freebsd-skq/sys/dev/bwi/if_bwi.c

4022 lines
97 KiB
C
Raw Normal View History

/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 2007 The DragonFly Project. All rights reserved.
*
* This code is derived from software contributed to The DragonFly Project
* by Sepherosa Ziehau <sepherosa@gmail.com>
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* 3. Neither the name of The DragonFly Project nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific, prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
* COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
* AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $DragonFly: src/sys/dev/netif/bwi/if_bwi.c,v 1.19 2008/02/15 11:15:38 sephe Exp $
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#include "opt_bwi.h"
#include "opt_wlan.h"
#include <sys/param.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/bus.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/rman.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/taskqueue.h>
#include <net/if.h>
#include <net/if_var.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_arp.h>
#include <net/ethernet.h>
#include <net/if_llc.h>
#include <net80211/ieee80211_var.h>
#include <net80211/ieee80211_radiotap.h>
#include <net80211/ieee80211_regdomain.h>
#include <net80211/ieee80211_phy.h>
#include <net80211/ieee80211_ratectl.h>
#include <net/bpf.h>
#ifdef INET
#include <netinet/in.h>
#include <netinet/if_ether.h>
#endif
#include <machine/bus.h>
#include <dev/pci/pcivar.h>
#include <dev/pci/pcireg.h>
#include <dev/bwi/bitops.h>
#include <dev/bwi/if_bwireg.h>
#include <dev/bwi/if_bwivar.h>
#include <dev/bwi/bwimac.h>
#include <dev/bwi/bwirf.h>
struct bwi_clock_freq {
u_int clkfreq_min;
u_int clkfreq_max;
};
struct bwi_myaddr_bssid {
uint8_t myaddr[IEEE80211_ADDR_LEN];
uint8_t bssid[IEEE80211_ADDR_LEN];
} __packed;
static struct ieee80211vap *bwi_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 bwi_vap_delete(struct ieee80211vap *);
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 bwi_init(struct bwi_softc *);
static void bwi_parent(struct ieee80211com *);
static int bwi_transmit(struct ieee80211com *, struct mbuf *);
static void bwi_start_locked(struct bwi_softc *);
static int bwi_raw_xmit(struct ieee80211_node *, struct mbuf *,
const struct ieee80211_bpf_params *);
static void bwi_watchdog(void *);
static void bwi_scan_start(struct ieee80211com *);
static void bwi_getradiocaps(struct ieee80211com *, int, int *,
struct ieee80211_channel[]);
static void bwi_set_channel(struct ieee80211com *);
static void bwi_scan_end(struct ieee80211com *);
static int bwi_newstate(struct ieee80211vap *, enum ieee80211_state, int);
static void bwi_updateslot(struct ieee80211com *);
static int bwi_media_change(struct ifnet *);
static void bwi_calibrate(void *);
static int bwi_calc_rssi(struct bwi_softc *, const struct bwi_rxbuf_hdr *);
static int bwi_calc_noise(struct bwi_softc *);
static __inline uint8_t bwi_plcp2rate(uint32_t, enum ieee80211_phytype);
static void bwi_rx_radiotap(struct bwi_softc *, struct mbuf *,
struct bwi_rxbuf_hdr *, const void *, int, int, int);
static void bwi_restart(void *, int);
static void bwi_init_statechg(struct bwi_softc *, int);
static void bwi_stop(struct bwi_softc *, int);
static void bwi_stop_locked(struct bwi_softc *, int);
static int bwi_newbuf(struct bwi_softc *, int, int);
static int bwi_encap(struct bwi_softc *, int, struct mbuf *,
struct ieee80211_node *);
static int bwi_encap_raw(struct bwi_softc *, int, struct mbuf *,
struct ieee80211_node *,
const struct ieee80211_bpf_params *);
static void bwi_init_rxdesc_ring32(struct bwi_softc *, uint32_t,
bus_addr_t, int, int);
static void bwi_reset_rx_ring32(struct bwi_softc *, uint32_t);
static int bwi_init_tx_ring32(struct bwi_softc *, int);
static int bwi_init_rx_ring32(struct bwi_softc *);
static int bwi_init_txstats32(struct bwi_softc *);
static void bwi_free_tx_ring32(struct bwi_softc *, int);
static void bwi_free_rx_ring32(struct bwi_softc *);
static void bwi_free_txstats32(struct bwi_softc *);
static void bwi_setup_rx_desc32(struct bwi_softc *, int, bus_addr_t, int);
static void bwi_setup_tx_desc32(struct bwi_softc *, struct bwi_ring_data *,
int, bus_addr_t, int);
static int bwi_rxeof32(struct bwi_softc *);
static void bwi_start_tx32(struct bwi_softc *, uint32_t, int);
static void bwi_txeof_status32(struct bwi_softc *);
static int bwi_init_tx_ring64(struct bwi_softc *, int);
static int bwi_init_rx_ring64(struct bwi_softc *);
static int bwi_init_txstats64(struct bwi_softc *);
static void bwi_free_tx_ring64(struct bwi_softc *, int);
static void bwi_free_rx_ring64(struct bwi_softc *);
static void bwi_free_txstats64(struct bwi_softc *);
static void bwi_setup_rx_desc64(struct bwi_softc *, int, bus_addr_t, int);
static void bwi_setup_tx_desc64(struct bwi_softc *, struct bwi_ring_data *,
int, bus_addr_t, int);
static int bwi_rxeof64(struct bwi_softc *);
static void bwi_start_tx64(struct bwi_softc *, uint32_t, int);
static void bwi_txeof_status64(struct bwi_softc *);
static int bwi_rxeof(struct bwi_softc *, int);
static void _bwi_txeof(struct bwi_softc *, uint16_t, int, int);
static void bwi_txeof(struct bwi_softc *);
static void bwi_txeof_status(struct bwi_softc *, int);
static void bwi_enable_intrs(struct bwi_softc *, uint32_t);
static void bwi_disable_intrs(struct bwi_softc *, uint32_t);
static int bwi_dma_alloc(struct bwi_softc *);
static void bwi_dma_free(struct bwi_softc *);
static int bwi_dma_ring_alloc(struct bwi_softc *, bus_dma_tag_t,
struct bwi_ring_data *, bus_size_t,
uint32_t);
static int bwi_dma_mbuf_create(struct bwi_softc *);
static void bwi_dma_mbuf_destroy(struct bwi_softc *, int, int);
static int bwi_dma_txstats_alloc(struct bwi_softc *, uint32_t, bus_size_t);
static void bwi_dma_txstats_free(struct bwi_softc *);
static void bwi_dma_ring_addr(void *, bus_dma_segment_t *, int, int);
static void bwi_dma_buf_addr(void *, bus_dma_segment_t *, int,
bus_size_t, int);
static void bwi_power_on(struct bwi_softc *, int);
static int bwi_power_off(struct bwi_softc *, int);
static int bwi_set_clock_mode(struct bwi_softc *, enum bwi_clock_mode);
static int bwi_set_clock_delay(struct bwi_softc *);
static void bwi_get_clock_freq(struct bwi_softc *, struct bwi_clock_freq *);
static int bwi_get_pwron_delay(struct bwi_softc *sc);
static void bwi_set_addr_filter(struct bwi_softc *, uint16_t,
const uint8_t *);
static void bwi_set_bssid(struct bwi_softc *, const uint8_t *);
static void bwi_get_card_flags(struct bwi_softc *);
static void bwi_get_eaddr(struct bwi_softc *, uint16_t, uint8_t *);
static int bwi_bus_attach(struct bwi_softc *);
static int bwi_bbp_attach(struct bwi_softc *);
static int bwi_bbp_power_on(struct bwi_softc *, enum bwi_clock_mode);
static void bwi_bbp_power_off(struct bwi_softc *);
static const char *bwi_regwin_name(const struct bwi_regwin *);
static uint32_t bwi_regwin_disable_bits(struct bwi_softc *);
static void bwi_regwin_info(struct bwi_softc *, uint16_t *, uint8_t *);
static int bwi_regwin_select(struct bwi_softc *, int);
static void bwi_led_attach(struct bwi_softc *);
static void bwi_led_newstate(struct bwi_softc *, enum ieee80211_state);
static void bwi_led_event(struct bwi_softc *, int);
static void bwi_led_blink_start(struct bwi_softc *, int, int);
static void bwi_led_blink_next(void *);
static void bwi_led_blink_end(void *);
static const struct {
uint16_t did_min;
uint16_t did_max;
uint16_t bbp_id;
} bwi_bbpid_map[] = {
{ 0x4301, 0x4301, 0x4301 },
{ 0x4305, 0x4307, 0x4307 },
{ 0x4402, 0x4403, 0x4402 },
{ 0x4610, 0x4615, 0x4610 },
{ 0x4710, 0x4715, 0x4710 },
{ 0x4720, 0x4725, 0x4309 }
};
static const struct {
uint16_t bbp_id;
int nregwin;
} bwi_regwin_count[] = {
{ 0x4301, 5 },
{ 0x4306, 6 },
{ 0x4307, 5 },
{ 0x4310, 8 },
{ 0x4401, 3 },
{ 0x4402, 3 },
{ 0x4610, 9 },
{ 0x4704, 9 },
{ 0x4710, 9 },
{ 0x5365, 7 }
};
#define CLKSRC(src) \
[BWI_CLKSRC_ ## src] = { \
.freq_min = BWI_CLKSRC_ ##src## _FMIN, \
.freq_max = BWI_CLKSRC_ ##src## _FMAX \
}
static const struct {
u_int freq_min;
u_int freq_max;
} bwi_clkfreq[BWI_CLKSRC_MAX] = {
CLKSRC(LP_OSC),
CLKSRC(CS_OSC),
CLKSRC(PCI)
};
#undef CLKSRC
#define VENDOR_LED_ACT(vendor) \
{ \
.vid = PCI_VENDOR_##vendor, \
.led_act = { BWI_VENDOR_LED_ACT_##vendor } \
}
static const struct {
#define PCI_VENDOR_COMPAQ 0x0e11
#define PCI_VENDOR_LINKSYS 0x1737
uint16_t vid;
uint8_t led_act[BWI_LED_MAX];
} bwi_vendor_led_act[] = {
VENDOR_LED_ACT(COMPAQ),
VENDOR_LED_ACT(LINKSYS)
#undef PCI_VENDOR_LINKSYS
#undef PCI_VENDOR_COMPAQ
};
static const uint8_t bwi_default_led_act[BWI_LED_MAX] =
{ BWI_VENDOR_LED_ACT_DEFAULT };
#undef VENDOR_LED_ACT
static const struct {
int on_dur;
int off_dur;
} bwi_led_duration[109] = {
[0] = { 400, 100 },
[2] = { 150, 75 },
[4] = { 90, 45 },
[11] = { 66, 34 },
[12] = { 53, 26 },
[18] = { 42, 21 },
[22] = { 35, 17 },
[24] = { 32, 16 },
[36] = { 21, 10 },
[48] = { 16, 8 },
[72] = { 11, 5 },
[96] = { 9, 4 },
[108] = { 7, 3 }
};
static const uint8_t bwi_chan_2ghz[] =
{ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 };
#ifdef BWI_DEBUG
#ifdef BWI_DEBUG_VERBOSE
