mbox series

[RFC,00/11] Bus cleanup for 22.11

Message ID 20220628144643.1213026-1-david.marchand@redhat.com (mailing list archive)
Headers show
Series Bus cleanup for 22.11 | expand

Message

David Marchand June 28, 2022, 2:46 p.m. UTC
This is a PoC for hiding the rte_bus object and mark associated API
as internal.

A good amount of the patches are preparation work on rte_bus.h,
rte_dev.h, rte_devargs.h and rte_eal.h headers, removing dependencies
between them. This is something I had in store for some time, maybe I
should have dropped it from the PoC, but I think those cleanups are
worth it in any case.

The last two patches do the actual job: add accessors and make the
rte_bus object opaque to non internal users.


Disclaimer: this series is a bit rushed (I brute forced compilation
tests in GHA so that it passes between patches, but there still may be
something broken...).
Not surprisingly, the ABI check in the CI is expected to fail.

Comments welcome.

I also hope we can do the same work on other generic objects
(rte_driver, rte_device), but this is another story.

Comments

David Marchand Sept. 5, 2022, 8:39 a.m. UTC | #1
The rte_bus, rte_driver and rte_device objects are part of the public
ABI and this has proved to be a problem when needing to extend them.
This series hides them, and mark associated driver only API as internal.

A good amount of the patches are preparation work on rte_bus.h,
rte_dev.h, rte_devargs.h and rte_eal.h headers, removing dependencies
between them.

PCI bus specific handling are removed from testpmd, unit tests and
examples.

After this series, driver-only API headers for registering to buses are
not exported anymore, unless the enable_driver_sdk meson option is
selected.

New accessors for rte_bus, rte_driver and rte_device have been added,
marked with an experimental tag first when introducing them, and later
in the series marked as stable since external users will want to use
those drop-in replacements right away.

A check is added to ensure we won't pollute app/ and examples/ again,
though some unit tests are left intentionnally untouched as they test
some internals of DPDK.

Changes since v4:
- (hopefully, because I can't test) fixed Windows build,
- dropped unneeded change in vdev unit test,

Changes since RFC v3:
- marked the series as non-RFC,
- rebased on v22.11-rc0,
- marked RTE_FUNC_PTR_* macros as deprecated,
- split the bus specific patch in two patches,

Changes since RFC v2:
- added check for additions of include .*_(driver|pmd)\.h in apps and
  examples,
- dropped legacy/debug testpmd commands to read PCI BAR0 registers,
- dropped patches on bbdev, ethdev, rawdev driver headers for now,
- reordered patches and separated changes per bus type to ease review,
- added more accessor for device,
- introduced rte_dev_bus_info to provide a Bus specific description of
  a device, a first use is for providing a PCI device vendor / device
  identifiers that are otherwise unavailable through a generic existing
  API,

Changes since RFC v1:
- added two more cleanups (new patch 3 and 4) for unit test and examples
  relying on PCI specific info,
- went on with masking rte_driver and rte_device too,