- 22 Jun, 2013 29 commits
-
-
Chen Gang authored
For NULL terminated string, need always let it ended by zero. Since have already called memcpy() to initialize 'ci', so need not redundant initialization. Better use ''if(session->hid) {} else if(session->input) {}"" instead of ''if(session->hid) {}; if(session->input) {};'' Signed-off-by:Chen Gang <gang.chen@asianux.com> Reviewed-by:
David Herrmann <dh.herrmann@gmail.com> Acked-by:
Jiri Kosina <jkosina@suse.cz> Signed-off-by:
Gustavo Padovan <gustavo.padovan@collabora.co.uk>
-
Andrei Emeltchenko authored
Remove HCI_LINK_KEYS flag since using HCI_MGMT is enough for test that user space expects the kernel managing link keys. Signed-off-by:
Andrei Emeltchenko <andrei.emeltchenko@intel.com> Acked-by:
Johan Hedberg <johan.hedberg@intel.com> Signed-off-by:
-
Andrei Emeltchenko authored
Use HCI_MGMT flag instead of HCI_LINK_KEYS flag. There is a problem with HCI_LINK_KEYS flag since it is set only when link keys are loaded. Otherwise kernel assumes that old interface is used. Signed-off-by:
-
Andre Guedes authored
We only want to send Mgmt Device Found Events if we are running the Device Discovery procedure (started by the MGMT Start Discovery Command). Inquiry or LE scanning triggered by HCI raw interface (e.g. hcitool) or kernel internals should not send Mgmt Device Found Events. Signed-off-by:
Andre Guedes <andre.guedes@openbossa.org> Acked-by:
-
Andre Guedes authored
This patch removes the hci_cc_le_set_scan_param event handler. This handler became empty because failures of this event are now handled by start_discovery_complete function in mgmt.c. Signed-off-by:
-
Andre Guedes authored
This patch removes hci_do_inquiry and hci_cancel_inquiry helpers. We now use the HCI request framework in device discovery functionality and these helpers are no longer needed. Signed-off-by:
-
Andre Guedes authored
This patch removes the LE scan helpers hci_le_scan and hci_cancel_ le_scan and all code related to it. We now use the HCI request framework in device discovery functionality and these helpers are no longer needed. Signed-off-by:
-
Andre Guedes authored
This patch does a trivial refactoring in hci_cc_le_set_scan_enable. Since start and stop discovery command failures are now handled in mgmt layer, the status check became empty. So, we can move it to outside the switch statement. Signed-off-by:
-
Andre Guedes authored
mgmt_stop_discovery_failed is now only used in mgmt.c so we can make it a local function. This patch also moves the mgmt_stop_ discovery_failed definition up in mgmt.c to avoid forward declaration. Signed-off-by:
-
Andre Guedes authored
Since all mgmt stop discovery command complete events are now handled in stop_discovery_complete callback in mgmt.c, we can remove this handling from hci_event.c. Signed-off-by:
-
Andre Guedes authored
This patch modifies the stop_discovery function so it uses the HCI request framework. The HCI request is built according to the current discovery state (inquiry, LE scanning or name resolving) and a complete callback is register to handle the command complete event for the stop discovery command. This way, we move all stop_discovery mgmt handling code spread in hci_event.c to a single place in mgmt.c. Signed-off-by:
-
Andre Guedes authored
In order to have a better HCI error handling in interleaved discovery functionality, we should use the HCI request framework. This patch updates le_scan_disable_work function so it uses the HCI request framework instead of the hci_send_cmd helper. A complete callback is registered (le_scan_disable_work_complete function) so we are able to trigger the inquiry procedure (if we are running the interleaved discovery) or to stop the discovery procedure (if we are running LE-only discovery). This patch also removes the extra logic in hci_cc_le_set_scan_enable to trigger the inquiry procedure and the mgmt_interleaved_discovery function since they become useless. Signed-off-by:
-
Andre Guedes authored
Some of discovery macros will be used in hci_core so we need to define them in common place such as hci_core.h. Thus, this patch moves discovery macros to hci_core.h and also adds the DISCOV_ prefix to them. Signed-off-by:
-
Andre Guedes authored
mgmt_start_discovery_failed is now only used in mgmt.c so we can make it a local function. This patch also moves the mgmt_start_ discovery_failed definition up in mgmt.c to avoid forward declaration. Signed-off-by:
-
Andre Guedes authored
Since all mgmt start discovery command complete events are now handled in start_discovery_complete callback in mgmt.c, we can remove this handling from hci_event.c. Signed-off-by:
-
Andre Guedes authored
