freebsd-dev/sys/dev/wpi/if_wpi.c

5596 lines
145 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2006,2007
* Damien Bergamini <damien.bergamini@free.fr>
* Benjamin Close <Benjamin.Close@clearchain.com>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
/*
* Driver for Intel PRO/Wireless 3945ABG 802.11 network adapters.
*
* The 3945ABG network adapter doesn't use traditional hardware as
* many other adaptors do. Instead at run time the eeprom is set into a known
* state and told to load boot firmware. The boot firmware loads an init and a
* main binary firmware image into SRAM on the card via DMA.
* Once the firmware is loaded, the driver/hw then
* communicate by way of circular dma rings via the SRAM to the firmware.
*
* There is 6 memory rings. 1 command ring, 1 rx data ring & 4 tx data rings.
* The 4 tx data rings allow for prioritization QoS.
*
* The rx data ring consists of 32 dma buffers. Two registers are used to
* indicate where in the ring the driver and the firmware are up to. The
* driver sets the initial read index (reg1) and the initial write index (reg2),
* the firmware updates the read index (reg1) on rx of a packet and fires an
* interrupt. The driver then processes the buffers starting at reg1 indicating
* to the firmware which buffers have been accessed by updating reg2. At the
* same time allocating new memory for the processed buffer.
*
* A similar thing happens with the tx rings. The difference is the firmware
* stop processing buffers once the queue is full and until confirmation
* of a successful transmition (tx_done) has occurred.
*
* The command ring operates in the same manner as the tx queues.
*
* All communication direct to the card (ie eeprom) is classed as Stage1
* communication
*
* All communication via the firmware to the card is classed as State2.
* The firmware consists of 2 parts. A bootstrap firmware and a runtime
* firmware. The bootstrap firmware and runtime firmware are loaded
* from host memory via dma to the card then told to execute. From this point
* on the majority of communications between the driver and the card goes
* via the firmware.
*/
#include "opt_wlan.h"
#include "opt_wpi.h"
#include <sys/param.h>
#include <sys/sysctl.h>
#include <sys/sockio.h>
#include <sys/mbuf.h>
#include <sys/kernel.h>
#include <sys/socket.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/queue.h>
#include <sys/taskqueue.h>
#include <sys/module.h>
#include <sys/bus.h>
#include <sys/endian.h>
#include <sys/linker.h>
#include <sys/firmware.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/rman.h>
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
#include <net/bpf.h>
#include <net/if.h>
#include <net/if_var.h>
#include <net/if_arp.h>
#include <net/ethernet.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/in_var.h>
#include <netinet/if_ether.h>
#include <netinet/ip.h>
#include <net80211/ieee80211_var.h>
#include <net80211/ieee80211_radiotap.h>
#include <net80211/ieee80211_regdomain.h>
#include <net80211/ieee80211_ratectl.h>
#include <dev/wpi/if_wpireg.h>
#include <dev/wpi/if_wpivar.h>
#include <dev/wpi/if_wpi_debug.h>
struct wpi_ident {
uint16_t vendor;
uint16_t device;
uint16_t subdevice;
const char *name;
};
static const struct wpi_ident wpi_ident_table[] = {
/* The below entries support ABG regardless of the subid */
{ 0x8086, 0x4222, 0x0, "Intel(R) PRO/Wireless 3945ABG" },
{ 0x8086, 0x4227, 0x0, "Intel(R) PRO/Wireless 3945ABG" },
/* The below entries only support BG */
{ 0x8086, 0x4222, 0x1005, "Intel(R) PRO/Wireless 3945BG" },
{ 0x8086, 0x4222, 0x1034, "Intel(R) PRO/Wireless 3945BG" },
{ 0x8086, 0x4227, 0x1014, "Intel(R) PRO/Wireless 3945BG" },
{ 0x8086, 0x4222, 0x1044, "Intel(R) PRO/Wireless 3945BG" },
{ 0, 0, 0, NULL }
};
static int wpi_probe(device_t);
static int wpi_attach(device_t);
static void wpi_radiotap_attach(struct wpi_softc *);
static void wpi_sysctlattach(struct wpi_softc *);
static void wpi_init_beacon(struct wpi_vap *);
static struct ieee80211vap *wpi_vap_create(struct ieee80211com *,
const char [IFNAMSIZ], int, enum ieee80211_opmode, int,
const uint8_t [IEEE80211_ADDR_LEN],
const uint8_t [IEEE80211_ADDR_LEN]);
static void wpi_vap_delete(struct ieee80211vap *);
static int wpi_detach(device_t);
static int wpi_shutdown(device_t);
static int wpi_suspend(device_t);
static int wpi_resume(device_t);
static int wpi_nic_lock(struct wpi_softc *);
static int wpi_read_prom_data(struct wpi_softc *, uint32_t, void *, int);
static void wpi_dma_map_addr(void *, bus_dma_segment_t *, int, int);
static int wpi_dma_contig_alloc(struct wpi_softc *, struct wpi_dma_info *,
void **, bus_size_t, bus_size_t);
static void wpi_dma_contig_free(struct wpi_dma_info *);
static int wpi_alloc_shared(struct wpi_softc *);
static void wpi_free_shared(struct wpi_softc *);
static int wpi_alloc_fwmem(struct wpi_softc *);
static void wpi_free_fwmem(struct wpi_softc *);
static int wpi_alloc_rx_ring(struct wpi_softc *);
static void wpi_update_rx_ring(struct wpi_softc *);
static void wpi_update_rx_ring_ps(struct wpi_softc *);
static void wpi_reset_rx_ring(struct wpi_softc *);
static void wpi_free_rx_ring(struct wpi_softc *);
static int wpi_alloc_tx_ring(struct wpi_softc *, struct wpi_tx_ring *,
uint8_t);
static void wpi_update_tx_ring(struct wpi_softc *, struct wpi_tx_ring *);
static void wpi_update_tx_ring_ps(struct wpi_softc *,
struct wpi_tx_ring *);
static void wpi_reset_tx_ring(struct wpi_softc *, struct wpi_tx_ring *);
static void wpi_free_tx_ring(struct wpi_softc *, struct wpi_tx_ring *);
static int wpi_read_eeprom(struct wpi_softc *,
uint8_t macaddr[IEEE80211_ADDR_LEN]);
static uint32_t wpi_eeprom_channel_flags(struct wpi_eeprom_chan *);
static void wpi_read_eeprom_band(struct wpi_softc *, uint8_t);
static int wpi_read_eeprom_channels(struct wpi_softc *, uint8_t);
static struct wpi_eeprom_chan *wpi_find_eeprom_channel(struct wpi_softc *,
struct ieee80211_channel *);
static int wpi_setregdomain(struct ieee80211com *,
struct ieee80211_regdomain *, int,
struct ieee80211_channel[]);
static int wpi_read_eeprom_group(struct wpi_softc *, uint8_t);
static struct ieee80211_node *wpi_node_alloc(struct ieee80211vap *,
const uint8_t mac[IEEE80211_ADDR_LEN]);
static void wpi_node_free(struct ieee80211_node *);
static void wpi_recv_mgmt(struct ieee80211_node *, struct mbuf *, int,
const struct ieee80211_rx_stats *,
int, int);
static void wpi_restore_node(void *, struct ieee80211_node *);
static void wpi_restore_node_table(struct wpi_softc *, struct wpi_vap *);
static int wpi_newstate(struct ieee80211vap *, enum ieee80211_state, int);
static void wpi_calib_timeout(void *);
static void wpi_rx_done(struct wpi_softc *, struct wpi_rx_desc *,
struct wpi_rx_data *);
static void wpi_rx_statistics(struct wpi_softc *, struct wpi_rx_desc *,
struct wpi_rx_data *);
static void wpi_tx_done(struct wpi_softc *, struct wpi_rx_desc *);
static void wpi_cmd_done(struct wpi_softc *, struct wpi_rx_desc *);
static void wpi_notif_intr(struct wpi_softc *);
static void wpi_wakeup_intr(struct wpi_softc *);
#ifdef WPI_DEBUG
static void wpi_debug_registers(struct wpi_softc *);
#endif
static void wpi_fatal_intr(struct wpi_softc *);
static void wpi_intr(void *);
static int wpi_cmd2(struct wpi_softc *, struct wpi_buf *);
static int wpi_tx_data(struct wpi_softc *, struct mbuf *,
struct ieee80211_node *);
static int wpi_tx_data_raw(struct wpi_softc *, struct mbuf *,
struct ieee80211_node *,
const struct ieee80211_bpf_params *);
static int wpi_raw_xmit(struct ieee80211_node *, struct mbuf *,
const struct ieee80211_bpf_params *);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static int wpi_transmit(struct ieee80211com *, struct mbuf *);
static void wpi_watchdog_rfkill(void *);
static void wpi_scan_timeout(void *);
static void wpi_tx_timeout(void *);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static void wpi_parent(struct ieee80211com *);
static int wpi_cmd(struct wpi_softc *, uint8_t, const void *, uint16_t,
int);
static int wpi_mrr_setup(struct wpi_softc *);
static int wpi_add_node(struct wpi_softc *, struct ieee80211_node *);
static int wpi_add_broadcast_node(struct wpi_softc *, int);
static int wpi_add_ibss_node(struct wpi_softc *, struct ieee80211_node *);
static void wpi_del_node(struct wpi_softc *, struct ieee80211_node *);
static int wpi_updateedca(struct ieee80211com *);
static void wpi_set_promisc(struct wpi_softc *);
static void wpi_update_promisc(struct ieee80211com *);
static void wpi_update_mcast(struct ieee80211com *);
static void wpi_set_led(struct wpi_softc *, uint8_t, uint8_t, uint8_t);
static int wpi_set_timing(struct wpi_softc *, struct ieee80211_node *);
static void wpi_power_calibration(struct wpi_softc *);
static int wpi_set_txpower(struct wpi_softc *, int);
static int wpi_get_power_index(struct wpi_softc *,
struct wpi_power_group *, uint8_t, int, int);
static int wpi_set_pslevel(struct wpi_softc *, uint8_t, int, int);
static int wpi_send_btcoex(struct wpi_softc *);
static int wpi_send_rxon(struct wpi_softc *, int, int);
static int wpi_config(struct wpi_softc *);
static uint16_t wpi_get_active_dwell_time(struct wpi_softc *,
struct ieee80211_channel *, uint8_t);
static uint16_t wpi_limit_dwell(struct wpi_softc *, uint16_t);
static uint16_t wpi_get_passive_dwell_time(struct wpi_softc *,
struct ieee80211_channel *);
static uint32_t wpi_get_scan_pause_time(uint32_t, uint16_t);
static int wpi_scan(struct wpi_softc *, struct ieee80211_channel *);
static int wpi_auth(struct wpi_softc *, struct ieee80211vap *);
static int wpi_config_beacon(struct wpi_vap *);
static int wpi_setup_beacon(struct wpi_softc *, struct ieee80211_node *);
static void wpi_update_beacon(struct ieee80211vap *, int);
static void wpi_newassoc(struct ieee80211_node *, int);
static int wpi_run(struct wpi_softc *, struct ieee80211vap *);
static int wpi_load_key(struct ieee80211_node *,
const struct ieee80211_key *);
static void wpi_load_key_cb(void *, struct ieee80211_node *);
static int wpi_set_global_keys(struct ieee80211_node *);
static int wpi_del_key(struct ieee80211_node *,
const struct ieee80211_key *);
static void wpi_del_key_cb(void *, struct ieee80211_node *);
static int wpi_process_key(struct ieee80211vap *,
const struct ieee80211_key *, int);
static int wpi_key_set(struct ieee80211vap *,
const struct ieee80211_key *);
static int wpi_key_delete(struct ieee80211vap *,
const struct ieee80211_key *);
static int wpi_post_alive(struct wpi_softc *);
static int wpi_load_bootcode(struct wpi_softc *, const uint8_t *,
uint32_t);
static int wpi_load_firmware(struct wpi_softc *);
static int wpi_read_firmware(struct wpi_softc *);
static void wpi_unload_firmware(struct wpi_softc *);
static int wpi_clock_wait(struct wpi_softc *);
static int wpi_apm_init(struct wpi_softc *);
static void wpi_apm_stop_master(struct wpi_softc *);
static void wpi_apm_stop(struct wpi_softc *);
static void wpi_nic_config(struct wpi_softc *);
static int wpi_hw_init(struct wpi_softc *);
static void wpi_hw_stop(struct wpi_softc *);
static void wpi_radio_on(void *, int);
static void wpi_radio_off(void *, int);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static int wpi_init(struct wpi_softc *);
static void wpi_stop_locked(struct wpi_softc *);
static void wpi_stop(struct wpi_softc *);
static void wpi_scan_start(struct ieee80211com *);
static void wpi_scan_end(struct ieee80211com *);
static void wpi_set_channel(struct ieee80211com *);
static void wpi_scan_curchan(struct ieee80211_scan_state *, unsigned long);
static void wpi_scan_mindwell(struct ieee80211_scan_state *);
static void wpi_hw_reset(void *, int);
static device_method_t wpi_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, wpi_probe),
DEVMETHOD(device_attach, wpi_attach),
DEVMETHOD(device_detach, wpi_detach),
DEVMETHOD(device_shutdown, wpi_shutdown),
DEVMETHOD(device_suspend, wpi_suspend),
DEVMETHOD(device_resume, wpi_resume),
DEVMETHOD_END
};
static driver_t wpi_driver = {
"wpi",
wpi_methods,
sizeof (struct wpi_softc)
};
static devclass_t wpi_devclass;
DRIVER_MODULE(wpi, pci, wpi_driver, wpi_devclass, NULL, NULL);
MODULE_VERSION(wpi, 1);
MODULE_DEPEND(wpi, pci, 1, 1, 1);
MODULE_DEPEND(wpi, wlan, 1, 1, 1);
MODULE_DEPEND(wpi, firmware, 1, 1, 1);
static int
wpi_probe(device_t dev)
{
const struct wpi_ident *ident;
for (ident = wpi_ident_table; ident->name != NULL; ident++) {
if (pci_get_vendor(dev) == ident->vendor &&
pci_get_device(dev) == ident->device) {
device_set_desc(dev, ident->name);
return (BUS_PROBE_DEFAULT);
}
}
return ENXIO;
}
static int
wpi_attach(device_t dev)
{
struct wpi_softc *sc = (struct wpi_softc *)device_get_softc(dev);
struct ieee80211com *ic;
uint8_t i;
int error, rid;
#ifdef WPI_DEBUG
int supportsa = 1;
const struct wpi_ident *ident;
#endif
sc->sc_dev = dev;
#ifdef WPI_DEBUG
error = resource_int_value(device_get_name(sc->sc_dev),
device_get_unit(sc->sc_dev), "debug", &(sc->sc_debug));
if (error != 0)
sc->sc_debug = 0;
#else
sc->sc_debug = 0;
#endif
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/*
* Get the offset of the PCI Express Capability Structure in PCI
* Configuration Space.
*/
error = pci_find_cap(dev, PCIY_EXPRESS, &sc->sc_cap_off);
if (error != 0) {
device_printf(dev, "PCIe capability structure not found!\n");
return error;
}
/*
* Some card's only support 802.11b/g not a, check to see if
* this is one such card. A 0x0 in the subdevice table indicates
* the entire subdevice range is to be ignored.
*/
#ifdef WPI_DEBUG
for (ident = wpi_ident_table; ident->name != NULL; ident++) {
if (ident->subdevice &&
pci_get_subdevice(dev) == ident->subdevice) {
supportsa = 0;
break;
}
}
#endif
/* Clear device-specific "PCI retry timeout" register (41h). */
pci_write_config(dev, 0x41, 0, 1);
/* Enable bus-mastering. */
pci_enable_busmaster(dev);
rid = PCIR_BAR(0);
sc->mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
RF_ACTIVE);
if (sc->mem == NULL) {
device_printf(dev, "can't map mem space\n");
return ENOMEM;
}
sc->sc_st = rman_get_bustag(sc->mem);
sc->sc_sh = rman_get_bushandle(sc->mem);
rid = 1;
if (pci_alloc_msi(dev, &rid) == 0)
rid = 1;
else
rid = 0;
/* Install interrupt handler. */
sc->irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid, RF_ACTIVE |
(rid != 0 ? 0 : RF_SHAREABLE));
if (sc->irq == NULL) {
device_printf(dev, "can't map interrupt\n");
error = ENOMEM;
goto fail;
}
WPI_LOCK_INIT(sc);
WPI_TX_LOCK_INIT(sc);
WPI_RXON_LOCK_INIT(sc);
WPI_NT_LOCK_INIT(sc);
WPI_TXQ_LOCK_INIT(sc);
WPI_TXQ_STATE_LOCK_INIT(sc);
/* Allocate DMA memory for firmware transfers. */
if ((error = wpi_alloc_fwmem(sc)) != 0) {
device_printf(dev,
"could not allocate memory for firmware, error %d\n",
error);
goto fail;
}
/* Allocate shared page. */
if ((error = wpi_alloc_shared(sc)) != 0) {
device_printf(dev, "could not allocate shared page\n");
goto fail;
}
/* Allocate TX rings - 4 for QoS purposes, 1 for commands. */
for (i = 0; i < WPI_NTXQUEUES; i++) {
if ((error = wpi_alloc_tx_ring(sc, &sc->txq[i], i)) != 0) {
device_printf(dev,
"could not allocate TX ring %d, error %d\n", i,
error);
goto fail;
}
}
/* Allocate RX ring. */
if ((error = wpi_alloc_rx_ring(sc)) != 0) {
device_printf(dev, "could not allocate RX ring, error %d\n",
error);
goto fail;
}
/* Clear pending interrupts. */
WPI_WRITE(sc, WPI_INT, 0xffffffff);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ic = &sc->sc_ic;
ic->ic_softc = sc;
ic->ic_name = device_get_nameunit(dev);
ic->ic_phytype = IEEE80211_T_OFDM; /* not only, but not used */
ic->ic_opmode = IEEE80211_M_STA; /* default to BSS mode */
/* Set device capabilities. */
ic->ic_caps =
IEEE80211_C_STA /* station mode supported */
| IEEE80211_C_IBSS /* IBSS mode supported */
| IEEE80211_C_HOSTAP /* Host access point mode */
| IEEE80211_C_MONITOR /* monitor mode supported */
| IEEE80211_C_AHDEMO /* adhoc demo mode */
| IEEE80211_C_BGSCAN /* capable of bg scanning */
| IEEE80211_C_TXPMGT /* tx power management */
| IEEE80211_C_SHSLOT /* short slot time supported */
| IEEE80211_C_WPA /* 802.11i */
| IEEE80211_C_SHPREAMBLE /* short preamble supported */
| IEEE80211_C_WME /* 802.11e */
| IEEE80211_C_PMGT /* Station-side power mgmt */
;
ic->ic_cryptocaps =
IEEE80211_CRYPTO_AES_CCM;
/*
* Read in the eeprom and also setup the channels for
* net80211. We don't set the rates as net80211 does this for us
*/
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if ((error = wpi_read_eeprom(sc, ic->ic_macaddr)) != 0) {
device_printf(dev, "could not read EEPROM, error %d\n",
error);
goto fail;
}
#ifdef WPI_DEBUG
if (bootverbose) {
device_printf(sc->sc_dev, "Regulatory Domain: %.4s\n",
sc->domain);
device_printf(sc->sc_dev, "Hardware Type: %c\n",
sc->type > 1 ? 'B': '?');
device_printf(sc->sc_dev, "Hardware Revision: %c\n",
((sc->rev & 0xf0) == 0xd0) ? 'D': '?');
device_printf(sc->sc_dev, "SKU %s support 802.11a\n",
supportsa ? "does" : "does not");
/* XXX hw_config uses the PCIDEV for the Hardware rev. Must
check what sc->rev really represents - benjsc 20070615 */
}
#endif
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ieee80211_ifattach(ic);
ic->ic_vap_create = wpi_vap_create;
ic->ic_vap_delete = wpi_vap_delete;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ic->ic_parent = wpi_parent;
ic->ic_raw_xmit = wpi_raw_xmit;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ic->ic_transmit = wpi_transmit;
ic->ic_node_alloc = wpi_node_alloc;
sc->sc_node_free = ic->ic_node_free;
ic->ic_node_free = wpi_node_free;
ic->ic_wme.wme_update = wpi_updateedca;
ic->ic_update_promisc = wpi_update_promisc;
ic->ic_update_mcast = wpi_update_mcast;
ic->ic_newassoc = wpi_newassoc;
ic->ic_scan_start = wpi_scan_start;
ic->ic_scan_end = wpi_scan_end;
ic->ic_set_channel = wpi_set_channel;
ic->ic_scan_curchan = wpi_scan_curchan;
ic->ic_scan_mindwell = wpi_scan_mindwell;
ic->ic_setregdomain = wpi_setregdomain;
sc->sc_update_rx_ring = wpi_update_rx_ring;
sc->sc_update_tx_ring = wpi_update_tx_ring;
wpi_radiotap_attach(sc);
callout_init_mtx(&sc->calib_to, &sc->rxon_mtx, 0);
callout_init_mtx(&sc->scan_timeout, &sc->rxon_mtx, 0);
callout_init_mtx(&sc->tx_timeout, &sc->txq_state_mtx, 0);
callout_init_mtx(&sc->watchdog_rfkill, &sc->sc_mtx, 0);
TASK_INIT(&sc->sc_reinittask, 0, wpi_hw_reset, sc);
TASK_INIT(&sc->sc_radiooff_task, 0, wpi_radio_off, sc);
TASK_INIT(&sc->sc_radioon_task, 0, wpi_radio_on, sc);
sc->sc_tq = taskqueue_create("wpi_taskq", M_WAITOK,
taskqueue_thread_enqueue, &sc->sc_tq);
error = taskqueue_start_threads(&sc->sc_tq, 1, 0, "wpi_taskq");
if (error != 0) {
device_printf(dev, "can't start threads, error %d\n", error);
goto fail;
}
wpi_sysctlattach(sc);
/*
* Hook our interrupt after all initialization is complete.
*/
error = bus_setup_intr(dev, sc->irq, INTR_TYPE_NET | INTR_MPSAFE,
NULL, wpi_intr, sc, &sc->sc_ih);
if (error != 0) {
device_printf(dev, "can't establish interrupt, error %d\n",
error);
goto fail;
}
if (bootverbose)
ieee80211_announce(ic);
#ifdef WPI_DEBUG
if (sc->sc_debug & WPI_DEBUG_HW)
ieee80211_announce_channels(ic);
#endif
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
fail: wpi_detach(dev);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
/*
* Attach the interface to 802.11 radiotap.
*/
static void
wpi_radiotap_attach(struct wpi_softc *sc)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct wpi_rx_radiotap_header *rxtap = &sc->sc_rxtap;
struct wpi_tx_radiotap_header *txtap = &sc->sc_txtap;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ieee80211_radiotap_attach(&sc->sc_ic,
&txtap->wt_ihdr, sizeof(*txtap), WPI_TX_RADIOTAP_PRESENT,
&rxtap->wr_ihdr, sizeof(*rxtap), WPI_RX_RADIOTAP_PRESENT);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
}
static void
wpi_sysctlattach(struct wpi_softc *sc)
{
#ifdef WPI_DEBUG
struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(sc->sc_dev);
struct sysctl_oid *tree = device_get_sysctl_tree(sc->sc_dev);
SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
"debug", CTLFLAG_RW, &sc->sc_debug, sc->sc_debug,
"control debugging printfs");
#endif
}
static void
wpi_init_beacon(struct wpi_vap *wvp)
{
struct wpi_buf *bcn = &wvp->wv_bcbuf;
struct wpi_cmd_beacon *cmd = (struct wpi_cmd_beacon *)&bcn->data;
cmd->id = WPI_ID_BROADCAST;
cmd->ofdm_mask = 0xff;
cmd->cck_mask = 0x0f;
cmd->lifetime = htole32(WPI_LIFETIME_INFINITE);
/*
* XXX WPI_TX_AUTO_SEQ seems to be ignored - workaround this issue
* XXX by using WPI_TX_NEED_ACK instead (with some side effects).
