Skip to contentSkip to conversation

NetLoop initialization bug

PATCH New Michael Zaidman 5 participants 18 messages Mar 22, 2009 – Apr 3, 2009 Networking

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 NetLoop initialization bug

[U-Boot] [PATCH] NetLoop initialization bug

Upon u-boot's start up the first ping command causes a failure of the consequent TFTP command. It happens in the recently added mechanism of the NetLoop initialization where initialization of global network parameters is separated in the NetInitLoop routine which is called per env_id change. Thus, ping request will initialize the network parameters necessary for ping operation only, afterwards the env_changed_id will be set to the env_id that will prevent all following initialization requests from other protocols. The problem is that the initialized by ping subset of network parameters is not sufficient for other protocols and particularly for TFTP
which requires the NetServerIp also.

Signed-off-by michael.zaidman at kodak.com
net/net.c +8 -58
index a89f6a0..dc98d0f 100644
@@ -288,64 +288,13 @@ NetInitLoop(proto_t protocol)
 	if (env_changed_id == env_id)
 		return 0;
 
-	switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-	case NFS:
-#endif
-#if defined(CONFIG_CMD_PING)
-	case PING:
-#endif
-#if defined(CONFIG_CMD_SNTP)
-	case SNTP:
-#endif
-	case NETCONS:
-	case TFTP:
-		NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
-		NetOurGatewayIP = getenv_IPaddr ("gatewayip");
-		NetOurSubnetMask= getenv_IPaddr ("netmask");
-		NetOurVLAN = getenv_VLAN("vlan");
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
+	NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
+	NetOurGatewayIP = getenv_IPaddr ("gatewayip");
+	NetOurSubnetMask= getenv_IPaddr ("netmask");
+	NetServerIP = getenv_IPaddr ("serverip");
+	NetOurNativeVLAN = getenv_VLAN("nvlan");
+	NetOurVLAN = getenv_VLAN("vlan");
 
-		switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-		case NFS:
-#endif
-		case NETCONS:
-		case TFTP:
-			NetServerIP = getenv_IPaddr ("serverip");
-			break;
-#if defined(CONFIG_CMD_PING)
-		case PING:
-			/* nothing */
-			break;
-#endif
-#if defined(CONFIG_CMD_SNTP)
-		case SNTP:
-			/* nothing */
-			break;
-#endif
-		default:
-			break;
-		}
-
-		break;
-	case BOOTP:
-	case RARP:
-		/*
-		 * initialize our IP addr to 0 in order to accept ANY
-		 * IP addr assigned to us by the BOOTP / RARP server
-		 */
-		NetOurIP = 0;
-		NetServerIP = getenv_IPaddr ("serverip");
-		NetOurVLAN = getenv_VLAN("vlan");	/* VLANs must be read */
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
-	case CDP:
-		NetOurVLAN = getenv_VLAN("vlan");	/* VLANs must be read */
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
-		break;
-	default:
-		break;
-	}
 	env_changed_id = env_id;
 	return 0;
 }
@@ -443,18 +392,19 @@ restart:
 			/* Start with a clean slate... */
 			BootpTry = 0;
 			NetOurIP = 0;
-			NetServerIP = getenv_IPaddr ("serverip");
 			DhcpRequest();		/* Basically same as BOOTP */
 			break;
 #endif
 
 		case BOOTP:
 			BootpTry = 0;
+			NetOurIP = 0;
 			BootpRequest ();
 			break;
 
 		case RARP:
 			RarpTry = 0;
+			NetOurIP = 0;
 			RarpRequest ();
 			break;
 #if defined(CONFIG_CMD_PING)

Regards,
Michael Zaidman,
michael.zaidman at kodak.com

#1 lore raw

PATCH 1/1 NetLoop initialization bug

Hi Heiko,

The patch "netloop: speed up NetLoop" you delivered into the u-boot-2009.03 introduced bug I have described below a few days ago. Could you please take a look at the proposed fix?

Thanks,
Michael

---------- Forwarded message ----------
From: Michael Zaidman <[email protected]>
Date: Sun, Mar 22, 2009 at 8:35 PM
Subject: [U-Boot] [PATCH] NetLoop initialization bug
To: u-boot at lists.denx.de

[U-Boot] [PATCH] NetLoop initialization bug

Upon u-boot's start up the first ping command causes a failure of the consequent TFTP command. It happens in the recently added mechanism of the NetLoop initialization where initialization of global network parameters is separated in the NetInitLoop routine which is called per env_id change. Thus, ping request will initialize the network parameters necessary for ping operation only, afterwards the env_changed_id will be set to the env_id that will prevent all following initialization requests from other protocols. The problem is that the initialized by ping subset of network parameters is not sufficient for other protocols and particularly for TFTP
which requires the NetServerIp also.

