public inbox for glibc-cvs@sourceware.org
help / color / mirror / Atom feed
* [glibc/google/grte/v5-2.27/master] Avoid ldbl-96 stack corruption from range reduction of pseudo-zero (bug 25487).
@ 2021-07-23 17:07 Stan Shebs
0 siblings, 0 replies; 2+ messages in thread
From: Stan Shebs @ 2021-07-23 17:07 UTC (permalink / raw)
To: glibc-cvs
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=ade59dc8ae92015be1ac41b463e17c9fd756a7a9
commit ade59dc8ae92015be1ac41b463e17c9fd756a7a9
Author: Joseph Myers <joseph@codesourcery.com>
Date: Wed Feb 12 23:31:56 2020 +0000
Avoid ldbl-96 stack corruption from range reduction of pseudo-zero (bug 25487).
Bug 25487 reports stack corruption in ldbl-96 sinl on a pseudo-zero
argument (an representation where all the significand bits, including
the explicit high bit, are zero, but the exponent is not zero, which
is not a valid representation for the long double type).
Although this is not a valid long double representation, existing
practice in this area (see bug 4586, originally marked invalid but
subsequently fixed) is that we still seek to avoid invalid memory
accesses as a result, in case of programs that treat arbitrary binary
data as long double representations, although the invalid
representations of the ldbl-96 format do not need to be consistently
handled the same as any particular valid representation.
This patch makes the range reduction detect pseudo-zero and unnormal
representations that would otherwise go to __kernel_rem_pio2, and
returns a NaN for them instead of continuing with the range reduction
process. (Pseudo-zero and unnormal representations whose unbiased
exponent is less than -1 have already been safely returned from the
function before this point without going through the rest of range
reduction.) Pseudo-zero representations would previously result in
the value passed to __kernel_rem_pio2 being all-zero, which is
definitely unsafe; unnormal representations would previously result in
a value passed whose high bit is zero, which might well be unsafe
since that is not a form of input expected by __kernel_rem_pio2.
Tested for x86_64.
Diff:
---
sysdeps/ieee754/ldbl-96/Makefile | 3 ++-
sysdeps/ieee754/ldbl-96/e_rem_pio2l.c | 12 +++++++++
sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c | 41 ++++++++++++++++++++++++++++++
3 files changed, 55 insertions(+), 1 deletion(-)
diff --git a/sysdeps/ieee754/ldbl-96/Makefile b/sysdeps/ieee754/ldbl-96/Makefile
index 790f670e44..99c596e3ce 100644
--- a/sysdeps/ieee754/ldbl-96/Makefile
+++ b/sysdeps/ieee754/ldbl-96/Makefile
@@ -17,5 +17,6 @@
# <http://www.gnu.org/licenses/>.
ifeq ($(subdir),math)
-tests += test-canonical-ldbl-96 test-totalorderl-ldbl-96
+tests += test-canonical-ldbl-96 test-totalorderl-ldbl-96 test-sinl-pseudo
+CFLAGS-test-sinl-pseudo.c += -fstack-protector-all
endif
diff --git a/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c b/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c
index f67805f2d3..c2ee56e71f 100644
--- a/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c
+++ b/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c
@@ -210,6 +210,18 @@ __ieee754_rem_pio2l (long double x, long double *y)
return 0;
}
+ if ((i0 & 0x80000000) == 0)
+ {
+ /* Pseudo-zero and unnormal representations are not valid
+ representations of long double. We need to avoid stack
+ corruption in __kernel_rem_pio2, which expects input in a
+ particular normal form, but those representations do not need
+ to be consistently handled like any particular floating-point
+ value. */
+ y[1] = y[0] = __builtin_nanl ("");
+ return 0;
+ }
+
/* Split the 64 bits of the mantissa into three 24-bit integers
stored in a double array. */
exp = j0 - 23;
diff --git a/sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c b/sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c
new file mode 100644
index 0000000000..f59b97769d
--- /dev/null
+++ b/sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c
@@ -0,0 +1,41 @@
+/* Test sinl for pseudo-zeros and unnormals for ldbl-96 (bug 25487).
+ Copyright (C) 2020 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
+ <https://www.gnu.org/licenses/>. */
+
+#include <math.h>
+#include <math_ldbl.h>
+#include <stdint.h>
+
+static int
+do_test (void)
+{
+ for (int i = 0; i < 64; i++)
+ {
+ uint64_t sig = i == 63 ? 0 : 1ULL << i;
+ long double ld;
+ SET_LDOUBLE_WORDS (ld, 0x4141,
+ sig >> 32, sig & 0xffffffffULL);
+ /* The requirement is that no stack overflow occurs when the
+ pseudo-zero or unnormal goes through range reduction. */
+ volatile long double ldr;
+ ldr = sinl (ld);
+ (void) ldr;
+ }
+ return 0;
+}
+
+#include <support/test-driver.c>
^ permalink raw reply [flat|nested] 2+ messages in thread
* [glibc/google/grte/v5-2.27/master] Avoid ldbl-96 stack corruption from range reduction of pseudo-zero (bug 25487).
