Hi Stephen,

Agreed. Since it gates nothing, I will delete it rather than document a
no-op constant.

On Mon, Sep 21, 2026 at 9:54 AM Stephen Hemminger <
[email protected]> wrote:

> On Sun, 20 Sep 2026 20:59:03 -0600
> Mohammad Shuab Siddique <[email protected]> wrote:
>
> > From: Mohammad Shuab Siddique <[email protected]>
> >
> > The driver only supported queue sizes up to 4096 for Tx and 8192 for
> > Rx. Raise both to 16384. The completion ring for a Rx ring is sized
> > at 2x the Rx ring size, further multiplied by 4 when the aggregation
> > ring is in use (8x total), so at 16384 it can reach 131072, above
> > uint16_t range - widen the ring index/counter variables touched by
> > that path to uint32_t.
> >
> > Signed-off-by: Keegan Freyhof <[email protected]>
> > Signed-off-by: Mohammad Shuab Siddique <
> [email protected]>
> > ---
> [PATCH v2] net/bnxt: add support for queue size of 16384
>
> Warning: the change of BNXT_MAX_MTU and BNXT_NUM_VLANS to UL is
> unrelated to queue size and is not described. It changes the type of
> BNXT_MAX_PKT_LEN to unsigned long. Drop it, or split it out with a
> reason.
>
> Warning: MAX_CP_DESC_CNT is bumped to 32K and gains a comment saying
> it is unused and the real worst case is 131072. If it is unused,
> delete it.
>


-- 
Regards,
Shuab

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to