summaryrefslogtreecommitdiffstats
path: root/net
Commit message (Collapse)AuthorAgeFilesLines
* Merge branch 'master' of ↵John W. Linville2013-06-2135-762/+1561
|\ | | | | | | | | | | | | git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next into for-davem Conflicts: net/wireless/nl80211.c
| * Merge branch 'master' of ↵John W. Linville2013-06-187-17/+105
| |\ | | | | | | | | | git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless
| * | NFC: llcp: Fix the well known services endiannessSamuel Ortiz2013-06-141-2/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | The WKS (Well Known Services) bitmask should be transmitted in big endian order. Picky implementations will refuse to establish an LLCP link when the WKS bit 0 is not set to 1. The vast majority of implementations out there are not that picky though... Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: llcp: Set the LLC Link Management well known service bitSamuel Ortiz2013-06-141-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | In order to advertise our LLCP support properly and to follow the LLCP specs requirements, we need to initialize the WKS (Well-Known Services) bitfield to 1 as SAP 0 is the only mandatory supported service. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: llcp: Do not send pending Tx frames when the remote is not readySamuel Ortiz2013-06-141-0/+3
| | | | | | | | | | | | | | | | | | | | | | | | When we receive a RNR, the remote is busy processing the last received frame. We set a local flag for that, and we should send a SYMM when it is set instead of sending any pending frame. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: llcp: Fix non blocking sockets connectionsSamuel Ortiz2013-06-142-3/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Without the new LLCP_CONNECTING state, non blocking sockets will be woken up with a POLLHUP right after calling connect() because their state is stuck at LLCP_CLOSED. That prevents userspace from implementing any proper non blocking socket based NFC p2p client. Cc: stable@vger.kernel.org Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Fix a potential memory leakThierry Escande2013-06-141-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | In nfc_llcp_tx_work() the sk_buff is not freed when the llcp_sock is null and the PDU is an I one. Signed-off-by: Thierry Escande <thierry.escande@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Keep socket alive until the DISC PDU is actually sentThierry Escande2013-06-143-0/+15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch keeps the socket alive and therefore does not remove it from the sockets list in the local until the DISC PDU has been actually sent. Otherwise we would reply with DM PDUs before sending the DISC one. Signed-off-by: Thierry Escande <thierry.escande@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Rename nfc_llcp_disconnect() to nfc_llcp_send_disconnect()Thierry Escande2013-06-143-24/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | nfc_llcp_send_disconnect() already exists but is not used. nfc_llcp_disconnect() naming is not consistent with other PDU sending functions. This patch removes nfc_llcp_send_disconnect() and renames nfc_llcp_disconnect() Signed-off-by: Thierry Escande <thierry.escande@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add secure element enablement netlink APISamuel Ortiz2013-06-141-0/+56
| | | | | | | | | | | | | | | | | | | | | | | | | | | Enabling or disabling an NFC accessible secure element through netlink requires giving both an NFC controller and a secure element indexes. Once enabled the secure element will handle card emulation once polling starts. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add secure element enablement internal APISamuel Ortiz2013-06-142-4/+109
| | | | | | | | | | | | | | | | | | | | | | | | | | | Called via netlink, this API will enable or disable a specific secure element. When a secure element is enabled, it will handle card emulation and more generically ISO-DEP target mode, i.e. all target mode cases except for p2p target mode. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Remove and free all SEs when releasing an NFC deviceSamuel Ortiz2013-06-141-0/+8
| | | | | | | | | | | | Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Send netlink events for secure elements additions and removalsSamuel Ortiz2013-06-143-0/+80
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | When an NFC driver or host controller stack discovers a secure element, it will call nfc_add_se(). In order for userspace applications to use these secure elements, a netlink event will then be sent with the SE index and its type. With that information userspace applications can decide wether or not to enable SEs, through their indexes. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add secure elements addition and removal APISamuel Ortiz2013-06-141-1/+44