*/
cmd->flags = htole32(WPI_TX_NEED_ACK | WPI_TX_INSERT_TSTAMP);
bcn->code = WPI_CMD_SET_BEACON;
bcn->ac = WPI_CMD_QUEUE_NUM;
bcn->size = sizeof(struct wpi_cmd_beacon);
}
static struct ieee80211vap *
wpi_vap_create(struct ieee80211com *ic, const char name[IFNAMSIZ], int unit,
enum ieee80211_opmode opmode, int flags,
const uint8_t bssid[IEEE80211_ADDR_LEN],
const uint8_t mac[IEEE80211_ADDR_LEN])
{
struct wpi_vap *wvp;
struct ieee80211vap *vap;
if (!TAILQ_EMPTY(&ic->ic_vaps)) /* only one at a time */
return NULL;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
wvp = malloc(sizeof(struct wpi_vap), M_80211_VAP, M_WAITOK | M_ZERO);
vap = &wvp->wv_vap;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ieee80211_vap_setup(ic, vap, name, unit, opmode, flags, bssid);
if (opmode == IEEE80211_M_IBSS || opmode == IEEE80211_M_HOSTAP) {
WPI_VAP_LOCK_INIT(wvp);
wpi_init_beacon(wvp);
}
/* Override with driver methods. */
vap->iv_key_set = wpi_key_set;
vap->iv_key_delete = wpi_key_delete;
wvp->wv_recv_mgmt = vap->iv_recv_mgmt;
vap->iv_recv_mgmt = wpi_recv_mgmt;
wvp->wv_newstate = vap->iv_newstate;
vap->iv_newstate = wpi_newstate;
vap->iv_update_beacon = wpi_update_beacon;
vap->iv_max_aid = WPI_ID_IBSS_MAX - WPI_ID_IBSS_MIN + 1;
ieee80211_ratectl_init(vap);
/* Complete setup. */
ieee80211_vap_attach(vap, ieee80211_media_change,
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ieee80211_media_status, mac);
ic->ic_opmode = opmode;
return vap;
}
static void
wpi_vap_delete(struct ieee80211vap *vap)
{
struct wpi_vap *wvp = WPI_VAP(vap);
struct wpi_buf *bcn = &wvp->wv_bcbuf;
enum ieee80211_opmode opmode = vap->iv_opmode;
ieee80211_ratectl_deinit(vap);
ieee80211_vap_detach(vap);
if (opmode == IEEE80211_M_IBSS || opmode == IEEE80211_M_HOSTAP) {
if (bcn->m != NULL)
m_freem(bcn->m);
WPI_VAP_LOCK_DESTROY(wvp);
}
free(wvp, M_80211_VAP);
}
static int
wpi_detach(device_t dev)
{
struct wpi_softc *sc = device_get_softc(dev);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
uint8_t qid;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (ic->ic_vap_create == wpi_vap_create) {
ieee80211_draintask(ic, &sc->sc_radioon_task);
wpi_stop(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (sc->sc_tq != NULL) {
taskqueue_drain_all(sc->sc_tq);
taskqueue_free(sc->sc_tq);
}
callout_drain(&sc->watchdog_rfkill);
callout_drain(&sc->tx_timeout);
callout_drain(&sc->scan_timeout);
callout_drain(&sc->calib_to);
ieee80211_ifdetach(ic);
}
/* Uninstall interrupt handler. */
if (sc->irq != NULL) {
bus_teardown_intr(dev, sc->irq, sc->sc_ih);
bus_release_resource(dev, SYS_RES_IRQ, rman_get_rid(sc->irq),
sc->irq);
pci_release_msi(dev);
}
if (sc->txq[0].data_dmat) {
/* Free DMA resources. */
for (qid = 0; qid < WPI_NTXQUEUES; qid++)
wpi_free_tx_ring(sc, &sc->txq[qid]);
wpi_free_rx_ring(sc);
wpi_free_shared(sc);
}
if (sc->fw_dma.tag)
wpi_free_fwmem(sc);
if (sc->mem != NULL)
bus_release_resource(dev, SYS_RES_MEMORY,
rman_get_rid(sc->mem), sc->mem);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
WPI_TXQ_STATE_LOCK_DESTROY(sc);
WPI_TXQ_LOCK_DESTROY(sc);
WPI_NT_LOCK_DESTROY(sc);
WPI_RXON_LOCK_DESTROY(sc);
WPI_TX_LOCK_DESTROY(sc);
WPI_LOCK_DESTROY(sc);
return 0;
}
static int
wpi_shutdown(device_t dev)
{
struct wpi_softc *sc = device_get_softc(dev);
wpi_stop(sc);
return 0;
}
static int
wpi_suspend(device_t dev)
{
struct wpi_softc *sc = device_get_softc(dev);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
ieee80211_suspend_all(ic);
return 0;
}
static int
wpi_resume(device_t dev)
{
struct wpi_softc *sc = device_get_softc(dev);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
/* Clear device-specific "PCI retry timeout" register (41h). */
pci_write_config(dev, 0x41, 0, 1);
ieee80211_resume_all(ic);
return 0;
}
/*
* Grab exclusive access to NIC memory.
*/
static int
wpi_nic_lock(struct wpi_softc *sc)
{
int ntries;
/* Request exclusive access to NIC. */
WPI_SETBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
/* Spin until we actually get the lock. */
for (ntries = 0; ntries < 1000; ntries++) {
if ((WPI_READ(sc, WPI_GP_CNTRL) &
(WPI_GP_CNTRL_MAC_ACCESS_ENA | WPI_GP_CNTRL_SLEEP)) ==
WPI_GP_CNTRL_MAC_ACCESS_ENA)
return 0;
DELAY(10);
}
device_printf(sc->sc_dev, "could not lock memory\n");
return ETIMEDOUT;
}
/*
* Release lock on NIC memory.
*/
static __inline void
wpi_nic_unlock(struct wpi_softc *sc)
{
WPI_CLRBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
}
static __inline uint32_t
wpi_prph_read(struct wpi_softc *sc, uint32_t addr)
{
WPI_WRITE(sc, WPI_PRPH_RADDR, WPI_PRPH_DWORD | addr);
WPI_BARRIER_READ_WRITE(sc);
return WPI_READ(sc, WPI_PRPH_RDATA);
}
static __inline void
wpi_prph_write(struct wpi_softc *sc, uint32_t addr, uint32_t data)
{
WPI_WRITE(sc, WPI_PRPH_WADDR, WPI_PRPH_DWORD | addr);
WPI_BARRIER_WRITE(sc);
WPI_WRITE(sc, WPI_PRPH_WDATA, data);
}
static __inline void
wpi_prph_setbits(struct wpi_softc *sc, uint32_t addr, uint32_t mask)
{
wpi_prph_write(sc, addr, wpi_prph_read(sc, addr) | mask);
}
static __inline void
wpi_prph_clrbits(struct wpi_softc *sc, uint32_t addr, uint32_t mask)
{
wpi_prph_write(sc, addr, wpi_prph_read(sc, addr) & ~mask);
}
static __inline void
wpi_prph_write_region_4(struct wpi_softc *sc, uint32_t addr,
const uint32_t *data, uint32_t count)
{
for (; count != 0; count--, data++, addr += 4)
wpi_prph_write(sc, addr, *data);
}
static __inline uint32_t
wpi_mem_read(struct wpi_softc *sc, uint32_t addr)
{
WPI_WRITE(sc, WPI_MEM_RADDR, addr);
WPI_BARRIER_READ_WRITE(sc);
return WPI_READ(sc, WPI_MEM_RDATA);
}
static __inline void
wpi_mem_read_region_4(struct wpi_softc *sc, uint32_t addr, uint32_t *data,
int count)
{
for (; count > 0; count--, addr += 4)
*data++ = wpi_mem_read(sc, addr);
}
static int
wpi_read_prom_data(struct wpi_softc *sc, uint32_t addr, void *data, int count)
{
uint8_t *out = data;
uint32_t val;
int error, ntries;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
if ((error = wpi_nic_lock(sc)) != 0)
return error;
for (; count > 0; count -= 2, addr++) {
WPI_WRITE(sc, WPI_EEPROM, addr << 2);
for (ntries = 0; ntries < 10; ntries++) {
val = WPI_READ(sc, WPI_EEPROM);
if (val & WPI_EEPROM_READ_VALID)
break;
DELAY(5);
}
if (ntries == 10) {
device_printf(sc->sc_dev,
"timeout reading ROM at 0x%x\n", addr);
return ETIMEDOUT;
}
*out++= val >> 16;
if (count > 1)
*out ++= val >> 24;
}
wpi_nic_unlock(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
static void
wpi_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nsegs, int error)
{
if (error != 0)
return;
KASSERT(nsegs == 1, ("too many DMA segments, %d should be 1", nsegs));
*(bus_addr_t *)arg = segs[0].ds_addr;
}
/*
* Allocates a contiguous block of dma memory of the requested size and
* alignment.
*/
static int
wpi_dma_contig_alloc(struct wpi_softc *sc, struct wpi_dma_info *dma,
void **kvap, bus_size_t size, bus_size_t alignment)
{
int error;
dma->tag = NULL;
dma->size = size;
error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), alignment,
0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, size,
1, size, BUS_DMA_NOWAIT, NULL, NULL, &dma->tag);
if (error != 0)
goto fail;
error = bus_dmamem_alloc(dma->tag, (void **)&dma->vaddr,
BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT, &dma->map);
if (error != 0)
goto fail;
error = bus_dmamap_load(dma->tag, dma->map, dma->vaddr, size,
wpi_dma_map_addr, &dma->paddr, BUS_DMA_NOWAIT);
if (error != 0)
goto fail;
bus_dmamap_sync(dma->tag, dma->map, BUS_DMASYNC_PREWRITE);
if (kvap != NULL)
*kvap = dma->vaddr;
return 0;
fail: wpi_dma_contig_free(dma);
return error;
}
static void
wpi_dma_contig_free(struct wpi_dma_info *dma)
{
if (dma->vaddr != NULL) {
bus_dmamap_sync(dma->tag, dma->map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(dma->tag, dma->map);
bus_dmamem_free(dma->tag, dma->vaddr, dma->map);
dma->vaddr = NULL;
}
if (dma->tag != NULL) {
bus_dma_tag_destroy(dma->tag);
dma->tag = NULL;
}
}
/*
* Allocate a shared page between host and NIC.
*/
static int
wpi_alloc_shared(struct wpi_softc *sc)
{
/* Shared buffer must be aligned on a 4KB boundary. */
return wpi_dma_contig_alloc(sc, &sc->shared_dma,
(void **)&sc->shared, sizeof (struct wpi_shared), 4096);
}
static void
wpi_free_shared(struct wpi_softc *sc)
{
wpi_dma_contig_free(&sc->shared_dma);
}
/*
* Allocate DMA-safe memory for firmware transfer.
*/
static int
wpi_alloc_fwmem(struct wpi_softc *sc)
{
/* Must be aligned on a 16-byte boundary. */
return wpi_dma_contig_alloc(sc, &sc->fw_dma, NULL,
WPI_FW_TEXT_MAXSZ + WPI_FW_DATA_MAXSZ, 16);
}
static void
wpi_free_fwmem(struct wpi_softc *sc)
{
wpi_dma_contig_free(&sc->fw_dma);
}
static int
wpi_alloc_rx_ring(struct wpi_softc *sc)
{
struct wpi_rx_ring *ring = &sc->rxq;
bus_size_t size;
int i, error;
ring->cur = 0;
ring->update = 0;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/* Allocate RX descriptors (16KB aligned.) */
size = WPI_RX_RING_COUNT * sizeof (uint32_t);
error = wpi_dma_contig_alloc(sc, &ring->desc_dma,
(void **)&ring->desc, size, WPI_RING_DMA_ALIGN);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not allocate RX ring DMA memory, error %d\n",
__func__, error);
goto fail;
}
/* Create RX buffer DMA tag. */
error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 1, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
MJUMPAGESIZE, 1, MJUMPAGESIZE, BUS_DMA_NOWAIT, NULL, NULL,
&ring->data_dmat);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not create RX buf DMA tag, error %d\n",
__func__, error);
goto fail;
}
/*
* Allocate and map RX buffers.
*/
for (i = 0; i < WPI_RX_RING_COUNT; i++) {
struct wpi_rx_data *data = &ring->data[i];
bus_addr_t paddr;
error = bus_dmamap_create(ring->data_dmat, 0, &data->map);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not create RX buf DMA map, error %d\n",
__func__, error);
goto fail;
}
data->m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUMPAGESIZE);
if (data->m == NULL) {
device_printf(sc->sc_dev,
"%s: could not allocate RX mbuf\n", __func__);
error = ENOBUFS;
goto fail;
}
error = bus_dmamap_load(ring->data_dmat, data->map,
mtod(data->m, void *), MJUMPAGESIZE, wpi_dma_map_addr,
&paddr, BUS_DMA_NOWAIT);
if (error != 0 && error != EFBIG) {
device_printf(sc->sc_dev,
"%s: can't map mbuf (error %d)\n", __func__,
error);
goto fail;
}
/* Set physical address of RX buffer. */
ring->desc[i] = htole32(paddr);
}
bus_dmamap_sync(ring->desc_dma.tag, ring->desc_dma.map,
BUS_DMASYNC_PREWRITE);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
fail: wpi_free_rx_ring(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
static void
wpi_update_rx_ring(struct wpi_softc *sc)
{
WPI_WRITE(sc, WPI_FH_RX_WPTR, sc->rxq.cur & ~7);
}
static void
wpi_update_rx_ring_ps(struct wpi_softc *sc)
{
struct wpi_rx_ring *ring = &sc->rxq;
if (ring->update != 0) {
/* Wait for INT_WAKEUP event. */
return;
}
WPI_SETBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
if (WPI_READ(sc, WPI_GP_CNTRL) & WPI_GP_CNTRL_SLEEP) {
DPRINTF(sc, WPI_DEBUG_PWRSAVE, "%s: wakeup request\n",
__func__);
ring->update = 1;
} else {
wpi_update_rx_ring(sc);
WPI_CLRBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
}
}
static void
wpi_reset_rx_ring(struct wpi_softc *sc)
{
struct wpi_rx_ring *ring = &sc->rxq;
int ntries;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (wpi_nic_lock(sc) == 0) {
WPI_WRITE(sc, WPI_FH_RX_CONFIG, 0);
for (ntries = 0; ntries < 1000; ntries++) {
if (WPI_READ(sc, WPI_FH_RX_STATUS) &
WPI_FH_RX_STATUS_IDLE)
break;
DELAY(10);
}
wpi_nic_unlock(sc);
}
ring->cur = 0;
ring->update = 0;
}
static void
wpi_free_rx_ring(struct wpi_softc *sc)
{
struct wpi_rx_ring *ring = &sc->rxq;
int i;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
wpi_dma_contig_free(&ring->desc_dma);
for (i = 0; i < WPI_RX_RING_COUNT; i++) {
struct wpi_rx_data *data = &ring->data[i];
if (data->m != NULL) {
bus_dmamap_sync(ring->data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(ring->data_dmat, data->map);
m_freem(data->m);
data->m = NULL;
}
if (data->map != NULL)
bus_dmamap_destroy(ring->data_dmat, data->map);
}
if (ring->data_dmat != NULL) {
bus_dma_tag_destroy(ring->data_dmat);
ring->data_dmat = NULL;
}
}
static int
wpi_alloc_tx_ring(struct wpi_softc *sc, struct wpi_tx_ring *ring, uint8_t qid)
{
bus_addr_t paddr;
bus_size_t size;
int i, error;
ring->qid = qid;
ring->queued = 0;
ring->cur = 0;
ring->update = 0;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/* Allocate TX descriptors (16KB aligned.) */
size = WPI_TX_RING_COUNT * sizeof (struct wpi_tx_desc);
error = wpi_dma_contig_alloc(sc, &ring->desc_dma, (void **)&ring->desc,
size, WPI_RING_DMA_ALIGN);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not allocate TX ring DMA memory, error %d\n",
__func__, error);
goto fail;
}
/* Update shared area with ring physical address. */
sc->shared->txbase[qid] = htole32(ring->desc_dma.paddr);
bus_dmamap_sync(sc->shared_dma.tag, sc->shared_dma.map,
BUS_DMASYNC_PREWRITE);
/*
* We only use rings 0 through 4 (4 EDCA + cmd) so there is no need
* to allocate commands space for other rings.
* XXX Do we really need to allocate descriptors for other rings?
*/
if (qid > WPI_CMD_QUEUE_NUM) {
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
size = WPI_TX_RING_COUNT * sizeof (struct wpi_tx_cmd);
error = wpi_dma_contig_alloc(sc, &ring->cmd_dma, (void **)&ring->cmd,
size, 4);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not allocate TX cmd DMA memory, error %d\n",
__func__, error);
goto fail;
}
error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 1, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES,
WPI_MAX_SCATTER - 1, MCLBYTES, BUS_DMA_NOWAIT, NULL, NULL,
&ring->data_dmat);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not create TX buf DMA tag, error %d\n",
__func__, error);
goto fail;
}
paddr = ring->cmd_dma.paddr;
for (i = 0; i < WPI_TX_RING_COUNT; i++) {
struct wpi_tx_data *data = &ring->data[i];
data->cmd_paddr = paddr;
paddr += sizeof (struct wpi_tx_cmd);
error = bus_dmamap_create(ring->data_dmat, 0, &data->map);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not create TX buf DMA map, error %d\n",
__func__, error);
goto fail;
}
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
fail: wpi_free_tx_ring(sc, ring);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
static void
wpi_update_tx_ring(struct wpi_softc *sc, struct wpi_tx_ring *ring)
{
WPI_WRITE(sc, WPI_HBUS_TARG_WRPTR, ring->qid << 8 | ring->cur);
}
static void
wpi_update_tx_ring_ps(struct wpi_softc *sc, struct wpi_tx_ring *ring)
{
if (ring->update != 0) {
/* Wait for INT_WAKEUP event. */
return;
}
WPI_SETBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
if (WPI_READ(sc, WPI_GP_CNTRL) & WPI_GP_CNTRL_SLEEP) {
DPRINTF(sc, WPI_DEBUG_PWRSAVE, "%s (%d): requesting wakeup\n",
__func__, ring->qid);
ring->update = 1;
} else {
wpi_update_tx_ring(sc, ring);
WPI_CLRBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
}
}
static void
wpi_reset_tx_ring(struct wpi_softc *sc, struct wpi_tx_ring *ring)
{
int i;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
for (i = 0; i < WPI_TX_RING_COUNT; i++) {
struct wpi_tx_data *data = &ring->data[i];
if (data->m != NULL) {
bus_dmamap_sync(ring->data_dmat, data->map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(ring->data_dmat, data->map);
m_freem(data->m);
data->m = NULL;
}
if (data->ni != NULL) {
ieee80211_free_node(data->ni);
data->ni = NULL;
}
}
/* Clear TX descriptors. */
memset(ring->desc, 0, ring->desc_dma.size);
bus_dmamap_sync(ring->desc_dma.tag, ring->desc_dma.map,
BUS_DMASYNC_PREWRITE);
ring->queued = 0;
ring->cur = 0;
ring->update = 0;
}
static void
wpi_free_tx_ring(struct wpi_softc *sc, struct wpi_tx_ring *ring)
{
int i;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
wpi_dma_contig_free(&ring->desc_dma);
wpi_dma_contig_free(&ring->cmd_dma);
for (i = 0; i < WPI_TX_RING_COUNT; i++) {
struct wpi_tx_data *data = &ring->data[i];
if (data->m != NULL) {
bus_dmamap_sync(ring->data_dmat, data->map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(ring->data_dmat, data->map);
m_freem(data->m);
}
if (data->map != NULL)
bus_dmamap_destroy(ring->data_dmat, data->map);
}
if (ring->data_dmat != NULL) {
bus_dma_tag_destroy(ring->data_dmat);
ring->data_dmat = NULL;
}
}
/*
* Extract various information from EEPROM.
*/
static int
wpi_read_eeprom(struct wpi_softc *sc, uint8_t macaddr[IEEE80211_ADDR_LEN])
{
#define WPI_CHK(res) do { \
if ((error = res) != 0) \
goto fail; \
} while (0)
uint8_t i;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/* Adapter has to be powered on for EEPROM access to work. */
if ((error = wpi_apm_init(sc)) != 0) {
device_printf(sc->sc_dev,
"%s: could not power ON adapter, error %d\n", __func__,
error);
return error;
}
if ((WPI_READ(sc, WPI_EEPROM_GP) & 0x6) == 0) {
device_printf(sc->sc_dev, "bad EEPROM signature\n");
error = EIO;
goto fail;
}
/* Clear HW ownership of EEPROM. */
WPI_CLRBITS(sc, WPI_EEPROM_GP, WPI_EEPROM_GP_IF_OWNER);
/* Read the hardware capabilities, revision and SKU type. */
WPI_CHK(wpi_read_prom_data(sc, WPI_EEPROM_SKU_CAP, &sc->cap,
sizeof(sc->cap)));
WPI_CHK(wpi_read_prom_data(sc, WPI_EEPROM_REVISION, &sc->rev,
sizeof(sc->rev)));
WPI_CHK(wpi_read_prom_data(sc, WPI_EEPROM_TYPE, &sc->type,
sizeof(sc->type)));
sc->rev = le16toh(sc->rev);
DPRINTF(sc, WPI_DEBUG_EEPROM, "cap=%x rev=%x type=%x\n", sc->cap,
sc->rev, sc->type);
/* Read the regulatory domain (4 ASCII characters.) */
WPI_CHK(wpi_read_prom_data(sc, WPI_EEPROM_DOMAIN, sc->domain,
sizeof(sc->domain)));
/* Read MAC address. */
WPI_CHK(wpi_read_prom_data(sc, WPI_EEPROM_MAC, macaddr,
IEEE80211_ADDR_LEN));
/* Read the list of authorized channels. */
for (i = 0; i < WPI_CHAN_BANDS_COUNT; i++)
WPI_CHK(wpi_read_eeprom_channels(sc, i));
/* Read the list of TX power groups. */
for (i = 0; i < WPI_POWER_GROUPS_COUNT; i++)
WPI_CHK(wpi_read_eeprom_group(sc, i));
fail: wpi_apm_stop(sc); /* Power OFF adapter. */
DPRINTF(sc, WPI_DEBUG_TRACE, error ? TRACE_STR_END_ERR : TRACE_STR_END,
__func__);
return error;
#undef WPI_CHK
}
/*
* Translate EEPROM flags to net80211.
