fix issue with longjmp out of signal handlers and cancellation

stale state information indicating that a thread was possibly blocked
at a cancellation point could get left behind if longjmp was used to
exit a signal handler that interrupted a cancellation point.

to fix the issue, we throw away the state information entirely and
simply compare the saved instruction pointer to a range of code
addresses in the __syscall_cp_asm function. all the ugly PIC work
(which becomes minimal anyway with this approach) is defered to
cancellation time instead of happening at every syscall, which should
improve performance too.

this commit also fixes cancellation on arm, which was mildly broken
(race condition, not checking cancellation flag once inside the
cancellation point zone). apparently i forgot to implement that. the
new arm code is untested, but appears correct; i'll test and fix it
later if there are problems.

1 files changed, 8 insertions, 10 deletions

diff --git a/src/thread/x86_64/syscall_cp.s b/src/thread/x86_64/syscall_cp.s
index b0363547..788c53cc 100644
--- a/src/thread/x86_64/syscall_cp.s
+++ b/src/thread/x86_64/syscall_cp.s
@@ -2,12 +2,12 @@
 .global __syscall_cp_asm
 .type   __syscall_cp_asm,@function
 __syscall_cp_asm:
-	lea 1f(%rip),%rax
-	mov %rax,8(%rdi)
-	mov %rsp,(%rdi)
-	mov 16(%rdi),%eax
+
+.global __cp_begin
+__cp_begin:
+	mov (%rdi),%eax
 	test %eax,%eax
-	jnz 2f
+	jnz __cancel
 	mov %rdi,%r11
 	mov %rsi,%rax
 	mov %rdx,%rdi
@@ -17,9 +17,7 @@ __syscall_cp_asm:
 	mov 8(%rsp),%r8
 	mov 16(%rsp),%r9
 	mov %r11,8(%rsp)
-1:	syscall
-	xor %ecx,%ecx
-	mov 8(%rsp),%rdi
-	mov %rcx,(%rdi)
+	syscall
+.global __cp_end
+__cp_end:
 	ret
-2:	call __cancel