de2104x tulip driver bug in 2.6.0-test9

de2104x tulip driver bug in 2.6.0-test9

[Russell Coker]

00:14.0 Ethernet controller: Digital Equipment Corporation DECchip 21041 
[Tulip Pass 3] (rev 11)

Above is the lspci output for my PCI Ethernet card.  Below is what happens 
when I try to boot 2.6.0-test9.  2.4.x kernels have been working well on the 
same card for a long time, so the hardware seems basically OK.

Configuring network interfaces... eth0: set link BNC
 eth0:    mode 0x7ffc0040, sia 0x10c4,0xffffef09,0xfffff7fd,0xffff0006
 eth0:    set mode 0x7ffc0000, set sia 0xef09,0xf7fd,0x6
 eth0: timeout expired stopping DMA
 ------------[ cut here ]------------
 kernel BUG at drivers/net/tulip/de2104x.c:926!
 invalid operand: 0000 [#1]
 CPU:    0
 EIP:    0060:[<d08a7f4c>]    Not tainted
 EFLAGS: 00010006
 EIP is at de_set_media+0x1c/0x140 [de2104x]
 eax: fc200100   ebx: cf172200   ecx: c02c6770   edx: d089f000
 esi: cf172200   edi: 00000000   ebp: cf021e50   esp: cf021e44
 ds: 007b   es: 007b   ss: 0068
 Process ethtool (pid: 478, threadinfo=cf020000 task=cfb0ccc0)
 Stack: cf172200 000008c3 00000000 cf021e84 d08a8fb6 cf172200 cf172200 
cf172200
        fffff73c 00000000 00000000 00000000 ffffffea cf020000 00000001 
00000000
        cf021e9c d08a9104 cf172200 cf021eb0 cf021eb0 cf172000 cf021eec 
c0224368
 Call Trace:
  [<d08a8fb6>] __de_set_settings+0x186/0x200 [de2104x]
  [<d08a9104>] de_set_settings+0x24/0x50 [de2104x]
  [<c0224368>] ethtool_set_settings+0x68/0x90
  [<c019e4fa>] capable+0x1a/0x40
  [<c022587b>] dev_ethtool+0xab/0x220
  [<c0223766>] dev_ioctl+0x156/0x350
  [<c02613b7>] inet_ioctl+0xb7/0xd0
  [<c021ae31>] sock_ioctl+0xf1/0x280
  [<c015eac2>] sys_ioctl+0x112/0x280
  [<c0109169>] sysenter_past_esp+0x52/0x79

 Code: 0f 0b 9e 03 04 99 8a d0 f6 86 a4 05 00 00 01 74 0a c7 42 58
  <7>eth0: tx err, status 0x7fffb178
 note: ethtool[478] exited with preempt_count 1
 bad: scheduling while atomic!
 Call Trace:
  [<c0118189>] schedule+0x5b9/0x5c0
  [<c013f1a3>] unmap_page_range+0x33/0x60
  [<c013f3c3>] unmap_vmas+0x1f3/0x240
  [<c01430a5>] exit_mmap+0x65/0x180
  [<c0119aaa>] mmput+0x6a/0xc0
  [<c011d938>] do_exit+0x128/0x410
  [<c0109cc0>] do_invalid_op+0x0/0xa0
  [<c0109a74>] die+0xc4/0xd0
  [<c0109d54>] do_invalid_op+0x94/0xa0
  [<d08a7f4c>] de_set_media+0x1c/0x140 [de2104x]
  [<c01d5ae7>] poke_blanked_console+0x57/0x70
  [<c01d4f4d>] vt_console_print+0x22d/0x310
  [<c011bcd5>] __call_console_drivers+0x55/0x60
  [<c011bdac>] call_console_drivers+0x5c/0x100
  [<c010942d>] error_code+0x2d/0x40
  [<d08a7f4c>] de_set_media+0x1c/0x140 [de2104x]
  [<d08a8fb6>] __de_set_settings+0x186/0x200 [de2104x]
  [<d08a9104>] de_set_settings+0x24/0x50 [de2104x]
  [<c0224368>] ethtool_set_settings+0x68/0x90
  [<c019e4fa>] capable+0x1a/0x40
  [<c022587b>] dev_ethtool+0xab/0x220
  [<c0223766>] dev_ioctl+0x156/0x350
  [<c02613b7>] inet_ioctl+0xb7/0xd0
  [<c021ae31>] sock_ioctl+0xf1/0x280
  [<c015eac2>] sys_ioctl+0x112/0x280
  [<c0109169>] sysenter_past_esp+0x52/0x79