*/
static uint32_t
wpi_eeprom_channel_flags(struct wpi_eeprom_chan *channel)
{
uint32_t nflags;
nflags = 0;
if ((channel->flags & WPI_EEPROM_CHAN_ACTIVE) == 0)
nflags |= IEEE80211_CHAN_PASSIVE;
if ((channel->flags & WPI_EEPROM_CHAN_IBSS) == 0)
nflags |= IEEE80211_CHAN_NOADHOC;
if (channel->flags & WPI_EEPROM_CHAN_RADAR) {
nflags |= IEEE80211_CHAN_DFS;
/* XXX apparently IBSS may still be marked */
nflags |= IEEE80211_CHAN_NOADHOC;
}
/* XXX HOSTAP uses WPI_MODE_IBSS */
if (nflags & IEEE80211_CHAN_NOADHOC)
nflags |= IEEE80211_CHAN_NOHOSTAP;
return nflags;
}
static void
wpi_read_eeprom_band(struct wpi_softc *sc, uint8_t n)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct wpi_eeprom_chan *channels = sc->eeprom_channels[n];
const struct wpi_chan_band *band = &wpi_bands[n];
struct ieee80211_channel *c;
uint32_t nflags;
uint8_t chan, i;
for (i = 0; i < band->nchan; i++) {
if (!(channels[i].flags & WPI_EEPROM_CHAN_VALID)) {
DPRINTF(sc, WPI_DEBUG_EEPROM,
"Channel Not Valid: %d, band %d\n",
band->chan[i],n);
continue;
}
chan = band->chan[i];
nflags = wpi_eeprom_channel_flags(&channels[i]);
c = &ic->ic_channels[ic->ic_nchans++];
c->ic_ieee = chan;
c->ic_maxregpower = channels[i].maxpwr;
c->ic_maxpower = 2*c->ic_maxregpower;
if (n == 0) { /* 2GHz band */
c->ic_freq = ieee80211_ieee2mhz(chan,
IEEE80211_CHAN_G);
/* G =>'s B is supported */
c->ic_flags = IEEE80211_CHAN_B | nflags;
c = &ic->ic_channels[ic->ic_nchans++];
c[0] = c[-1];
c->ic_flags = IEEE80211_CHAN_G | nflags;
} else { /* 5GHz band */
c->ic_freq = ieee80211_ieee2mhz(chan,
IEEE80211_CHAN_A);
c->ic_flags = IEEE80211_CHAN_A | nflags;
}
/* Save maximum allowed TX power for this channel. */
sc->maxpwr[chan] = channels[i].maxpwr;
DPRINTF(sc, WPI_DEBUG_EEPROM,
"adding chan %d (%dMHz) flags=0x%x maxpwr=%d passive=%d,"
" offset %d\n", chan, c->ic_freq,
channels[i].flags, sc->maxpwr[chan],
IEEE80211_IS_CHAN_PASSIVE(c), ic->ic_nchans);
}
}
/**
* Read the eeprom to find out what channels are valid for the given
* band and update net80211 with what we find.
*/
static int
wpi_read_eeprom_channels(struct wpi_softc *sc, uint8_t n)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
const struct wpi_chan_band *band = &wpi_bands[n];
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
error = wpi_read_prom_data(sc, band->addr, &sc->eeprom_channels[n],
band->nchan * sizeof (struct wpi_eeprom_chan));
if (error != 0) {
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
wpi_read_eeprom_band(sc, n);
ieee80211_sort_channels(ic->ic_channels, ic->ic_nchans);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
static struct wpi_eeprom_chan *
wpi_find_eeprom_channel(struct wpi_softc *sc, struct ieee80211_channel *c)
{
int i, j;
for (j = 0; j < WPI_CHAN_BANDS_COUNT; j++)
for (i = 0; i < wpi_bands[j].nchan; i++)
if (wpi_bands[j].chan[i] == c->ic_ieee)
return &sc->eeprom_channels[j][i];
return NULL;
}
/*
* Enforce flags read from EEPROM.
*/
static int
wpi_setregdomain(struct ieee80211com *ic, struct ieee80211_regdomain *rd,
int nchan, struct ieee80211_channel chans[])
{
struct wpi_softc *sc = ic->ic_softc;
int i;
for (i = 0; i < nchan; i++) {
struct ieee80211_channel *c = &chans[i];
struct wpi_eeprom_chan *channel;
channel = wpi_find_eeprom_channel(sc, c);
if (channel == NULL) {
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ic_printf(ic, "%s: invalid channel %u freq %u/0x%x\n",
__func__, c->ic_ieee, c->ic_freq, c->ic_flags);
return EINVAL;
}
c->ic_flags |= wpi_eeprom_channel_flags(channel);
}
return 0;
}
static int
wpi_read_eeprom_group(struct wpi_softc *sc, uint8_t n)
{
struct wpi_power_group *group = &sc->groups[n];
struct wpi_eeprom_group rgroup;
int i, error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
if ((error = wpi_read_prom_data(sc, WPI_EEPROM_POWER_GRP + n * 32,
&rgroup, sizeof rgroup)) != 0) {
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
/* Save TX power group information. */
group->chan = rgroup.chan;
group->maxpwr = rgroup.maxpwr;
/* Retrieve temperature at which the samples were taken. */
group->temp = (int16_t)le16toh(rgroup.temp);
DPRINTF(sc, WPI_DEBUG_EEPROM,
"power group %d: chan=%d maxpwr=%d temp=%d\n", n, group->chan,
group->maxpwr, group->temp);
for (i = 0; i < WPI_SAMPLES_COUNT; i++) {
group->samples[i].index = rgroup.samples[i].index;
group->samples[i].power = rgroup.samples[i].power;
DPRINTF(sc, WPI_DEBUG_EEPROM,
"\tsample %d: index=%d power=%d\n", i,
group->samples[i].index, group->samples[i].power);
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
static __inline uint8_t
wpi_add_node_entry_adhoc(struct wpi_softc *sc)
{
uint8_t newid = WPI_ID_IBSS_MIN;
for (; newid <= WPI_ID_IBSS_MAX; newid++) {
if ((sc->nodesmsk & (1 << newid)) == 0) {
sc->nodesmsk |= 1 << newid;
return newid;
}
}
return WPI_ID_UNDEFINED;
}
static __inline uint8_t
wpi_add_node_entry_sta(struct wpi_softc *sc)
{
sc->nodesmsk |= 1 << WPI_ID_BSS;
return WPI_ID_BSS;
}
static __inline int
wpi_check_node_entry(struct wpi_softc *sc, uint8_t id)
{
if (id == WPI_ID_UNDEFINED)
return 0;
return (sc->nodesmsk >> id) & 1;
}
static __inline void
wpi_clear_node_table(struct wpi_softc *sc)
{
sc->nodesmsk = 0;
}
static __inline void
wpi_del_node_entry(struct wpi_softc *sc, uint8_t id)
{
sc->nodesmsk &= ~(1 << id);
}
static struct ieee80211_node *
wpi_node_alloc(struct ieee80211vap *vap, const uint8_t mac[IEEE80211_ADDR_LEN])
{
struct wpi_node *wn;
wn = malloc(sizeof (struct wpi_node), M_80211_NODE,
M_NOWAIT | M_ZERO);
if (wn == NULL)
return NULL;
wn->id = WPI_ID_UNDEFINED;
return &wn->ni;
}
static void
wpi_node_free(struct ieee80211_node *ni)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct wpi_softc *sc = ni->ni_ic->ic_softc;
struct wpi_node *wn = WPI_NODE(ni);
if (wn->id != WPI_ID_UNDEFINED) {
WPI_NT_LOCK(sc);
if (wpi_check_node_entry(sc, wn->id)) {
wpi_del_node_entry(sc, wn->id);
wpi_del_node(sc, ni);
}
WPI_NT_UNLOCK(sc);
}
sc->sc_node_free(ni);
}
static __inline int
wpi_check_bss_filter(struct wpi_softc *sc)
{
return (sc->rxon.filter & htole32(WPI_FILTER_BSS)) != 0;
}
static void
wpi_recv_mgmt(struct ieee80211_node *ni, struct mbuf *m, int subtype,
const struct ieee80211_rx_stats *rxs,
int rssi, int nf)
{
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_softc *sc = vap->iv_ic->ic_softc;
struct wpi_vap *wvp = WPI_VAP(vap);
uint64_t ni_tstamp, rx_tstamp;
wvp->wv_recv_mgmt(ni, m, subtype, rxs, rssi, nf);
if (vap->iv_opmode == IEEE80211_M_IBSS &&
vap->iv_state == IEEE80211_S_RUN &&
(subtype == IEEE80211_FC0_SUBTYPE_BEACON ||
subtype == IEEE80211_FC0_SUBTYPE_PROBE_RESP)) {
ni_tstamp = le64toh(ni->ni_tstamp.tsf);
rx_tstamp = le64toh(sc->rx_tstamp);
if (ni_tstamp >= rx_tstamp) {
DPRINTF(sc, WPI_DEBUG_STATE,
"ibss merge, tsf %ju tstamp %ju\n",
(uintmax_t)rx_tstamp, (uintmax_t)ni_tstamp);
(void) ieee80211_ibss_merge(ni);
}
}
}
static void
wpi_restore_node(void *arg, struct ieee80211_node *ni)
{
struct wpi_softc *sc = arg;
struct wpi_node *wn = WPI_NODE(ni);
int error;
WPI_NT_LOCK(sc);
if (wn->id != WPI_ID_UNDEFINED) {
wn->id = WPI_ID_UNDEFINED;
if ((error = wpi_add_ibss_node(sc, ni)) != 0) {
device_printf(sc->sc_dev,
"%s: could not add IBSS node, error %d\n",
__func__, error);
}
}
WPI_NT_UNLOCK(sc);
}
static void
wpi_restore_node_table(struct wpi_softc *sc, struct wpi_vap *wvp)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
/* Set group keys once. */
WPI_NT_LOCK(sc);
wvp->wv_gtk = 0;
WPI_NT_UNLOCK(sc);
ieee80211_iterate_nodes(&ic->ic_sta, wpi_restore_node, sc);
ieee80211_crypto_reload_keys(ic);
}
/**
* Called by net80211 when ever there is a change to 80211 state machine
*/
static int
wpi_newstate(struct ieee80211vap *vap, enum ieee80211_state nstate, int arg)
{
struct wpi_vap *wvp = WPI_VAP(vap);
struct ieee80211com *ic = vap->iv_ic;
struct wpi_softc *sc = ic->ic_softc;
int error = 0;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
WPI_TXQ_LOCK(sc);
if (nstate > IEEE80211_S_INIT && sc->sc_running == 0) {
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
WPI_TXQ_UNLOCK(sc);
return ENXIO;
}
WPI_TXQ_UNLOCK(sc);
DPRINTF(sc, WPI_DEBUG_STATE, "%s: %s -> %s\n", __func__,
ieee80211_state_name[vap->iv_state],
ieee80211_state_name[nstate]);
if (vap->iv_state == IEEE80211_S_RUN && nstate < IEEE80211_S_RUN) {
if ((error = wpi_set_pslevel(sc, 0, 0, 1)) != 0) {
device_printf(sc->sc_dev,
"%s: could not set power saving level\n",
__func__);
return error;
}
wpi_set_led(sc, WPI_LED_LINK, 1, 0);
}
switch (nstate) {
case IEEE80211_S_SCAN:
WPI_RXON_LOCK(sc);
if (wpi_check_bss_filter(sc) != 0) {
sc->rxon.filter &= ~htole32(WPI_FILTER_BSS);
if ((error = wpi_send_rxon(sc, 0, 1)) != 0) {
device_printf(sc->sc_dev,
"%s: could not send RXON\n", __func__);
}
}
WPI_RXON_UNLOCK(sc);
break;
case IEEE80211_S_ASSOC:
if (vap->iv_state != IEEE80211_S_RUN)
break;
/* FALLTHROUGH */
case IEEE80211_S_AUTH:
/*
* NB: do not optimize AUTH -> AUTH state transmission -
* this will break powersave with non-QoS AP!
*/
/*
* The node must be registered in the firmware before auth.
* Also the associd must be cleared on RUN -> ASSOC
* transitions.
*/
if ((error = wpi_auth(sc, vap)) != 0) {
device_printf(sc->sc_dev,
"%s: could not move to AUTH state, error %d\n",
__func__, error);
}
break;
case IEEE80211_S_RUN:
/*
* RUN -> RUN transition:
* STA mode: Just restart the timers.
* IBSS mode: Process IBSS merge.
*/
if (vap->iv_state == IEEE80211_S_RUN) {
if (vap->iv_opmode != IEEE80211_M_IBSS) {
WPI_RXON_LOCK(sc);
wpi_calib_timeout(sc);
WPI_RXON_UNLOCK(sc);
break;
} else {
/*
* Drop the BSS_FILTER bit
* (there is no another way to change bssid).
*/
WPI_RXON_LOCK(sc);
sc->rxon.filter &= ~htole32(WPI_FILTER_BSS);
if ((error = wpi_send_rxon(sc, 0, 1)) != 0) {
device_printf(sc->sc_dev,
"%s: could not send RXON\n",
__func__);
}
WPI_RXON_UNLOCK(sc);
/* Restore all what was lost. */
wpi_restore_node_table(sc, wvp);
/* XXX set conditionally? */
wpi_updateedca(ic);
}
}
/*
* !RUN -> RUN requires setting the association id
* which is done with a firmware cmd. We also defer
* starting the timers until that work is done.
*/
if ((error = wpi_run(sc, vap)) != 0) {
device_printf(sc->sc_dev,
"%s: could not move to RUN state\n", __func__);
}
break;
default:
break;
}
if (error != 0) {
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return wvp->wv_newstate(vap, nstate, arg);
}
static void
wpi_calib_timeout(void *arg)
{
struct wpi_softc *sc = arg;
if (wpi_check_bss_filter(sc) == 0)
return;
wpi_power_calibration(sc);
callout_reset(&sc->calib_to, 60*hz, wpi_calib_timeout, sc);
}
static __inline uint8_t
rate2plcp(const uint8_t rate)
{
switch (rate) {
case 12: return 0xd;
case 18: return 0xf;
case 24: return 0x5;
case 36: return 0x7;
case 48: return 0x9;
case 72: return 0xb;
case 96: return 0x1;
case 108: return 0x3;
case 2: return 10;
case 4: return 20;
case 11: return 55;
case 22: return 110;
default: return 0;
}
}
static __inline uint8_t
plcp2rate(const uint8_t plcp)
{
switch (plcp) {
case 0xd: return 12;
case 0xf: return 18;
case 0x5: return 24;
case 0x7: return 36;
case 0x9: return 48;
case 0xb: return 72;
case 0x1: return 96;
case 0x3: return 108;
case 10: return 2;
case 20: return 4;
case 55: return 11;
case 110: return 22;
default: return 0;
}
}
/* Quickly determine if a given rate is CCK or OFDM. */
#define WPI_RATE_IS_OFDM(rate) ((rate) >= 12 && (rate) != 22)
static void
wpi_rx_done(struct wpi_softc *sc, struct wpi_rx_desc *desc,
struct wpi_rx_data *data)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct wpi_rx_ring *ring = &sc->rxq;
struct wpi_rx_stat *stat;
struct wpi_rx_head *head;
struct wpi_rx_tail *tail;
struct ieee80211_frame *wh;
struct ieee80211_node *ni;
struct mbuf *m, *m1;
bus_addr_t paddr;
uint32_t flags;
uint16_t len;
int error;
stat = (struct wpi_rx_stat *)(desc + 1);
if (__predict_false(stat->len > WPI_STAT_MAXLEN)) {
device_printf(sc->sc_dev, "invalid RX statistic header\n");
goto fail1;
}
bus_dmamap_sync(ring->data_dmat, data->map, BUS_DMASYNC_POSTREAD);
head = (struct wpi_rx_head *)((caddr_t)(stat + 1) + stat->len);
len = le16toh(head->len);
tail = (struct wpi_rx_tail *)((caddr_t)(head + 1) + len);
flags = le32toh(tail->flags);
DPRINTF(sc, WPI_DEBUG_RECV, "%s: idx %d len %d stat len %u rssi %d"
" rate %x chan %d tstamp %ju\n", __func__, ring->cur,
le32toh(desc->len), len, (int8_t)stat->rssi,
head->plcp, head->chan, (uintmax_t)le64toh(tail->tstamp));
/* Discard frames with a bad FCS early. */
if ((flags & WPI_RX_NOERROR) != WPI_RX_NOERROR) {
DPRINTF(sc, WPI_DEBUG_RECV, "%s: RX flags error %x\n",
__func__, flags);
goto fail1;
}
/* Discard frames that are too short. */
if (len < sizeof (struct ieee80211_frame_ack)) {
DPRINTF(sc, WPI_DEBUG_RECV, "%s: frame too short: %d\n",
__func__, len);
goto fail1;
}
m1 = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUMPAGESIZE);
if (__predict_false(m1 == NULL)) {
DPRINTF(sc, WPI_DEBUG_ANY, "%s: no mbuf to restock ring\n",
__func__);
goto fail1;
}
bus_dmamap_unload(ring->data_dmat, data->map);
error = bus_dmamap_load(ring->data_dmat, data->map, mtod(m1, void *),
MJUMPAGESIZE, wpi_dma_map_addr, &paddr, BUS_DMA_NOWAIT);
if (__predict_false(error != 0 && error != EFBIG)) {
device_printf(sc->sc_dev,
"%s: bus_dmamap_load failed, error %d\n", __func__, error);
m_freem(m1);
/* Try to reload the old mbuf. */
error = bus_dmamap_load(ring->data_dmat, data->map,
mtod(data->m, void *), MJUMPAGESIZE, wpi_dma_map_addr,
&paddr, BUS_DMA_NOWAIT);
if (error != 0 && error != EFBIG) {
panic("%s: could not load old RX mbuf", __func__);
}
/* Physical address may have changed. */
ring->desc[ring->cur] = htole32(paddr);
bus_dmamap_sync(ring->data_dmat, ring->desc_dma.map,
BUS_DMASYNC_PREWRITE);
goto fail1;
}
m = data->m;
data->m = m1;
/* Update RX descriptor. */
ring->desc[ring->cur] = htole32(paddr);
bus_dmamap_sync(ring->desc_dma.tag, ring->desc_dma.map,
BUS_DMASYNC_PREWRITE);
/* Finalize mbuf. */
m->m_data = (caddr_t)(head + 1);
m->m_pkthdr.len = m->m_len = len;
/* Grab a reference to the source node. */
wh = mtod(m, struct ieee80211_frame *);
if ((wh->i_fc[1] & IEEE80211_FC1_PROTECTED) &&
(flags & WPI_RX_CIPHER_MASK) == WPI_RX_CIPHER_CCMP) {
/* Check whether decryption was successful or not. */
if ((flags & WPI_RX_DECRYPT_MASK) != WPI_RX_DECRYPT_OK) {
DPRINTF(sc, WPI_DEBUG_RECV,
"CCMP decryption failed 0x%x\n", flags);
goto fail2;
}
m->m_flags |= M_WEP;
}
if (len >= sizeof(struct ieee80211_frame_min))
ni = ieee80211_find_rxnode(ic, (struct ieee80211_frame_min *)wh);
else
ni = NULL;
sc->rx_tstamp = tail->tstamp;
if (ieee80211_radiotap_active(ic)) {
struct wpi_rx_radiotap_header *tap = &sc->sc_rxtap;
tap->wr_flags = 0;
if (head->flags & htole16(WPI_STAT_FLAG_SHPREAMBLE))
tap->wr_flags |= IEEE80211_RADIOTAP_F_SHORTPRE;
tap->wr_dbm_antsignal = (int8_t)(stat->rssi + WPI_RSSI_OFFSET);
tap->wr_dbm_antnoise = WPI_RSSI_OFFSET;
tap->wr_tsft = tail->tstamp;
tap->wr_antenna = (le16toh(head->flags) >> 4) & 0xf;
tap->wr_rate = plcp2rate(head->plcp);
}
WPI_UNLOCK(sc);
/* Send the frame to the 802.11 layer. */
if (ni != NULL) {
(void)ieee80211_input(ni, m, stat->rssi, WPI_RSSI_OFFSET);
/* Node is no longer needed. */
ieee80211_free_node(ni);
} else
(void)ieee80211_input_all(ic, m, stat->rssi, WPI_RSSI_OFFSET);
WPI_LOCK(sc);
return;
fail2: m_freem(m);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
fail1: counter_u64_add(ic->ic_ierrors, 1);
}
static void
wpi_rx_statistics(struct wpi_softc *sc, struct wpi_rx_desc *desc,
struct wpi_rx_data *data)
{
/* Ignore */
}
static void
wpi_tx_done(struct wpi_softc *sc, struct wpi_rx_desc *desc)
{
struct wpi_tx_ring *ring = &sc->txq[desc->qid & 0x3];
struct wpi_tx_data *data = &ring->data[desc->idx];
struct wpi_tx_stat *stat = (struct wpi_tx_stat *)(desc + 1);
struct mbuf *m;
struct ieee80211_node *ni;
struct ieee80211vap *vap;
struct ieee80211com *ic;
uint32_t status = le32toh(stat->status);
int ackfailcnt = stat->ackfailcnt / WPI_NTRIES_DEFAULT;
KASSERT(data->ni != NULL, ("no node"));
KASSERT(data->m != NULL, ("no mbuf"));
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
DPRINTF(sc, WPI_DEBUG_XMIT, "%s: "
"qid %d idx %d retries %d btkillcnt %d rate %x duration %d "
"status %x\n", __func__, desc->qid, desc->idx, stat->ackfailcnt,
stat->btkillcnt, stat->rate, le32toh(stat->duration), status);
/* Unmap and free mbuf. */
bus_dmamap_sync(ring->data_dmat, data->map, BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(ring->data_dmat, data->map);
m = data->m, data->m = NULL;
ni = data->ni, data->ni = NULL;
vap = ni->ni_vap;
ic = vap->iv_ic;
/*
* Update rate control statistics for the node.
*/
if (status & WPI_TX_STATUS_FAIL) {
ieee80211_ratectl_tx_complete(vap, ni,
IEEE80211_RATECTL_TX_FAILURE, &ackfailcnt, NULL);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
} else
ieee80211_ratectl_tx_complete(vap, ni,
IEEE80211_RATECTL_TX_SUCCESS, &ackfailcnt, NULL);
ieee80211_tx_complete(ni, m, (status & WPI_TX_STATUS_FAIL) != 0);
WPI_TXQ_STATE_LOCK(sc);
if (--ring->queued > 0)
callout_reset(&sc->tx_timeout, 5*hz, wpi_tx_timeout, sc);
else
callout_stop(&sc->tx_timeout);
WPI_TXQ_STATE_UNLOCK(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
}
/*
* Process a "command done" firmware notification. This is where we wakeup
* processes waiting for a synchronous command completion.
