 1; } @@ -380,7 +412,7 @@ do_test (void) /* strncpy/wcsncpy tests */ adr[nchars - 1] = L('T'); - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars; outer >= max128; --outer) { size_t len; @@ -389,17 +421,19 @@ do_test (void) if (STRNCPY (dest, &adr[outer], len) != dest || MEMCMP (dest, &adr[outer], len) != 0) { - printf ("outer %s flunked for outer = %d, len = %Zd\n", + printf ("outer %s flunked for outer = %zu, len = %zu\n", STRINGIFY (STRNCPY), outer, len); result = 1; } } + if (outer == 0) + break; } adr[nchars - 1] = L('\0'); - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (inner = MAX (outer, nchars - 64); inner < nchars; ++inner) + for (inner = MAX (outer, nchars64); inner < nchars; ++inner) { size_t len; @@ -413,8 +447,8 @@ do_test (void) || (inner - outer < len && STRLEN (dest) != (inner - outer))) { - printf ("%s flunked for outer = %d, inner = %d, " - "len = %Zd\n", + printf ("%s flunked for outer = %zu, inner = %zu, " + "len = %zu\n", STRINGIFY (STRNCPY), outer, inner, len); result = 1; } @@ -424,8 +458,8 @@ do_test (void) || (inner - outer < len && STRLEN (dest + 1) != (inner - outer))) { - printf ("%s+1 flunked for outer = %d, inner = %d, " - "len = %Zd\n", + printf ("%s+1 flunked for outer = %zu, inner = %zu, " + "len = %zu\n", STRINGIFY (STRNCPY), outer, inner, len); result = 1; } @@ -433,29 +467,33 @@ do_test (void) adr[inner] = L('T'); } + if (outer == 0) + break; } /* stpcpy/wcpcpy test */ - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (inner = MAX (outer, nchars - 64); inner < nchars; ++inner) + for (inner = MAX (outer, nchars64); inner < nchars; ++inner) { adr[inner] = L('\0'); if ((STPCPY (dest, &adr[outer]) - dest) != inner - outer) { - printf ("%s flunked for outer = %d, inner = %d\n", + printf ("%s flunked for outer = %zu, inner = %zu\n", STRINGIFY (STPCPY), outer, inner); result = 1; } adr[inner] = L('T'); } + if (outer == 0) + break; } /* stpncpy/wcpncpy test */ adr[nchars - 1] = L('T'); - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars; outer >= max128; --outer) { size_t len; @@ -464,17 +502,19 @@ do_test (void) if (STPNCPY (dest, &adr[outer], len) != dest + len || MEMCMP (dest, &adr[outer], len) != 0) { - printf ("outer %s flunked for outer = %d, len = %Zd\n", + printf ("outer %s flunked for outer = %zu, len = %zu\n", STRINGIFY (STPNCPY), outer, len); result = 1; } } + if (outer == 0) + break; } adr[nchars - 1] = L('\0'); - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) + for (outer = nchars - 1; outer >= max128; --outer) { - for (middle = MAX (outer, nchars - 64); middle < nchars; ++middle) + for (middle = MAX (outer, nchars64); middle < nchars; ++middle) { adr[middle] = L('\0'); @@ -483,8 +523,8 @@ do_test (void) if ((STPNCPY (dest, &adr[outer], inner) - dest) != MIN (inner, middle - outer)) { - printf ("%s flunked for outer = %d, middle = %d, " - "inner = %d\n", + printf ("%s flunked for outer = %zu, middle = %zu, " + "inner = %zu\n", STRINGIFY (STPNCPY), outer, middle, inner); result = 1; } @@ -492,66 +532,84 @@ do_test (void) adr[middle] = L('T'); } + if (outer == 0) + break; } /* memcpy/wmemcpy test */ - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) - for (inner = 0; inner < nchars - outer; ++inner) - if (MEMCPY (dest, &adr[outer], inner) != dest) - { - printf ("%s flunked for outer = %d, inner = %d\n", - STRINGIFY (MEMCPY), outer, inner); - result = 1; - } + for (outer = nchars; outer >= max128; --outer) + { + for (inner = 0; inner < nchars - outer; ++inner) + if (MEMCPY (dest, &adr[outer], inner) != dest) + { + printf ("%s flunked for outer = %zu, inner = %zu\n", + STRINGIFY (MEMCPY), outer, inner); + result = 1; + } + if (outer == 0) + break; + } /* mempcpy/wmempcpy test */ - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) - for (inner = 0; inner < nchars - outer; ++inner) - if (MEMPCPY (dest, &adr[outer], inner) != dest + inner) - { - printf ("%s flunked for outer = %d, inner = %d\n", - STRINGIFY (MEMPCPY), outer, inner); - result = 1; - } + for (outer = nchars; outer >= max128; --outer) + { + for (inner = 0; inner < nchars - outer; ++inner) + if (MEMPCPY (dest, &adr[outer], inner) != dest + inner) + { + printf ("%s flunked for outer = %zu, inner = %zu\n", + STRINGIFY (MEMPCPY), outer, inner); + result = 1; + } + if (outer == 0) + break; + } /* This function only exists for single-byte characters. */ #ifndef WCSTEST /* memccpy test */ memset (adr, '\0', nchars); - for (outer = nchars; outer >= MAX (0, nchars - 128); --outer) - for (inner = 0; inner < nchars - outer; ++inner) - if (memccpy (dest, &adr[outer], L('\1'), inner) != NULL) - { - printf ("memccpy flunked full copy for outer = %d, inner = %d\n", - outer, inner); - result = 1; - } - for (outer = nchars - 1; outer >= MAX (0, nchars - 128); --outer) - for (middle = 0; middle < nchars - outer; ++middle) - { - memset (dest, L('\2'), middle + 1); - for (inner = 0; inner < middle; ++inner) + for (outer = nchars; outer >= max128; --outer) + { + for (inner = 0; inner < nchars - outer; ++inner) + if (memccpy (dest, &adr[outer], L('\1'), inner) != NULL) { - adr[outer + inner] = L('\1'); - - if (memccpy (dest, &adr[outer], '\1', middle + 128) - != dest + inner + 1) - { - printf ("\ -memccpy flunked partial copy for outer = %d, middle = %d, inner = %d\n", - outer, middle, inner); - result = 1; - } - else if (dest[inner + 1] != L('\2')) - { - printf ("\ -memccpy copied too much for outer = %d, middle = %d, inner = %d\n", - outer, middle, inner); - result = 1; - } - adr[outer + inner] = L('\0'); + printf ("memccpy flunked full copy for outer = %zu, inner = %zu\n", + outer, inner); + result = 1; } - } + if (outer == 0) + break; + } + for (outer = nchars - 1; outer >= max128; --outer) + { + for (middle = 0; middle < nchars - outer; ++middle) + { + memset (dest, L('\2'), middle + 1); + for (inner = 0; inner < middle; ++inner) + { + adr[outer + inner] = L('\1'); + + if (memccpy (dest, &adr[outer], '\1', middle + 128) + != dest + inner + 1) + { + printf ("\ + memccpy flunked partial copy for outer = %zu, middle = %zu, inner = %zu\n", + outer, middle, inner); + result = 1; + } + else if (dest[inner + 1] != L('\2')) + { + printf ("\ + memccpy copied too much for outer = %zu, middle = %zu, inner = %zu\n", + outer, middle, inner); + result = 1; + } + adr[outer + inner] = L('\0'); + } + } + if (outer == 0) + break; + } #endif } commit 2422c6032fd5e95b1c29df5b8671455ead51314e Author: Alan Modra Date: Thu Aug 3 15:39:21 2017 +0930 tst-tlsopt-powerpc as a shared lib This makes the __tls_get_addr_opt test run as a shared library, and so actually test that DTPMOD64/DTPREL64 pairs are processed by ld.so to support the __tls_get_adfr_opt call stub fast return. After a 2017-01-24 patch (binutils f0158f4416) ld.bfd no longer emitted unnecessary dynamic relocations against local thread variables, instead setting up the __tls_index GOT entries for the call stub fast return. This meant tst-tlsopt-powerpc passed but did not check ld.so relocation support. After a 2017-07-16 patch (binutils 676ee2b5fa) ld.bfd no longer set up the __tls_index GOT entries for the call stub fast return, and tst-tlsopt-powerpc failed. Compiling mod-tlsopt-powerpc.c with -DSHARED exposed a bug in powerpc64/tls-macros.h, which defines a __TLS_GET_ADDR macro that clashes with one defined in dl-tls.h. The tls-macros.h version is only used in that file, so delete it and expand. * sysdeps/powerpc/mod-tlsopt-powerpc.c: Extract from tst-tlsopt-powerpc.c with function name change and no test harness. * sysdeps/powerpc/tst-tlsopt-powerpc.c: Remove body of test. Call tls_get_addr_opt_test. * sysdeps/powerpc/Makefile (LDFLAGS-tst-tlsopt-powerpc): Don't define. (modules-names): Add mod-tlsopt-powerpc. (mod-tlsopt-powerpc.so-no-z-defs): Define. (tst-tlsopt-powerpc): Depend on .so. * sysdeps/powerpc/powerpc64/tls-macros.h (__TLS_GET_ADDR): Don't define. Expand use in TLS_GD and TLS_LD. (cherry picked from commit e98c925fa4f6486447eb20676755dbf9eb36a110) diff --git a/ChangeLog b/ChangeLog index 84721f6265..b53d8830a2 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,16 @@ +2017-08-03 Alan Modra + + * sysdeps/powerpc/mod-tlsopt-powerpc.c: Extract from + tst-tlsopt-powerpc.c with function name change and no test harness. + * sysdeps/powerpc/tst-tlsopt-powerpc.c: Remove body of test. + Call tls_get_addr_opt_test. + * sysdeps/powerpc/Makefile (LDFLAGS-tst-tlsopt-powerpc): Don't define. + (modules-names): Add mod-tlsopt-powerpc. + (mod-tlsopt-powerpc.so-no-z-defs): Define. + (tst-tlsopt-powerpc): Depend on .so. + * sysdeps/powerpc/powerpc64/tls-macros.h (__TLS_GET_ADDR): Don't + define. Expand use in TLS_GD and TLS_LD. + 2017-09-11 H.J. Lu [BZ #21982] diff --git a/sysdeps/powerpc/Makefile b/sysdeps/powerpc/Makefile index 0d9206bec4..6aa683b03f 100644 --- a/sysdeps/powerpc/Makefile +++ b/sysdeps/powerpc/Makefile @@ -8,9 +8,11 @@ sysdep-dl-routines += dl-machine hwcapinfo sysdep_routines += dl-machine hwcapinfo # extra shared linker files to link only into dl-allobjs.so sysdep-rtld-routines += dl-machine hwcapinfo -# Don't optimize GD tls sequence to LE. -LDFLAGS-tst-tlsopt-powerpc += -Wl,--no-tls-optimize + +modules-names += mod-tlsopt-powerpc +mod-tlsopt-powerpc.so-no-z-defs = yes tests += tst-tlsopt-powerpc +$(objpfx)tst-tlsopt-powerpc: $(objpfx)mod-tlsopt-powerpc.so ifneq (no,$(multi-arch)) tests-static += tst-tlsifunc-static diff --git a/sysdeps/powerpc/mod-tlsopt-powerpc.c b/sysdeps/powerpc/mod-tlsopt-powerpc.c new file mode 100644 index 0000000000..ee0db12a73 --- /dev/null +++ b/sysdeps/powerpc/mod-tlsopt-powerpc.c @@ -0,0 +1,49 @@ +/* shared library to test for __tls_get_addr optimization. */ +#include + +#include "../../elf/tls-macros.h" +#include "dl-tls.h" + +/* common 'int' variable in TLS. */ +COMMON_INT_DEF(foo); + + +int +tls_get_addr_opt_test (void) +{ + int result = 0; + + /* Get variable using general dynamic model. */ + int *ap = TLS_GD (foo); + if (*ap != 0) + { + printf ("foo = %d\n", *ap); + result = 1; + } + + tls_index *tls_arg; +#ifdef __powerpc64__ + register unsigned long thread_pointer __asm__ ("r13"); + asm ("addi %0,2,foo@got@tlsgd" : "=r" (tls_arg)); +#else + register unsigned long thread_pointer __asm__ ("r2"); + asm ("bcl 20,31,1f\n1:\t" + "mflr %0\n\t" + "addis %0,%0,_GLOBAL_OFFSET_TABLE_-1b@ha\n\t" + "addi %0,%0,_GLOBAL_OFFSET_TABLE_-1b@l\n\t" + "addi %0,%0,foo@got@tlsgd" : "=b" (tls_arg)); +#endif + + if (tls_arg->ti_module != 0) + { + printf ("tls_index not optimized, binutils too old?\n"); + result = 1; + } + else if (tls_arg->ti_offset + thread_pointer != (unsigned long) ap) + { + printf ("tls_index->ti_offset wrong value\n"); + result = 1; + } + + return result; +} diff --git a/sysdeps/powerpc/powerpc64/tls-macros.h b/sysdeps/powerpc/powerpc64/tls-macros.h index 42a95ec5c1..79a0b2579c 100644 --- a/sysdeps/powerpc/powerpc64/tls-macros.h +++ b/sysdeps/powerpc/powerpc64/tls-macros.h @@ -18,13 +18,11 @@ __result; \ }) -#define __TLS_GET_ADDR "__tls_get_addr" - /* PowerPC64 Local Dynamic TLS access. */ #define TLS_LD(x) \ ({ int * __result; \ asm ("addi 3,2," #x "@got@tlsld\n\t" \ - "bl " __TLS_GET_ADDR "\n\t" \ + "bl __tls_get_addr\n\t" \ "nop \n\t" \ "addis %0,3," #x "@dtprel@ha\n\t" \ "addi %0,%0," #x "@dtprel@l" \ @@ -36,7 +34,7 @@ #define TLS_GD(x) \ ({ register int *__result __asm__ ("r3"); \ asm ("addi 3,2," #x "@got@tlsgd\n\t" \ - "bl " __TLS_GET_ADDR "\n\t" \ + "bl __tls_get_addr\n\t" \ "nop " \ : "=r" (__result) : \ : __TLS_CALL_CLOBBERS); \ diff --git a/sysdeps/powerpc/tst-tlsopt-powerpc.c b/sysdeps/powerpc/tst-tlsopt-powerpc.c index 8ae928a3f4..cc682b2ed0 100644 --- a/sysdeps/powerpc/tst-tlsopt-powerpc.c +++ b/sysdeps/powerpc/tst-tlsopt-powerpc.c @@ -1,51 +1,11 @@ /* glibc test for __tls_get_addr optimization. */ -#include - -#include "../../elf/tls-macros.h" -#include "dl-tls.h" - -/* common 'int' variable in TLS. */ -COMMON_INT_DEF(foo); - static int do_test (void) { - int result = 0; - - /* Get variable using general dynamic model. */ - int *ap = TLS_GD (foo); - if (*ap != 0) - { - printf ("foo = %d\n", *ap); - result = 1; - } - - tls_index *tls_arg; -#ifdef __powerpc64__ - register unsigned long thread_pointer __asm__ ("r13"); - asm ("addi %0,2,foo@got@tlsgd" : "=r" (tls_arg)); -#else - register unsigned long thread_pointer __asm__ ("r2"); - asm ("bcl 20,31,1f\n1:\t" - "mflr %0\n\t" - "addis %0,%0,_GLOBAL_OFFSET_TABLE_-1b@ha\n\t" - "addi %0,%0,_GLOBAL_OFFSET_TABLE_-1b@l\n\t" - "addi %0,%0,foo@got@tlsgd" : "=b" (tls_arg)); -#endif - - if (tls_arg->ti_module != 0) - { - printf ("tls_index not optimized, binutils too old?\n"); - result = 1; - } - else if (tls_arg->ti_offset + thread_pointer != (unsigned long) ap) - { - printf ("tls_index->ti_offset wrong value\n"); - result = 1; - } + extern int tls_get_addr_opt_test (void); - return result; + return tls_get_addr_opt_test (); } #include commit ac6113cb0146494af7ecacbbdc6957ce501a40ec Author: Martin Sebor Date: Tue Aug 22 09:35:23 2017 -0600 Declare ifunc resolver to return a pointer to the same type as the target function to help GCC detect incompatibilities between the two when it's enhanced to do so. (cherry picked from commit ee4e992ebe5f9712faedeefe8958b67d61eaa0f2) diff --git a/ChangeLog b/ChangeLog index b53d8830a2..ec9b9f5edb 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-08-22 Martin Sebor + + * include/libc-symbols.h (__ifunc_resolver): Declare resolver + to return a pointer to the same type as the target function. + 2017-08-03 Alan Modra * sysdeps/powerpc/mod-tlsopt-powerpc.c: Extract from diff --git a/include/libc-symbols.h b/include/libc-symbols.h index 3310e3a678..c50f9adec2 100644 --- a/include/libc-symbols.h +++ b/include/libc-symbols.h @@ -782,7 +782,8 @@ for linking") /* Helper / base macros for indirect function symbols. */ #define __ifunc_resolver(type_name, name, expr, arg, init, classifier) \ - classifier inhibit_stack_protector void *name##_ifunc (arg) \ + classifier inhibit_stack_protector \ + __typeof (type_name) *name##_ifunc (arg) \ { \ init (); \ __typeof (type_name) *res = expr; \ commit 37d4262a7a35886cf8ac856457bbad8c0498c8d6 Author: Gabriel F. T. Gomes Date: Tue Aug 22 14:01:07 2017 -0300 Fix remaining return type of ifunc resolver declaration Since Martin Sebor's commit commit ee4e992ebe5f9712faedeefe8958b67d61eaa0f2 Author: Martin Sebor Date: Tue Aug 22 09:35:23 2017 -0600 Declare ifunc resolver to return a pointer to the same type as the target function to help GCC detect incompatibilities between the two when it's enhanced to do so. builds for powerpc64le fail in the declaration of some ifunc resolvers, because the ifunc is declared with unmatching return types. One of the declarations comes from the __ifunc_resolver macro, which was patched by the aforementioned commit: /* Helper / base macros for indirect function symbols. */ #define __ifunc_resolver(type_name, name, expr, arg, init, classifier) \ classifier inhibit_stack_protector \ __typeof (type_name) *name##_ifunc (arg) \ whereas the other comes from the unpatched __ifunc macro when HAVE_GCC_IFUNC is not defined: # define __ifunc(type_name, name, expr, arg, init) \ extern __typeof (type_name) name; \ void *name##_ifunc (arg) __asm__ (#name); \ This patch changes the return type of the ifunc resolver in the __ifunc macro, so that it matches the return type of the target function, similarly to what the aforementioned commit does. Tested for powerpc64le and s390x with unpatched GCC. * include/libc-symbols.h: [!defined HAVE_GCC_IFUNC] (__ifunc): Change the return type of the ifunc resolver to match the return type of the target function. (cherry picked from commit b513da7e80febbbfb8e58282075018652b6f7273) diff --git a/ChangeLog b/ChangeLog index ec9b9f5edb..6b09c61d5a 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-08-22 Gabriel F. T. Gomes + + * include/libc-symbols.h: [!defined HAVE_GCC_IFUNC] (__ifunc): + Change the return type of the ifunc resolver to match the return + type of the target function. + 2017-08-22 Martin Sebor * include/libc-symbols.h (__ifunc_resolver): Declare resolver diff --git a/include/libc-symbols.h b/include/libc-symbols.h index c50f9adec2..5bf57703a9 100644 --- a/include/libc-symbols.h +++ b/include/libc-symbols.h @@ -810,7 +810,7 @@ for linking") # define __ifunc(type_name, name, expr, arg, init) \ extern __typeof (type_name) name; \ - void *name##_ifunc (arg) __asm__ (#name); \ + __typeof (type_name) *name##_ifunc (arg) __asm__ (#name); \ __ifunc_resolver (type_name, name, expr, arg, init,) \ __asm__ (".type " #name ", %gnu_indirect_function"); commit d37c951fde57e8acb320a9a7d437ba50a1fc3c8a Author: Gabriel F. T. Gomes Date: Wed Sep 20 15:10:26 2017 -0300 Let fpclassify use the builtin when optimizing for size in C++ mode (bug 22146) When optimization for size is on (-Os), fpclassify does not use the type-generic __builtin_fpclassify builtin, instead it uses __MATH_TG. However, when library support for float128 is available, __MATH_TG uses __builtin_types_compatible_p, which is not available in C++ mode. On the other hand, libstdc++ undefines (in cmath) many macros from math.h, including fpclassify, so that it can provide its own functions. However, during its configure tests, libstdc++ just tests for the availability of the macros (it does not undefine them, nor does it provide its own functions). Finally, when libstdc++ is configured with optimization for size enabled, its configure tests include math.h and get the definition of fpclassify that uses __MATH_TG (and __builtin_types_compatible_p). Since