[PATCH] net/failsafe: Fix crash due to global devargs syntax parsing from secondary process

[PATCH] net/failsafe: Fix crash due to global devargs syntax parsing from secondary process

[madhuker.mythri]

From: Madhuker Mythri <madhuker.mythri@oracle.com>

Failsafe pmd started crashing with global devargs syntax as devargs is
not memset to zero. Access it to in rte_devargs_parse resulted in a
crash when called from secondary process.

Bugzilla Id: 933

Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
---
 drivers/net/failsafe/failsafe.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/failsafe/failsafe.c b/drivers/net/failsafe/failsafe.c
index 3c754a5f66..aa93cc6000 100644
--- a/drivers/net/failsafe/failsafe.c
+++ b/drivers/net/failsafe/failsafe.c
@@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
 			if (sdev->devargs.name[0] == '\0')
 				continue;
 
+			memset(&devargs, 0, sizeof(devargs));
 			/* rebuild devargs to be able to get the bus name. */
 			ret = rte_devargs_parse(&devargs,
 						sdev->devargs.name);
-- 
2.32.0.windows.1

(no subject)

[Ferruh Yigit]

On 2/10/2022 7:10 AM, madhuker.mythri@oracle.com wrote:
> From: Madhuker Mythri <madhuker.mythri@oracle.com>
> 
> Failsafe pmd started crashing with global devargs syntax as devargs is
> not memset to zero. Access it to in rte_devargs_parse resulted in a
> crash when called from secondary process.
> 
> Bugzilla Id: 933
> 
> Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
> ---
>   drivers/net/failsafe/failsafe.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/net/failsafe/failsafe.c b/drivers/net/failsafe/failsafe.c
> index 3c754a5f66..aa93cc6000 100644
> --- a/drivers/net/failsafe/failsafe.c
> +++ b/drivers/net/failsafe/failsafe.c
> @@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
>   			if (sdev->devargs.name[0] == '\0')
>   				continue;
>   
> +			memset(&devargs, 0, sizeof(devargs));
>   			/* rebuild devargs to be able to get the bus name. */
>   			ret = rte_devargs_parse(&devargs,
>   						sdev->devargs.name);

if 'rte_devargs_parse()' requires 'devargs' parameter to be memset,
what do you think memset it in the API?
This prevents forgotten cases like this.

Re:

[Gaëtan Rivet]

On Thu, Feb 10, 2022, at 16:00, Ferruh Yigit wrote:
> On 2/10/2022 7:10 AM, madhuker.mythri@oracle.com wrote:
>> From: Madhuker Mythri <madhuker.mythri@oracle.com>
>> 
>> Failsafe pmd started crashing with global devargs syntax as devargs is
>> not memset to zero. Access it to in rte_devargs_parse resulted in a
>> crash when called from secondary process.
>> 
>> Bugzilla Id: 933
>> 
>> Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
>> ---
>>   drivers/net/failsafe/failsafe.c | 1 +
>>   1 file changed, 1 insertion(+)
>> 
>> diff --git a/drivers/net/failsafe/failsafe.c b/drivers/net/failsafe/failsafe.c
>> index 3c754a5f66..aa93cc6000 100644
>> --- a/drivers/net/failsafe/failsafe.c
>> +++ b/drivers/net/failsafe/failsafe.c
>> @@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
>>   			if (sdev->devargs.name[0] == '\0')
>>   				continue;
>>   
>> +			memset(&devargs, 0, sizeof(devargs));
>>   			/* rebuild devargs to be able to get the bus name. */
>>   			ret = rte_devargs_parse(&devargs,
>>   						sdev->devargs.name);
>
> if 'rte_devargs_parse()' requires 'devargs' parameter to be memset,
> what do you think memset it in the API?
> This prevents forgotten cases like this.

Hi,

I was looking at it this morning.
Before the last release, rte_devargs_parse() was only supporting legacy syntax.
It never read from the devargs structure, only wrote to it. So it was safe to
use with a non-memset devargs.

The rte_devargs_layer_parse() however is more complex. To allow rte_dev_iterator_init() to call it without doing memory allocation, it reads parts of the devargs to make decisions.