*/
static void
wpi_cmd_done(struct wpi_softc *sc, struct wpi_rx_desc *desc)
{
struct wpi_tx_ring *ring = &sc->txq[WPI_CMD_QUEUE_NUM];
struct wpi_tx_data *data;
struct wpi_tx_cmd *cmd;
DPRINTF(sc, WPI_DEBUG_CMD, "cmd notification qid %x idx %d flags %x "
"type %s len %d\n", desc->qid, desc->idx,
desc->flags, wpi_cmd_str(desc->type),
le32toh(desc->len));
if ((desc->qid & WPI_RX_DESC_QID_MSK) != WPI_CMD_QUEUE_NUM)
return; /* Not a command ack. */
KASSERT(ring->queued == 0, ("ring->queued must be 0"));
data = &ring->data[desc->idx];
cmd = &ring->cmd[desc->idx];
/* If the command was mapped in an mbuf, free it. */
if (data->m != NULL) {
bus_dmamap_sync(ring->data_dmat, data->map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(ring->data_dmat, data->map);
m_freem(data->m);
data->m = NULL;
}
wakeup(cmd);
if (desc->type == WPI_CMD_SET_POWER_MODE) {
struct wpi_pmgt_cmd *pcmd = (struct wpi_pmgt_cmd *)cmd->data;
bus_dmamap_sync(ring->data_dmat, ring->cmd_dma.map,
BUS_DMASYNC_POSTREAD);
WPI_TXQ_LOCK(sc);
if (le16toh(pcmd->flags) & WPI_PS_ALLOW_SLEEP) {
sc->sc_update_rx_ring = wpi_update_rx_ring_ps;
sc->sc_update_tx_ring = wpi_update_tx_ring_ps;
} else {
sc->sc_update_rx_ring = wpi_update_rx_ring;
sc->sc_update_tx_ring = wpi_update_tx_ring;
}
WPI_TXQ_UNLOCK(sc);
}
}
static void
wpi_notif_intr(struct wpi_softc *sc)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
uint32_t hw;
bus_dmamap_sync(sc->shared_dma.tag, sc->shared_dma.map,
BUS_DMASYNC_POSTREAD);
hw = le32toh(sc->shared->next) & 0xfff;
hw = (hw == 0) ? WPI_RX_RING_COUNT - 1 : hw - 1;
while (sc->rxq.cur != hw) {
sc->rxq.cur = (sc->rxq.cur + 1) % WPI_RX_RING_COUNT;
struct wpi_rx_data *data = &sc->rxq.data[sc->rxq.cur];
struct wpi_rx_desc *desc;
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
desc = mtod(data->m, struct wpi_rx_desc *);
DPRINTF(sc, WPI_DEBUG_NOTIFY,
"%s: cur=%d; qid %x idx %d flags %x type %d(%s) len %d\n",
__func__, sc->rxq.cur, desc->qid, desc->idx, desc->flags,
desc->type, wpi_cmd_str(desc->type), le32toh(desc->len));
if (!(desc->qid & WPI_UNSOLICITED_RX_NOTIF)) {
/* Reply to a command. */
wpi_cmd_done(sc, desc);
}
switch (desc->type) {
case WPI_RX_DONE:
/* An 802.11 frame has been received. */
wpi_rx_done(sc, desc, data);
if (__predict_false(sc->sc_running == 0)) {
/* wpi_stop() was called. */
return;
}
break;
case WPI_TX_DONE:
/* An 802.11 frame has been transmitted. */
wpi_tx_done(sc, desc);
break;
case WPI_RX_STATISTICS:
case WPI_BEACON_STATISTICS:
wpi_rx_statistics(sc, desc, data);
break;
case WPI_BEACON_MISSED:
{
struct wpi_beacon_missed *miss =
(struct wpi_beacon_missed *)(desc + 1);
uint32_t expected, misses, received, threshold;
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
misses = le32toh(miss->consecutive);
expected = le32toh(miss->expected);
received = le32toh(miss->received);
threshold = MAX(2, vap->iv_bmissthreshold);
DPRINTF(sc, WPI_DEBUG_BMISS,
"%s: beacons missed %u(%u) (received %u/%u)\n",
__func__, misses, le32toh(miss->total), received,
expected);
if (misses >= threshold ||
(received == 0 && expected >= threshold)) {
WPI_RXON_LOCK(sc);
if (callout_pending(&sc->scan_timeout)) {
wpi_cmd(sc, WPI_CMD_SCAN_ABORT, NULL,
0, 1);
}
WPI_RXON_UNLOCK(sc);
if (vap->iv_state == IEEE80211_S_RUN &&
(ic->ic_flags & IEEE80211_F_SCAN) == 0)
ieee80211_beacon_miss(ic);
}
break;
}
#ifdef WPI_DEBUG
case WPI_BEACON_SENT:
{
struct wpi_tx_stat *stat =
(struct wpi_tx_stat *)(desc + 1);
uint64_t *tsf = (uint64_t *)(stat + 1);
uint32_t *mode = (uint32_t *)(tsf + 1);
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
DPRINTF(sc, WPI_DEBUG_BEACON,
"beacon sent: rts %u, ack %u, btkill %u, rate %u, "
"duration %u, status %x, tsf %ju, mode %x\n",
stat->rtsfailcnt, stat->ackfailcnt,
stat->btkillcnt, stat->rate, le32toh(stat->duration),
le32toh(stat->status), le64toh(*tsf),
le32toh(*mode));
break;
}
#endif
case WPI_UC_READY:
{
struct wpi_ucode_info *uc =
(struct wpi_ucode_info *)(desc + 1);
/* The microcontroller is ready. */
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
DPRINTF(sc, WPI_DEBUG_RESET,
"microcode alive notification version=%d.%d "
"subtype=%x alive=%x\n", uc->major, uc->minor,
uc->subtype, le32toh(uc->valid));
if (le32toh(uc->valid) != 1) {
device_printf(sc->sc_dev,
"microcontroller initialization failed\n");
wpi_stop_locked(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
return;
}
/* Save the address of the error log in SRAM. */
sc->errptr = le32toh(uc->errptr);
break;
}
case WPI_STATE_CHANGED:
{
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
uint32_t *status = (uint32_t *)(desc + 1);
DPRINTF(sc, WPI_DEBUG_STATE, "state changed to %x\n",
le32toh(*status));
if (le32toh(*status) & 1) {
WPI_NT_LOCK(sc);
wpi_clear_node_table(sc);
WPI_NT_UNLOCK(sc);
taskqueue_enqueue(sc->sc_tq,
&sc->sc_radiooff_task);
return;
}
break;
}
#ifdef WPI_DEBUG
case WPI_START_SCAN:
{
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
struct wpi_start_scan *scan =
(struct wpi_start_scan *)(desc + 1);
DPRINTF(sc, WPI_DEBUG_SCAN,
"%s: scanning channel %d status %x\n",
__func__, scan->chan, le32toh(scan->status));
break;
}
#endif
case WPI_STOP_SCAN:
{
bus_dmamap_sync(sc->rxq.data_dmat, data->map,
BUS_DMASYNC_POSTREAD);
struct wpi_stop_scan *scan =
(struct wpi_stop_scan *)(desc + 1);
DPRINTF(sc, WPI_DEBUG_SCAN,
"scan finished nchan=%d status=%d chan=%d\n",
scan->nchan, scan->status, scan->chan);
WPI_RXON_LOCK(sc);
callout_stop(&sc->scan_timeout);
WPI_RXON_UNLOCK(sc);
if (scan->status == WPI_SCAN_ABORTED)
ieee80211_cancel_scan(vap);
else
ieee80211_scan_next(vap);
break;
}
}
if (sc->rxq.cur % 8 == 0) {
/* Tell the firmware what we have processed. */
sc->sc_update_rx_ring(sc);
}
}
}
/*
* Process an INT_WAKEUP interrupt raised when the microcontroller wakes up
* from power-down sleep mode.
*/
static void
wpi_wakeup_intr(struct wpi_softc *sc)
{
int qid;
DPRINTF(sc, WPI_DEBUG_PWRSAVE,
"%s: ucode wakeup from power-down sleep\n", __func__);
/* Wakeup RX and TX rings. */
if (sc->rxq.update) {
sc->rxq.update = 0;
wpi_update_rx_ring(sc);
}
WPI_TXQ_LOCK(sc);
for (qid = 0; qid < WPI_DRV_NTXQUEUES; qid++) {
struct wpi_tx_ring *ring = &sc->txq[qid];
if (ring->update) {
ring->update = 0;
wpi_update_tx_ring(sc, ring);
}
}
WPI_CLRBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_MAC_ACCESS_REQ);
WPI_TXQ_UNLOCK(sc);
}
/*
* This function prints firmware registers
*/
#ifdef WPI_DEBUG
static void
wpi_debug_registers(struct wpi_softc *sc)
{
size_t i;
static const uint32_t csr_tbl[] = {
WPI_HW_IF_CONFIG,
WPI_INT,
WPI_INT_MASK,
WPI_FH_INT,
WPI_GPIO_IN,
WPI_RESET,
WPI_GP_CNTRL,
WPI_EEPROM,
WPI_EEPROM_GP,
WPI_GIO,
WPI_UCODE_GP1,
WPI_UCODE_GP2,
WPI_GIO_CHICKEN,
WPI_ANA_PLL,
WPI_DBG_HPET_MEM,
};
static const uint32_t prph_tbl[] = {
WPI_APMG_CLK_CTRL,
WPI_APMG_PS,
WPI_APMG_PCI_STT,
WPI_APMG_RFKILL,
};
DPRINTF(sc, WPI_DEBUG_REGISTER,"%s","\n");
for (i = 0; i < nitems(csr_tbl); i++) {
DPRINTF(sc, WPI_DEBUG_REGISTER, " %-18s: 0x%08x ",
wpi_get_csr_string(csr_tbl[i]), WPI_READ(sc, csr_tbl[i]));
if ((i + 1) % 2 == 0)
DPRINTF(sc, WPI_DEBUG_REGISTER, "\n");
}
DPRINTF(sc, WPI_DEBUG_REGISTER, "\n\n");
if (wpi_nic_lock(sc) == 0) {
for (i = 0; i < nitems(prph_tbl); i++) {
DPRINTF(sc, WPI_DEBUG_REGISTER, " %-18s: 0x%08x ",
wpi_get_prph_string(prph_tbl[i]),
wpi_prph_read(sc, prph_tbl[i]));
if ((i + 1) % 2 == 0)
DPRINTF(sc, WPI_DEBUG_REGISTER, "\n");
}
DPRINTF(sc, WPI_DEBUG_REGISTER, "\n");
wpi_nic_unlock(sc);
} else {
DPRINTF(sc, WPI_DEBUG_REGISTER,
"Cannot access internal registers.\n");
}
}
#endif
/*
* Dump the error log of the firmware when a firmware panic occurs. Although
* we can't debug the firmware because it is neither open source nor free, it
* can help us to identify certain classes of problems.
*/
static void
wpi_fatal_intr(struct wpi_softc *sc)
{
struct wpi_fw_dump dump;
uint32_t i, offset, count;
/* Check that the error log address is valid. */
if (sc->errptr < WPI_FW_DATA_BASE ||
sc->errptr + sizeof (dump) >
WPI_FW_DATA_BASE + WPI_FW_DATA_MAXSZ) {
printf("%s: bad firmware error log address 0x%08x\n", __func__,
sc->errptr);
return;
}
if (wpi_nic_lock(sc) != 0) {
printf("%s: could not read firmware error log\n", __func__);
return;
}
/* Read number of entries in the log. */
count = wpi_mem_read(sc, sc->errptr);
if (count == 0 || count * sizeof (dump) > WPI_FW_DATA_MAXSZ) {
printf("%s: invalid count field (count = %u)\n", __func__,
count);
wpi_nic_unlock(sc);
return;
}
/* Skip "count" field. */
offset = sc->errptr + sizeof (uint32_t);
printf("firmware error log (count = %u):\n", count);
for (i = 0; i < count; i++) {
wpi_mem_read_region_4(sc, offset, (uint32_t *)&dump,
sizeof (dump) / sizeof (uint32_t));
printf(" error type = \"%s\" (0x%08X)\n",
(dump.desc < nitems(wpi_fw_errmsg)) ?
wpi_fw_errmsg[dump.desc] : "UNKNOWN",
dump.desc);
printf(" error data = 0x%08X\n",
dump.data);
printf(" branch link = 0x%08X%08X\n",
dump.blink[0], dump.blink[1]);
printf(" interrupt link = 0x%08X%08X\n",
dump.ilink[0], dump.ilink[1]);
printf(" time = %u\n", dump.time);
offset += sizeof (dump);
}
wpi_nic_unlock(sc);
/* Dump driver status (TX and RX rings) while we're here. */
printf("driver status:\n");
WPI_TXQ_LOCK(sc);
for (i = 0; i < WPI_DRV_NTXQUEUES; i++) {
struct wpi_tx_ring *ring = &sc->txq[i];
printf(" tx ring %2d: qid=%-2d cur=%-3d queued=%-3d\n",
i, ring->qid, ring->cur, ring->queued);
}
WPI_TXQ_UNLOCK(sc);
printf(" rx ring: cur=%d\n", sc->rxq.cur);
}
static void
wpi_intr(void *arg)
{
struct wpi_softc *sc = arg;
uint32_t r1, r2;
WPI_LOCK(sc);
/* Disable interrupts. */
WPI_WRITE(sc, WPI_INT_MASK, 0);
r1 = WPI_READ(sc, WPI_INT);
if (__predict_false(r1 == 0xffffffff ||
(r1 & 0xfffffff0) == 0xa5a5a5a0))
goto end; /* Hardware gone! */
r2 = WPI_READ(sc, WPI_FH_INT);
DPRINTF(sc, WPI_DEBUG_INTR, "%s: reg1=0x%08x reg2=0x%08x\n", __func__,
r1, r2);
if (r1 == 0 && r2 == 0)
goto done; /* Interrupt not for us. */
/* Acknowledge interrupts. */
WPI_WRITE(sc, WPI_INT, r1);
WPI_WRITE(sc, WPI_FH_INT, r2);
if (__predict_false(r1 & (WPI_INT_SW_ERR | WPI_INT_HW_ERR))) {
device_printf(sc->sc_dev, "fatal firmware error\n");
#ifdef WPI_DEBUG
wpi_debug_registers(sc);
#endif
wpi_fatal_intr(sc);
DPRINTF(sc, WPI_DEBUG_HW,
"(%s)\n", (r1 & WPI_INT_SW_ERR) ? "(Software Error)" :
"(Hardware Error)");
taskqueue_enqueue(sc->sc_tq, &sc->sc_reinittask);
goto end;
}
if ((r1 & (WPI_INT_FH_RX | WPI_INT_SW_RX)) ||
(r2 & WPI_FH_INT_RX))
wpi_notif_intr(sc);
if (r1 & WPI_INT_ALIVE)
wakeup(sc); /* Firmware is alive. */
if (r1 & WPI_INT_WAKEUP)
wpi_wakeup_intr(sc);
done:
/* Re-enable interrupts. */
if (__predict_true(sc->sc_running))
WPI_WRITE(sc, WPI_INT_MASK, WPI_INT_MASK_DEF);
end: WPI_UNLOCK(sc);
}
static int
wpi_cmd2(struct wpi_softc *sc, struct wpi_buf *buf)
{
struct ieee80211_frame *wh;
struct wpi_tx_cmd *cmd;
struct wpi_tx_data *data;
struct wpi_tx_desc *desc;
struct wpi_tx_ring *ring;
struct mbuf *m1;
bus_dma_segment_t *seg, segs[WPI_MAX_SCATTER];
uint8_t pad;
uint16_t hdrlen;
int error, i, nsegs, totlen;
WPI_TXQ_LOCK(sc);
KASSERT(buf->size <= sizeof(buf->data), ("buffer overflow"));
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
if (__predict_false(sc->sc_running == 0)) {
/* wpi_stop() was called */
error = ENETDOWN;
goto fail;
}
wh = mtod(buf->m, struct ieee80211_frame *);
hdrlen = ieee80211_anyhdrsize(wh);
totlen = buf->m->m_pkthdr.len;
if (__predict_false(totlen < sizeof(struct ieee80211_frame_min))) {
error = EINVAL;
goto fail;
}
if (hdrlen & 3) {
/* First segment length must be a multiple of 4. */
pad = 4 - (hdrlen & 3);
} else
pad = 0;
ring = &sc->txq[buf->ac];
desc = &ring->desc[ring->cur];
data = &ring->data[ring->cur];
/* Prepare TX firmware command. */
cmd = &ring->cmd[ring->cur];
cmd->code = buf->code;
cmd->flags = 0;
cmd->qid = ring->qid;
cmd->idx = ring->cur;
memcpy(cmd->data, buf->data, buf->size);
/* Save and trim IEEE802.11 header. */
memcpy((uint8_t *)(cmd->data + buf->size), wh, hdrlen);
m_adj(buf->m, hdrlen);
error = bus_dmamap_load_mbuf_sg(ring->data_dmat, data->map, buf->m,
segs, &nsegs, BUS_DMA_NOWAIT);
if (error != 0 && error != EFBIG) {
device_printf(sc->sc_dev,
"%s: can't map mbuf (error %d)\n", __func__, error);
goto fail;
}
if (error != 0) {
/* Too many DMA segments, linearize mbuf. */
m1 = m_collapse(buf->m, M_NOWAIT, WPI_MAX_SCATTER - 1);
if (m1 == NULL) {
device_printf(sc->sc_dev,
"%s: could not defrag mbuf\n", __func__);
error = ENOBUFS;
goto fail;
}
buf->m = m1;
error = bus_dmamap_load_mbuf_sg(ring->data_dmat, data->map,
buf->m, segs, &nsegs, BUS_DMA_NOWAIT);
if (__predict_false(error != 0)) {
device_printf(sc->sc_dev,
"%s: can't map mbuf (error %d)\n", __func__,
error);
goto fail;
}
}
KASSERT(nsegs < WPI_MAX_SCATTER,
("too many DMA segments, nsegs (%d) should be less than %d",
nsegs, WPI_MAX_SCATTER));
data->m = buf->m;
data->ni = buf->ni;
DPRINTF(sc, WPI_DEBUG_XMIT, "%s: qid %d idx %d len %d nsegs %d\n",
__func__, ring->qid, ring->cur, totlen, nsegs);
/* Fill TX descriptor. */
desc->nsegs = WPI_PAD32(totlen + pad) << 4 | (1 + nsegs);
/* First DMA segment is used by the TX command. */
desc->segs[0].addr = htole32(data->cmd_paddr);
desc->segs[0].len = htole32(4 + buf->size + hdrlen + pad);
/* Other DMA segments are for data payload. */
seg = &segs[0];
for (i = 1; i <= nsegs; i++) {
desc->segs[i].addr = htole32(seg->ds_addr);
desc->segs[i].len = htole32(seg->ds_len);
seg++;
}
bus_dmamap_sync(ring->data_dmat, data->map, BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(ring->data_dmat, ring->cmd_dma.map,
BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(ring->desc_dma.tag, ring->desc_dma.map,
BUS_DMASYNC_PREWRITE);
/* Kick TX ring. */
ring->cur = (ring->cur + 1) % WPI_TX_RING_COUNT;
sc->sc_update_tx_ring(sc, ring);
if (ring->qid < WPI_CMD_QUEUE_NUM) {
WPI_TXQ_STATE_LOCK(sc);
ring->queued++;
callout_reset(&sc->tx_timeout, 5*hz, wpi_tx_timeout, sc);
WPI_TXQ_STATE_UNLOCK(sc);
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
WPI_TXQ_UNLOCK(sc);
return 0;
fail: m_freem(buf->m);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
WPI_TXQ_UNLOCK(sc);
return error;
}
/*
* Construct the data packet for a transmit buffer.
*/
static int
wpi_tx_data(struct wpi_softc *sc, struct mbuf *m, struct ieee80211_node *ni)
{
const struct ieee80211_txparam *tp;
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211com *ic = ni->ni_ic;
struct wpi_node *wn = WPI_NODE(ni);
struct ieee80211_channel *chan;
struct ieee80211_frame *wh;
struct ieee80211_key *k = NULL;
struct wpi_buf tx_data;
struct wpi_cmd_data *tx = (struct wpi_cmd_data *)&tx_data.data;
uint32_t flags;
uint16_t ac, qos;
uint8_t tid, type, rate;
int error, swcrypt, ismcast, totlen;
wh = mtod(m, struct ieee80211_frame *);
type = wh->i_fc[0] & IEEE80211_FC0_TYPE_MASK;
ismcast = IEEE80211_IS_MULTICAST(wh->i_addr1);
swcrypt = 1;
/* Select EDCA Access Category and TX ring for this frame. */
if (IEEE80211_QOS_HAS_SEQ(wh)) {
qos = ((const struct ieee80211_qosframe *)wh)->i_qos[0];
tid = qos & IEEE80211_QOS_TID;
} else {
qos = 0;
tid = 0;
}
ac = M_WME_GETAC(m);
chan = (ni->ni_chan != IEEE80211_CHAN_ANYC) ?
