- 19 5月, 2009 8 次提交
-
-
由 Johann Baudy 提交于
New packet socket feature that makes packet socket more efficient for transmission. - It reduces number of system call through a PACKET_TX_RING mechanism, based on PACKET_RX_RING (Circular buffer allocated in kernel space which is mmapped from user space). - It minimizes CPU copy using fragmented SKB (almost zero copy). Signed-off-by: NJohann Baudy <johann.baudy@gnu-log.net> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
gen_estimator can overflow bps (bytes per second) with Gb links, while it was designed with a u32 API, with a theorical limit of 34360Mbit (2^32 bytes) Using 64 bit intermediate avbps/brate counters can allow us to reach this theorical limit. Signed-off-by: NEric Dumazet <dada1@cosmosbay.com> Signed-off-by: NJarek Poplawski <jarkao2@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Rami Rosen 提交于
There is no need for net/icmp.h header in net/ipv4/fib_frontend.c. This patch removes the #include net/icmp.h from it. Signed-off-by: NRami Rosen <ramirose@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Rami Rosen 提交于
Signed-off-by: NRami Rosen <ramirose@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
We can update netdev_queue tx_bytes/tx_packets/tx_dropped counters instead of dev->stats ones, to reduce number of cache lines dirtied in xmit path. This fixes a performance problem on SMP when many different cpus take vlan tx path. Signed-off-by: NEric Dumazet <dada1@cosmosbay.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
offsetof(struct net_device, features)=0x44 offsetof(struct net_device, stats.tx_packets)=0x54 offsetof(struct net_device, stats.tx_bytes)=0x5c offsetof(struct net_device, stats.tx_dropped)=0x6c Network drivers that touch dev->stats.tx_packets/stats.tx_bytes in their tx path can slow down SMP operations, since they dirty a cache line that should stay shared (dev->features is needed in rx and tx paths) We could move away stats field in net_device but it wont help that much. (Two cache lines dirtied in tx path, we can do one only) Better solution is to add tx_packets/tx_bytes/tx_dropped in struct netdev_queue because this structure is already touched in tx path and counters updates will then be free (no increase in size) Signed-off-by: NEric Dumazet <dada1@cosmosbay.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
It is illegal to dereference a skb after a successful ndo_start_xmit() call. We must store skb length in a local variable instead. Bug was introduced in 2.6.27 by commit 0abf77e5 (net_sched: Add accessor function for packet length for qdiscs) Signed-off-by: NEric Dumazet <dada1@cosmosbay.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Ilpo Järvinen 提交于
Commit 518a09ef (tcp: Fix recvmsg MSG_PEEK influence of blocking behavior) lets the loop run longer than the race check did previously expect, so we need to be more careful with this check and consider the work we have been doing. I tried my best to deal with urg hole madness too which happens here: if (!sock_flag(sk, SOCK_URGINLINE)) { ++*seq; ... by using additional offset by one but I certainly have very little interest in testing that part. Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi> Tested-by: NFrans Pop <elendil@planet.nl> Tested-by: NIan Zimmermann <itz@buug.org> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
- 18 5月, 2009 10 次提交
-
-
由 Stephen Hemminger 提交于
If bridge is configured with no STP and forwarding delay of 0 (which is typical for virtualization) then when link starts it will flood all packets for the first 20 seconds. This bug was introduced by a combination of earlier changes: * forwarding database uses hold time of zero to indicate user wants to always flood packets * optimzation of the case of forwarding delay of 0 avoids the initial timer tick The fix is to just skip all the topology change detection code if kernel STP is not being used. Signed-off-by: NStephen Hemminger <shemminger@vyatta.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Stephen Hemminger 提交于
Currently the bridge catches all STP packets; even if STP is turned off. This prevents other systems (which do have STP turned on) from being able to detect loops in the network. With this patch, if STP is off, then any packet sent to the STP multicast group address is forwarded to all ports. Based on earlier patch by Joakim Tjernlund with changes to go through forwarding (not local chain), and optimization that only last octet needs to be checked. Signed-off-by: NStephen Hemminger <shemminger@vyatta.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
struct net_device trans_start field is a hot spot on SMP and high performance devices, particularly multi queues ones, because every transmitter dirties it. Is main use is tx watchdog and bonding alive checks. But as most devices dont use NETIF_F_LLTX, we have to lock a netdev_queue before calling their ndo_start_xmit(). So it makes sense to move trans_start from net_device to netdev_queue. Its update will occur on a already present (and in exclusive state) cache line, for free. We can do this transition smoothly. An old driver continue to update dev->trans_start, while an updated one updates txq->trans_start. Further patches could also put tx_bytes/tx_packets counters in netdev_queue to avoid dirtying dev->stats (vlan device comes to mind) Signed-off-by: NEric Dumazet <dada1@cosmosbay.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 John Dykstra 提交于
When TCP frees up write buffer space, avoid waking up tasks that have done a poll() or select() on the same socket specifying read-side events. This is an extension of a read-side patch by Eric Dumazet. Signed-off-by: NJohn Dykstra <john.dykstra1@gmail.com> Acked-by: NEric Dumazet <dada1@cosmosbay.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Chris Friesen 提交于
