aboutsummaryrefslogtreecommitdiff
path: root/net
AgeCommit message (Collapse)AuthorFilesLines
2016-02-24Merge tag 'mac80211-for-davem-2016-02-23' of ↵David S. Miller7-20/+60
git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211 Johannes Berg says: ==================== Another small set of fixes: * stop critical protocol session on disconnect to avoid it getting stuck * wext: fix two RTNL message ordering issues * fix an uninitialized value (found by KASAN) * fix an out-of-bounds access (also found by KASAN) * clear connection keys when freeing them in all cases (IBSS, all other places already did so) * fix expected throughput unit to get consistent values * set default TX aggregation timeout to 0 in minstrel to avoid (really just hide) issues and perform better ==================== Signed-off-by: David S. Miller <[email protected]>
2016-02-24bpf: fix csum setting for bpf_set_tunnel_keyDaniel Borkmann1-2/+4
The fix in 35e2d1152b22 ("tunnels: Allow IPv6 UDP checksums to be correctly controlled.") changed behavior for bpf_set_tunnel_key() when in use with IPv6 and thus uncovered a bug that TUNNEL_CSUM needed to be set but wasn't. As a result, the stack dropped ingress vxlan IPv6 packets, that have been sent via eBPF through collect meta data mode due to checksum now being zero. Since after LCO, we enable IPv4 checksum by default, so make that analogous and only provide a flag BPF_F_ZERO_CSUM_TX for the user to turn it off in IPv4 case. Fixes: 35e2d1152b22 ("tunnels: Allow IPv6 UDP checksums to be correctly controlled.") Fixes: c6c33454072f ("bpf: support ipv6 for bpf_skb_{set,get}_tunnel_key") Signed-off-by: Daniel Borkmann <[email protected]> Acked-by: Alexei Starovoitov <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24netfilter: ipset: Fix set:list type crash when flush/dump set in parallelJozsef Kadlecsik2-30/+28
Flushing/listing entries was not RCU safe, so parallel flush/dump could lead to kernel crash. Bug reported by Deniz Eren. Fixes netfilter bugzilla id #1050. Signed-off-by: Jozsef Kadlecsik <[email protected]>
2016-02-24libceph: don't spam dmesg with stray reply warningsIlya Dryomov1-2/+2
Commit d15f9d694b77 ("libceph: check data_len in ->alloc_msg()") mistakenly bumped the log level on the "tid %llu unknown, skipping" message. Turn it back into a dout() - stray replies are perfectly normal when OSDs flap, crash, get killed for testing purposes, etc. Cc: [email protected] # 4.3+ Signed-off-by: Ilya Dryomov <[email protected]> Reviewed-by: Alex Elder <[email protected]>
2016-02-24libceph: use the right footer size when skipping a messageIlya Dryomov1-2/+9
ceph_msg_footer is 21 bytes long, while ceph_msg_footer_old is only 13. Don't skip too much when CEPH_FEATURE_MSG_AUTH isn't negotiated. Cc: [email protected] # 3.19+ Signed-off-by: Ilya Dryomov <[email protected]> Reviewed-by: Alex Elder <[email protected]>
2016-02-24libceph: don't bail early from try_read() when skipping a messageIlya Dryomov1-2/+2
The contract between try_read() and try_write() is that when called each processes as much data as possible. When instructed by osd_client to skip a message, try_read() is violating this contract by returning after receiving and discarding a single message instead of checking for more. try_write() then gets a chance to write out more requests, generating more replies/skips for try_read() to handle, forcing the messenger into a starvation loop. Cc: [email protected] # 3.10+ Reported-by: Varada Kari <[email protected]> Signed-off-by: Ilya Dryomov <[email protected]> Tested-by: Varada Kari <[email protected]> Reviewed-by: Alex Elder <[email protected]>
2016-02-24ipv4: only create late gso-skb if skb is already set up with CHECKSUM_PARTIALHannes Frederic Sowa1-1/+4
Otherwise we break the contract with GSO to only pass CHECKSUM_PARTIAL skbs down. This can easily happen with UDP+IPv4 sockets with the first MSG_MORE write smaller than the MTU, second write is a sendfile. Returning -EOPNOTSUPP lets the callers fall back into normal sendmsg path, were we calculate the checksum manually during copying. Commit d749c9cbffd6 ("ipv4: no CHECKSUM_PARTIAL on MSG_MORE corked sockets") started to exposes this bug. Fixes: d749c9cbffd6 ("ipv4: no CHECKSUM_PARTIAL on MSG_MORE corked sockets") Reported-by: Jiri Benc <[email protected]> Cc: Jiri Benc <[email protected]> Reported-by: Wakko Warner <[email protected]> Cc: Wakko Warner <[email protected]> Signed-off-by: Hannes Frederic Sowa <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24eth: Pull header from first fragment via eth_get_headlenAlexander Duyck1-1/+2
We want to try and pull the L4 header in if it is available in the first fragment. As such add the flag to indicate we want to pull the headers on the first fragment in. Signed-off-by: Alexander Duyck <[email protected]> Acked-by: Tom Herbert <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24flow_dissector: Use same pointer for IPv4 and IPv6 addressesAlexander Duyck1-6/+5