Doing a first call to rte_devargs_layer_parse() as part of rte_devargs_parse() thus
modified the contract it had with the users, that it would never read from devargs.

It is not possible to completely avoid reading from devargs in rte_devargs_layer_parse().
It is necessary for RTE_DEV_FOREACH() to be safe to interrupt without having to do iterator cleanup.

This is my current understanding. In that context, yes I think it is preferrable to
do memset() within rte_devargs_parse(). It will restore the previous part of the API saying that calling it with non-memset devargs was safe to do.

Thanks,
-- 
Gaetan Rivet

RE: [External] : Re:

[Madhuker Mythri]

Hi Gaetan and  Ferruh,

>
>-----Original Message-----
>From: Gaëtan Rivet <grive@u256.net> 
>Sent: 10 फरवरी 2022 21:39
>To: Ferruh Yigit <ferruh.yigit@intel.com>; Madhuker Mythri <madhuker.mythri@oracle.com>
>Cc: dev@dpdk.org
>Subject: [External] : Re: 
>
>On Thu, Feb 10, 2022, at 16:00, Ferruh Yigit wrote:
>> On 2/10/2022 7:10 AM, madhuker.mythri@oracle.com wrote:
>>> From: Madhuker Mythri <madhuker.mythri@oracle.com>
>>> 
>>> Failsafe pmd started crashing with global devargs syntax as devargs 
>>> is not memset to zero. Access it to in rte_devargs_parse resulted in 
>>> a crash when called from secondary process.
>>> 
>>> Bugzilla Id: 933
>>> 
>>> Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
>>> ---
>>>   drivers/net/failsafe/failsafe.c | 1 +
>>>   1 file changed, 1 insertion(+)
>>> 
>>> diff --git a/drivers/net/failsafe/failsafe.c 
>>> b/drivers/net/failsafe/failsafe.c index 3c754a5f66..aa93cc6000 100644
>>> --- a/drivers/net/failsafe/failsafe.c
>>> +++ b/drivers/net/failsafe/failsafe.c
>>> @@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
>>>   			if (sdev->devargs.name[0] == '\0')
>>>   				continue;
>>>   
>>> +			memset(&devargs, 0, sizeof(devargs));
>>>   			/* rebuild devargs to be able to get the bus name. */
>>>   			ret = rte_devargs_parse(&devargs,
>>>   						sdev->devargs.name);
>>
>> if 'rte_devargs_parse()' requires 'devargs' parameter to be memset, 
>> what do you think memset it in the API?
>> This prevents forgotten cases like this.
>
>Hi,
>
>I was looking at it this morning.
>Before the last release, rte_devargs_parse() was only supporting legacy syntax.
>It never read from the devargs structure, only wrote to it. So it was safe to use with a non-memset devargs.
>
>The rte_devargs_layer_parse() however is more complex. To allow rte_dev_iterator_init() to call it without doing memory allocation, it reads parts of the devargs to make decisions.
>
>Doing a first call to rte_devargs_layer_parse() as part of rte_devargs_parse() thus modified the contract it had with the users, that it would never read from devargs.
>
>It is not possible to completely avoid reading from devargs in rte_devargs_layer_parse().
>It is necessary for RTE_DEV_FOREACH() to be safe to interrupt without having to do iterator cleanup.
>
>This is my current understanding. In that context, yes I think it is preferrable to do memset() within rte_devargs_parse(). It will restore the previous part of the API saying that calling it with non-memset >devargs was safe to do.
>
>Thanks,
>--
>Gaetan Rivet

Thanks for your comments.
The rte_devargs_parse() is used in other 'netvsc' PMD also in netvsc_hotadd_callback().
In this netvsc_hotadd_callback(), it was assigning the devargs with some other instance pointer(not sure, whether its just a pointer or with data values) before calling this rte_devargs_parse(), so if we memset inside this API, then the devargs data values will be nullified right.
I'm not fully familiar with this parsing functionality. So, please let me know, doing memset() inside this rte_devargs_parse() is valid or not, as this is a generic function for all the PMD's.

Thanks,
Madhuker.