If a DHCP server is delayed, it's possible for the client to receive the DHCPOFFER after it has already sent out a new DHCPDISCOVER message from a second interface. The client then sends out a DHCPREQUEST from the second interface, but the server doesn't recognize the device and rejects the request. This patch simply tracks the current device being configured and throws away the OFFER if it is not intended for the current device. A more sophisticated approach would be to put the OFFER information into the struct ic_device rather than storing it globally. Signed-off-by: NChris Friesen <cfriesen@nortel.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Pavel Emelyanov 提交于
It looks like the dev in netpoll_poll can be NULL - at lease it's checked at the function beginning. Thus the dev->netde_ops dereference looks dangerous. Signed-off-by: NPavel Emelyanov <xemul@openvz.org> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Jiri Pirko 提交于
Missed part of "&" removal. Signed-off-by: NJiri Pirko <jpirko@redhat.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Li Zefan 提交于
We can remove this lock here, since we are in cgroup write handler and thus the cgrp is guaranteed to be valid, and no lock is needed when writing a u32 variable. Signed-off-by: NLi Zefan <lizf@cn.fujitsuc.com> Acked-by: NPaul Menage <menage@google.com> Signed-off-by: NAndrew Morton <akpm@linux-foundation.org> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Jiri Pirko 提交于
Patch fixes issues with dev->dev_addr changing from array to pointer. Hopefully there are no others. Signed-off-by: NJiri Pirko <jpirko@redhat.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Rami Rosen 提交于
Signed-off-by: NRami Rosen <ramirose@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
- 14 5月, 2009 11 次提交
-
-
由 Johannes Berg 提交于
There's this internal wifi_wme_noack_test variable that we use to set the QoS control if set. For one, it is unlikely that it is set. Secondly, if set it needs to influence the IEEE80211_TX_CTL_NO_ACK TX control flag, and finally we should also be able to set it at all, so make it available in debugfs. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
Currently mac80211 announces a rate set with no basic rates, this fixes it to use 1/2 or 6/9 Mbit as basic rates by default. Additionally, mac80211 will currently adopt the peer's entire rate set, rather than just the basic rate set; fix that too. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
Even when we find an IBSS with the SSID we're looking for, we may not be able to connect to it because it has a key and we don't, or vice versa. Avoid such situations by checking the privacy capability bit. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
The time we wait for a probe response after probing an AP due to beacon loss is currently the same as the monitoring interval, 2s. This is far too long, APs should respond to probes within a fraction of that time. To be able to adjust both values, add a new constant IEEE80211_PROBE_WAIT, use it for checking the probe response, and adjust it down to 200ms instead of 2 seconds. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
The driver might keep reporting beacon loss until we disassociate -- catch that and don't respond to any subsequent events until the probe is either successful or we disassociate. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
Even though they are true, they cause sparse to complain because it doesn't see the __acquires(dev_base_lock) on dev_seq_start() because it is only added to the function in net/core/dev.c, not the header file. To keep track of the nesting correctly we should probably annotate those functions publically, but for now let's just remove the annotation I added to wext. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Jouni Malinen 提交于
When setting a key with NL80211_CMD_NEW_KEY, we should allow the key sequence number (RSC) to be set in order to allow replay protection to work correctly for group keys. This patch documents this use for nl80211 and adds the couple of missing pieces in nl80211/cfg80211 and mac80211 to support this. In addition, WEXT SIOCSIWENCODEEXT compat processing in cfg80211 is extended to handle the RSC (this was already specified in WEXT, but just not implemented in cfg80211/mac80211). Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Jouni Malinen 提交于
Add a new NL80211_ATTR_CONTROL_PORT flag for NL80211_CMD_ASSOCIATE to allow user space to indicate that it will control the IEEE 802.1X port in station mode. Previously, mac80211 was always marking the port authorized in station mode. This was enough when drop_unencrypted flag was set. However, drop_unencrypted can currently be controlled only with WEXT and the current nl80211 design does not allow fully secure configuration. Fix this by providing a mechanism for user space to control the IEEE 802.1X port in station mode (i.e., do the same that we are already doing in AP mode). Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
It is currently not possible to modify station flags, but that capability would be very useful. This patch introduces a new nl80211 attribute that contains a set/mask for station flags, and updates the internal API (and mac80211) to mirror that. The new attribute is parsed before falling back to the old so that userspace can specify both (if it can) to work on all kernels. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Jouni Malinen 提交于
NL80211_STA_FLAG_MFP was forgotten from sta_flags_policy. The previous version added the flag due to the loop used in parse_station_flags, but the proper behavior would be to allow nla_parse_nested() to go through the policy for all flags. Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
Move key handling wireless extension ioctls from mac80211 to cfg80211 so that all drivers that implement the cfg80211 operations get wext compatibility. Note that this drops the SIOCGIWENCODE ioctl support for getting IW_ENCODE_RESTRICTED/IW_ENCODE_OPEN. This means that iwconfig will no longer report "Security mode:open" or "Security mode:restricted" for mac80211. However, what we displayed there (the authentication algo used) was actually wrong -- linux/wireless.h states that this setting is meant to differentiate between "Refuse non-encoded packets" and "Accept non-encoded packets". (Combined with "cfg80211: fix a couple of bugs with key ioctls". -- JWL) Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
