- 24 Aug, 2009 4 commits
-
-
Amit Kumar Salecha authored
Implement pci driver shutdown functionality, this helps quiescing all PCI transaction before chipset is reset. Amit Kumar Salecha <amit@qlogic.com> Signed-off-by:
Dhananjay Phadke <dhananjay@netxen.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Dhananjay Phadke authored
Move all tx skb mapping code into netxen_map_tx_skb(). Signed-off-by:
-
Dhananjay Phadke authored
If netxen_need_fw_reset() return 0 [ implies firmware is up and running], still go through dma mask check, etc. Signed-off-by:
-
Narender Kumar authored
Add ethtool -K knob to control LRO in firmware. LRO path is completely separated from GRO, LRO packets are still fed with netif_receive_skb(). Also fix ethtool statistics to include LRO packets. Also use correct message type while configuring interrupt coalescing. Signed-off-by:
Narender Kumar <narender.kumar@qlogic.com> Signed-off-by:
-
- 23 Aug, 2009 8 commits
-
-
Michael Chan authored
Signed-off-by:
Michael Chan <mchan@broadcom.com> Signed-off-by:
-
Michael Chan authored
The structure, once initialized, never changes. Signed-off-by:
Benjamin Li <benli@broadcom.com> Signed-off-by:
-
Michael Chan authored
- Set the USE_INT_PARAM bit so the rx-frames-irq and tx-frames-irq will take effect on 5709. - Increase the default rx-frames to reduce interrupt count. - Decrease the default rx-frames-irq and tx-frames-irq to catch more events during NAPI poll. All these will reduce interrupts without affecting latency. Signed-off-by:
-
Michael Chan authored
Report this counter to ethtool -S and include it in netstat. Signed-off-by:
-
Michael Chan authored
Add flag to expand the workaround to both chips. Signed-off-by:
-
Michael Chan authored
When unable to allocate memory for new MTU or new ring size, we need to close the device to prevent it from crashing. Signed-off-by:
-
Michael Chan authored
Add this check to bnx2_netif_stop() and bnx2_vlan_rx_register() to prevent bus lockups on some systems when the chip is in low power state. Signed-off-by:
-
Michael Chan authored
In case IRQs are shared, we will not mistakenly start processing the ring based on old status block indices. Signed-off-by:
-
- 21 Aug, 2009 1 commit
-
-
- 20 Aug, 2009 27 commits
-
-
Lennert Buytenhek authored
Signed-off-by:
Lennert Buytenhek <buytenh@marvell.com> Signed-off-by:
John W. Linville <linville@tuxdriver.com>
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
The current mwl8k_priv->fw_lock spinlock doesn't actually protect against multiple commands being submitted at once, as it is not kept held over the entire firmware command submission. And since waiting for command completion sleeps, we can't use a spinlock anyway. To fix mwl8k firmware command serialisation properly, we have the following requirements: - Some commands require that the packet transmit path is idle when the command is issued. (For simplicity, we'll just quiesce the transmit path for every command.) - There are certain sequences of commands that need to be issued to the hardware sequentially, with no other intervening commands. This leads to an implementation of a "firmware lock" as a mutex that can be taken recursively, and which is taken by both the low-level command submission function (mwl8k_post_cmd) as well as any users of that function that require issuing of an atomic sequence of commands, and quiesces the transmit path whenever it's taken. Signed-off-by:
-
Lennert Buytenhek authored
Delete most of the mwl8k_work_struct fields and options, since most of them are unused or never changed from their defaults: - We always use priv->config_wq, so delete the wqueue argument from mwl8k_queue_work(). - MWL8K_WQ_SPIN and MWL8K_WQ_POST_REQUEST are never used, as all callers sleep for request completion, so sleep unconditionally. - MWL8K_WQ_FREE_WORKSTRUCT is never used. - MWL8K_WQ_TX_WAIT_EMPTY is always set, so assume it unconditionally. - timeout_ms/txwait_attempts/tx_timeout_ms are never changed from their defaults, so just hardcode these in the workqueue worker. - step is never used. Signed-off-by:
-
Lennert Buytenhek authored
Various mwl8k_txq_xmit changes: - Extract the QoS field before adding the DMA header. - Only write to tx->status once, and only after all the other descriptor fields have been set. - Do all tx state manipulation under the tx spinlock. - Remove the priv->inconfig check, as all transmit queues will be frozen during config cycles, so we won't ever be asked to transmit if a config cycle is running. - Remove some more dead code. Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Use MWL8K_TX_QUEUES instead of a hardcoded "4" in a couple of places. Signed-off-by:
-
Lennert Buytenhek authored
Since firmware load commands don't generate 'command done' interrupts like normal commands do, polling for command done interrupts just unnecessarily slows down the firmware load process. Removing this bit of code speeds up loading a typical firmware image from 840 msec to 180 msec. Signed-off-by:
-
Lennert Buytenhek authored
Adding and removing the DMA header that the mwl8k hardware requires on tx and provides on rx can never fail, since we are guaranteed to have enough headroom on the tx path to expand the packet, and we only ever shrink the packet on the rx path. (And on both paths we are guaranteed to be the only user of the skb we are handling.) This allows removing all of the skb clone handling in the tx and tx reclaim paths, and eliminates error checks in both the tx and rx paths, simplifying the code a bit more. Signed-off-by:
-
Lennert Buytenhek authored
There is no need to check for NL80211_IFTYPE_MONITOR in ->add_interface(), as this function is never called for monitor interfaces. Also, there is no need to advertise this bit in our wiphy's ->interface_modes. Signed-off-by:
-
Lennert Buytenhek authored
mwl8k_configure_filter() passes pointers to total_flags and the multicast address list to a workqueue function, while there is no guarantee that those pointers will still be valid by the time the workqueue function runs. Solve this by passing total_flags by value, and by passing an already built multicast address setup command packet to the workqueue function so that we don't have to look at the multicast address list itself outside of mwl8k_configure_filter(). Also, since ->configure_filter() can sleep now, wait synchronously for the worker to finish. Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Instead of passing a flag bitmask to mwl8k_cmd_802_11_radio_control, pass the 'enable' and 'force' arguments as separate parameters, and introduce wrappers for the common cases of enabling and disabling without forcing. Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
Lennert Buytenhek authored
Signed-off-by:
-