static uint32_t bwi_debug = BWI_DBG_ATTACH | BWI_DBG_INIT | BWI_DBG_TXPOWER;
#else
static uint32_t bwi_debug;
#endif
TUNABLE_INT("hw.bwi.debug", (int *)&bwi_debug);
#endif /* BWI_DEBUG */
static const uint8_t bwi_zero_addr[IEEE80211_ADDR_LEN];
uint16_t
bwi_read_sprom(struct bwi_softc *sc, uint16_t ofs)
{
return CSR_READ_2(sc, ofs + BWI_SPROM_START);
}
static __inline void
bwi_setup_desc32(struct bwi_softc *sc, struct bwi_desc32 *desc_array,
int ndesc, int desc_idx, bus_addr_t paddr, int buf_len,
int tx)
{
struct bwi_desc32 *desc = &desc_array[desc_idx];
uint32_t ctrl, addr, addr_hi, addr_lo;
addr_lo = __SHIFTOUT(paddr, BWI_DESC32_A_ADDR_MASK);
addr_hi = __SHIFTOUT(paddr, BWI_DESC32_A_FUNC_MASK);
addr = __SHIFTIN(addr_lo, BWI_DESC32_A_ADDR_MASK) |
__SHIFTIN(BWI_DESC32_A_FUNC_TXRX, BWI_DESC32_A_FUNC_MASK);
ctrl = __SHIFTIN(buf_len, BWI_DESC32_C_BUFLEN_MASK) |
__SHIFTIN(addr_hi, BWI_DESC32_C_ADDRHI_MASK);
if (desc_idx == ndesc - 1)
ctrl |= BWI_DESC32_C_EOR;
if (tx) {
/* XXX */
ctrl |= BWI_DESC32_C_FRAME_START |
BWI_DESC32_C_FRAME_END |
BWI_DESC32_C_INTR;
}
desc->addr = htole32(addr);
desc->ctrl = htole32(ctrl);
}
int
bwi_attach(struct bwi_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;
device_t dev = sc->sc_dev;
struct bwi_mac *mac;
struct bwi_phy *phy;
int i, error;
BWI_LOCK_INIT(sc);
/*
* Initialize taskq and various tasks
*/
sc->sc_tq = taskqueue_create("bwi_taskq", M_NOWAIT | M_ZERO,
taskqueue_thread_enqueue, &sc->sc_tq);
taskqueue_start_threads(&sc->sc_tq, 1, PI_NET, "%s taskq",
device_get_nameunit(dev));
TASK_INIT(&sc->sc_restart_task, 0, bwi_restart, sc);
callout_init_mtx(&sc->sc_calib_ch, &sc->sc_mtx, 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
mbufq_init(&sc->sc_snd, ifqmaxlen);
/*
* Initialize sysctl variables
*/
sc->sc_fw_version = BWI_FW_VERSION3;
sc->sc_led_idle = (2350 * hz) / 1000;
sc->sc_led_ticks = ticks - sc->sc_led_idle;
sc->sc_led_blink = 1;
sc->sc_txpwr_calib = 1;
#ifdef BWI_DEBUG
sc->sc_debug = bwi_debug;
#endif
bwi_power_on(sc, 1);
error = bwi_bbp_attach(sc);
if (error)
goto fail;
error = bwi_bbp_power_on(sc, BWI_CLOCK_MODE_FAST);
if (error)
goto fail;
if (BWI_REGWIN_EXIST(&sc->sc_com_regwin)) {
error = bwi_set_clock_delay(sc);
if (error)
goto fail;
error = bwi_set_clock_mode(sc, BWI_CLOCK_MODE_FAST);
if (error)
goto fail;
error = bwi_get_pwron_delay(sc);
if (error)
goto fail;
}
error = bwi_bus_attach(sc);
if (error)
goto fail;
bwi_get_card_flags(sc);
bwi_led_attach(sc);
for (i = 0; i < sc->sc_nmac; ++i) {
struct bwi_regwin *old;
mac = &sc->sc_mac[i];
error = bwi_regwin_switch(sc, &mac->mac_regwin, &old);
if (error)
goto fail;
error = bwi_mac_lateattach(mac);
if (error)
goto fail;
error = bwi_regwin_switch(sc, old, NULL);
if (error)
goto fail;
}
/*
* XXX First MAC is known to exist
* TODO2
*/
mac = &sc->sc_mac[0];
phy = &mac->mac_phy;
bwi_bbp_power_off(sc);
error = bwi_dma_alloc(sc);
if (error)
goto fail;
error = bwi_mac_fw_alloc(mac);
if (error)
goto fail;
callout_init_mtx(&sc->sc_watchdog_timer, &sc->sc_mtx, 0);
/*
* Setup ratesets, phytype, channels and get MAC address
*/
if (phy->phy_mode == IEEE80211_MODE_11B ||
phy->phy_mode == IEEE80211_MODE_11G) {
if (phy->phy_mode == IEEE80211_MODE_11B) {
ic->ic_phytype = IEEE80211_T_DS;
} else {
ic->ic_phytype = IEEE80211_T_OFDM;
}
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
bwi_get_eaddr(sc, BWI_SPROM_11BG_EADDR, ic->ic_macaddr);
if (IEEE80211_IS_MULTICAST(ic->ic_macaddr)) {
bwi_get_eaddr(sc, BWI_SPROM_11A_EADDR, ic->ic_macaddr);
if (IEEE80211_IS_MULTICAST(ic->ic_macaddr)) {
device_printf(dev,
"invalid MAC address: %6D\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
ic->ic_macaddr, ":");
}
}
} else if (phy->phy_mode == IEEE80211_MODE_11A) {
/* TODO:11A */
error = ENXIO;
goto fail;
} else {
panic("unknown phymode %d\n", phy->phy_mode);
}
/* Get locale */
sc->sc_locale = __SHIFTOUT(bwi_read_sprom(sc, BWI_SPROM_CARD_INFO),
BWI_SPROM_CARD_INFO_LOCALE);
DPRINTF(sc, BWI_DBG_ATTACH, "locale: %d\n", sc->sc_locale);
/* XXX use locale */
bwi_getradiocaps(ic, IEEE80211_CHAN_MAX, &ic->ic_nchans,
ic->ic_channels);
ic->ic_softc = sc;
ic->ic_name = device_get_nameunit(dev);
ic->ic_caps = IEEE80211_C_STA |
IEEE80211_C_SHSLOT |
IEEE80211_C_SHPREAMBLE |
IEEE80211_C_WPA |
IEEE80211_C_BGSCAN |
IEEE80211_C_MONITOR;
ic->ic_opmode = IEEE80211_M_STA;
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_headroom = sizeof(struct bwi_txbuf_hdr);
/* override default methods */
ic->ic_vap_create = bwi_vap_create;
ic->ic_vap_delete = bwi_vap_delete;
ic->ic_raw_xmit = bwi_raw_xmit;
ic->ic_updateslot = bwi_updateslot;
ic->ic_scan_start = bwi_scan_start;
ic->ic_scan_end = bwi_scan_end;
ic->ic_getradiocaps = bwi_getradiocaps;
ic->ic_set_channel = bwi_set_channel;
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 = bwi_transmit;
ic->ic_parent = bwi_parent;
sc->sc_rates = ieee80211_get_ratetable(ic->ic_curchan);
ieee80211_radiotap_attach(ic,
&sc->sc_tx_th.wt_ihdr, sizeof(sc->sc_tx_th),
BWI_TX_RADIOTAP_PRESENT,
&sc->sc_rx_th.wr_ihdr, sizeof(sc->sc_rx_th),
BWI_RX_RADIOTAP_PRESENT);
/*
* Add sysctl nodes
*/
SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO,
"fw_version", CTLFLAG_RD, &sc->sc_fw_version, 0,
"Firmware version");
SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO,
"led_idle", CTLFLAG_RW, &sc->sc_led_idle, 0,
"# ticks before LED enters idle state");
SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO,
"led_blink", CTLFLAG_RW, &sc->sc_led_blink, 0,
"Allow LED to blink");
SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO,
"txpwr_calib", CTLFLAG_RW, &sc->sc_txpwr_calib, 0,
"Enable software TX power calibration");
#ifdef BWI_DEBUG
SYSCTL_ADD_UINT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO,
"debug", CTLFLAG_RW, &sc->sc_debug, 0, "Debug flags");
#endif
if (bootverbose)
ieee80211_announce(ic);
return (0);
fail:
BWI_LOCK_DESTROY(sc);
return (error);
}
int
bwi_detach(struct bwi_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;
int i;
bwi_stop(sc, 1);
callout_drain(&sc->sc_led_blink_ch);
callout_drain(&sc->sc_calib_ch);
callout_drain(&sc->sc_watchdog_timer);
ieee80211_ifdetach(ic);
for (i = 0; i < sc->sc_nmac; ++i)
bwi_mac_detach(&sc->sc_mac[i]);
bwi_dma_free(sc);
taskqueue_free(sc->sc_tq);
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
mbufq_drain(&sc->sc_snd);
BWI_LOCK_DESTROY(sc);
return (0);
}
static struct ieee80211vap *
bwi_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 bwi_vap *bvp;
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
bvp = malloc(sizeof(struct bwi_vap), M_80211_VAP, M_WAITOK | M_ZERO);
vap = &bvp->bv_vap;
/* enable s/w bmiss handling for sta mode */
ieee80211_vap_setup(ic, vap, name, unit, opmode,
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
flags | IEEE80211_CLONE_NOBEACONS, bssid);
/* override default methods */
bvp->bv_newstate = vap->iv_newstate;
vap->iv_newstate = bwi_newstate;
#if 0
vap->iv_update_beacon = bwi_beacon_update;
#endif
ieee80211_ratectl_init(vap);
/* complete setup */
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_attach(vap, bwi_media_change, ieee80211_media_status,
mac);
ic->ic_opmode = opmode;
return vap;
}
static void
bwi_vap_delete(struct ieee80211vap *vap)
{
struct bwi_vap *bvp = BWI_VAP(vap);
ieee80211_ratectl_deinit(vap);
ieee80211_vap_detach(vap);
free(bvp, M_80211_VAP);
}
void
bwi_suspend(struct bwi_softc *sc)
{
bwi_stop(sc, 1);
}
void
bwi_resume(struct bwi_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
if (sc->sc_ic.ic_nrunning > 0)
bwi_init(sc);
}
int
bwi_shutdown(struct bwi_softc *sc)
{
bwi_stop(sc, 1);
return 0;
}
static void
bwi_power_on(struct bwi_softc *sc, int with_pll)
{
uint32_t gpio_in, gpio_out, gpio_en;
uint16_t status;
gpio_in = pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_IN, 4);
if (gpio_in & BWI_PCIM_GPIO_PWR_ON)
goto back;
gpio_out = pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_OUT, 4);
gpio_en = pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_ENABLE, 4);
gpio_out |= BWI_PCIM_GPIO_PWR_ON;
gpio_en |= BWI_PCIM_GPIO_PWR_ON;
if (with_pll) {
/* Turn off PLL first */
gpio_out |= BWI_PCIM_GPIO_PLL_PWR_OFF;
gpio_en |= BWI_PCIM_GPIO_PLL_PWR_OFF;
}
pci_write_config(sc->sc_dev, BWI_PCIR_GPIO_OUT, gpio_out, 4);
pci_write_config(sc->sc_dev, BWI_PCIR_GPIO_ENABLE, gpio_en, 4);
DELAY(1000);
if (with_pll) {
/* Turn on PLL */
gpio_out &= ~BWI_PCIM_GPIO_PLL_PWR_OFF;
pci_write_config(sc->sc_dev, BWI_PCIR_GPIO_OUT, gpio_out, 4);
DELAY(5000);
}
back:
/* Clear "Signaled Target Abort" */
status = pci_read_config(sc->sc_dev, PCIR_STATUS, 2);
status &= ~PCIM_STATUS_STABORT;
pci_write_config(sc->sc_dev, PCIR_STATUS, status, 2);
}
static int
bwi_power_off(struct bwi_softc *sc, int with_pll)
{
uint32_t gpio_out, gpio_en;
pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_IN, 4); /* dummy read */
gpio_out = pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_OUT, 4);
gpio_en = pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_ENABLE, 4);
gpio_out &= ~BWI_PCIM_GPIO_PWR_ON;
gpio_en |= BWI_PCIM_GPIO_PWR_ON;
if (with_pll) {
gpio_out |= BWI_PCIM_GPIO_PLL_PWR_OFF;
gpio_en |= BWI_PCIM_GPIO_PLL_PWR_OFF;
}
