Skip to contentSkip to conversation

cifs: remove the sockopt= mount option

PATCH In review Jeff Layton 6 participants 11 messages Feb 21, 2013 – Mar 6, 2013 5 messages not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cifs: remove the sockopt= mount option

...as promised for 3.9.

---
 fs/cifs/connect.c | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)
fs/cifs/connect.c +1 -15
index d997737..8609c42 100644
@@ -97,7 +97,7 @@ enum {
 	Opt_user, Opt_pass, Opt_ip,
 	Opt_unc, Opt_domain,
 	Opt_srcaddr, Opt_prefixpath,
-	Opt_iocharset, Opt_sockopt,
+	Opt_iocharset,
 	Opt_netbiosname, Opt_servern,
 	Opt_ver, Opt_vers, Opt_sec, Opt_cache,
 
@@ -202,7 +202,6 @@ static const match_table_t cifs_mount_option_tokens = {
 	{ Opt_srcaddr, "srcaddr=%s" },
 	{ Opt_prefixpath, "prefixpath=%s" },
 	{ Opt_iocharset, "iocharset=%s" },
-	{ Opt_sockopt, "sockopt=%s" },
 	{ Opt_netbiosname, "netbiosname=%s" },
 	{ Opt_servern, "servern=%s" },
 	{ Opt_ver, "ver=%s" },
@@ -1722,19 +1721,6 @@ cifs_parse_mount_options(const char *mountdata, const char *devname,
 			 */
 			cFYI(1, "iocharset set to %s", string);
 			break;
-		case Opt_sockopt:
-			string = match_strdup(args);
-			if (string == NULL)
-				goto out_nomem;
-
-			if (strnicmp(string, "TCP_NODELAY", 11) == 0) {
-				printk(KERN_WARNING "CIFS: the "
-					"sockopt=TCP_NODELAY option has been "
-					"deprecated and will be removed "
-					"in 3.9\n");
-				vol->sockopt_tcp_nodelay = 1;
-			}
-			break;
 		case Opt_netbiosname:
 			string = match_strdup(args);
 			if (string == NULL)
Signature · 1 line
1.7.11.7

A message that is not in the archive (sent off-list or not yet received): 1361446379-7970-1-git-send-email-jlayton-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org

#2 lore raw
Quoted 44 lines · Jeff Layton

...as promised for 3.9.

---
 fs/cifs/connect.c | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)
fs/cifs/connect.c +1 -2
index d997737..8609c42 100644
@@ -97,7 +97,7 @@ enum {
        Opt_user, Opt_pass, Opt_ip,
        Opt_unc, Opt_domain,
        Opt_srcaddr, Opt_prefixpath,
-       Opt_iocharset, Opt_sockopt,
+       Opt_iocharset,
        Opt_netbiosname, Opt_servern,
        Opt_ver, Opt_vers, Opt_sec, Opt_cache,
 
@@ -202,7 +202,6 @@ static const match_table_t cifs_mount_option_tokens =
 {
        { Opt_srcaddr, "srcaddr=%s" },
        { Opt_prefixpath, "prefixpath=%s" },
        { Opt_iocharset, "iocharset=%s" },
-       { Opt_sockopt, "sockopt=%s" },
        { Opt_netbiosname, "netbiosname=%s" },
        { Opt_servern, "servern=%s" },
        { Opt_ver, "ver=%s" },
fs/cifs/connect.c +0 -8
@@ -1722,19 +1721,6 @@ cifs_parse_mount_options(const char *mountdata,
 const char *devname,
                         */
                        cFYI(1, "iocharset set to %s", string);
                        break;
-               case Opt_sockopt:
-                       string = match_strdup(args);
-                       if (string == NULL)
-                               goto out_nomem;
-
-                       if (strnicmp(string, "TCP_NODELAY", 11) == 0) {
-                               printk(KERN_WARNING "CIFS: the "
-                                       "sockopt=TCP_NODELAY option has

been "

fs/cifs/connect.c cifs_parse_mount_options(const char *mountdata, const char *devname, · line 1726

