-----Original Message-----
> Date: Tue, 11 Jul 2017 15:19:26 +0100
> From: Harry van Haaren <harry.van.haa...@intel.com>
> To: dev@dpdk.org
> CC: tho...@monjalon.net, jerin.ja...@caviumnetworks.com,
> keith.wi...@intel.com, bruce.richard...@intel.com, Harry van Haaren
> <harry.van.haa...@intel.com>
> Subject: [PATCH v5 0/7] service cores: cover letter
> X-Mailer: git-send-email 2.7.4
>
> This patchset introduces service cores to DPDK. A service core
> is an lcore that performs functions to abstract away details of
> differences in environment of the application.
>
> An example is using the eventdev API, where either a software or hardware
> PMD performs scheduling. In the case of the software PMD an lcore is
> required to perform scheduling, which means application logic would have
> to be aware of the PMD running under the API. To abstract away the
> differences in HW / SW PMDs, service cores can run the SW PMD service
> without application logic specifying the exact cores to use. Note that
> eventdev is only one API that benefits; timers, interrupts handling,
> statistics and monitoring, and a range of other infrastructure that
> requires a slice of CPU time may all benefit from service cores.
>
> The application is not obliged to manually use the service cores API,
> however if an application wishes to use the service cores API for fine
> grained control over how the services are run, this is possible. Deciding
> between a performance threading-profile and scaled-down profile can be
> achieved by advanced usage of service cores and setting the lcore mappings.
>
> Patch 5/7 shows how a PMD can register a service to run a function. This
> is then available (along with any other registered services) to be run by
> the service cores. Patches 6/7 and 7/7 add documentation, and claim
> maintainership.
>
> Regards, -Harry
Nice feature. This series Looks good to me.
Series Acked-by: Jerin Jacob <jerin.ja...@caviumnetworks.com>
IMO, We can add this series in RC2 if there are no objections.
Two points:
1) There is a check-git log warning. It could be fixed it on apply.
Wrong headline uppercase:
service cores: EAL init changes
2) Based on the mail[1], It is agreed that we need to refactor
eal_parse_service_coremask() and
eal_parse__coremask() code in future to avoid code duplication in coremask
parsing
logic.
[1]
http://dpdk.org/ml/archives/dev/2017-July/070610.html
>
> v5:
> Jerin:
> - Fix documentation warnings
> - Rename variables to better names
> - Enable statistics per-service
> - Improve atomic operation flag checks
> - Reworked function to rte_service_start_with_defaults()
> - Added memory barriers to lcore_add() and lcore_del()
> - Simplified EAL code, reduced duplication and makes it more maintainable
>
> Jerin/Thomas:
> - Rename component header to rte_service_component.h
>
> v4:
> - Range of fixes as suggested by Jerin
> - Improved unit tests, ensuring ex-service cores become available to app
> - Added functions to EXPERIMENTAL tag in .map files (Thomas)
> - Added @warning experimental notes to Doxygen API documentation (Thomas)
> - Various smaller fixes / cleanups
> - See commit notes for details
>
> v3:
> - Added docs
> - Added release notes
> - Updated maintainers file
> - Compile checks with devtools/test-build.sh
> - Validated patches apply to latest dpdk/master
> - Based on discussion, rte_service_iterate() is *not* included,
> but could be adding at a later date if use-cases require it.
> - Future work includes enabling the eventdev_pipeline sample app, but there
> is still some churn there to enable both HW/SW PMDs seamlessly. Once sample
> app is enabled a service core walk-through with that sample app can be added
> to the docs, to provide a tutorial on service-core usage.
>
>
> Harry van Haaren (7):
> service cores: header and implementation
> service cores: EAL init changes
> service cores: coremask parsing
> service cores: add unit tests
> event/sw: enable SW PMD with service capability
> doc: add service cores to doc and release notes
> maintainers: claim service cores
>
> MAINTAINERS | 7 +
> doc/api/doxy-api-index.md | 1 +
> doc/guides/eventdevs/sw.rst | 4 +-
> doc/guides/prog_guide/index.rst | 1 +
> doc/guides/prog_guide/service_cores.rst | 81 +++
> doc/guides/rel_notes/release_17_08.rst | 8 +
> drivers/event/sw/sw_evdev.c | 32 +
> drivers/event/sw/sw_evdev.h | 3 +
> lib/librte_eal/bsdapp/eal/Makefile | 1 +
> lib/librte_eal/bsdapp/eal/eal.c | 18 +
> lib/librte_eal/bsdapp/eal/rte_eal_version.map | 23 +
> lib/librte_eal/common/Makefile | 1 +
> lib/librte_eal/common/eal_common_lcore.c | 1 +
> lib/librte_eal/common/eal_common_options.c | 91 ++-
> lib/librte_eal/common/include/rte_eal.h | 4 +
> lib/librte_eal/common/include/rte_lcore.h | 3 +-
> lib/librte_eal/common/include/rte_service.h | 387 +++++++++++
> .../common/include/rte_service_component.h | 144 +++++
> lib/librte_eal/common/rte_service.c | 704
> +++++++++++++++++++++
> lib/librte_eal/linuxapp/eal/Makefile | 1 +
> lib/librte_eal/linuxapp/eal/eal.c | 18 +
> lib/librte_eal/linuxapp/eal/eal_thread.c | 9 +-
> lib/librte_eal/linuxapp/eal/rte_eal_version.map | 23 +
> test/test/Makefile | 2 +
> test/test/test_service_cores.c | 599 ++++++++++++++++++
> 25 files changed, 2162 insertions(+), 4 deletions(-)
> create mode 100644 doc/guides/prog_guide/service_cores.rst
> create mode 100644 lib/librte_eal/common/include/rte_service.h
> create mode 100644 lib/librte_eal/common/include/rte_service_component.h
> create mode 100644 lib/librte_eal/common/rte_service.c
> create mode 100644 test/test/test_service_cores.c
>
> --
> 2.7.4
>
