- 29 Aug, 2009 11 commits
-
-
Stephen Hemminger authored
Signed-off-by:
Stephen Hemminger <shemminger@vyatta.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Stephen Hemminger authored
Don't force inlining where not needed. Gcc does better job of deciding to inline local functions. Signed-off-by:
-
Stephen Hemminger authored
A couple of minor functions can be written more compactly. Signed-off-by:
-
Ron Mercer authored
TX completions were running in a workqueue queued by the ISR. This patch moves the processing of TX completions to an existing RSS NAPI context. Now each irq vector runs NAPI for one RSS ring and one or more TX completion rings. Signed-off-by:
Ron Mercer <ron.mercer@qlogic.com> Signed-off-by:
-
Ron Mercer authored
Currently we downshift to MSI/Legacy if we don't get enough vectors for cpu_count RSS rings plus cpu_count TX completion rings. This patch allows running MSIX with the vector count that the platform provides. Signed-off-by:
-
Ron Mercer authored
Currently we have three types of RX rings. 1) Default ring - services rx_ring for broadcast/multicast, handles firmware events, and errors. 2) TX completion ring - handles only outbound completions. 3) RSS ring - handles only inbound completions. This patch gets rid of the default ring type and moves it's functionality into the first RSS ring. This makes better use of MSIX vectors since they are a limited resource on some platforms. Signed-off-by:
-
-
Petri Gynther authored
bonding: Have bond_check_dev_link examine netif_running Some network devices do not call netif_carrier_off when they are set administratively down. Have the bonding link check function also inspect the netif_running state. Ignore netif_running if the bond_check_dev_link function is called with "reporting" set, as in that case it's inspecting the capabilities of the non-netif_carrier device driver. Signed-off-by:
Petri Gynther <pgynther@google.com> Signed-off-by:
Jay Vosburgh <fubar@us.ibm.com> Signed-off-by:
-
Nicolas de Pesloüan authored
max_bonds is of type int and cannot be greater than INT_MAX. Signed-off-by:
Nicolas de Pesloüan <nicolas.2p.debian@free.fr> Signed-off-by:
-
Stephen Hemminger authored
Bonding can use compare_ether_addr() in bond_release. Signed-off-by:
-
Jay Vosburgh authored
Propogate the vlan_features of the slave devices to the bonding master device, using the same logic as for regular features. Tested by Or Gerlitz <ogerlitz@voltaire.com>, who also removed the debug logic from the original test patch. Signed-off-by:
Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by:
-
- 28 Aug, 2009 29 commits
-
-
Kiran Divekar authored
Most of the places in debugfs.c are missing a NULL check on the return value of get_zeroed_page API call. Added required NULL check at appropriate places. Signed-off-by:
Kiran Divekar <kirandivekar@gmail.com> Signed-off-by:
John W. Linville <linville@tuxdriver.com>
-
Jussi Kivilinna authored
Now that cfg80211 functions are added and wext converted to use wext-compat functions, remove wext structures and disabled code. Signed-off-by:
Jussi Kivilinna <jussi.kivilinna@mbnet.fi> Signed-off-by:
-
Jussi Kivilinna authored
Signed-off-by:
-
Jussi Kivilinna authored
Signed-off-by:
-
Jussi Kivilinna authored
Signed-off-by:
-
Jussi Kivilinna authored
Stats worker no longer poll stats from device anymore. It's still needed to poll device control channel for connect/disconnect events, so rename stats worker as device poller. Signed-off-by:
-
Jussi Kivilinna authored
Signed-off-by:
-
Jussi Kivilinna authored
Add cfg80211 get_station and convert SIOCGIWRATE and get_wireless_stats to cfg80211. Signed-off-by:
-
Jussi Kivilinna authored
Add cfg80211 add_key/del_key/set_default_key and convert wext to use theim. Signed-off-by:
-
Jussi Kivilinna authored
Add cfg80211 set_channel and convert wext to use it. Signed-off-by:
-
Jussi Kivilinna authored
Add cfg80211 connect functions for station and ad-hoc modes and convert wext to use theim. Signed-off-by:
-
Jussi Kivilinna authored
Random essid must be set to turn on radio when not connected. If device is in ad-hoc mode, this results 'media connect' indications with the random essid which should be ignored. Signed-off-by:
-
Jussi Kivilinna authored
Signed-off-by:
-
Jussi Kivilinna authored
Use is_zero_ether_addr() and is_broadcast_ether_addr() instead of memcmp against ffff_bssid/zero_bssid. Signed-off-by:
-
Jussi Kivilinna authored
Move link up/down work to separate functions and use local array for allocating memory for info structure instead of kzmalloc. Signed-off-by:
-
Jussi Kivilinna authored
Increase scan delay from 1 sec to 6 sec. Spec says that scan by OID_802_11_BSSID_LIST_SCAN completes in 6 seconds. Before rfkill patch too short delay was not problem as device was always active (radio on) and performing background scanning. Signed-off-by:
