Re: [dpdk-dev] [PATCH 3/3] ethdev: fix build warning on 64-bit value

["Morten Brørup" <mb@smartsharesystems.com>]

> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> Sent: Tuesday, April 28, 2020 11:28 AM
> 
> 27/04/2020 16:00, David Marchand:
> > On Mon, Apr 27, 2020 at 3:47 PM Ananyev, Konstantin wrote:
> > > From: dev <dev-bounces@dpdk.org> On Behalf Of David Marchand
> > > > On Mon, Apr 27, 2020 at 3:34 PM Bruce Richardson
> > > > <bruce.richardson@intel.com> wrote:
> > > > > On Mon, Apr 27, 2020 at 03:23:41PM +0200, David Marchand wrote:
> > > > > > Building OVS with dpdk, sparse complains about 64-bit
> constant being
> > > > > > passed as a normal integer that can't fit it:
> > > > > > error: constant 0xffffffffffffffff is so big it is unsigned
> long
> > > > > >
> > > > > > Fixes: ecbc8570131d ("ethdev: add PFCP header to flow API")
> > > > > >
> > > > > > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > > > > > ---
> > > > > > --- a/lib/librte_ethdev/rte_flow.h
> > > > > > +++ b/lib/librte_ethdev/rte_flow.h
> > > > > >  static const struct rte_flow_item_pfcp
> rte_flow_item_pfcp_mask = {
> > > > > >       .s_field = 0x01,
> > > > > > -     .seid = RTE_BE64(0xffffffffffffffff),
> > > > > > +     .seid = RTE_BE64(UINT64_C(0xffffffffffffffff)),
> > > > >
> > > > > Rather than cast, why not put "ULL" at the end. If we are going
> to cast,
> > > > > why not just put "-1" in to save some digits.
> > > >
> > > > I preferred this form in the hope future developers who want
> > > > 0x0fffffffffffffff will copy/paste this.
> > > >
> > >
> > > As I remember there should be UINT64_MAX in stdint.h.
> >
> > Yes, we could go with:
> > +     .seid = RTE_BE64(UINT64_MAX),
> >
> > And then next time, for any value like 0x0fff ffff ffff ffff (had to
> > group the digits of what I had written), pretty sure we will miss
> this
> > and I will catch it only when building ovs.
> 
> I agree with David (in general and especially here).
> 
> RTE_BE64 is required for static analyzers and is an explicit info.
> 
> UINT64_C is better than ULL suffix because it
> 	- is generic
> 	- gives size explicitly

Certainly. Explicit is preferred in code for embedded systems.
"unsigned long long" means 64 bit or more, which also applies to the "ULL" postfix.
"uint64_t" and "UINT64_C" means exactly 64 bit.

> 
> UINT64_C(0xffffffffffffffff) is better than UINT64_MAX because
> 	- rte_flow.h has a lot of bitmasks
> 	- it is copy/paste safe
> 
> Acked-by: Thomas Monjalon <thomas@monjalon.net>
> 

And shouldn't the struct rte_flow_item_pfcp be packed? I would expect the compiler to add 32 bit padding before seid to ensure its 64 bit alignment.


Med venlig hilsen / kind regards
- Morten Brørup