Skip to contentSkip to conversation

net: eth-uclass: call stop only for active devices

PATCH New Keerthy 5 participants 8 messages Feb 20, 2019 – Mar 29, 2019 Networking

lore
Share & tools
lore

Keerthy

author
#0 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

---
 net/eth-uclass.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/eth-uclass.c +2 -1
index 2ef20df192..26f7e0b8cd 100644
@@ -531,7 +531,8 @@ static int eth_pre_remove(struct udevice *dev)
 {
 	struct eth_pdata *pdata = dev->platdata;
 
-	eth_get_ops(dev)->stop(dev);
+	if (eth_is_active(dev))
+		eth_get_ops(dev)->stop(dev);
 
 	/* clear the MAC address */
 	memset(pdata->enetaddr, 0, ARP_HLEN);
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

Quoted 28 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

---
 net/eth-uclass.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/eth-uclass.c +2 -2
index 2ef20df192..26f7e0b8cd 100644
@@ -531,7 +531,8 @@ static int eth_pre_remove(struct udevice *dev)
 {
        struct eth_pdata *pdata = dev->platdata;
 
-       eth_get_ops(dev)->stop(dev);
+       if (eth_is_active(dev))
+               eth_get_ops(dev)->stop(dev);
 
        /* clear the MAC address */
        memset(pdata->enetaddr, 0, ARP_HLEN);
--
Quoted 2 lines

2.17.1

#2 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

In reply to Keerthy
Quoted 7 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

Quoted 5 lines
---
 net/eth-uclass.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Could you please provide a test case?

Regards,
Bin

#3 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

Quoted 9 lines · Bin Meng
Quoted 7 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

Quoted 7 lines
Quoted 5 lines
---
 net/eth-uclass.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Could you please provide a test case?

Presumably that test should live in test/dm/eth.c with some potential support needed by drivers/net/sandbox.c

Thanks,
-Joe

Quoted 6 lines

Regards,
Bin
_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot

#4 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

In reply to Keerthy
Quoted 17 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

---
 net/eth-uclass.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/eth-uclass.c +0 -0
index 2ef20df192..26f7e0b8cd 100644

net/eth-uclass.c static int eth_pre_remove(struct udevice *dev) · line 535

Quoted 6 lines
net/eth-uclass.c +2 -1
@@ -531,7 +531,8 @@ static int eth_pre_remove(struct udevice *dev)
 {
        struct eth_pdata *pdata = dev->platdata;
 
-       eth_get_ops(dev)->stop(dev);
+       if (eth_is_active(dev))
+               eth_get_ops(dev)->stop(dev);

This seems reasonable... What was the case that provoked an issue? Which driver was having trouble?

Quoted 10 lines
net/eth-uclass.c +0 -1
        /* clear the MAC address */
        memset(pdata->enetaddr, 0, ARP_HLEN);
--

2.17.1

_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot

#5 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

Quoted 21 lines · Joe Hershberger
Quoted 20 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

---
  net/eth-uclass.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
net/eth-uclass.c +0 -0
index 2ef20df192..26f7e0b8cd 100644
@@ -531,7 +531,8 @@ static int eth_pre_remove(struct udevice *dev)
  {
         struct eth_pdata *pdata = dev->platdata;
 
Quoted 6 lines

net/eth-uclass.c static int eth_pre_remove(struct udevice *dev) · line 535

Quoted 3 lines
net/eth-uclass.c +2 -1
-       eth_get_ops(dev)->stop(dev);
+       if (eth_is_active(dev))
+               eth_get_ops(dev)->stop(dev);

This seems reasonable... What was the case that provoked an issue? Which driver was having trouble?

I am trying to implement pru based ethernet driver(multiple instances) with cpsw driver on top on am654-evm. I observed that even though i had not started
a particular instance of my pru based device stop was getting called and it was crashing right at the end.

Quoted 11 lines
Quoted 11 lines
net/eth-uclass.c +0 -1
         /* clear the MAC address */
         memset(pdata->enetaddr, 0, ARP_HLEN);
--

2.17.1

_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot

#6 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

Quoted 32 lines · Keerthy
Quoted 21 lines · Joe Hershberger
Quoted 20 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

---
  net/eth-uclass.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
net/eth-uclass.c +0 -0
index 2ef20df192..26f7e0b8cd 100644
@@ -531,7 +531,8 @@ static int eth_pre_remove(struct udevice *dev)
  {
         struct eth_pdata *pdata = dev->platdata;
 
Quoted 6 lines

net/eth-uclass.c static int eth_pre_remove(struct udevice *dev) · line 535

Quoted 3 lines
net/eth-uclass.c +2 -1
-       eth_get_ops(dev)->stop(dev);
+       if (eth_is_active(dev))
+               eth_get_ops(dev)->stop(dev);

This seems reasonable... What was the case that provoked an issue? Which driver was having trouble?

Quoted 6 lines

I am trying to implement pru based ethernet driver(multiple instances) with cpsw driver on top on am654-evm. I observed that even though i had not started
a particular instance of my pru based device stop was getting called and it was crashing right at the end.

Please add a proper test case in test/dm/eth.c to cover such scenarios. Thanks!

Regards,
Bin

author
#7 lore raw

PATCH 1/1 net: eth-uclass: call stop only for active devices

In reply to Keerthy
Quoted 31 lines · Keerthy
Quoted 21 lines · Joe Hershberger
Quoted 20 lines · Keerthy

Currently stop is being called unconditionally without even checking if start is called. In case of multiple instances eth being present many devices might just be initialized without a start call in such cases stop might lead unpredictable behaviors including aborts and crashes. Hence add a check before calling stop.

---
  net/eth-uclass.c | 3 ++-

  1 file changed, 2 insertions(+), 1 deletion(-)
net/eth-uclass.c +0 -0
index 2ef20df192..26f7e0b8cd 100644
@@ -531,7 +531,8 @@ static int eth_pre_remove(struct udevice *dev)
   {
          struct eth_pdata *pdata = dev->platdata;
 
Quoted 6 lines

net/eth-uclass.c static int eth_pre_remove(struct udevice *dev) · line 535

Quoted 3 lines
net/eth-uclass.c +2 -1
-       eth_get_ops(dev)->stop(dev);
+       if (eth_is_active(dev))
+               eth_get_ops(dev)->stop(dev);

This seems reasonable... What was the case that provoked an issue? Which driver was having trouble?

Quoted 5 lines

I am trying to implement pru based ethernet driver(multiple instances) with cpsw driver on top on am654-evm. I observed that even though i had not started
a particular instance of my pru based device stop was getting called and it was crashing right at the end.

Joe,

I am yet to implement a test case for this. Can this patch be applied independently?

Regards,
Keerthy

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
net/eth-uclass.c +0 -0
          /* clear the MAC address */
          memset(pdata->enetaddr, 0, ARP_HLEN);
Signature · 6 lines
2.17.1

_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot