The TX DMA engine issues upstream read requests when there is room in
the TX FIFO for the completion. However, the fetches for the rest of
the packet might be delayed by any back pressure. Since a flush must
wait for an EOP, the entire flush may be delayed by back pressure.
Mitigate this by disabling flow control before the flushes are
started. Since PF and VF flushes run in parallel introduce
fc_disable, a reference count of the number of flushes outstanding.
The same principle could be applied to Falcon, but that
would bring with it its own testing.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
}
if (efx->wanted_fc & EFX_FC_AUTO)
fcntl = MC_CMD_FCNTL_AUTO;
+ if (efx->fc_disable)
+ fcntl = MC_CMD_FCNTL_OFF;
MCDI_SET_DWORD(cmdbytes, SET_MAC_IN_FCNTL, fcntl);
* @promiscuous: Promiscuous flag. Protected by netif_tx_lock.
* @multicast_hash: Multicast hash table
* @wanted_fc: Wanted flow control flags
+ * @fc_disable: When non-zero flow control is disabled. Typically used to
+ * ensure that network back pressure doesn't delay dma queue flushes.
+ * Serialised by the rtnl lock.
* @mac_work: Work item for changing MAC promiscuity and multicast hash
* @loopback_mode: Loopback status
* @loopback_modes: Supported loopback mode bitmask
bool promiscuous;
union efx_multicast_hash multicast_hash;
u8 wanted_fc;
+ unsigned fc_disable;
atomic_t rx_reset;
enum efx_loopback_mode loopback_mode;
struct efx_tx_queue *tx_queue;
int rc = 0;
+ efx->fc_disable++;
efx->type->prepare_flush(efx);
efx_for_each_channel(channel, efx) {
atomic_set(&efx->rxq_flush_outstanding, 0);
}
+ efx->fc_disable--;
+
return rc;
}