[PATCH net-next] Revert "net: rtnetlink: remove local list in __linkwatch_run_queue()"

[PATCH net-next] Revert "net: rtnetlink: remove local list in __linkwatch_run_queue()"

[Johannes Berg]

From: Johannes Berg <johannes.berg@intel.com>

This reverts commit b8dbbbc535a9 ("net: rtnetlink: remove local list
in __linkwatch_run_queue()"). It's evidently broken when there's a
non-urgent work that gets added back, and then the loop can never
finish.

While reverting, add a note about that.

Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Fixes: b8dbbbc535a9 ("net: rtnetlink: remove local list in __linkwatch_run_queue()")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
---
Clear case of me being asleep at the wheel ... sorry about that!
---
 net/core/link_watch.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/net/core/link_watch.c b/net/core/link_watch.c
index 7be5b3ab32bd..429571c258da 100644
--- a/net/core/link_watch.c
+++ b/net/core/link_watch.c
@@ -192,6 +192,11 @@ static void __linkwatch_run_queue(int urgent_only)
 #define MAX_DO_DEV_PER_LOOP	100
 
 	int do_dev = MAX_DO_DEV_PER_LOOP;
+	/* Use a local list here since we add non-urgent
+	 * events back to the global one when called with
+	 * urgent_only=1.
+	 */
+	LIST_HEAD(wrk);
 
 	/* Give urgent case more budget */
 	if (urgent_only)
@@ -213,11 +218,12 @@ static void __linkwatch_run_queue(int urgent_only)
 	clear_bit(LW_URGENT, &linkwatch_flags);
 
 	spin_lock_irq(&lweventlist_lock);
-	while (!list_empty(&lweventlist) && do_dev > 0) {
+	list_splice_init(&lweventlist, &wrk);
+
+	while (!list_empty(&wrk) && do_dev > 0) {
 		struct net_device *dev;
 
-		dev = list_first_entry(&lweventlist, struct net_device,
-				       link_watch_list);
+		dev = list_first_entry(&wrk, struct net_device, link_watch_list);
 		list_del_init(&dev->link_watch_list);
 
 		if (!netif_device_present(dev) ||
@@ -235,6 +241,9 @@ static void __linkwatch_run_queue(int urgent_only)
 		spin_lock_irq(&lweventlist_lock);
 	}
 
+	/* Add the remaining work back to lweventlist */
+	list_splice_init(&wrk, &lweventlist);
+
 	if (!list_empty(&lweventlist))
 		linkwatch_schedule_work(0);
 	spin_unlock_irq(&lweventlist_lock);
-- 
2.43.0

Re: [PATCH net-next] Revert "net: rtnetlink: remove local list in __linkwatch_run_queue()"

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by David S. Miller <davem@davemloft.net>:

On Fri,  8 Dec 2023 10:52:15 +0100 you wrote:
> From: Johannes Berg <johannes.berg@intel.com>
> 
> This reverts commit b8dbbbc535a9 ("net: rtnetlink: remove local list
> in __linkwatch_run_queue()"). It's evidently broken when there's a
> non-urgent work that gets added back, and then the loop can never
> finish.
> 
> [...]

Here is the summary with links:
  - [net-next] Revert "net: rtnetlink: remove local list in __linkwatch_run_queue()"
    https://git.kernel.org/netdev/net-next/c/9a64d4c93eee

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html