Re: [PATCH net v2 2/2] net: ethernet: oa_tc6: Improvement in buffer overflow handling

[Jakub Kicinski]

On Fri, 26 Jun 2026 08:35:19 -0700 Selvamani Rajagopal via B4 Relay
wrote:
> When oversubscribed traffic causes lot of buffer overflow errors,
> probably due to loss of data chunks, driver fails to find a
> data chunk with end_valid bit set, before it runs out of sk buffer
> space. As a result, assert is seen during skb_put.
> 
> Now check is made if tail + len > end, driver abandons the current
> data and starts look for a data chunk with start_valid bit,
> that is a new frame.

This sounds rather scary. The driver seems to have no length
information to confirm it got all the chunks. So if it missed
a middle chunk we will never know? At the very least this seems
like something we should increment rx_error for, not just rx_dropped?

Regarding the patch itself, I'm not clear on why we need to look
for new frame. Will we not notice the start bit immediately and
call oa_tc6_allocate_rx_skb() (if there is indeed a start bit in the
stream?)

So handling skb-already-exists in oa_tc6_allocate_rx_skb() seems
like enough to start a new frame.

Sashiko has another comment:

https://sashiko.dev/#/patchset/20260626-fix-race-condition-and-crash-v2-1-b6c5c10e604f@xxxxxxxxxx
-- 
pw-bot: cr