The IPv6 parsing was using a local pointer when it could use the same pointer as the IPv4 portion of the code since the key_addrs can support both IPv4 and IPv6 as it is just a pointer. Signed-off-by: Alexander Duyck <[email protected]> Acked-by: Tom Herbert <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24flow_dissector: Correctly handle parsing FCoEAlexander Duyck1-2/+5
The flow dissector bits handling FCoE didn't bother to actually validate that the space there was enough for the FCoE header. So we need to update things so that if there is room we add the header and report a good result, otherwise we do not add the header, and report the bad result. Signed-off-by: Alexander Duyck <[email protected]> Acked-by: Tom Herbert <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24flow_dissector: Fix fragment handling for header length computationAlexander Duyck1-3/+7
It turns out that for IPv4 we were reporting the ip_proto of the fragment, and for IPv6 we were not. This patch updates that behavior so that we always report the IP protocol of the fragment. In addition it takes the steps of updating the payload offset code so that we will determine the start of the payload not including the L4 header for any fragment after the first. Signed-off-by: Alexander Duyck <[email protected]> Acked-by: Tom Herbert <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24flow_dissector: Check for IP fragmentation even if not using IPv4 addressAlexander Duyck1-8/+9
This patch corrects the logic for the IPv4 parsing so that it is consistent with how we handle IPv6. Specifically if we do not have the flow key indicating we want the addresses we still may need to take a look at the IP fragmentation bits and to see if we should stop after we have recognized the L3 header. Fixes: 807e165dc44f ("flow_dissector: Add control/reporting of fragmentation") Signed-off-by: Alexander Duyck <[email protected]> Acked-by: Tom Herbert <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-24soreuseport: fix merge conflict in tcp bindCraig Gallek1-0/+1
One of the validation checks for the new array-based TCP SO_REUSEPORT validation was unintentionally dropped in ea8add2b1903. This adds it back. Lack of this check allows the user to allocate multiple sock_reuseport structures (leaking all but the first). Fixes: ea8add2b1903 ("tcp/dccp: better use of ephemeral ports in bind()") Signed-off-by: Craig Gallek <[email protected]> Signed-off-by: David S. Miller <[email protected]>
2016-02-246lowpan: fix error checking codeAndrzej Hajda1-1/+1
Bool variable 'fail' is always non-negative, it indicates an error if it is true. The problem has been detected using coccinelle script scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci Signed-off-by: Andrzej Hajda <[email protected]> Acked-by: Alexander Aring <[email protected]> Signed-off-by: Marcel Holtmann <[email protected]>
2016-02-246lowpan: iphc: fix stateful multicast compressionAlexander Aring1-1/+2
In case of multicast address we need to set always the LOWPAN_IPHC_M bit and if a destination context identifier was found for a multicast address then we need to set the LOWPAN_IPHC_DAC as well. Signed-off-by: Alexander Aring <[email protected]> Signed-off-by: Marcel Holtmann <[email protected]>
2016-02-24rfkill: Factor rfkill_global_states[].cur assignmentsJoão Paulo Rechi Vita1-21/+17
Factor all assignments to rfkill_global_states[].cur into a single function rfkill_update_global_state(). Signed-off-by: João Paulo Rechi Vita <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24rfkill: Remove extra blank lineJoão Paulo Rechi Vita1-1/+0
Signed-off-by: João Paulo Rechi Vita <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24rfkill: Improve documentation languageJoão Paulo Rechi Vita1-2/+2
Signed-off-by: João Paulo Rechi Vita <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24net: rfkill: gpio: remove rfkill_gpio_platform_dataHeikki Krogerus2-10/+1
No more users for it. Signed-off-by: Heikki Krogerus <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24net: rfkill: gpio: get the name and type from device propertyHeikki Krogerus1-5/+11
This prepares the driver for removal of platform data. Signed-off-by: Heikki Krogerus <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24net: rfkill: add rfkill_find_type functionHeikki Krogerus1-28/+30
Helper for finding the type based on name. Useful if the type needs to be determined based on device property. Signed-off-by: Heikki Krogerus <[email protected]> [modify rfkill_types array and BUILD_BUG_ON to not cause errors] Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: add radiotap VHT info to rtap_namespace_sizesLorenzo Bianconi1-0/+1