Signed-off-by michael.zaidman at gmail.com
net/net.c +0 -0
index a89f6a0..dc98d0f 100644
@@ -288,64 +288,13 @@ NetInitLoop(proto_t protocol)

? ? ? ?if (env_changed_id == env_id)
? ? ? ? ? ? ? ?return 0;

- ? ? ? switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
- ? ? ? case NFS:
-#endif
-#if defined(CONFIG_CMD_PING)
- ? ? ? case PING:
-#endif
-#if defined(CONFIG_CMD_SNTP)
- ? ? ? case SNTP:
-#endif
- ? ? ? case NETCONS:
- ? ? ? case TFTP:
- ? ? ? ? ? ? ? NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
- ? ? ? ? ? ? ? NetOurGatewayIP = getenv_IPaddr ("gatewayip");
- ? ? ? ? ? ? ? NetOurSubnetMask= getenv_IPaddr ("netmask");
- ? ? ? ? ? ? ? NetOurVLAN = getenv_VLAN("vlan");
- ? ? ? ? ? ? ? NetOurNativeVLAN = getenv_VLAN("nvlan");
+ ? ? ? NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
+ ? ? ? NetOurGatewayIP = getenv_IPaddr ("gatewayip");
+ ? ? ? NetOurSubnetMask= getenv_IPaddr ("netmask");
+ ? ? ? NetServerIP = getenv_IPaddr ("serverip");
+ ? ? ? NetOurNativeVLAN = getenv_VLAN("nvlan");
+ ? ? ? NetOurVLAN = getenv_VLAN("vlan");

- ? ? ? ? ? ? ? switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
- ? ? ? ? ? ? ? case NFS:
-#endif
- ? ? ? ? ? ? ? case NETCONS:
- ? ? ? ? ? ? ? case TFTP:
- ? ? ? ? ? ? ? ? ? ? ? NetServerIP = getenv_IPaddr ("serverip");
- ? ? ? ? ? ? ? ? ? ? ? break;
-#if defined(CONFIG_CMD_PING)
- ? ? ? ? ? ? ? case PING:
- ? ? ? ? ? ? ? ? ? ? ? /* nothing */
- ? ? ? ? ? ? ? ? ? ? ? break;
-#endif
-#if defined(CONFIG_CMD_SNTP)
- ? ? ? ? ? ? ? case SNTP:
- ? ? ? ? ? ? ? ? ? ? ? /* nothing */
- ? ? ? ? ? ? ? ? ? ? ? break;
-#endif
- ? ? ? ? ? ? ? default:
- ? ? ? ? ? ? ? ? ? ? ? break;
- ? ? ? ? ? ? ? }
-
- ? ? ? ? ? ? ? break;
- ? ? ? case BOOTP:
- ? ? ? case RARP:
- ? ? ? ? ? ? ? /*
- ? ? ? ? ? ? ? ?* initialize our IP addr to 0 in order to accept ANY
- ? ? ? ? ? ? ? ?* IP addr assigned to us by the BOOTP / RARP server
- ? ? ? ? ? ? ? ?*/
- ? ? ? ? ? ? ? NetOurIP = 0;
- ? ? ? ? ? ? ? NetServerIP = getenv_IPaddr ("serverip");
- ? ? ? ? ? ? ? NetOurVLAN = getenv_VLAN("vlan"); ? ? ? /* VLANs must be read */
- ? ? ? ? ? ? ? NetOurNativeVLAN = getenv_VLAN("nvlan");
- ? ? ? case CDP:
- ? ? ? ? ? ? ? NetOurVLAN = getenv_VLAN("vlan"); ? ? ? /* VLANs must be read */
- ? ? ? ? ? ? ? NetOurNativeVLAN = getenv_VLAN("nvlan");
- ? ? ? ? ? ? ? break;
- ? ? ? default:
- ? ? ? ? ? ? ? break;
- ? ? ? }
? ? ? ?env_changed_id = env_id;
? ? ? ?return 0;
?}
@@ -443,18 +392,19 @@ restart:
? ? ? ? ? ? ? ? ? ? ? ?/* Start with a clean slate... */
? ? ? ? ? ? ? ? ? ? ? ?BootpTry = 0;
? ? ? ? ? ? ? ? ? ? ? ?NetOurIP = 0;
- ? ? ? ? ? ? ? ? ? ? ? NetServerIP = getenv_IPaddr ("serverip");
? ? ? ? ? ? ? ? ? ? ? ?DhcpRequest(); ? ? ? ? ?/* Basically same as BOOTP */ ? ? ? ? ? ? ? ? ? ? ? ?break;
?#endif

? ? ? ? ? ? ? ?case BOOTP:
? ? ? ? ? ? ? ? ? ? ? ?BootpTry = 0;
+ ? ? ? ? ? ? ? ? ? ? ? NetOurIP = 0;
? ? ? ? ? ? ? ? ? ? ? ?BootpRequest ();
? ? ? ? ? ? ? ? ? ? ? ?break;

? ? ? ? ? ? ? ?case RARP:
? ? ? ? ? ? ? ? ? ? ? ?RarpTry = 0;
+ ? ? ? ? ? ? ? ? ? ? ? NetOurIP = 0;
? ? ? ? ? ? ? ? ? ? ? ?RarpRequest ();
? ? ? ? ? ? ? ? ? ? ? ?break;
?#if defined(CONFIG_CMD_PING)

Regards,
Michael

#2 lore raw

PATCH 1/1 NetLoop initialization bug

Hello Michael,

Quoted 7 lines · Michael Zaidman

Hi Heiko,

The patch "netloop: speed up NetLoop" you delivered into the u-boot-2009.03 introduced bug I have described below a few days ago. Could you please take a look at the proposed fix?