pci_write_config(sc->sc_dev, BWI_PCIR_GPIO_OUT, gpio_out, 4);
pci_write_config(sc->sc_dev, BWI_PCIR_GPIO_ENABLE, gpio_en, 4);
return 0;
}
int
bwi_regwin_switch(struct bwi_softc *sc, struct bwi_regwin *rw,
struct bwi_regwin **old_rw)
{
int error;
if (old_rw != NULL)
*old_rw = NULL;
if (!BWI_REGWIN_EXIST(rw))
return EINVAL;
if (sc->sc_cur_regwin != rw) {
error = bwi_regwin_select(sc, rw->rw_id);
if (error) {
device_printf(sc->sc_dev, "can't select regwin %d\n",
rw->rw_id);
return error;
}
}
if (old_rw != NULL)
*old_rw = sc->sc_cur_regwin;
sc->sc_cur_regwin = rw;
return 0;
}
static int
bwi_regwin_select(struct bwi_softc *sc, int id)
{
uint32_t win = BWI_PCIM_REGWIN(id);
int i;
#define RETRY_MAX 50
for (i = 0; i < RETRY_MAX; ++i) {
pci_write_config(sc->sc_dev, BWI_PCIR_SEL_REGWIN, win, 4);
if (pci_read_config(sc->sc_dev, BWI_PCIR_SEL_REGWIN, 4) == win)
return 0;
DELAY(10);
}
#undef RETRY_MAX
return ENXIO;
}
static void
bwi_regwin_info(struct bwi_softc *sc, uint16_t *type, uint8_t *rev)
{
uint32_t val;
val = CSR_READ_4(sc, BWI_ID_HI);
*type = BWI_ID_HI_REGWIN_TYPE(val);
*rev = BWI_ID_HI_REGWIN_REV(val);
DPRINTF(sc, BWI_DBG_ATTACH, "regwin: type 0x%03x, rev %d, "
"vendor 0x%04x\n", *type, *rev,
__SHIFTOUT(val, BWI_ID_HI_REGWIN_VENDOR_MASK));
}
static int
bwi_bbp_attach(struct bwi_softc *sc)
{
uint16_t bbp_id, rw_type;
uint8_t rw_rev;
uint32_t info;
int error, nregwin, i;
/*
* Get 0th regwin information
* NOTE: 0th regwin should exist
*/
error = bwi_regwin_select(sc, 0);
if (error) {
device_printf(sc->sc_dev, "can't select regwin 0\n");
return error;
}
bwi_regwin_info(sc, &rw_type, &rw_rev);
/*
* Find out BBP id
*/
bbp_id = 0;
info = 0;
if (rw_type == BWI_REGWIN_T_COM) {
info = CSR_READ_4(sc, BWI_INFO);
bbp_id = __SHIFTOUT(info, BWI_INFO_BBPID_MASK);
BWI_CREATE_REGWIN(&sc->sc_com_regwin, 0, rw_type, rw_rev);
sc->sc_cap = CSR_READ_4(sc, BWI_CAPABILITY);
} else {
for (i = 0; i < nitems(bwi_bbpid_map); ++i) {
if (sc->sc_pci_did >= bwi_bbpid_map[i].did_min &&
sc->sc_pci_did <= bwi_bbpid_map[i].did_max) {
bbp_id = bwi_bbpid_map[i].bbp_id;
break;
}
}
if (bbp_id == 0) {
device_printf(sc->sc_dev, "no BBP id for device id "
"0x%04x\n", sc->sc_pci_did);
return ENXIO;
}
info = __SHIFTIN(sc->sc_pci_revid, BWI_INFO_BBPREV_MASK) |
__SHIFTIN(0, BWI_INFO_BBPPKG_MASK);
}
/*
* Find out number of regwins
*/
nregwin = 0;
if (rw_type == BWI_REGWIN_T_COM && rw_rev >= 4) {
nregwin = __SHIFTOUT(info, BWI_INFO_NREGWIN_MASK);
} else {
for (i = 0; i < nitems(bwi_regwin_count); ++i) {
if (bwi_regwin_count[i].bbp_id == bbp_id) {
nregwin = bwi_regwin_count[i].nregwin;
break;
}
}
if (nregwin == 0) {
device_printf(sc->sc_dev, "no number of win for "
"BBP id 0x%04x\n", bbp_id);
return ENXIO;
}
}
/* Record BBP id/rev for later using */
sc->sc_bbp_id = bbp_id;
sc->sc_bbp_rev = __SHIFTOUT(info, BWI_INFO_BBPREV_MASK);
sc->sc_bbp_pkg = __SHIFTOUT(info, BWI_INFO_BBPPKG_MASK);
device_printf(sc->sc_dev, "BBP: id 0x%04x, rev 0x%x, pkg %d\n",
sc->sc_bbp_id, sc->sc_bbp_rev, sc->sc_bbp_pkg);
DPRINTF(sc, BWI_DBG_ATTACH, "nregwin %d, cap 0x%08x\n",
nregwin, sc->sc_cap);
/*
* Create rest of the regwins
*/
/* Don't re-create common regwin, if it is already created */
i = BWI_REGWIN_EXIST(&sc->sc_com_regwin) ? 1 : 0;
for (; i < nregwin; ++i) {
/*
* Get regwin information
*/
error = bwi_regwin_select(sc, i);
if (error) {
device_printf(sc->sc_dev,
"can't select regwin %d\n", i);
return error;
}
bwi_regwin_info(sc, &rw_type, &rw_rev);
/*
* Try attach:
* 1) Bus (PCI/PCIE) regwin
* 2) MAC regwin
* Ignore rest types of regwin
*/
if (rw_type == BWI_REGWIN_T_BUSPCI ||
rw_type == BWI_REGWIN_T_BUSPCIE) {
if (BWI_REGWIN_EXIST(&sc->sc_bus_regwin)) {
device_printf(sc->sc_dev,
"bus regwin already exists\n");
} else {
BWI_CREATE_REGWIN(&sc->sc_bus_regwin, i,
rw_type, rw_rev);
}
} else if (rw_type == BWI_REGWIN_T_MAC) {
/* XXX ignore return value */
bwi_mac_attach(sc, i, rw_rev);
}
}
/* At least one MAC shold exist */
if (!BWI_REGWIN_EXIST(&sc->sc_mac[0].mac_regwin)) {
device_printf(sc->sc_dev, "no MAC was found\n");
return ENXIO;
}
KASSERT(sc->sc_nmac > 0, ("no mac's"));
/* Bus regwin must exist */
if (!BWI_REGWIN_EXIST(&sc->sc_bus_regwin)) {
device_printf(sc->sc_dev, "no bus regwin was found\n");
return ENXIO;
}
/* Start with first MAC */
error = bwi_regwin_switch(sc, &sc->sc_mac[0].mac_regwin, NULL);
if (error)
return error;
return 0;
}
int
bwi_bus_init(struct bwi_softc *sc, struct bwi_mac *mac)
{
struct bwi_regwin *old, *bus;
uint32_t val;
int error;
bus = &sc->sc_bus_regwin;
KASSERT(sc->sc_cur_regwin == &mac->mac_regwin, ("not cur regwin"));
/*
* Tell bus to generate requested interrupts
*/
if (bus->rw_rev < 6 && bus->rw_type == BWI_REGWIN_T_BUSPCI) {
/*
* NOTE: Read BWI_FLAGS from MAC regwin
*/
val = CSR_READ_4(sc, BWI_FLAGS);
error = bwi_regwin_switch(sc, bus, &old);
if (error)
return error;
CSR_SETBITS_4(sc, BWI_INTRVEC, (val & BWI_FLAGS_INTR_MASK));
} else {
uint32_t mac_mask;
mac_mask = 1 << mac->mac_id;
error = bwi_regwin_switch(sc, bus, &old);
if (error)
return error;
val = pci_read_config(sc->sc_dev, BWI_PCIR_INTCTL, 4);
val |= mac_mask << 8;
pci_write_config(sc->sc_dev, BWI_PCIR_INTCTL, val, 4);
}
if (sc->sc_flags & BWI_F_BUS_INITED)
goto back;
if (bus->rw_type == BWI_REGWIN_T_BUSPCI) {
/*
* Enable prefetch and burst
*/
CSR_SETBITS_4(sc, BWI_BUS_CONFIG,
BWI_BUS_CONFIG_PREFETCH | BWI_BUS_CONFIG_BURST);
if (bus->rw_rev < 5) {
struct bwi_regwin *com = &sc->sc_com_regwin;
/*
* Configure timeouts for bus operation
*/
/*
* Set service timeout and request timeout
*/
CSR_SETBITS_4(sc, BWI_CONF_LO,
__SHIFTIN(BWI_CONF_LO_SERVTO, BWI_CONF_LO_SERVTO_MASK) |
__SHIFTIN(BWI_CONF_LO_REQTO, BWI_CONF_LO_REQTO_MASK));
/*
* If there is common regwin, we switch to that regwin
* and switch back to bus regwin once we have done.
*/
if (BWI_REGWIN_EXIST(com)) {
error = bwi_regwin_switch(sc, com, NULL);
if (error)
return error;
}
/* Let bus know what we have changed */
CSR_WRITE_4(sc, BWI_BUS_ADDR, BWI_BUS_ADDR_MAGIC);
CSR_READ_4(sc, BWI_BUS_ADDR); /* Flush */
CSR_WRITE_4(sc, BWI_BUS_DATA, 0);
CSR_READ_4(sc, BWI_BUS_DATA); /* Flush */
if (BWI_REGWIN_EXIST(com)) {
error = bwi_regwin_switch(sc, bus, NULL);
if (error)
return error;
}
} else if (bus->rw_rev >= 11) {
/*
* Enable memory read multiple
*/
CSR_SETBITS_4(sc, BWI_BUS_CONFIG, BWI_BUS_CONFIG_MRM);
}
} else {
/* TODO:PCIE */
}
sc->sc_flags |= BWI_F_BUS_INITED;
back:
return bwi_regwin_switch(sc, old, NULL);
}
static void
bwi_get_card_flags(struct bwi_softc *sc)
{
#define PCI_VENDOR_APPLE 0x106b
#define PCI_VENDOR_DELL 0x1028
sc->sc_card_flags = bwi_read_sprom(sc, BWI_SPROM_CARD_FLAGS);
if (sc->sc_card_flags == 0xffff)
sc->sc_card_flags = 0;
if (sc->sc_pci_subvid == PCI_VENDOR_DELL &&
sc->sc_bbp_id == BWI_BBPID_BCM4301 &&
sc->sc_pci_revid == 0x74)
sc->sc_card_flags |= BWI_CARD_F_BT_COEXIST;
if (sc->sc_pci_subvid == PCI_VENDOR_APPLE &&
sc->sc_pci_subdid == 0x4e && /* XXX */
sc->sc_pci_revid > 0x40)
sc->sc_card_flags |= BWI_CARD_F_PA_GPIO9;
DPRINTF(sc, BWI_DBG_ATTACH, "card flags 0x%04x\n", sc->sc_card_flags);
#undef PCI_VENDOR_DELL
#undef PCI_VENDOR_APPLE
}
static void
bwi_get_eaddr(struct bwi_softc *sc, uint16_t eaddr_ofs, uint8_t *eaddr)
{
int i;
for (i = 0; i < 3; ++i) {
*((uint16_t *)eaddr + i) =
htobe16(bwi_read_sprom(sc, eaddr_ofs + 2 * i));
}
}
static void
bwi_get_clock_freq(struct bwi_softc *sc, struct bwi_clock_freq *freq)
{
struct bwi_regwin *com;
uint32_t val;
u_int div;
int src;
bzero(freq, sizeof(*freq));
com = &sc->sc_com_regwin;
KASSERT(BWI_REGWIN_EXIST(com), ("regwin does not exist"));
KASSERT(sc->sc_cur_regwin == com, ("wrong regwin"));
KASSERT(sc->sc_cap & BWI_CAP_CLKMODE, ("wrong clock mode"));
/*
* Calculate clock frequency
*/
src = -1;
div = 0;
if (com->rw_rev < 6) {
val = pci_read_config(sc->sc_dev, BWI_PCIR_GPIO_OUT, 4);
if (val & BWI_PCIM_GPIO_OUT_CLKSRC) {
src = BWI_CLKSRC_PCI;
div = 64;
} else {
src = BWI_CLKSRC_CS_OSC;
div = 32;
}
} else if (com->rw_rev < 10) {
val = CSR_READ_4(sc, BWI_CLOCK_CTRL);
src = __SHIFTOUT(val, BWI_CLOCK_CTRL_CLKSRC);
if (src == BWI_CLKSRC_LP_OSC) {
div = 1;
} else {
div = (__SHIFTOUT(val, BWI_CLOCK_CTRL_FDIV) + 1) << 2;
/* Unknown source */
if (src >= BWI_CLKSRC_MAX)
src = BWI_CLKSRC_CS_OSC;
}
} else {
val = CSR_READ_4(sc, BWI_CLOCK_INFO);
src = BWI_CLKSRC_CS_OSC;
div = (__SHIFTOUT(val, BWI_CLOCK_INFO_FDIV) + 1) << 2;
}
KASSERT(src >= 0 && src < BWI_CLKSRC_MAX, ("bad src %d", src));
KASSERT(div != 0, ("div zero"));
DPRINTF(sc, BWI_DBG_ATTACH, "clksrc %s\n",
src == BWI_CLKSRC_PCI ? "PCI" :
(src == BWI_CLKSRC_LP_OSC ? "LP_OSC" : "CS_OSC"));
freq->clkfreq_min = bwi_clkfreq[src].freq_min / div;
freq->clkfreq_max = bwi_clkfreq[src].freq_max / div;
DPRINTF(sc, BWI_DBG_ATTACH, "clkfreq min %u, max %u\n",
freq->clkfreq_min, freq->clkfreq_max);
}
static int
bwi_set_clock_mode(struct bwi_softc *sc, enum bwi_clock_mode clk_mode)
{
struct bwi_regwin *old, *com;
uint32_t clk_ctrl, clk_src;
int error, pwr_off = 0;
com = &sc->sc_com_regwin;
if (!BWI_REGWIN_EXIST(com))
return 0;
if (com->rw_rev >= 10 || com->rw_rev < 6)
return 0;
/*
* For common regwin whose rev is [6, 10), the chip
* must be capable to change clock mode.