ni->ni_chan : ic->ic_curchan;
tp = &vap->iv_txparms[ieee80211_chan2mode(chan)];
/* Choose a TX rate index. */
if (type == IEEE80211_FC0_TYPE_MGT)
rate = tp->mgmtrate;
else if (ismcast)
rate = tp->mcastrate;
else if (tp->ucastrate != IEEE80211_FIXED_RATE_NONE)
rate = tp->ucastrate;
else if (m->m_flags & M_EAPOL)
rate = tp->mgmtrate;
else {
/* XXX pass pktlen */
(void) ieee80211_ratectl_rate(ni, NULL, 0);
rate = ni->ni_txrate;
}
/* Encrypt the frame if need be. */
if (wh->i_fc[1] & IEEE80211_FC1_PROTECTED) {
/* Retrieve key for TX. */
k = ieee80211_crypto_encap(ni, m);
if (k == NULL) {
error = ENOBUFS;
goto fail;
}
swcrypt = k->wk_flags & IEEE80211_KEY_SWCRYPT;
/* 802.11 header may have moved. */
wh = mtod(m, struct ieee80211_frame *);
}
totlen = m->m_pkthdr.len;
if (ieee80211_radiotap_active_vap(vap)) {
struct wpi_tx_radiotap_header *tap = &sc->sc_txtap;
tap->wt_flags = 0;
tap->wt_rate = rate;
if (k != NULL)
tap->wt_flags |= IEEE80211_RADIOTAP_F_WEP;
ieee80211_radiotap_tx(vap, m);
}
flags = 0;
if (!ismcast) {
/* Unicast frame, check if an ACK is expected. */
if (!qos || (qos & IEEE80211_QOS_ACKPOLICY) !=
IEEE80211_QOS_ACKPOLICY_NOACK)
flags |= WPI_TX_NEED_ACK;
}
if (!IEEE80211_QOS_HAS_SEQ(wh))
flags |= WPI_TX_AUTO_SEQ;
if (wh->i_fc[1] & IEEE80211_FC1_MORE_FRAG)
flags |= WPI_TX_MORE_FRAG; /* Cannot happen yet. */
/* Check if frame must be protected using RTS/CTS or CTS-to-self. */
if (!ismcast) {
/* NB: Group frames are sent using CCK in 802.11b/g. */
if (totlen + IEEE80211_CRC_LEN > vap->iv_rtsthreshold) {
flags |= WPI_TX_NEED_RTS;
} else if ((ic->ic_flags & IEEE80211_F_USEPROT) &&
WPI_RATE_IS_OFDM(rate)) {
if (ic->ic_protmode == IEEE80211_PROT_CTSONLY)
flags |= WPI_TX_NEED_CTS;
else if (ic->ic_protmode == IEEE80211_PROT_RTSCTS)
flags |= WPI_TX_NEED_RTS;
}
if (flags & (WPI_TX_NEED_RTS | WPI_TX_NEED_CTS))
flags |= WPI_TX_FULL_TXOP;
}
memset(tx, 0, sizeof (struct wpi_cmd_data));
if (type == IEEE80211_FC0_TYPE_MGT) {
uint8_t subtype = wh->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK;
/* Tell HW to set timestamp in probe responses. */
if (subtype == IEEE80211_FC0_SUBTYPE_PROBE_RESP)
flags |= WPI_TX_INSERT_TSTAMP;
if (subtype == IEEE80211_FC0_SUBTYPE_ASSOC_REQ ||
subtype == IEEE80211_FC0_SUBTYPE_REASSOC_REQ)
tx->timeout = htole16(3);
else
tx->timeout = htole16(2);
}
if (ismcast || type != IEEE80211_FC0_TYPE_DATA)
tx->id = WPI_ID_BROADCAST;
else {
if (wn->id == WPI_ID_UNDEFINED) {
device_printf(sc->sc_dev,
"%s: undefined node id\n", __func__);
error = EINVAL;
goto fail;
}
tx->id = wn->id;
}
if (!swcrypt) {
switch (k->wk_cipher->ic_cipher) {
case IEEE80211_CIPHER_AES_CCM:
tx->security = WPI_CIPHER_CCMP;
break;
default:
break;
}
memcpy(tx->key, k->wk_key, k->wk_keylen);
}
tx->len = htole16(totlen);
tx->flags = htole32(flags);
tx->plcp = rate2plcp(rate);
tx->tid = tid;
tx->lifetime = htole32(WPI_LIFETIME_INFINITE);
tx->ofdm_mask = 0xff;
tx->cck_mask = 0x0f;
tx->rts_ntries = 7;
tx->data_ntries = tp->maxretry;
tx_data.ni = ni;
tx_data.m = m;
tx_data.size = sizeof(struct wpi_cmd_data);
tx_data.code = WPI_CMD_TX_DATA;
tx_data.ac = ac;
return wpi_cmd2(sc, &tx_data);
fail: m_freem(m);
return error;
}
static int
wpi_tx_data_raw(struct wpi_softc *sc, struct mbuf *m,
struct ieee80211_node *ni, const struct ieee80211_bpf_params *params)
{
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211_key *k = NULL;
struct ieee80211_frame *wh;
struct wpi_buf tx_data;
struct wpi_cmd_data *tx = (struct wpi_cmd_data *)&tx_data.data;
uint32_t flags;
uint8_t ac, type, rate;
int swcrypt, totlen;
wh = mtod(m, struct ieee80211_frame *);
type = wh->i_fc[0] & IEEE80211_FC0_TYPE_MASK;
swcrypt = 1;
ac = params->ibp_pri & 3;
/* Choose a TX rate index. */
rate = params->ibp_rate0;
flags = 0;
if (!IEEE80211_QOS_HAS_SEQ(wh))
flags |= WPI_TX_AUTO_SEQ;
if ((params->ibp_flags & IEEE80211_BPF_NOACK) == 0)
flags |= WPI_TX_NEED_ACK;
if (params->ibp_flags & IEEE80211_BPF_RTS)
flags |= WPI_TX_NEED_RTS;
if (params->ibp_flags & IEEE80211_BPF_CTS)
flags |= WPI_TX_NEED_CTS;
if (flags & (WPI_TX_NEED_RTS | WPI_TX_NEED_CTS))
flags |= WPI_TX_FULL_TXOP;
/* Encrypt the frame if need be. */
if (params->ibp_flags & IEEE80211_BPF_CRYPTO) {
/* Retrieve key for TX. */
k = ieee80211_crypto_encap(ni, m);
if (k == NULL) {
m_freem(m);
return ENOBUFS;
}
swcrypt = k->wk_flags & IEEE80211_KEY_SWCRYPT;
/* 802.11 header may have moved. */
wh = mtod(m, struct ieee80211_frame *);
}
totlen = m->m_pkthdr.len;
if (ieee80211_radiotap_active_vap(vap)) {
struct wpi_tx_radiotap_header *tap = &sc->sc_txtap;
tap->wt_flags = 0;
tap->wt_rate = rate;
if (params->ibp_flags & IEEE80211_BPF_CRYPTO)
tap->wt_flags |= IEEE80211_RADIOTAP_F_WEP;
ieee80211_radiotap_tx(vap, m);
}
memset(tx, 0, sizeof (struct wpi_cmd_data));
if (type == IEEE80211_FC0_TYPE_MGT) {
uint8_t subtype = wh->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK;
/* Tell HW to set timestamp in probe responses. */
if (subtype == IEEE80211_FC0_SUBTYPE_PROBE_RESP)
flags |= WPI_TX_INSERT_TSTAMP;
if (subtype == IEEE80211_FC0_SUBTYPE_ASSOC_REQ ||
subtype == IEEE80211_FC0_SUBTYPE_REASSOC_REQ)
tx->timeout = htole16(3);
else
tx->timeout = htole16(2);
}
if (!swcrypt) {
switch (k->wk_cipher->ic_cipher) {
case IEEE80211_CIPHER_AES_CCM:
tx->security = WPI_CIPHER_CCMP;
break;
default:
break;
}
memcpy(tx->key, k->wk_key, k->wk_keylen);
}
tx->len = htole16(totlen);
tx->flags = htole32(flags);
tx->plcp = rate2plcp(rate);
tx->id = WPI_ID_BROADCAST;
tx->lifetime = htole32(WPI_LIFETIME_INFINITE);
tx->rts_ntries = params->ibp_try1;
tx->data_ntries = params->ibp_try0;
tx_data.ni = ni;
tx_data.m = m;
tx_data.size = sizeof(struct wpi_cmd_data);
tx_data.code = WPI_CMD_TX_DATA;
tx_data.ac = ac;
return wpi_cmd2(sc, &tx_data);
}
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static __inline int
wpi_tx_ring_is_full(struct wpi_softc *sc, uint16_t ac)
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
{
struct wpi_tx_ring *ring = &sc->txq[ac];
int retval;
WPI_TXQ_STATE_LOCK(sc);
retval = (ring->queued > WPI_TX_RING_HIMARK);
WPI_TXQ_STATE_UNLOCK(sc);
return retval;
}
static __inline void
wpi_handle_tx_failure(struct ieee80211_node *ni)
{
/* NB: m is reclaimed on tx failure */
if_inc_counter(ni->ni_vap->iv_ifp, IFCOUNTER_OERRORS, 1);
ieee80211_free_node(ni);
}
static int
wpi_raw_xmit(struct ieee80211_node *ni, struct mbuf *m,
const struct ieee80211_bpf_params *params)
{
struct ieee80211com *ic = ni->ni_ic;
struct wpi_softc *sc = ic->ic_softc;
uint16_t ac;
int error = 0;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ac = M_WME_GETAC(m);
WPI_TX_LOCK(sc);
if (sc->sc_running == 0 || wpi_tx_ring_is_full(sc, ac)) {
m_freem(m);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
error = sc->sc_running ? ENOBUFS : ENETDOWN;
goto unlock;
}
if (params == NULL) {
/*
* Legacy path; interpret frame contents to decide
* precisely how to send the frame.
*/
error = wpi_tx_data(sc, m, ni);
} else {
/*
* Caller supplied explicit parameters to use in
* sending the frame.
*/
error = wpi_tx_data_raw(sc, m, ni, params);
}
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
unlock: WPI_TX_UNLOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (error != 0) {
wpi_handle_tx_failure(ni);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static int
wpi_transmit(struct ieee80211com *ic, struct mbuf *m)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct wpi_softc *sc = ic->ic_softc;
struct ieee80211_node *ni;
uint16_t ac;
int error;
WPI_TX_LOCK(sc);
DPRINTF(sc, WPI_DEBUG_XMIT, "%s: called\n", __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
/* Check if interface is up & running. */
if (__predict_false(sc->sc_running == 0)) {
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
error = ENXIO;
goto unlock;
}
Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Details here: https://wiki.freebsd.org/projects/ifnet/net80211 Still, drivers: ndis, wtap, mwl, ipw, bwn, wi, upgt, uath were not tested. Changes to mwl, ipw, bwn, wi, upgt are trivial and chances of problems are low. The wtap wasn't compilable even before this change. But the ndis driver is complex, and it is likely to be broken with this commit. Help with testing and debugging it is appreciated. Differential Revision: D2655, D2740 Sponsored by: Nginx, Inc. Sponsored by: Netflix
2015-08-07 11:43:14 +00:00
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
/* Check for available space. */
ac = M_WME_GETAC(m);
if (wpi_tx_ring_is_full(sc, ac)) {
error = ENOBUFS;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
goto unlock;
}
error = 0;
ni = (struct ieee80211_node *)m->m_pkthdr.rcvif;
if (wpi_tx_data(sc, m, ni) != 0) {
wpi_handle_tx_failure(ni);
}
DPRINTF(sc, WPI_DEBUG_XMIT, "%s: done\n", __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
unlock: WPI_TX_UNLOCK(sc);
return (error);
}
static void
wpi_watchdog_rfkill(void *arg)
{
struct wpi_softc *sc = arg;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
DPRINTF(sc, WPI_DEBUG_WATCHDOG, "RFkill Watchdog: tick\n");
/* No need to lock firmware memory. */
if ((wpi_prph_read(sc, WPI_APMG_RFKILL) & 0x1) == 0) {
/* Radio kill switch is still off. */
callout_reset(&sc->watchdog_rfkill, hz, wpi_watchdog_rfkill,
sc);
} else
ieee80211_runtask(ic, &sc->sc_radioon_task);
}
static void
wpi_scan_timeout(void *arg)
{
struct wpi_softc *sc = arg;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ic_printf(ic, "scan timeout\n");
taskqueue_enqueue(sc->sc_tq, &sc->sc_reinittask);
}
static void
wpi_tx_timeout(void *arg)
{
struct wpi_softc *sc = arg;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ic_printf(ic, "device timeout\n");
taskqueue_enqueue(sc->sc_tq, &sc->sc_reinittask);
}
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static void
wpi_parent(struct ieee80211com *ic)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct wpi_softc *sc = ic->ic_softc;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (ic->ic_nrunning > 0) {
if (wpi_init(sc) == 0) {
ieee80211_notify_radio(ic, 1);
ieee80211_start_all(ic);
} else {
ieee80211_notify_radio(ic, 0);
ieee80211_stop(vap);
}
} else
wpi_stop(sc);
}
/*
* Send a command to the firmware.
*/
static int
wpi_cmd(struct wpi_softc *sc, uint8_t code, const void *buf, uint16_t size,
int async)
{
struct wpi_tx_ring *ring = &sc->txq[WPI_CMD_QUEUE_NUM];
struct wpi_tx_desc *desc;
struct wpi_tx_data *data;
struct wpi_tx_cmd *cmd;
struct mbuf *m;
bus_addr_t paddr;
uint16_t totlen;
int error;
WPI_TXQ_LOCK(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
if (__predict_false(sc->sc_running == 0)) {
/* wpi_stop() was called */
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (code == WPI_CMD_SCAN)
error = ENETDOWN;
else
error = 0;
goto fail;
}
if (async == 0)
WPI_LOCK_ASSERT(sc);
DPRINTF(sc, WPI_DEBUG_CMD, "%s: cmd %s size %u async %d\n",
__func__, wpi_cmd_str(code), size, async);
desc = &ring->desc[ring->cur];
data = &ring->data[ring->cur];
totlen = 4 + size;
if (size > sizeof cmd->data) {
/* Command is too large to fit in a descriptor. */
if (totlen > MCLBYTES) {
error = EINVAL;
goto fail;
}
m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUMPAGESIZE);
if (m == NULL) {
error = ENOMEM;
goto fail;
}
cmd = mtod(m, struct wpi_tx_cmd *);
error = bus_dmamap_load(ring->data_dmat, data->map, cmd,
totlen, wpi_dma_map_addr, &paddr, BUS_DMA_NOWAIT);
if (error != 0) {
m_freem(m);
goto fail;
}
data->m = m;
} else {
cmd = &ring->cmd[ring->cur];
paddr = data->cmd_paddr;
}
cmd->code = code;
cmd->flags = 0;
cmd->qid = ring->qid;
cmd->idx = ring->cur;
memcpy(cmd->data, buf, size);
desc->nsegs = 1 + (WPI_PAD32(size) << 4);
desc->segs[0].addr = htole32(paddr);
desc->segs[0].len = htole32(totlen);
if (size > sizeof cmd->data) {
bus_dmamap_sync(ring->data_dmat, data->map,
BUS_DMASYNC_PREWRITE);
} else {
bus_dmamap_sync(ring->data_dmat, ring->cmd_dma.map,
BUS_DMASYNC_PREWRITE);
}
bus_dmamap_sync(ring->desc_dma.tag, ring->desc_dma.map,
BUS_DMASYNC_PREWRITE);
/* Kick command ring. */
ring->cur = (ring->cur + 1) % WPI_TX_RING_COUNT;
sc->sc_update_tx_ring(sc, ring);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
WPI_TXQ_UNLOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
return async ? 0 : mtx_sleep(cmd, &sc->sc_mtx, PCATCH, "wpicmd", hz);
fail: DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
WPI_TXQ_UNLOCK(sc);
return error;
}
/*
* Configure HW multi-rate retries.
*/
static int
wpi_mrr_setup(struct wpi_softc *sc)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct wpi_mrr_setup mrr;
uint8_t i;
int error;
/* CCK rates (not used with 802.11a). */
for (i = WPI_RIDX_CCK1; i <= WPI_RIDX_CCK11; i++) {
mrr.rates[i].flags = 0;
mrr.rates[i].plcp = wpi_ridx_to_plcp[i];
/* Fallback to the immediate lower CCK rate (if any.) */
mrr.rates[i].next =
(i == WPI_RIDX_CCK1) ? WPI_RIDX_CCK1 : i - 1;
/* Try twice at this rate before falling back to "next". */
mrr.rates[i].ntries = WPI_NTRIES_DEFAULT;
}
/* OFDM rates (not used with 802.11b). */
for (i = WPI_RIDX_OFDM6; i <= WPI_RIDX_OFDM54; i++) {
mrr.rates[i].flags = 0;
mrr.rates[i].plcp = wpi_ridx_to_plcp[i];
/* Fallback to the immediate lower rate (if any.) */
/* We allow fallback from OFDM/6 to CCK/2 in 11b/g mode. */
mrr.rates[i].next = (i == WPI_RIDX_OFDM6) ?
((ic->ic_curmode == IEEE80211_MODE_11A) ?
WPI_RIDX_OFDM6 : WPI_RIDX_CCK2) :
i - 1;
/* Try twice at this rate before falling back to "next". */
mrr.rates[i].ntries = WPI_NTRIES_DEFAULT;
}
/* Setup MRR for control frames. */
mrr.which = htole32(WPI_MRR_CTL);
error = wpi_cmd(sc, WPI_CMD_MRR_SETUP, &mrr, sizeof mrr, 0);
if (error != 0) {
device_printf(sc->sc_dev,
"could not setup MRR for control frames\n");
return error;
}
/* Setup MRR for data frames. */
mrr.which = htole32(WPI_MRR_DATA);
error = wpi_cmd(sc, WPI_CMD_MRR_SETUP, &mrr, sizeof mrr, 0);
if (error != 0) {
device_printf(sc->sc_dev,
"could not setup MRR for data frames\n");
return error;
}
return 0;
}
static int
wpi_add_node(struct wpi_softc *sc, struct ieee80211_node *ni)
{
struct ieee80211com *ic = ni->ni_ic;
struct wpi_vap *wvp = WPI_VAP(ni->ni_vap);
struct wpi_node *wn = WPI_NODE(ni);
struct wpi_node_info node;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (wn->id == WPI_ID_UNDEFINED)
return EINVAL;
memset(&node, 0, sizeof node);
IEEE80211_ADDR_COPY(node.macaddr, ni->ni_macaddr);
node.id = wn->id;
node.plcp = (ic->ic_curmode == IEEE80211_MODE_11A) ?
wpi_ridx_to_plcp[WPI_RIDX_OFDM6] : wpi_ridx_to_plcp[WPI_RIDX_CCK1];
node.action = htole32(WPI_ACTION_SET_RATE);
node.antenna = WPI_ANTENNA_BOTH;
DPRINTF(sc, WPI_DEBUG_NODE, "%s: adding node %d (%s)\n", __func__,
wn->id, ether_sprintf(ni->ni_macaddr));
error = wpi_cmd(sc, WPI_CMD_ADD_NODE, &node, sizeof node, 1);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: wpi_cmd() call failed with error code %d\n", __func__,
error);
return error;
}
if (wvp->wv_gtk != 0) {
error = wpi_set_global_keys(ni);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: error while setting global keys\n", __func__);
return ENXIO;
}
}
return 0;
}
/*
* Broadcast node is used to send group-addressed and management frames.
*/
static int
wpi_add_broadcast_node(struct wpi_softc *sc, int async)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct wpi_node_info node;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
memset(&node, 0, sizeof node);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
IEEE80211_ADDR_COPY(node.macaddr, ieee80211broadcastaddr);
node.id = WPI_ID_BROADCAST;
node.plcp = (ic->ic_curmode == IEEE80211_MODE_11A) ?
wpi_ridx_to_plcp[WPI_RIDX_OFDM6] : wpi_ridx_to_plcp[WPI_RIDX_CCK1];
node.action = htole32(WPI_ACTION_SET_RATE);
node.antenna = WPI_ANTENNA_BOTH;
DPRINTF(sc, WPI_DEBUG_NODE, "%s: adding broadcast node\n", __func__);
return wpi_cmd(sc, WPI_CMD_ADD_NODE, &node, sizeof node, async);
}
static int
wpi_add_sta_node(struct wpi_softc *sc, struct ieee80211_node *ni)
{
struct wpi_node *wn = WPI_NODE(ni);
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
wn->id = wpi_add_node_entry_sta(sc);
if ((error = wpi_add_node(sc, ni)) != 0) {
wpi_del_node_entry(sc, wn->id);
wn->id = WPI_ID_UNDEFINED;
return error;
}
return 0;
}
static int
wpi_add_ibss_node(struct wpi_softc *sc, struct ieee80211_node *ni)
{
struct wpi_node *wn = WPI_NODE(ni);
int error;
KASSERT(wn->id == WPI_ID_UNDEFINED,
("the node %d was added before", wn->id));
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if ((wn->id = wpi_add_node_entry_adhoc(sc)) == WPI_ID_UNDEFINED) {
device_printf(sc->sc_dev, "%s: h/w table is full\n", __func__);
return ENOMEM;
}
if ((error = wpi_add_node(sc, ni)) != 0) {
wpi_del_node_entry(sc, wn->id);
wn->id = WPI_ID_UNDEFINED;
return error;
}
return 0;
}
static void
wpi_del_node(struct wpi_softc *sc, struct ieee80211_node *ni)
{
struct wpi_node *wn = WPI_NODE(ni);
struct wpi_cmd_del_node node;
int error;
KASSERT(wn->id != WPI_ID_UNDEFINED, ("undefined node id passed"));
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
memset(&node, 0, sizeof node);
IEEE80211_ADDR_COPY(node.macaddr, ni->ni_macaddr);
node.count = 1;
DPRINTF(sc, WPI_DEBUG_NODE, "%s: deleting node %d (%s)\n", __func__,
wn->id, ether_sprintf(ni->ni_macaddr));
error = wpi_cmd(sc, WPI_CMD_DEL_NODE, &node, sizeof node, 1);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not delete node %u, error %d\n", __func__,
wn->id, error);
}
}
static int
wpi_updateedca(struct ieee80211com *ic)
{
#define WPI_EXP2(x) ((1 << (x)) - 1) /* CWmin = 2^ECWmin - 1 */
struct wpi_softc *sc = ic->ic_softc;
struct wpi_edca_params cmd;
int aci, error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
memset(&cmd, 0, sizeof cmd);
cmd.flags = htole32(WPI_EDCA_UPDATE);
for (aci = 0; aci < WME_NUM_AC; aci++) {
const struct wmeParams *ac =
&ic->ic_wme.wme_chanParams.cap_wmeParams[aci];
cmd.ac[aci].aifsn = ac->wmep_aifsn;
cmd.ac[aci].cwmin = htole16(WPI_EXP2(ac->wmep_logcwmin));
cmd.ac[aci].cwmax = htole16(WPI_EXP2(ac->wmep_logcwmax));
cmd.ac[aci].txoplimit =
htole16(IEEE80211_TXOP_TO_US(ac->wmep_txopLimit));
DPRINTF(sc, WPI_DEBUG_EDCA,
"setting WME for queue %d aifsn=%d cwmin=%d cwmax=%d "
"txoplimit=%d\n", aci, cmd.ac[aci].aifsn,
cmd.ac[aci].cwmin, cmd.ac[aci].cwmax,
cmd.ac[aci].txoplimit);
}
error = wpi_cmd(sc, WPI_CMD_EDCA_PARAMS, &cmd, sizeof cmd, 1);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return error;
#undef WPI_EXP2
}
static void
wpi_set_promisc(struct wpi_softc *sc)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
uint32_t promisc_filter;
promisc_filter = WPI_FILTER_CTL;
if (vap != NULL && vap->iv_opmode != IEEE80211_M_HOSTAP)
promisc_filter |= WPI_FILTER_PROMISC;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (ic->ic_promisc > 0)
sc->rxon.filter |= htole32(promisc_filter);
else
sc->rxon.filter &= ~htole32(promisc_filter);
}
static void
wpi_update_promisc(struct ieee80211com *ic)
{
struct wpi_softc *sc = ic->ic_softc;
WPI_RXON_LOCK(sc);
wpi_set_promisc(sc);
if (wpi_send_rxon(sc, 1, 1) != 0) {
device_printf(sc->sc_dev, "%s: could not send RXON\n",
__func__);
}
WPI_RXON_UNLOCK(sc);
}
static void
wpi_update_mcast(struct ieee80211com *ic)
{
/* Ignore */
}
static void
wpi_set_led(struct wpi_softc *sc, uint8_t which, uint8_t off, uint8_t on)
{
struct wpi_cmd_led led;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
led.which = which;
led.unit = htole32(100000); /* on/off in unit of 100ms */
led.off = off;
led.on = on;
(void)wpi_cmd(sc, WPI_CMD_SET_LED, &led, sizeof led, 1);
}
static int
wpi_set_timing(struct wpi_softc *sc, struct ieee80211_node *ni)
{
struct wpi_cmd_timing cmd;
uint64_t val, mod;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
memset(&cmd, 0, sizeof cmd);
memcpy(&cmd.tstamp, ni->ni_tstamp.data, sizeof (uint64_t));
cmd.bintval = htole16(ni->ni_intval);
cmd.lintval = htole16(10);
/* Compute remaining time until next beacon. */
val = (uint64_t)ni->ni_intval * IEEE80211_DUR_TU;
mod = le64toh(cmd.tstamp) % val;
cmd.binitval = htole32((uint32_t)(val - mod));
DPRINTF(sc, WPI_DEBUG_RESET, "timing bintval=%u tstamp=%ju, init=%u\n",
ni->ni_intval, le64toh(cmd.tstamp), (uint32_t)(val - mod));
return wpi_cmd(sc, WPI_CMD_TIMING, &cmd, sizeof cmd, 1);
}
/*
* This function is called periodically (every 60 seconds) to adjust output
* power to temperature changes.
*/
static void
wpi_power_calibration(struct wpi_softc *sc)
{
int temp;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
/* Update sensor data. */
temp = (int)WPI_READ(sc, WPI_UCODE_GP2);
DPRINTF(sc, WPI_DEBUG_TEMP, "Temp in calibration is: %d\n", temp);
/* Sanity-check read value. */
if (temp < -260 || temp > 25) {
/* This can't be correct, ignore. */
DPRINTF(sc, WPI_DEBUG_TEMP,
"out-of-range temperature reported: %d\n", temp);
return;
}
DPRINTF(sc, WPI_DEBUG_TEMP, "temperature %d->%d\n", sc->temp, temp);
/* Adjust Tx power if need be. */
if (abs(temp - sc->temp) <= 6)
return;
sc->temp = temp;
if (wpi_set_txpower(sc, 1) != 0) {
/* just warn, too bad for the automatic calibration... */
device_printf(sc->sc_dev,"could not adjust Tx power\n");
}
}
/*
* Set TX power for current channel.