Quoted 8 lines
fs/cifs/connect.c +0 -5
-                                       "deprecated and will be removed "
-                                       "in 3.9\n");
-                               vol->sockopt_tcp_nodelay = 1;
-                       }
-                       break;
                case Opt_netbiosname:
                        string = match_strdup(args);
                        if (string == NULL)

Gar. Why was TCP_NODELAY pulled? I only ask because I'm updating the socket options section of the smb.conf man page. Supposedly NODELAY can have great results in some network configurations[citation needed].

--
Peace and Blessings,
-Scott.

A message that is not in the archive (sent off-list or not yet received): CAFB9KM0bJo4NpePoe9aWMZGJ_jfk0W41Thefg7MD=_V+VP98Eg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#4 lore raw

We should do some quick performance runs to verify this - but with corking and uncorking the socket explicitly presumably TCP_NODELAY should not be needed.

Quoted 62 lines · Scott Lovenberg
Quoted 44 lines · Jeff Layton

...as promised for 3.9.

---
 fs/cifs/connect.c | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)
fs/cifs/connect.c +1 -2
index d997737..8609c42 100644
@@ -97,7 +97,7 @@ enum {
        Opt_user, Opt_pass, Opt_ip,
        Opt_unc, Opt_domain,
        Opt_srcaddr, Opt_prefixpath,
-       Opt_iocharset, Opt_sockopt,
+       Opt_iocharset,
        Opt_netbiosname, Opt_servern,
        Opt_ver, Opt_vers, Opt_sec, Opt_cache,
 
@@ -202,7 +202,6 @@ static const match_table_t cifs_mount_option_tokens =
 {
        { Opt_srcaddr, "srcaddr=%s" },
        { Opt_prefixpath, "prefixpath=%s" },
        { Opt_iocharset, "iocharset=%s" },
-       { Opt_sockopt, "sockopt=%s" },
        { Opt_netbiosname, "netbiosname=%s" },
        { Opt_servern, "servern=%s" },
        { Opt_ver, "ver=%s" },
fs/cifs/connect.c +0 -8
@@ -1722,19 +1721,6 @@ cifs_parse_mount_options(const char *mountdata,
 const char *devname,
                         */
                        cFYI(1, "iocharset set to %s", string);
                        break;
-               case Opt_sockopt:
-                       string = match_strdup(args);
-                       if (string == NULL)
-                               goto out_nomem;
-
-                       if (strnicmp(string, "TCP_NODELAY", 11) == 0) {
-                               printk(KERN_WARNING "CIFS: the "
-                                       "sockopt=TCP_NODELAY option has

been "

fs/cifs/connect.c cifs_parse_mount_options(const char *mountdata, const char *devname, · line 1726

Quoted 8 lines
fs/cifs/connect.c +0 -5
-                                       "deprecated and will be removed "
-                                       "in 3.9\n");
-                               vol->sockopt_tcp_nodelay = 1;
-                       }
-                       break;
                case Opt_netbiosname:
                        string = match_strdup(args);
                        if (string == NULL)

Gar. Why was TCP_NODELAY pulled? I only ask because I'm updating the socket options section of the smb.conf man page. Supposedly NODELAY can have great results in some network configurations[citation needed].

--
Peace and Blessings,
-Scott.

Signature · 3 lines
Thanks,

Steve

A message that is not in the archive (sent off-list or not yet received): CAH2r5mtHsOjXjXx5NdHWNzoTmVmny8gSh_Gd1XfHdNfCOHJDyA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#6 lore raw
Quoted 4 lines · On Thu, 21 Feb 2013 10:00:07 -0600 Steve French…

We should do some quick performance runs to verify this - but with corking and uncorking the socket explicitly presumably TCP_NODELAY should not be needed.

I would think time for that sort of investigation was before you merged the patch that said we were going to remove it. I assumed that since you took that patch, you were OK with it.