*/
if ((sc->sc_cap & BWI_CAP_CLKMODE) == 0)
return 0;
error = bwi_regwin_switch(sc, com, &old);
if (error)
return error;
if (clk_mode == BWI_CLOCK_MODE_FAST)
bwi_power_on(sc, 0); /* Don't turn on PLL */
clk_ctrl = CSR_READ_4(sc, BWI_CLOCK_CTRL);
clk_src = __SHIFTOUT(clk_ctrl, BWI_CLOCK_CTRL_CLKSRC);
switch (clk_mode) {
case BWI_CLOCK_MODE_FAST:
clk_ctrl &= ~BWI_CLOCK_CTRL_SLOW;
clk_ctrl |= BWI_CLOCK_CTRL_IGNPLL;
break;
case BWI_CLOCK_MODE_SLOW:
clk_ctrl |= BWI_CLOCK_CTRL_SLOW;
break;
case BWI_CLOCK_MODE_DYN:
clk_ctrl &= ~(BWI_CLOCK_CTRL_SLOW |
BWI_CLOCK_CTRL_IGNPLL |
BWI_CLOCK_CTRL_NODYN);
if (clk_src != BWI_CLKSRC_CS_OSC) {
clk_ctrl |= BWI_CLOCK_CTRL_NODYN;
pwr_off = 1;
}
break;
}
CSR_WRITE_4(sc, BWI_CLOCK_CTRL, clk_ctrl);
if (pwr_off)
bwi_power_off(sc, 0); /* Leave PLL as it is */
return bwi_regwin_switch(sc, old, NULL);
}
static int
bwi_set_clock_delay(struct bwi_softc *sc)
{
struct bwi_regwin *old, *com;
int error;
com = &sc->sc_com_regwin;
if (!BWI_REGWIN_EXIST(com))
return 0;
error = bwi_regwin_switch(sc, com, &old);
if (error)
return error;
if (sc->sc_bbp_id == BWI_BBPID_BCM4321) {
if (sc->sc_bbp_rev == 0)
CSR_WRITE_4(sc, BWI_CONTROL, BWI_CONTROL_MAGIC0);
else if (sc->sc_bbp_rev == 1)
CSR_WRITE_4(sc, BWI_CONTROL, BWI_CONTROL_MAGIC1);
}
if (sc->sc_cap & BWI_CAP_CLKMODE) {
if (com->rw_rev >= 10) {
CSR_FILT_SETBITS_4(sc, BWI_CLOCK_INFO, 0xffff, 0x40000);
} else {
struct bwi_clock_freq freq;
bwi_get_clock_freq(sc, &freq);
CSR_WRITE_4(sc, BWI_PLL_ON_DELAY,
howmany(freq.clkfreq_max * 150, 1000000));
CSR_WRITE_4(sc, BWI_FREQ_SEL_DELAY,
howmany(freq.clkfreq_max * 15, 1000000));
}
}
return bwi_regwin_switch(sc, old, NULL);
}
static 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
bwi_init(struct bwi_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;
BWI_LOCK(sc);
bwi_init_statechg(sc, 1);
BWI_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 (sc->sc_flags & BWI_F_RUNNING)
ieee80211_start_all(ic); /* start all vap's */
}
static void
bwi_init_statechg(struct bwi_softc *sc, int statechg)
{
struct bwi_mac *mac;
int 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
BWI_ASSERT_LOCKED(sc);
bwi_stop_locked(sc, statechg);
bwi_bbp_power_on(sc, BWI_CLOCK_MODE_FAST);
/* TODO: 2 MAC */
mac = &sc->sc_mac[0];
error = bwi_regwin_switch(sc, &mac->mac_regwin, NULL);
if (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
device_printf(sc->sc_dev, "%s: error %d on regwin switch\n",
__func__, error);
goto bad;
}
error = bwi_mac_init(mac);
if (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
device_printf(sc->sc_dev, "%s: error %d on MAC init\n",
__func__, error);
goto bad;
}
bwi_bbp_power_on(sc, BWI_CLOCK_MODE_DYN);
bwi_set_bssid(sc, bwi_zero_addr); /* Clear BSSID */
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
bwi_set_addr_filter(sc, BWI_ADDR_FILTER_MYADDR, sc->sc_ic.ic_macaddr);
bwi_mac_reset_hwkeys(mac);
if ((mac->mac_flags & BWI_MAC_F_HAS_TXSTATS) == 0) {
int i;
#define NRETRY 1000
/*
* Drain any possible pending TX status
*/
for (i = 0; i < NRETRY; ++i) {
if ((CSR_READ_4(sc, BWI_TXSTATUS0) &
BWI_TXSTATUS0_VALID) == 0)
break;
CSR_READ_4(sc, BWI_TXSTATUS1);
}
if (i == NRETRY)
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
device_printf(sc->sc_dev,
"%s: can't drain TX status\n", __func__);
#undef NRETRY
}
if (mac->mac_phy.phy_mode == IEEE80211_MODE_11G)
bwi_mac_updateslot(mac, 1);
/* Start MAC */
error = bwi_mac_start(mac);
if (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
device_printf(sc->sc_dev, "%s: error %d starting MAC\n",
__func__, error);
goto bad;
}
/* Clear stop flag before enabling interrupt */
sc->sc_flags &= ~BWI_F_STOP;
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_flags |= BWI_F_RUNNING;
callout_reset(&sc->sc_watchdog_timer, hz, bwi_watchdog, sc);
/* Enable intrs */
bwi_enable_intrs(sc, BWI_INIT_INTRS);
return;
bad:
bwi_stop_locked(sc, 1);
}
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
bwi_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 bwi_softc *sc = ic->ic_softc;
int startall = 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
BWI_LOCK(sc);
if (ic->ic_nrunning > 0) {
struct bwi_mac *mac;
int promisc = -1;
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
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d",
sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
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
if (ic->ic_promisc > 0 && (sc->sc_flags & BWI_F_PROMISC) == 0) {
promisc = 1;
sc->sc_flags |= BWI_F_PROMISC;
} else if (ic->ic_promisc == 0 &&
(sc->sc_flags & BWI_F_PROMISC) != 0) {
promisc = 0;
sc->sc_flags &= ~BWI_F_PROMISC;
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
if (promisc >= 0)
bwi_mac_set_promisc(mac, 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_nrunning > 0) {
if ((sc->sc_flags & BWI_F_RUNNING) == 0) {
bwi_init_statechg(sc, 1);
startall = 1;
}
} else if (sc->sc_flags & BWI_F_RUNNING)
bwi_stop_locked(sc, 1);
BWI_UNLOCK(sc);
if (startall)
ieee80211_start_all(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
static int
bwi_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 bwi_softc *sc = ic->ic_softc;
int error;
BWI_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
if ((sc->sc_flags & BWI_F_RUNNING) == 0) {
BWI_UNLOCK(sc);
return (ENXIO);
}
error = mbufq_enqueue(&sc->sc_snd, m);
if (error) {
BWI_UNLOCK(sc);
return (error);
}
bwi_start_locked(sc);
BWI_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);
}
static 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
bwi_start_locked(struct bwi_softc *sc)
{
struct bwi_txbuf_data *tbd = &sc->sc_tx_bdata[BWI_TX_DATA_RING];
struct ieee80211_frame *wh;
struct ieee80211_node *ni;
struct mbuf *m;
int trans, idx;
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
BWI_ASSERT_LOCKED(sc);
trans = 0;
idx = tbd->tbd_idx;
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
while (tbd->tbd_buf[idx].tb_mbuf == NULL &&
tbd->tbd_used + BWI_TX_NSPRDESC < BWI_TX_NDESC &&
(m = mbufq_dequeue(&sc->sc_snd)) != NULL) {
ni = (struct ieee80211_node *) m->m_pkthdr.rcvif;
wh = mtod(m, struct ieee80211_frame *);
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 ((wh->i_fc[1] & IEEE80211_FC1_PROTECTED) != 0 &&
ieee80211_crypto_encap(ni, m) == NULL) {
if_inc_counter(ni->ni_vap->iv_ifp,
IFCOUNTER_OERRORS, 1);
ieee80211_free_node(ni);
m_freem(m);
continue;
}
if (bwi_encap(sc, idx, m, ni) != 0) {
/* 'm' is freed in bwi_encap() if we reach here */
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 (ni != NULL) {
if_inc_counter(ni->ni_vap->iv_ifp,
IFCOUNTER_OERRORS, 1);
ieee80211_free_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
} else
counter_u64_add(sc->sc_ic.ic_oerrors, 1);
continue;
}
trans = 1;
tbd->tbd_used++;
idx = (idx + 1) % BWI_TX_NDESC;
}
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
tbd->tbd_idx = idx;
if (trans)
sc->sc_tx_timer = 5;
}
static int
bwi_raw_xmit(struct ieee80211_node *ni, struct mbuf *m,
const struct ieee80211_bpf_params *params)
{
struct ieee80211com *ic = ni->ni_ic;
struct bwi_softc *sc = ic->ic_softc;
/* XXX wme? */
struct bwi_txbuf_data *tbd = &sc->sc_tx_bdata[BWI_TX_DATA_RING];
int idx, 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
if ((sc->sc_flags & BWI_F_RUNNING) == 0) {
m_freem(m);
return ENETDOWN;
}
BWI_LOCK(sc);
idx = tbd->tbd_idx;
KASSERT(tbd->tbd_buf[idx].tb_mbuf == NULL, ("slot %d not empty", idx));
if (params == NULL) {
/*
* Legacy path; interpret frame contents to decide
* precisely how to send the frame.
*/
error = bwi_encap(sc, idx, m, ni);
} else {
/*
* Caller supplied explicit parameters to use in
* sending the frame.