*/
static int
wpi_set_txpower(struct wpi_softc *sc, int async)
{
struct wpi_power_group *group;
struct wpi_cmd_txpower cmd;
uint8_t chan;
int idx, is_chan_5ghz, i;
/* Retrieve current channel from last RXON. */
chan = sc->rxon.chan;
is_chan_5ghz = (sc->rxon.flags & htole32(WPI_RXON_24GHZ)) == 0;
/* Find the TX power group to which this channel belongs. */
if (is_chan_5ghz) {
for (group = &sc->groups[1]; group < &sc->groups[4]; group++)
if (chan <= group->chan)
break;
} else
group = &sc->groups[0];
memset(&cmd, 0, sizeof cmd);
cmd.band = is_chan_5ghz ? WPI_BAND_5GHZ : WPI_BAND_2GHZ;
cmd.chan = htole16(chan);
/* Set TX power for all OFDM and CCK rates. */
for (i = 0; i <= WPI_RIDX_MAX ; i++) {
/* Retrieve TX power for this channel/rate. */
idx = wpi_get_power_index(sc, group, chan, is_chan_5ghz, i);
cmd.rates[i].plcp = wpi_ridx_to_plcp[i];
if (is_chan_5ghz) {
cmd.rates[i].rf_gain = wpi_rf_gain_5ghz[idx];
cmd.rates[i].dsp_gain = wpi_dsp_gain_5ghz[idx];
} else {
cmd.rates[i].rf_gain = wpi_rf_gain_2ghz[idx];
cmd.rates[i].dsp_gain = wpi_dsp_gain_2ghz[idx];
}
DPRINTF(sc, WPI_DEBUG_TEMP,
"chan %d/ridx %d: power index %d\n", chan, i, idx);
}
return wpi_cmd(sc, WPI_CMD_TXPOWER, &cmd, sizeof cmd, async);
}
/*
* Determine Tx power index for a given channel/rate combination.
* This takes into account the regulatory information from EEPROM and the
* current temperature.
*/
static int
wpi_get_power_index(struct wpi_softc *sc, struct wpi_power_group *group,
uint8_t chan, int is_chan_5ghz, int ridx)
{
/* Fixed-point arithmetic division using a n-bit fractional part. */
#define fdivround(a, b, n) \
((((1 << n) * (a)) / (b) + (1 << n) / 2) / (1 << n))
/* Linear interpolation. */
#define interpolate(x, x1, y1, x2, y2, n) \
((y1) + fdivround(((x) - (x1)) * ((y2) - (y1)), (x2) - (x1), n))
struct wpi_power_sample *sample;
int pwr, idx;
/* Default TX power is group maximum TX power minus 3dB. */
pwr = group->maxpwr / 2;
/* Decrease TX power for highest OFDM rates to reduce distortion. */
switch (ridx) {
case WPI_RIDX_OFDM36:
pwr -= is_chan_5ghz ? 5 : 0;
break;
case WPI_RIDX_OFDM48:
pwr -= is_chan_5ghz ? 10 : 7;
break;
case WPI_RIDX_OFDM54:
pwr -= is_chan_5ghz ? 12 : 9;
break;
}
/* Never exceed the channel maximum allowed TX power. */
pwr = min(pwr, sc->maxpwr[chan]);
/* Retrieve TX power index into gain tables from samples. */
for (sample = group->samples; sample < &group->samples[3]; sample++)
if (pwr > sample[1].power)
break;
/* Fixed-point linear interpolation using a 19-bit fractional part. */
idx = interpolate(pwr, sample[0].power, sample[0].index,
sample[1].power, sample[1].index, 19);
/*-
* Adjust power index based on current temperature:
* - if cooler than factory-calibrated: decrease output power
* - if warmer than factory-calibrated: increase output power
*/
idx -= (sc->temp - group->temp) * 11 / 100;
/* Decrease TX power for CCK rates (-5dB). */
if (ridx >= WPI_RIDX_CCK1)
idx += 10;
/* Make sure idx stays in a valid range. */
if (idx < 0)
return 0;
if (idx > WPI_MAX_PWR_INDEX)
return WPI_MAX_PWR_INDEX;
return idx;
#undef interpolate
#undef fdivround
}
/*
* Set STA mode power saving level (between 0 and 5).
* Level 0 is CAM (Continuously Aware Mode), 5 is for maximum power saving.
*/
static int
wpi_set_pslevel(struct wpi_softc *sc, uint8_t dtim, int level, int async)
{
struct wpi_pmgt_cmd cmd;
const struct wpi_pmgt *pmgt;
uint32_t max, reg;
uint8_t skip_dtim;
int i;
DPRINTF(sc, WPI_DEBUG_PWRSAVE,
"%s: dtim=%d, level=%d, async=%d\n",
__func__, dtim, level, async);
/* Select which PS parameters to use. */
if (dtim <= 10)
pmgt = &wpi_pmgt[0][level];
else
pmgt = &wpi_pmgt[1][level];
memset(&cmd, 0, sizeof cmd);
if (level != 0) /* not CAM */
cmd.flags |= htole16(WPI_PS_ALLOW_SLEEP);
/* Retrieve PCIe Active State Power Management (ASPM). */
reg = pci_read_config(sc->sc_dev, sc->sc_cap_off + 0x10, 1);
if (!(reg & 0x1)) /* L0s Entry disabled. */
cmd.flags |= htole16(WPI_PS_PCI_PMGT);
cmd.rxtimeout = htole32(pmgt->rxtimeout * IEEE80211_DUR_TU);
cmd.txtimeout = htole32(pmgt->txtimeout * IEEE80211_DUR_TU);
if (dtim == 0) {
dtim = 1;
skip_dtim = 0;
} else
skip_dtim = pmgt->skip_dtim;
if (skip_dtim != 0) {
cmd.flags |= htole16(WPI_PS_SLEEP_OVER_DTIM);
max = pmgt->intval[4];
if (max == (uint32_t)-1)
max = dtim * (skip_dtim + 1);
else if (max > dtim)
max = (max / dtim) * dtim;
} else
max = dtim;
for (i = 0; i < 5; i++)
cmd.intval[i] = htole32(MIN(max, pmgt->intval[i]));
return wpi_cmd(sc, WPI_CMD_SET_POWER_MODE, &cmd, sizeof cmd, async);
}
static int
wpi_send_btcoex(struct wpi_softc *sc)
{
struct wpi_bluetooth cmd;
memset(&cmd, 0, sizeof cmd);
cmd.flags = WPI_BT_COEX_MODE_4WIRE;
cmd.lead_time = WPI_BT_LEAD_TIME_DEF;
cmd.max_kill = WPI_BT_MAX_KILL_DEF;
DPRINTF(sc, WPI_DEBUG_RESET, "%s: configuring bluetooth coexistence\n",
__func__);
return wpi_cmd(sc, WPI_CMD_BT_COEX, &cmd, sizeof(cmd), 0);
}
static int
wpi_send_rxon(struct wpi_softc *sc, int assoc, int async)
{
int error;
if (async)
WPI_RXON_LOCK_ASSERT(sc);
if (assoc && wpi_check_bss_filter(sc) != 0) {
struct wpi_assoc rxon_assoc;
rxon_assoc.flags = sc->rxon.flags;
rxon_assoc.filter = sc->rxon.filter;
rxon_assoc.ofdm_mask = sc->rxon.ofdm_mask;
rxon_assoc.cck_mask = sc->rxon.cck_mask;
rxon_assoc.reserved = 0;
error = wpi_cmd(sc, WPI_CMD_RXON_ASSOC, &rxon_assoc,
sizeof (struct wpi_assoc), async);
if (error != 0) {
device_printf(sc->sc_dev,
"RXON_ASSOC command failed, error %d\n", error);
return error;
}
} else {
if (async) {
WPI_NT_LOCK(sc);
error = wpi_cmd(sc, WPI_CMD_RXON, &sc->rxon,
sizeof (struct wpi_rxon), async);
if (error == 0)
wpi_clear_node_table(sc);
WPI_NT_UNLOCK(sc);
} else {
error = wpi_cmd(sc, WPI_CMD_RXON, &sc->rxon,
sizeof (struct wpi_rxon), async);
if (error == 0)
wpi_clear_node_table(sc);
}
if (error != 0) {
device_printf(sc->sc_dev,
"RXON command failed, error %d\n", error);
return error;
}
/* Add broadcast node. */
error = wpi_add_broadcast_node(sc, async);
if (error != 0) {
device_printf(sc->sc_dev,
"could not add broadcast node, error %d\n", error);
return error;
}
}
/* Configuration has changed, set Tx power accordingly. */
if ((error = wpi_set_txpower(sc, async)) != 0) {
device_printf(sc->sc_dev,
"%s: could not set TX power, error %d\n", __func__, error);
return error;
}
return 0;
}
/**
* Configure the card to listen to a particular channel, this transisions the
* card in to being able to receive frames from remote devices.
*/
static int
wpi_config(struct wpi_softc *sc)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
struct ieee80211_channel *c = ic->ic_curchan;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/* Set power saving level to CAM during initialization. */
if ((error = wpi_set_pslevel(sc, 0, 0, 0)) != 0) {
device_printf(sc->sc_dev,
"%s: could not set power saving level\n", __func__);
return error;
}
/* Configure bluetooth coexistence. */
if ((error = wpi_send_btcoex(sc)) != 0) {
device_printf(sc->sc_dev,
"could not configure bluetooth coexistence\n");
return error;
}
/* Configure adapter. */
memset(&sc->rxon, 0, sizeof (struct wpi_rxon));
IEEE80211_ADDR_COPY(sc->rxon.myaddr, vap->iv_myaddr);
/* Set default channel. */
sc->rxon.chan = ieee80211_chan2ieee(ic, c);
sc->rxon.flags = htole32(WPI_RXON_TSF | WPI_RXON_CTS_TO_SELF);
if (IEEE80211_IS_CHAN_2GHZ(c))
sc->rxon.flags |= htole32(WPI_RXON_AUTO | WPI_RXON_24GHZ);
sc->rxon.filter = WPI_FILTER_MULTICAST;
switch (ic->ic_opmode) {
case IEEE80211_M_STA:
sc->rxon.mode = WPI_MODE_STA;
break;
case IEEE80211_M_IBSS:
sc->rxon.mode = WPI_MODE_IBSS;
sc->rxon.filter |= WPI_FILTER_BEACON;
break;
case IEEE80211_M_HOSTAP:
/* XXX workaround for beaconing */
sc->rxon.mode = WPI_MODE_IBSS;
sc->rxon.filter |= WPI_FILTER_ASSOC | WPI_FILTER_PROMISC;
break;
case IEEE80211_M_AHDEMO:
sc->rxon.mode = WPI_MODE_HOSTAP;
break;
case IEEE80211_M_MONITOR:
sc->rxon.mode = WPI_MODE_MONITOR;
break;
default:
device_printf(sc->sc_dev, "unknown opmode %d\n",
ic->ic_opmode);
return EINVAL;
}
sc->rxon.filter = htole32(sc->rxon.filter);
wpi_set_promisc(sc);
sc->rxon.cck_mask = 0x0f; /* not yet negotiated */
sc->rxon.ofdm_mask = 0xff; /* not yet negotiated */
if ((error = wpi_send_rxon(sc, 0, 0)) != 0) {
device_printf(sc->sc_dev, "%s: could not send RXON\n",
__func__);
return error;
}
/* Setup rate scalling. */
if ((error = wpi_mrr_setup(sc)) != 0) {
device_printf(sc->sc_dev, "could not setup MRR, error %d\n",
error);
return error;
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
static uint16_t
wpi_get_active_dwell_time(struct wpi_softc *sc,
struct ieee80211_channel *c, uint8_t n_probes)
{
/* No channel? Default to 2GHz settings. */
if (c == NULL || IEEE80211_IS_CHAN_2GHZ(c)) {
return (WPI_ACTIVE_DWELL_TIME_2GHZ +
WPI_ACTIVE_DWELL_FACTOR_2GHZ * (n_probes + 1));
}
/* 5GHz dwell time. */
return (WPI_ACTIVE_DWELL_TIME_5GHZ +
WPI_ACTIVE_DWELL_FACTOR_5GHZ * (n_probes + 1));
}
/*
* Limit the total dwell time.
*
* Returns the dwell time in milliseconds.
*/
static uint16_t
wpi_limit_dwell(struct wpi_softc *sc, uint16_t dwell_time)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
uint16_t bintval = 0;
/* bintval is in TU (1.024mS) */
if (vap != NULL)
bintval = vap->iv_bss->ni_intval;
/*
* If it's non-zero, we should calculate the minimum of
* it and the DWELL_BASE.
*
* XXX Yes, the math should take into account that bintval
* is 1.024mS, not 1mS..
*/
if (bintval > 0) {
DPRINTF(sc, WPI_DEBUG_SCAN, "%s: bintval=%d\n", __func__,
bintval);
return (MIN(dwell_time, bintval - WPI_CHANNEL_TUNE_TIME * 2));
}
/* No association context? Default. */
return dwell_time;
}
static uint16_t
wpi_get_passive_dwell_time(struct wpi_softc *sc, struct ieee80211_channel *c)
{
uint16_t passive;
if (c == NULL || IEEE80211_IS_CHAN_2GHZ(c))
passive = WPI_PASSIVE_DWELL_BASE + WPI_PASSIVE_DWELL_TIME_2GHZ;
else
passive = WPI_PASSIVE_DWELL_BASE + WPI_PASSIVE_DWELL_TIME_5GHZ;
/* Clamp to the beacon interval if we're associated. */
return (wpi_limit_dwell(sc, passive));
}
static uint32_t
wpi_get_scan_pause_time(uint32_t time, uint16_t bintval)
{
uint32_t mod = (time % bintval) * IEEE80211_DUR_TU;
uint32_t nbeacons = time / bintval;
if (mod > WPI_PAUSE_MAX_TIME)
mod = WPI_PAUSE_MAX_TIME;
return WPI_PAUSE_SCAN(nbeacons, mod);
}
/*
* Send a scan request to the firmware.
*/
static int
wpi_scan(struct wpi_softc *sc, struct ieee80211_channel *c)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211_scan_state *ss = ic->ic_scan;
struct ieee80211vap *vap = ss->ss_vap;
struct wpi_scan_hdr *hdr;
struct wpi_cmd_data *tx;
struct wpi_scan_essid *essids;
struct wpi_scan_chan *chan;
struct ieee80211_frame *wh;
struct ieee80211_rateset *rs;
uint16_t bintval, buflen, dwell_active, dwell_passive;
uint8_t *buf, *frm, i, nssid;
int bgscan, error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/*
* We are absolutely not allowed to send a scan command when another
* scan command is pending.
*/
if (callout_pending(&sc->scan_timeout)) {
device_printf(sc->sc_dev, "%s: called whilst scanning!\n",
__func__);
error = EAGAIN;
goto fail;
}
bgscan = wpi_check_bss_filter(sc);
bintval = vap->iv_bss->ni_intval;
if (bgscan != 0 &&
bintval < WPI_QUIET_TIME_DEFAULT + WPI_CHANNEL_TUNE_TIME * 2) {
error = EOPNOTSUPP;
goto fail;
}
buf = malloc(WPI_SCAN_MAXSZ, M_DEVBUF, M_NOWAIT | M_ZERO);
if (buf == NULL) {
device_printf(sc->sc_dev,
"%s: could not allocate buffer for scan command\n",
__func__);
error = ENOMEM;
goto fail;
}
hdr = (struct wpi_scan_hdr *)buf;
/*
* Move to the next channel if no packets are received within 10 msecs
* after sending the probe request.
*/
hdr->quiet_time = htole16(WPI_QUIET_TIME_DEFAULT);
hdr->quiet_threshold = htole16(1);
if (bgscan != 0) {
/*
* Max needs to be greater than active and passive and quiet!
* It's also in microseconds!
*/
hdr->max_svc = htole32(250 * IEEE80211_DUR_TU);
hdr->pause_svc = htole32(wpi_get_scan_pause_time(100,
bintval));
}
hdr->filter = htole32(WPI_FILTER_MULTICAST | WPI_FILTER_BEACON);
tx = (struct wpi_cmd_data *)(hdr + 1);
tx->flags = htole32(WPI_TX_AUTO_SEQ);
tx->id = WPI_ID_BROADCAST;
tx->lifetime = htole32(WPI_LIFETIME_INFINITE);
if (IEEE80211_IS_CHAN_5GHZ(c)) {
/* Send probe requests at 6Mbps. */
tx->plcp = wpi_ridx_to_plcp[WPI_RIDX_OFDM6];
rs = &ic->ic_sup_rates[IEEE80211_MODE_11A];
} else {
hdr->flags = htole32(WPI_RXON_24GHZ | WPI_RXON_AUTO);
/* Send probe requests at 1Mbps. */
tx->plcp = wpi_ridx_to_plcp[WPI_RIDX_CCK1];
rs = &ic->ic_sup_rates[IEEE80211_MODE_11G];
}
essids = (struct wpi_scan_essid *)(tx + 1);
nssid = MIN(ss->ss_nssid, WPI_SCAN_MAX_ESSIDS);
for (i = 0; i < nssid; i++) {
essids[i].id = IEEE80211_ELEMID_SSID;
essids[i].len = MIN(ss->ss_ssid[i].len, IEEE80211_NWID_LEN);
memcpy(essids[i].data, ss->ss_ssid[i].ssid, essids[i].len);
#ifdef WPI_DEBUG
if (sc->sc_debug & WPI_DEBUG_SCAN) {
printf("Scanning Essid: ");
ieee80211_print_essid(essids[i].data, essids[i].len);
printf("\n");
}
#endif
}
/*
* Build a probe request frame. Most of the following code is a
* copy & paste of what is done in net80211.
*/
wh = (struct ieee80211_frame *)(essids + WPI_SCAN_MAX_ESSIDS);
wh->i_fc[0] = IEEE80211_FC0_VERSION_0 | IEEE80211_FC0_TYPE_MGT |
IEEE80211_FC0_SUBTYPE_PROBE_REQ;
wh->i_fc[1] = IEEE80211_FC1_DIR_NODS;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
IEEE80211_ADDR_COPY(wh->i_addr1, ieee80211broadcastaddr);
IEEE80211_ADDR_COPY(wh->i_addr2, vap->iv_myaddr);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
IEEE80211_ADDR_COPY(wh->i_addr3, ieee80211broadcastaddr);
frm = (uint8_t *)(wh + 1);
frm = ieee80211_add_ssid(frm, NULL, 0);
frm = ieee80211_add_rates(frm, rs);
if (rs->rs_nrates > IEEE80211_RATE_SIZE)
frm = ieee80211_add_xrates(frm, rs);
/* Set length of probe request. */
tx->len = htole16(frm - (uint8_t *)wh);
/*
* Construct information about the channel that we
* want to scan. The firmware expects this to be directly
* after the scan probe request
*/
chan = (struct wpi_scan_chan *)frm;
chan->chan = ieee80211_chan2ieee(ic, c);
chan->flags = 0;
if (nssid) {
hdr->crc_threshold = WPI_SCAN_CRC_TH_DEFAULT;
chan->flags |= WPI_CHAN_NPBREQS(nssid);
} else
hdr->crc_threshold = WPI_SCAN_CRC_TH_NEVER;
if (!IEEE80211_IS_CHAN_PASSIVE(c))
chan->flags |= WPI_CHAN_ACTIVE;
/*
* Calculate the active/passive dwell times.
*/
dwell_active = wpi_get_active_dwell_time(sc, c, nssid);
dwell_passive = wpi_get_passive_dwell_time(sc, c);
/* Make sure they're valid. */
if (dwell_active > dwell_passive)
dwell_active = dwell_passive;
chan->active = htole16(dwell_active);
chan->passive = htole16(dwell_passive);
chan->dsp_gain = 0x6e; /* Default level */
if (IEEE80211_IS_CHAN_5GHZ(c))
chan->rf_gain = 0x3b;
else
chan->rf_gain = 0x28;
DPRINTF(sc, WPI_DEBUG_SCAN, "Scanning %u Passive: %d\n",
chan->chan, IEEE80211_IS_CHAN_PASSIVE(c));
hdr->nchan++;
if (hdr->nchan == 1 && sc->rxon.chan == chan->chan) {
/* XXX Force probe request transmission. */
memcpy(chan + 1, chan, sizeof (struct wpi_scan_chan));
chan++;
/* Reduce unnecessary delay. */
chan->flags = 0;
chan->passive = chan->active = hdr->quiet_time;
hdr->nchan++;
}
chan++;
buflen = (uint8_t *)chan - buf;
hdr->len = htole16(buflen);
DPRINTF(sc, WPI_DEBUG_CMD, "sending scan command nchan=%d\n",
hdr->nchan);
error = wpi_cmd(sc, WPI_CMD_SCAN, buf, buflen, 1);
free(buf, M_DEVBUF);
if (error != 0)
goto fail;
callout_reset(&sc->scan_timeout, 5*hz, wpi_scan_timeout, sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
fail: DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
return error;
}
static int
wpi_auth(struct wpi_softc *sc, struct ieee80211vap *vap)
{
struct ieee80211com *ic = vap->iv_ic;
struct ieee80211_node *ni = vap->iv_bss;
struct ieee80211_channel *c = ni->ni_chan;
int error;
WPI_RXON_LOCK(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/* Update adapter configuration. */
sc->rxon.associd = 0;
sc->rxon.filter &= ~htole32(WPI_FILTER_BSS);
IEEE80211_ADDR_COPY(sc->rxon.bssid, ni->ni_bssid);
sc->rxon.chan = ieee80211_chan2ieee(ic, c);
sc->rxon.flags = htole32(WPI_RXON_TSF | WPI_RXON_CTS_TO_SELF);
if (IEEE80211_IS_CHAN_2GHZ(c))
sc->rxon.flags |= htole32(WPI_RXON_AUTO | WPI_RXON_24GHZ);
if (ic->ic_flags & IEEE80211_F_SHSLOT)
sc->rxon.flags |= htole32(WPI_RXON_SHSLOT);
if (ic->ic_flags & IEEE80211_F_SHPREAMBLE)
sc->rxon.flags |= htole32(WPI_RXON_SHPREAMBLE);
if (IEEE80211_IS_CHAN_A(c)) {
sc->rxon.cck_mask = 0;
sc->rxon.ofdm_mask = 0x15;
} else if (IEEE80211_IS_CHAN_B(c)) {
sc->rxon.cck_mask = 0x03;
sc->rxon.ofdm_mask = 0;
} else {
/* Assume 802.11b/g. */
sc->rxon.cck_mask = 0x0f;
sc->rxon.ofdm_mask = 0x15;
}
DPRINTF(sc, WPI_DEBUG_STATE, "rxon chan %d flags %x cck %x ofdm %x\n",
sc->rxon.chan, sc->rxon.flags, sc->rxon.cck_mask,
sc->rxon.ofdm_mask);
if ((error = wpi_send_rxon(sc, 0, 1)) != 0) {
device_printf(sc->sc_dev, "%s: could not send RXON\n",
__func__);
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
WPI_RXON_UNLOCK(sc);
return error;
}
static int
wpi_config_beacon(struct wpi_vap *wvp)
{
struct ieee80211vap *vap = &wvp->wv_vap;
struct ieee80211com *ic = vap->iv_ic;
struct ieee80211_beacon_offsets *bo = &vap->iv_bcn_off;
struct wpi_buf *bcn = &wvp->wv_bcbuf;
struct wpi_softc *sc = ic->ic_softc;
struct wpi_cmd_beacon *cmd = (struct wpi_cmd_beacon *)&bcn->data;
struct ieee80211_tim_ie *tie;
struct mbuf *m;
uint8_t *ptr;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
WPI_VAP_LOCK_ASSERT(wvp);
cmd->len = htole16(bcn->m->m_pkthdr.len);
cmd->plcp = (ic->ic_curmode == IEEE80211_MODE_11A) ?
