- 17 Feb, 2010 40 commits
-
-
Joe Perches authored
These macros are too similar to the dev_<level> equivalents but take a usbnet * argument. Convert them to the recently introduced netdev_<level> macros and remove the old macros. The old macros had "\n" appended to the format string. Add the "\n" to the converted uses. Some existing uses of the dev<foo> macros in cdc_eem.c probably mistakenly had trailing "\n". No "\n" added there. Fix net1080 this/other log message inversion. Signed-off-by:
Joe Perches <joe@perches.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Sathya Perla authored
Signed-off-by:
Sathya Perla <sathyap@serverengines.com> Signed-off-by:
-
Sathya Perla authored
The flush compl (compl with numfrags == 0; no data) is rcvd from hw to indicate completion of RXQ destory operation. Fix the RX path to not process it as RX data. Signed-off-by:
-
Sathya Perla authored
When an MCC cmd is issued (via a netdev/ethtool op) while the device is not open, the MCC CQ gets processed but the EQ is not processed (as isr is not registered.) This can cause the EQ to become full. So, while the device is not open, CQ must not be re-armed to prevent EQ entries. Signed-off-by:
-
Sathya Perla authored
Introduce unlikely() for skb alloc failure and vlanf checks... Signed-off-by:
-
Ron Mercer authored
On some servers we see the cleaning of the RX queue finish before all the loopback packets are sent out. This delay allows the queues to settle before checking for successful completion. Also, delay completion so link has time to come back up. Signed-off-by:
Ron Mercer <ron.mercer@qlogic.com> Signed-off-by:
-
Ron Mercer authored
Use local copy of current mac address when initializing. In bonding testing we saw cases where dev_addr was out of data causing failover errors. Signed-off-by:
-
Ron Mercer authored
This was found with pings that were large enough to span multiple rx buffers. The split frame logic handles this case nicely. Signed-off-by:
-
Steven J. Magnani authored
A userland command to set the LLTEMAC MAC address, i.e. "ifconfig eth0 hw addr xx:yy:zz:pp:dd:qq", results in a device address of 00:01:xx:yy:zz:pp. Correct this. Signed-off-by:
Steven J. Magnani <steve@digidescorp.com> Signed-off-by:
-
Steven J. Magnani authored
LLTEMAC V1 cores place only received packet length in the app4 word. V2 cores place additional information in app4. Mask out the additional information when retrieving the packet length. Signed-off-by:
-
Jiri Pirko authored
Signed-off-by:
Jiri Pirko <jpirko@redhat.com> Signed-off-by:
-
Jiri Pirko authored
Introduced a new function to do the mc list processing, makes code clearer after transition. Signed-off-by:
-
Alexey Dobriyan authored
kasprintf() makes code smaller. Signed-off-by:
Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by:
-
Eric Dumazet authored
Commit 2eff25c1 (netfilter: xt_hashlimit: fix race condition and simplify locking) added a mutex deadlock : htable_create() is called with hashlimit_mutex already locked Signed-off-by:
Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by:
-
Pavel Emelyanov authored
The net being checked there is dev_net(dev) and thus this if is always false. Fits both net and net-next trees. Signed-off-by:
Pavel Emelyanov <xemul@openvz.org> Signed-off-by:
-
Nick Nunley authored
This updates the tx DMA mapping error handling code to resemble e1000e/ixgbe. Signed-off-by:
Nicholas Nunley <nicholasx.d.nunley@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by:
-
Nick Nunley authored
This change simplifies the code by setting RX_PTHRESH to 8 for all devices, as it was unlikely that there was any advantage to set it at 16 for earlier cards. Additionally TX_WTHRESH is set to 1 for the 82576 NIC to improve performance by enabling a minimal amount of write combining when writing back descriptors. Signed-off-by:
-
Nick Nunley authored
This change makes it so that igb_free_q_vectors is reused in igb_alloc_q_vectors to handle the cleanup instead of unwinding through the allocations. Signed-off-by:
-
Nick Nunley authored
In an earlier version of igb_write_mc_addr_list() the vmolr register was modified. This register is no longer accessed, although the variable still exists. This patch removes it from the function. Signed-off-by:
-
Nick Nunley authored
This change cleans up some instances where unsigned int and u32 were being used interchangeably, and cleans up hdr_len which was being set to 0 twice. Signed-off-by:
-
Nick Nunley authored
This change moves gso_segs into the buffer_info structure to avoid a possible cache line miss in clean_tx_irq. Signed-off-by:
-
Nick Nunley authored
igb_maybe_stop_tx() is extremely small and appears in several spots in the tx hotpath. This change inlines the function for a possible performance boost. Signed-off-by:
-
Nick Nunley authored
