Found these two bugs while reading the bonding code. The first bug is mostly usability: if a user gives an incorrect port id as a slave, it will error-out later. The error will simply be less clear.
I'm partially responsible for the second one. I don't see why we did not wrote it properly at the time: the comment about lacking the proper rte_bus_pci impl was already not correct anymore when the check on the kdrv was removed. The necessary functions were already available I think. Gaetan Rivet (2): bonding: fix port id validity check on parsing bonding: fix PCI address comparison on non-pci ports drivers/net/bonding/rte_eth_bond_args.c | 63 +++++++++++-------------- 1 file changed, 27 insertions(+), 36 deletions(-) -- 2.26.0
