linux-kernel - Question: How to customize retransmission timeout of unacknowledged NFS v3 TCP packet?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAPKjjnoeLkkvSZwpWw_bTCOCTfWYgjCyvjpoa95E4yroQR5zyA@mail.gmail.com>
Date: Wed, 29 May 2024 18:10:29 +0800
From: Zhitao Li <zhitao.li@...rtx.com>
To: Trond Myklebust <trond.myklebust@...merspace.com>, Anna Schumaker <anna@...nel.org>, 
	Chuck Lever <chuck.lever@...cle.com>, Jeff Layton <jlayton@...nel.org>, Neil Brown <neilb@...e.de>, 
	Olga Kornievskaia <kolga@...app.com>, Dai Ngo <Dai.Ngo@...cle.com>, Tom Talpey <tom@...pey.com>
Cc: linux-nfs@...r.kernel.org, linux-kernel@...r.kernel.org, 
	Ping Huang <huangping@...rtx.com>
Subject: Question: How to customize retransmission timeout of unacknowledged
 NFS v3 TCP packet?

Hi, dear community,

In our NFS environment, NFS client mounts remote NFS export with its
VIP. The VIP can be assigned to another server node for failover.
However, the NFS client sends the unacknowledged packet 50s+ after the
VIP is ready on the new node, which is because of the exponential
backoff retransmission algorithm.  I tried to set this parameter
"tcp_retries2" smaller so that the NFS client can reconnect with the
new node more quickly, but this parameter didn't take effect. From
tcpdump entries as follows,
  1. At "2024-05-29 11:47:00",  ARP is updated.
  2. At "2024-05-29 11:47:52" ,  the NFS client retried to send the packet.
  3. Then the connection is reset and a new connection starts.

I guess the parameter just takes effect for applications and doesn't
take effect for kernel modules like the NFS client. Could anyone give
some advice to customize  retransmission timeout of unacknowledged NFS
v3 TCP packet?


OS: Linux kernel v6.7.0
NFS mount options:
vers=3,nolock,proto=tcp,rsize=1048576,wsize=1048576,hard,timeo=600,retrans=2,noresvport

tcp_retries2:
[root@...play zhitaoli]# sysctl -w net.ipv4.tcp_retries2=5
net.ipv4.tcp_retries2 = 5
[root@...play zhitaoli]# cat /proc/sys/net/ipv4/tcp_retries2
5

tcpdump entries:

2024-05-29 11:46:02.331891 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973659245 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:02.542836 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973659456 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:02.751013 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973659664 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:03.166958 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973660080 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:04.046882 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973660960 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:05.710910 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973662624 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:09.039310 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973665952 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:16.017889 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973672930 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:29.326891 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973686240 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:46:55.950915 52:54:00:1d:a4:24 > 52:54:00:a0:93:93,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973712864 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:47:00.379844 52:54:00:13:1f:34 > Broadcast, ethertype
ARP (0x0806), length 60: Reply 10.125.1.85 is-at 52:54:00:13:1f:34,
length 46

2024-05-29 11:47:52.271192 52:54:00:1d:a4:24 > 52:54:00:13:1f:34,
ethertype IPv4 (0x0800), length 190: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [P.], seq 129897:130021, ack 171633, win 2356,
options [nop,nop,TS val 1973769184 ecr 28456
58566], length 124: NFS request xid 1954624602 120 access fh
Unknown/43000001180100000000000000DE40020000000000F439000000000000000000
NFS_ACCESS_READ|NFS_ACCESS_LOOKUP|NFS_ACCESS_MODIFY|NFS_ACCESS_EXTEND|NFS_ACCESS_DELETE

2024-05-29 11:47:52.272041 52:54:00:13:1f:34 > 52:54:00:1d:a4:24,
ethertype IPv4 (0x0800), length 54: 10.125.1.85.nfs >
10.125.1.214.58428: Flags [R], seq 1148562527, win 0, length 0

2024-05-29 11:47:52.272909 52:54:00:1d:a4:24 > 52:54:00:13:1f:34,
ethertype IPv4 (0x0800), length 74: 10.125.1.214.58428 >
10.125.1.85.nfs: Flags [S], seq 1734997801, win 32120, options [mss
1460,sackOK,TS val 1973769186 ecr 0,nop,wscale 7], length 0

2024-05-29 11:47:52.273503 52:54:00:13:1f:34 > 52:54:00:1d:a4:24,
ethertype IPv4 (0x0800), length 74: 10.125.1.85.nfs >
10.125.1.214.58428: Flags [S.], seq 1078843840, ack 1734997802, win
28960, options [mss 1460,sackOK,TS val 2235915769 ecr
1973769186,nop,wscale 7], length 0


Best regards,
Zhitao Li