*/
error = bwi_encap_raw(sc, idx, m, ni, params);
}
if (error == 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
tbd->tbd_used++;
tbd->tbd_idx = (idx + 1) % BWI_TX_NDESC;
sc->sc_tx_timer = 5;
}
BWI_UNLOCK(sc);
return error;
}
static void
bwi_watchdog(void *arg)
{
struct bwi_softc *sc;
sc = arg;
BWI_ASSERT_LOCKED(sc);
if (sc->sc_tx_timer != 0 && --sc->sc_tx_timer == 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
device_printf(sc->sc_dev, "watchdog timeout\n");
counter_u64_add(sc->sc_ic.ic_oerrors, 1);
taskqueue_enqueue(sc->sc_tq, &sc->sc_restart_task);
}
callout_reset(&sc->sc_watchdog_timer, hz, bwi_watchdog, sc);
}
static void
bwi_stop(struct bwi_softc *sc, int statechg)
{
BWI_LOCK(sc);
bwi_stop_locked(sc, statechg);
BWI_UNLOCK(sc);
}
static void
bwi_stop_locked(struct bwi_softc *sc, int statechg)
{
struct bwi_mac *mac;
int i, error, pwr_off = 0;
BWI_ASSERT_LOCKED(sc);
callout_stop(&sc->sc_calib_ch);
callout_stop(&sc->sc_led_blink_ch);
sc->sc_led_blinking = 0;
sc->sc_flags |= BWI_F_STOP;
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_flags & BWI_F_RUNNING) {
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
bwi_disable_intrs(sc, BWI_ALL_INTRS);
CSR_READ_4(sc, BWI_MAC_INTR_MASK);
bwi_mac_stop(mac);
}
for (i = 0; i < sc->sc_nmac; ++i) {
struct bwi_regwin *old_rw;
mac = &sc->sc_mac[i];
if ((mac->mac_flags & BWI_MAC_F_INITED) == 0)
continue;
error = bwi_regwin_switch(sc, &mac->mac_regwin, &old_rw);
if (error)
continue;
bwi_mac_shutdown(mac);
pwr_off = 1;
bwi_regwin_switch(sc, old_rw, NULL);
}
if (pwr_off)
bwi_bbp_power_off(sc);
sc->sc_tx_timer = 0;
callout_stop(&sc->sc_watchdog_timer);
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_flags &= ~BWI_F_RUNNING;
}
void
bwi_intr(void *xsc)
{
struct bwi_softc *sc = xsc;
struct bwi_mac *mac;
uint32_t intr_status;
uint32_t txrx_intr_status[BWI_TXRX_NRING];
int i, txrx_error, tx = 0, rx_data = -1;
BWI_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
if ((sc->sc_flags & BWI_F_RUNNING) == 0 ||
(sc->sc_flags & BWI_F_STOP)) {
BWI_UNLOCK(sc);
return;
}
/*
* Get interrupt status
*/
intr_status = CSR_READ_4(sc, BWI_MAC_INTR_STATUS);
if (intr_status == 0xffffffff) { /* Not for us */
BWI_UNLOCK(sc);
return;
}
DPRINTF(sc, BWI_DBG_INTR, "intr status 0x%08x\n", intr_status);
intr_status &= CSR_READ_4(sc, BWI_MAC_INTR_MASK);
if (intr_status == 0) { /* Nothing is interesting */
BWI_UNLOCK(sc);
return;
}
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
txrx_error = 0;
DPRINTF(sc, BWI_DBG_INTR, "%s\n", "TX/RX intr");
for (i = 0; i < BWI_TXRX_NRING; ++i) {
uint32_t mask;
if (BWI_TXRX_IS_RX(i))
mask = BWI_TXRX_RX_INTRS;
else
mask = BWI_TXRX_TX_INTRS;
txrx_intr_status[i] =
CSR_READ_4(sc, BWI_TXRX_INTR_STATUS(i)) & mask;
_DPRINTF(sc, BWI_DBG_INTR, ", %d 0x%08x",
i, txrx_intr_status[i]);
if (txrx_intr_status[i] & BWI_TXRX_INTR_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
device_printf(sc->sc_dev,
"%s: intr fatal TX/RX (%d) error 0x%08x\n",
__func__, i, txrx_intr_status[i]);
txrx_error = 1;
}
}
_DPRINTF(sc, BWI_DBG_INTR, "%s\n", "");
/*
* Acknowledge interrupt
*/
CSR_WRITE_4(sc, BWI_MAC_INTR_STATUS, intr_status);
for (i = 0; i < BWI_TXRX_NRING; ++i)
CSR_WRITE_4(sc, BWI_TXRX_INTR_STATUS(i), txrx_intr_status[i]);
/* Disable all interrupts */
bwi_disable_intrs(sc, BWI_ALL_INTRS);
/*
* http://bcm-specs.sipsolutions.net/Interrupts
* Says for this bit (0x800):
* "Fatal Error
*
* We got this one while testing things when by accident the
* template ram wasn't set to big endian when it should have
* been after writing the initial values. It keeps on being
* triggered, the only way to stop it seems to shut down the
* chip."
*
* Suggesting that we should never get it and if we do we're not
* feeding TX packets into the MAC correctly if we do... Apparently,
* it is valid only on mac version 5 and higher, but I couldn't
* find a reference for that... Since I see them from time to time
* on my card, this suggests an error in the tx path still...
*/
if (intr_status & BWI_INTR_PHY_TXERR) {
if (mac->mac_flags & BWI_MAC_F_PHYE_RESET) {
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
device_printf(sc->sc_dev, "%s: intr PHY TX error\n",
__func__);
taskqueue_enqueue(sc->sc_tq, &sc->sc_restart_task);
BWI_UNLOCK(sc);
return;
}
}
if (txrx_error) {
/* TODO: reset device */
}
if (intr_status & BWI_INTR_TBTT)
bwi_mac_config_ps(mac);
if (intr_status & BWI_INTR_EO_ATIM)
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
device_printf(sc->sc_dev, "EO_ATIM\n");
if (intr_status & BWI_INTR_PMQ) {
for (;;) {
if ((CSR_READ_4(sc, BWI_MAC_PS_STATUS) & 0x8) == 0)
break;
}
CSR_WRITE_2(sc, BWI_MAC_PS_STATUS, 0x2);
}
if (intr_status & BWI_INTR_NOISE)
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
device_printf(sc->sc_dev, "intr noise\n");
if (txrx_intr_status[0] & BWI_TXRX_INTR_RX) {
rx_data = sc->sc_rxeof(sc);
if (sc->sc_flags & BWI_F_STOP) {
BWI_UNLOCK(sc);
return;
}
}
if (txrx_intr_status[3] & BWI_TXRX_INTR_RX) {
sc->sc_txeof_status(sc);
tx = 1;
}
if (intr_status & BWI_INTR_TX_DONE) {
bwi_txeof(sc);
tx = 1;
}
/* Re-enable interrupts */
bwi_enable_intrs(sc, BWI_INIT_INTRS);
if (sc->sc_blink_led != NULL && sc->sc_led_blink) {
int evt = BWI_LED_EVENT_NONE;
if (tx && rx_data > 0) {
if (sc->sc_rx_rate > sc->sc_tx_rate)
evt = BWI_LED_EVENT_RX;
else
evt = BWI_LED_EVENT_TX;
} else if (tx) {
evt = BWI_LED_EVENT_TX;
} else if (rx_data > 0) {
evt = BWI_LED_EVENT_RX;
} else if (rx_data == 0) {
evt = BWI_LED_EVENT_POLL;
}
if (evt != BWI_LED_EVENT_NONE)
bwi_led_event(sc, evt);
}
BWI_UNLOCK(sc);
}
static void
bwi_scan_start(struct ieee80211com *ic)
{
struct bwi_softc *sc = ic->ic_softc;
BWI_LOCK(sc);
/* Enable MAC beacon promiscuity */
CSR_SETBITS_4(sc, BWI_MAC_STATUS, BWI_MAC_STATUS_PASS_BCN);
BWI_UNLOCK(sc);
}
static void
bwi_getradiocaps(struct ieee80211com *ic,
int maxchans, int *nchans, struct ieee80211_channel chans[])
{
struct bwi_softc *sc = ic->ic_softc;
struct bwi_mac *mac;
struct bwi_phy *phy;
uint8_t bands[IEEE80211_MODE_BYTES];
/*
* XXX First MAC is known to exist
* TODO2
*/
mac = &sc->sc_mac[0];
phy = &mac->mac_phy;
memset(bands, 0, sizeof(bands));
switch (phy->phy_mode) {
case IEEE80211_MODE_11G:
setbit(bands, IEEE80211_MODE_11G);
/* FALLTHROUGH */
case IEEE80211_MODE_11B:
setbit(bands, IEEE80211_MODE_11B);
break;
case IEEE80211_MODE_11A:
/* TODO:11A */
setbit(bands, IEEE80211_MODE_11A);
device_printf(sc->sc_dev, "no 11a support\n");
return;
default:
panic("unknown phymode %d\n", phy->phy_mode);
}
ieee80211_add_channel_list_2ghz(chans, maxchans, nchans,
bwi_chan_2ghz, nitems(bwi_chan_2ghz), bands, 0);
}
static void
bwi_set_channel(struct ieee80211com *ic)
{
struct bwi_softc *sc = ic->ic_softc;
struct ieee80211_channel *c = ic->ic_curchan;
struct bwi_mac *mac;
BWI_LOCK(sc);
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
bwi_rf_set_chan(mac, ieee80211_chan2ieee(ic, c), 0);
sc->sc_rates = ieee80211_get_ratetable(c);
/*
* Setup radio tap channel freq and flags
*/
sc->sc_tx_th.wt_chan_freq = sc->sc_rx_th.wr_chan_freq =
htole16(c->ic_freq);
sc->sc_tx_th.wt_chan_flags = sc->sc_rx_th.wr_chan_flags =
htole16(c->ic_flags & 0xffff);
BWI_UNLOCK(sc);
}
static void
bwi_scan_end(struct ieee80211com *ic)
{
struct bwi_softc *sc = ic->ic_softc;
BWI_LOCK(sc);
CSR_CLRBITS_4(sc, BWI_MAC_STATUS, BWI_MAC_STATUS_PASS_BCN);
BWI_UNLOCK(sc);
}
static int
bwi_newstate(struct ieee80211vap *vap, enum ieee80211_state nstate, int arg)
{
struct bwi_vap *bvp = BWI_VAP(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
struct ieee80211com *ic= vap->iv_ic;
struct bwi_softc *sc = ic->ic_softc;
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
enum ieee80211_state ostate = vap->iv_state;
struct bwi_mac *mac;
int error;
BWI_LOCK(sc);
callout_stop(&sc->sc_calib_ch);
if (nstate == IEEE80211_S_INIT)
sc->sc_txpwrcb_type = BWI_TXPWR_INIT;
bwi_led_newstate(sc, nstate);
error = bvp->bv_newstate(vap, nstate, arg);
if (error != 0)
goto back;
/*
* Clear the BSSID when we stop a STA
*/
if (vap->iv_opmode == IEEE80211_M_STA) {
if (ostate == IEEE80211_S_RUN && nstate != IEEE80211_S_RUN) {
/*
* Clear out the BSSID. If we reassociate to
* the same AP, this will reinialize things
* correctly...