-- 
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page

Re: de2104x tulip driver bug in 2.6.0-test9

[Christian Kujau]

Russell Coker wrote:
> 00:14.0 Ethernet controller: Digital Equipment Corporation DECchip 21041 
> [Tulip Pass 3] (rev 11)
> 
> Above is the lspci output for my PCI Ethernet card.  Below is what happens 
> when I try to boot 2.6.0-test9.  2.4.x kernels have been working well on the 
> same card for a long time, so the hardware seems basically OK.
> 
> Configuring network interfaces... eth0: set link BNC
>  eth0:    mode 0x7ffc0040, sia 0x10c4,0xffffef09,0xfffff7fd,0xffff0006
>  eth0:    set mode 0x7ffc0000, set sia 0xef09,0xf7fd,0x6
>  eth0: timeout expired stopping DMA

could this be anyhow related to this:

http://marc.theaimsgroup.com/?l=linux-kernel&m=106766135110165&w=2

there is a thread on linuxppc-dev too, as this is ppc specific:

http://lists.linuxppc.org/linuxppc-dev/200311/msg00001.html

it looks similar, but here on ppc32 i got no oops :-(

Christian.
-- 
BOFH excuse #27:

radiosity depletion

Re: de2104x tulip driver bug in 2.6.0-test9

[Russell Coker]

On Sat, 22 Nov 2003 02:32, Christian Kujau <evil@g-house.de> wrote:
> Russell Coker wrote:
> > 00:14.0 Ethernet controller: Digital Equipment Corporation DECchip 21041
> > [Tulip Pass 3] (rev 11)
> >
> > Above is the lspci output for my PCI Ethernet card.  Below is what
> > happens when I try to boot 2.6.0-test9.  2.4.x kernels have been working
> > well on the same card for a long time, so the hardware seems basically
> > OK.
> >
> > Configuring network interfaces... eth0: set link BNC
> >  eth0:    mode 0x7ffc0040, sia 0x10c4,0xffffef09,0xfffff7fd,0xffff0006
> >  eth0:    set mode 0x7ffc0000, set sia 0xef09,0xf7fd,0x6
> >  eth0: timeout expired stopping DMA
>
> could this be anyhow related to this:
>
> http://marc.theaimsgroup.com/?l=linux-kernel&m=106766135110165&w=2
>
> there is a thread on linuxppc-dev too, as this is ppc specific:

It is a bit different.  If I don't load the module during the boot sequence 
then I can get it to the stage of being pingable.  I think that timing is the 
issue, the boot scripts do everything quickly.  If I modprobe it, then 
ifconfig it, then change the media type, then I have succeeded once in 
getting ping responses, but I couldn't ssh.

-- 
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page

[PATCH/CFT] de2104x fixes

[Jeff Garzik]

[-- Attachment #1: Type: text/plain, Size: 247 bytes --]

So, can people give this a test?  It includes a change that, I hope, 
addresses Russell's problem, as well as a patch from Rask.

	Jeff


P.S.  It would be great if people cc'd me on such bug reports ;-) 
Quicker to find and respond, these days.


[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 1792 bytes --]

===== drivers/net/tulip/de2104x.c 1.25 vs edited =====
--- 1.25/drivers/net/tulip/de2104x.c	Thu Sep 11 18:46:11 2003
+++ edited/drivers/net/tulip/de2104x.c	Fri Nov 21 13:48:17 2003
@@ -28,8 +28,8 @@
  */
 
 #define DRV_NAME		"de2104x"
-#define DRV_VERSION		"0.6"
-#define DRV_RELDATE		"Sep 1, 2003"
+#define DRV_VERSION		"0.9"
+#define DRV_RELDATE		"Nov 21, 2003"
 
 #include <linux/config.h>
 #include <linux/module.h>
@@ -1380,18 +1380,18 @@
 		return rc;
 	}
 