In any case, I'll turn this question around...

TCP_NODELAY says: Send frames as soon as possible after a sendmsg (or kernel_sendmsg in this case)

TCP_CORK says: hold sending until I uncork the socket

If the socket is CORK'ed then that overrides TCP_NODELAY (at least in my reading of the code). Since we always cork/uncork the socket now, what possible benefit could there be to keeping the sockopt=TCP_NODELAY option around?

Quoted 67 lines
Quoted 62 lines · Scott Lovenberg
Quoted 44 lines · Jeff Layton

...as promised for 3.9.

---
 fs/cifs/connect.c | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)
fs/cifs/connect.c +1 -2
index d997737..8609c42 100644
@@ -97,7 +97,7 @@ enum {
        Opt_user, Opt_pass, Opt_ip,
        Opt_unc, Opt_domain,
        Opt_srcaddr, Opt_prefixpath,
-       Opt_iocharset, Opt_sockopt,
+       Opt_iocharset,
        Opt_netbiosname, Opt_servern,
        Opt_ver, Opt_vers, Opt_sec, Opt_cache,
 
@@ -202,7 +202,6 @@ static const match_table_t cifs_mount_option_tokens =
 {
        { Opt_srcaddr, "srcaddr=%s" },
        { Opt_prefixpath, "prefixpath=%s" },
        { Opt_iocharset, "iocharset=%s" },
-       { Opt_sockopt, "sockopt=%s" },
        { Opt_netbiosname, "netbiosname=%s" },
        { Opt_servern, "servern=%s" },
        { Opt_ver, "ver=%s" },
fs/cifs/connect.c +0 -8
@@ -1722,19 +1721,6 @@ cifs_parse_mount_options(const char *mountdata,
 const char *devname,
                         */
                        cFYI(1, "iocharset set to %s", string);
                        break;
-               case Opt_sockopt:
-                       string = match_strdup(args);
-                       if (string == NULL)
-                               goto out_nomem;
-
-                       if (strnicmp(string, "TCP_NODELAY", 11) == 0) {
-                               printk(KERN_WARNING "CIFS: the "
-                                       "sockopt=TCP_NODELAY option has

been "

fs/cifs/connect.c cifs_parse_mount_options(const char *mountdata, const char *devname, · line 1726

Quoted 8 lines
fs/cifs/connect.c +0 -5
-                                       "deprecated and will be removed "
-                                       "in 3.9\n");
-                               vol->sockopt_tcp_nodelay = 1;
-                       }
-                       break;
                case Opt_netbiosname:
                        string = match_strdup(args);
                        if (string == NULL)

Gar. Why was TCP_NODELAY pulled? I only ask because I'm updating the socket options section of the smb.conf man page. Supposedly NODELAY can have great results in some network configurations[citation needed].

--
Peace and Blessings,
-Scott.

Signature · 1 line
Jeff Layton <[email protected]>

A message that is not in the archive (sent off-list or not yet received): 20130221112719.2ed40ecf-9yPaYZwiELC+kQycOl6kW4xkIHaj4LzF@public.gmane.org

#8 lore raw
Quoted 11 lines · Jeff Layton
Quoted 4 lines · On Thu, 21 Feb 2013 10:00:07 -0600 Steve French…

We should do some quick performance runs to verify this - but with corking and uncorking the socket explicitly presumably TCP_NODELAY should not be needed.

I would think time for that sort of investigation was before you merged the patch that said we were going to remove it. I assumed that since you took that patch, you were OK with it.

Yes - I am ok with it and it didn't seem to hurt performance. Corking/uncorking explicitly makes more sense if the complex network layers work as expected - but the problem is that they change so best to be safe and at least try some quick tests.

Quoted 11 lines

In any case, I'll turn this question around...

TCP_NODELAY says: Send frames as soon as possible after a sendmsg (or kernel_sendmsg in this case)

TCP_CORK says: hold sending until I uncork the socket

If the socket is CORK'ed then that overrides TCP_NODELAY (at least in my reading of the code). Since we always cork/uncork the socket now, what possible benefit could there be to keeping the sockopt=TCP_NODELAY option around?