libstdc++ does not undefine the macros during its configure tests, they fail. This patch lets fpclassify use the builtin in C++ mode, even when optimization for size is on. This allows the configure test in libstdc++ to work. Tested for powerpc64le and x86_64. [BZ #22146] math/math.h: Let fpclassify use the builtin in C++ mode, even when optimazing for size. (cherry picked from commit c5c4a626098ec884b8527356abdf2a4bb7b6bf27) diff --git a/ChangeLog b/ChangeLog index 6b09c61d5a..382674d5a3 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,9 @@ +2017-09-22 Gabriel F. T. Gomes + + [BZ #22146] + math/math.h: Let fpclassify use the builtin in C++ mode, even + when optimazing for size. + 2017-08-22 Gabriel F. T. Gomes * include/libc-symbols.h: [!defined HAVE_GCC_IFUNC] (__ifunc): diff --git a/NEWS b/NEWS index 9bcb176171..48e2064380 100644 --- a/NEWS +++ b/NEWS @@ -30,6 +30,7 @@ The following bugs are resolved with this release: [21972] assert macro requires operator== (int) for its argument type [22095] resolv: Fix memory leak with OOM during resolv.conf parsing [22096] resolv: __resolv_conf_attach must not free passed conf object + [22146] Let fpclassify use the builtin when optimizing for size in C++ mode Version 2.26 diff --git a/math/math.h b/math/math.h index 7c0fc6dbb3..f9348ec3ea 100644 --- a/math/math.h +++ b/math/math.h @@ -402,7 +402,13 @@ enum /* Return number of classification appropriate for X. */ # if __GNUC_PREREQ (4,4) && !defined __SUPPORT_SNAN__ \ - && !defined __OPTIMIZE_SIZE__ + && (!defined __OPTIMIZE_SIZE__ || defined __cplusplus) + /* The check for __cplusplus allows the use of the builtin, even + when optimization for size is on. This is provided for + libstdc++, only to let its configure test work when it is built + with -Os. No further use of this definition of fpclassify is + expected in C++ mode, since libstdc++ provides its own version + of fpclassify in cmath (which undefines fpclassify). */ # define fpclassify(x) __builtin_fpclassify (FP_NAN, FP_INFINITE, \ FP_NORMAL, FP_SUBNORMAL, FP_ZERO, x) # else commit 548cc83c38a91852b1e44045ead3d20ccd5db4cf Author: Joseph Myers Date: Thu Sep 28 01:59:02 2017 +0000 Fix nearbyint arithmetic moved before feholdexcept (bug 22225). In I remarked on the possibility of arithmetic in various nearbyint implementations being scheduled before feholdexcept calls, resulting in spurious "inexact" exceptions. I'm now actually observing this occurring in glibc built for ARM with GCC 7 (in fact, both copies of the same addition/subtraction sequence being combined and moved out before the conditionals and feholdexcept/fesetenv pairs), resulting in test failures. This patch makes the nearbyint implementations with this particular feholdexcept / arithmetic / fesetenv pattern consistently use math_opt_barrier on the function argument when first used in arithmetic, and also consistently use math_force_eval before fesetenv (the latter was generally already done, but the dbl-64/wordsize-64 implementation used math_opt_barrier instead, and as math_opt_barrier's intended effect is through its output value being used, such a use that doesn't use the return value is suspect). Tested for x86_64 (--disable-multi-arch so more of these implementations get used), and for ARM in a configuration where I saw the problem scheduling. [BZ #22225] * sysdeps/ieee754/dbl-64/s_nearbyint.c (__nearbyint): Use math_opt_barrier on argument when doing arithmetic on it. * sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c (__nearbyint): Likewise. Use math_force_eval not math_opt_barrier after arithmetic. * sysdeps/ieee754/flt-32/s_nearbyintf.c (__nearbyintf): Use math_opt_barrier on argument when doing arithmetic on it. * sysdeps/ieee754/ldbl-128/s_nearbyintl.c (__nearbyintl): Likewise. (cherry picked from commit f124cb381116b5809de198327690ad0bd8d1478e) diff --git a/ChangeLog b/ChangeLog index 382674d5a3..992722acdb 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,16 @@ +2017-09-28 Joseph Myers + + [BZ #22225] + * sysdeps/ieee754/dbl-64/s_nearbyint.c (__nearbyint): Use + math_opt_barrier on argument when doing arithmetic on it. + * sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c (__nearbyint): + Likewise. Use math_force_eval not math_opt_barrier after + arithmetic. + * sysdeps/ieee754/flt-32/s_nearbyintf.c (__nearbyintf): Use + math_opt_barrier on argument when doing arithmetic on it. + * sysdeps/ieee754/ldbl-128/s_nearbyintl.c (__nearbyintl): + Likewise. + 2017-09-22 Gabriel F. T. Gomes [BZ #22146] diff --git a/sysdeps/ieee754/dbl-64/s_nearbyint.c b/sysdeps/ieee754/dbl-64/s_nearbyint.c index dec0c5d6ee..6e3f8316b1 100644 --- a/sysdeps/ieee754/dbl-64/s_nearbyint.c +++ b/sysdeps/ieee754/dbl-64/s_nearbyint.c @@ -48,7 +48,7 @@ __nearbyint (double x) if (j0 < 0) { libc_feholdexcept (&env); - w = TWO52[sx] + x; + w = TWO52[sx] + math_opt_barrier (x); t = w - TWO52[sx]; math_force_eval (t); libc_fesetenv (&env); @@ -65,7 +65,7 @@ __nearbyint (double x) return x; /* x is integral */ } libc_feholdexcept (&env); - w = TWO52[sx] + x; + w = TWO52[sx] + math_opt_barrier (x); t = w - TWO52[sx]; math_force_eval (t); libc_fesetenv (&env); diff --git a/sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c b/sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c index 8293819981..7d135b54e4 100644 --- a/sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c +++ b/sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c @@ -42,9 +42,9 @@ __nearbyint(double x) if(__builtin_expect(j0<52, 1)) { if(j0<0) { libc_feholdexcept (&env); - double w = TWO52[sx]+x; + double w = TWO52[sx] + math_opt_barrier (x); double t = w-TWO52[sx]; - math_opt_barrier(t); + math_force_eval (t); libc_fesetenv (&env); return __copysign (t, x); } @@ -53,9 +53,9 @@ __nearbyint(double x) else return x; /* x is integral */ } libc_feholdexcept (&env); - double w = TWO52[sx]+x; + double w = TWO52[sx] + math_opt_barrier (x); double t = w-TWO52[sx]; - math_opt_barrier (t); + math_force_eval (t); libc_fesetenv (&env); return t; } diff --git a/sysdeps/ieee754/flt-32/s_nearbyintf.c b/sysdeps/ieee754/flt-32/s_nearbyintf.c index 5aebefafcf..b06df6b3c8 100644 --- a/sysdeps/ieee754/flt-32/s_nearbyintf.c +++ b/sysdeps/ieee754/flt-32/s_nearbyintf.c @@ -37,7 +37,7 @@ __nearbyintf(float x) if(j0<23) { if(j0<0) { libc_feholdexceptf (&env); - w = TWO23[sx]+x; + w = TWO23[sx] + math_opt_barrier (x); t = w-TWO23[sx]; math_force_eval (t); libc_fesetenvf (&env); @@ -50,7 +50,7 @@ __nearbyintf(float x) else return x; /* x is integral */ } libc_feholdexceptf (&env); - w = TWO23[sx]+x; + w = TWO23[sx] + math_opt_barrier (x); t = w-TWO23[sx]; math_force_eval (t); libc_fesetenvf (&env); diff --git a/sysdeps/ieee754/ldbl-128/s_nearbyintl.c b/sysdeps/ieee754/ldbl-128/s_nearbyintl.c index 1565a8183f..98a33d24a7 100644 --- a/sysdeps/ieee754/ldbl-128/s_nearbyintl.c +++ b/sysdeps/ieee754/ldbl-128/s_nearbyintl.c @@ -45,7 +45,7 @@ _Float128 __nearbyintl(_Float128 x) if(j0<112) { if(j0<0) { feholdexcept (&env); - w = TWO112[sx]+x; + w = TWO112[sx] + math_opt_barrier (x); t = w-TWO112[sx]; math_force_eval (t); fesetenv (&env); @@ -58,7 +58,7 @@ _Float128 __nearbyintl(_Float128 x) else return x; /* x is integral */ } feholdexcept (&env); - w = TWO112[sx]+x; + w = TWO112[sx] + math_opt_barrier (x); t = w-TWO112[sx]; math_force_eval (t); fesetenv (&env); commit fdf58ebc60ce0eb459fd616241b52872b3571ac1 Author: Joseph Myers Date: Sat Sep 30 00:48:07 2017 +0000 Add missing bug fixes to list in NEWS file. diff --git a/NEWS b/NEWS index 48e2064380..8e468c0369 100644 --- a/NEWS +++ b/NEWS @@ -28,9 +28,13 @@ The following bugs are resolved with this release: [21932] Unpaired __resolv_context_get in generic get*_r implementation [21941] powerpc: Restrict xssqrtqp operands to Vector Registers [21972] assert macro requires operator== (int) for its argument type + [21982] string: stratcliff.c: error: assuming signed overflow does not + occur with -O3 + [22051] libc: zero terminator in the middle of glibc's .eh_frame [22095] resolv: Fix memory leak with OOM during resolv.conf parsing [22096] resolv: __resolv_conf_attach must not free passed conf object [22146] Let fpclassify use the builtin when optimizing for size in C++ mode + [22225] math: nearbyint arithmetic moved before feholdexcept Version 2.26 commit 3f68c5c9b61600f0f85c75bac15b1520d5059359 Author: Joseph Myers Date: Tue Aug 22 00:59:43 2017 +0000 Fix sparc32 bits/long-double.h (bug 21987). My refactoring of long double information commit 0acb8a2a855395c25b1feef2470f4d7ca4bed589 Author: Joseph Myers Date: Wed Dec 14 18:27:56 2016 +0000 Refactor long double information into bits/long-double.h. resulted in sparc32 configurations installing the ldbl-opt version of bits/long-double.h instead of the intended sysdeps/unix/sysv/linux/sparc version. For sparc32 by itself, this is not a problem, since the ldbl-opt version is correct for sparc32. However, both sparc32 and sparc64 are supposed to install sets of headers that work for both of them, so that a single sysroot, whichever order the libraries are built and installed in, works for both. The effect of having the wrong version installed is that you end up with a miscompiled sparc64 libstdc++ which fails glibc's configure tests for the C++ compiler. This patch moves the header from sysdeps/unix/sysv/linux/sparc to separate copies of the same file for sparc32 and sparc64, to ensure it comes before ldbl-opt in the sysdeps directory ordering. Tested with build-many-glibcs.py for sparc64-linux-gnu and sparcv9-linux-gnu. [BZ #21987] * sysdeps/unix/sysv/linux/sparc/bits/long-double.h: Remove file and copy to ... * sysdeps/unix/sysv/linux/sparc/sparc32/bits/long-double.h: ... here. * sysdeps/unix/sysv/linux/sparc/sparc64/bits/long-double.h: ... and here. (cherry picked from commit 80f91666fed71fa3dd5eb5618739147cc731bc89) diff --git a/ChangeLog b/ChangeLog index 992722acdb..cc763ac065 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,13 @@ +2017-08-22 Joseph Myers + + [BZ #21987] + * sysdeps/unix/sysv/linux/sparc/bits/long-double.h: Remove file + and copy to ... + * sysdeps/unix/sysv/linux/sparc/sparc32/bits/long-double.h: + ... here. + * sysdeps/unix/sysv/linux/sparc/sparc64/bits/long-double.h: + ... and here. + 2017-09-28 Joseph Myers [BZ #22225] diff --git a/sysdeps/unix/sysv/linux/sparc/bits/long-double.h b/sysdeps/unix/sysv/linux/sparc/sparc32/bits/long-double.h similarity index 100% rename from sysdeps/unix/sysv/linux/sparc/bits/long-double.h rename to sysdeps/unix/sysv/linux/sparc/sparc32/bits/long-double.h diff --git a/sysdeps/unix/sysv/linux/sparc/sparc64/bits/long-double.h b/sysdeps/unix/sysv/linux/sparc/sparc64/bits/long-double.h new file mode 100644 index 0000000000..094e05124b --- /dev/null +++ b/sysdeps/unix/sysv/linux/sparc/sparc64/bits/long-double.h @@ -0,0 +1,26 @@ +/* Properties of long double type. SPARC version. + Copyright (C) 2016-2017 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 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 + . */ + +#include + +#if !defined __NO_LONG_DOUBLE_MATH && __WORDSIZE == 32 +# define __LONG_DOUBLE_MATH_OPTIONAL 1 +# ifndef __LONG_DOUBLE_128__ +# define __NO_LONG_DOUBLE_MATH 1 +# endif +#endif commit 3b10c5d2abb0392d5ecfd865e2eb911ac109e36f Author: Gabriel F. T. Gomes Date: Mon Oct 2 14:46:35 2017 -0300 Add C++ versions of iscanonical for ldbl-96 and ldbl-128ibm (bug 22235) All representations of floating-point numbers in types with IEC 60559 binary exchange format are canonical. On the other hand, types with IEC 60559 extended formats, such as those implemented under ldbl-96 and ldbl-128ibm, contain representations that are not