diff mbox series

[v7,7/7] app/testpmd: check not detaching device twice

Message ID 20181023082842.7963-8-thomas@monjalon.net (mailing list archive)
State Superseded, archived
Delegated to: Ferruh Yigit
Headers
Series replace attach/detach functions |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK

Commit Message

Thomas Monjalon Oct. 23, 2018, 8:28 a.m. UTC 
  The command "port detach" is removing the EAL rte_device
of the ethdev port specified as parameter.

After detaching, the pointer, which maps a port to its device,
is resetted. This way, it is possible to check whether a port
is still associated to a (not removed) device.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
---
 app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)

Comments

Iremonger, Bernard Oct. 23, 2018, 10:01 a.m. UTC | #1 
Hi Thomas,

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas@monjalon.net]
> Sent: Tuesday, October 23, 2018 9:29 AM
> To: dev@dpdk.org
> Cc: gaetan.rivet@6wind.com; ophirmu@mellanox.com;
> wisamm@mellanox.com; Yigit, Ferruh <ferruh.yigit@intel.com>;
> arybchenko@solarflare.com; Iremonger, Bernard
> <bernard.iremonger@intel.com>
> Subject: [PATCH v7 7/7] app/testpmd: check not detaching device twice
> 
> The command "port detach" is removing the EAL rte_device of the ethdev
> port specified as parameter.
> 
> After detaching, the pointer, which maps a port to its device, is resetted. This

Typo:  "resetted" should be "reset"

> way, it is possible to check whether a port is still associated to a (not
> removed) device.
> 
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> ---
>  app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
>  1 file changed, 21 insertions(+), 3 deletions(-)
> 
> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> 14647fa19..d5998fddc 100644
> --- a/app/test-pmd/testpmd.c
> +++ b/app/test-pmd/testpmd.c
> @@ -2353,8 +2353,17 @@ setup_attached_port(portid_t pi)  void
> detach_port(portid_t port_id)  {
> +	struct rte_device *dev;
> +	portid_t sibling;
> +
>  	printf("Removing a device...\n");

The functionality of the detach_port() function has changed now to removing a device, should the function name be changed to reflect the new functionality.
How about detach_device() instead of detach detach_port().

> 
> +	dev = rte_eth_devices[port_id].device;
> +	if (dev == NULL) {
> +		printf("Device already removed\n");
> +		return;
> +	}
> +
>  	if (ports[port_id].port_status != RTE_PORT_CLOSED) {
>  		if (ports[port_id].port_status != RTE_PORT_STOPPED) {
>  			printf("Port not stopped\n");
> @@ -2365,15 +2374,24 @@ detach_port(portid_t port_id)
>  			port_flow_flush(port_id);
>  	}
> 
> -	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
> +	if (rte_dev_remove(dev) != 0) {
>  		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);

Should the log message be ( ERR "Failed to detach device %s\n", dev->name) ?

>  		return;
>  	}
> 
> +	/* reset mapping between old ports and removed device */
> +	for (sibling = 0; sibling < RTE_MAX_ETHPORTS; sibling++)
> +		if (rte_eth_devices[sibling].device == dev) {
> +			rte_eth_devices[sibling].device = NULL;
> +			if (ports[sibling].port_status != RTE_PORT_CLOSED) {
> +				ports[sibling].port_status =
> RTE_PORT_CLOSED;
> +				printf("Port %u is closed\n", sibling);
> +			}
> +		}
> +
>  	remove_unused_fwd_ports();
> 
> -	printf("Port %u is detached. Now total ports is %d\n",
> -			port_id, nb_ports);

How about printf("Device %s is detached, Now total ports is %d\n"
	dev->name, nb_ports);
 
> +	printf("Now total ports is %d\n", nb_ports);
>  	printf("Done\n");
>  	return;
>  }
> --
> 2.19.0

Regards,

Bernard.
Thomas Monjalon Oct. 23, 2018, 12:03 p.m. UTC | #2 
23/10/2018 12:01, Iremonger, Bernard:
> From: Thomas Monjalon [mailto:thomas@monjalon.net]
> > The command "port detach" is removing the EAL rte_device of the ethdev
> > port specified as parameter.
> > 
> > After detaching, the pointer, which maps a port to its device, is resetted. This
> 
> Typo:  "resetted" should be "reset"
> 
> > way, it is possible to check whether a port is still associated to a (not
> > removed) device.
> > 
> > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > ---
> >  app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
> >  1 file changed, 21 insertions(+), 3 deletions(-)
> > 
> > diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> > 14647fa19..d5998fddc 100644
> > --- a/app/test-pmd/testpmd.c
> > +++ b/app/test-pmd/testpmd.c
> > @@ -2353,8 +2353,17 @@ setup_attached_port(portid_t pi)  void
> > detach_port(portid_t port_id)  {
> > +	struct rte_device *dev;
> > +	portid_t sibling;
> > +
> >  	printf("Removing a device...\n");
> 
> The functionality of the detach_port() function has changed now to
> removing a device, should the function name be changed to reflect
> the new functionality.

