- 14 Nov, 2011 15 commits
-
-
Eric Dumazet authored
One of the thing we discussed during netdev 2011 conference was the idea to change some network drivers to allocate/populate their skb at RX completion time, right before feeding the skb to network stack. In old days, we allocated skbs when populating the RX ring. This means bringing into cpu cache sk_buff and skb_shared_info cache lines (since we clear/initialize them), then 'queue' skb->data to NIC. By the time NIC fills a frame in skb->data buffer and host can process it, cpu probably threw away the cache lines from its caches, because lot of things happened between the allocation and final use. So the deal would be to allocate only the data buffer for the NIC to populate its RX ring buffer. And use build_skb() at RX completion to attach a data buffer (now filled with an ethernet frame) to a new skb, initialize the skb_shared_info portion, and give the hot skb to network stack. build_skb() is the function to allocate an skb, caller providing the data buffer that should be attached to it. Drivers are expected to call skb_reserve() right after build_skb() to adjust skb->data to the Ethernet frame (usually skipping NET_SKB_PAD and NET_IP_ALIGN, but some drivers might add a hardware provided alignment) Data provided to build_skb() MUST have been allocated by a prior kmalloc() call, with enough room to add SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) bytes at the end of the data without corrupting incoming frame. data = kmalloc(NET_SKB_PAD + NET_IP_ALIGN + 1536 + SKB_DATA_ALIGN(sizeof(struct skb_shared_info)), GFP_ATOMIC); ... skb = build_skb(data); if (!skb) { recycle_data(data); } else { skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN); ... } Signed-off-by:Eric Dumazet <eric.dumazet@gmail.com> CC: Eilon Greenstein <eilong@broadcom.com> CC: Ben Hutchings <bhutchings@solarflare.com> CC: Tom Herbert <therbert@google.com> CC: Jamal Hadi Salim <hadi@mojatatu.com> CC: Stephen Hemminger <shemminger@vyatta.com> CC: Thomas Graf <tgraf@infradead.org> CC: Herbert Xu <herbert@gondor.apana.org.au> CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Baruch Siach authored
Since 952c5ca1 (fsl_pq_mdio: Clean up tbi address configuration) .probe returns -EBUSY when the "tbi-phy" node is missing. Fix this. Cc: Andy Fleming <afleming@freescale.com> Signed-off-by:
Baruch Siach <baruch@tkos.co.il> Signed-off-by:
-
Marc Kleine-Budde authored
This patch adds listen only mode to the mscan controller. Signed-off-by:
Marc Kleine-Budde <mkl@pengutronix.de> Acked-by:
Wolfgang Grandegger <wg@grandegger.com> Signed-off-by:
-
Eric Dumazet authored
Le mercredi 09 novembre 2011 à 16:21 -0500, David Miller a écrit : > From: David Miller <davem@davemloft.net> > Date: Wed, 09 Nov 2011 16:16:44 -0500 (EST) > > > From: Eric Dumazet <eric.dumazet@gmail.com> > > Date: Wed, 09 Nov 2011 12:14:09 +0100 > > > >> unres_qlen is the number of frames we are able to queue per unresolved > >> neighbour. Its default value (3) was never changed and is responsible > >> for strange drops, especially if IP fragments are used, or multiple > >> sessions start in parallel. Even a single tcp flow can hit this limit. > > ... > > > > Ok, I've applied this, let's see what happens :-) > > Early answer, build fails. > > Please test build this patch with DECNET enabled and resubmit. The > decnet neigh layer still refers to the removed ->queue_len member. > > Thanks. Ouch, this was fixed on one machine yesterday, but not the other one I used this morning, sorry. [PATCH V5 net-next] neigh: new unresolved queue limits unres_qlen is the number of frames we are able to queue per unresolved neighbour. Its default value (3) was never changed and is responsible for strange drops, especially if IP fragments are used, or multiple sessions start in parallel. Even a single tcp flow can hit this limit. $ arp -d 192.168.20.108 ; ping -c 2 -s 8000 192.168.20.108 PING 192.168.20.108 (192.168.20.108) 8000(8028) bytes of data. 8008 bytes from 192.168.20.108: icmp_seq=2 ttl=64 time=0.322 ms Signed-off-by: -
stephen hemminger authored
More changes to the recent code to support control of forwarding database via netlink. * Support NTF_USE like neighbour table * Validate state bits from application * Only send notifications (and change bits) if new entry is different. Signed-off-by:Stephen Hemminger <shemminger@vyatta.com> Signed-off-by:
-
Rick Jones authored
Perform another round of floor sweeping, converting the .get_drvinfo routines of additional drivers from strcpy to strlcpy along with some conversion of sprintf to snprintf. Signed-off-by:
Rick Jones <rick.jones2@hp.com> Signed-off-by:
-
Andy Fleming authored