Sorry, seems I missed it.

Quoted 12 lines

Thanks,
Michael

---------- Forwarded message ----------
From: Michael Zaidman <[email protected]>
Date: Sun, Mar 22, 2009 at 8:35 PM
Subject: [U-Boot] [PATCH] NetLoop initialization bug
To: u-boot at lists.denx.de

[U-Boot] [PATCH] NetLoop initialization bug

Thanks for catching this.
Your patch doesn;t apply, I get "malformed patch". Before sending a new one, please have a look at my comments:

Quoted 35 lines

Upon u-boot's start up the first ping command causes a failure of the consequent TFTP command. It happens in the recently added mechanism of the NetLoop initialization where initialization of global network parameters is separated in the NetInitLoop routine which is called per env_id change. Thus, ping request will initialize the network parameters necessary for ping operation only, afterwards the env_changed_id will be set to the env_id that will prevent all following initialization requests from other protocols. The problem is that the initialized by ping subset of network parameters is not sufficient for other protocols and particularly for TFTP
which requires the NetServerIp also.

Signed-off-by michael.zaidman at gmail.com
net/net.c +0 -12
index a89f6a0..dc98d0f 100644
@@ -288,64 +288,13 @@ NetInitLoop(proto_t protocol)
       if (env_changed_id == env_id)
               return 0;
 
-       switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-       case NFS:
-#endif
-#if defined(CONFIG_CMD_PING)
-       case PING:
-#endif
-#if defined(CONFIG_CMD_SNTP)
-       case SNTP:
-#endif
-       case NETCONS:
-       case TFTP:

net/net.c NetInitLoop(proto_t protocol) · line 295

Quoted 10 lines
net/net.c +4 -5
-               NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
-               NetOurGatewayIP = getenv_IPaddr ("gatewayip");
-               NetOurSubnetMask= getenv_IPaddr ("netmask");
-               NetOurVLAN = getenv_VLAN("vlan");
-               NetOurNativeVLAN = getenv_VLAN("nvlan");
+       NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
+       NetOurGatewayIP = getenv_IPaddr ("gatewayip");
+       NetOurSubnetMask= getenv_IPaddr ("netmask");
+       NetServerIP = getenv_IPaddr ("serverip");
 

The following 2 vars are just used, if CONFIG_CMD_CDP is used, can we do a "#if defined" around it?

Quoted 3 lines
net/net.c +2 -0
+       NetOurNativeVLAN = getenv_VLAN("nvlan");
+       NetOurVLAN = getenv_VLAN("vlan");
 

[...]

net/net.c restart: · line 402

Quoted 13 lines
net/net.c +1 -1
@@ -443,18 +392,19 @@ restart:
                       /* Start with a clean slate... */
                       BootpTry = 0;
                       NetOurIP = 0;
-                       NetServerIP = getenv_IPaddr ("serverip");
                       DhcpRequest();          /* Basically same as BOOTP */
                       break;
 #endif
 
               case BOOTP:
                       BootpTry = 0;
+                       NetOurIP = 0;
 

why we need this here?

net/net.c restart: · line 409

Quoted 7 lines
net/net.c +1 -0
                       BootpRequest ();
                       break;
 
               case RARP:
                       RarpTry = 0;
+                       NetOurIP = 0;
 

also here?
Has this to do something with the bugfix?

bye
Heiko

Signature · 2 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany 
#3 lore raw

PATCH 1/1 NetLoop initialization bug

Dear Heiko,

Quoted 5 lines · In message <[email protected]> you wrote:

The following 2 vars are just used, if CONFIG_CMD_CDP is used, can we do a "#if defined" around it?

Quoted 2 lines
(quoted diff) +2 -0
+       NetOurNativeVLAN = getenv_VLAN("nvlan");
+       NetOurVLAN = getenv_VLAN("vlan");

We probably could, but we then had to do this in several places, and the code would become less readable, I think.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Q:  How many DEC repairman does it take to fix a flat ?
A:  Five; four to hold the car up and one to swap tires.
#4 lore raw

PATCH 1/1 NetLoop initialization bug

In reply to Heiko Schocher

Hello Heiko,

Please see my comments and updated patch below.

Quoted 70 lines · Heiko Schocher

Hello Michael,

Quoted 7 lines · Michael Zaidman

Hi Heiko,

The patch "netloop: speed up NetLoop" you delivered into the u-boot-2009.03 introduced bug I have described below a few days ago. Could you please take a look at the proposed fix?

