On Friday 16 December 2016 06:40 PM, Shreyansh Jain wrote:
Register a PCI bus with Scan/match and probe callbacks. Necessary changes
in EAL layer for enabling bus interfaces. PCI devices and drivers now
reside within the Bus object.

Now that PCI bus handles the scan/probe methods, independent calls to
PCI scan and probe can be removed from the code.
PCI device and driver list are also removed.

rte_device and rte_driver list continue to exist. As does the VDEV lists.

Changes to test_pci:
- use a dummy test_pci_bus for all PCI test driver registrations
- this reduces the need for cleaning global list
- add necessary callbacks for invoking scan and probing/matching
  using EAL PCI scan code

Note: With this patch, all PCI PMDs would cease to work because of lack
      rte_driver->probe/remove implementations. Next patch would do that.

Signed-off-by: Shreyansh Jain <shreyansh.j...@nxp.com>
---
 app/test/test_pci.c                             | 154 +++++++++++------
 lib/librte_eal/bsdapp/eal/eal.c                 |   7 -
 lib/librte_eal/bsdapp/eal/eal_pci.c             |  52 +++---
 lib/librte_eal/bsdapp/eal/rte_eal_version.map   |   7 +-
 lib/librte_eal/common/eal_common_pci.c          | 212 ++++++++++++++----------
 lib/librte_eal/common/eal_private.h             |  14 +-
 lib/librte_eal/common/include/rte_pci.h         |  53 +++---
 lib/librte_eal/linuxapp/eal/eal.c               |   7 -
 lib/librte_eal/linuxapp/eal/eal_pci.c           |  57 ++++---
 lib/librte_eal/linuxapp/eal/rte_eal_version.map |   7 +-
 10 files changed, 330 insertions(+), 240 deletions(-)

<snip>

This is a relatively large patch. I would love to split it but
currently I am unable to find a nice/clean way. Any suggestions would
be really appreciated:

This currently does 3 broad things:
 - Move the PCI device/driver registration to PCI Bus
  -- So, introduce PCI bus
  -- change EAL to point to this new bus
 - remove/refactor existing EAL/*/*pci* code to accommodate this shift
 - Test PCI framework.

I could have split test_pci changes into a new patch, but that breaks
compilation after this patchset if test_pci compilation is enabled
(disabled by default). Is that acceptable? (if not, I am stuck with
keeping this set of changes into a single patch).

And, I would appreciate some help in confirming if the changes to
test_pci are OK or not from PCI testing perspective.

-
Shreyansh

Reply via email to