github-mirror/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-05-30 10:04:04 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity Graph

cifs: add mount parameter tcpnodelay

Browse Source 
Although corking and uncorking the socket (which cifs.ko already
does) should usually have the desired benefit, using the new
tcpnodelay mount option causes tcp_sock_set_nodelay() to be set
on the socket which may be useful in order to ensure that we don't
ever have cases where the network stack is waiting on sending an
SMB request until multiple SMB requests have been added to the
send queue (since this could lead to long latencies).

To enable it simply append "tcpnodelay" it to the mount options

Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>
This commit is contained in:
Steve French 2021-10-14 15:54:26 -05:00
parent 7be3248f31
commit 7ae5e588b0
2 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
fs/cifs/fs_context.c
View File

@ -116,6 +116,7 @@ const struct fs_parameter_spec smb3_fs_parameters[] = {
fsparam_flag("nosharesock", Opt_nosharesock),
fsparam_flag_no("persistenthandles", Opt_persistent),
fsparam_flag_no("resilienthandles", Opt_resilient),
fsparam_flag_no("tcpnodelay", Opt_tcp_nodelay),
fsparam_flag("domainauto", Opt_domainauto),
fsparam_flag("rdma", Opt_rdma),
fsparam_flag("modesid", Opt_modesid),
@ -1389,6 +1390,13 @@ static int smb3_fs_context_parse_param(struct fs_context *fc,
}
}
break;
case Opt_tcp_nodelay:
/* tcp nodelay should not usually be needed since we CORK/UNCORK the socket */
if (result.negated)
ctx->sockopt_tcp_nodelay = false;
else
ctx->sockopt_tcp_nodelay = true;
break;
case Opt_domainauto:
ctx->domainauto = true;
break;

1
fs/cifs/fs_context.h
View File

@ -98,6 +98,7 @@ enum cifs_param {
Opt_nosharesock,
Opt_persistent,
Opt_resilient,
Opt_tcp_nodelay,
Opt_domainauto,
Opt_rdma,
Opt_modesid,