-
Jussi Kivilinna authored
Buffer used for bssid list might be too small. Change rndis_query_oid() to return required buffer length to caller and make rndis_check_bssid_list() resize buffer when needed. Signed-off-by:
-
Jussi Kivilinna authored
OID_802_11_BSSID_LIST_SCAN clears device's bssid list, so retrieve current bssid list from device before issuing new scan. Signed-off-by:
-
Jussi Kivilinna authored
Setting WPA keys with OID_802_11_ADD_KEY sometimes trigger instant media connect indication. These indications are extranous and should be ignored, as otherwise driver would send reassociation event to userspace which in this case is not needed. Signed-off-by:
-
Gábor Stefanik authored
The A/G-PHY changes are fallout fixes from the enum change, which in turn allows the LP-PHY code to be much simpler. The antenna_to_phyctl change is a fix for a potential existing bug that this patch may otherwise trigger. Signed-off-by:
Gábor Stefanik <netrolller.3d@gmail.com> Signed-off-by:
-
Gábor Stefanik authored
The most common LP-PHY device, BCM4312, is now fully functional. So, no need to say "probably won't work for you" anymore. It's also not "for debuggers and developers only", as it is perfectly usable for end-users now (at least for BCM4312). Signed-off-by:
-
Pavel Roskin authored
AR5K_SREV is available even if the chip has been put to sleep. Relying on the chip register allows binding non-standard PCI IDs by echo VENDOR_ID PRODUCT_ID >/sys/bus/pci/drivers/ath5k/new_id without having to specify the driver data as well. Signed-off-by:
Pavel Roskin <proski@gnu.org> Acked-by:
Bob Copeland <me@bobcopeland.com> Acked-by:
Nick Kossifidis <mickflemm@gmail.com> Signed-off-by:
-
Pavel Roskin authored
The `val' variable in ath5k_eeprom_read_turbo_modes() is used uninitialized. gcc 4.4.1 with -fno-inline-functions-called-once reports it: eeprom.c: In function 'ath5k_eeprom_read_turbo_modes': eeprom.c:441: warning: 'val' may be used uninitialized in this function Comparing the code to the Atheros HAL, it's clear that the split between ath5k_eeprom_read_modes() and ath5k_eeprom_read_turbo_modes() was incorrect. The Atheros HAL reads both turbo and non-turbo data from EEPROM in one function. Some turbo mode parameters are derived from the same EEPROM values as non-turbo parameters, just from different bits. Merge ath5k_eeprom_read_turbo_modes() into ath5k_eeprom_read_modes() to fix the warning. The actual values and offsets have been cross-checked against Atheros HAL. Signed-off-by:
-
Bob Copeland authored
As Pavel Roskin noted, the check for mac version as copied from legacy_hal made no sense. This replaces it with the equivalent and makes up a suitable #define for the mac version legacy_hal checked. Signed-off-by:
-
Gábor Stefanik authored
After some discussion on IRC about the PHY register write change, I am not sure anymore if this is the right thing to do. Signed-off-by:
-
Johannes Berg authored
The ICT IRQ table is a set of __le32 values, not u32 values, so when reading it we need to take into account that it has to be converted to CPU endianness. This was causing a lot of trouble on my powerpc box where various things would simply not work for no apparent reason with 5xxx cards, but worked with 4965 -- which doesn't use the ICT table. Signed-off-by:
Johannes Berg <johannes@sipsolutions.net> Acked-by:
Reinette Chatre <reinette.chatre@intel.com> Signed-off-by:
-
Ivo van Doorn authored
Since patch "rt2x00: bss_info_changed() callback is allowed to sleep" the variable delayed wasn't used anymore. This means it can be removed along with the call to schedule_work which depended on that variable. Signed-off-by:
Ivo van Doorn <IvDoorn@gmail.com> Signed-off-by:
-
Gábor Stefanik authored
Implement the "PMU LDO set voltage" and "PMU LDO PA ref enable" functions, and use them during LP-PHY baseband init in b43. Signed-off-by:
-
Gábor Stefanik authored
-Fix a few nasty typos (b43_phy_* operations instead of b43_radio_*) in the channel tune routines. -Fix some typos & spec errors found by MMIO tracing. -Optimize b43_phy_write & b43_phy_mask/set/maskset to use only the minimal number of MMIO accesses. (Write is possible using a single 32-bit MMIO write, while set/mask/maskset can be done in 3 16-bit MMIOs). -Set the default channel back to 1, as the bug forcing us to use channel 7 is now fixed. With this, the device comes up, scans, associates, transmits, receives, monitors and injects on all channels - in other words, it's fully functional. Sensitivity and TX power are still sub-optimal, due to the lack of calibration (that's next on my list). Signed-off-by:
-