Add IEEE80211_RADIOTAP_VHT entry to rtap_namespace_sizes array in order to define alignment and size of VHT info in tx radiotap Signed-off-by: Lorenzo Bianconi <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: Add global RRM capabilityBeni Lev1-7/+11
Today, the supplicant will add the RRM capabilities Information Element in the association request only if Quiet period is supported (NL80211_FEATURE_QUIET). Quiet is one of many RRM features, and there are other RRM features that are not related to Quiet (e.g. neighbor report). Therefore, requiring Quiet to enable RRM is too restrictive. Some of the features, like neighbor report, can be supported by user space without any help from the kernel. Hence adding the RRM capabilities IE to association request should be the sole user space's decision. Removing the RRM dependency on Quiet in the driver solves this problem, but using an old driver with a user space tool that would not require Quiet feature would be problematic: the user space would add NL80211_ATTR_USE_RRM in the association request even if the kernel doesn't advertize NL80211_FEATURE_QUIET and the association would be denied by the kernel. This solution adds a global RRM capability, that tells user space that it can request RRM capabilities IE publishment without any specific feature support in the kernel. Signed-off-by: Beni Lev <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: move MU_MIMO_OWNER flag to ieee80211_vifSara Sharon4-12/+10
Drivers may need to track which vif is using VHT MU-MIMO. Move the flag indicationg the ownership of MU_MIMO to ieee80211_vif. Signed-off-by: Sara Sharon <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: provide interface to driver to set VHT MU-MIMO dataSara Sharon1-1/+15
Provide an interface to the lower level driver to set the VHT MU-MIMO data. This is needed for example when there is an update of the group data during low power state, where the management frame will not be passed to the host at all. Signed-off-by: Sara Sharon <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: remove ieee80211_get_key_tx_seq/ieee80211_set_key_tx_seqEliad Peller1-87/+0
Since the PNs of all the tx keys are now tracked in the public part of the key struct (with atomic counter), we no longer need these functions. dvm and vt665{5,6} are currently the only users of these functions, so update them accordingly. Signed-off-by: Eliad Peller <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: move TKIP TX IVs to public part of key structEliad Peller7-38/+40
Some drivers/devices might want to set the IVs by themselves (and still let mac80211 generate MMIC). Specifically, this is needed when the device does offloading at certain times, and the driver has to make sure that the IVs of new tx frames (from the host) are synchronized with IVs that were potentially used during the offloading. Similarly to CCMP, move the TX IVs of TKIP keys to the public part of the key struct, and export a function to add the IV right into the crypto header. The public tx_pn field is defined as atomic64, so define TKIP_PN_TO_IV16/32 helper macros to convert it to iv16/32 when needed. Since the iv32 used for the p1k cache is taken directly from the frame, we can safely remove iv16/32 from being protected by tkip.txlock. Signed-off-by: Eliad Peller <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: fix wiphy supported_band accessLorenzo Bianconi1-4/+5
Fix wiphy supported_band access in tx radiotap parsing introduced in commit 5ec3aed9ba4c ("mac80211: Parse legacy and HT rate in injected frames"). In particular, info->band is always set to 0 (IEEE80211_BAND_2GHZ) since it has not assigned yet. This cause a kernel crash on 5GHz only devices. Move ieee80211_parse_tx_radiotap() after info->band assignment Signed-off-by: Lorenzo Bianconi <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: reuse existing page fragments in A-MSDU rxFelix Fietkau1-5/+77
This massively reduces data copying and thus improves rx performance Signed-off-by: Felix Fietkau <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: fix faulty variable initialization in ieee80211_amsdu_to_8023sFelix Fietkau1-1/+1
reuse_skb is set to true if the code decides to use the last segment. Fixes a memory leak Signed-off-by: Felix Fietkau <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: fix VHT MU-MIMO frame processingSara Sharon1-4/+6
The source and destintation addresses in the memcpy arguments are flipped. Fix that. Fixes: 23a1f8d44c0b("mac80211: process and save VHT MU-MIMO group frame") Signed-off-by: Sara Sharon <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: Unify mesh and mpp path removal functionHenning Rogge1-36/+41
mpp_path_del() and mesh_path_del() are mostly the same function. Move common code into a new static function. Acked-by: Bob Copeland <[email protected]> Signed-off-by: Henning Rogge <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: let unused MPP table entries timeoutHenning Rogge3-1/+56