-	rc = de_init_hw(de);
+	rc = request_irq(dev->irq, de_interrupt, SA_SHIRQ, dev->name, dev);
 	if (rc) {
-		printk(KERN_ERR "%s: h/w init failure, err=%d\n",
-		       dev->name, rc);
+		printk(KERN_ERR "%s: IRQ %d request failure, err=%d\n",
+		       dev->name, dev->irq, rc);
 		goto err_out_free;
 	}
 
-	rc = request_irq(dev->irq, de_interrupt, SA_SHIRQ, dev->name, dev);
+	rc = de_init_hw(de);
 	if (rc) {
-		printk(KERN_ERR "%s: IRQ %d request failure, err=%d\n",
-		       dev->name, dev->irq, rc);
-		goto err_out_hw;
+		printk(KERN_ERR "%s: h/w init failure, err=%d\n",
+		       dev->name, rc);
+		goto err_out_free_irq;
 	}
 
 	netif_start_queue(dev);
@@ -1399,10 +1399,8 @@
 
 	return 0;
 
-err_out_hw:
-	spin_lock_irqsave(&de->lock, flags);
-	de_stop_hw(de);
-	spin_unlock_irqrestore(&de->lock, flags);
+err_out_free_irq:
+	free_irq (dev->irq, dev);
 
 err_out_free:
 	de_free_rings(de);
@@ -1571,13 +1569,17 @@
 	    (ecmd->advertising == de->media_advertise))
 		return 0; /* nothing to change */
 	    
-	de_link_down(de);
-	de_stop_rxtx(de);
+	if (netif_running(dev)) {
+		de_link_down(de);
+		de_stop_rxtx(de);
+	}
 	
 	de->media_type = new_media;
 	de->media_lock = media_lock;
 	de->media_advertise = ecmd->advertising;
-	de_set_media(de);
+
+	if (netif_running(dev))
+		de_set_media(de);
 	
 	return 0;
 }

[PATCH/CFT] de2104x fixes

[Jeff Garzik]

[-- Attachment #1: Type: text/plain, Size: 907 bytes --]

Russell Coker wrote:
> On Sat, 22 Nov 2003 05:50, Jeff Garzik <jgarzik@pobox.com> wrote:
> 
>>So, can people give this a test?  It includes a change that, I hope,
>>addresses Russell's problem, as well as a patch from Rask.
> 
> 
> It does not compile for me, below are the warnings and errors.  I am using 
> 2.6.0-test9 with no patches applied other than your patch.
> 
> drivers/net/tulip/de2104x.c: In function `de_open':
> drivers/net/tulip/de2104x.c:1369: warning: unused variable `flags'
> drivers/net/tulip/de2104x.c: In function `__de_set_settings':
> drivers/net/tulip/de2104x.c:1572: error: `dev' undeclared (first use in this 
> function)
> drivers/net/tulip/de2104x.c:1572: error: (Each undeclared identifier is 
> reported only once
> drivers/net/tulip/de2104x.c:1572: error: for each function it appears in.)

Ok, here's a better patch.  It's against 2.6.0-test9-BK-latest 
(currently bk25).

[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 2437 bytes --]

===== drivers/net/tulip/de2104x.c 1.25 vs edited =====
--- 1.25/drivers/net/tulip/de2104x.c	Thu Sep 11 18:46:11 2003
+++ edited/drivers/net/tulip/de2104x.c	Sat Nov 22 07:05:40 2003
@@ -28,8 +28,8 @@
  */
 
 #define DRV_NAME		"de2104x"
-#define DRV_VERSION		"0.6"
-#define DRV_RELDATE		"Sep 1, 2003"
+#define DRV_VERSION		"0.9"
+#define DRV_RELDATE		"Nov 21, 2003"
 
 #include <linux/config.h>
 #include <linux/module.h>
@@ -1366,7 +1366,6 @@
 {
 	struct de_private *de = dev->priv;
 	int rc;
-	unsigned long flags;
 
 	if (netif_msg_ifup(de))
 		printk(KERN_DEBUG "%s: enabling interface\n", dev->name);
@@ -1380,18 +1379,18 @@
 		return rc;
 	}
 
-	rc = de_init_hw(de);
+	rc = request_irq(dev->irq, de_interrupt, SA_SHIRQ, dev->name, dev);
 	if (rc) {
-		printk(KERN_ERR "%s: h/w init failure, err=%d\n",
-		       dev->name, rc);
+		printk(KERN_ERR "%s: IRQ %d request failure, err=%d\n",
+		       dev->name, dev->irq, rc);
 		goto err_out_free;
 	}
 