This patch modifies the start_discovery function so it uses the HCI request framework. We build the HCI request according to the discovery type (add inquiry or LE scan HCI commands) and run the HCI request. We also register the start_discovery_complete callback which handles mgmt command complete events for this command. This way, we move all start_ discovery mgmt handling code spread in hci_event.c to a single place in mgmt.c. This patch also merges the LE-only and interleaved discovery type cases since these cases are pretty much the same now. Signed-off-by:
-
Andre Guedes authored
In order to use HCI request framework in start_discovery, we'll need to call inquiry_cache_flush in mgmt.c. Therefore, this patch adds the hci_ prefix to inquiry_cache_flush and makes it non-static. Signed-off-by:
-
Johan Hedberg authored
The LE ATT server socket needs to be superseded by any ATT client sockets. Previously this was done by looking at the hcon->out variable (indicating whether the connection is outgoing or incoming) which is a too crude way of determining whether the server socket needs to be picked or not (an outgoing connection doesn't necessarily mean that an ATT client socket has triggered it). This patch extends the ATT server socket lookup function (l2cap_le_conn_ready) to be used for all LE connections (regardless of the hcon->out value) and adds an internal check into the function for the existence of any ATT client sockets (in which case the server socket should be skipped). For this to work reliably all lookups must be done while the l2cap_conn->chan_lock is held, meaning also that the call to l2cap_chan_add needs to be changed to its lockless __l2cap_chan_add counterpart. Signed-off-by:
Marcel Holtmann <marcel@holtmann.org> Signed-off-by:
-
Johan Hedberg authored
There's no need to reset disc_timeout in l2cap_le_conn_ready since HCI_DISCONN_TIMEOUT is the default when the hci_conn is created and there should be no way for it to get changed between creation and l2cap_le_conn_ready being called. Signed-off-by:
-
Johan Hedberg authored
The L2CAP code has been incrementing the hci_conn reference for each l2cap_chan instance in the l2cap_conn list. Likewise, the reference is dropped each time an l2cap_chan is removed from the list. The reference counting policy with respect to removal has been clear and explicit in the l2cap_chan_del function, however for addition the function calling 2cap_chan_add has always had to do a separate hci_conn_hold call. What made the counting even more hard to follow is that the hci_connect() procedure increments the reference and the L2CAP layer making this call took advantage of it to use it as its own reference. This patch aims to clarify things by having the call to hci_conn_hold inside __l2cap_chan_add, thereby removing the need to do it in the functions calling __l2cap_chan_add. The reference count for hci_connect is still kept as it's necessary for users such as mgmt_pair_device, however for the L2CAP layer it means that an extra call to hci_conn_drop must be performed once l2cap_chan_add has been done. Signed-off-by:
-
Johan Hedberg authored
In l2cap_att_channel() we're only interested in the BT_CONNECTED state so this state can directly be passed to l2cap_global_chan_by_scid(). This way there's no need to do any additional state check later. Signed-off-by:
-
Johan Hedberg authored
The sk variable is of quite little use since it's only used to simplify access in the two bt_sk() calls. Signed-off-by:
-
Johan Hedberg authored
In l2cap_le_conn_ready() after doing l2cap_chann_add() the LE channel is part of the list which is subsequently iterated in l2cap_conn_ready() in this loop each channel will get l2cap_chan_ready() called which would result in trying to set the channel two times into BT_CONNECTED state. Instead it makes sense to just add the channel but not call chan_ready in l2cap_le_conn_ready, which is what this patch does. Signed-off-by:
-
Johan Hedberg authored
There is an extra call to smp_conn_security() for outgoing LE connections from l2cap_conn_ready() but the reason for this call is far from clear. After a bit of commit history research and using git blame I found out that this extra call is for socket-less pairing processes added by commit 160dc6ac. This patch adds a clarifying comment to the code for this. Signed-off-by:
-
Johan Hedberg authored
Since in the future more than the ATT CID may be permissible we should not be hardcoding it for all LE connections in __l2cap_chan_add(). Instead, the source ATT CID should only be set if the destination is also ATT, and in other cases we should just use the existing dynamic CID allocation function. Assigning scid based on dcid means that whenever __l2cap_chan_add() is called that chan->dcid is properly initialized. l2cap_le_conn_ready() wasn't initializing is properly so this is also taken care of in this patch. Signed-off-by:
