Message ID | 20240312190352.3455451-1-simon.chopin@canonical.com |
---|---|
State | New |
Headers | show |
Series | libquadmath: printf: fix misaligned access on args | expand |
* Simon Chopin: > On x86, this compiles into movdqa which segfaults on unaligned access. > > This kind of failure has been seen when running against glibc 2.39, > which incidentally changed the printf implementation to move away from > alloca() for this data to instead append it at the end of an existing > "scratch buffer", with arbitrary alignement, whereas alloca() was > probably more likely to be naturally aligned. This glibc change appears to be incorrect. I think we need to preserve ABI alignment for types than can be passed through the vararg interface. I'm not sure if this easily possible, though. Certainly needs a discussion on libc-alpha. Thanks, Florian
diff --git a/libquadmath/printf/printf_fp.c b/libquadmath/printf/printf_fp.c index 8effcee88fa..d86aa650d38 100644 --- a/libquadmath/printf/printf_fp.c +++ b/libquadmath/printf/printf_fp.c @@ -363,7 +363,7 @@ __quadmath_printf_fp (struct __quadmath_printf_file *fp, /* Fetch the argument value. */ { - fpnum = **(const __float128 **) args[0]; + memcpy(&fpnum, *(void* const *) args[0], sizeof(fpnum)); /* Check for special values: not a number or infinity. */ if (isnanq (fpnum))
On x86, this compiles into movdqa which segfaults on unaligned access. This kind of failure has been seen when running against glibc 2.39, which incidentally changed the printf implementation to move away from alloca() for this data to instead append it at the end of an existing "scratch buffer", with arbitrary alignement, whereas alloca() was probably more likely to be naturally aligned. Tested by adding the patch to the Ubuntu gcc-14 package in https://launchpad.net/~schopin/+archive/ubuntu/libquadmath Signed-off-by: Simon Chopin <simon.chopin@canonical.com> --- libquadmath/printf/printf_fp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) base-commit: 39737cdf002637c7a652e9c3e36f369cfce581e5