canonical. TS 18661-1 introduced the type-generic macro iscanonical, which returns whether a floating-point value is canonical or not. In Glibc, this type-generic macro is implemented using the macro __MATH_TG, which, when support for float128 is enabled, relies on __builtin_types_compatible_p to select between floating-point types. However, this use of iscanonical breaks C++ applications, because the builtin is only available in C mode. This patch provides a C++ implementation of iscanonical that relies on function overloading, rather than builtins, to select between floating-point types. Unlike the C++ implementations for iszero and issignaling, this implementation ignores __NO_LONG_DOUBLE_MATH. The double type always matches IEC 60559 double format, which is always canonical. Thus, when double and long double are the same (__NO_LONG_DOUBLE_MATH), iscanonical always returns 1 and is not implemented with __MATH_TG. Tested for powerpc64, powerpc64le and x86_64. [BZ #22235] * math/math.h: Trivial fix for unbalanced parentheses in comment. * math/Makefile [CXX] (tests): Add test-math-iscanonical.cc. (CFLAGS-test-math-iscanonical.cc): New variable. * math/test-math-iscanonical.cc: New file. * sysdeps/ieee754/ldbl-96/bits/iscanonical.h (iscanonical): Provide a C++ implementation based on function overloading, rather than using __MATH_TG, which uses C-only builtins. * sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h (iscanonical): Likewise. * sysdeps/powerpc/powerpc64le/Makefile (CFLAGS-test-math-iscanonical.cc): New variable. (cherry picked from commit aa0235dfdebffe9b338deba51f3ba563ee9b433d) diff --git a/ChangeLog b/ChangeLog index cc763ac065..fea4fd0cd6 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,18 @@ +2017-10-03 Gabriel F. T. Gomes + + [BZ #22235] + * math/math.h: Trivial fix for unbalanced parentheses in comment. + * math/Makefile [CXX] (tests): Add test-math-iscanonical.cc. + (CFLAGS-test-math-iscanonical.cc): New variable. + * math/test-math-iscanonical.cc: New file. + * sysdeps/ieee754/ldbl-96/bits/iscanonical.h (iscanonical): + Provide a C++ implementation based on function overloading, + rather than using __MATH_TG, which uses C-only builtins. + * sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h (iscanonical): + Likewise. + * sysdeps/powerpc/powerpc64le/Makefile + (CFLAGS-test-math-iscanonical.cc): New variable. + 2017-08-22 Joseph Myers [BZ #21987] diff --git a/NEWS b/NEWS index 8e468c0369..5b05edeb4f 100644 --- a/NEWS +++ b/NEWS @@ -35,6 +35,7 @@ The following bugs are resolved with this release: [22096] resolv: __resolv_conf_attach must not free passed conf object [22146] Let fpclassify use the builtin when optimizing for size in C++ mode [22225] math: nearbyint arithmetic moved before feholdexcept + [22235] Add C++ versions of iscanonical for ldbl-96 and ldbl-128ibm Version 2.26 diff --git a/math/Makefile b/math/Makefile index 0130fcf38b..2c17c68eda 100644 --- a/math/Makefile +++ b/math/Makefile @@ -203,7 +203,8 @@ tests-static = test-fpucw-static test-fpucw-ieee-static \ test-signgam-ullong-static test-signgam-ullong-init-static ifneq (,$(CXX)) -tests += test-math-isinff test-math-iszero test-math-issignaling +tests += test-math-isinff test-math-iszero test-math-issignaling \ + test-math-iscanonical endif ifneq (no,$(PERL)) @@ -351,6 +352,7 @@ CFLAGS-test-signgam-ullong-init-static.c = -std=c99 CFLAGS-test-math-isinff.cc = -std=gnu++11 CFLAGS-test-math-iszero.cc = -std=gnu++11 CFLAGS-test-math-issignaling.cc = -std=gnu++11 +CFLAGS-test-math-iscanonical.cc = -std=gnu++11 CFLAGS-test-iszero-excess-precision.c = -fexcess-precision=standard CFLAGS-test-iseqsig-excess-precision.c = -fexcess-precision=standard diff --git a/math/math.h b/math/math.h index f9348ec3ea..2b216c6da1 100644 --- a/math/math.h +++ b/math/math.h @@ -488,7 +488,7 @@ enum other hand, overloading provides the means to distinguish between the floating-point types. The overloading resolution will match the correct parameter (regardless of type qualifiers (i.e.: const - and volatile). */ + and volatile)). */ extern "C++" { inline int issignaling (float __val) { return __issignalingf (__val); } inline int issignaling (double __val) { return __issignaling (__val); } diff --git a/math/test-math-iscanonical.cc b/math/test-math-iscanonical.cc new file mode 100644 index 0000000000..aba68acb4f --- /dev/null +++ b/math/test-math-iscanonical.cc @@ -0,0 +1,48 @@ +/* Test for the C++ implementation of iscanonical. + Copyright (C) 2017 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 + . */ + +#define _GNU_SOURCE 1 +#include +#include + +static bool errors; + +template +static void +check_type () +{ + T val = 0; + + /* Check if iscanonical is available in C++ mode (bug 22235). */ + if (iscanonical (val) == 0) + errors++; +} + +static int +do_test (void) +{ + check_type (); + check_type (); + check_type (); +#if __HAVE_DISTINCT_FLOAT128 + check_type<_Float128> (); +#endif + return errors; +} + +#include diff --git a/sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h b/sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h index 7ddb368d26..f756857c03 100644 --- a/sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h +++ b/sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h @@ -37,5 +37,22 @@ extern int __iscanonicall (long double __x) conversion, before being discarded; in IBM long double, there are encodings that are not consistently handled as corresponding to any particular value of the type, and we return 0 for those. */ -# define iscanonical(x) __MATH_TG ((x), __iscanonical, (x)) -#endif +# ifndef __cplusplus +# define iscanonical(x) __MATH_TG ((x), __iscanonical, (x)) +# else +/* In C++ mode, __MATH_TG cannot be used, because it relies on + __builtin_types_compatible_p, which is a C-only builtin. On the + other hand, overloading