From a3d2f4261ef9a953904e3e21abafba0dad7daa77 Mon Sep 17 00:00:00 2001
From: gaoyi <gaoyi15@huawei.com>
Date: Mon, 28 Sep 2020 22:36:37 +0800
Subject: [PATCH] Retry to handle the uevent when worker is terminated abnormal

When processing uevent events fails, retry it.
---
 src/udev/udevd.c | 41 ++++++++++++++++++++++++++++++++++++-----
 1 file changed, 36 insertions(+), 5 deletions(-)

diff --git a/src/udev/udevd.c b/src/udev/udevd.c
index eb94ed3..5b743ad 100644
--- a/src/udev/udevd.c
+++ b/src/udev/udevd.c
@@ -70,6 +70,7 @@
 #include "version.h"
 
 #define WORKER_NUM_MAX 2048U
+#define UEVENT_MAX_RETRY_TIMES 3
 
 static bool arg_debug = false;
 static int arg_daemonize = false;
@@ -114,6 +115,7 @@ typedef struct Event {
         Manager *manager;
         Worker *worker;
         EventState state;
+        int retry;
 
         sd_device *dev;
         sd_device *dev_kernel; /* clone of originally received device */
@@ -148,6 +150,32 @@ typedef struct Worker {
 typedef struct WorkerMessage {
 } WorkerMessage;
 
+static bool event_retry(Event *event) {
+        if (!event)
+                return false;
+
+        assert(event->manager);
+
+        if (--event->retry < 0) {
+                log_device_error(event->dev, "Retry failed.");
+                return false;
+        }
+
+        log_device_info(event->dev, "Retry %d times.", UEVENT_MAX_RETRY_TIMES - event->retry);
+
+        event->timeout_warning_event = sd_event_source_unref(event->timeout_warning_event);
+        event->timeout_event = sd_event_source_unref(event->timeout_event);
+
+        if (event->worker) {
+                event->worker->event = NULL;
+                event->worker = NULL;
+        }
+
+        event->state = EVENT_QUEUED;
+
+        return true;
+}
+
 static void event_free(Event *event) {
         if (!event)
                 return;
@@ -638,6 +666,7 @@ static int event_queue_insert(Manager *manager, sd_device *dev) {
                 .dev_kernel = TAKE_PTR(clone),
                 .seqnum = seqnum,
                 .state = EVENT_QUEUED,
+                .retry = UEVENT_MAX_RETRY_TIMES,
         };
 
         if (LIST_IS_EMPTY(manager->events)) {
@@ -1314,11 +1343,13 @@ static int on_sigchld(sd_event_source *s, const struct signalfd_siginfo *si, voi
                         device_delete_db(worker->event->dev);
                         device_tag_index(worker->event->dev, NULL, false);
 
-                        if (manager->monitor) {
-                                /* forward kernel event without amending it */
-                                r = device_monitor_send_device(manager->monitor, NULL, worker->event->dev_kernel);
-                                if (r < 0)
-                                        log_device_error_errno(worker->event->dev_kernel, r, "Failed to send back device to kernel: %m");
+                        if (event_retry(worker->event) == false) {
+                                if (manager->monitor) {
+                                        /* forward kernel event without amending it */
+                                        r = device_monitor_send_device(manager->monitor, NULL, worker->event->dev_kernel);
+                                        if (r < 0)
+                                                log_device_error_errno(worker->event->dev_kernel, r, "Failed to send back device to kernel: %m");
+                                }
                         }
                 }
 
-- 
2.23.0