Sorry, seems I missed it.

Quoted 12 lines

Thanks,
Michael

---------- Forwarded message ----------
From: Michael Zaidman <[email protected]>
Date: Sun, Mar 22, 2009 at 8:35 PM
Subject: [U-Boot] [PATCH] NetLoop initialization bug
To: u-boot at lists.denx.de

[U-Boot] [PATCH] NetLoop initialization bug

Thanks for catching this.
Your patch doesn;t apply, I get "malformed patch". Before sending a new one, please have a look at my comments:

Quoted 40 lines

Upon u-boot's start up the first ping command causes a failure of the consequent TFTP command. It happens in the recently added mechanism of the NetLoop initialization where initialization of global network parameters is separated in the NetInitLoop routine which is called per env_id change. Thus, ping request will initialize the network parameters necessary for ping operation only, afterwards the env_changed_id will be set to the env_id that will prevent all following initialization requests from other protocols. The problem is that the initialized by ping subset of network parameters is not sufficient for other protocols and particularly for TFTP
which requires the NetServerIp also.

Signed-off-by michael.zaidman at gmail.com
net/net.c +0 -17
index a89f6a0..dc98d0f 100644
@@ -288,64 +288,13 @@ NetInitLoop(proto_t protocol)
 ? ? ? ?if (env_changed_id == env_id)
 ? ? ? ? ? ? ? ?return 0;
 
- ? ? ? switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
- ? ? ? case NFS:
-#endif
-#if defined(CONFIG_CMD_PING)
- ? ? ? case PING:
-#endif
-#if defined(CONFIG_CMD_SNTP)
- ? ? ? case SNTP:
-#endif
- ? ? ? case NETCONS:
- ? ? ? case TFTP:
- ? ? ? ? ? ? ? NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
- ? ? ? ? ? ? ? NetOurGatewayIP = getenv_IPaddr ("gatewayip");
- ? ? ? ? ? ? ? NetOurSubnetMask= getenv_IPaddr ("netmask");
- ? ? ? ? ? ? ? NetOurVLAN = getenv_VLAN("vlan");
- ? ? ? ? ? ? ? NetOurNativeVLAN = getenv_VLAN("nvlan");
Quoted 10 lines

net/net.c NetInitLoop(proto_t protocol) · line 295

Quoted 5 lines
net/net.c +4 -0
+ ? ? ? NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
+ ? ? ? NetOurGatewayIP = getenv_IPaddr ("gatewayip");
+ ? ? ? NetOurSubnetMask= getenv_IPaddr ("netmask");
+ ? ? ? NetServerIP = getenv_IPaddr ("serverip");
 

The following 2 vars are just used, if CONFIG_CMD_CDP is used, can we do a "#if defined" around it?

Quoted 2 lines
net/net.c +2 -0
+ ? ? ? NetOurNativeVLAN = getenv_VLAN("nvlan");
+ ? ? ? NetOurVLAN = getenv_VLAN("vlan");

These two variables have been initialized in original code for all protocols supported by u-boot. As I can see, at least the NetOurVLAN is used by most of them. This was the reason for keeping them in place.

Quoted 11 lines
Quoted 1 line

[...]

Quoted 9 lines
net/net.c +0 -0
@@ -443,18 +392,19 @@ restart:

? ? ? ? ? ? ? ? ? ? ? ?/* Start with a clean slate... */
? ? ? ? ? ? ? ? ? ? ? ?BootpTry = 0;
? ? ? ? ? ? ? ? ? ? ? ?NetOurIP = 0;
- ? ? ? ? ? ? ? ? ? ? ? NetServerIP = getenv_IPaddr ("serverip"); ? ? ? ? ? ? ? ? ? ? ? ?DhcpRequest(); ? ? ? ? ?/* Basically same as BOOTP */ ? ? ? ? ? ? ? ? ? ? ? ?break;
?#endif

Quoted 6 lines
Quoted 4 lines

? ? ? ? ? ? ? ?case BOOTP:
? ? ? ? ? ? ? ? ? ? ? ?BootpTry = 0;
+ ? ? ? ? ? ? ? ? ? ? ? NetOurIP = 0;

why we need this here?

Generally, for the same reason the DHCP does - "Start with a clean state..." On the other hand you are right - we do not need to clear the NetOurIP address for BOOTP, DHCP and RARP because the current implementation does not check it. So it has been removed in my new patch below.

Quoted 10 lines
Quoted 8 lines

? ? ? ? ? ? ? ? ? ? ? ?BootpRequest ();
? ? ? ? ? ? ? ? ? ? ? ?break;

? ? ? ? ? ? ? ?case RARP:
? ? ? ? ? ? ? ? ? ? ? ?RarpTry = 0;
+ ? ? ? ? ? ? ? ? ? ? ? NetOurIP = 0;

also here?
Has this to do something with the bugfix?

The same as before.