Re: [External] : Re: [PATCH] net/failsafe: Fix crash due to global devargs syntax parsing from secondary process

[Ferruh Yigit]

On 2/11/2022 9:37 AM, Madhuker Mythri wrote:
> Hi Gaetan and  Ferruh,
> 
>>
>> -----Original Message-----
>> From: Gaëtan Rivet <grive@u256.net>
>> Sent: 10 फरवरी 2022 21:39
>> To: Ferruh Yigit <ferruh.yigit@intel.com>; Madhuker Mythri <madhuker.mythri@oracle.com>
>> Cc: dev@dpdk.org
>> Subject: [External] : Re:
>>
>> On Thu, Feb 10, 2022, at 16:00, Ferruh Yigit wrote:
>>> On 2/10/2022 7:10 AM, madhuker.mythri@oracle.com wrote:
>>>> From: Madhuker Mythri <madhuker.mythri@oracle.com>
>>>>
>>>> Failsafe pmd started crashing with global devargs syntax as devargs
>>>> is not memset to zero. Access it to in rte_devargs_parse resulted in
>>>> a crash when called from secondary process.
>>>>
>>>> Bugzilla Id: 933
>>>>
>>>> Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
>>>> ---
>>>>    drivers/net/failsafe/failsafe.c | 1 +
>>>>    1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/drivers/net/failsafe/failsafe.c
>>>> b/drivers/net/failsafe/failsafe.c index 3c754a5f66..aa93cc6000 100644
>>>> --- a/drivers/net/failsafe/failsafe.c
>>>> +++ b/drivers/net/failsafe/failsafe.c
>>>> @@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
>>>>    			if (sdev->devargs.name[0] == '\0')
>>>>    				continue;
>>>>    
>>>> +			memset(&devargs, 0, sizeof(devargs));
>>>>    			/* rebuild devargs to be able to get the bus name. */
>>>>    			ret = rte_devargs_parse(&devargs,
>>>>    						sdev->devargs.name);
>>>
>>> if 'rte_devargs_parse()' requires 'devargs' parameter to be memset,
>>> what do you think memset it in the API?
>>> This prevents forgotten cases like this.
>>
>> Hi,
>>
>> I was looking at it this morning.
>> Before the last release, rte_devargs_parse() was only supporting legacy syntax.
>> It never read from the devargs structure, only wrote to it. So it was safe to use with a non-memset devargs.
>>
>> The rte_devargs_layer_parse() however is more complex. To allow rte_dev_iterator_init() to call it without doing memory allocation, it reads parts of the devargs to make decisions.
>>
>> Doing a first call to rte_devargs_layer_parse() as part of rte_devargs_parse() thus modified the contract it had with the users, that it would never read from devargs.
>>
>> It is not possible to completely avoid reading from devargs in rte_devargs_layer_parse().
>> It is necessary for RTE_DEV_FOREACH() to be safe to interrupt without having to do iterator cleanup.
>>
>> This is my current understanding. In that context, yes I think it is preferrable to do memset() within rte_devargs_parse(). It will restore the previous part of the API saying that calling it with non-memset >devargs was safe to do.
>>
>> Thanks,
>> --
>> Gaetan Rivet
> 
> Thanks for your comments.
> The rte_devargs_parse() is used in other 'netvsc' PMD also in netvsc_hotadd_callback().
> In this netvsc_hotadd_callback(), it was assigning the devargs with some other instance pointer(not sure, whether its just a pointer or with data values) before calling this rte_devargs_parse(), so if we memset inside this API, then the devargs data values will be nullified right.
> I'm not fully familiar with this parsing functionality. So, please let me know, doing memset() inside this rte_devargs_parse() is valid or not, as this is a generic function for all the PMD's.
> 

Hi Madhuker,

Gaetan already sent a patch for it:
https://patches.dpdk.org/project/dpdk/patch/20220210170131.2199922-1-grive@u256.net/

Re: [External] : Re:

[Gaëtan Rivet]