*/
if (ic->ic_opmode == IEEE80211_M_STA &&
!(sc->sc_flags & BWI_F_STOP))
bwi_set_bssid(sc, bwi_zero_addr);
}
}
if (vap->iv_opmode == IEEE80211_M_MONITOR) {
/* Nothing to do */
} else if (nstate == IEEE80211_S_RUN) {
bwi_set_bssid(sc, vap->iv_bss->ni_bssid);
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
/* Initial TX power calibration */
bwi_mac_calibrate_txpower(mac, BWI_TXPWR_INIT);
#ifdef notyet
sc->sc_txpwrcb_type = BWI_TXPWR_FORCE;
#else
sc->sc_txpwrcb_type = BWI_TXPWR_CALIB;
#endif
callout_reset(&sc->sc_calib_ch, hz, bwi_calibrate, sc);
}
back:
BWI_UNLOCK(sc);
return error;
}
static int
bwi_media_change(struct ifnet *ifp)
{
int error = ieee80211_media_change(ifp);
/* NB: only the fixed rate can change and that doesn't need a reset */
return (error == ENETRESET ? 0 : error);
}
static int
bwi_dma_alloc(struct bwi_softc *sc)
{
int error, i, has_txstats;
bus_addr_t lowaddr = 0;
bus_size_t tx_ring_sz, rx_ring_sz, desc_sz = 0;
uint32_t txrx_ctrl_step = 0;
has_txstats = 0;
for (i = 0; i < sc->sc_nmac; ++i) {
if (sc->sc_mac[i].mac_flags & BWI_MAC_F_HAS_TXSTATS) {
has_txstats = 1;
break;
}
}
switch (sc->sc_bus_space) {
case BWI_BUS_SPACE_30BIT:
case BWI_BUS_SPACE_32BIT:
if (sc->sc_bus_space == BWI_BUS_SPACE_30BIT)
lowaddr = BWI_BUS_SPACE_MAXADDR;
else
lowaddr = BUS_SPACE_MAXADDR_32BIT;
desc_sz = sizeof(struct bwi_desc32);
txrx_ctrl_step = 0x20;
sc->sc_init_tx_ring = bwi_init_tx_ring32;
sc->sc_free_tx_ring = bwi_free_tx_ring32;
sc->sc_init_rx_ring = bwi_init_rx_ring32;
sc->sc_free_rx_ring = bwi_free_rx_ring32;
sc->sc_setup_rxdesc = bwi_setup_rx_desc32;
sc->sc_setup_txdesc = bwi_setup_tx_desc32;
sc->sc_rxeof = bwi_rxeof32;
sc->sc_start_tx = bwi_start_tx32;
if (has_txstats) {
sc->sc_init_txstats = bwi_init_txstats32;
sc->sc_free_txstats = bwi_free_txstats32;
sc->sc_txeof_status = bwi_txeof_status32;
}
break;
case BWI_BUS_SPACE_64BIT:
lowaddr = BUS_SPACE_MAXADDR; /* XXX */
desc_sz = sizeof(struct bwi_desc64);
txrx_ctrl_step = 0x40;
sc->sc_init_tx_ring = bwi_init_tx_ring64;
sc->sc_free_tx_ring = bwi_free_tx_ring64;
sc->sc_init_rx_ring = bwi_init_rx_ring64;
sc->sc_free_rx_ring = bwi_free_rx_ring64;
sc->sc_setup_rxdesc = bwi_setup_rx_desc64;
sc->sc_setup_txdesc = bwi_setup_tx_desc64;
sc->sc_rxeof = bwi_rxeof64;
sc->sc_start_tx = bwi_start_tx64;
if (has_txstats) {
sc->sc_init_txstats = bwi_init_txstats64;
sc->sc_free_txstats = bwi_free_txstats64;
sc->sc_txeof_status = bwi_txeof_status64;
}
break;
}
KASSERT(lowaddr != 0, ("lowaddr zero"));
KASSERT(desc_sz != 0, ("desc_sz zero"));
KASSERT(txrx_ctrl_step != 0, ("txrx_ctrl_step zero"));
tx_ring_sz = roundup(desc_sz * BWI_TX_NDESC, BWI_RING_ALIGN);
rx_ring_sz = roundup(desc_sz * BWI_RX_NDESC, BWI_RING_ALIGN);
/*
* Create top level DMA tag
*/
error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), /* parent */
BWI_ALIGN, 0, /* alignment, bounds */
lowaddr, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
BUS_SPACE_MAXSIZE, /* maxsize */
BUS_SPACE_UNRESTRICTED, /* nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sc_parent_dtag);
if (error) {
device_printf(sc->sc_dev, "can't create parent DMA tag\n");
return error;
}
#define TXRX_CTRL(idx) (BWI_TXRX_CTRL_BASE + (idx) * txrx_ctrl_step)
/*
* Create TX ring DMA stuffs
*/
error = bus_dma_tag_create(sc->sc_parent_dtag,
BWI_RING_ALIGN, 0,
BUS_SPACE_MAXADDR,
BUS_SPACE_MAXADDR,
NULL, NULL,
tx_ring_sz,
1,
tx_ring_sz,
0,
NULL, NULL,
&sc->sc_txring_dtag);
if (error) {
device_printf(sc->sc_dev, "can't create TX ring DMA tag\n");
return error;
}
for (i = 0; i < BWI_TX_NRING; ++i) {
error = bwi_dma_ring_alloc(sc, sc->sc_txring_dtag,
&sc->sc_tx_rdata[i], tx_ring_sz,
TXRX_CTRL(i));
if (error) {
device_printf(sc->sc_dev, "%dth TX ring "
"DMA alloc failed\n", i);
return error;
}
}
/*
* Create RX ring DMA stuffs
*/
error = bus_dma_tag_create(sc->sc_parent_dtag,
BWI_RING_ALIGN, 0,
BUS_SPACE_MAXADDR,
BUS_SPACE_MAXADDR,
NULL, NULL,
rx_ring_sz,
1,
rx_ring_sz,
0,
NULL, NULL,
&sc->sc_rxring_dtag);
if (error) {
device_printf(sc->sc_dev, "can't create RX ring DMA tag\n");
return error;
}
error = bwi_dma_ring_alloc(sc, sc->sc_rxring_dtag, &sc->sc_rx_rdata,
rx_ring_sz, TXRX_CTRL(0));
if (error) {
device_printf(sc->sc_dev, "RX ring DMA alloc failed\n");
return error;
}
if (has_txstats) {
error = bwi_dma_txstats_alloc(sc, TXRX_CTRL(3), desc_sz);
if (error) {
device_printf(sc->sc_dev,
"TX stats DMA alloc failed\n");
return error;
}
}
#undef TXRX_CTRL
return bwi_dma_mbuf_create(sc);
}
static void
bwi_dma_free(struct bwi_softc *sc)
{
if (sc->sc_txring_dtag != NULL) {
int i;
for (i = 0; i < BWI_TX_NRING; ++i) {
struct bwi_ring_data *rd = &sc->sc_tx_rdata[i];
if (rd->rdata_desc != NULL) {
bus_dmamap_unload(sc->sc_txring_dtag,
rd->rdata_dmap);
bus_dmamem_free(sc->sc_txring_dtag,
rd->rdata_desc,
rd->rdata_dmap);
}
}
bus_dma_tag_destroy(sc->sc_txring_dtag);
}
if (sc->sc_rxring_dtag != NULL) {
struct bwi_ring_data *rd = &sc->sc_rx_rdata;
if (rd->rdata_desc != NULL) {
bus_dmamap_unload(sc->sc_rxring_dtag, rd->rdata_dmap);
bus_dmamem_free(sc->sc_rxring_dtag, rd->rdata_desc,
rd->rdata_dmap);
}
bus_dma_tag_destroy(sc->sc_rxring_dtag);
}
bwi_dma_txstats_free(sc);
bwi_dma_mbuf_destroy(sc, BWI_TX_NRING, 1);
if (sc->sc_parent_dtag != NULL)
bus_dma_tag_destroy(sc->sc_parent_dtag);
}
static int
bwi_dma_ring_alloc(struct bwi_softc *sc, bus_dma_tag_t dtag,
struct bwi_ring_data *rd, bus_size_t size,
uint32_t txrx_ctrl)
{
int error;
error = bus_dmamem_alloc(dtag, &rd->rdata_desc,
BUS_DMA_WAITOK | BUS_DMA_ZERO,
&rd->rdata_dmap);
if (error) {
device_printf(sc->sc_dev, "can't allocate DMA mem\n");
return error;
}
error = bus_dmamap_load(dtag, rd->rdata_dmap, rd->rdata_desc, size,
bwi_dma_ring_addr, &rd->rdata_paddr,
BUS_DMA_NOWAIT);
if (error) {
device_printf(sc->sc_dev, "can't load DMA mem\n");
bus_dmamem_free(dtag, rd->rdata_desc, rd->rdata_dmap);
rd->rdata_desc = NULL;
return error;
}
rd->rdata_txrx_ctrl = txrx_ctrl;
return 0;
}
static int
bwi_dma_txstats_alloc(struct bwi_softc *sc, uint32_t ctrl_base,
bus_size_t desc_sz)
{
struct bwi_txstats_data *st;
bus_size_t dma_size;
int error;
st = malloc(sizeof(*st), M_DEVBUF, M_NOWAIT | M_ZERO);
if (st == NULL) {
device_printf(sc->sc_dev, "can't allocate txstats data\n");
return ENOMEM;
}
sc->sc_txstats = st;
/*
* Create TX stats descriptor DMA stuffs
*/
dma_size = roundup(desc_sz * BWI_TXSTATS_NDESC, BWI_RING_ALIGN);
error = bus_dma_tag_create(sc->sc_parent_dtag,
BWI_RING_ALIGN,
0,
BUS_SPACE_MAXADDR,
BUS_SPACE_MAXADDR,
NULL, NULL,
dma_size,
1,
dma_size,
0,
NULL, NULL,
&st->stats_ring_dtag);
if (error) {
device_printf(sc->sc_dev, "can't create txstats ring "
"DMA tag\n");
return error;
}
error = bus_dmamem_alloc(st->stats_ring_dtag, &st->stats_ring,
BUS_DMA_WAITOK | BUS_DMA_ZERO,
&st->stats_ring_dmap);
if (error) {
device_printf(sc->sc_dev, "can't allocate txstats ring "
"DMA mem\n");
bus_dma_tag_destroy(st->stats_ring_dtag);
st->stats_ring_dtag = NULL;
return error;
}
error = bus_dmamap_load(st->stats_ring_dtag, st->stats_ring_dmap,
st->stats_ring, dma_size,
bwi_dma_ring_addr, &st->stats_ring_paddr,
BUS_DMA_NOWAIT);
if (error) {
device_printf(sc->sc_dev, "can't load txstats ring DMA mem\n");
bus_dmamem_free(st->stats_ring_dtag, st->stats_ring,
st->stats_ring_dmap);
bus_dma_tag_destroy(st->stats_ring_dtag);
st->stats_ring_dtag = NULL;
return error;
}
/*
* Create TX stats DMA stuffs
*/
dma_size = roundup(sizeof(struct bwi_txstats) * BWI_TXSTATS_NDESC,
BWI_ALIGN);
error = bus_dma_tag_create(sc->sc_parent_dtag,
BWI_ALIGN,
0,
BUS_SPACE_MAXADDR,
BUS_SPACE_MAXADDR,
NULL, NULL,
dma_size,
1,
dma_size,
0,
NULL, NULL,
&st->stats_dtag);
if (error) {
device_printf(sc->sc_dev, "can't create txstats DMA tag\n");
return error;
}
error = bus_dmamem_alloc(st->stats_dtag, (void **)&st->stats,
BUS_DMA_WAITOK | BUS_DMA_ZERO,
&st->stats_dmap);
if (error) {
device_printf(sc->sc_dev, "can't allocate txstats DMA mem\n");
bus_dma_tag_destroy(st->stats_dtag);
st->stats_dtag = NULL;
return error;
}
error = bus_dmamap_load(st->stats_dtag, st->stats_dmap, st->stats,
dma_size, bwi_dma_ring_addr, &st->stats_paddr,
BUS_DMA_NOWAIT);
if (error) {
device_printf(sc->sc_dev, "can't load txstats DMA mem\n");
bus_dmamem_free(st->stats_dtag, st->stats, st->stats_dmap);
bus_dma_tag_destroy(st->stats_dtag);