Quoted 9 lines

bye
Heiko

--
DENX Software Engineering GmbH, ? ? MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany

Here is the updated patch:

Subject: [U-Boot] [PATCH] NetLoop initialization bug

---
 net/net.c |   66 +++++-------------------------------------------------------
 1 files changed, 6 insertions(+), 60 deletions(-)
net/net.c +6 -60
index a89f6a0..91ed0d7 100644
@@ -288,64 +288,13 @@ NetInitLoop(proto_t protocol)
 	if (env_changed_id == env_id)
 		return 0;
 
-	switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-	case NFS:
-#endif
-#if defined(CONFIG_CMD_PING)
-	case PING:
-#endif
-#if defined(CONFIG_CMD_SNTP)
-	case SNTP:
-#endif
-	case NETCONS:
-	case TFTP:
-		NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
-		NetOurGatewayIP = getenv_IPaddr ("gatewayip");
-		NetOurSubnetMask= getenv_IPaddr ("netmask");
-		NetOurVLAN = getenv_VLAN("vlan");
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
+	NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
+	NetOurGatewayIP = getenv_IPaddr ("gatewayip");
+	NetOurSubnetMask= getenv_IPaddr ("netmask");
+	NetServerIP = getenv_IPaddr ("serverip");
+	NetOurNativeVLAN = getenv_VLAN("nvlan");
+	NetOurVLAN = getenv_VLAN("vlan");
 
-		switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-		case NFS:
-#endif
-		case NETCONS:
-		case TFTP:
-			NetServerIP = getenv_IPaddr ("serverip");
-			break;
-#if defined(CONFIG_CMD_PING)
-		case PING:
-			/* nothing */
-			break;
-#endif
-#if defined(CONFIG_CMD_SNTP)
-		case SNTP:
-			/* nothing */
-			break;
-#endif
-		default:
-			break;
-		}
-
-		break;
-	case BOOTP:
-	case RARP:
-		/*
-		 * initialize our IP addr to 0 in order to accept ANY
-		 * IP addr assigned to us by the BOOTP / RARP server
-		 */
-		NetOurIP = 0;
-		NetServerIP = getenv_IPaddr ("serverip");
-		NetOurVLAN = getenv_VLAN("vlan");	/* VLANs must be read */
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
-	case CDP:
-		NetOurVLAN = getenv_VLAN("vlan");	/* VLANs must be read */
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
-		break;
-	default:
-		break;
-	}
 	env_changed_id = env_id;
 	return 0;
 }
@@ -440,10 +389,7 @@ restart:
 
 #if defined(CONFIG_CMD_DHCP)
 		case DHCP:
-			/* Start with a clean slate... */
 			BootpTry = 0;
-			NetOurIP = 0;
-			NetServerIP = getenv_IPaddr ("serverip");
 			DhcpRequest();		/* Basically same as BOOTP */
 			break;
 #endif
Signature · 4 lines
1.5.6.3

Thanks,
Michael
#5 lore raw

PATCH 1/1 NetLoop initialization bug

Hello Michael,

Quoted 6 lines · Michael Zaidman

Please see my comments and updated patch below.

Quoted 3 lines · Heiko Schocher

Hello Michael,

Michael Zaidman wrote:

[...]

Quoted 8 lines
Quoted 4 lines

The following 2 vars are just used, if CONFIG_CMD_CDP is used, can we do a "#if defined" around it?

Quoted 2 lines
net/net.c +2 -0
+       NetOurNativeVLAN = getenv_VLAN("nvlan");
+       NetOurVLAN = getenv_VLAN("vlan");

These two variables have been initialized in original code for all protocols supported by u-boot. As I can see, at least the NetOurVLAN is used by most of them. This was the reason for keeping them in place.

Ok.

Quoted 16 lines
Quoted 10 lines

[...]

Quoted 9 lines
net/net.c +0 -1
@@ -443,18 +392,19 @@ restart:
                       /* Start with a clean slate... */
                       BootpTry = 0;
                       NetOurIP = 0;
-                       NetServerIP = getenv_IPaddr ("serverip");
                       DhcpRequest();          /* Basically same as BOOTP */
                       break;
 #endif
 
Quoted 5 lines

Review restart: · line 402

Quoted 4 lines
(quoted diff) +1 -0
               case BOOTP:
                       BootpTry = 0;
+                       NetOurIP = 0;
 

why we need this here?

Quoted 4 lines

Generally, for the same reason the DHCP does - "Start with a clean state..." On the other hand you are right - we do not need to clear the NetOurIP address for BOOTP, DHCP and RARP because the current implementation does not check it. So it has been removed in my new patch below.

thanks.

Quoted 3 lines

Here is the updated patch:

Subject: [U-Boot] [PATCH] NetLoop initialization bug

Hmm.. can you please add your commit message and your Signed-off-by

thanks
Heiko

Signature · 2 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
#6 lore raw

PATCH 1/1 NetLoop initialization bug

In reply to Michael Zaidman

Hi Michael,

Quoted 1 line

Please see my comments and updated patch below.

As a side note, please send your patch as an inline attachment also adding your signed-off-by line. It's probably easiest to actually use git to apply your changes to a branch and use "git-format-patch" to create the patch.