wpi_ridx_to_plcp[WPI_RIDX_OFDM6] : wpi_ridx_to_plcp[WPI_RIDX_CCK1];
/* XXX seems to be unused */
if (*(bo->bo_tim) == IEEE80211_ELEMID_TIM) {
tie = (struct ieee80211_tim_ie *) bo->bo_tim;
ptr = mtod(bcn->m, uint8_t *);
cmd->tim = htole16(bo->bo_tim - ptr);
cmd->timsz = tie->tim_len;
}
/* Necessary for recursion in ieee80211_beacon_update(). */
m = bcn->m;
bcn->m = m_dup(m, M_NOWAIT);
if (bcn->m == NULL) {
device_printf(sc->sc_dev,
"%s: could not copy beacon frame\n", __func__);
error = ENOMEM;
goto end;
}
if ((error = wpi_cmd2(sc, bcn)) != 0) {
device_printf(sc->sc_dev,
"%s: could not update beacon frame, error %d", __func__,
error);
}
/* Restore mbuf. */
end: bcn->m = m;
return error;
}
static int
wpi_setup_beacon(struct wpi_softc *sc, struct ieee80211_node *ni)
{
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_vap *wvp = WPI_VAP(vap);
struct wpi_buf *bcn = &wvp->wv_bcbuf;
struct mbuf *m;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (ni->ni_chan == IEEE80211_CHAN_ANYC)
return EINVAL;
m = ieee80211_beacon_alloc(ni);
if (m == NULL) {
device_printf(sc->sc_dev,
"%s: could not allocate beacon frame\n", __func__);
return ENOMEM;
}
WPI_VAP_LOCK(wvp);
if (bcn->m != NULL)
m_freem(bcn->m);
bcn->m = m;
error = wpi_config_beacon(wvp);
WPI_VAP_UNLOCK(wvp);
return error;
}
static void
wpi_update_beacon(struct ieee80211vap *vap, int item)
{
struct wpi_softc *sc = vap->iv_ic->ic_softc;
struct wpi_vap *wvp = WPI_VAP(vap);
struct wpi_buf *bcn = &wvp->wv_bcbuf;
struct ieee80211_beacon_offsets *bo = &vap->iv_bcn_off;
struct ieee80211_node *ni = vap->iv_bss;
int mcast = 0;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
WPI_VAP_LOCK(wvp);
if (bcn->m == NULL) {
bcn->m = ieee80211_beacon_alloc(ni);
if (bcn->m == NULL) {
device_printf(sc->sc_dev,
"%s: could not allocate beacon frame\n", __func__);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR,
__func__);
WPI_VAP_UNLOCK(wvp);
return;
}
}
WPI_VAP_UNLOCK(wvp);
if (item == IEEE80211_BEACON_TIM)
mcast = 1; /* TODO */
setbit(bo->bo_flags, item);
ieee80211_beacon_update(ni, bcn->m, mcast);
WPI_VAP_LOCK(wvp);
wpi_config_beacon(wvp);
WPI_VAP_UNLOCK(wvp);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
}
static void
wpi_newassoc(struct ieee80211_node *ni, int isnew)
{
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_softc *sc = ni->ni_ic->ic_softc;
struct wpi_node *wn = WPI_NODE(ni);
int error;
WPI_NT_LOCK(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (vap->iv_opmode != IEEE80211_M_STA && wn->id == WPI_ID_UNDEFINED) {
if ((error = wpi_add_ibss_node(sc, ni)) != 0) {
device_printf(sc->sc_dev,
"%s: could not add IBSS node, error %d\n",
__func__, error);
}
}
WPI_NT_UNLOCK(sc);
}
static int
wpi_run(struct wpi_softc *sc, struct ieee80211vap *vap)
{
struct ieee80211com *ic = vap->iv_ic;
struct ieee80211_node *ni = vap->iv_bss;
struct ieee80211_channel *c = ni->ni_chan;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
if (vap->iv_opmode == IEEE80211_M_MONITOR) {
/* Link LED blinks while monitoring. */
wpi_set_led(sc, WPI_LED_LINK, 5, 5);
return 0;
}
/* XXX kernel panic workaround */
if (c == IEEE80211_CHAN_ANYC) {
device_printf(sc->sc_dev, "%s: incomplete configuration\n",
__func__);
return EINVAL;
}
if ((error = wpi_set_timing(sc, ni)) != 0) {
device_printf(sc->sc_dev,
"%s: could not set timing, error %d\n", __func__, error);
return error;
}
/* Update adapter configuration. */
WPI_RXON_LOCK(sc);
IEEE80211_ADDR_COPY(sc->rxon.bssid, ni->ni_bssid);
sc->rxon.associd = htole16(IEEE80211_NODE_AID(ni));
sc->rxon.chan = ieee80211_chan2ieee(ic, c);
sc->rxon.flags = htole32(WPI_RXON_TSF | WPI_RXON_CTS_TO_SELF);
if (IEEE80211_IS_CHAN_2GHZ(c))
sc->rxon.flags |= htole32(WPI_RXON_AUTO | WPI_RXON_24GHZ);
if (ic->ic_flags & IEEE80211_F_SHSLOT)
sc->rxon.flags |= htole32(WPI_RXON_SHSLOT);
if (ic->ic_flags & IEEE80211_F_SHPREAMBLE)
sc->rxon.flags |= htole32(WPI_RXON_SHPREAMBLE);
if (IEEE80211_IS_CHAN_A(c)) {
sc->rxon.cck_mask = 0;
sc->rxon.ofdm_mask = 0x15;
} else if (IEEE80211_IS_CHAN_B(c)) {
sc->rxon.cck_mask = 0x03;
sc->rxon.ofdm_mask = 0;
} else {
/* Assume 802.11b/g. */
sc->rxon.cck_mask = 0x0f;
sc->rxon.ofdm_mask = 0x15;
}
sc->rxon.filter |= htole32(WPI_FILTER_BSS);
DPRINTF(sc, WPI_DEBUG_STATE, "rxon chan %d flags %x\n",
sc->rxon.chan, sc->rxon.flags);
if ((error = wpi_send_rxon(sc, 0, 1)) != 0) {
device_printf(sc->sc_dev, "%s: could not send RXON\n",
__func__);
return error;
}
/* Start periodic calibration timer. */
callout_reset(&sc->calib_to, 60*hz, wpi_calib_timeout, sc);
WPI_RXON_UNLOCK(sc);
if (vap->iv_opmode == IEEE80211_M_IBSS ||
vap->iv_opmode == IEEE80211_M_HOSTAP) {
if ((error = wpi_setup_beacon(sc, ni)) != 0) {
device_printf(sc->sc_dev,
"%s: could not setup beacon, error %d\n", __func__,
error);
return error;
}
}
if (vap->iv_opmode == IEEE80211_M_STA) {
/* Add BSS node. */
WPI_NT_LOCK(sc);
error = wpi_add_sta_node(sc, ni);
WPI_NT_UNLOCK(sc);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not add BSS node, error %d\n", __func__,
error);
return error;
}
}
/* Link LED always on while associated. */
wpi_set_led(sc, WPI_LED_LINK, 0, 1);
/* Enable power-saving mode if requested by user. */
if ((vap->iv_flags & IEEE80211_F_PMGTON) &&
vap->iv_opmode != IEEE80211_M_IBSS)
(void)wpi_set_pslevel(sc, 0, 3, 1);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
return 0;
}
static int
wpi_load_key(struct ieee80211_node *ni, const struct ieee80211_key *k)
{
const struct ieee80211_cipher *cip = k->wk_cipher;
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_softc *sc = ni->ni_ic->ic_softc;
struct wpi_node *wn = WPI_NODE(ni);
struct wpi_node_info node;
uint16_t kflags;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (wpi_check_node_entry(sc, wn->id) == 0) {
device_printf(sc->sc_dev, "%s: node does not exist\n",
__func__);
return 0;
}
switch (cip->ic_cipher) {
case IEEE80211_CIPHER_AES_CCM:
kflags = WPI_KFLAG_CCMP;
break;
default:
device_printf(sc->sc_dev, "%s: unknown cipher %d\n", __func__,
cip->ic_cipher);
return 0;
}
kflags |= WPI_KFLAG_KID(k->wk_keyix);
if (k->wk_flags & IEEE80211_KEY_GROUP)
kflags |= WPI_KFLAG_MULTICAST;
memset(&node, 0, sizeof node);
node.id = wn->id;
node.control = WPI_NODE_UPDATE;
node.flags = WPI_FLAG_KEY_SET;
node.kflags = htole16(kflags);
memcpy(node.key, k->wk_key, k->wk_keylen);
again:
DPRINTF(sc, WPI_DEBUG_KEY,
"%s: setting %s key id %d for node %d (%s)\n", __func__,
(kflags & WPI_KFLAG_MULTICAST) ? "group" : "ucast", k->wk_keyix,
node.id, ether_sprintf(ni->ni_macaddr));
error = wpi_cmd(sc, WPI_CMD_ADD_NODE, &node, sizeof node, 1);
if (error != 0) {
device_printf(sc->sc_dev, "can't update node info, error %d\n",
error);
return !error;
}
if (!(kflags & WPI_KFLAG_MULTICAST) && &vap->iv_nw_keys[0] <= k &&
k < &vap->iv_nw_keys[IEEE80211_WEP_NKID]) {
kflags |= WPI_KFLAG_MULTICAST;
node.kflags = htole16(kflags);
goto again;
}
return 1;
}
static void
wpi_load_key_cb(void *arg, struct ieee80211_node *ni)
{
const struct ieee80211_key *k = arg;
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_softc *sc = ni->ni_ic->ic_softc;
struct wpi_node *wn = WPI_NODE(ni);
int error;
if (vap->iv_bss == ni && wn->id == WPI_ID_UNDEFINED)
return;
WPI_NT_LOCK(sc);
error = wpi_load_key(ni, k);
WPI_NT_UNLOCK(sc);
if (error == 0) {
device_printf(sc->sc_dev, "%s: error while setting key\n",
__func__);
}
}
static int
wpi_set_global_keys(struct ieee80211_node *ni)
{
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211_key *wk = &vap->iv_nw_keys[0];
int error = 1;
for (; wk < &vap->iv_nw_keys[IEEE80211_WEP_NKID] && error; wk++)
if (wk->wk_keyix != IEEE80211_KEYIX_NONE)
error = wpi_load_key(ni, wk);
return !error;
}
static int
wpi_del_key(struct ieee80211_node *ni, const struct ieee80211_key *k)
{
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_softc *sc = ni->ni_ic->ic_softc;
struct wpi_node *wn = WPI_NODE(ni);
struct wpi_node_info node;
uint16_t kflags;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (wpi_check_node_entry(sc, wn->id) == 0) {
DPRINTF(sc, WPI_DEBUG_KEY, "%s: node was removed\n", __func__);
return 1; /* Nothing to do. */
}
kflags = WPI_KFLAG_KID(k->wk_keyix);
if (k->wk_flags & IEEE80211_KEY_GROUP)
kflags |= WPI_KFLAG_MULTICAST;
memset(&node, 0, sizeof node);
node.id = wn->id;
node.control = WPI_NODE_UPDATE;
node.flags = WPI_FLAG_KEY_SET;
node.kflags = htole16(kflags);
again:
DPRINTF(sc, WPI_DEBUG_KEY, "%s: deleting %s key %d for node %d (%s)\n",
__func__, (kflags & WPI_KFLAG_MULTICAST) ? "group" : "ucast",
k->wk_keyix, node.id, ether_sprintf(ni->ni_macaddr));
error = wpi_cmd(sc, WPI_CMD_ADD_NODE, &node, sizeof node, 1);
if (error != 0) {
device_printf(sc->sc_dev, "can't update node info, error %d\n",
error);
return !error;
}
if (!(kflags & WPI_KFLAG_MULTICAST) && &vap->iv_nw_keys[0] <= k &&
k < &vap->iv_nw_keys[IEEE80211_WEP_NKID]) {
kflags |= WPI_KFLAG_MULTICAST;
node.kflags = htole16(kflags);
goto again;
}
return 1;
}
static void
wpi_del_key_cb(void *arg, struct ieee80211_node *ni)
{
const struct ieee80211_key *k = arg;
struct ieee80211vap *vap = ni->ni_vap;
struct wpi_softc *sc = ni->ni_ic->ic_softc;
struct wpi_node *wn = WPI_NODE(ni);
int error;
if (vap->iv_bss == ni && wn->id == WPI_ID_UNDEFINED)
return;
WPI_NT_LOCK(sc);
error = wpi_del_key(ni, k);
WPI_NT_UNLOCK(sc);
if (error == 0) {
device_printf(sc->sc_dev, "%s: error while deleting key\n",
__func__);
}
}
static int
wpi_process_key(struct ieee80211vap *vap, const struct ieee80211_key *k,
int set)
{
struct ieee80211com *ic = vap->iv_ic;
struct wpi_softc *sc = ic->ic_softc;
struct wpi_vap *wvp = WPI_VAP(vap);
struct ieee80211_node *ni;
int error, ni_ref = 0;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (k->wk_flags & IEEE80211_KEY_SWCRYPT) {
/* Not for us. */
return 1;
}
if (!(k->wk_flags & IEEE80211_KEY_RECV)) {
/* XMIT keys are handled in wpi_tx_data(). */
return 1;
}
/* Handle group keys. */
if (&vap->iv_nw_keys[0] <= k &&
k < &vap->iv_nw_keys[IEEE80211_WEP_NKID]) {
WPI_NT_LOCK(sc);
if (set)
wvp->wv_gtk |= WPI_VAP_KEY(k->wk_keyix);
else
wvp->wv_gtk &= ~WPI_VAP_KEY(k->wk_keyix);
WPI_NT_UNLOCK(sc);
if (vap->iv_state == IEEE80211_S_RUN) {
ieee80211_iterate_nodes(&ic->ic_sta,
set ? wpi_load_key_cb : wpi_del_key_cb,
__DECONST(void *, k));
}
return 1;
}
switch (vap->iv_opmode) {
case IEEE80211_M_STA:
ni = vap->iv_bss;
break;
case IEEE80211_M_IBSS:
case IEEE80211_M_AHDEMO:
case IEEE80211_M_HOSTAP:
ni = ieee80211_find_vap_node(&ic->ic_sta, vap, k->wk_macaddr);
if (ni == NULL)
return 0; /* should not happen */
ni_ref = 1;
break;
default:
device_printf(sc->sc_dev, "%s: unknown opmode %d\n", __func__,
vap->iv_opmode);
return 0;
}
WPI_NT_LOCK(sc);
if (set)
error = wpi_load_key(ni, k);
else
error = wpi_del_key(ni, k);
WPI_NT_UNLOCK(sc);
if (ni_ref)
ieee80211_node_decref(ni);
return error;
}
static int
wpi_key_set(struct ieee80211vap *vap, const struct ieee80211_key *k)
{
return wpi_process_key(vap, k, 1);
}
static int
wpi_key_delete(struct ieee80211vap *vap, const struct ieee80211_key *k)
{
return wpi_process_key(vap, k, 0);
}
/*
* This function is called after the runtime firmware notifies us of its
* readiness (called in a process context).
*/
static int
wpi_post_alive(struct wpi_softc *sc)
{
int ntries, error;
/* Check (again) that the radio is not disabled. */
if ((error = wpi_nic_lock(sc)) != 0)
return error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
/* NB: Runtime firmware must be up and running. */
if (!(wpi_prph_read(sc, WPI_APMG_RFKILL) & 1)) {
device_printf(sc->sc_dev,
"RF switch: radio disabled (%s)\n", __func__);
wpi_nic_unlock(sc);
return EPERM; /* :-) */
}
wpi_nic_unlock(sc);
/* Wait for thermal sensor to calibrate. */
for (ntries = 0; ntries < 1000; ntries++) {
if ((sc->temp = (int)WPI_READ(sc, WPI_UCODE_GP2)) != 0)
break;
DELAY(10);
}
if (ntries == 1000) {
device_printf(sc->sc_dev,
"timeout waiting for thermal sensor calibration\n");
return ETIMEDOUT;
}
DPRINTF(sc, WPI_DEBUG_TEMP, "temperature %d\n", sc->temp);
return 0;
}
/*
* The firmware boot code is small and is intended to be copied directly into
* the NIC internal memory (no DMA transfer).
*/
static int
wpi_load_bootcode(struct wpi_softc *sc, const uint8_t *ucode, uint32_t size)
{
int error, ntries;
DPRINTF(sc, WPI_DEBUG_HW, "Loading microcode size 0x%x\n", size);
size /= sizeof (uint32_t);
if ((error = wpi_nic_lock(sc)) != 0)
return error;
/* Copy microcode image into NIC memory. */
wpi_prph_write_region_4(sc, WPI_BSM_SRAM_BASE,
(const uint32_t *)ucode, size);
wpi_prph_write(sc, WPI_BSM_WR_MEM_SRC, 0);
wpi_prph_write(sc, WPI_BSM_WR_MEM_DST, WPI_FW_TEXT_BASE);
wpi_prph_write(sc, WPI_BSM_WR_DWCOUNT, size);
/* Start boot load now. */
wpi_prph_write(sc, WPI_BSM_WR_CTRL, WPI_BSM_WR_CTRL_START);
/* Wait for transfer to complete. */
for (ntries = 0; ntries < 1000; ntries++) {
uint32_t status = WPI_READ(sc, WPI_FH_TX_STATUS);
DPRINTF(sc, WPI_DEBUG_HW,
"firmware status=0x%x, val=0x%x, result=0x%x\n", status,
WPI_FH_TX_STATUS_IDLE(6),
status & WPI_FH_TX_STATUS_IDLE(6));
if (status & WPI_FH_TX_STATUS_IDLE(6)) {
DPRINTF(sc, WPI_DEBUG_HW,
"Status Match! - ntries = %d\n", ntries);
break;
}
DELAY(10);
}
if (ntries == 1000) {
device_printf(sc->sc_dev, "%s: could not load boot firmware\n",
__func__);
wpi_nic_unlock(sc);
return ETIMEDOUT;
}
/* Enable boot after power up. */
wpi_prph_write(sc, WPI_BSM_WR_CTRL, WPI_BSM_WR_CTRL_START_EN);
wpi_nic_unlock(sc);
return 0;
}
static int
wpi_load_firmware(struct wpi_softc *sc)
{
struct wpi_fw_info *fw = &sc->fw;
struct wpi_dma_info *dma = &sc->fw_dma;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
/* Copy initialization sections into pre-allocated DMA-safe memory. */
memcpy(dma->vaddr, fw->init.data, fw->init.datasz);
bus_dmamap_sync(dma->tag, dma->map, BUS_DMASYNC_PREWRITE);
memcpy(dma->vaddr + WPI_FW_DATA_MAXSZ, fw->init.text, fw->init.textsz);
bus_dmamap_sync(dma->tag, dma->map, BUS_DMASYNC_PREWRITE);
/* Tell adapter where to find initialization sections. */
if ((error = wpi_nic_lock(sc)) != 0)
return error;
wpi_prph_write(sc, WPI_BSM_DRAM_DATA_ADDR, dma->paddr);
wpi_prph_write(sc, WPI_BSM_DRAM_DATA_SIZE, fw->init.datasz);
wpi_prph_write(sc, WPI_BSM_DRAM_TEXT_ADDR,
dma->paddr + WPI_FW_DATA_MAXSZ);
wpi_prph_write(sc, WPI_BSM_DRAM_TEXT_SIZE, fw->init.textsz);
wpi_nic_unlock(sc);
/* Load firmware boot code. */
error = wpi_load_bootcode(sc, fw->boot.text, fw->boot.textsz);
if (error != 0) {
device_printf(sc->sc_dev, "%s: could not load boot firmware\n",
__func__);
return error;
}
/* Now press "execute". */
WPI_WRITE(sc, WPI_RESET, 0);
/* Wait at most one second for first alive notification. */
if ((error = mtx_sleep(sc, &sc->sc_mtx, PCATCH, "wpiinit", hz)) != 0) {
device_printf(sc->sc_dev,
"%s: timeout waiting for adapter to initialize, error %d\n",
__func__, error);
return error;
}
/* Copy runtime sections into pre-allocated DMA-safe memory. */
memcpy(dma->vaddr, fw->main.data, fw->main.datasz);
bus_dmamap_sync(dma->tag, dma->map, BUS_DMASYNC_PREWRITE);
memcpy(dma->vaddr + WPI_FW_DATA_MAXSZ, fw->main.text, fw->main.textsz);
bus_dmamap_sync(dma->tag, dma->map, BUS_DMASYNC_PREWRITE);
/* Tell adapter where to find runtime sections. */
if ((error = wpi_nic_lock(sc)) != 0)
return error;
wpi_prph_write(sc, WPI_BSM_DRAM_DATA_ADDR, dma->paddr);
wpi_prph_write(sc, WPI_BSM_DRAM_DATA_SIZE, fw->main.datasz);
wpi_prph_write(sc, WPI_BSM_DRAM_TEXT_ADDR,
dma->paddr + WPI_FW_DATA_MAXSZ);
wpi_prph_write(sc, WPI_BSM_DRAM_TEXT_SIZE,
WPI_FW_UPDATED | fw->main.textsz);
wpi_nic_unlock(sc);
return 0;
}
static int
wpi_read_firmware(struct wpi_softc *sc)
{
const struct firmware *fp;
struct wpi_fw_info *fw = &sc->fw;
const struct wpi_firmware_hdr *hdr;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
DPRINTF(sc, WPI_DEBUG_FIRMWARE,
"Attempting Loading Firmware from %s module\n", WPI_FW_NAME);
WPI_UNLOCK(sc);
fp = firmware_get(WPI_FW_NAME);
WPI_LOCK(sc);
if (fp == NULL) {
device_printf(sc->sc_dev,
"could not load firmware image '%s'\n", WPI_FW_NAME);
return EINVAL;
}
sc->fw_fp = fp;
if (fp->datasize < sizeof (struct wpi_firmware_hdr)) {
device_printf(sc->sc_dev,
"firmware file too short: %zu bytes\n", fp->datasize);
error = EINVAL;
goto fail;
}
fw->size = fp->datasize;
fw->data = (const uint8_t *)fp->data;
/* Extract firmware header information. */
hdr = (const struct wpi_firmware_hdr *)fw->data;
/* | RUNTIME FIRMWARE | INIT FIRMWARE | BOOT FW |
|HDR|<--TEXT-->|<--DATA-->|<--TEXT-->|<--DATA-->|<--TEXT-->| */
fw->main.textsz = le32toh(hdr->rtextsz);
fw->main.datasz = le32toh(hdr->rdatasz);
fw->init.textsz = le32toh(hdr->itextsz);
fw->init.datasz = le32toh(hdr->idatasz);
fw->boot.textsz = le32toh(hdr->btextsz);
fw->boot.datasz = 0;
/* Sanity-check firmware header. */
if (fw->main.textsz > WPI_FW_TEXT_MAXSZ ||
fw->main.datasz > WPI_FW_DATA_MAXSZ ||
fw->init.textsz > WPI_FW_TEXT_MAXSZ ||
fw->init.datasz > WPI_FW_DATA_MAXSZ ||
fw->boot.textsz > WPI_FW_BOOT_TEXT_MAXSZ ||
(fw->boot.textsz & 3) != 0) {
device_printf(sc->sc_dev, "invalid firmware header\n");
error = EINVAL;
goto fail;
}
/* Check that all firmware sections fit. */
if (fw->size < sizeof (*hdr) + fw->main.textsz + fw->main.datasz +
fw->init.textsz + fw->init.datasz + fw->boot.textsz) {
device_printf(sc->sc_dev,
"firmware file too short: %zu bytes\n", fw->size);
error = EINVAL;
goto fail;
}
/* Get pointers to firmware sections. */
fw->main.text = (const uint8_t *)(hdr + 1);
fw->main.data = fw->main.text + fw->main.textsz;
fw->init.text = fw->main.data + fw->main.datasz;
fw->init.data = fw->init.text + fw->init.textsz;
fw->boot.text = fw->init.data + fw->init.datasz;
DPRINTF(sc, WPI_DEBUG_FIRMWARE,
"Firmware Version: Major %d, Minor %d, Driver %d, \n"