@ 2021-08-28 0:39 Fangrui Song
0 siblings, 0 replies; 2+ messages in thread
From: Fangrui Song @ 2021-08-28 0:39 UTC (permalink / raw)
To: glibc-cvs
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=8c5a6bf8130c9e94d015eb6005ad19a16c1f6f0a
commit 8c5a6bf8130c9e94d015eb6005ad19a16c1f6f0a
Author: Joseph Myers <joseph@codesourcery.com>
Date: Wed Feb 12 23:31:56 2020 +0000
Avoid ldbl-96 stack corruption from range reduction of pseudo-zero (bug 25487).
Bug 25487 reports stack corruption in ldbl-96 sinl on a pseudo-zero
argument (an representation where all the significand bits, including
the explicit high bit, are zero, but the exponent is not zero, which
is not a valid representation for the long double type).
Although this is not a valid long double representation, existing
practice in this area (see bug 4586, originally marked invalid but
subsequently fixed) is that we still seek to avoid invalid memory
accesses as a result, in case of programs that treat arbitrary binary
data as long double representations, although the invalid
representations of the ldbl-96 format do not need to be consistently
handled the same as any particular valid representation.
This patch makes the range reduction detect pseudo-zero and unnormal
representations that would otherwise go to __kernel_rem_pio2, and
returns a NaN for them instead of continuing with the range reduction
process. (Pseudo-zero and unnormal representations whose unbiased
exponent is less than -1 have already been safely returned from the
function before this point without going through the rest of range
reduction.) Pseudo-zero representations would previously result in
the value passed to __kernel_rem_pio2 being all-zero, which is
definitely unsafe; unnormal representations would previously result in
a value passed whose high bit is zero, which might well be unsafe
since that is not a form of input expected by __kernel_rem_pio2.
Tested for x86_64.
Diff:
---
sysdeps/ieee754/ldbl-96/Makefile | 3 ++-
sysdeps/ieee754/ldbl-96/e_rem_pio2l.c | 12 +++++++++
sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c | 41 ++++++++++++++++++++++++++++++
3 files changed, 55 insertions(+), 1 deletion(-)
diff --git a/sysdeps/ieee754/ldbl-96/Makefile b/sysdeps/ieee754/ldbl-96/Makefile
index 790f670e44..99c596e3ce 100644
--- a/sysdeps/ieee754/ldbl-96/Makefile
+++ b/sysdeps/ieee754/ldbl-96/Makefile
@@ -17,5 +17,6 @@
# <http://www.gnu.org/licenses/>.
ifeq ($(subdir),math)
-tests += test-canonical-ldbl-96 test-totalorderl-ldbl-96
+tests += test-canonical-ldbl-96 test-totalorderl-ldbl-96 test-sinl-pseudo
+CFLAGS-test-sinl-pseudo.c += -fstack-protector-all
endif
diff --git a/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c b/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c
index f67805f2d3..c2ee56e71f 100644
--- a/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c
+++ b/sysdeps/ieee754/ldbl-96/e_rem_pio2l.c
@@ -210,6 +210,18 @@ __ieee754_rem_pio2l (long double x, long double *y)
return 0;
}
+ if ((i0 & 0x80000000) == 0)
+ {
+ /* Pseudo-zero and unnormal representations are not valid
+ representations of long double. We need to avoid stack
+ corruption in __kernel_rem_pio2, which expects input in a
+ particular normal form, but those representations do not need
+ to be consistently handled like any particular floating-point
+ value. */
+ y[1] = y[0] = __builtin_nanl ("");
+ return 0;
+ }
+
/* Split the 64 bits of the mantissa into three 24-bit integers
stored in a double array. */
exp = j0 - 23;
diff --git a/sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c b/sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c
new file mode 100644
index 0000000000..f59b97769d
--- /dev/null
+++ b/sysdeps/ieee754/ldbl-96/test-sinl-pseudo.c
@@ -0,0 +1,41 @@
+/* Test sinl for pseudo-zeros and unnormals for ldbl-96 (bug 25487).
+ Copyright (C) 2020 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
+ <https://www.gnu.org/licenses/>. */
+
+#include <math.h>
+#include <math_ldbl.h>
+#include <stdint.h>
+
+static int
+do_test (void)
+{
+ for (int i = 0; i < 64; i++)
+ {
+ uint64_t sig = i == 63 ? 0 : 1ULL << i;
+ long double ld;
+ SET_LDOUBLE_WORDS (ld, 0x4141,
+ sig >> 32, sig & 0xffffffffULL);
+ /* The requirement is that no stack overflow occurs when the
+ pseudo-zero or unnormal goes through range reduction. */
+ volatile long double ldr;
+ ldr = sinl (ld);
+ (void) ldr;
+ }
+ return 0;
+}
+
+#include <support/test-driver.c>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2021-08-28 0:39 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-23 17:07 [glibc/google/grte/v5-2.27/master] Avoid ldbl-96 stack corruption from range reduction of pseudo-zero (bug 25487) Stan Shebs
2021-08-28 0:39 Fangrui Song
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).