Please also provide more dsecriptive commit text explaining what the patch does.

The patch in your mail does still not apply. Neither git-am (after fixing the patch with a valid e-mail) nor patch can do anything with it.

Thanks
Detlev

Signature · 7 lines
;; Self-replicator in ELisp
((lambda (l) (prin1-to-string (list l (list (quote quote) l))))
 (quote (lambda (l) (prin1-to-string (list l (list (quote quote) l))))))
--
DENX Software Engineering GmbH,      MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich,  Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: dzu at denx.de
#7 lore raw

PATCH 1/1 NetLoop initialization bug

Quoted 18 lines · Detlev Zundel

Hi Michael,

Quoted 1 line

Please see my comments and updated patch below.

As a side note, please send your patch as an inline attachment also adding your signed-off-by line. ?It's probably easiest to actually use git to apply your changes to a branch and use "git-format-patch" to create the patch.

Please also provide more dsecriptive commit text explaining what the patch does.

The patch in your mail does still not apply. ?Neither git-am (after fixing the patch with a valid e-mail) nor patch can do anything with it.

Thanks
?Detlev

Quoted 9 lines

--
;; Self-replicator in ELisp
((lambda (l) (prin1-to-string (list l (list (quote quote) l)))) ?(quote (lambda (l) (prin1-to-string (list l (list (quote quote) l)))))) --
DENX Software Engineering GmbH, ? ? ?MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, ?Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: dzu at denx.de

Hi Detlev,

Please try this one. I used the git-format-patch for its creation.

Thanks,
Michael

#8 lore raw

PATCH 1/1 NetLoop initialization bug

In reply to Detlev Zundel
Quoted 1 line · Detlev Zundel

Hi Michael,

[...]

Quoted 6 lines

The patch in your mail does still not apply. ?Neither git-am (after fixing the patch with a valid e-mail) nor patch can do anything with it.

Thanks
?Detlev

Sorry for the previous mail,
I have to enable the git operation via gmail on my workstation.

Meanwhile I created the patch again by diff.

Thanks,
Michael

The patch fixes the bug of partial initialization of global network parameters.

Upon u-boot's start up the first ping command causes a failure of the consequent TFTP transfers. It happens in the recently added mechanism of NetLoop initialization where initialization of global network parameters is separated in the NetInitLoop routine which is called per env_id change. Thus, ping request will initialize the network parameters necessary for ping operation only, afterwards the env_changed_id will be set to the env_id that will prevent all following initialization requests from other protocols.
The problem is that the initialized by ping subset of network parameters is not sufficient for other protocols and particularly for TFTP which requires the NetServerIp also.

michael at michaelz:~/develop$ diff -u u-boot-git/net/net.c u-boot-git-test1/net/net.c

net/net.c +6 -60
@@ -288,64 +288,13 @@
 	if (env_changed_id == env_id)
 		return 0;
 
-	switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-	case NFS:
-#endif
-#if defined(CONFIG_CMD_PING)
-	case PING:
-#endif
-#if defined(CONFIG_CMD_SNTP)
-	case SNTP:
-#endif
-	case NETCONS:
-	case TFTP:
-		NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
-		NetOurGatewayIP = getenv_IPaddr ("gatewayip");
-		NetOurSubnetMask= getenv_IPaddr ("netmask");
-		NetOurVLAN = getenv_VLAN("vlan");
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
+	NetCopyIP(&NetOurIP, &bd->bi_ip_addr);
+	NetOurGatewayIP = getenv_IPaddr ("gatewayip");
+	NetOurSubnetMask= getenv_IPaddr ("netmask");
+	NetServerIP = getenv_IPaddr ("serverip");
+	NetOurNativeVLAN = getenv_VLAN("nvlan");
+	NetOurVLAN = getenv_VLAN("vlan");
 
-		switch (protocol) {
-#if defined(CONFIG_CMD_NFS)
-		case NFS:
-#endif
-		case NETCONS:
-		case TFTP:
-			NetServerIP = getenv_IPaddr ("serverip");
-			break;
-#if defined(CONFIG_CMD_PING)
-		case PING:
-			/* nothing */
-			break;
-#endif
-#if defined(CONFIG_CMD_SNTP)
-		case SNTP:
-			/* nothing */
-			break;
-#endif
-		default:
-			break;
-		}
-
-		break;
-	case BOOTP:
-	case RARP:
-		/*
-		 * initialize our IP addr to 0 in order to accept ANY
-		 * IP addr assigned to us by the BOOTP / RARP server
-		 */
-		NetOurIP = 0;
-		NetServerIP = getenv_IPaddr ("serverip");
-		NetOurVLAN = getenv_VLAN("vlan");	/* VLANs must be read */
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
-	case CDP:
-		NetOurVLAN = getenv_VLAN("vlan");	/* VLANs must be read */
-		NetOurNativeVLAN = getenv_VLAN("nvlan");
-		break;
-	default:
-		break;
-	}
 	env_changed_id = env_id;
 	return 0;
 }
@@ -440,10 +389,7 @@
 
 #if defined(CONFIG_CMD_DHCP)
 		case DHCP:
-			/* Start with a clean slate... */
 			BootpTry = 0;
-			NetOurIP = 0;
-			NetServerIP = getenv_IPaddr ("serverip");
 			DhcpRequest();		/* Basically same as BOOTP */
 			break;
 #endif
#9 lore raw

PATCH 1/1 NetLoop initialization bug

Hi Michael,

Quoted 2 lines

Sorry for the previous mail,
I have to enable the git operation via gmail on my workstation.

