[dpdk-dev] [PATCH v1] Move rte_mbuf macros to common header file

[Ananyev, Konstantin]

Hi Ravi,

> -----Original Message-----
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Ravi Kerur
> Sent: Saturday, September 26, 2015 3:47 AM
> To: Stephen Hemminger; Olivier Matz
> Cc: dev at dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v1] Move rte_mbuf macros to common header file
> 
> On Thu, Sep 24, 2015 at 4:25 PM, Stephen Hemminger <
> stephen at networkplumber.org> wrote:
> 
> > On Thu, 24 Sep 2015 15:50:41 -0700
> > Ravi Kerur <rkerur at gmail.com> wrote:
> >
> > > Macros RTE_MBUF_DATA_DMA_ADDR and RTE_MBUF_DATA_DMA_ADDR_DEFAULT
> > > are defined in each PMD driver file. Move those macros into common
> > > lib/librte_mbuf/rte_mbuf.h file. All PMD drivers include rte_mbuf.h
> > > file directly/indirectly hence no additionl header file inclusion
> > > is necessary.
> > >
> > > Compiled for:
> > >     > x86_64-native-linuxapp-clang
> > >     > x86_64-native-linuxapp-gcc
> > >     > i686-native-linuxapp-gcc
> > >     > x86_64-native-bsdapp-gcc
> > >     > x86_64-native-bsdapp-clang
> > >
> > > Tested on:
> > >     > x86_64 Ubuntu 14.04, testpmd and 'make test'
> > >     > FreeBSD 10.1, testpmd
> > >
> > > Signed-off-by: Ravi Kerur <rkerur at gmail.com>
> >
> > I like the idea, should have been done long ago.
> >
> > My only gripe is that you should do this as inline functions
> > rather than macros. Inline functions are type safe, macros are not.
> >
> 
> Agreed. However, I see another variation of the macro, users are primarily
> from "app" directory and lone user from drivers/net/xenvirt/virtqueue.h
> 
> #define RTE_MBUF_DATA_DMA_ADDR(mb) \
>         rte_pktmbuf_mtod(mb, uint64_t)


As I can see, it is used only in one place inside xenvirt:

drivers/net/xenvirt/virtqueue.h:        start_dp[idx].addr  = RTE_MBUF_DATA_DMA_ADDR(cookie);

So we probably can remove that macro definition here and use
rte_pktmbuf_mtod(mb, uint64_t) directly.

Konstantin

> 
> #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
> 
> #define rte_pktmbuf_mtod_offset(m, t, o)        \
>         ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
> 
> Let me know should I still go ahead and do inline variation for drivers or
> use above macro?