provides the means to distinguish between + the floating-point types. The overloading resolution will match + the correct parameter (regardless of type qualifiers (i.e.: const + and volatile)). */ +extern "C++" { +inline int iscanonical (float __val) { return __iscanonicalf (__val); } +inline int iscanonical (double __val) { return __iscanonical (__val); } +inline int iscanonical (long double __val) { return __iscanonicall (__val); } +# if __HAVE_DISTINCT_FLOAT128 +inline int iscanonical (_Float128 __val) { return __iscanonicalf128 (__val); } +# endif +} +# endif /* __cplusplus */ +#endif /* __NO_LONG_DOUBLE_MATH */ diff --git a/sysdeps/ieee754/ldbl-96/bits/iscanonical.h b/sysdeps/ieee754/ldbl-96/bits/iscanonical.h index 4a4f4ad024..cfa36a0c2a 100644 --- a/sysdeps/ieee754/ldbl-96/bits/iscanonical.h +++ b/sysdeps/ieee754/ldbl-96/bits/iscanonical.h @@ -34,4 +34,21 @@ extern int __iscanonicall (long double __x) conversion, before being discarded; in extended precision, there are encodings that are not consistently handled as corresponding to any particular value of the type, and we return 0 for those. */ -#define iscanonical(x) __MATH_TG ((x), __iscanonical, (x)) +#ifndef __cplusplus +# define iscanonical(x) __MATH_TG ((x), __iscanonical, (x)) +#else +/* In C++ mode, __MATH_TG cannot be used, because it relies on + __builtin_types_compatible_p, which is a C-only builtin. On the + other hand, overloading provides the means to distinguish between + the floating-point types. The overloading resolution will match + the correct parameter (regardless of type qualifiers (i.e.: const + and volatile)). */ +extern "C++" { +inline int iscanonical (float __val) { return __iscanonicalf (__val); } +inline int iscanonical (double __val) { return __iscanonical (__val); } +inline int iscanonical (long double __val) { return __iscanonicall (__val); } +# if __HAVE_DISTINCT_FLOAT128 +inline int iscanonical (_Float128 __val) { return __iscanonicalf128 (__val); } +# endif +} +#endif /* __cplusplus */ diff --git a/sysdeps/powerpc/powerpc64le/Makefile b/sysdeps/powerpc/powerpc64le/Makefile index dea2290736..cf2dbfb673 100644 --- a/sysdeps/powerpc/powerpc64le/Makefile +++ b/sysdeps/powerpc/powerpc64le/Makefile @@ -16,6 +16,7 @@ $(foreach suf,$(all-object-suffixes),%f128_r$(suf)): CFLAGS += -mfloat128 $(foreach suf,$(all-object-suffixes),$(objpfx)test-float128%$(suf)): CFLAGS += -mfloat128 $(foreach suf,$(all-object-suffixes),$(objpfx)test-ifloat128%$(suf)): CFLAGS += -mfloat128 CFLAGS-libm-test-support-float128.c += -mfloat128 +CFLAGS-test-math-iscanonical.cc += -mfloat128 CFLAGS-test-math-issignaling.cc += -mfloat128 CFLAGS-test-math-iszero.cc += -mfloat128 $(objpfx)test-float128% $(objpfx)test-ifloat128% $(objpfx)test-math-iszero: \ commit dd3a7239fddff81ac31373d69978d7aa1902c65f Author: H.J. Lu Date: Tue Oct 3 17:41:32 2017 -0700 test-math-iscanonical.cc: Replace bool with int Fix GCC 7 compilation error: test-math-iscanonical.cc: In function ‘void check_type()’: test-math-iscanonical.cc:33:11: error: use of an operand of type ‘bool’ in ‘operator++’ is deprecated [-Werror=deprecated] errors++; ^~ Since not all non-zero error counts are errors, return errors != 0 instead. * math/test-math-iscanonical.cc (error): Replace bool with int. (do_test): Return errors != 0. (cherry picked from commit cdd4155d6c527c00a89606385859984e35bd2910 and commit 758f1bfa2a1bccb52f1b3e97444a367d35aceaee) diff --git a/ChangeLog b/ChangeLog index fea4fd0cd6..d7a185e99d 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,8 @@ +2017-10-04 H.J. Lu + + * math/test-math-iscanonical.cc (error): Replace bool with int. + (do_test): Return errors != 0. + 2017-10-03 Gabriel F. T. Gomes [BZ #22235] diff --git a/math/test-math-iscanonical.cc b/math/test-math-iscanonical.cc index aba68acb4f..4cfb1c5055 100644 --- a/math/test-math-iscanonical.cc +++ b/math/test-math-iscanonical.cc @@ -20,7 +20,7 @@ #include #include -static bool errors; +static int errors; template static void @@ -42,7 +42,7 @@ do_test (void) #if __HAVE_DISTINCT_FLOAT128 check_type<_Float128> (); #endif - return errors; + return errors != 0; } #include commit 6e1ea21501eac981204c3cc8212d45998f74983c Author: Carlos O'Donell Date: Thu Sep 28 11:05:18 2017 -0600 malloc: Fix tcache leak after thread destruction [BZ #22111] The malloc tcache added in 2.26 will leak all of the elements remaining in the cache and the cache structure itself when a thread exits. The defect is that we do not set tcache_shutting_down early enough, and the thread simply recreates the tcache and places the elements back onto a new tcache which is subsequently lost as the thread exits (unfreed memory). The fix is relatively simple, move the setting of tcache_shutting_down earlier in tcache_thread_freeres. We add a test case which uses mallinfo and some heuristics to look for unaccounted for memory usage between the start and end of a thread start/join loop. It is very reliable at detecting that there is a leak given the number of iterations. Without the fix the test will consume 122MiB of leaked memory. (cherry picked from commit 1e26d35193efbb29239c710a4c46a64708643320) diff --git a/ChangeLog b/ChangeLog index d7a185e99d..bfcdc1ebbf 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,12 @@ +2017-10-06 Carlos O'Donell + + [BZ #22111] + * malloc/malloc.c (tcache_shutting_down): Use bool type. + (tcache_thread_freeres): Set tcache_shutting_down before + freeing the tcache. + * malloc/Makefile (tests): Add tst-malloc-tcache-leak. + * malloc/tst-malloc-tcache-leak.c: New file. + 2017-10-04 H.J. Lu * math/test-math-iscanonical.cc (error): Replace bool with int. diff --git a/malloc/Makefile b/malloc/Makefile index 3fa395b949..9e23db9343 100644 --- a/malloc/Makefile +++ b/malloc/Makefile @@ -34,6 +34,7 @@ tests := mallocbug tst-malloc tst-valloc