"runtime (text: %u, data: %u) init (text: %u, data %u) "
"boot (text %u)\n", hdr->major, hdr->minor, le32toh(hdr->driver),
fw->main.textsz, fw->main.datasz,
fw->init.textsz, fw->init.datasz, fw->boot.textsz);
DPRINTF(sc, WPI_DEBUG_FIRMWARE, "fw->main.text %p\n", fw->main.text);
DPRINTF(sc, WPI_DEBUG_FIRMWARE, "fw->main.data %p\n", fw->main.data);
DPRINTF(sc, WPI_DEBUG_FIRMWARE, "fw->init.text %p\n", fw->init.text);
DPRINTF(sc, WPI_DEBUG_FIRMWARE, "fw->init.data %p\n", fw->init.data);
DPRINTF(sc, WPI_DEBUG_FIRMWARE, "fw->boot.text %p\n", fw->boot.text);
return 0;
fail: wpi_unload_firmware(sc);
return error;
}
/**
* Free the referenced firmware image
*/
static void
wpi_unload_firmware(struct wpi_softc *sc)
{
if (sc->fw_fp != NULL) {
firmware_put(sc->fw_fp, FIRMWARE_UNLOAD);
sc->fw_fp = NULL;
}
}
static int
wpi_clock_wait(struct wpi_softc *sc)
{
int ntries;
/* Set "initialization complete" bit. */
WPI_SETBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_INIT_DONE);
/* Wait for clock stabilization. */
for (ntries = 0; ntries < 2500; ntries++) {
if (WPI_READ(sc, WPI_GP_CNTRL) & WPI_GP_CNTRL_MAC_CLOCK_READY)
return 0;
DELAY(100);
}
device_printf(sc->sc_dev,
"%s: timeout waiting for clock stabilization\n", __func__);
return ETIMEDOUT;
}
static int
wpi_apm_init(struct wpi_softc *sc)
{
uint32_t reg;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
/* Disable L0s exit timer (NMI bug workaround). */
WPI_SETBITS(sc, WPI_GIO_CHICKEN, WPI_GIO_CHICKEN_DIS_L0S_TIMER);
/* Don't wait for ICH L0s (ICH bug workaround). */
WPI_SETBITS(sc, WPI_GIO_CHICKEN, WPI_GIO_CHICKEN_L1A_NO_L0S_RX);
/* Set FH wait threshold to max (HW bug under stress workaround). */
WPI_SETBITS(sc, WPI_DBG_HPET_MEM, 0xffff0000);
/* Retrieve PCIe Active State Power Management (ASPM). */
reg = pci_read_config(sc->sc_dev, sc->sc_cap_off + 0x10, 1);
/* Workaround for HW instability in PCIe L0->L0s->L1 transition. */
if (reg & 0x02) /* L1 Entry enabled. */
WPI_SETBITS(sc, WPI_GIO, WPI_GIO_L0S_ENA);
else
WPI_CLRBITS(sc, WPI_GIO, WPI_GIO_L0S_ENA);
WPI_SETBITS(sc, WPI_ANA_PLL, WPI_ANA_PLL_INIT);
/* Wait for clock stabilization before accessing prph. */
if ((error = wpi_clock_wait(sc)) != 0)
return error;
if ((error = wpi_nic_lock(sc)) != 0)
return error;
/* Cleanup. */
wpi_prph_write(sc, WPI_APMG_CLK_DIS, 0x00000400);
wpi_prph_clrbits(sc, WPI_APMG_PS, 0x00000200);
/* Enable DMA and BSM (Bootstrap State Machine). */
wpi_prph_write(sc, WPI_APMG_CLK_EN,
WPI_APMG_CLK_CTRL_DMA_CLK_RQT | WPI_APMG_CLK_CTRL_BSM_CLK_RQT);
DELAY(20);
/* Disable L1-Active. */
wpi_prph_setbits(sc, WPI_APMG_PCI_STT, WPI_APMG_PCI_STT_L1A_DIS);
wpi_nic_unlock(sc);
return 0;
}
static void
wpi_apm_stop_master(struct wpi_softc *sc)
{
int ntries;
/* Stop busmaster DMA activity. */
WPI_SETBITS(sc, WPI_RESET, WPI_RESET_STOP_MASTER);
if ((WPI_READ(sc, WPI_GP_CNTRL) & WPI_GP_CNTRL_PS_MASK) ==
WPI_GP_CNTRL_MAC_PS)
return; /* Already asleep. */
for (ntries = 0; ntries < 100; ntries++) {
if (WPI_READ(sc, WPI_RESET) & WPI_RESET_MASTER_DISABLED)
return;
DELAY(10);
}
device_printf(sc->sc_dev, "%s: timeout waiting for master\n",
__func__);
}
static void
wpi_apm_stop(struct wpi_softc *sc)
{
wpi_apm_stop_master(sc);
/* Reset the entire device. */
WPI_SETBITS(sc, WPI_RESET, WPI_RESET_SW);
DELAY(10);
/* Clear "initialization complete" bit. */
WPI_CLRBITS(sc, WPI_GP_CNTRL, WPI_GP_CNTRL_INIT_DONE);
}
static void
wpi_nic_config(struct wpi_softc *sc)
{
uint32_t rev;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
/* voodoo from the Linux "driver".. */
rev = pci_read_config(sc->sc_dev, PCIR_REVID, 1);
if ((rev & 0xc0) == 0x40)
WPI_SETBITS(sc, WPI_HW_IF_CONFIG, WPI_HW_IF_CONFIG_ALM_MB);
else if (!(rev & 0x80))
WPI_SETBITS(sc, WPI_HW_IF_CONFIG, WPI_HW_IF_CONFIG_ALM_MM);
if (sc->cap == 0x80)
WPI_SETBITS(sc, WPI_HW_IF_CONFIG, WPI_HW_IF_CONFIG_SKU_MRC);
if ((sc->rev & 0xf0) == 0xd0)
WPI_SETBITS(sc, WPI_HW_IF_CONFIG, WPI_HW_IF_CONFIG_REV_D);
else
WPI_CLRBITS(sc, WPI_HW_IF_CONFIG, WPI_HW_IF_CONFIG_REV_D);
if (sc->type > 1)
WPI_SETBITS(sc, WPI_HW_IF_CONFIG, WPI_HW_IF_CONFIG_TYPE_B);
}
static int
wpi_hw_init(struct wpi_softc *sc)
{
uint8_t chnl;
int ntries, error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
/* Clear pending interrupts. */
WPI_WRITE(sc, WPI_INT, 0xffffffff);
if ((error = wpi_apm_init(sc)) != 0) {
device_printf(sc->sc_dev,
"%s: could not power ON adapter, error %d\n", __func__,
error);
return error;
}
/* Select VMAIN power source. */
if ((error = wpi_nic_lock(sc)) != 0)
return error;
wpi_prph_clrbits(sc, WPI_APMG_PS, WPI_APMG_PS_PWR_SRC_MASK);
wpi_nic_unlock(sc);
/* Spin until VMAIN gets selected. */
for (ntries = 0; ntries < 5000; ntries++) {
if (WPI_READ(sc, WPI_GPIO_IN) & WPI_GPIO_IN_VMAIN)
break;
DELAY(10);
}
if (ntries == 5000) {
device_printf(sc->sc_dev, "timeout selecting power source\n");
return ETIMEDOUT;
}
/* Perform adapter initialization. */
wpi_nic_config(sc);
/* Initialize RX ring. */
if ((error = wpi_nic_lock(sc)) != 0)
return error;
/* Set physical address of RX ring. */
WPI_WRITE(sc, WPI_FH_RX_BASE, sc->rxq.desc_dma.paddr);
/* Set physical address of RX read pointer. */
WPI_WRITE(sc, WPI_FH_RX_RPTR_ADDR, sc->shared_dma.paddr +
offsetof(struct wpi_shared, next));
WPI_WRITE(sc, WPI_FH_RX_WPTR, 0);
/* Enable RX. */
WPI_WRITE(sc, WPI_FH_RX_CONFIG,
WPI_FH_RX_CONFIG_DMA_ENA |
WPI_FH_RX_CONFIG_RDRBD_ENA |
WPI_FH_RX_CONFIG_WRSTATUS_ENA |
WPI_FH_RX_CONFIG_MAXFRAG |
WPI_FH_RX_CONFIG_NRBD(WPI_RX_RING_COUNT_LOG) |
WPI_FH_RX_CONFIG_IRQ_DST_HOST |
WPI_FH_RX_CONFIG_IRQ_TIMEOUT(1));
(void)WPI_READ(sc, WPI_FH_RSSR_TBL); /* barrier */
wpi_nic_unlock(sc);
WPI_WRITE(sc, WPI_FH_RX_WPTR, (WPI_RX_RING_COUNT - 1) & ~7);
/* Initialize TX rings. */
if ((error = wpi_nic_lock(sc)) != 0)
return error;
wpi_prph_write(sc, WPI_ALM_SCHED_MODE, 2); /* bypass mode */
wpi_prph_write(sc, WPI_ALM_SCHED_ARASTAT, 1); /* enable RA0 */
/* Enable all 6 TX rings. */
wpi_prph_write(sc, WPI_ALM_SCHED_TXFACT, 0x3f);
wpi_prph_write(sc, WPI_ALM_SCHED_SBYPASS_MODE1, 0x10000);
wpi_prph_write(sc, WPI_ALM_SCHED_SBYPASS_MODE2, 0x30002);
wpi_prph_write(sc, WPI_ALM_SCHED_TXF4MF, 4);
wpi_prph_write(sc, WPI_ALM_SCHED_TXF5MF, 5);
/* Set physical address of TX rings. */
WPI_WRITE(sc, WPI_FH_TX_BASE, sc->shared_dma.paddr);
WPI_WRITE(sc, WPI_FH_MSG_CONFIG, 0xffff05a5);
/* Enable all DMA channels. */
for (chnl = 0; chnl < WPI_NDMACHNLS; chnl++) {
WPI_WRITE(sc, WPI_FH_CBBC_CTRL(chnl), 0);
WPI_WRITE(sc, WPI_FH_CBBC_BASE(chnl), 0);
WPI_WRITE(sc, WPI_FH_TX_CONFIG(chnl), 0x80200008);
}
wpi_nic_unlock(sc);
(void)WPI_READ(sc, WPI_FH_TX_BASE); /* barrier */
/* Clear "radio off" and "commands blocked" bits. */
WPI_WRITE(sc, WPI_UCODE_GP1_CLR, WPI_UCODE_GP1_RFKILL);
WPI_WRITE(sc, WPI_UCODE_GP1_CLR, WPI_UCODE_GP1_CMD_BLOCKED);
/* Clear pending interrupts. */
WPI_WRITE(sc, WPI_INT, 0xffffffff);
/* Enable interrupts. */
WPI_WRITE(sc, WPI_INT_MASK, WPI_INT_MASK_DEF);
/* _Really_ make sure "radio off" bit is cleared! */
WPI_WRITE(sc, WPI_UCODE_GP1_CLR, WPI_UCODE_GP1_RFKILL);
WPI_WRITE(sc, WPI_UCODE_GP1_CLR, WPI_UCODE_GP1_RFKILL);
if ((error = wpi_load_firmware(sc)) != 0) {
device_printf(sc->sc_dev,
"%s: could not load firmware, error %d\n", __func__,
error);
return error;
}
/* Wait at most one second for firmware alive notification. */
if ((error = mtx_sleep(sc, &sc->sc_mtx, PCATCH, "wpiinit", hz)) != 0) {
device_printf(sc->sc_dev,
"%s: timeout waiting for adapter to initialize, error %d\n",
__func__, error);
return error;
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
/* Do post-firmware initialization. */
return wpi_post_alive(sc);
}
static void
wpi_hw_stop(struct wpi_softc *sc)
{
uint8_t chnl, qid;
int ntries;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
if (WPI_READ(sc, WPI_UCODE_GP1) & WPI_UCODE_GP1_MAC_SLEEP)
wpi_nic_lock(sc);
WPI_WRITE(sc, WPI_RESET, WPI_RESET_NEVO);
/* Disable interrupts. */
WPI_WRITE(sc, WPI_INT_MASK, 0);
WPI_WRITE(sc, WPI_INT, 0xffffffff);
WPI_WRITE(sc, WPI_FH_INT, 0xffffffff);
/* Make sure we no longer hold the NIC lock. */
wpi_nic_unlock(sc);
if (wpi_nic_lock(sc) == 0) {
/* Stop TX scheduler. */
wpi_prph_write(sc, WPI_ALM_SCHED_MODE, 0);
wpi_prph_write(sc, WPI_ALM_SCHED_TXFACT, 0);
/* Stop all DMA channels. */
for (chnl = 0; chnl < WPI_NDMACHNLS; chnl++) {
WPI_WRITE(sc, WPI_FH_TX_CONFIG(chnl), 0);
for (ntries = 0; ntries < 200; ntries++) {
if (WPI_READ(sc, WPI_FH_TX_STATUS) &
WPI_FH_TX_STATUS_IDLE(chnl))
break;
DELAY(10);
}
}
wpi_nic_unlock(sc);
}
/* Stop RX ring. */
wpi_reset_rx_ring(sc);
/* Reset all TX rings. */
for (qid = 0; qid < WPI_NTXQUEUES; qid++)
wpi_reset_tx_ring(sc, &sc->txq[qid]);
if (wpi_nic_lock(sc) == 0) {
wpi_prph_write(sc, WPI_APMG_CLK_DIS,
WPI_APMG_CLK_CTRL_DMA_CLK_RQT);
wpi_nic_unlock(sc);
}
DELAY(5);
/* Power OFF adapter. */
wpi_apm_stop(sc);
}
static void
wpi_radio_on(void *arg0, int pending)
{
struct wpi_softc *sc = arg0;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
device_printf(sc->sc_dev, "RF switch: radio enabled\n");
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
WPI_LOCK(sc);
callout_stop(&sc->watchdog_rfkill);
WPI_UNLOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (vap != NULL)
ieee80211_init(vap);
}
static void
wpi_radio_off(void *arg0, int pending)
{
struct wpi_softc *sc = arg0;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
device_printf(sc->sc_dev, "RF switch: radio disabled\n");
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ieee80211_notify_radio(ic, 0);
wpi_stop(sc);
if (vap != NULL)
ieee80211_stop(vap);
WPI_LOCK(sc);
callout_reset(&sc->watchdog_rfkill, hz, wpi_watchdog_rfkill, sc);
WPI_UNLOCK(sc);
}
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
static int
wpi_init(struct wpi_softc *sc)
{
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
int error = 0;
WPI_LOCK(sc);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_BEGIN, __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (sc->sc_running != 0)
goto end;
/* Check that the radio is not disabled by hardware switch. */
if (!(WPI_READ(sc, WPI_GP_CNTRL) & WPI_GP_CNTRL_RFKILL)) {
device_printf(sc->sc_dev,
"RF switch: radio disabled (%s)\n", __func__);
callout_reset(&sc->watchdog_rfkill, hz, wpi_watchdog_rfkill,
sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
error = EINPROGRESS;
goto end;
}
/* Read firmware images from the filesystem. */
if ((error = wpi_read_firmware(sc)) != 0) {
device_printf(sc->sc_dev,
"%s: could not read firmware, error %d\n", __func__,
error);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
goto end;
}
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
sc->sc_running = 1;
/* Initialize hardware and upload firmware. */
error = wpi_hw_init(sc);
wpi_unload_firmware(sc);
if (error != 0) {
device_printf(sc->sc_dev,
"%s: could not initialize hardware, error %d\n", __func__,
error);
goto fail;
}
/* Configure adapter now that it is ready. */
if ((error = wpi_config(sc)) != 0) {
device_printf(sc->sc_dev,
"%s: could not configure device, error %d\n", __func__,
error);
goto fail;
}
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END, __func__);
WPI_UNLOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
return 0;
Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Details here: https://wiki.freebsd.org/projects/ifnet/net80211 Still, drivers: ndis, wtap, mwl, ipw, bwn, wi, upgt, uath were not tested. Changes to mwl, ipw, bwn, wi, upgt are trivial and chances of problems are low. The wtap wasn't compilable even before this change. But the ndis driver is complex, and it is likely to be broken with this commit. Help with testing and debugging it is appreciated. Differential Revision: D2655, D2740 Sponsored by: Nginx, Inc. Sponsored by: Netflix
2015-08-07 11:43:14 +00:00
fail: wpi_stop_locked(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
end: DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_END_ERR, __func__);
WPI_UNLOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
return error;
}
static void
wpi_stop_locked(struct wpi_softc *sc)
{
WPI_LOCK_ASSERT(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (sc->sc_running == 0)
return;
WPI_TX_LOCK(sc);
WPI_TXQ_LOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
sc->sc_running = 0;
WPI_TXQ_UNLOCK(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
WPI_TX_UNLOCK(sc);
WPI_TXQ_STATE_LOCK(sc);
callout_stop(&sc->tx_timeout);
WPI_TXQ_STATE_UNLOCK(sc);
WPI_RXON_LOCK(sc);
callout_stop(&sc->scan_timeout);
callout_stop(&sc->calib_to);
WPI_RXON_UNLOCK(sc);
/* Power OFF hardware. */
wpi_hw_stop(sc);
}
static void
wpi_stop(struct wpi_softc *sc)
{
WPI_LOCK(sc);
wpi_stop_locked(sc);
WPI_UNLOCK(sc);
}
/*
* Callback from net80211 to start a scan.
*/
static void
wpi_scan_start(struct ieee80211com *ic)
{
struct wpi_softc *sc = ic->ic_softc;
wpi_set_led(sc, WPI_LED_LINK, 20, 2);
}
/*
* Callback from net80211 to terminate a scan.
*/
static void
wpi_scan_end(struct ieee80211com *ic)
{
struct wpi_softc *sc = ic->ic_softc;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
if (vap->iv_state == IEEE80211_S_RUN)
wpi_set_led(sc, WPI_LED_LINK, 0, 1);
}
/**
* Called by the net80211 framework to indicate to the driver
* that the channel should be changed
*/
static void
wpi_set_channel(struct ieee80211com *ic)
{
const struct ieee80211_channel *c = ic->ic_curchan;
struct wpi_softc *sc = ic->ic_softc;
int error;
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
WPI_LOCK(sc);
sc->sc_rxtap.wr_chan_freq = htole16(c->ic_freq);
sc->sc_rxtap.wr_chan_flags = htole16(c->ic_flags);
WPI_UNLOCK(sc);
WPI_TX_LOCK(sc);
sc->sc_txtap.wt_chan_freq = htole16(c->ic_freq);
sc->sc_txtap.wt_chan_flags = htole16(c->ic_flags);
WPI_TX_UNLOCK(sc);
/*
* Only need to set the channel in Monitor mode. AP scanning and auth
* are already taken care of by their respective firmware commands.
*/
if (ic->ic_opmode == IEEE80211_M_MONITOR) {
WPI_RXON_LOCK(sc);
sc->rxon.chan = ieee80211_chan2ieee(ic, c);
if (IEEE80211_IS_CHAN_2GHZ(c)) {
sc->rxon.flags |= htole32(WPI_RXON_AUTO |
WPI_RXON_24GHZ);
} else {
sc->rxon.flags &= ~htole32(WPI_RXON_AUTO |
WPI_RXON_24GHZ);
}
if ((error = wpi_send_rxon(sc, 0, 1)) != 0)
device_printf(sc->sc_dev,
"%s: error %d setting channel\n", __func__,
error);
WPI_RXON_UNLOCK(sc);
}
}
/**
* Called by net80211 to indicate that we need to scan the current
* channel. The channel is previously be set via the wpi_set_channel
* callback.
*/
static void
wpi_scan_curchan(struct ieee80211_scan_state *ss, unsigned long maxdwell)
{
struct ieee80211vap *vap = ss->ss_vap;
struct ieee80211com *ic = vap->iv_ic;
struct wpi_softc *sc = ic->ic_softc;
int error;
WPI_RXON_LOCK(sc);
error = wpi_scan(sc, ic->ic_curchan);
WPI_RXON_UNLOCK(sc);
if (error != 0)
ieee80211_cancel_scan(vap);
}
/**
* Called by the net80211 framework to indicate
* the minimum dwell time has been met, terminate the scan.
* We don't actually terminate the scan as the firmware will notify
* us when it's finished and we have no way to interrupt it.
*/
static void
wpi_scan_mindwell(struct ieee80211_scan_state *ss)
{
/* NB: don't try to abort scan; wait for firmware to finish */
}
static void
wpi_hw_reset(void *arg, int pending)
{
struct wpi_softc *sc = arg;
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps);
DPRINTF(sc, WPI_DEBUG_TRACE, TRACE_STR_DOING, __func__);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
ieee80211_notify_radio(ic, 0);
if (vap != NULL && (ic->ic_flags & IEEE80211_F_SCAN))
ieee80211_cancel_scan(vap);
wpi_stop(sc);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
if (vap != NULL) {
ieee80211_stop(vap);
ieee80211_init(vap);
Replay r286410. Change KPI of how device drivers that provide wireless connectivity interact with the net80211 stack. Historical background: originally wireless devices created an interface, just like Ethernet devices do. Name of an interface matched the name of the driver that created. Later, wlan(4) layer was introduced, and the wlanX interfaces become the actual interface, leaving original ones as "a parent interface" of wlanX. Kernelwise, the KPI between net80211 layer and a driver became a mix of methods that pass a pointer to struct ifnet as identifier and methods that pass pointer to struct ieee80211com. From user point of view, the parent interface just hangs on in the ifconfig list, and user can't do anything useful with it. Now, the struct ifnet goes away. The struct ieee80211com is the only KPI between a device driver and net80211. Details: - The struct ieee80211com is embedded into drivers softc. - Packets are sent via new ic_transmit method, which is very much like the previous if_transmit. - Bringing parent up/down is done via new ic_parent method, which notifies driver about any changes: number of wlan(4) interfaces, number of them in promisc or allmulti state. - Device specific ioctls (if any) are received on new ic_ioctl method. - Packets/errors accounting are done by the stack. In certain cases, when driver experiences errors and can not attribute them to any specific interface, driver updates ic_oerrors or ic_ierrors counters. Details on interface configuration with new world order: - A sequence of commands needed to bring up wireless DOESN"T change. - /etc/rc.conf parameters DON'T change. - List of devices that can be used to create wlan(4) interfaces is now provided by net.wlan.devices sysctl. Most drivers in this change were converted by me, except of wpi(4), that was done by Andriy Voskoboinyk. Big thanks to Kevin Lo for testing changes to at least 8 drivers. Thanks to pluknet@, Oliver Hartmann, Olivier Cochard, gjb@, mmoll@, op@ and lev@, who also participated in testing. Reviewed by: adrian Sponsored by: Netflix Sponsored by: Nginx, Inc.
2015-08-27 08:56:39 +00:00
}
}