There are a couple statistics registers that are not meant to be read when in SGMII/serdes mode. This patch adds a check to verify mode before reading and updating these statistics. Signed-off-by:
-
Nick Nunley authored
The SRRCTL.DROP_EN bit should only be set when we are supporting multiple queues. This bit is meant to prevent head of line blocking and is unnecessary in the single queue case. Signed-off-by:
-
Nick Nunley authored
In order to maintain similar performance between MSI-X and legacy/MSI interrupts, this patch reduces the number of interrupts when receiving small packets to 20K when in interrupt throttle rate mode 3. Signed-off-by:
-
Nick Nunley authored
Adaptive IFS support has been included in the igb driver since its initial release, but it is not a feature on any igb NICs. This patch removes it from the driver. Signed-off-by:
-
Alexander Duyck authored
This change makes it so that the rings are allocated seperately. As a result we can allocate them on seperate nodes at some point in the future if we so desire. Signed-off-by:
Alexander Duyck <alexander.h.duyck@intel.com> Signed-off-by:
-
Nick Nunley authored
This patch adds calls to pci_save_state() immediately after calls to pci_restore_state(). Due to a change in the behavior of pci_restore_state() it is necessary to call pci_save_state() to keep the state_saved flag. This patch is based on a similar patch for ixgbe. Signed-off-by:
-
Nick Nunley authored
This changes the behavior of the driver to power down the link when the associated interface is down, unless management is enabled. Signed-off-by:
-
Nick Nunley authored
Checking the EEPROM APME state unnecessarily prevents the link from shutting down. The standard power down routines should be sufficient to determine whether the serdes link can power down when going into D3. Signed-off-by:
-
Nick Nunley authored
With this change ethtool will correctly report link status when the interface is down. Currently ethtool reports the link as not detected when the interface is down. Signed-off-by:
-
Nick Nunley authored
This adds support for wake-on-link/phy activity to the ethtool interface. Signed-off-by:
-
Alexander Duyck authored
This change removes the use of itr_shift since a mac type call can be used just as easily to identify the only HW that needs to have the itr shifted. In addition it removes two unecessary declarations of a q_vector pointer from the initialization path. Signed-off-by:
-
Joe Perches authored
On Tue, 2010-02-16 at 16:47 +0100, Patrick McHardy wrote: > Joe Perches wrote: > >> @@ -246,6 +246,8 @@ extern int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb); > >> int ip6_frag_nqueues(struct net *net); > >> int ip6_frag_mem(struct net *net); > >> > >> +#define IPV6_FRAG_HIGH_THRESH 262144 /* == 256*1024 */ > >> +#define IPV6_FRAG_LOW_THRESH 196608 /* == 192*1024 */ > >> #define IPV6_FRAG_TIMEOUT (60*HZ) /* 60 seconds */ > > > > 196608 isn't a number I want to remember. > > Is this better as: > > > > #define IPV6_FRAG_HIGH_THRESH (256 * 1024) /* 262144 */ > > #define IPV6_FRAG_LOW_THRESH (192 * 1024) /* 196608 */ > > Please send a patch, I'll apply it once these patches are in Dave's > tree. Signed-off-by:
-
Alexey Dobriyan authored
call_rcu() will unconditionally reinitialize RCU head anyway. Signed-off-by:
Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by:
-
Jochen Friedrich authored
Signed-off-by:
Jochen Friedrich <jochen@scram.de> Signed-off-by:
-
Tejun Heo authored
Add __percpu sparse annotations to net drivers. These annotations are to make sparse consider percpu variables to be in a different address space and warn if accessed without going through percpu accessors. This patch doesn't affect normal builds. Signed-off-by:
Tejun Heo <tj@kernel.org> Acked-by:
-
Tejun Heo authored
Add __percpu sparse annotations to net. These annotations are to make sparse consider percpu variables to be in a different address space and warn if accessed without going through percpu accessors. This patch doesn't affect normal builds. The macro and type tricks around snmp stats make things a bit interesting. DEFINE/DECLARE_SNMP_STAT() macros mark the target field as __percpu and SNMP_UPD_PO_STATS() macro is updated accordingly. All snmp_mib_*() users which used to cast the argument to (void **) are updated to cast it to (void __percpu **). Signed-off-by:
-
David S. Miller authored
-
jamal authored
Eric's version fixed it for pfkey. This one is for xfrm user. I thought about amortizing those two get_acqseq()s but it seems reasonable to have two of these sequence spaces for the two different interfaces. cheers, jamal commit d5168d5addbc999c94aacda8f28a4a173756a72b Author: Jamal Hadi Salim <hadi@cyberus.ca> Date: Tue Feb 16 06:51:22 2010 -0500 xfrm: avoid spinlock in get_acqseq() used by xfrm user This is in the same spirit as commit 28aecb9d by Eric Dumazet. Use atomic_inc_return() in get_acqseq() to avoid taking a spinlock Signed-off-by:Jamal Hadi Salim <hadi@cyberus.ca> Acked-by:
-