Remember the last time when a mpp table entry is used for rx or tx and remove them after MESH_PATH_EXPIRE time. Acked-by: Bob Copeland <[email protected]> Signed-off-by: Henning Rogge <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: Remove MPP table entries with MPathHenning Rogge1-0/+26
Make the mesh_path_del() function remove all mpp table entries that are proxied by the removed mesh path. Acked-by: Bob Copeland <[email protected]> Signed-off-by: Henning Rogge <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: basic support for PBSS network typeLior David2-3/+17
PBSS (Personal Basic Service Set) is a new BSS type for DMG networks. It is similar to infrastructure BSS, having an AP-like entity called PCP (PBSS Control Point), but it has few differences. PBSS support is mandatory for 11ad devices. Add support for PBSS by introducing a new PBSS flag attribute. The PBSS flag is used in the START_AP command to request starting a PCP instead of an AP, and in the CONNECT command to request connecting to a PCP instead of an AP. Signed-off-by: Lior David <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: add support for non-linear skbs in ieee80211_amsdu_to_8023sFelix Fietkau1-51/+49
Signed-off-by: Felix Fietkau <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24cfg80211: add function for 802.3 conversion with separate output bufferFelix Fietkau1-53/+53
Use skb_copy_bits in preparation for allowing fragmented skbs Signed-off-by: Felix Fietkau <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: move A-MSDU skb_linearize call to ieee80211_amsdu_to_8023sFelix Fietkau2-3/+3
Prepararation for zero-copy A-MSDU support with page fragment SKBs Signed-off-by: Felix Fietkau <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: fix memory leakSudip Mukherjee1-0/+1
On error we jumped to the error label and returned the error code but we missed releasing sinfo. Fixes: 5fe74014172d ("mac80211: avoid excessive stack usage in sta_info") Reviewed-by: Julian Calaby <[email protected]> Signed-off-by: Sudip Mukherjee <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: Interoperability workaround for 80+80 and 160 MHz channelsJouni Malinen1-2/+29
Number of deployed 80 MHz capable VHT stations that do not support 80+80 and 160 MHz bandwidths seem to misbehave when trying to connect to an AP that advertises 80+80 or 160 MHz channel bandwidth in the VHT Operation element. To avoid such issues with deployed devices, modify the design based on recently accepted IEEE 802.11 standard changes (*). This allows poorly implemented VHT 80 MHz stations to connect with the AP in 80 MHz mode. 80+80 and 160 MHz capable stations need to support the new workaround mechanism to allow full bandwidth to be used. However, there are more or less no impacted station with 80+80/160 capability deployed. The rebased version of this patch is based on the updated version from Johannes Berg to take the HT/VHT chandef refactoring into account. (*) Changes in https://mentor.ieee.org/802.11/dcn/15/11-15-1530-04-000m-vht160-operation-signaling-through-non-zero-ccfs1.docx were accepted during the IEEE 802.11 January 2016 meeting. Signed-off-by: Jouni Malinen <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: add API to allow filtering frames in BA sessionsSara Sharon3-0/+88
If any frames are dropped that are part of a BA session, the reorder buffer will "indefinitely" (until the timeout) wait for them to come in (or a BAR moving the window) and won't release frames after them. This means it isn't possible to filter frames within a BA session in firmware. Introduce an API function that allows such filtering. Calling this function will move the BA window forward to the new SSN, and allows marking frames after the SSN as having been filtered, so any future reordering activity will release frames while skipping the holes. Signed-off-by: Sara Sharon <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: change ieee80211_rx_reorder_ready() argumentsJohannes Berg2-23/+24
Clean up ieee80211_rx_reorder_ready() callers by passing the RX TID struct and the index, instead of the frames list. This will make it more extensible as well. While at it, move the inline to rx.c as it's only used there. Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: expose txq queue depth and size to driversMichal Kazior5-1/+24
This will allow drivers to make more educated decisions whether to defer transmission or not. Relying on wake_tx_queue() call count implicitly was not possible because it could be called without queued frame count actually changing on software tx aggregation start/stop code paths. It was also not possible to know how long byte-wise queue was without dequeueing. Signed-off-by: Michal Kazior <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: Parse legacy and HT rate in injected framesSven Eckelmann1-2/+70