None if corking/uncorking works as one would intuitively think - but the tcp network layers under us are very complicated.

(I am not objecting to the patch to remove sockopt setting - just want to do the simple checks to make sure we don't miss something)

Signature · 3 lines
Thanks,

Steve

A message that is not in the archive (sent off-list or not yet received): CAH2r5mszydSWqVsGTT3cqe_J2oK+k=4=Ly+hiiPs9eNWsZaVZw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#10 lore raw
Quoted 30 lines · On Thu, 21 Feb 2013 10:31:24 -0600 Steve French…
Quoted 11 lines · Jeff Layton
Quoted 4 lines · On Thu, 21 Feb 2013 10:00:07 -0600 Steve French…

We should do some quick performance runs to verify this - but with corking and uncorking the socket explicitly presumably TCP_NODELAY should not be needed.

I would think time for that sort of investigation was before you merged the patch that said we were going to remove it. I assumed that since you took that patch, you were OK with it.

Yes - I am ok with it and it didn't seem to hurt performance. Corking/uncorking explicitly makes more sense if the complex network layers work as expected - but the problem is that they change so best to be safe and at least try some quick tests.

Quoted 11 lines

In any case, I'll turn this question around...

TCP_NODELAY says: Send frames as soon as possible after a sendmsg (or kernel_sendmsg in this case)

TCP_CORK says: hold sending until I uncork the socket

If the socket is CORK'ed then that overrides TCP_NODELAY (at least in my reading of the code). Since we always cork/uncork the socket now, what possible benefit could there be to keeping the sockopt=TCP_NODELAY option around?

Quoted 6 lines

None if corking/uncorking works as one would intuitively think - but the tcp network layers under us are very complicated.

(I am not objecting to the patch to remove sockopt setting - just want to do the simple checks to make sure we don't miss something)

Fair enough, let me know what you find.

Thanks,

Signature · 1 line
Jeff Layton <[email protected]>
#11 lore raw
In reply to a message not in the archive
Quoted 3 lines · Steve French

We should do some quick performance runs to verify this - but with corking and uncorking the socket explicitly presumably TCP_NODELAY should not be needed.

I got that info from the old Samba HOWTO and Reference Guide, but it's very, very old info. I'm guessing that when matched with an older kernel and possibly older TCP/IP congestion algorithm, it might have made more of a difference than it does now.

author Acked-by Pavel Shilovsky
#12 lore raw
In reply to a message not in the archive
Quoted 50 lines · On Thu, 21 Feb 2013 10:37:21 -0500 Scott Lovenberg…
Quoted 49 lines · Jeff Layton

...as promised for 3.9.

---
 fs/cifs/connect.c | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)
fs/cifs/connect.c +1 -2
index d997737..8609c42 100644
@@ -97,7 +97,7 @@ enum {
        Opt_user, Opt_pass, Opt_ip,
        Opt_unc, Opt_domain,
        Opt_srcaddr, Opt_prefixpath,
-       Opt_iocharset, Opt_sockopt,
+       Opt_iocharset,
        Opt_netbiosname, Opt_servern,
        Opt_ver, Opt_vers, Opt_sec, Opt_cache,
 
@@ -202,7 +202,6 @@ static const match_table_t cifs_mount_option_tokens =
 {
        { Opt_srcaddr, "srcaddr=%s" },
        { Opt_prefixpath, "prefixpath=%s" },
        { Opt_iocharset, "iocharset=%s" },
-       { Opt_sockopt, "sockopt=%s" },
        { Opt_netbiosname, "netbiosname=%s" },
        { Opt_servern, "servern=%s" },
        { Opt_ver, "ver=%s" },
fs/cifs/connect.c +0 -8
@@ -1722,19 +1721,6 @@ cifs_parse_mount_options(const char *mountdata,
 const char *devname,
                         */
                        cFYI(1, "iocharset set to %s", string);
                        break;
-               case Opt_sockopt:
-                       string = match_strdup(args);
-                       if (string == NULL)
-                               goto out_nomem;
-
-                       if (strnicmp(string, "TCP_NODELAY", 11) == 0) {
-                               printk(KERN_WARNING "CIFS: the "
-                                       "sockopt=TCP_NODELAY option has

