Re: [dpdk-dev] [PATCH v2] mbuf: support dynamic fields and flags

[Olivier Matz <olivier.matz@6wind.com>]

On Fri, Oct 18, 2019 at 08:28:02AM +0000, Wang, Haiyue wrote:
> Hi Olivier,
> 
> > -----Original Message-----
> > From: Olivier Matz [mailto:olivier.matz@6wind.com]
> > Sent: Friday, October 18, 2019 15:54
> > To: Wang, Haiyue <haiyue.wang@intel.com>
> > Cc: dev@dpdk.org; Andrew Rybchenko <arybchenko@solarflare.com>; Richardson, Bruce
> > <bruce.richardson@intel.com>; Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Wiles, Keith
> > <keith.wiles@intel.com>; Ananyev, Konstantin <konstantin.ananyev@intel.com>; Morten Brørup
> > <mb@smartsharesystems.com>; Stephen Hemminger <stephen@networkplumber.org>; Thomas Monjalon
> > <thomas@monjalon.net>
> > Subject: Re: [PATCH v2] mbuf: support dynamic fields and flags
> > 
> > Hi Haiyue,
> > 
> > On Fri, Oct 18, 2019 at 02:47:50AM +0000, Wang, Haiyue wrote:
> > > Hi Olivier
> > >
> > > > -----Original Message-----
> > > > From: Olivier Matz [mailto:olivier.matz@6wind.com]
> > > > Sent: Thursday, October 17, 2019 22:42
> > > > To: dev@dpdk.org
> > > > Cc: Andrew Rybchenko <arybchenko@solarflare.com>; Richardson, Bruce <bruce.richardson@intel.com>;
> > Wang,
> > > > Haiyue <haiyue.wang@intel.com>; Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Wiles, Keith
> > > > <keith.wiles@intel.com>; Ananyev, Konstantin <konstantin.ananyev@intel.com>; Morten Brørup
> > > > <mb@smartsharesystems.com>; Stephen Hemminger <stephen@networkplumber.org>; Thomas Monjalon
> > > > <thomas@monjalon.net>
> > > > Subject: [PATCH v2] mbuf: support dynamic fields and flags
> > > >
> > > > Many features require to store data inside the mbuf. As the room in mbuf
> > > > structure is limited, it is not possible to have a field for each
> > > > feature. Also, changing fields in the mbuf structure can break the API
> > > > or ABI.
> > > >
> > > > This commit addresses these issues, by enabling the dynamic registration
> > > > of fields or flags:
> > > >
> > > > - a dynamic field is a named area in the rte_mbuf structure, with a
> > > >   given size (>= 1 byte) and alignment constraint.
> > > > - a dynamic flag is a named bit in the rte_mbuf structure.
> > > >
> > > > The typical use case is a PMD that registers space for an offload
> > > > feature, when the application requests to enable this feature.  As
> > > > the space in mbuf is limited, the space should only be reserved if it
> > > > is going to be used (i.e when the application explicitly asks for it).
> > > >
> > > > The registration can be done at any moment, but it is not possible
> > > > to unregister fields or flags for now.
> > > >
> > > > Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> > > > Acked-by: Thomas Monjalon <thomas@monjalon.net>
> > > > ---
> > > >
> > > > v2
> > > >
> > > > * Rebase on top of master: solve conflict with Stephen's patchset
> > > >   (packet copy)
> > > > * Add new apis to register a dynamic field/flag at a specific place
> > > > * Add a dump function (sugg by David)
> > > > * Enhance field registration function to select the best offset, keeping
> > > >   large aligned zones as much as possible (sugg by Konstantin)
> > > > * Use a size_t and unsigned int instead of int when relevant
> > > >   (sugg by Konstantin)
> > > > * Use "uint64_t dynfield1[2]" in mbuf instead of 2 uint64_t fields
> > > >   (sugg by Konstantin)
> > > > * Remove unused argument in private function (sugg by Konstantin)
> > > > * Fix and simplify locking (sugg by Konstantin)
> > > > * Fix minor typo
> > > >
> > > > rfc -> v1
> > > >
> > > > * Rebase on top of master
> > > > * Change registration API to use a structure instead of
> > > >   variables, getting rid of #defines (Stephen's comment)
> > > > * Update flag registration to use a similar API as fields.
> > > > * Change max name length from 32 to 64 (sugg. by Thomas)
> > > > * Enhance API documentation (Haiyue's and Andrew's comments)
> > > > * Add a debug log at registration
> > > > * Add some words in release note
> > > > * Did some performance tests (sugg. by Andrew):
> > > >   On my platform, reading a dynamic field takes ~3 cycles more
> > > >   than a static field, and ~2 cycles more for writing.
> > > >
> > > >  app/test/test_mbuf.c                   | 145 ++++++-
> > > >  doc/guides/rel_notes/release_19_11.rst |   7 +
> > > >  lib/librte_mbuf/Makefile               |   2 +
> > > >  lib/librte_mbuf/meson.build            |   6 +-
> > > >  lib/librte_mbuf/rte_mbuf.h             |  23 +-
> > > >  lib/librte_mbuf/rte_mbuf_dyn.c         | 548 +++++++++++++++++++++++++
> > > >  lib/librte_mbuf/rte_mbuf_dyn.h         | 226 ++++++++++
> > > >  lib/librte_mbuf/rte_mbuf_version.map   |   7 +
> > > >  8 files changed, 959 insertions(+), 5 deletions(-)
> > > >  create mode 100644 lib/librte_mbuf/rte_mbuf_dyn.c
> > > >  create mode 100644 lib/librte_mbuf/rte_mbuf_dyn.h
> > > >
> > > > diff --git a/app/test/test_mbuf.c b/app/test/test_mbuf.c
> > > > index b9c2b2500..01cafad59 100644
> > > > --- a/app/test/test_mbuf.c
> > > > +++ b/app/test/test_mbuf.c
> > > > @@ -28,6 +28,7 @@
> > > >  #include <rte_random.h>
> > >
> > > [snip]
> > >
> > > > +/**
> > > > + * Helper macro to access to a dynamic field.
> > > > + */
> > > > +#define RTE_MBUF_DYNFIELD(m, offset, type) ((type)((uintptr_t)(m) + (offset)))
> > > > +
> > >
> > > The suggested macro is missed ? ;-)
> > > 	/**
> > > 	 * Helper macro to access to a dynamic flag.
> > > 	 */
> > > 	#define RTE_MBUF_DYNFLAG(offset) (1ULL << (offset))
> > 
> > Yes, sorry.
> > 
> > Thinking a bit more about it, I wonder if the macros below aren't
> > more consistent with the dynamic field (because they take the mbuf
> > as parameter)?
> > 
> >   #define RTE_MBUF_SET_DYNFLAG(m, bitnum, val) ...
> >   #define RTE_MBUF_GET_DYNFLAG(m, bitnum) ...
> > 
> > They could even be static inline functions.
> > 
> > On the other hand, these helpers would be generic to ol_flags, not only
> > for dynamic flags. Today, we use (1ULL << bit) for ol_flags, which makes
> > me wonder... is the macro really needed after all? :)
> > 
> 
> I used as this:
> 	1). 	in PMD:
> 		mb->ol_flags |= RTE_MBUF_DYNFLAG(ol_offset); 
> 
> 
> 	2). In testpmd
> 		if (mb->ol_flags & RTE_MBUF_DYNFLAG(ol_offset))
> 			...
> 
> The above two macros look better in real use.