- 12 5月, 2009 11 次提交
-
-
由 Johannes Berg 提交于
Currently, get_wireless_stats is racy by _design_. This is because it returns a buffer, which needs to be statically allocated since it cannot be freed if it was allocated dynamically. Also, SIOCGIWSTATS and /proc/net/wireless use no common lock, and /proc/net/wireless accesses are not synchronised against each other. This is a design flaw in get_wireless_stats since the beginning. This patch fixes it by wrapping /proc/net/wireless accesses with the RTNL so they are protected against each other and SIOCGIWSTATS. The more correct method of fixing this would be to pass in the buffer instead of returning it and have the caller take care of synchronisation of the buffer, but even then most drivers probably assume that their callback is protected by the RTNL like all other wext callbacks. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
On non-AP interfaces userspace has no business interfering with the station management, this can confuse mac80211 (and other drivers probably wouldn't support it anyway). Allow adding and removing stations only on AP interfaces. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
To make it more apparent in the code what is for wext only (and needs to be #ifdef'ed) put all the info for wext into a substruct in each wireless_dev. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
The address pointed to by mac_addr can be marked as const. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
When we disassociate, we set the channel to non-HT which obviously invalidates any ht_operation_mode setting. But when we then associate with the next AP again, we might still have the ht_operation_mode from the previous AP cached and fail to configure the hardware with the new (but unchanged) operation mode. This patch fixes it by separately tracking whether our cache is valid. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
There really is no need to have a separate struct for a single variable. The fact that it exists is due to the code legacy, but we can remove that now. Very simple. Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
The call to ieee80211_hw_config() is supposed to apply changes synchronously, so once it returns the parameters are applied to the hardware. Thus, there really is no need to delay the probing by the channel switch time again since the channel switch has already happened once we get to this code. Additionally, there is no need to wait for a NAV update (probe delay) when the channel is passively scanned. Remove that extra time too. This cuts scanning time from over 7 seconds to under 4 on ar9170, which is due to the number of channels scanned and ar9170's switch time being advertised as 135ms (my test now indicates it is about 77ms with the current driver, but the difference might also be due to using a different machine with different USB controllers). Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Jouni Malinen 提交于
When management frame protection (IEEE 802.11w) is used, unprotected Robust Action frames are not allowed prior to key configuration. However, unprotected Deauthentication and Disassociation frames are allowed at that point, but not after key configuration. Make ieee80211_drop_unencrypted() handle the special cases for MFP by separating the basic Data frame case from Management frame processing and handle the Management frames only if MFP has been negotiated. In addition, do not use sdata->drop_unencrypted for Management frames since the decision on whether to accept the frame depends on the key being configured. Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Jouni Malinen 提交于
When using nl80211, we do not have a mechanism to set sdata->drop_unencrypted. Currently, this breaks code that is supposed to drop unencrypted frames when protection is expected since ieee80211_rx_h_decrypt() is optimized to not set rx->key when the frame is not protected. This patch modifies ieee80211_rx_h_decrypt() to set rx->key for all frames and only skip decryption if the frame is not protected. This allows ieee80211_drop_unencrypted() to correctly drop frames even if drop_unencrypted is not set. The changes here are not enough to handle all cases, though. Additional patches will be needed to implement proper IEEE 802.1X PAE for station mode (currently, this is only used for AP mode) and some additional rules are needed for MFP to drop unprotected Robust Action frames prior to having PTK and IGTK configured. In theory, the unprotected frames could and should be dropped in ieee80211_rx_h_decrypt(). However, due to the special case with EAPOL frames that have to be allowed to be received unprotected even when keys are set, it is simpler to only set rx->key and allow the ieee80211_frame_allowed() function to handle the actual dropping of data frames after 802.11->802.3 header conversion. In addition, unprotected robust management frames are dropped before they are processed. Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
The function to parse a struct iw_freq has a stupid bug, it returns NULL when the channel cannot be found at all, but NULL is supposed to mean "auto". Fix this by checking the return value of ieee80211_get_channel() and returning ERR_PTR(-EINVAL) if it returned NULL (channel not found). This fixes an issue where you could say (in IBSS mode) iwconfig wlan0 channel 21 and it would use channel 1 instead because that's the first available channel with IBSS allowed (which is what the "auto" setting uses). Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-
由 Johannes Berg 提交于
We've never really cared about the default QoS (WMM) values, but we really should if the AP doesn't send any. This patch makes mac80211 use the default values according to 802.11-2007, and additionally syncs the default values when we disassociate so whatever the last AP said gets "unconfigured". Signed-off-by: NJohannes Berg <johannes@sipsolutions.net> Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
-