On Mon, Nov 9, 2026 at 9:45 AM Joe Damato [off-list ref] wrote:
On Sat, Nov 07, 2026 at 07:58:49PM +0000, netdev-bot+sashiko@kernel.org wrote:
quoted
By the end of the series, bnxt_hwrm_ring_free() counts stuck rings only
for TX, RX and AGG failures. If a CP or NQ RING_FREE fails, would
bnxt_stop_dma() be skipped, with __bnxt_close_nic() going on to free
that ring's memory in bnxt_free_mem()?
This also seems possible, but this is not what I observed in the field. On
three hosts, the NQ rings failed alongside the RX/TX ring free failures and so
if this series were applied DMA would be stopped.
I suppose it is possible that only the NQ or completion rings fail and this
series do not handle that case.
The NQ and CQ free operations differ slightly from the TX/RX rings.
The TX/RX ring free operation requires a feedback loop from the CQ
using interrupts, and it may fail after an IOMMU fault. Your patches
target this mode of ring free. Perhaps you can elaborate on this in
the commit message? Thanks.