From: Etienne Martineau <etmartin4313@xxxxxxxxx>
If hung task checking is enabled and FUSE server stops responding for a
long period of time, the hung task timer may fire towards the FUSE clients
and trigger stack dumps that unnecessarily alarm the user.
So, if hung task checking is enabled, we should wake up periodically to
prevent it from triggering stack dumps. This patch uses a wake-up interval
equal to half the hung_task_timeout_secs timer period, which keeps overhead
low.
Without this patch, an unresponsive FUSE server can leave the FUSE clients
in D state and produce stack dumps like below when the hung task timer
expire.
INFO: task cp:2780 blocked for more than 30 seconds.
Not tainted 6.13.0-rc1 #4
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
task:cat state:D stack:0 pid:2598 tgid:2598 ppid:2583 flags:0x00004006
......
<TASK>
__schedule+0x443/0x16b0
schedule+0x2b/0x140
request_wait_answer+0x143/0x220
__fuse_simple_request+0xd8/0x2c0
fuse_send_open+0xc5/0x130
fuse_file_open+0x117/0x1a0
fuse_open+0x92/0x2f0
do_dentry_open+0x25d/0x5c0
vfs_open+0x2a/0x100
path_openat+0x2f5/0x11d0
do_filp_open+0xbe/0x180
do_sys_openat2+0xa1/0xd0
__x64_sys_openat+0x55/0xa0
x64_sys_call+0x1998/0x26f0
do_syscall_64+0x7c/0x170
......
</TASK>
Signed-off-by: Etienne Martineau <etmartin4313@xxxxxxxxx>
---
fs/fuse/dev.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
index 27ccae63495d..29e0c9adb799 100644
--- a/fs/fuse/dev.c
+++ b/fs/fuse/dev.c
@@ -21,6 +21,7 @@
#include <linux/swap.h>
#include <linux/splice.h>
#include <linux/sched.h>
+#include <linux/sched/sysctl.h>
#define CREATE_TRACE_POINTS
#include "fuse_trace.h"
@@ -422,6 +423,8 @@ static void request_wait_answer(struct fuse_req *req)
{
struct fuse_conn *fc = req->fm->fc;
struct fuse_iqueue *fiq = &fc->iq;
+ /* Prevent hung task timer from firing at us */
+ unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2;
int err;
if (!fc->no_interrupt) {
@@ -461,7 +464,12 @@ static void request_wait_answer(struct fuse_req *req)
* Either request is already in userspace, or it was forced.
* Wait it out.
*/
- wait_event(req->waitq, test_bit(FR_FINISHED, &req->flags));
+ if (timeout)
+ while (!wait_event_timeout(req->waitq,
+ test_bit(FR_FINISHED, &req->flags), timeout))
+ ;
+ else
+ wait_event(req->waitq, test_bit(FR_FINISHED, &req->flags));
}
static void __fuse_request_send(struct fuse_req *req)
--
2.34.1