-
Johan Hedberg authored
The current test in l2cap_chan_connect is intended to protect against multiple conflicting connect attempts. However, it assumes that there will ever only be a single CID that is connected to, which is not true. We do need to check for conflicts with connect attempts to the same destination CID but this check is not in anyway specific to LE but can be applied to BR/EDR as well. Signed-off-by:
-
Johan Hedberg authored
The choice between LE and BR/EDR should be made on the destination address type instead of the destination CID. This is particularly important when in the future more than one CID will be allowed for LE. Signed-off-by:
-
Johan Hedberg authored
In future Core Specification versions the ATT CID will be just one of many possible CIDs that can be used for data transfer. Therefore, it makes sense to rename the define for the ATT CID to something less ambigous. Signed-off-by:
-
Johan Hedberg authored
The LE L2CAP signalling channel follows its own rules and will continue to evolve independently from the BR/EDR signalling channel. Therefore, it makes sense to have a clear split from BR/EDR by having a dedicated function for handling LE signalling commands. Signed-off-by:
-
- 19 Jun, 2013 5 commits
-
-
Amitkumar Karwar authored
After receiving channel switch announcement from AP, scan and association on that channel is blocked for DFS_CHAN_MOVE_TIME (10 seconds). Hence station will be able to connect to the AP, once it is moved to new channel. Signed-off-by:
Amitkumar Karwar <akarwar@marvell.com> Signed-off-by:
Bing Zhao <bzhao@marvell.com> Signed-off-by:
Paul Stewart <pstew@chromium.org> Signed-off-by:
John W. Linville <linville@tuxdriver.com>
-
Amitkumar Karwar authored
This patch adds code to parse requested AP's 11h capabilities and add 11h information in association request. Also, deauth is sent to the AP after receiving channel switch announcement event from firmware. This happens when AP advertises WLAN_EID_CHANNEL_SWITCH IE in it's beacon. Signed-off-by:
-
Antonio Quartulli authored
Do not unconditionally access the chan variable in brcmf_cfg80211_mgmt_tx() as it may be NULL. Use freq instead. Introduced by c2ff8cad ("brcm80211: make mgmt_tx in brcmfmac accept a NULL channel") Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
Antonio Quartulli <antonio@open-mesh.com> Signed-off-by:
-
Amitkumar Karwar authored
Use negative check (if(!bss_desc)) and return failure instead of failing a NULL check later in mwifiex_check_network_compatibility() routine. Signed-off-by:
-
Larry Finger authored
A typo causes routine rtl92cu_phy_rf6052_set_cck_txpower() to test the same condition twice. The problem was found using cppcheck-1.49, and the proper fix was verified against the pre-mac80211 version of the code. This patch was originally included as commit 1288aa4e, but was accidentally reverted in a later patch. Reported-by: David Binderman <dcb314@hotmail.com> [original report] Reported-by: Andrea Morello <andrea.merello@gmail.com> [report of accidental reversion] Signed-off-by:
Larry Finger <Larry.Finger@lwfinger.net> Cc: stable <stable@vger.kernel.org> [back to 2.6.39] Signed-off-by:
-
- 18 Jun, 2013 6 commits
-
-
Franky Lin authored
Add scatter gather list support for better rx glom performance. Reviewed-by:
Hante Meuleman <meuleman@broadcom.com> Reviewed-by:
Pieter-Paul Giesberts <pieterpg@broadcom.com> Reviewed-by:
Arend van Spriel <arend@broadcom.com> Signed-off-by:
Franky Lin <frankyl@broadcom.com> Signed-off-by:
-
Franky Lin authored
Introducing a new SDIO block data access interface function brcmf_sdio_buffrw. It will act as a unified interface function for any block data access to WiFi dongle through SDIO interface. This patch enables the support for single skb transmission. Reviewed-by:
-
Franky Lin authored
There is an ID table registered to SDIO stack which consists of all SDIO devices supported by brcmfmac. It is not necessary to have an extra check. Reviewed-by:
-
Arend van Spriel authored
Instead of allocating an empty list item and queue that for the dpc data worker to dequeue an atomic counter is used. Reviewed-by:
-
Arend van Spriel authored
Just cleaning up and being consistent in naming operations related to struct brcmf_fws_mac_descriptor objects. Reviewed-by:
-
Arend van Spriel authored
The firmware-signalling code used NL80211_NUM_ACS, but it has its own definition enum brcmf_fws_fifo, which is more appropriate to use. This effectively removes the need to include nl80211.h. Reviewed-by:
-