On Fri, Feb 11, 2022, at 10:37, Madhuker Mythri wrote:
> Hi Gaetan and  Ferruh,
>
>>
>>-----Original Message-----
>>From: Gaëtan Rivet <grive@u256.net> 
>>Sent: 10 फरवरी 2022 21:39
>>To: Ferruh Yigit <ferruh.yigit@intel.com>; Madhuker Mythri <madhuker.mythri@oracle.com>
>>Cc: dev@dpdk.org
>>Subject: [External] : Re: 
>>
>>On Thu, Feb 10, 2022, at 16:00, Ferruh Yigit wrote:
>>> On 2/10/2022 7:10 AM, madhuker.mythri@oracle.com wrote:
>>>> From: Madhuker Mythri <madhuker.mythri@oracle.com>
>>>> 
>>>> Failsafe pmd started crashing with global devargs syntax as devargs 
>>>> is not memset to zero. Access it to in rte_devargs_parse resulted in 
>>>> a crash when called from secondary process.
>>>> 
>>>> Bugzilla Id: 933
>>>> 
>>>> Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
>>>> ---
>>>>   drivers/net/failsafe/failsafe.c | 1 +
>>>>   1 file changed, 1 insertion(+)
>>>> 
>>>> diff --git a/drivers/net/failsafe/failsafe.c 
>>>> b/drivers/net/failsafe/failsafe.c index 3c754a5f66..aa93cc6000 100644
>>>> --- a/drivers/net/failsafe/failsafe.c
>>>> +++ b/drivers/net/failsafe/failsafe.c
>>>> @@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
>>>>   			if (sdev->devargs.name[0] == '\0')
>>>>   				continue;
>>>>   
>>>> +			memset(&devargs, 0, sizeof(devargs));
>>>>   			/* rebuild devargs to be able to get the bus name. */
>>>>   			ret = rte_devargs_parse(&devargs,
>>>>   						sdev->devargs.name);
>>>
>>> if 'rte_devargs_parse()' requires 'devargs' parameter to be memset, 
>>> what do you think memset it in the API?
>>> This prevents forgotten cases like this.
>>
>>Hi,
>>
>>I was looking at it this morning.
>>Before the last release, rte_devargs_parse() was only supporting legacy syntax.
>>It never read from the devargs structure, only wrote to it. So it was safe to use with a non-memset devargs.
>>
>>The rte_devargs_layer_parse() however is more complex. To allow rte_dev_iterator_init() to call it without doing memory allocation, it reads parts of the devargs to make decisions.
>>
>>Doing a first call to rte_devargs_layer_parse() as part of rte_devargs_parse() thus modified the contract it had with the users, that it would never read from devargs.
>>
>>It is not possible to completely avoid reading from devargs in rte_devargs_layer_parse().
>>It is necessary for RTE_DEV_FOREACH() to be safe to interrupt without having to do iterator cleanup.
>>
>>This is my current understanding. In that context, yes I think it is preferrable to do memset() within rte_devargs_parse(). It will restore the previous part of the API saying that calling it with non-memset >devargs was safe to do.
>>
>>Thanks,
>>--
>>Gaetan Rivet
>
> Thanks for your comments.
> The rte_devargs_parse() is used in other 'netvsc' PMD also in 
> netvsc_hotadd_callback().
> In this netvsc_hotadd_callback(), it was assigning the devargs with 
> some other instance pointer(not sure, whether its just a pointer or 
> with data values) before calling this rte_devargs_parse(), so if we 
> memset inside this API, then the devargs data values will be nullified 
> right.
> I'm not fully familiar with this parsing functionality. So, please let 
> me know, doing memset() inside this rte_devargs_parse() is valid or 
> not, as this is a generic function for all the PMD's.
>
> Thanks,
> Madhuker.

Hi Madhuker,

I have just checked the code, it does not seem that netvsc relies on values
being kept in the devargs structure across calls, so it seems that it would be
safe. It would be better to check with the maintainers however, I'm adding them to
this thread.