tst-calloc tst-obstack \ tst-interpose-nothread \ tst-interpose-thread \ tst-alloc_buffer \ + tst-malloc-tcache-leak \ tests-static := \ tst-interpose-static-nothread \ @@ -242,3 +243,5 @@ tst-dynarray-fail-ENV = MALLOC_TRACE=$(objpfx)tst-dynarray-fail.mtrace $(objpfx)tst-dynarray-fail-mem.out: $(objpfx)tst-dynarray-fail.out $(common-objpfx)malloc/mtrace $(objpfx)tst-dynarray-fail.mtrace > $@; \ $(evaluate-test) + +$(objpfx)tst-malloc-tcache-leak: $(shared-thread-library) diff --git a/malloc/malloc.c b/malloc/malloc.c index e3ff778113..01ec1571b9 100644 --- a/malloc/malloc.c +++ b/malloc/malloc.c @@ -2952,7 +2952,7 @@ typedef struct tcache_perthread_struct tcache_entry *entries[TCACHE_MAX_BINS]; } tcache_perthread_struct; -static __thread char tcache_shutting_down = 0; +static __thread bool tcache_shutting_down = false; static __thread tcache_perthread_struct *tcache = NULL; /* Caller must ensure that we know tc_idx is valid and there's room @@ -2989,8 +2989,12 @@ tcache_thread_freeres (void) if (!tcache) return; + /* Disable the tcache and prevent it from being reinitialized. */ tcache = NULL; + tcache_shutting_down = true; + /* Free all of the entries and the tcache itself back to the arena + heap for coalescing. */ for (i = 0; i < TCACHE_MAX_BINS; ++i) { while (tcache_tmp->entries[i]) @@ -3002,8 +3006,6 @@ tcache_thread_freeres (void) } __libc_free (tcache_tmp); - - tcache_shutting_down = 1; } text_set_element (__libc_thread_subfreeres, tcache_thread_freeres); diff --git a/malloc/tst-malloc-tcache-leak.c b/malloc/tst-malloc-tcache-leak.c new file mode 100644 index 0000000000..22c679b65b --- /dev/null +++ b/malloc/tst-malloc-tcache-leak.c @@ -0,0 +1,112 @@ +/* Bug 22111: Test that threads do not leak their per thread cache. + Copyright (C) 2015-2017 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 + . */ + +/* The point of this test is to start and exit a large number of + threads, while at the same time looking to see if the used + memory grows with each round of threads run. If the memory + grows above some linear bound we declare the test failed and + that the malloc implementation is leaking memory with each + thread. This is a good indicator that the thread local cache + is leaking chunks. */ + +#include +#include +#include +#include +#include + +#include +#include +#include + +void * +worker (void *data) +{ + void *ret; + /* Allocate an arbitrary amount of memory that is known to fit into + the thread local cache (tcache). If we have at least 64 bins + (default e.g. TCACHE_MAX_BINS) we should be able to allocate 32 + bytes and force malloc to fill the tcache. We are assuming tcahce + init happens at the first small alloc, but it might in the future + be deferred to some other point. Therefore to future proof this + test we include a full alloc/free/alloc cycle for the thread. We + need a compiler barrier to avoid the removal of the useless + alloc/free. We send some memory back to main to have the memory + freed after the thread dies, as just another check that the chunks + that were previously in the tcache are still OK to free after + thread death. */ + ret = xmalloc (32); + __asm__ volatile ("" ::: "memory"); + free (ret); + return (void *) xmalloc (32); +} + +static int +do_test (void) +{ + pthread_t *thread; + struct mallinfo info_before, info_after; + void *retval; + + /* This is an arbitrary choice. We choose a total of THREADS + threads created and joined. This gives us enough iterations to + show a leak. */ + int threads = 100000; + + /* Avoid there being 0 malloc'd data at this point by allocating the + pthread_t required to run the test. */ + thread = (pthread_t *) xcalloc (1, sizeof (pthread_t)); + + info_before = mallinfo (); + + assert (info_before.uordblks != 0); + + printf ("INFO: %d (bytes) are in use before starting threads.\n", + info_before.uordblks); + + for (int loop = 0; loop < threads; loop++) + { + *thread = xpthread_create (NULL, worker, NULL); + retval = xpthread_join (*thread); + free (retval); + } + + info_after = mallinfo (); + printf ("INFO: %d (bytes) are in use after all threads joined.\n", + info_after.uordblks); + + /* We need to compare the memory in use before and the memory in use + after starting and joining THREADS threads. We almost always grow + memory slightly, but not much. Consider that if even 1-byte leaked + per thread we'd have THREADS bytes of additional memory, and in + general the in-use at the start of main is quite low. We will + always leak a full malloc chunk, and never just 1-byte, therefore + anything above "+ threads" from the start (constant offset) is a + leak. Obviously this assumes no thread-related malloc'd internal + libc data structures persist beyond the thread death, and any that + did would limit the number of times you could call pthread_create, + which is a QoI we'd want to detect and fix. */ + if (info_after.uordblks > (info_before.uordblks + threads)) + FAIL_EXIT1 ("Memory usage after threads is too high.\n"); + + /* Did not detect excessive memory usage. */ + free (thread); + exit (0); +} + +#include commit d5c6dea2d5b4b5c64625c5386f6baec7bf2d89b3 Author: Carlos O'Donell Date: Fri Oct 6 13:31:05 2017 -0700 Update NEWS for bug 22111. diff --git a/NEWS b/NEWS index 5b05edeb4f..d6d1f90008 100644 --- a/NEWS +++ b/NEWS @@ -33,6 +33,7 @@ The following bugs are resolved with this release: [22051] libc: zero terminator in the middle of glibc's .eh_frame [22095] resolv: Fix memory leak with OOM during resolv.conf parsing [22096] resolv: __resolv_conf_attach must not free passed conf object + [22111] malloc: per thread cache is not returned when thread exits [22146] Let fpclassify use the builtin when optimizing for size in C++ mode [22225] math: nearbyint arithmetic moved before feholdexcept [22235] Add C++ versions of iscanonical for ldbl-96 and ldbl-128ibm