-	rc = request_irq(dev->irq, de_interrupt, SA_SHIRQ, dev->name, dev);
+	rc = de_init_hw(de);
 	if (rc) {
-		printk(KERN_ERR "%s: IRQ %d request failure, err=%d\n",
-		       dev->name, dev->irq, rc);
-		goto err_out_hw;
+		printk(KERN_ERR "%s: h/w init failure, err=%d\n",
+		       dev->name, rc);
+		goto err_out_free_irq;
 	}
 
 	netif_start_queue(dev);
@@ -1399,10 +1398,8 @@
 
 	return 0;
 
-err_out_hw:
-	spin_lock_irqsave(&de->lock, flags);
-	de_stop_hw(de);
-	spin_unlock_irqrestore(&de->lock, flags);
+err_out_free_irq:
+	free_irq (dev->irq, dev);
 
 err_out_free:
 	de_free_rings(de);
@@ -1514,7 +1511,8 @@
 	return 0;
 }
 
-static int __de_set_settings(struct de_private *de, struct ethtool_cmd *ecmd)
+static int __de_set_settings(struct net_device *dev, struct de_private *de,
+			     struct ethtool_cmd *ecmd)
 {
 	u32 new_media;
 	unsigned int media_lock;
@@ -1571,13 +1569,17 @@
 	    (ecmd->advertising == de->media_advertise))
 		return 0; /* nothing to change */
 	    
-	de_link_down(de);
-	de_stop_rxtx(de);
+	if (netif_running(dev)) {
+		de_link_down(de);
+		de_stop_rxtx(de);
+	}
 	
 	de->media_type = new_media;
 	de->media_lock = media_lock;
 	de->media_advertise = ecmd->advertising;
-	de_set_media(de);
+
+	if (netif_running(dev))
+		de_set_media(de);
 	
 	return 0;
 }
@@ -1615,7 +1617,7 @@
 	int rc;
 
 	spin_lock_irq(&de->lock);
-	rc = __de_set_settings(de, ecmd);
+	rc = __de_set_settings(dev, de, ecmd);
 	spin_unlock_irq(&de->lock);
 
 	return rc;

Re: [PATCH/CFT] de2104x fixes

[Rask Ingemann Lambertsen]

[-- Attachment #1: Type: text/plain, Size: 550 bytes --]

On Fri, Nov 21, 2003 at 01:50:24PM -0500, Jeff Garzik wrote:
> So, can people give this a test?  It includes a change that, I hope, 
> addresses Russell's problem, as well as a patch from Rask.

I have attached a patch which fixes two problems I found during compilation:
1) de_open() no longer uses the flags variable because the spinlocking is
   gone, but I forgot to remove the variable.
2) __de_set_settings() now references dev which is undefined.

The patch should be applied on top of your patch, Jeff.

-- 
Regards,
Rask Ingemann Lambertsen

[-- Attachment #2: de2104x-fix-for-jeff.patch --]
[-- Type: text/plain, Size: 874 bytes --]

--- linux-2.6.0-test8/drivers/net/tulip/de2104x.c-orig	Tue Nov 25 13:20:25 2003
+++ linux-2.6.0-test8/drivers/net/tulip/de2104x.c	Tue Nov 25 13:20:25 2003
@@ -1384,7 +1384,6 @@ static int de_open (struct net_device *d
 {
 	struct de_private *de = dev->priv;
 	int rc;
-	unsigned long flags;
 
 	if (netif_msg_ifup(de))
 		printk(KERN_DEBUG "%s: enabling interface\n", dev->name);
@@ -1601,7 +1600,7 @@ static int __de_set_settings(struct de_p
 	    (ecmd->advertising == de->media_advertise))
 		return 0; /* nothing to change */
 	    
-	if (netif_running(dev)) {
+	if (netif_running(de->dev)) {
 		de_link_down(de);
 		de_stop_rxtx(de);
 	}
@@ -1610,7 +1609,7 @@ static int __de_set_settings(struct de_p
 	de->media_lock = media_lock;
 	de->media_advertise = ecmd->advertising;
 
-	if (netif_running(dev))
+	if (netif_running(de->dev))
 		de_set_media(de);
 	
 	return 0;