I just looked at http://patchwork.dpdk.org/patch/60908/
In the patch, a mask is used instead of a bit number, which is indeed
better in terms of performance. This makes the macro not that useful,
given there is a specific helper.


> > > BTW, should we have a place to put the registered dynamic fields and flags
> > > names together (a name overview -- detail Link to --> PMD's help page) ?
> > 
> > The centralized place will be in rte_mbuf_dyn.h for fields/flags that can
> > are shared between several dpdk areas. Some libraries/pmd could have private
> > dynamic fields/flags. In any case, I think the same namespace than functions
> > should be used. Probably something like this:
> >  - "rte_mbuf_dynfield_<name>" in mbuf lib
> >  - "rte_<libname>_dynfield_<name>" in other libs
> >  - "rte_net_<pmd>_dynfield_<name>" in pmds
> >  - "<name>" in apps
> > 
> > > Since rte_mbuf_dynfield:name & rte_mbuf_dynflag:name work as a API style,
> > > users can check how many 'names' registered, developers can check whether
> > > the names they want to use are registered or not ? They don't need to have
> > > to check the rte_errno ... Just a suggestion for user experience.
> > 
> > I did not get you point. Does my response above answers to your question?
> > 
> 
> Yes, the name conversation you mentioned above is a good practice, then no doc
> needed any more, thanks!
> 
> > Regards,
> > Olivier