crash-stack: fixed attaching to thread

It is not necessary to attach to parent process main thread to
resolve stack in another thread. This commit also solves hanging up
on 'D' (and 'Z') processes - main thread is not attachable,
when a crash occurs in another thread.

Change-Id: Ic3a085951acb4057fcdd4a476693954d1d3a522b

1 files changed, 23 insertions, 28 deletions

diff --git a/src/crash-stack/crash-stack.c b/src/crash-stack/crash-stack.c
index 63a2b4a..894e876 100644
--- a/src/crash-stack/crash-stack.c
+++ b/src/crash-stack/crash-stack.c
@@ -290,14 +290,14 @@ void __find_symbol_in_elf(ProcInfo *proc_info, Dwarf_Addr mapping_start)
 	close(fd);
 }
 
-static int __attachable(pid_t pid)
+static int __attachable(pid_t pid, pid_t tid)
 {
 	/* read /proc/<pid>/stat */
-	char buf[20];
+	char buf[40];
 	FILE *f;
 	char status;
 
-	snprintf(buf, sizeof(buf), "/proc/%d/stat", pid);
+	snprintf(buf, sizeof(buf), "/proc/%d/task/%d/stat", pid, tid);
 
 	f = fopen(buf, "r");
 	if (NULL == f)
@@ -312,13 +312,13 @@ static int __attachable(pid_t pid)
 	return status != 'D';
 }
 
-static void __print_proc_file(pid_t pid, const char *name)
+static void __print_proc_file(pid_t pid, pid_t tid, const char *name)
 {
 	char buf[1024];
 	FILE *f;
 	int r;
 
-	snprintf(buf, sizeof(buf), "/proc/%d/%s", pid, name);
+	snprintf(buf, sizeof(buf), "/proc/%d/task/%d/%s", pid, tid, name);
 
 	fprintf(outputfile, "%s:\n", buf);
 
@@ -339,14 +339,14 @@ static void __print_proc_file(pid_t pid, const char *name)
 	fprintf(outputfile, "\n");
 }
 
-static void __print_not_attachable_process_info(pid_t pid)
+static void __print_not_attachable_process_info(pid_t pid, pid_t tid)
 {
-	fprintf(outputfile, "ERROR: can't attach to process %d - process is in uninterruptible sleep state\n", pid);
+	fprintf(outputfile, "ERROR: can't attach to process %d, thread %d - thread is in uninterruptible sleep state\n", pid, tid);
 	fprintf(outputfile, "Giving some /proc info instead:\n\n");
-	__print_proc_file(pid, "wchan");
+	__print_proc_file(pid, tid, "wchan");
 	fprintf(outputfile, "\n");
-	__print_proc_file(pid, "syscall");
-	__print_proc_file(pid, "stack");
+	__print_proc_file(pid, tid, "syscall");
+	__print_proc_file(pid, tid, "stack");
 }
 
 /**
@@ -360,16 +360,16 @@ static Dwfl *__open_dwfl_with_pid(pid_t pid, pid_t tid)
 	int status;
 	pid_t stopped_pid;
 
-	status = __attachable(pid);
+	status = __attachable(pid, tid);
 	if (-1 == status)
 	{
-		fprintf(errfile, "failed to read /proc/%d/stat: %m\n", pid);
+		fprintf(errfile, "failed to read /proc/%d/task/%d/stat: %m\n", pid, tid);
 		return NULL;
 	}
 
 	if (!status)
 	{
-		__print_not_attachable_process_info(pid);
+		__print_not_attachable_process_info(pid, tid);
 		return NULL;
 	}
 
@@ -378,23 +378,18 @@ static Dwfl *__open_dwfl_with_pid(pid_t pid, pid_t tid)
 		return NULL;
 	}
 
-	if (pid != tid)
-		if (ptrace(PTRACE_SEIZE, pid, NULL, PTRACE_O_TRACEEXIT) != 0) {
-			fprintf(errfile, "PTRACE_SEIZE failed on PID %d: %m\n", pid);
-			return NULL;
-		}
-
 	ptrace(PTRACE_INTERRUPT, tid, 0, 0);
-	ptrace(PTRACE_INTERRUPT, pid, 0, 0);
 
-	stopped_pid = waitpid(pid, &status, 0);
-	if (stopped_pid == -1 || stopped_pid != pid || !WIFSTOPPED(status)) {
+	stopped_pid = waitpid(tid, &status, __WALL);
+	if (stopped_pid == -1 || stopped_pid != tid || !WIFSTOPPED(status)) {
 		fprintf(errfile, "waitpid failed: %m, stopped_pid=%d, status=%d\n", stopped_pid, status);
 		return NULL;
 	}
 
-	if (ptrace(PTRACE_GETSIGINFO, pid, NULL, &__siginfo) != 0)
+	if (ptrace(PTRACE_GETSIGINFO, tid, NULL, &__siginfo) != 0) {
+		fprintf(errfile, "ptrace GETSIGINFO failed: %m, pid=%d, tid=%d\n", pid, tid);
 		return NULL;
+	}
 
 	static const Dwfl_Callbacks proc_callbacks = {
 		.find_elf = dwfl_linux_proc_find_elf,
@@ -405,19 +400,19 @@ static Dwfl *__open_dwfl_with_pid(pid_t pid, pid_t tid)
 
 	Dwfl *dwfl = dwfl_begin(&proc_callbacks);
 	if (dwfl == NULL) {
-		fprintf(errfile, "process %d : Can't start dwfl (%s)\n", pid, dwfl_errmsg(-1));
+		fprintf(errfile, "process %d : Can't start dwfl (%s)\n", tid, dwfl_errmsg(-1));
 		return NULL;
 	}
 
-	if (dwfl_linux_proc_report(dwfl, pid) < 0) {
-		fprintf(errfile, "process %d : dwfl report failed (%s)\n", pid, dwfl_errmsg(-1));
+	if (dwfl_linux_proc_report(dwfl, tid) < 0) {
+		fprintf(errfile, "process %d : dwfl report failed (%s)\n", tid, dwfl_errmsg(-1));
 		dwfl_end(dwfl);
 		return NULL;
 	}
 
 #if _ELFUTILS_PREREQ(0,158)
-	if (dwfl_linux_proc_attach(dwfl, pid, true) < 0) {
-		fprintf(errfile, "process %d : dwfl attach failed (%s)\n", pid, dwfl_errmsg(-1));
+	if (dwfl_linux_proc_attach(dwfl, tid, true) < 0) {
+		fprintf(errfile, "process %d : dwfl attach failed (%s)\n", tid, dwfl_errmsg(-1));
 		dwfl_end(dwfl);
 		return NULL;
 	}