| | | | | | | | | | | | | | | | | | | | | | | | This API will allow NFC drivers to add and remove the secure elements they know about or detect. Typically this should be called (asynchronously or not) from the driver or the host interface stack detect_se hook. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Extend and fix the internal secure element APISamuel Ortiz2013-06-143-0/+58
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Secure elements need to be discovered after enabling the NFC controller. This is typically done by the NCI core and the HCI drivers (HCI does not specify how to discover SEs, it is left to the specific drivers). Also, the SE enable/disable API explicitely takes a SE index as its argument. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Remove the static supported_se fieldSamuel Ortiz2013-06-145-9/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Supported secure elements are typically found during a discovery process initiated when the NFC controller is up and running. For a given NFC chipset there can be many configurations (embedded SE or not, with or without a SIM card wired to the NFC controller SWP interface, etc...) and thus driver code will never know before hand which SEs are available. So we remove this field, it will be replaced by a real SE discovery mechanism. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add NCI over SPI receiveFrederic Danis2013-06-141-0/+174
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Before any operation, driver interruption is de-asserted to prevent race condition between TX and RX. Transaction starts by emitting "Direct read" and acknowledged mode bytes. Then packet length is read allowing to allocate correct NCI socket buffer. After that payload is retrieved. A delay after the transaction can be added. This delay is determined by the driver during nci_spi_allocate_device() call and can be 0. If acknowledged mode is set: - CRC of header and payload is checked - if frame reception fails (CRC error): NACK is sent - if received frame has ACK or NACK flag: unblock nci_spi_send() Payload is passed to NCI module. At the end, driver interruption is re asserted. Signed-off-by: Frederic Danis <frederic.danis@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add NCI over SPI sendFrederic Danis2013-06-141-1/+70
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Before any operation, driver interruption is de-asserted to prevent race condition between TX and RX. The NCI over SPI header is added in front of NCI packet. If acknowledged mode is set, CRC-16-CCITT is added to the packet. Then the packet is forwarded to SPI module to be sent. A delay after the transaction is added. This delay is determined by the driver during nci_spi_allocate_device() call and can be 0. After data has been sent, driver interruption is re-asserted. If acknowledged mode is set, nci_spi_send will block until acknowledgment is received. Signed-off-by: Frederic Danis <frederic.danis@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add basic NCI over SPIFrederic Danis2013-06-143-1/+149
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The NFC Forum defines a transport interface based on Serial Peripheral Interface (SPI) for the NFC Controller Interface (NCI). This module implements the SPI transport of NCI, calling SPI module directly to read/write data to NFC controller (NFCC). NFCC driver should provide functions performing device open and close. It should also provide functions asserting/de-asserting interruption to prevent TX/RX race conditions. NFCC driver can also fix a delay between transactions if needed by the hardware. Signed-off-by: Frederic Danis <frederic.danis@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: HCI: Follow a positive code path in the HCI ops implementationsSamuel Ortiz2013-06-141-18/+18
| | | | | | | | | | | | | | | | | | Exiting on the error case is more typical to the kernel coding style. Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: HCI: Implement fw_upload opsEric Lapuyade2013-06-141-0/+11
| | | | | | | | | | | | | | | | | | | | | | | | | | | This is a simple forward to the HCI driver. When driver is done with the operation, it shall directly notify NFC Core by calling nfc_fw_upload_done(). Signed-off-by: Eric Lapuyade <eric.lapuyade@intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: Add firmware upload netlink commandEric Lapuyade2013-06-143-0/+114
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | As several NFC chipsets can have their firmwares upgraded and reflashed, this patchset adds a new netlink command to trigger that the driver loads or flashes a new firmware. This will allows userspace triggered firmware upgrade through netlink. The firmware name or hint is passed as a parameter, and the driver will eventually fetch the firmware binary through the request_firmware API. The cmd can only be executed when the nfc dev is not in use. Actual firmware loading/flashing is an asynchronous operation. Result of the operation shall send a new event up to user space through the nfc dev multicast socket. During operation, the nfc dev is not openable and thus not usable. Signed-off-by: Eric Lapuyade <eric.lapuyade@intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | NFC: NCI: Fix skb->dev usageFrederic Danis2013-06-142-13/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | skb->dev is used for carrying a net_device pointer and not an nci_dev pointer. Remove usage of skb-dev to carry nci_dev and replace it by parameter in nci_recv_frame(), nci_send_frame() and driver send() functions. NfcWilink driver is also updated to use those functions. Signed-off-by: Frederic Danis <frederic.danis@linux.intel.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