The code for setting the address of the internal TBI PHY was convoluted enough without a maze of ifdefs. Clean it up a bit so we allow the logic to fail down to -ENODEV at the end of the if/else ladder, rather than using ifdefs to repeat the same failure code over and over. Also, remove the support for the auto-configuration. I'm not aware of anyone using it, and it ends up using the bus mutex before it's been initialized. Signed-off-by:
Andy Fleming <afleming@freescale.com> Signed-off-by:
-
Sanjay Hortikar authored
Support enabling/disabling/querying internal loopback mode for forcedeth NICs using ethtool. Signed-off-by:
Sanjay Hortikar <horti@google.com> Signed-off-by:
Mahesh Bandewar <maheshb@google.com> Signed-off-by:
-
alex.bluesman.smirnov@gmail.com authored
This patch adds chapter to documentation which describes how to use 6lowpan technology. Signed-off-by:
Alexander Smirnov <alex.bluesman.smirnov@gmail.com> Signed-off-by:
-
alex.bluesman.smirnov@gmail.com authored
This patch provides possibility to decompress UDP headers. Derived from Contiki OS. Signed-off-by:
-
alex.bluesman.smirnov@gmail.com authored
This patch adds support for UDP header compression. Derived from Contiki OS. Signed-off-by:
-
alex.bluesman.smirnov@gmail.com authored
This patch fixes settings for device initialization which makes possible to use NDISC and TCP. Signed-off-by:
Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Signed-off-by:
-
alex.bluesman.smirnov@gmail.com authored
This patch disables debug output enabled by default. Signed-off-by:
-
alex.bluesman.smirnov@gmail.com authored
This patch adds support for frame fragmentation. Signed-off-by:
-
Eric Dumazet authored
Reading /proc/net/snmp6 on a machine with a lot of cpus is very expensive (can be ~88000 us). This is because ICMPV6MSG MIB uses 4096 bytes per cpu, and folding values for all possible cpus can read 16 Mbytes of memory (32MBytes on non x86 arches) ICMP messages are not considered as fast path on a typical server, and eventually few cpus handle them anyway. We can afford an atomic operation instead of using percpu data. This saves 4096 bytes per cpu and per network namespace. Signed-off-by:
-
- 13 Nov, 2011 13 commits
-
-
Jiri Pirko authored
This patch introduces new network device called team. It supposes to be very fast, simple, userspace-driven alternative to existing bonding driver. Userspace library called libteam with couple of demo apps is available here: https://github.com/jpirko/libteam Note it's still in its dipers atm. team<->libteam use generic netlink for communication. That and rtnl suppose to be the only way to configure team device, no sysfs etc. Python binding of libteam was recently introduced. Daemon providing arpmon/miimon active-backup functionality will be introduced shortly. All what's necessary is already implemented in kernel team driver. v7->v8: - check ndo_ndo_vlan_rx_[add/kill]_vid functions before calling them. - use dev_kfree_skb_any() instead of dev_kfree_skb() v6->v7: - transmit and receive functions are not checked in hot paths. That also resolves memory leak on transmit when no port is present v5->v6: - changed couple of _rcu calls to non _rcu ones in non-readers v4->v5: - team_change_mtu() uses team->lock while travesing though port list - mac address changes are moved completely to jurisdiction of userspace daemon. This way the daemon can do FOM1, FOM2 and possibly other weird things with mac addresses. Only round-robin mode sets up all ports to bond's address then enslaved. - Extended Kconfig text v3->v4: - remove redundant synchronize_rcu from __team_change_mode() - revert "set and clear of mode_ops happens per pointer, not per byte" - extend comment of function __team_change_mode() v2->v3: - team_change_mtu() uses rcu version of list traversal to unwind - set and clear of mode_ops happens per pointer, not per byte - port hashlist changed to be embedded into team structure - error branch in team_port_enter() does cleanup now - fixed rtln->rtnl v1->v2: - modes are made as modules. Makes team more modular and extendable. - several commenters' nitpicks found on v1 were fixed - several other bugs were fixed. - note I ignored Eric's comment about roundrobin port selector as Eric's way may be easily implemented as another mode (mode "random") in future. Signed-off-by:
Jiri Pirko <jpirko@redhat.com> Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
Dmitry Kravkov <dmitry@broadcom.com> Signed-off-by:
Eilon Greenstein <eilong@broadcom.com> Signed-off-by:
-
Ariel Elior authored
Signed-off-by:
Ariel Elior <ariele@broadcom.com> Signed-off-by:
-
Dmitry Kravkov authored
The race may cause access of registers while MAC hw block is in reset state. As a result syslog will show error messages. We can prevent this by using state from local variable. Signed-off-by:
-
Ariel Elior authored
