fix the case High Sierra's mincore(2) may return -128 [Bug #13895]

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@59893 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This commit is contained in:
naruse 2017-09-14 09:21:48 +00:00
parent ea49381e51
commit fff6809b30

View File

@ -442,7 +442,7 @@ darwin_sigtramp:
ucontext_t *uctx;
char vec[1];
int r;
/* get _sigtramp's ucontext_t and set values to cursor
/* get previous frame information from %rbx at _sigtramp and set values to cursor
* http://www.opensource.apple.com/source/Libc/Libc-825.25/i386/sys/_sigtramp.s
* http://www.opensource.apple.com/source/libunwind/libunwind-35.1/src/unw_getcontext.s
*/
@ -465,8 +465,35 @@ darwin_sigtramp:
unw_set_reg(&cursor, UNW_X86_64_R14, uctx->uc_mcontext->__ss.__r14);
unw_set_reg(&cursor, UNW_X86_64_R15, uctx->uc_mcontext->__ss.__r15);
ip = uctx->uc_mcontext->__ss.__rip;
/* There's 4 cases for SEGV:
* (1) called invalid address
* (2) read or write invalid address
* (3) received signal
*
* Detail:
* (1) called invalid address
* In this case, saved ip is invalid address.
* It needs to just save the address for the information,
* skip the frame, and restore the frame calling the
* invalid address from %rsp.
* The problem is how to check whether the ip is valid or not.
* This code uses mincore(2) and assume the address's page is
* incore/referenced or not reflects the problem.
* Note that High Sierra's mincore(2) may return -128.
* (2) read or write invalid address
* saved ip is valid. just restart backtracing.
* (3) received signal in user space
* Same as (2).
* (4) received signal in kernel
* In this case saved ip points just after syscall, but registers are
* already overwriten by kernel. To fix register consistency,
* skip libc's kernel wrapper.
* To detect this case, just previous two bytes of ip is "\x0f\x05",
* syscall instruction of x86_64.
*/
r = mincore((const void *)ip, 1, vec);
if (r || !vec[0] || memcmp((const char *)ip-2, "\x0f\x05", 2) == 0) {
if (r || vec[0] <= 0 || memcmp((const char *)ip-2, "\x0f\x05", 2) == 0) {
/* if segv is caused by invalid call or signal received in syscall */
/* the frame is invalid; skip */
trace[n++] = (void *)ip;