-- 
Gaetan Rivet

RE: [External] : Re:

[Long Li]

> Subject: Re: [External] : Re:
> 
> On Fri, Feb 11, 2022, at 10:37, Madhuker Mythri wrote:
> > Hi Gaetan and  Ferruh,
> >
> >>
> >>-----Original Message-----
> >>From: Gaëtan Rivet <grive@u256.net>
> >>Sent: 10 फरवरी 2022 21:39
> >>To: Ferruh Yigit <ferruh.yigit@intel.com>; Madhuker Mythri
> >><madhuker.mythri@oracle.com>
> >>Cc: dev@dpdk.org
> >>Subject: [External] : Re:
> >>
> >>On Thu, Feb 10, 2022, at 16:00, Ferruh Yigit wrote:
> >>> On 2/10/2022 7:10 AM, madhuker.mythri@oracle.com wrote:
> >>>> From: Madhuker Mythri <madhuker.mythri@oracle.com>
> >>>>
> >>>> Failsafe pmd started crashing with global devargs syntax as devargs
> >>>> is not memset to zero. Access it to in rte_devargs_parse resulted
> >>>> in a crash when called from secondary process.
> >>>>
> >>>> Bugzilla Id: 933
> >>>>
> >>>> Signed-off-by: Madhuker Mythri <madhuker.mythri@oracle.com>
> >>>> ---
> >>>>   drivers/net/failsafe/failsafe.c | 1 +
> >>>>   1 file changed, 1 insertion(+)
> >>>>
> >>>> diff --git a/drivers/net/failsafe/failsafe.c
> >>>> b/drivers/net/failsafe/failsafe.c index 3c754a5f66..aa93cc6000
> >>>> 100644
> >>>> --- a/drivers/net/failsafe/failsafe.c
> >>>> +++ b/drivers/net/failsafe/failsafe.c
> >>>> @@ -360,6 +360,7 @@ rte_pmd_failsafe_probe(struct
> rte_vdev_device *vdev)
> >>>>   			if (sdev->devargs.name[0] == '\0')
> >>>>   				continue;
> >>>>
> >>>> +			memset(&devargs, 0, sizeof(devargs));
> >>>>   			/* rebuild devargs to be able to get the bus name. */
> >>>>   			ret = rte_devargs_parse(&devargs,
> >>>>   						sdev->devargs.name);
> >>>
> >>> if 'rte_devargs_parse()' requires 'devargs' parameter to be memset,
> >>> what do you think memset it in the API?
> >>> This prevents forgotten cases like this.
> >>
> >>Hi,
> >>
> >>I was looking at it this morning.
> >>Before the last release, rte_devargs_parse() was only supporting legacy
> syntax.
> >>It never read from the devargs structure, only wrote to it. So it was safe to
> use with a non-memset devargs.
> >>
> >>The rte_devargs_layer_parse() however is more complex. To allow
> rte_dev_iterator_init() to call it without doing memory allocation, it reads
> parts of the devargs to make decisions.
> >>
> >>Doing a first call to rte_devargs_layer_parse() as part of
> rte_devargs_parse() thus modified the contract it had with the users, that it
> would never read from devargs.
> >>
> >>It is not possible to completely avoid reading from devargs in
> rte_devargs_layer_parse().
> >>It is necessary for RTE_DEV_FOREACH() to be safe to interrupt without
> having to do iterator cleanup.
> >>
> >>This is my current understanding. In that context, yes I think it is
> preferrable to do memset() within rte_devargs_parse(). It will restore the
> previous part of the API saying that calling it with non-memset >devargs was
> safe to do.
> >>
> >>Thanks,
> >>--
> >>Gaetan Rivet
> >
> > Thanks for your comments.
> > The rte_devargs_parse() is used in other 'netvsc' PMD also in
> > netvsc_hotadd_callback().
> > In this netvsc_hotadd_callback(), it was assigning the devargs with
> > some other instance pointer(not sure, whether its just a pointer or
> > with data values) before calling this rte_devargs_parse(), so if we
> > memset inside this API, then the devargs data values will be nullified
> > right.
> > I'm not fully familiar with this parsing functionality. So, please let
> > me know, doing memset() inside this rte_devargs_parse() is valid or
> > not, as this is a generic function for all the PMD's.
> >
> > Thanks,
> > Madhuker.
> 
> Hi Madhuker,
> 
> I have just checked the code, it does not seem that netvsc relies on values
> being kept in the devargs structure across calls, so it seems that it would be
> safe. It would be better to check with the maintainers however, I'm adding
> them to this thread.
> 
> --
> Gaetan Rivet

When netvsc calls rte_devargs_parse, it assumes the value in rte_devargs will be overwritten. There is an issue with netvsc dealing multiple devices, that is on my todo list.

