On 27/06/2024 20:55, Jakub Kicinski wrote:
On Thu, 27 Jun 2024 11:50:40 +0200 Alexander Lobakin wrote:
quoted
quoted
I don't think we should group them indiscriminately. Better to add the
asserts flag by flag. Neither of the flags you're breaking out in this
patch are used on the fast path.
Or is the problem that CACHELINE_ASSERT_GROUP_MEMBER doesn't work on
bitfields?
It generates sizeof(bitfield) which the compilers don't like and don't
want to compile ._.
Mm. Okay, I have no better ideas then.
Do consider moving the cold flags next to wol_enabled, tho?
My RSS series moves wol_enabled out to struct ethtool_netdev_state [1] so
this may not be worthwhile?
-ed
[1]: https://lore.kernel.org/netdev/293a562278371de7534ed1eb17531838ca090633.1719502239.git.ecree.xilinx@gmail.com/ (local)