st->stats_dtag = NULL;
return error;
}
st->stats_ctrl_base = ctrl_base;
return 0;
}
static void
bwi_dma_txstats_free(struct bwi_softc *sc)
{
struct bwi_txstats_data *st;
if (sc->sc_txstats == NULL)
return;
st = sc->sc_txstats;
if (st->stats_ring_dtag != NULL) {
bus_dmamap_unload(st->stats_ring_dtag, st->stats_ring_dmap);
bus_dmamem_free(st->stats_ring_dtag, st->stats_ring,
st->stats_ring_dmap);
bus_dma_tag_destroy(st->stats_ring_dtag);
}
if (st->stats_dtag != NULL) {
bus_dmamap_unload(st->stats_dtag, st->stats_dmap);
bus_dmamem_free(st->stats_dtag, st->stats, st->stats_dmap);
bus_dma_tag_destroy(st->stats_dtag);
}
free(st, M_DEVBUF);
}
static void
bwi_dma_ring_addr(void *arg, bus_dma_segment_t *seg, int nseg, int error)
{
KASSERT(nseg == 1, ("too many segments\n"));
*((bus_addr_t *)arg) = seg->ds_addr;
}
static int
bwi_dma_mbuf_create(struct bwi_softc *sc)
{
struct bwi_rxbuf_data *rbd = &sc->sc_rx_bdata;
int i, j, k, ntx, error;
/*
* Create TX/RX mbuf DMA tag
*/
error = bus_dma_tag_create(sc->sc_parent_dtag,
1,
0,
BUS_SPACE_MAXADDR,
BUS_SPACE_MAXADDR,
NULL, NULL,
MCLBYTES,
1,
MCLBYTES,
BUS_DMA_ALLOCNOW,
NULL, NULL,
&sc->sc_buf_dtag);
if (error) {
device_printf(sc->sc_dev, "can't create mbuf DMA tag\n");
return error;
}
ntx = 0;
/*
* Create TX mbuf DMA map
*/
for (i = 0; i < BWI_TX_NRING; ++i) {
struct bwi_txbuf_data *tbd = &sc->sc_tx_bdata[i];
for (j = 0; j < BWI_TX_NDESC; ++j) {
error = bus_dmamap_create(sc->sc_buf_dtag, 0,
&tbd->tbd_buf[j].tb_dmap);
if (error) {
device_printf(sc->sc_dev, "can't create "
"%dth tbd, %dth DMA map\n", i, j);
ntx = i;
for (k = 0; k < j; ++k) {
bus_dmamap_destroy(sc->sc_buf_dtag,
tbd->tbd_buf[k].tb_dmap);
}
goto fail;
}
}
}
ntx = BWI_TX_NRING;
/*
* Create RX mbuf DMA map and a spare DMA map
*/
error = bus_dmamap_create(sc->sc_buf_dtag, 0,
&rbd->rbd_tmp_dmap);
if (error) {
device_printf(sc->sc_dev,
"can't create spare RX buf DMA map\n");
goto fail;
}
for (j = 0; j < BWI_RX_NDESC; ++j) {
error = bus_dmamap_create(sc->sc_buf_dtag, 0,
&rbd->rbd_buf[j].rb_dmap);
if (error) {
device_printf(sc->sc_dev, "can't create %dth "
"RX buf DMA map\n", j);
for (k = 0; k < j; ++k) {
bus_dmamap_destroy(sc->sc_buf_dtag,
rbd->rbd_buf[j].rb_dmap);
}
bus_dmamap_destroy(sc->sc_buf_dtag,
rbd->rbd_tmp_dmap);
goto fail;
}
}
return 0;
fail:
bwi_dma_mbuf_destroy(sc, ntx, 0);
return error;
}
static void
bwi_dma_mbuf_destroy(struct bwi_softc *sc, int ntx, int nrx)
{
int i, j;
if (sc->sc_buf_dtag == NULL)
return;
for (i = 0; i < ntx; ++i) {
struct bwi_txbuf_data *tbd = &sc->sc_tx_bdata[i];
for (j = 0; j < BWI_TX_NDESC; ++j) {
struct bwi_txbuf *tb = &tbd->tbd_buf[j];
if (tb->tb_mbuf != NULL) {
bus_dmamap_unload(sc->sc_buf_dtag,
tb->tb_dmap);
m_freem(tb->tb_mbuf);
}
if (tb->tb_ni != NULL)
ieee80211_free_node(tb->tb_ni);
bus_dmamap_destroy(sc->sc_buf_dtag, tb->tb_dmap);
}
}
if (nrx) {
struct bwi_rxbuf_data *rbd = &sc->sc_rx_bdata;
bus_dmamap_destroy(sc->sc_buf_dtag, rbd->rbd_tmp_dmap);
for (j = 0; j < BWI_RX_NDESC; ++j) {
struct bwi_rxbuf *rb = &rbd->rbd_buf[j];
if (rb->rb_mbuf != NULL) {
bus_dmamap_unload(sc->sc_buf_dtag,
rb->rb_dmap);
m_freem(rb->rb_mbuf);
}
bus_dmamap_destroy(sc->sc_buf_dtag, rb->rb_dmap);
}
}
bus_dma_tag_destroy(sc->sc_buf_dtag);
sc->sc_buf_dtag = NULL;
}
static void
bwi_enable_intrs(struct bwi_softc *sc, uint32_t enable_intrs)
{
CSR_SETBITS_4(sc, BWI_MAC_INTR_MASK, enable_intrs);
}
static void
bwi_disable_intrs(struct bwi_softc *sc, uint32_t disable_intrs)
{
CSR_CLRBITS_4(sc, BWI_MAC_INTR_MASK, disable_intrs);
}
static int
bwi_init_tx_ring32(struct bwi_softc *sc, int ring_idx)
{
struct bwi_ring_data *rd;
struct bwi_txbuf_data *tbd;
uint32_t val, addr_hi, addr_lo;
KASSERT(ring_idx < BWI_TX_NRING, ("ring_idx %d", ring_idx));
rd = &sc->sc_tx_rdata[ring_idx];
tbd = &sc->sc_tx_bdata[ring_idx];
tbd->tbd_idx = 0;
tbd->tbd_used = 0;
bzero(rd->rdata_desc, sizeof(struct bwi_desc32) * BWI_TX_NDESC);
bus_dmamap_sync(sc->sc_txring_dtag, rd->rdata_dmap,
BUS_DMASYNC_PREWRITE);
addr_lo = __SHIFTOUT(rd->rdata_paddr, BWI_TXRX32_RINGINFO_ADDR_MASK);
addr_hi = __SHIFTOUT(rd->rdata_paddr, BWI_TXRX32_RINGINFO_FUNC_MASK);
val = __SHIFTIN(addr_lo, BWI_TXRX32_RINGINFO_ADDR_MASK) |
__SHIFTIN(BWI_TXRX32_RINGINFO_FUNC_TXRX,
BWI_TXRX32_RINGINFO_FUNC_MASK);
CSR_WRITE_4(sc, rd->rdata_txrx_ctrl + BWI_TX32_RINGINFO, val);
val = __SHIFTIN(addr_hi, BWI_TXRX32_CTRL_ADDRHI_MASK) |
BWI_TXRX32_CTRL_ENABLE;
CSR_WRITE_4(sc, rd->rdata_txrx_ctrl + BWI_TX32_CTRL, val);
return 0;
}
static void
bwi_init_rxdesc_ring32(struct bwi_softc *sc, uint32_t ctrl_base,
bus_addr_t paddr, int hdr_size, int ndesc)
{
uint32_t val, addr_hi, addr_lo;
addr_lo = __SHIFTOUT(paddr, BWI_TXRX32_RINGINFO_ADDR_MASK);
addr_hi = __SHIFTOUT(paddr, BWI_TXRX32_RINGINFO_FUNC_MASK);
val = __SHIFTIN(addr_lo, BWI_TXRX32_RINGINFO_ADDR_MASK) |
__SHIFTIN(BWI_TXRX32_RINGINFO_FUNC_TXRX,
BWI_TXRX32_RINGINFO_FUNC_MASK);
CSR_WRITE_4(sc, ctrl_base + BWI_RX32_RINGINFO, val);
val = __SHIFTIN(hdr_size, BWI_RX32_CTRL_HDRSZ_MASK) |
__SHIFTIN(addr_hi, BWI_TXRX32_CTRL_ADDRHI_MASK) |
BWI_TXRX32_CTRL_ENABLE;
CSR_WRITE_4(sc, ctrl_base + BWI_RX32_CTRL, val);
CSR_WRITE_4(sc, ctrl_base + BWI_RX32_INDEX,
(ndesc - 1) * sizeof(struct bwi_desc32));
}
static int
bwi_init_rx_ring32(struct bwi_softc *sc)
{
struct bwi_ring_data *rd = &sc->sc_rx_rdata;
int i, error;
sc->sc_rx_bdata.rbd_idx = 0;
for (i = 0; i < BWI_RX_NDESC; ++i) {
error = bwi_newbuf(sc, i, 1);
if (error) {
device_printf(sc->sc_dev,
"can't allocate %dth RX buffer\n", i);
return error;
}
}
bus_dmamap_sync(sc->sc_rxring_dtag, rd->rdata_dmap,
BUS_DMASYNC_PREWRITE);
bwi_init_rxdesc_ring32(sc, rd->rdata_txrx_ctrl, rd->rdata_paddr,
sizeof(struct bwi_rxbuf_hdr), BWI_RX_NDESC);
return 0;
}
static int
bwi_init_txstats32(struct bwi_softc *sc)
{
struct bwi_txstats_data *st = sc->sc_txstats;
bus_addr_t stats_paddr;
int i;
bzero(st->stats, BWI_TXSTATS_NDESC * sizeof(struct bwi_txstats));
bus_dmamap_sync(st->stats_dtag, st->stats_dmap, BUS_DMASYNC_PREWRITE);
st->stats_idx = 0;
stats_paddr = st->stats_paddr;
for (i = 0; i < BWI_TXSTATS_NDESC; ++i) {
bwi_setup_desc32(sc, st->stats_ring, BWI_TXSTATS_NDESC, i,
stats_paddr, sizeof(struct bwi_txstats), 0);
stats_paddr += sizeof(struct bwi_txstats);
}
bus_dmamap_sync(st->stats_ring_dtag, st->stats_ring_dmap,
BUS_DMASYNC_PREWRITE);
bwi_init_rxdesc_ring32(sc, st->stats_ctrl_base,
st->stats_ring_paddr, 0, BWI_TXSTATS_NDESC);
return 0;
}
static void
bwi_setup_rx_desc32(struct bwi_softc *sc, int buf_idx, bus_addr_t paddr,
int buf_len)
{
struct bwi_ring_data *rd = &sc->sc_rx_rdata;
KASSERT(buf_idx < BWI_RX_NDESC, ("buf_idx %d", buf_idx));
bwi_setup_desc32(sc, rd->rdata_desc, BWI_RX_NDESC, buf_idx,
paddr, buf_len, 0);
}
static void
bwi_setup_tx_desc32(struct bwi_softc *sc, struct bwi_ring_data *rd,
int buf_idx, bus_addr_t paddr, int buf_len)
{
KASSERT(buf_idx < BWI_TX_NDESC, ("buf_idx %d", buf_idx));
bwi_setup_desc32(sc, rd->rdata_desc, BWI_TX_NDESC, buf_idx,
paddr, buf_len, 1);
}
static int
bwi_init_tx_ring64(struct bwi_softc *sc, int ring_idx)
{
/* TODO:64 */
return EOPNOTSUPP;
}
static int
bwi_init_rx_ring64(struct bwi_softc *sc)
{
/* TODO:64 */
return EOPNOTSUPP;
}
static int
bwi_init_txstats64(struct bwi_softc *sc)
{
/* TODO:64 */
return EOPNOTSUPP;
}
static void
bwi_setup_rx_desc64(struct bwi_softc *sc, int buf_idx, bus_addr_t paddr,
int buf_len)
{
/* TODO:64 */
}
static void
bwi_setup_tx_desc64(struct bwi_softc *sc, struct bwi_ring_data *rd,
int buf_idx, bus_addr_t paddr, int buf_len)
{
/* TODO:64 */
}
static void
bwi_dma_buf_addr(void *arg, bus_dma_segment_t *seg, int nseg,
bus_size_t mapsz __unused, int error)
{
if (!error) {
KASSERT(nseg == 1, ("too many segments(%d)\n", nseg));
*((bus_addr_t *)arg) = seg->ds_addr;
}
}
static int
bwi_newbuf(struct bwi_softc *sc, int buf_idx, int init)
{
struct bwi_rxbuf_data *rbd = &sc->sc_rx_bdata;
struct bwi_rxbuf *rxbuf = &rbd->rbd_buf[buf_idx];
struct bwi_rxbuf_hdr *hdr;
bus_dmamap_t map;
bus_addr_t paddr;
struct mbuf *m;
int error;
KASSERT(buf_idx < BWI_RX_NDESC, ("buf_idx %d", buf_idx));
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
if (m == NULL) {
error = ENOBUFS;
/*
* If the NIC is up and running, we need to:
* - Clear RX buffer's header.
* - Restore RX descriptor settings.