What do you mean by this? Usually you only need a mailer which is capable to attach things inline. Even better - if you have a direct access to an mta or an smtp server, use git-send-email after git-format-patch.

Quoted 1 line

Meanwhile I created the patch again by diff.

Well this is really more work for the people importing the patches, so if at all possible, get tghe setup working with the automatically generated patches.

Quoted 15 lines

The patch fixes the bug of partial initialization of global network parameters.

Upon u-boot's start up the first ping command causes a failure of the consequent TFTP transfers. It happens in the recently added mechanism of NetLoop initialization where initialization of global network parameters is separated in the NetInitLoop routine which is called per env_id change. Thus, ping request will initialize the network parameters necessary for ping operation only, afterwards the env_changed_id will be set to the env_id that will prevent all following initialization requests from other protocols.
The problem is that the initialized by ping subset of network parameters is not sufficient for other protocols and particularly for TFTP which requires the NetServerIp also.

Quoted 1 line

Please don't elide your e-mail address here. The consent is that we use real e-mail addresses in the signed-off-by lines (as you also did in your previous post). One negative impact for example is that our statistic scripts will fail for such entries.

Apart from that you have to nudge Ben or Heiko as to the real content. I only wanted to help get the formalities straight ;)

Cheers
Detlev

Signature · 5 lines
Number theorists do it perfectly and rationally.
--
DENX Software Engineering GmbH,      MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich,  Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: dzu at denx.de
#10 lore raw

PATCH 1/1 NetLoop initialization bug

Quoted 6 lines · Detlev Zundel
Quoted 1 line

I have to enable the git operation via gmail on my workstation.

What do you mean by this? ?Usually you only need a mailer which is capable to attach things inline. ?Even better - if you have a direct access to an mta or an smtp server, use git-send-email after git-format-patch.

Ok, I tried the msmtp to access my www.gmail.com account with no success nether directly (is prohibited by corporate policy) nor via corporate email relay (the gmail smtp server uses non standard port which requires reflection in the email relay configuration and no one is willing to do it for me :-).

Quoted 5 lines

Apart from that you have to nudge Ben or Heiko as to the real content. I only wanted to help get the formalities straight ;)

Cheers
Detlev

Dear Heiko,

Could you please do an additional try to apply the patch produced
with git-format-patch. I hope it is OK now.

Thanks,
Michael

#11 lore raw

PATCH 1/1 NetLoop initialization bug

Hi Michael,

Quoted 26 lines · Michael Zaidman
Quoted 7 lines · Detlev Zundel
Quoted 2 lines

I have to enable the git operation via gmail on my workstation.

What do you mean by this? Usually you only need a mailer which is capable to attach things inline. Even better - if you have a direct access to an mta or an smtp server, use git-send-email after git-format-patch.

Ok, I tried the msmtp to access my www.gmail.com account with no success nether directly (is prohibited by corporate policy) nor via corporate email relay (the gmail smtp server uses non standard port which requires reflection in the email relay configuration and no one is willing to do it for me :-).

Quoted 6 lines

Apart from that you have to nudge Ben or Heiko as to the real content. I only wanted to help get the formalities straight ;)

Cheers
Detlev

Quoted 5 lines

Dear Heiko,

Could you please do an additional try to apply the patch produced
with git-format-patch. I hope it is OK now.

I tried to apply it but it didn't work, but this isn't my usual computer. I'll try on my other machine tonight. BTW - what mailer are you using?

regards,
Ben

#12 lore raw

PATCH 1/1 NetLoop initialization bug

Dear Ben Warren,

Quoted 7 lines · In message <[email protected]> you wrote:
Quoted 4 lines

Could you please do an additional try to apply the patch produced
with git-format-patch. I hope it is OK now.

I tried to apply it but it didn't work, but this isn't my usual computer. I'll try on my other machine tonight. BTW - what mailer are you using?

It fails for me, too.

Best regards,

Wolfgang Denk

Signature · 6 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"Ja, mach' nur einen Plan,    sei nur ein grosses Licht
und mach' dann noch 'nen zweiten Plan,    geh'n tun sie beide nicht."
                                     -- Bert Brecht, Dreigroschenoper
#13 lore raw

PATCH 1/1 NetLoop initialization bug

In reply to Ben Warren

Hi Ben,

Quoted 4 lines · Ben Warren

I tried to apply it but it didn't work, but this isn't my usual computer.
I'll try on my other machine tonight. BTW - what mailer are you using?