| * | Merge branch 'for-john' of ↵John W. Linville2013-06-1222-698/+646
| |\ \ | | | | | | | | | | | | | | | | | | | | | | | | git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next Conflicts: drivers/net/wireless/iwlwifi/mvm/mac80211.c
| | * | cfg80211: fix rtnl leak in wiphy dump error casesJohannes Berg2013-06-111-1/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In two wiphy dump error cases, most often when the dump allocation must be increased, the RTNL is leaked. This quickly results in a complete system lockup. Release the RTNL correctly. Reported-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | nl80211: allow sending CMD_FRAME without specifying any frequencyAntonio Quartulli2013-06-111-4/+15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Users may want to send a frame on the current channel without specifying it. This is particularly useful for the correct implementation of the IBSS/RSN support in wpa_supplicant which requires to receive and send AUTH frames. Make mgmt_tx pass a NULL channel to the driver if none has been specified by the user. Signed-off-by: Antonio Quartulli <antonio@open-mesh.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: make mgmt_tx accept a NULL channelAntonio Quartulli2013-06-111-3/+14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | cfg80211 passes a NULL channel to mgmt_tx if the frame has to be sent on the one currently in use by the device. Make the implementation of mgmt_tx correctly handle this case. Fail if offchan is required. Signed-off-by: Antonio Quartulli <antonio@open-mesh.com> [fix RCU locking] Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | cfg80211: fix VHT TDLS peer AID verificationJouni Malinen2013-06-111-5/+7
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | I (Johannes) accidentally applied the first version of the patch ("Allow TDLS peer AID to be configured for VHT"). Now apply just the changes between v1 and v2 to get the AID verification and prefer the new attribute over the old one. Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | {nl,mac,cfg}80211: Allow user to configure basic rates for meshAshok Nagarajan2013-06-114-4/+28
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Currently mesh uses mandatory rates as the default basic rates. Allow basic rates to be configured during mesh join. Basic rates are applied only if channel is also provided with mesh join command. Signed-off-by: Ashok Nagarajan <ashok@cozybit.com> [some whitespace fixes, refuse basic rates w/o channel] Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: expire mesh peers based on mesh configurationColleen Twitty2013-06-113-2/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The time it takes to see the peer link expire may differ by a minute since sta_expire() is run once a minute as a mesh housekeeping task. Signed-off-by: Colleen Twitty <colleen@cozybit.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | {nl,cfg}80211: make peer link expiration time configurableColleen Twitty2013-06-112-1/+9
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | If a STA has a peer that it hasn't seen any tx activity from for a certain length of time, the peer link is expired. This means the inactive STA is removed from the list of peers and that STA is not considered a peer again unless it re-peers. Previously, this inactivity time was always 30 minutes. Now, add it to the mesh configuration and allow it to be configured. Retain 30 minutes as a default value. Signed-off-by: Colleen Twitty <colleen@cozybit.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: fix mesh deadlockThomas Pedersen2013-06-112-18/+18
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The patch "cfg80211/mac80211: use cfg80211 wdev mutex in mac80211" introduced several deadlocks by converting the ifmsh->mtx to wdev->mtx. Solve these by: 1. drop the cancel_work_sync() in ieee80211_stop_mesh(). Instead make the mesh work conditional on whether the mesh is running or not. 2. lock the mesh work with sdata_lock() to protect beacon updates and prevent races with wdev->mesh_id_len or cfg80211. Signed-off-by: Thomas Pedersen <thomas@cozybit.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | wireless: fix kernel-docJohannes Berg2013-06-051-0/+3
