summary refs log tree commit diff
path: root/nptl/pthread_create.c
diff options
context:
space:
mode:
authorH.J. Lu <hjl.tools@gmail.com>2018-05-02 06:17:20 -0700
committerH.J. Lu <hjl.tools@gmail.com>2018-05-02 06:17:41 -0700
commitd6cc1829aa31b6fb060f24dffd28aa6705cdd33a (patch)
tree197e7983f66d43765505f94a5a29d254d27f6d02 /nptl/pthread_create.c
parentb109fbfe4dd2ca77656157ddaded773e0da630a2 (diff)
downloadglibc-d6cc1829aa31b6fb060f24dffd28aa6705cdd33a.tar.gz
glibc-d6cc1829aa31b6fb060f24dffd28aa6705cdd33a.tar.xz
glibc-d6cc1829aa31b6fb060f24dffd28aa6705cdd33a.zip
x86: Use pad in pthread_unwind_buf to preserve shadow stack register
The pad array in struct pthread_unwind_buf is used by setjmp to save
shadow stack register.  We assert that size of struct pthread_unwind_buf
is no less than offset of shadow stack pointer + shadow stack pointer
size.

Since functions, like LIBC_START_MAIN, START_THREAD_DEFN as well as
these with thread cancellation, call setjmp, but never return after
__libc_unwind_longjmp, __libc_unwind_longjmp, which is defined as
__libc_longjmp on x86, doesn't need to restore shadow stack register.
__libc_longjmp, which is a private interface for thread cancellation
implementation in libpthread, is changed to call __longjmp_cancel,
instead of __longjmp.  __longjmp_cancel is a new internal function
in libc, which is similar to __longjmp, but doesn't restore shadow
stack register.

The compatibility longjmp and siglongjmp in libpthread.so are changed
to call __libc_siglongjmp, instead of __libc_longjmp, so that they will
restore shadow stack register.

Tested with build-many-glibcs.py.

Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>

	* nptl/pthread_create.c (START_THREAD_DEFN): Clear previous
	handlers after setjmp.
	* setjmp/longjmp.c (__libc_longjmp): Don't define alias if
	defined.
	* sysdeps/unix/sysv/linux/x86/setjmpP.h: Include
	<libc-pointer-arith.h>.
	(_JUMP_BUF_SIGSET_BITS_PER_WORD): New.
	(_JUMP_BUF_SIGSET_NSIG): Changed to 96.
	(_JUMP_BUF_SIGSET_NWORDS): Changed to use ALIGN_UP and
	_JUMP_BUF_SIGSET_BITS_PER_WORD.
	* sysdeps/x86/Makefile (sysdep_routines): Add __longjmp_cancel.
	* sysdeps/x86/__longjmp_cancel.S: New file.
	* sysdeps/x86/longjmp.c: Likewise.
	* sysdeps/x86/nptl/pt-longjmp.c: Likewise.
Diffstat (limited to 'nptl/pthread_create.c')
-rw-r--r--nptl/pthread_create.c17
1 files changed, 14 insertions, 3 deletions
diff --git a/nptl/pthread_create.c b/nptl/pthread_create.c
index caaf07c134..92c945b12b 100644
--- a/nptl/pthread_create.c
+++ b/nptl/pthread_create.c
@@ -427,12 +427,23 @@ START_THREAD_DEFN
      compilers without that support we do use setjmp.  */
   struct pthread_unwind_buf unwind_buf;
 
-  /* No previous handlers.  */
+  int not_first_call;
+  not_first_call = setjmp ((struct __jmp_buf_tag *) unwind_buf.cancel_jmp_buf);
+
+  /* No previous handlers.  NB: This must be done after setjmp since the
+     private space in the unwind jump buffer may overlap space used by
+     setjmp to store extra architecture-specific information which is
+     never used by the cancellation-specific __libc_unwind_longjmp.
+
+     The private space is allowed to overlap because the unwinder never
+     has to return through any of the jumped-to call frames, and thus
+     only a minimum amount of saved data need be stored, and for example,
+     need not include the process signal mask information. This is all
+     an optimization to reduce stack usage when pushing cancellation
+     handlers.  */
   unwind_buf.priv.data.prev = NULL;
   unwind_buf.priv.data.cleanup = NULL;
 
-  int not_first_call;
-  not_first_call = setjmp ((struct __jmp_buf_tag *) unwind_buf.cancel_jmp_buf);
   if (__glibc_likely (! not_first_call))
     {
       /* Store the new cleanup handler info.  */