In this case, I think it's safe to do memset() within rte_devargs_parse().

Long

[PATCH] devargs: Fix rte_devargs_parse uninitialized calls

[Gaetan Rivet]

The function rte_devargs_parse() previously was safe to call with
non-initialized devargs structure as parameter.

When adding the support for the global device syntax,
this assumption was broken. Restore it by forcing memset as part of
the call itself.

Bugzilla Id: 933
Fixes: b344eb5d941a ("devargs: parse global device syntax")
Signed-off-by: Gaetan Rivet <grive@u256.net>
---
 lib/eal/common/eal_common_devargs.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/lib/eal/common/eal_common_devargs.c b/lib/eal/common/eal_common_devargs.c
index 8c7650cf6c..184fe676aa 100644
--- a/lib/eal/common/eal_common_devargs.c
+++ b/lib/eal/common/eal_common_devargs.c
@@ -191,6 +191,7 @@ rte_devargs_parse(struct rte_devargs *da, const char *dev)
 
 	if (da == NULL)
 		return -EINVAL;
+	memset(da, 0, sizeof(*da));
 
 	/* First parse according global device syntax. */
 	if (rte_devargs_layers_parse(da, dev) == 0) {
-- 
2.31.1

Re: [PATCH] devargs: Fix rte_devargs_parse uninitialized calls

[Ferruh Yigit]

On 2/10/2022 5:01 PM, Gaetan Rivet wrote:
> The function rte_devargs_parse() previously was safe to call with
> non-initialized devargs structure as parameter.
> 
> When adding the support for the global device syntax,
> this assumption was broken. Restore it by forcing memset as part of
> the call itself.
> 
> Bugzilla Id: 933
> Fixes: b344eb5d941a ("devargs: parse global device syntax")
> Signed-off-by: Gaetan Rivet <grive@u256.net>

Reported-by: Madhuker Mythri <madhuker.mythri@oracle.com>

Since Madhuker did the initial analysis and the patch
I think fair to give credit for the work.

Thanks,
ferruh

> ---
>   lib/eal/common/eal_common_devargs.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/lib/eal/common/eal_common_devargs.c b/lib/eal/common/eal_common_devargs.c
> index 8c7650cf6c..184fe676aa 100644
> --- a/lib/eal/common/eal_common_devargs.c
> +++ b/lib/eal/common/eal_common_devargs.c
> @@ -191,6 +191,7 @@ rte_devargs_parse(struct rte_devargs *da, const char *dev)
>   
>   	if (da == NULL)
>   		return -EINVAL;
> +	memset(da, 0, sizeof(*da));
>   
>   	/* First parse according global device syntax. */
>   	if (rte_devargs_layers_parse(da, dev) == 0) {

Re: [PATCH] devargs: Fix rte_devargs_parse uninitialized calls

[Gaëtan Rivet]

On Tue, Feb 15, 2022, at 13:51, Ferruh Yigit wrote:
> On 2/10/2022 5:01 PM, Gaetan Rivet wrote:
>> The function rte_devargs_parse() previously was safe to call with
>> non-initialized devargs structure as parameter.
>> 
>> When adding the support for the global device syntax,
>> this assumption was broken. Restore it by forcing memset as part of
>> the call itself.
>> 
>> Bugzilla Id: 933
>> Fixes: b344eb5d941a ("devargs: parse global device syntax")
>> Signed-off-by: Gaetan Rivet <grive@u256.net>
>
> Reported-by: Madhuker Mythri <madhuker.mythri@oracle.com>
>
> Since Madhuker did the initial analysis and the patch
> I think fair to give credit for the work.
>
> Thanks,
> ferruh
>

Sure, thanks Ferruh. Whichever patch is fine by me.

-- 
Gaetan Rivet

Re: [PATCH] devargs: Fix rte_devargs_parse uninitialized calls

[David Marchand]

On Thu, Feb 10, 2022 at 6:01 PM Gaetan Rivet <grive@u256.net> wrote:
>
> The function rte_devargs_parse() previously was safe to call with
> non-initialized devargs structure as parameter.
>
> When adding the support for the global device syntax,
> this assumption was broken. Restore it by forcing memset as part of
> the call itself.
>
> Bugzilla Id: 933

Nit: Bugzilla ID*

> Fixes: b344eb5d941a ("devargs: parse global device syntax")

We need a backport in 21.11, right?


