diff options
author | Edward Cree <ecree@solarflare.com> | 2017-01-14 00:20:14 +0300 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2017-01-16 22:00:46 +0300 |
commit | de1deff994286f464d5cad152d97830a5155f9ba (patch) | |
tree | e2977c4ce0b4ef87eaa226226b0c52c4ca9de2de /drivers/net/ethernet/sfc/nic.h | |
parent | cdfb1a9f30aaf43cfcecbb6c4061cd3807a4e086 (diff) | |
download | linux-de1deff994286f464d5cad152d97830a5155f9ba.tar.xz |
sfc: allow PIO more often
If an option descriptor has been sent on a queue but not followed by a
packet, there will have been no completion event, so the read and write
counts won't match and we'll think we can't do PIO. This combines with
the fact that we have two TX queues (for en/disable checksum offload),
and that both must be empty for PIO to happen.
This patch adds a separate "packet_write_count" that tracks the most
recent write_count we expect to see a completion event for; this excludes
option descriptors but _includes_ PIO descriptors (even though they look
like option descriptors). This is then used, rather than write_count,
in efx_nic_tx_is_empty().
We only bother to maintain packet_write_count on EF10, since on Siena
(a) there are no option descriptors and it always equals write_count, and
(b) there's no PIO, so we don't need it anyway.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'drivers/net/ethernet/sfc/nic.h')
-rw-r--r-- | drivers/net/ethernet/sfc/nic.h | 17 |
1 files changed, 14 insertions, 3 deletions
diff --git a/drivers/net/ethernet/sfc/nic.h b/drivers/net/ethernet/sfc/nic.h index 223774635cba..6a69aa3b0129 100644 --- a/drivers/net/ethernet/sfc/nic.h +++ b/drivers/net/ethernet/sfc/nic.h @@ -85,6 +85,17 @@ static inline bool __efx_nic_tx_is_empty(struct efx_tx_queue *tx_queue, return ((empty_read_count ^ write_count) & ~EFX_EMPTY_COUNT_VALID) == 0; } +/* Report whether the NIC considers this TX queue empty, using + * packet_write_count (the write count recorded for the last completable + * doorbell push). May return false negative. EF10 only, which is OK + * because only EF10 supports PIO. + */ +static inline bool efx_nic_tx_is_empty(struct efx_tx_queue *tx_queue) +{ + EFX_WARN_ON_ONCE_PARANOID(!tx_queue->efx->type->option_descriptors); + return __efx_nic_tx_is_empty(tx_queue, tx_queue->packet_write_count); +} + /* Decide whether we can use TX PIO, ie. write packet data directly into * a buffer on the device. This can reduce latency at the expense of * throughput, so we only do this if both hardware and software TX rings @@ -94,9 +105,9 @@ static inline bool __efx_nic_tx_is_empty(struct efx_tx_queue *tx_queue, static inline bool efx_nic_may_tx_pio(struct efx_tx_queue *tx_queue) { struct efx_tx_queue *partner = efx_tx_queue_partner(tx_queue); - return tx_queue->piobuf && - __efx_nic_tx_is_empty(tx_queue, tx_queue->insert_count) && - __efx_nic_tx_is_empty(partner, partner->insert_count); + + return tx_queue->piobuf && efx_nic_tx_is_empty(tx_queue) && + efx_nic_tx_is_empty(partner); } /* Decide whether to push a TX descriptor to the NIC vs merely writing |