On Mon, Dec 12, 2016 at 01:53:29PM +0300, Dmitry V. Levin wrote: > On Thu, Dec 08, 2016 at 10:00:59AM -1000, Sean Stangl wrote: > > There is a segfault if strace is called as `strace -p 2260 -k`. > > > > The -k flag sets a global bool stack_trace_enabled. But -p already consults > > stack_trace_enabled during tcb initialization. So if -p is passed first, > > the libunwind components of the tcb are uninitialized. > > Good catch, thanks! > > > The attached patch fixes this issue by having -k ensure that the libunwind > > components are initialized. > > I think the patch needs a guard against multiple -k options to avoid > multiple late_unwind_tcb_init() invocations.
This is the patch I've applied:
diff --git a/strace.c b/strace.c
index 1b8b5cc..56fdfde 100644
--- a/strace.c
+++ b/strace.c
@@ -1769,8 +1769,14 @@ init(int argc, char *argv[])
}
#ifdef USE_LIBUNWIND
- if (stack_trace_enabled)
+ if (stack_trace_enabled) {
+ unsigned int tcbi;
+
unwind_init();
+ for (tcbi = 0; tcbi < tcbtabsize; ++tcbi) {
+ unwind_tcb_init(tcbtab[tcbi]);
+ }
+ }
#endif
/* See if they want to run as another user. */
diff --git a/unwind.c b/unwind.c
index 238e215..326d184 100644
--- a/unwind.c
+++ b/unwind.c
@@ -103,6 +103,9 @@ unwind_init(void)
void
unwind_tcb_init(struct tcb *tcp)
{
+ if (tcp->libunwind_ui)
+ return;
+
tcp->libunwind_ui = _UPT_create(tcp->pid);
if (!tcp->libunwind_ui)
die_out_of_memory();
--
ldv
pgpCX5yYcqlZc.pgp
Description: PGP signature
------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________ Strace-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/strace-devel
