This patch adds split-stack support pointer guard on glibc for aarch64. Different from other architectures (powerpc, s390) where the memory is placed on TCB, aarch64 one is placed before thread pointer initial position. It has an advantage over extending TCB because for aarch64 TLS variable placement take in consideration tcbhead_t size and by changing its value would require to also update the static linker (and it would also add incompatibility with glibc and older linkers). For aarch64 tcb direct access is fastest for thread local variable on all mode and related TLS access. It requires just a direct load with displacement of -8 (since thread pointer points to tcbhead_t). It also adds a loader symbol (__tcb_private_ss) to signal the existence of the split stack guard area. Checked on aarch64-linux-gnu. * sysdeps/aarch64/Makefile [$(subdir) = elf] (sysdeps-dl-routines): Add tcb-version. * sysdeps/aarch64/Versions [ld] (GLIBC_2.26): Add __libc_tcb_private_ss. * sysdeps/aarch64/nptl/tls.h (tcbprehead_t): New struct. (TLS_PRE_TCB_SIZE): Take tcbprehead_t in consideration. (TLS_DEFINE_INIT_TP): Likewise. (THREAD_SELF): Likewise. (DB_THREAD_SELF): Likewise. * sysdeps/aarch64/tcb-version.c: New file. * sysdeps/unix/sysv/linux/aarch64/ld.abilist (GLIBC_2.26): Add. (__libc_tcb_private_ss): Likewise. * sysdeps/aarch64/nptl/tcb-offsets.sym (PTHREAD_SIZEOF): Rename to PTHREAD_PRE_TCB_SIZE.
The x86_64 backtrace implementation is used as a generic implementation (unwinding via unwind info and _Unwind_Backtrace) by various other architectures. This patch makes it more generic by making it use LIBGCC_S_SO from gnu/lib-names.h instead of hardcoding the libgcc_s.so.1 name, so that it can also be used on hppa which uses libgcc_s.so.4. Tested for x86_64. * sysdeps/x86_64/backtrace.c: Include <gnu/lib-names.h>. (init): Use LIBGCC_S_SO not hardcoded "libgcc_s.so.1".
Define new HWCAP bits and add their name to dl-procinfo.c following the linux definitions. Synchronizing with v4.15-rc8 version of linux, these are not expected to change before the 4.15 release. * sysdeps/unix/sysv/linux/aarch64/bits/hwcap.h (HWCAP_SHA3): Define. (HWCAP_SM3, HWCAP_SM4, HWCAP_ASIMDDP, HWCAP_SHA512, HWCAP_SVE): Define. * sysdeps/unix/sysv/linux/aarch64/dl-procinfo.c (_dl_aarch64_cap_flags): Update. (_DL_HWCAP_COUNT): Update.
Remove unused _DL_HWCAP_LAST definition and move _DL_HWCAP_COUNT where it is needed (dl-procinfo.h always includes dl-procinfo.c). * sysdeps/unix/sysv/linux/aarch64/dl-procinfo.h (_DL_HWCAP_LAST): Remove. (_DL_HWCAP_COUNT): Move to ... * sysdeps/unix/sysv/linux/aarch64/dl-procinfo.c (_DL_HWCAP_COUNT): ... here.
This issue is similar to BZ #19235, where spurious exceptions are created from adding 0.5 then converting to an integer. The solution is based on Joseph's fix for BZ #19235. [BZ #22697] * sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S (__llround): Do not add 0.5 to integer or out-of-range arguments.
In the static pie enabled libc, crt1.o uses the same position independent code as rcrt1.o and crt1.o is used instead of Scrt1.o when -no-pie executables are linked. When main is not defined in the executable, but in a shared library crt1.o is currently broken, it assumes main is local. (glibc has a test for this but i missed it in my previous testing.) To make both rcrt1.o and crt1.o happy with the same code, a wrapper is introduced around main: with this crt1.o works with extern main symbol while rcrt1.o does not depend on GOT relocations. (The change only affects static pie enabled libc. Further simplification of start.S is possible in the future by using the same approach for Scrt1.o too.) * aarch64/start.S (_start): Use __wrap_main. (__wrap_main): New local symbol.
Currently getcwd(3) can succeed without returning an absolute path because the underlying getcwd syscall, starting with linux commit v2.6.36-rc1~96^2~2, may succeed without returning an absolute path. This is a conformance issue because "The getcwd() function shall place an absolute pathname of the current working directory in the array pointed to by buf, and return buf". This is also a security issue because a non-absolute path returned by getcwd(3) causes a buffer underflow in realpath(3). Fix this by checking the path returned by getcwd syscall and falling back to generic_getcwd if the path is not absolute, effectively making getcwd(3) fail with ENOENT. The error code is chosen for consistency with the case when the current directory is unlinked. [BZ #22679] CVE-2018-1000001 * sysdeps/unix/sysv/linux/getcwd.c (__getcwd): Fall back to generic_getcwd if the path returned by getcwd syscall is not absolute. * io/tst-getcwd-abspath.c: New test. * io/Makefile (tests): Add tst-getcwd-abspath.
My fix for bug 22702 introduced linknamespace test failures on s390x-linux-gnu and s390-linux-gnu because it made remainder call __feholdexcept, and the s390 __feholdexcept calls fegetenv, and remainder is in Unix98 and XPG4.2 but fegetenv isn't. This patch makes __feholdexcept call __fegetenv instead to avoid that namespace issue. Tested (compilation) with build-many-glibcs.py for s390x-linux-gnu, where it resolves the test failures. * sysdeps/s390/fpu/feholdexcpt.c (__feholdexcept): Call __fegetenv instead of fegetenv.
For soft-float powerpc, the math/test-nearbyint-except-2 test fails because nearbyintl traps when traps on "inexact" are enabled on entry (and an "inexact" exception is generated internally, though cleared for the final return). The problem is the default implementation of libc_feholdsetround_noex_ctx, which does not disable exception traps. There is some ambiguity about whether the *noex* interfaces are required to do so or only permitted to do so. But given that we support fe* interfaces to enable and disable traps (on architectures with that functionality), functions that must not raise an exception (must not leave the flag set on exit if not set on entry) should also not trap on it when traps on that exception are enabled. So it is appropriate to define these interfaces to have the feholdexcept effect of disabling exception traps; this patch updates the default implementation and comments accordingly. At least some architecture versions already disable traps; there are few uses of the *noex* interfaces at all, and while it's possible there are bugs on any architecture versions failing to disable traps that appear in the exp2 and remainder implementations, there are currently no tests, other than this one for nearbyintl (where only the ldbl-128ibm implementation uses SET_RESTORE_ROUND_NOEX), that would fail as a result of such a bug. (Hard-float powerpc does disable traps here, hence the nearbyintl failure not appearing there.) Tested for powerpc (soft-float). This brings that configuration to clean math/ test results, provided you build with GCC 8 to get the fix for GCC bug 64811. [BZ #22702] * sysdeps/generic/math_private.h (libc_feresetround_noex): Update comment to say exceptions are discarded. (libc_feholdsetround_noex_ctx): Use __feholdexcept instead of __fegetenv. (SET_RESTORE_ROUND_NOEX): Update comment to say non-stop mode must be enabled.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>