On Sat June 12 2010 01:16:19 you wrote:
> What's the purpose of ath5k_rx_stop setting sc->rxlink to NULL?
good question. it has the comment /* just in case */ and this is in the HAL
too, so i guess it was just copied here "just in case" and because nobody is
sure we dont need it.
> As Bob point
What's the purpose of ath5k_rx_stop setting sc->rxlink to NULL?
As Bob points out, it causes problems if it's not done while holding the
rx buffer spinlock.
There's code in ath5k_intr that sets sc->rxlink to NULL, apparently
to handle a chip bug. It needs to hold the spinlock as well to avoid
sel
On Fri, Jun 11, 2010 at 5:50 AM, Bruno Randolf wrote:
> Add an extra check to be sure we never process the self-linked rx descriptor
> at the end of the list. This should not happen since in this case the RXDP
Here's a way this might happen frequently BTW:
cpu 0:
Add an extra check to be sure we never process the self-linked rx descriptor at
the end of the list. This should not happen since in this case the RXDP should
also point at the same descriptor and the previous check should catch that. But
we don't trust RXDP completely (might it be NULL or undefine