| | | | | | | | | | | | | | | | | | | | | | | | Some kernel-doc fixes for forgotten fields and renamed things. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: Use suitable semantics for beacon availability indicationAlexander Bondar2013-06-053-12/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Currently beacon availability upon association is marked by have_beacon flag of assoc_data structure that becomes unavailable when association completes. However beacon availability indication is required also after association to inform a driver. Currently dtim_period parameter is used for this purpose. Move have_beacon flag to another structure, persistant throughout a interface's life cycle. Use suitable sematics for beacon availability indication. Signed-off-by: Alexander Bondar <alexander.bondar@intel.com> [fix another instance of BSS_CHANGED_DTIM_PERIOD in docs] Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: fix powersave bug and clean up ieee80211_rx_bss_infoAlexander Bondar2013-06-051-21/+9
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | ieee80211_rx_bss_info() deals with dtim_period setting and PS update when associated. Move all these to another locations cleaning this function. Also, the current implementation is buggy because when it calls ieee80211_recalc_ps() bss_conf->dtim_period is notset properly yet and thus nothing will happen. Signed-off-by: Alexander Bondar <alexander.bondar@intel.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | cfg80211: make wiphy index start at 0 againJohannes Berg2013-06-041-0/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The change to use atomic_inc_return() for assigning the wiphy index made the first wiphy index 1 instead of 0. This is fine, but we all habitually type "phy0" when we're testing, so make it go back to 0 instead of 1 by subtracting 1 from the index. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | cfg80211: fix potential deadlock regressionJohannes Berg2013-06-041-15/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | My big locking cleanups caused a problem by registering the rfkill instance with the RTNL held, while the callback also acquires the RTNL. This potentially causes a deadlock since the two locks used (rfkill mutex and RTNL) can be acquired in two different orders. Fix this by (un)registering rfkill without holding the RTNL. This needs to be done after the device struct is registered, but that can also be done w/o holding the RTNL. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | regulatory: use proper enum return valueJohannes Berg2013-06-041-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | get_reg_request_treatment() returns 0 in one case but is defined to return an enum, use the proper value REG_REQ_OK. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | cfg80211: separate internal SME implementationJohannes Berg2013-06-047-448/+335
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The current internal SME implementation in cfg80211 is very mixed up with the MLME handling, which has been causing issues for a long time. There are three things that the implementation has to provide: * a basic SME implementation for nl80211's connect() call (for drivers implementing auth/assoc, which is really just mac80211) and wireless extensions * MLME events for the userspace SME * SME events (connected, disconnected etc.) for all different SME implementation possibilities (driver, cfg80211 and userspace) To achieve these goals it isn't necessary to track the software SME's connection status outside of it's state (which is the part that caused many issues.) Instead, track it only in the SME data (wdev->conn) and in the general case only track whether the wdev is connected or not (via wdev->current_bss.) Also separate the internal implementation to not have callbacks from the SME events, but rather call it from the API functions that the driver (or rather mac80211) calls. This separates the code better. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | cfg80211/mac80211: clean up cfg80211 SME APIsJohannes Berg2013-06-045-98/+134
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Do some cleanups in the cfg80211 SME APIs, which are only used by mac80211. Most of these functions get a frame passed, and there isn't really any reason to export multiple functions as cfg80211 can check the frame type instead, do that. Additionally, the API functions have confusing names like cfg80211_send_...() which was meant to indicate that it sends an event to userspace, but gets a bit confusing when there's both TX and RX and they're not all clearly labeled. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: set IEEE80211_TX_CTL_REQ_TX_STATUS on nullframesPontus Fuchs2013-06-041-0/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The connection monitor needs to know the tx status of nullframes to work properly. Signed-off-by: Pontus Fuchs <pontus.fuchs@gmail.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | nl80211: remove bogus genlmsg_end() error checkingJohannes Berg2013-06-041-33/+7
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | genlmsg_end() can't return an error since it returns the skb length so remove checks treating the return value as an error code. Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: add a tx control flag to indicate PS-Poll/uAPSD responseFelix Fietkau2013-06-041-1/+3
| | | | | | | | | | | | | | | | | | | | Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | cfg80211: take WoWLAN support information out of wiphy structJohannes Berg2013-06-033-39/+40
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | There's no need to take up the space for devices that don't support WoWLAN, and most drivers can even make the support data static const (except where it's modified at runtime.) Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: set mesh formation field properlyJacob Minshall2013-06-032-4/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Cap max peerings at 63 in accordance with IEEE-2012 8.4.2.100.7. Triggers a beacon regeneration every time the number of peerings changes. Previously this would only happen if the "accepting peerings" bit changed. Signed-off-by: Jacob Minshall <jacob@cozybit.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| | * | mac80211: don't check local mesh TTL on TXThomas Pedersen2013-06-031-6/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | nl80211 has already verified the mesh TTL on setting the mesh config, so no need to check it again in mac80211. Signed-off-by: Thomas Pedersen <thomas@cozybit.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
| * | | Merge branch 'master' of ↵John W. Linville2013-06-126-30/+102
| |\ \ \ | | |/ / | |/| | | | | | | | | | | | | | | | | | | | | | git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless Conflicts: drivers/net/wireless/ath/ath9k/Kconfig net/mac80211/iface.c
* | | | inet: frag , remove an empty ifdef.Rami Rosen2013-06-191-2/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch removes an empty ifdef from inet_frag_intern() in net/ipv4/inet_fragment.c. commit b67bfe0d42cac56c512dd5da4b1b347a23f4b70a (hlist: drop the node parameter from iterators) removed hlist from net/ipv4/inet_fragment.c, but did not remove the enclosing ifdef command, which is now empty. Signed-off-by: Rami Rosen <ramirose@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | htb: refactor struct htb_sched fields for performanceEric Dumazet2013-06-191-86/+95
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | htb_sched structures are big, and source of false sharing on SMP. Every time a packet is queued or dequeue, many cache lines must be touched because structures are not lay out properly. By carefully splitting htb_sched in two parts, and define sub structures to increase data locality, we can improve performance dramatically on SMP. New htb_prio structure can also be used in htb_class to increase data locality. I got 26 % performance increase on a 24 threads machine, with 200 concurrent netperf in TCP_RR mode, using a HTB hierarchy of 4 classes. Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Tom Herbert <therbert@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | tcp: introduce a per-route knob for quick ackCong Wang2013-06-192-3/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In previous discussions, I tried to find some reasonable heuristics for delayed ACK, however this seems not possible, according to Eric: "ACKS might also be delayed because of bidirectional traffic, and is more controlled by the application response time. TCP stack can not easily estimate it." "ACK can be incredibly useful to recover from losses in a short time. The vast majority of TCP sessions are small lived, and we send one ACK per received segment anyway at beginning or retransmits to let the sender smoothly increase its cwnd, so an auto-tuning facility wont help them that much." and according to David: "ACKs are the only information we have to detect loss. And, for the same reasons that TCP VEGAS is fundamentally broken, we cannot measure the pipe or some other receiver-side-visible piece of information to determine when it's "safe" to stretch ACK. And even if it's "safe", we should not do it so that losses are accurately detected and we don't spuriously retransmit. The only way to know when the bandwidth increases is to "test" it, by sending more and more packets until drops happen. That's why all successful congestion control algorithms must operate on explicited tested pieces of information. Similarly, it's not really possible to universally know if it's safe to stretch ACK or not." It still makes sense to enable or disable quick ack mode like what TCP_QUICK_ACK does. Similar to TCP_QUICK_ACK option, but for people who can't modify the source code and still wants to control TCP delayed ACK behavior. As David suggested, this should belong to per-path scope, since different pathes may want different behaviors. Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Rick Jones <rick.jones2@hp.com> Cc: Stephen Hemminger <stephen@networkplumber.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Thomas Graf <tgraf@suug.ch> CC: David Laight <David.Laight@ACULAB.COM> Signed-off-by: Cong Wang <amwang@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
OpenPOWER on IntegriCloud