Re: [dpdk-dev] [PATCH 2/2] ethdev: fix shallow copy of flow API RAW item

[Adrien Mazarguil <adrien.mazarguil@6wind.com>]

On Mon, May 21, 2018 at 11:44:33AM +0100, Ferruh Yigit wrote:
> On 5/21/2018 9:24 AM, Adrien Mazarguil wrote:
> > On Sat, May 19, 2018 at 04:25:15PM +0200, Thomas Monjalon wrote:
> >> 18/05/2018 19:06, Ferruh Yigit:
> >>> On 5/16/2018 4:41 PM, Adrien Mazarguil wrote:
> >>>> Like original commit mentioned below, this fix synchronizes flow rule copy
> >>>> function with testpmd's own implementation following "app/testpmd: fix copy
> >>>> of raw flow item (revisited)".
> >>>>
> >>>> Fixes: d0ad8648b1c5 ("ethdev: fix shallow copy of flow API RSS action")
> >>>> Cc: stable@dpdk.org
> >>>> Cc: Qi Zhang <qi.z.zhang@intel.com>
> >>>>
> >>>> Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
> >>>
> >>> Hi Thomas,
> >>>
> >>> What do you suggest about this one?
> >>> Scope is limited to rte_flow but still many features are now relies on rte_flow,
> >>> what is your comment on getting this in rc5?
> >>
> >> We need to know exactly what is broken.
> >> If nothing serious, it can wait 18.08.
> >>
> >> Adrien, please can you describe the use case, the issue and the impact?
> > 
> > A prior patch [1] (applied as "app/testpmd: fix copy of raw flow item"),
> > addresses a crash in testpmd's flow copy function.
> > 
> > The first patch of the present series [2] addresses remaining issues with
> > its behavior which is, in fact, what caused the original issue.
> > 
> > While both patches focus on testpmd, rte_flow also exposes its own public
> > copy function with the exact same code that breaks when encountering a RAW
> > pattern item. Primary users for this function are bonding and failsafe
> > PMDs. 
> > 
> > This patch therefore addresses both [1] and [2] at once for rte_flow_copy().
> 
> Hi Adrien,
> 
> What is the effect of _not_ getting this patch, just trying to understand if
> this is something to get for this release or postpone to next one.

Hi Ferruh,

*Not* getting this patch means rte_flow_copy() crashes when user creates a
flow rule that involves the RAW pattern item on top of either bonding or
failsafe PMDs.

Ditto for any external application that relies on rte_flow_copy() combined
with the RAW pattern item.

I'll send v2 to provide a bit more info and fix the wrong commit ID on the
"Fixes" line.

> > [1] "app/testpmd: fix invalid memory access"
> >     http://dpdk.org/ml/archives/dev/2018-May/100364.html
> > 
> > [2] "app/testpmd: fix copy of raw flow item (revisited)"
> >     http://dpdk.org/ml/archives/dev/2018-May/101994.html

-- 
Adrien Mazarguil
6WIND