Shut down the device in case of fan failure to prevent HW damage. Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
-
Dmitry Kravkov authored
We need propagate the DCBX results from PMF to other functions on the same port, in order to properly update netdev structure and allow following new ETS and PFC configurations. Signed-off-by:
-
Dmitry Kravkov authored
FCoE license info must be initialized at probe(), but iSCSI at open(). Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
-
Dmitry Kravkov authored
Signed-off-by:
-
- 12 Nov, 2011 4 commits
-
-
Sathya Perla authored
A FW cmd timeout (with a sufficiently large timeout value in the order of tens of seconds) indicates an unresponsive FW. In this state issuing further cmds and waiting for a completion will only stall the process. Signed-off-by:
Sathya Perla <sathya.perla@emulex.com> Signed-off-by:
-
Sathya Perla authored
Currently we're spamming error messages each time a FW cmd call is made while in EEH/UE error state. One log msg on error detection is enough. Signed-off-by:
-
Sathya Perla authored
Signed-off-by:
-
Sathya Perla authored
Initialize if_handle, vf_if_handle and vf_pmac_id with "-1" so that in failure cases when be_clear() is called, we can skip over if_destroy/pmac_del cmds if they have not been created. Signed-off-by:
-
- 09 Nov, 2011 2 commits
-
-
Eric Dumazet authored
Le lundi 07 novembre 2011 à 15:33 +0100, Eric Dumazet a écrit : > At least, in recent kernels we dont change dst->refcnt in forwarding > patch (usinf NOREF skb->dst) > > One particular point is the atomic_inc(dst->refcnt) we have to perform > when queuing an UDP packet if socket asked PKTINFO stuff (for example a > typical DNS server has to setup this option) > > I have one patch somewhere that stores the information in skb->cb[] and > avoid the atomic_{inc|dec}(dst->refcnt). > OK I found it, I did some extra tests and believe its ready. [PATCH net-next] ipv4: IP_PKTINFO doesnt need dst reference When a socket uses IP_PKTINFO notifications, we currently force a dst reference for each received skb. Reader has to access dst to get needed information (rt_iif & rt_spec_dst) and must release dst reference. We also forced a dst reference if skb was put in socket backlog, even without IP_PKTINFO handling. This happens under stress/load. We can instead store the needed information in skb->cb[], so that only softirq handler really access dst, improving cache hit ratios. This removes two atomic operations per packet, and false sharing as well. On a benchmark using a mono threaded receiver (doing only recvmsg() calls), I can reach 720.000 pps instead of 570.000 pps. IP_PKTINFO is typically used by DNS servers, and any multihomed aware UDP application. Signed-off-by: -
Eric Dumazet authored
Reading /proc/net/snmp on a machine with a lot of cpus is very expensive (can be ~88000 us). This is because ICMPMSG MIB uses 4096 bytes per cpu, and folding values for all possible cpus can read 16 Mbytes of memory. ICMP messages are not considered as fast path on a typical server, and eventually few cpus handle them anyway. We can afford an atomic operation instead of using percpu data. This saves 4096 bytes per cpu and per network namespace. Signed-off-by:
-
- 08 Nov, 2011 6 commits
-
-
Eric Dumazet authored
Make clear that sk_clone() and inet_csk_clone() return a locked socket. Add _lock() prefix and kerneldoc. Suggested-by:
Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by:
-
Eric Dumazet authored
Remove the assumption that skb_get_rxhash() makes IP header and ports linear, and use skb_header_pointer() instead in choke_match_flow() This permits __skb_get_rxhash() to use skb_header_pointer() eventually. Signed-off-by:
-
Ricardo Ribalda authored
This patch enables the ioctl support for the driver. So userspace programs like mii-tool can work. Resend in merge window Signed-off-by:
Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com> Signed-off-by:
-
Maciej Żenczykowski authored
This just makes it possible to spoof source IPv6 address on a socket without having to create and bind a new socket for every source IP we wish to spoof. Signed-off-by:
Maciej Żenczykowski <maze@google.com> Signed-off-by:
-
Maciej Żenczykowski authored
This makes native ipv6 bind follow the precedent set by: - native ipv4 bind behaviour - dual stack ipv4-mapped ipv6 bind behaviour. This does allow an unpriviledged process to spoof its source IPv6 address, just like it currently can spoof its source IPv4 address (for example when using UDP). Signed-off-by:
-
Rick Jones authored
Per the mention made by Ben Hutchings that strlcpy is now the preferred string copy routine for a .get_drvinfo routine, do a bit of floor sweeping and convert some of the as-yet unconverted ethernet drivers to it. Signed-off-by:
-