No it doesn't change, it has always removed the rte_device of the port.
But the naming is a bit strange, I agree.
I just changed the log to make it a bit clearer.

> How about detach_device() instead of detach detach_port().

The strange thing with testpmd is that every commands take a port id.
The rte_device is hidden in testpmd.
So the detach command is detaching the underlying device of the port,
and all its sibling ports of course.

What about detach_device_of_port() ?

[...]
> > -	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
> > +	if (rte_dev_remove(dev) != 0) {
> >  		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);
> 
> Should the log message be ( ERR "Failed to detach device %s\n", dev->name) ?

Yes!

[...]
> > -	printf("Port %u is detached. Now total ports is %d\n",
> > -			port_id, nb_ports);
> 
> How about printf("Device %s is detached, Now total ports is %d\n"
> 	dev->name, nb_ports);

The issue is that we cannot get the device name after detach.
I can reword it differently:
	Device of port %u is detached, Now total ports is %d
Thomas Monjalon Oct. 23, 2018, 12:13 p.m. UTC | #3 
23/10/2018 14:03, Thomas Monjalon:
> 23/10/2018 12:01, Iremonger, Bernard:
> > From: Thomas Monjalon [mailto:thomas@monjalon.net]
> > > The command "port detach" is removing the EAL rte_device of the ethdev
> > > port specified as parameter.
> > > 
> > > After detaching, the pointer, which maps a port to its device, is resetted. This
> > 
> > Typo:  "resetted" should be "reset"
> > 
> > > way, it is possible to check whether a port is still associated to a (not
> > > removed) device.
> > > 
> > > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > > ---
> > >  app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
> > >  1 file changed, 21 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> > > 14647fa19..d5998fddc 100644
> > > --- a/app/test-pmd/testpmd.c
> > > +++ b/app/test-pmd/testpmd.c
> > > @@ -2353,8 +2353,17 @@ setup_attached_port(portid_t pi)  void
> > > detach_port(portid_t port_id)  {
> > > +	struct rte_device *dev;
> > > +	portid_t sibling;
> > > +
> > >  	printf("Removing a device...\n");
> > 
> > The functionality of the detach_port() function has changed now to
> > removing a device, should the function name be changed to reflect
> > the new functionality.
> 
> No it doesn't change, it has always removed the rte_device of the port.
> But the naming is a bit strange, I agree.
> I just changed the log to make it a bit clearer.
> 
> > How about detach_device() instead of detach detach_port().
> 
> The strange thing with testpmd is that every commands take a port id.
> The rte_device is hidden in testpmd.
> So the detach command is detaching the underlying device of the port,
> and all its sibling ports of course.
> 
> What about detach_device_of_port() ?

Or detach_port_device()?

> [...]
> > > -	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
> > > +	if (rte_dev_remove(dev) != 0) {
> > >  		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);
> > 
> > Should the log message be ( ERR "Failed to detach device %s\n", dev->name) ?
> 
> Yes!
> 
> [...]
> > > -	printf("Port %u is detached. Now total ports is %d\n",
> > > -			port_id, nb_ports);
> > 
> > How about printf("Device %s is detached, Now total ports is %d\n"
> > 	dev->name, nb_ports);
> 
> The issue is that we cannot get the device name after detach.
> I can reword it differently:
> 	Device of port %u is detached, Now total ports is %d
> 
> 
> 
>
Thomas Monjalon Oct. 23, 2018, 12:37 p.m. UTC | #4 
I want to submit two more patches to clean testpmd for attach/detach.

I propose to drop this patch from this series,
and I will submit a new series dedicated to testpmd cleanup,
including this patch.