> Signed-off-by: Gaetan Rivet <grive@u256.net>
> ---
>  lib/eal/common/eal_common_devargs.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/lib/eal/common/eal_common_devargs.c b/lib/eal/common/eal_common_devargs.c
> index 8c7650cf6c..184fe676aa 100644
> --- a/lib/eal/common/eal_common_devargs.c
> +++ b/lib/eal/common/eal_common_devargs.c
> @@ -191,6 +191,7 @@ rte_devargs_parse(struct rte_devargs *da, const char *dev)
>
>         if (da == NULL)
>                 return -EINVAL;
> +       memset(da, 0, sizeof(*da));
>
>         /* First parse according global device syntax. */
>         if (rte_devargs_layers_parse(da, dev) == 0) {
> --
> 2.31.1
>

Otherwise lgtm.


-- 
David Marchand

Re: [PATCH] devargs: Fix rte_devargs_parse uninitialized calls

[Gaëtan Rivet]

On Tue, Feb 15, 2022, at 16:27, David Marchand wrote:
> On Thu, Feb 10, 2022 at 6:01 PM Gaetan Rivet <grive@u256.net> wrote:
>>
>> The function rte_devargs_parse() previously was safe to call with
>> non-initialized devargs structure as parameter.
>>
>> When adding the support for the global device syntax,
>> this assumption was broken. Restore it by forcing memset as part of
>> the call itself.
>>
>> Bugzilla Id: 933
>
> Nit: Bugzilla ID*
>
>> Fixes: b344eb5d941a ("devargs: parse global device syntax")
>
> We need a backport in 21.11, right?

Hi David, yes it should be in 21.11 as well.

-- 
Gaetan Rivet

[PATCH v3 0/7] net/gve: RSS Support for GVE Driver

[Joshua Washington]

This patch series introduces RSS support for the GVE poll-mode driver.
This series includes implementations of the following eth_dev_ops:

1) rss_hash_update
2) rss_hash_conf_get
3) reta_query
4) reta_update

In rss_hash_update, the GVE driver supports the following RSS hash
types:

* RTE_ETH_RSS_IPV4
* RTE_ETH_RSS_NONFRAG_IPV4_TCP
* RTE_ETH_RSS_NONFRAG_IPV4_UDP
* RTE_ETH_RSS_IPV6
* RTE_ETH_RSS_IPV6_EX
* RTE_ETH_RSS_NONFRAG_IPV6_TCP
* RTE_ETH_RSS_NONFRAG_IPV6_UDP
* RTE_ETH_RSS_IPV6_TCP_EX
* RTE_ETH_RSS_IPV6_UDP_EX

The hash key is 40B, and the lookup table has 128 entries. These values
are not configurable in this implementation.

In general, the DPDK driver expects the RSS hash configuration to be set
with a key before the redriection table is set up. When the RSS hash is
configured, a default redirection table is generated based on the number
of queues. When the device is re-configured, the redirection table is
reset to the default value based on the queue count.

An important note is that the gVNIC device expects 32 bit integers for
RSS redirection table entries, while the RTE API uses 16 bit integers.
However, this is unlikely to be an issue, as these values represent
receive queues, and the gVNIC device does not support anywhere near 64K
queues.

This series also updates the corresponding feature matrix ertries and
documentation as it pertains to RSS support in the GVE driver.

v2:
Add commmit messages for patches with it missing, and other checkpatches
fixes.

Note: There is a warning about complex macros being parenthesized that
does not seem to be well-founded.

v3:
Fix build warnings that come up on certain distros.

Joshua Washington (7):
  net/gve: fully expose RSS offload support in dev_info
  net/gve: RSS adminq command changes
  net/gve: add gve_rss library for handling RSS-related behaviors
  net/gve: RSS configuration update support
  net/gve: RSS redirection table update support
  net/gve: update gve.ini with RSS capabilities
  net/gve: update GVE documentation with RSS support

 doc/guides/nics/features/gve.ini  |   3 +
 doc/guides/nics/gve.rst           |  16 ++-
 drivers/net/gve/base/gve.h        |  15 ++
 drivers/net/gve/base/gve_adminq.c |  59 ++++++++
 drivers/net/gve/base/gve_adminq.h |  21 +++
 drivers/net/gve/gve_ethdev.c      | 231 +++++++++++++++++++++++++++++-
 drivers/net/gve/gve_ethdev.h      |  17 +++
 drivers/net/gve/gve_rss.c         | 206 ++++++++++++++++++++++++++
 drivers/net/gve/gve_rss.h         | 107 ++++++++++++++
 drivers/net/gve/meson.build       |   1 +
 10 files changed, 667 insertions(+), 9 deletions(-)
 create mode 100644 drivers/net/gve/gve_rss.c
 create mode 100644 drivers/net/gve/gve_rss.h

-- 
2.43.0.429.g432eaa2c6b-goog

Re:

[Ferruh Yigit]

On 1/26/2024 5:33 PM, Joshua Washington wrote:
> Subject: [PATCH v4 0/7] net/gve: RSS Support for GVE Driver
> 
> This patch series introduces RSS support for the GVE poll-mode driver.
> This series includes implementations of the following eth_dev_ops:
> 
> 1) rss_hash_update
> 2) rss_hash_conf_get
> 3) reta_query
> 4) reta_update
> 
> In rss_hash_update, the GVE driver supports the following RSS hash
> types:
> 
> * RTE_ETH_RSS_IPV4
> * RTE_ETH_RSS_NONFRAG_IPV4_TCP
> * RTE_ETH_RSS_NONFRAG_IPV4_UDP
> * RTE_ETH_RSS_IPV6
> * RTE_ETH_RSS_IPV6_EX
> * RTE_ETH_RSS_NONFRAG_IPV6_TCP
> * RTE_ETH_RSS_NONFRAG_IPV6_UDP
> * RTE_ETH_RSS_IPV6_TCP_EX
> * RTE_ETH_RSS_IPV6_UDP_EX
> 
> The hash key is 40B, and the lookup table has 128 entries. These values
> are not configurable in this implementation.
> 
> In general, the DPDK driver expects the RSS hash configuration to be set
> with a key before the redriection table is set up. When the RSS hash is
> configured, a default redirection table is generated based on the number
> of queues. When the device is re-configured, the redirection table is
> reset to the default value based on the queue count.
> 
> An important note is that the gVNIC device expects 32 bit integers for
> RSS redirection table entries, while the RTE API uses 16 bit integers.
> However, this is unlikely to be an issue, as these values represent
> receive queues, and the gVNIC device does not support anywhere near 64K
> queues.
> 
> This series also updates the corresponding feature matrix ertries and
> documentation as it pertains to RSS support in the GVE driver.
> 
> v2:
> Add commmit messages for patches with it missing, and other checkpatches
> fixes.
> 
> Note: There is a warning about complex macros being parenthesized that
> does not seem to be well-founded.
> 
> v3:
> Fix build warnings that come up on certain distros.
> 
> v4:
> Fix formatting in gve_adminq.c
> 
> Joshua Washington (7):
>   net/gve: fully expose RSS offload support in dev_info
>   net/gve: RSS adminq command changes
>   net/gve: add gve_rss library for handling RSS-related behaviors
>   net/gve: RSS configuration update support
>   net/gve: RSS redirection table update support
>   net/gve: update gve.ini with RSS capabilities
>   net/gve: update GVE documentation with RSS support
> 
> 

'./devtools/check-git-log.sh' script is giving warnings [1]:

Expected patch title format is:
net/gve: <verb> <object>

<verb> should start with lowercase.


[1]
- check-git-log:




Wrong headline format:




        net/gve: fully expose RSS offload support in dev_info




        net/gve: add gve_rss library for handling RSS-related behaviors




Wrong headline uppercase:




        net/gve: RSS adminq command changes




        net/gve: RSS configuration update support




        net/gve: RSS redirection table update support




Headline too long:




        net/gve: add gve_rss library for handling RSS-related behaviors