[PATCH net v2 0/4] net: axienet: fix TX queue handling around a DMA error reset

[Sagi Maimon]

I am withdrawing this series.

The review of v2 found a real gap in 3/4: axienet_open() and
axienet_resume() can still wake the queue while the error handler is
running, so the claim that the handler owns the queue state for its
whole duration does not hold.  4/4 has a smaller hole in the same resume
window.  The other findings are older problems on the same paths:
axienet_open() ignores the result of axienet_device_reset(),
axienet_resume() ignores the result of axienet_open(), and the dmaengine
stop path has the same suspend race as the legacy one.

Fixing that properly means reworking open, suspend and resume together,
and I cannot test suspend and resume on the hardware I have.  I would
rather not keep posting partial fixes for paths I cannot test, so I am
dropping the series.  The TX buffer leak in axienet_dma_bd_release() is
posted as a separate patch.

pw-bot: cr