been "

fs/cifs/connect.c +0 -5
-                                       "deprecated and will be removed "
-                                       "in 3.9\n");
-                               vol->sockopt_tcp_nodelay = 1;
-                       }
-                       break;
Quoted 9 lines
Quoted 3 lines
                case Opt_netbiosname:
                        string = match_strdup(args);
                        if (string == NULL)

Gar. Why was TCP_NODELAY pulled? I only ask because I'm updating the socket options section of the smb.conf man page. Supposedly NODELAY can have great results in some network configurations[citation needed].

commit b8eed28375a43e1c9aaa9d15af2a052aae0d0725

Author: Jeff Layton <[email protected]>
Date:   Tue Sep 18 16:20:35 2012 -0700

cifs: cork the socket before a send and uncork it afterward

We want to send SMBs as "atomically" as possible. Prior to sending any data on the socket, cork it to make sure that no non-full frames go out. Afterward, uncork it to make sure all of the data gets pushed out to the wire.

Note that this more or less renders the socket=TCP_NODELAY mount option obsolete. When TCP_CORK and TCP_NODELAY are used on the same socket, TCP_NODELAY is essentially ignored.

Signature · 1 line
Jeff Layton <[email protected]>
#13 lore raw
In reply to Jeff Layton

Resending patch to a slightly broader list for last minute check if anyone objects. Although setting this particular socket option (TCP_NODELAY) may not be as useful when corking/uncorking explicitly, I want to doublecheck before removing them because there has been some utility to the server allowing override of various sockopt options. Samba server has long supported at least the following set of settable socket options (although I don't know if the defaults are frequently overridden now, by setting sockopts in smb.conf as used to be common for the server).

SO_KEEPALIVE
SO_REUSEADDR
SO_BROADCAST
TCP_NODELAY
IPTOS_LOWDELAY
IPTOS_THROUGHPUT
SO_SNDBUF *
SO_RCVBUF *
SO_SNDLOWAT *
SO_RCVLOWAT *

* takes an integer argument rather than a boolean on/off

Any objections to removing the ability to set socket options explicitly for the cifs network file system client?

Quoted 53 lines · Jeff Layton

...as promised for 3.9.

---
 fs/cifs/connect.c | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)
fs/cifs/connect.c +1 -2
index d997737..8609c42 100644
@@ -97,7 +97,7 @@ enum {
        Opt_user, Opt_pass, Opt_ip,
        Opt_unc, Opt_domain,
        Opt_srcaddr, Opt_prefixpath,
-       Opt_iocharset, Opt_sockopt,
+       Opt_iocharset,
        Opt_netbiosname, Opt_servern,
        Opt_ver, Opt_vers, Opt_sec, Opt_cache,
 
@@ -202,7 +202,6 @@ static const match_table_t cifs_mount_option_tokens =
 {
        { Opt_srcaddr, "srcaddr=%s" },
        { Opt_prefixpath, "prefixpath=%s" },
        { Opt_iocharset, "iocharset=%s" },
-       { Opt_sockopt, "sockopt=%s" },
        { Opt_netbiosname, "netbiosname=%s" },
        { Opt_servern, "servern=%s" },
        { Opt_ver, "ver=%s" },
fs/cifs/connect.c +0 -8
@@ -1722,19 +1721,6 @@ cifs_parse_mount_options(const char *mountdata,
 const char *devname,
                         */
                        cFYI(1, "iocharset set to %s", string);
                        break;
-               case Opt_sockopt:
-                       string = match_strdup(args);
-                       if (string == NULL)
-                               goto out_nomem;
-
-                       if (strnicmp(string, "TCP_NODELAY", 11) == 0) {
-                               printk(KERN_WARNING "CIFS: the "
-                                       "sockopt=TCP_NODELAY option has