I used Firefox as front-end web interface to communicate with gmail server - the only option I have at my work place. It obviously caused the problem. The http://kerneltrap.org/Linux/Email_Clients_and_Patches
perfectly describes this. Today I created the patch on my home's Linux with git-format-patch and sent it by git-send-email via msmtp. I would like to apologize for inconvenience and ask you to try patch again.

BTW - what mailer do you use to communicate to the gmail server?

Thanks,
Michael

#14 lore raw

PATCH 1/1 NetLoop initialization bug

Dear Michael & Ben,

Quoted 7 lines · In message <[email protected]> you…

I used Firefox as front-end web interface to communicate with gmail server - the only option I have at my work place. It obviously caused the problem. The http://kerneltrap.org/Linux/Email_Clients_and_Patches
perfectly describes this. Today I created the patch on my home's Linux with git-format-patch and sent it by git-send-email via msmtp. I would like to apologize for inconvenience and ask you to try patch again.

I confirm that this patch applies cleanly.

[I cannot comment on the content, though.]

Best regards,

Wolfgang Denk

Signature · 6 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Each kiss is as the first.
	-- Miramanee, Kirk's wife, "The Paradise Syndrome",
	   stardate 4842.6
#15 lore raw

PATCH 1/1 NetLoop initialization bug

Quoted 12 lines · Wolfgang Denk

Dear Michael & Ben,

Quoted 7 lines · In message <[email protected]> you…

I used Firefox as front-end web interface to communicate with gmail server - the only option I have at my work place. It obviously caused the problem. The http://kerneltrap.org/Linux/Email_Clients_and_Patches
perfectly describes this. Today I created the patch on my home's Linux with git-format-patch and sent it by git-send-email via msmtp. I would like to apologize for inconvenience and ask you to try patch again.

Quoted 9 lines

I confirm that this patch applies cleanly.

[I cannot comment on the content, though.]

Best regards,

Wolfgang Denk

Cool - I'll try it out soon.

regards,
Ben

#16 lore raw

PATCH 1/1 NetLoop initialization bug

In reply to Michael Zaidman

Hi Michael,

Quoted 8 lines · Michael Zaidman

Hi Ben,

Quoted 4 lines · Ben Warren

I tried to apply it but it didn't work, but this isn't my usual computer.
I'll try on my other machine tonight. BTW - what mailer are you using?

Quoted 9 lines

I used Firefox as front-end web interface to communicate with gmail server - the only option I have at my work place. It obviously caused the problem. The http://kerneltrap.org/Linux/Email_Clients_and_Patches
perfectly describes this. Today I created the patch on my home's Linux with git-format-patch and sent it by git-send-email via msmtp. I would like to apologize for inconvenience and ask you to try patch again.

BTW - what mailer do you use to communicate to the gmail server?

I use Thunderbird via IMAP. IMAP is nice because the messages all stay on the server so if you have a few different machines and OSs you don't have to worry about keeping things coherent. Sending patches through TBird usually works, as long as you cut&paste into a 'preformat' block (available from a pull-down menu). But it's not bulletproof. The only sure way to send a patch without worrying about line wrapping etc. is to use git-send-email. You can find instructions online on how to do this with gmail, IIRC you need to have some sendmail-like relay running on your Linux box. I don't remember the details and am not at that computer right now.

cheers,
Ben

#17 lore raw

PATCH 1/1 NetLoop initialization bug

Hi Ben,

Quoted 35 lines · Ben Warren

Hi Michael,

Quoted 12 lines · Michael Zaidman

Hi Ben,

Quoted 5 lines · Ben Warren

I tried to apply it but it didn't work, but this isn't my usual computer. ?I'll try on my other machine tonight. ?BTW - what mailer are you using?

Quoted 9 lines

I used Firefox as front-end web interface to communicate with gmail server - the only option I have at my work place. It obviously caused the problem. The http://kerneltrap.org/Linux/Email_Clients_and_Patches
perfectly describes this. Today I created the patch on my home's Linux with git-format-patch and sent it by git-send-email via msmtp. I would like to apologize for inconvenience and ask you to try patch again.

BTW - what mailer do you use to communicate to the gmail server?

I use Thunderbird via IMAP. ?IMAP is nice because the messages all stay on the server so if you have a few different machines and OSs you don't have to worry about keeping things coherent. ?Sending patches through TBird usually works, as long as you cut&paste into a 'preformat' block (available from a pull-down menu). ?But it's not bulletproof. ?The only sure way to send a patch without worrying about line wrapping etc. is to use git-send-email. ?You can find instructions online on how to do this with gmail, IIRC you need to have some sendmail-like relay running on your Linux box. ?I don't remember the details and am not at that computer right now.

Quoted 3 lines

cheers,
Ben

Thanks for the info. I played today with the Evolution which also keeps mails on the server and has 'preformated' option trying to compose correct inline patch attachment but with no success.

Good night,
Michael