Hi,
I recently tested nfs and found that after a network failure on the server, the client umount shared folder experienced a timeout phenomenon, lasting
approximately 100 seconds.
Reproduction method:
client:
mount -t nfs server_ip:folder folder -o soft,timeo=30,retry=3,intr
After successfully mounting,
server:
ifconfig ether down
client:
date
umount folder
date
At this point, umount will be stuck for 100 seconds.
By debugging, in the call_timeout function, will enter the following statement and directly return:
（e4ec48d3cc6139f4c1a934ff25d440cd4d50279f：SUNRPC: Make "no retrans timeout" soft tasks behave like softconn for timeouts）
if ((task->tk_flags & RPC_TASK_NO_RETRANS_TIMEOUT) &&
rpc_check_connected(task->tk_rqstp)) {
return;
}
When i revert this patch for testing and found that the umount time was approximately 20 seconds.
In addition, opening nfs debugging will repeat the following logs, echo 7 > /proc/sys/sunrpc/rpc_debug:
[ 1131.100062] RPC: 61 call_timeout (minor)
[ 1131.100062] RPC: 61 call_bind (status 0)
[ 1131.100064] RPC: 61 call_connect xprt 0000000025fae2c5 is connected
[ 1131.100065] RPC: 61 call_transmit (status 0)
[ 1131.100065] RPC: 61 xprt_prepare_transmit
[ 1140.316000] RPC: 61 call_status (status -110)
[ 1140.316003] RPC: 61 call_timeout (major)
[ 1140.316003] RPC: 61 call_timeout (minor)
[ 1140.316004] RPC: 61 call_bind (status 0)
[ 1140.316005] RPC: 61 call_connect xprt 0000000025fae2c5 is connected
[ 1140.316005] RPC: 61 call_transmit (status 0)
[ 1140.316006] RPC: 61 xprt_prepare_transmit
I am newer to nfs, may I ask whether this timeout phenomenon is a normal？