23/10/2018 14:13, Thomas Monjalon:
> 23/10/2018 14:03, Thomas Monjalon:
> > 23/10/2018 12:01, Iremonger, Bernard:
> > > From: Thomas Monjalon [mailto:thomas@monjalon.net]
> > > > The command "port detach" is removing the EAL rte_device of the ethdev
> > > > port specified as parameter.
> > > > 
> > > > After detaching, the pointer, which maps a port to its device, is resetted. This
> > > 
> > > Typo:  "resetted" should be "reset"
> > > 
> > > > way, it is possible to check whether a port is still associated to a (not
> > > > removed) device.
> > > > 
> > > > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > > > ---
> > > >  app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
> > > >  1 file changed, 21 insertions(+), 3 deletions(-)
> > > > 
> > > > diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> > > > 14647fa19..d5998fddc 100644
> > > > --- a/app/test-pmd/testpmd.c
> > > > +++ b/app/test-pmd/testpmd.c
> > > > @@ -2353,8 +2353,17 @@ setup_attached_port(portid_t pi)  void
> > > > detach_port(portid_t port_id)  {
> > > > +	struct rte_device *dev;
> > > > +	portid_t sibling;
> > > > +
> > > >  	printf("Removing a device...\n");
> > > 
> > > The functionality of the detach_port() function has changed now to
> > > removing a device, should the function name be changed to reflect
> > > the new functionality.
> > 
> > No it doesn't change, it has always removed the rte_device of the port.
> > But the naming is a bit strange, I agree.
> > I just changed the log to make it a bit clearer.
> > 
> > > How about detach_device() instead of detach detach_port().
> > 
> > The strange thing with testpmd is that every commands take a port id.
> > The rte_device is hidden in testpmd.
> > So the detach command is detaching the underlying device of the port,
> > and all its sibling ports of course.
> > 
> > What about detach_device_of_port() ?
> 
> Or detach_port_device()?
> 
> > [...]
> > > > -	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
> > > > +	if (rte_dev_remove(dev) != 0) {
> > > >  		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);
> > > 
> > > Should the log message be ( ERR "Failed to detach device %s\n", dev->name) ?
> > 
> > Yes!
> > 
> > [...]
> > > > -	printf("Port %u is detached. Now total ports is %d\n",
> > > > -			port_id, nb_ports);
> > > 
> > > How about printf("Device %s is detached, Now total ports is %d\n"
> > > 	dev->name, nb_ports);
> > 
> > The issue is that we cannot get the device name after detach.
> > I can reword it differently:
> > 	Device of port %u is detached, Now total ports is %d
Iremonger, Bernard Oct. 23, 2018, 12:39 p.m. UTC | #5 
Hi Thomas

<snip>
> Subject: Re: [dpdk-dev] [PATCH v7 7/7] app/testpmd: check not detaching
> device twice
> 
> 23/10/2018 14:03, Thomas Monjalon:
> > 23/10/2018 12:01, Iremonger, Bernard:
> > > From: Thomas Monjalon [mailto:thomas@monjalon.net]
> > > > The command "port detach" is removing the EAL rte_device of the
> > > > ethdev port specified as parameter.
> > > >
> > > > After detaching, the pointer, which maps a port to its device, is
> > > > resetted. This
> > >
> > > Typo:  "resetted" should be "reset"
> > >
> > > > way, it is possible to check whether a port is still associated to
> > > > a (not
> > > > removed) device.
> > > >
> > > > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > > > ---
> > > >  app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
> > > >  1 file changed, 21 insertions(+), 3 deletions(-)
> > > >
> > > > diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> > > > 14647fa19..d5998fddc 100644
> > > > --- a/app/test-pmd/testpmd.c
> > > > +++ b/app/test-pmd/testpmd.c
> > > > @@ -2353,8 +2353,17 @@ setup_attached_port(portid_t pi)  void
> > > > detach_port(portid_t port_id)  {
> > > > +	struct rte_device *dev;
> > > > +	portid_t sibling;
> > > > +
> > > >  	printf("Removing a device...\n");
> > >
> > > The functionality of the detach_port() function has changed now to
> > > removing a device, should the function name be changed to reflect
> > > the new functionality.
> >
> > No it doesn't change, it has always removed the rte_device of the port.
> > But the naming is a bit strange, I agree.
> > I just changed the log to make it a bit clearer.
> >
> > > How about detach_device() instead of detach detach_port().
> >
> > The strange thing with testpmd is that every commands take a port id.
> > The rte_device is hidden in testpmd.
> > So the detach command is detaching the underlying device of the port,
> > and all its sibling ports of course.
> >
> > What about detach_device_of_port() ?
> 
> Or detach_port_device()?

detach_port_device() looks fine to me.

> 
> > [...]
> > > > -	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
> > > > +	if (rte_dev_remove(dev) != 0) {
> > > >  		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);
> > >
> > > Should the log message be ( ERR "Failed to detach device %s\n", dev-
> >name) ?
> >
> > Yes!
> >
> > [...]
> > > > -	printf("Port %u is detached. Now total ports is %d\n",
> > > > -			port_id, nb_ports);
> > >
> > > How about printf("Device %s is detached, Now total ports is %d\n"
> > > 	dev->name, nb_ports);
> >
> > The issue is that we cannot get the device name after detach.
> > I can reword it differently:
> > 	Device of port %u is detached, Now total ports is %d

Looks fine to me.

Regards,

Bernard
Ferruh Yigit Oct. 23, 2018, 2:06 p.m. UTC | #6 
On 10/23/2018 1:37 PM, Thomas Monjalon wrote:
> I want to submit two more patches to clean testpmd for attach/detach.
> 
> I propose to drop this patch from this series,
> and I will submit a new series dedicated to testpmd cleanup,
> including this patch.