Drivers/devices without their own rate control algorithm can get the information what rates they should use from either the radiotap header of injected frames or from the rate control algorithm. But the parsing of the legacy rate information from the radiotap header was removed in commit e6a9854b05c1 ("mac80211/drivers: rewrite the rate control API"). The removal of this feature heavily reduced the usefulness of frame injection when wanting to simulate specific transmission behavior. Having rate parsing together with MCS rates and retry support allows a fine grained selection of the tx behavior of injected frames for these kind of tests. Signed-off-by: Sven Eckelmann <[email protected]> Cc: Simon Wunderlich <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: mesh: drop constant field mean_chain_lenBob Copeland2-9/+3
The mean_chain_len field in struct mesh_table is copied whenever a new mesh table is allocated, but only ever has the value 2 and is never otherwise updated, so just remove it and use the related define instead. Signed-off-by: Bob Copeland <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: avoid excessive stack usage in sta_infoArnd Bergmann1-8/+16
When CONFIG_OPTIMIZE_INLINING is set, the sta_info_insert_finish function consumes more stack than normally, exceeding the 1024 byte limit on ARM: net/mac80211/sta_info.c: In function 'sta_info_insert_finish': net/mac80211/sta_info.c:561:1: error: the frame size of 1080 bytes is larger than 1024 bytes [-Werror=frame-larger-than=] It turns out that there are two functions that put a 'struct station_info' on the stack: __sta_info_destroy_part2 and sta_info_insert_finish, and this structure alone requires up to 792 bytes. Hoping that both are called rarely enough, this replaces the on-stack structure with a dynamic allocation, which unfortunately requires some suboptimal error handling for out-of-memory. The __sta_info_destroy_part2 function is actually affected by the stack usage twice because it calls cfg80211_del_sta_sinfo(), which has another instance of struct station_info on its stack. Signed-off-by: Arnd Bergmann <[email protected]> Fixes: 98b6218388e3 ("mac80211/cfg80211: add station events") Fixes: 6f7a8d26e266 ("mac80211: send statistics with delete station event") Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: always print a message when disconnectingJohannes Berg1-5/+12
Make sure there's at least a debug message whenever the connection to the AP is terminated. Also change one message from wiphy_debug() to the common mlme_dbg(). Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: fix ibss scan parametersSara Sharon1-7/+15
When joining IBSS a full scan should be initiated in order to search for existing cell, unless the fixed_channel parameter was set. A default channel to create the IBSS on if no cell was found is provided as well. However - a scan is initiated only on the default channel provided regardless of whether ifibss->fixed_channel is set or not, with the obvious result of the cell not joining existing IBSS cell that is on another channel. Fixes: 76bed0f43b27 ("mac80211: IBSS fix scan request") Signed-off-by: Sara Sharon <[email protected]> Signed-off-by: Emmanuel Grumbach <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: add RX_FLAG_MACTIME_PLCP_STARTJohannes Berg2-2/+20
The timestamp given by iwlwifi is at the beginning of the frame over the air, at (or during) the SYNC field. Allow such timestamps to be given to mac80211, at least (for now) for frames with non-HT/VHT preambles. Signed-off-by: Johannes Berg <[email protected]>
2016-02-24mac80211: fix unnecessary frame drops in mesh fwdingMichal Kazior1-2/+3
The ieee80211_queue_stopped() expects hw queue number but it was given raw WMM AC number instead. This could cause frame drops and problems with traffic in some cases - most notably if driver doesn't map AC numbers to queue numbers 1:1 and uses ieee80211_stop_queues() and ieee80211_wake_queue() only without ever calling ieee80211_wake_queues(). On ath10k it was possible to hit this problem in the following case: 1. wlan0 uses queue 0 (ath10k maps queues per vif) 2. offchannel uses queue 15 3. queues 1-14 are unused 4. ieee80211_stop_queues() 5. ieee80211_wake_queue(q=0) 6. ieee80211_wake_queue(q=15) (other queues are not woken up because both driver and mac80211 know other queues are unused) 7. ieee80211_rx_h_mesh_fwding() 8. ieee80211_select_queue_80211() returns 2 9. ieee80211_queue_stopped(q=2) returns true 10. frame is dropped (oops!) Fixes: d3c1597b8d1b ("mac80211: fix forwarded mesh frame queue mapping") Signed-off-by: Michal Kazior <[email protected]> Signed-off-by: Johannes Berg <[email protected]>
Powered by cgit, Gruvboxed by Blaster4385.