been "

fs/cifs/connect.c +0 -6
-                                       "deprecated and will be removed "
-                                       "in 3.9\n");
-                               vol->sockopt_tcp_nodelay = 1;
-                       }
-                       break;
                case Opt_netbiosname:
                        string = match_strdup(args);
                        if (string == NULL)
--
Quoted 2 lines

1.7.11.7

--
Thanks,

Steve

#14 lore raw
Quoted 21 lines · On Mon, 4 Mar 2013 16:08:30 -0600 Steve French…

Resending patch to a slightly broader list for last minute check if anyone objects. Although setting this particular socket option (TCP_NODELAY) may not be as useful when corking/uncorking explicitly, I want to doublecheck before removing them because there has been some utility to the server allowing override of various sockopt options. Samba server has long supported at least the following set of settable socket options (although I don't know if the defaults are frequently overridden now, by setting sockopts in smb.conf as used to be common for the server).

SO_KEEPALIVE
SO_REUSEADDR
SO_BROADCAST
TCP_NODELAY
IPTOS_LOWDELAY
IPTOS_THROUGHPUT
SO_SNDBUF *
SO_RCVBUF *
SO_SNDLOWAT *
SO_RCVLOWAT *

Quoted 5 lines

* takes an integer argument rather than a boolean on/off

Any objections to removing the ability to set socket options explicitly for the cifs network file system client?

A couple of points...

The sockopt= option was never documented in the mount.cifs manpage and the only value it ever accepted was TCP_NODELAY. Now that we're explicitly corking the socket, TCP_NODELAY has no effect. I don't think there's any value in leaving in a "placeholder" socket= option.

Signature · 1 line
Jeff Layton <[email protected]>
#15 lore raw
Quoted 30 lines · Jeff Layton
Quoted 21 lines · On Mon, 4 Mar 2013 16:08:30 -0600 Steve French…

Resending patch to a slightly broader list for last minute check if anyone objects. Although setting this particular socket option (TCP_NODELAY) may not be as useful when corking/uncorking explicitly, I want to doublecheck before removing them because there has been some utility to the server allowing override of various sockopt options. Samba server has long supported at least the following set of settable socket options (although I don't know if the defaults are frequently overridden now, by setting sockopts in smb.conf as used to be common for the server).

SO_KEEPALIVE
SO_REUSEADDR
SO_BROADCAST
TCP_NODELAY
IPTOS_LOWDELAY
IPTOS_THROUGHPUT
SO_SNDBUF *
SO_RCVBUF *
SO_SNDLOWAT *
SO_RCVLOWAT *

Quoted 5 lines

* takes an integer argument rather than a boolean on/off

Any objections to removing the ability to set socket options explicitly for the cifs network file system client?

Quoted 7 lines

A couple of points...

The sockopt= option was never documented in the mount.cifs manpage and the only value it ever accepted was TCP_NODELAY. Now that we're explicitly corking the socket, TCP_NODELAY has no effect. I don't think there's any value in leaving in a "placeholder" socket= option.

After doing some research, I agree. I've been updating the socket options section of the Samba smb.conf man page and the performance section of the Samba HOWTO. The more research I do the more convinced that there is almost no reason to set most socket options on a modern Linux kernel or in Samba (server or client side).

One thing pointed out to me in another thread is that the smbd server sets TCP_NODELAY by default. AIUI this was because the TCP_NODELAY option used to almost double the speed of SMB in some cases due to the way Microsoft's old TCP/IP stack sent ACKs.

This is all to say that I think the "sockopt" option can go altogether since it's unlikely that any socket options will be needed in the future and the only one that was supported is now useless.

Signature · 2 lines
Peace and Blessings,
-Scott.