On 11/16/26 5:42 PM, Jakub Kicinski wrote:
On Mon,  9 Nov 2026 15:55:41 -0800 Mingming Cao wrote:
  arch/powerpc/include/asm/hvcall.h   |    6 +-
  drivers/net/ethernet/ibm/ibmveth.c  | 5007 ++++++++++++++++++++++-----
  drivers/net/ethernet/ibm/ibmveth.h  |  240 +-
  tools/perf/python/powerpc-hcalls.py |    4 +
  4 files changed, 4318 insertions(+), 939 deletions(-)
This is way too large, I spent half an hour reviewing and only gotten
to patch 8. There's no reason for the series to be this big. The 15
patch limit doesn't mean you should try to cram as much as you can into
15 patches. Please split this up into (3?) smaller series.
Thank you for the time, and for being direct. You're right, and
point taken on the commit messages too.

The plan:

  - net-next: two fixes for bugs that exist today, posted on their own
  - series 1: prepare the RX path, 8 patches (928+/431-, net +497).
    IBMVETH_MAX_RX_QUEUES stays 1; same hypercalls, no new ABI.
    Patches 1-5 move open/close onto helpers, 6 register/teardown,
    7 a queue index through replenish and poll, 8 splits poll
     into helpers. Your 07/08/09 comments are all addressed there

- series 2: the multi-queue work itself: the new hypercalls,
    registering the extra RX queues and their interrupts, per-queue
    statistics, turning multi-queue on, and ethtool -L for the RX
    count. Posted once series 1 is applied

I also narrowed the scope to what multi-queue needs to go in. v8
resized the RX queues live, without stopping traffic. Changing the
queue count is a rare admin operation, so for now ethtool -L closes
and reopens the device, as an MTU change does. That could remove
lots of complexity code. (Live resize can come back later if users need it.)

On the three series you suggested: with close/reopen the resize could be a
single small patch, so I'd like to end series 2 with it rather than
post a one-patch third series. If you'd rather have it separate, I'll
split it out.

I just mailed out the - net-next: two fixes and series 1: prepare the RX path.

Thanks,

Mingming



Reply via email to