commit a10de7d35f9024bfa90268d2afac6c62d7f5bab3
author Mark Salyzyn <salyzyn@google.com> Tue Dec 04 10:30:45 2018 -0800
committer Mark Salyzyn <salyzyn@google.com> Mon Dec 10 13:34:02 2018 -0800
tree 15d9a12eaf78e6dec28b9078bb8233618ea7b64b
parent 8d9985f387c269f680b50757d735f27691be657a

llkd: make 100% sure process that triggers panic still exists

There is time between inspection, filtration, determination and
dumping before triggering panic, make 100% sure the process still
exists.  If we had one false start, but another process triggers
and panic in the same pass, then recognize that we have already
dumped the data and skip it on the later ones.

Test: llkd unit test.
Bug: 33808187
Bug: 120378563
Bug: 120229612
Change-Id: Iacaf82a3d58e5a3c18edcff3c8fa540b21da36f1

llkd/libllkd.cpp

diff --git a/llkd/libllkd.cpp b/llkd/libllkd.cpp
index 427dace..5fca1ee 100644
--- a/llkd/libllkd.cpp
+++ b/llkd/libllkd.cpp
@@ -510,9 +510,7 @@
     return android::base::Trim(content) == string;
 }
 
-void llkPanicKernel(bool dump, pid_t tid, const char* state,
-                    const std::string& message = "") __noreturn;
-void llkPanicKernel(bool dump, pid_t tid, const char* state, const std::string& message) {
+void llkPanicKernel(bool dump, pid_t tid, const char* state, const std::string& message = "") {
     if (!message.empty()) LOG(ERROR) << message;
     auto sysrqTriggerFd = llkFileToWriteFd("/proc/sysrq-trigger");
     if (sysrqTriggerFd < 0) {
@@ -521,6 +519,7 @@
         // The answer to life, the universe and everything
         ::exit(42);
         // NOTREACHED
+        return;
     }
     ::sync();
     if (dump) {
@@ -544,6 +543,13 @@
     llkWriteStringToFile(message + (message.empty() ? "" : "\n") +
                                  "SysRq : Trigger a crash : 'livelock,"s + state + "'\n",
                          "/dev/kmsg");
+    // Because panic is such a serious thing to do, let us
+    // make sure that the tid being inspected still exists!
+    auto piddir = procdir + std::to_string(tid) + "/stat";
+    if (access(piddir.c_str(), F_OK) != 0) {
+        PLOG(WARNING) << piddir;
+        return;
+    }
     android::base::WriteStringToFd("c", sysrqTriggerFd);
     // NOTREACHED
     // DYB
@@ -909,6 +915,7 @@
     ms -= llkCycle;
     auto myPid = ::getpid();
     auto myTid = ::gettid();
+    auto dump = true;
     for (auto dp = llkTopDirectory.read(); dp != nullptr; dp = llkTopDirectory.read()) {
         std::string piddir;
 
@@ -1109,9 +1116,10 @@
             const auto message = state + " "s + llkFormat(procp->count) + " " +
                                  std::to_string(ppid) + "->" + std::to_string(pid) + "->" +
                                  std::to_string(tid) + " " + procp->getComm() + " [panic]";
-            llkPanicKernel(true, tid,
+            llkPanicKernel(dump, tid,
                            (state == 'Z') ? "zombie" : (state == 'D') ? "driver" : "sleeping",
                            message);
+            dump = false;
         }
         LOG(VERBOSE) << "+closedir()";
     }