Got the set without this patch, please sent it separately.

> 
> 
> 23/10/2018 14:13, Thomas Monjalon:
>> 23/10/2018 14:03, Thomas Monjalon:
>>> 23/10/2018 12:01, Iremonger, Bernard:
>>>> From: Thomas Monjalon [mailto:thomas@monjalon.net]
>>>>> The command "port detach" is removing the EAL rte_device of the ethdev
>>>>> port specified as parameter.
>>>>>
>>>>> After detaching, the pointer, which maps a port to its device, is resetted. This
>>>>
>>>> Typo:  "resetted" should be "reset"
>>>>
>>>>> way, it is possible to check whether a port is still associated to a (not
>>>>> removed) device.
>>>>>
>>>>> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
>>>>> ---
>>>>>  app/test-pmd/testpmd.c | 24 +++++++++++++++++++++---
>>>>>  1 file changed, 21 insertions(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
>>>>> 14647fa19..d5998fddc 100644
>>>>> --- a/app/test-pmd/testpmd.c
>>>>> +++ b/app/test-pmd/testpmd.c
>>>>> @@ -2353,8 +2353,17 @@ setup_attached_port(portid_t pi)  void
>>>>> detach_port(portid_t port_id)  {
>>>>> +	struct rte_device *dev;
>>>>> +	portid_t sibling;
>>>>> +
>>>>>  	printf("Removing a device...\n");
>>>>
>>>> The functionality of the detach_port() function has changed now to
>>>> removing a device, should the function name be changed to reflect
>>>> the new functionality.
>>>
>>> No it doesn't change, it has always removed the rte_device of the port.
>>> But the naming is a bit strange, I agree.
>>> I just changed the log to make it a bit clearer.
>>>
>>>> How about detach_device() instead of detach detach_port().
>>>
>>> The strange thing with testpmd is that every commands take a port id.
>>> The rte_device is hidden in testpmd.
>>> So the detach command is detaching the underlying device of the port,
>>> and all its sibling ports of course.
>>>
>>> What about detach_device_of_port() ?
>>
>> Or detach_port_device()?
>>
>>> [...]
>>>>> -	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
>>>>> +	if (rte_dev_remove(dev) != 0) {
>>>>>  		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);
>>>>
>>>> Should the log message be ( ERR "Failed to detach device %s\n", dev->name) ?
>>>
>>> Yes!
>>>
>>> [...]
>>>>> -	printf("Port %u is detached. Now total ports is %d\n",
>>>>> -			port_id, nb_ports);
>>>>
>>>> How about printf("Device %s is detached, Now total ports is %d\n"
>>>> 	dev->name, nb_ports);
>>>
>>> The issue is that we cannot get the device name after detach.
>>> I can reword it differently:
>>> 	Device of port %u is detached, Now total ports is %d
> 
> 
> 
>
diff mbox series

Patch

diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index 14647fa19..d5998fddc 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -2353,8 +2353,17 @@  setup_attached_port(portid_t pi)
 void
 detach_port(portid_t port_id)
 {
+	struct rte_device *dev;
+	portid_t sibling;
+
 	printf("Removing a device...\n");
 
+	dev = rte_eth_devices[port_id].device;
+	if (dev == NULL) {
+		printf("Device already removed\n");
+		return;
+	}
+
 	if (ports[port_id].port_status != RTE_PORT_CLOSED) {
 		if (ports[port_id].port_status != RTE_PORT_STOPPED) {
 			printf("Port not stopped\n");
@@ -2365,15 +2374,24 @@  detach_port(portid_t port_id)
 			port_flow_flush(port_id);
 	}
 
-	if (rte_dev_remove(rte_eth_devices[port_id].device) != 0) {
+	if (rte_dev_remove(dev) != 0) {
 		TESTPMD_LOG(ERR, "Failed to detach port %u\n", port_id);
 		return;
 	}
 
+	/* reset mapping between old ports and removed device */
+	for (sibling = 0; sibling < RTE_MAX_ETHPORTS; sibling++)
+		if (rte_eth_devices[sibling].device == dev) {
+			rte_eth_devices[sibling].device = NULL;
+			if (ports[sibling].port_status != RTE_PORT_CLOSED) {
+				ports[sibling].port_status = RTE_PORT_CLOSED;
+				printf("Port %u is closed\n", sibling);
+			}
+		}
+
 	remove_unused_fwd_ports();
 
-	printf("Port %u is detached. Now total ports is %d\n",
-			port_id, nb_ports);
+	printf("Now total ports is %d\n", nb_ports);
 	printf("Done\n");
 	return;
 }