*/
if (init)
return error;
else
goto back;
}
m->m_len = m->m_pkthdr.len = MCLBYTES;
/*
* Try to load RX buf into temporary DMA map
*/
error = bus_dmamap_load_mbuf(sc->sc_buf_dtag, rbd->rbd_tmp_dmap, m,
bwi_dma_buf_addr, &paddr, BUS_DMA_NOWAIT);
if (error) {
m_freem(m);
/*
* See the comment above
*/
if (init)
return error;
else
goto back;
}
if (!init)
bus_dmamap_unload(sc->sc_buf_dtag, rxbuf->rb_dmap);
rxbuf->rb_mbuf = m;
rxbuf->rb_paddr = paddr;
/*
* Swap RX buf's DMA map with the loaded temporary one
*/
map = rxbuf->rb_dmap;
rxbuf->rb_dmap = rbd->rbd_tmp_dmap;
rbd->rbd_tmp_dmap = map;
back:
/*
* Clear RX buf header
*/
hdr = mtod(rxbuf->rb_mbuf, struct bwi_rxbuf_hdr *);
bzero(hdr, sizeof(*hdr));
bus_dmamap_sync(sc->sc_buf_dtag, rxbuf->rb_dmap, BUS_DMASYNC_PREWRITE);
/*
* Setup RX buf descriptor
*/
sc->sc_setup_rxdesc(sc, buf_idx, rxbuf->rb_paddr,
rxbuf->rb_mbuf->m_len - sizeof(*hdr));
return error;
}
static void
bwi_set_addr_filter(struct bwi_softc *sc, uint16_t addr_ofs,
const uint8_t *addr)
{
int i;
CSR_WRITE_2(sc, BWI_ADDR_FILTER_CTRL,
BWI_ADDR_FILTER_CTRL_SET | addr_ofs);
for (i = 0; i < (IEEE80211_ADDR_LEN / 2); ++i) {
uint16_t addr_val;
addr_val = (uint16_t)addr[i * 2] |
(((uint16_t)addr[(i * 2) + 1]) << 8);
CSR_WRITE_2(sc, BWI_ADDR_FILTER_DATA, addr_val);
}
}
static int
bwi_rxeof(struct bwi_softc *sc, int end_idx)
{
struct bwi_ring_data *rd = &sc->sc_rx_rdata;
struct bwi_rxbuf_data *rbd = &sc->sc_rx_bdata;
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;
int idx, rx_data = 0;
idx = rbd->rbd_idx;
while (idx != end_idx) {
struct bwi_rxbuf *rb = &rbd->rbd_buf[idx];
struct bwi_rxbuf_hdr *hdr;
struct ieee80211_frame_min *wh;
struct ieee80211_node *ni;
struct mbuf *m;
uint32_t plcp;
uint16_t flags2;
int buflen, wh_ofs, hdr_extra, rssi, noise, type, rate;
m = rb->rb_mbuf;
bus_dmamap_sync(sc->sc_buf_dtag, rb->rb_dmap,
BUS_DMASYNC_POSTREAD);
if (bwi_newbuf(sc, idx, 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
counter_u64_add(ic->ic_ierrors, 1);
goto next;
}
hdr = mtod(m, struct bwi_rxbuf_hdr *);
flags2 = le16toh(hdr->rxh_flags2);
hdr_extra = 0;
if (flags2 & BWI_RXH_F2_TYPE2FRAME)
hdr_extra = 2;
wh_ofs = hdr_extra + 6; /* XXX magic number */
buflen = le16toh(hdr->rxh_buflen);
if (buflen < BWI_FRAME_MIN_LEN(wh_ofs)) {
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
device_printf(sc->sc_dev,
"%s: zero length data, hdr_extra %d\n",
__func__, hdr_extra);
counter_u64_add(ic->ic_ierrors, 1);
m_freem(m);
goto next;
}
bcopy((uint8_t *)(hdr + 1) + hdr_extra, &plcp, sizeof(plcp));
rssi = bwi_calc_rssi(sc, hdr);
noise = bwi_calc_noise(sc);
m->m_len = m->m_pkthdr.len = buflen + sizeof(*hdr);
m_adj(m, sizeof(*hdr) + wh_ofs);
if (htole16(hdr->rxh_flags1) & BWI_RXH_F1_OFDM)
rate = bwi_plcp2rate(plcp, IEEE80211_T_OFDM);
else
rate = bwi_plcp2rate(plcp, IEEE80211_T_CCK);
/* RX radio tap */
if (ieee80211_radiotap_active(ic))
bwi_rx_radiotap(sc, m, hdr, &plcp, rate, rssi, noise);
m_adj(m, -IEEE80211_CRC_LEN);
BWI_UNLOCK(sc);
wh = mtod(m, struct ieee80211_frame_min *);
ni = ieee80211_find_rxnode(ic, wh);
if (ni != NULL) {
type = ieee80211_input(ni, m, rssi - noise, noise);
ieee80211_free_node(ni);
} else
type = ieee80211_input_all(ic, m, rssi - noise, noise);
if (type == IEEE80211_FC0_TYPE_DATA) {
rx_data = 1;
sc->sc_rx_rate = rate;
}
BWI_LOCK(sc);
next:
idx = (idx + 1) % BWI_RX_NDESC;
if (sc->sc_flags & BWI_F_STOP) {
/*
* Take the fast lane, don't do
* any damage to softc
*/
return -1;
}
}
rbd->rbd_idx = idx;
bus_dmamap_sync(sc->sc_rxring_dtag, rd->rdata_dmap,
BUS_DMASYNC_PREWRITE);
return rx_data;
}
static int
bwi_rxeof32(struct bwi_softc *sc)
{
uint32_t val, rx_ctrl;
int end_idx, rx_data;
rx_ctrl = sc->sc_rx_rdata.rdata_txrx_ctrl;
val = CSR_READ_4(sc, rx_ctrl + BWI_RX32_STATUS);
end_idx = __SHIFTOUT(val, BWI_RX32_STATUS_INDEX_MASK) /
sizeof(struct bwi_desc32);
rx_data = bwi_rxeof(sc, end_idx);
if (rx_data >= 0) {
CSR_WRITE_4(sc, rx_ctrl + BWI_RX32_INDEX,
end_idx * sizeof(struct bwi_desc32));
}
return rx_data;
}
static int
bwi_rxeof64(struct bwi_softc *sc)
{
/* TODO:64 */
return 0;
}
static void
bwi_reset_rx_ring32(struct bwi_softc *sc, uint32_t rx_ctrl)
{
int i;
CSR_WRITE_4(sc, rx_ctrl + BWI_RX32_CTRL, 0);
#define NRETRY 10
for (i = 0; i < NRETRY; ++i) {
uint32_t status;
status = CSR_READ_4(sc, rx_ctrl + BWI_RX32_STATUS);
if (__SHIFTOUT(status, BWI_RX32_STATUS_STATE_MASK) ==
BWI_RX32_STATUS_STATE_DISABLED)
break;
DELAY(1000);
}
if (i == NRETRY)
device_printf(sc->sc_dev, "reset rx ring timedout\n");
#undef NRETRY
CSR_WRITE_4(sc, rx_ctrl + BWI_RX32_RINGINFO, 0);
}
static void
bwi_free_txstats32(struct bwi_softc *sc)
{
bwi_reset_rx_ring32(sc, sc->sc_txstats->stats_ctrl_base);
}
static void
bwi_free_rx_ring32(struct bwi_softc *sc)
{
struct bwi_ring_data *rd = &sc->sc_rx_rdata;
struct bwi_rxbuf_data *rbd = &sc->sc_rx_bdata;
int i;
bwi_reset_rx_ring32(sc, rd->rdata_txrx_ctrl);
for (i = 0; i < BWI_RX_NDESC; ++i) {
struct bwi_rxbuf *rb = &rbd->rbd_buf[i];
if (rb->rb_mbuf != NULL) {
bus_dmamap_unload(sc->sc_buf_dtag, rb->rb_dmap);
m_freem(rb->rb_mbuf);
rb->rb_mbuf = NULL;
}
}
}
static void
bwi_free_tx_ring32(struct bwi_softc *sc, int ring_idx)
{
struct bwi_ring_data *rd;
struct bwi_txbuf_data *tbd;
uint32_t state, val;
int i;
KASSERT(ring_idx < BWI_TX_NRING, ("ring_idx %d", ring_idx));
rd = &sc->sc_tx_rdata[ring_idx];
tbd = &sc->sc_tx_bdata[ring_idx];
#define NRETRY 10
for (i = 0; i < NRETRY; ++i) {
val = CSR_READ_4(sc, rd->rdata_txrx_ctrl + BWI_TX32_STATUS);
state = __SHIFTOUT(val, BWI_TX32_STATUS_STATE_MASK);
if (state == BWI_TX32_STATUS_STATE_DISABLED ||
state == BWI_TX32_STATUS_STATE_IDLE ||
state == BWI_TX32_STATUS_STATE_STOPPED)
break;
DELAY(1000);
}
if (i == NRETRY) {
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
device_printf(sc->sc_dev,
"%s: wait for TX ring(%d) stable timed out\n",
__func__, ring_idx);
}
CSR_WRITE_4(sc, rd->rdata_txrx_ctrl + BWI_TX32_CTRL, 0);
for (i = 0; i < NRETRY; ++i) {
val = CSR_READ_4(sc, rd->rdata_txrx_ctrl + BWI_TX32_STATUS);
state = __SHIFTOUT(val, BWI_TX32_STATUS_STATE_MASK);
if (state == BWI_TX32_STATUS_STATE_DISABLED)
break;
DELAY(1000);
}
if (i == NRETRY)
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
device_printf(sc->sc_dev, "%s: reset TX ring (%d) timed out\n",
__func__, ring_idx);
#undef NRETRY
DELAY(1000);
CSR_WRITE_4(sc, rd->rdata_txrx_ctrl + BWI_TX32_RINGINFO, 0);
for (i = 0; i < BWI_TX_NDESC; ++i) {
struct bwi_txbuf *tb = &tbd->tbd_buf[i];
if (tb->tb_mbuf != NULL) {
bus_dmamap_unload(sc->sc_buf_dtag, tb->tb_dmap);
m_freem(tb->tb_mbuf);
tb->tb_mbuf = NULL;
}
if (tb->tb_ni != NULL) {
ieee80211_free_node(tb->tb_ni);
tb->tb_ni = NULL;
}
}
}
static void
bwi_free_txstats64(struct bwi_softc *sc)
{
/* TODO:64 */
}
static void
bwi_free_rx_ring64(struct bwi_softc *sc)
{
/* TODO:64 */
}
static void
bwi_free_tx_ring64(struct bwi_softc *sc, int ring_idx)
{
/* TODO:64 */
}
/* XXX does not belong here */
#define IEEE80211_OFDM_PLCP_RATE_MASK __BITS(3, 0)
#define IEEE80211_OFDM_PLCP_LEN_MASK __BITS(16, 5)
static __inline void
bwi_ofdm_plcp_header(uint32_t *plcp0, int pkt_len, uint8_t rate)
{
uint32_t plcp;
plcp = __SHIFTIN(ieee80211_rate2plcp(rate, IEEE80211_T_OFDM),
IEEE80211_OFDM_PLCP_RATE_MASK) |
__SHIFTIN(pkt_len, IEEE80211_OFDM_PLCP_LEN_MASK);
*plcp0 = htole32(plcp);
}
static __inline void
bwi_ds_plcp_header(struct ieee80211_ds_plcp_hdr *plcp, int pkt_len,
uint8_t rate)
{
int len, service, pkt_bitlen;
pkt_bitlen = pkt_len * NBBY;
len = howmany(pkt_bitlen * 2, rate);
service = IEEE80211_PLCP_SERVICE_LOCKED;
if (rate == (11 * 2)) {
int pkt_bitlen1;
/*
* PLCP service field needs to be adjusted,
* if TX rate is 11Mbytes/s
*/
pkt_bitlen1 = len * 11;
if (pkt_bitlen1 - pkt_bitlen >= NBBY)
service |= IEEE80211_PLCP_SERVICE_LENEXT7;
}
plcp->i_signal = ieee80211_rate2plcp(rate, IEEE80211_T_CCK);
plcp->i_service = service;
plcp->i_length = htole16(len);
/* NOTE: do NOT touch i_crc */
}
static __inline void
bwi_plcp_header(const struct ieee80211_rate_table *rt,
void *plcp, int pkt_len, uint8_t rate)
{
enum ieee80211_phytype modtype;
/*
* Assume caller has zeroed 'plcp'
*/
modtype = ieee80211_rate2phytype(rt, rate);
if (modtype == IEEE80211_T_OFDM)
bwi_ofdm_plcp_header(plcp, pkt_len, rate);
else if (modtype == IEEE80211_T_DS)
bwi_ds_plcp_header(plcp, pkt_len, rate);
else
panic("unsupport modulation type %u\n", modtype);
}
static int
bwi_encap(struct bwi_softc *sc, int idx, struct mbuf *m,
struct ieee80211_node *ni)
{
struct ieee80211vap *vap = ni->ni_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
struct ieee80211com *ic = &sc->sc_ic;
struct bwi_ring_data *rd = &sc->sc_tx_rdata[BWI_TX_DATA_RING];
struct bwi_txbuf_data *tbd = &sc->sc_tx_bdata[BWI_TX_DATA_RING];
struct bwi_txbuf *tb = &tbd->tbd_buf[idx];
struct bwi_mac *mac;
struct bwi_txbuf_hdr *hdr;
struct ieee80211_frame *wh;
const struct ieee80211_txparam *tp = ni->ni_txparms;
uint8_t rate, rate_fb;
uint32_t mac_ctrl;
uint16_t phy_ctrl;
bus_addr_t paddr;
int type, ismcast, pkt_len, error, rix;
#if 0
const uint8_t *p;
int i;
#endif
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
wh = mtod(m, struct ieee80211_frame *);
type = wh->i_fc[0] & IEEE80211_FC0_TYPE_MASK;
ismcast = IEEE80211_IS_MULTICAST(wh->i_addr1);
/* Get 802.11 frame len before prepending TX header */
pkt_len = m->m_pkthdr.len + IEEE80211_CRC_LEN;
/*
* Find TX rate
*/
if (type != IEEE80211_FC0_TYPE_DATA || (m->m_flags & M_EAPOL)) {
rate = rate_fb = tp->mgmtrate;
} else if (ismcast) {
rate = rate_fb = tp->mcastrate;
} else if (tp->ucastrate != IEEE80211_FIXED_RATE_NONE) {
rate = rate_fb = tp->ucastrate;
} else {
rix = ieee80211_ratectl_rate(ni, NULL, pkt_len);
rate = ni->ni_txrate;
if (rix > 0) {
rate_fb = ni->ni_rates.rs_rates[rix-1] &
IEEE80211_RATE_VAL;
} else {
rate_fb = rate;
}
}
tb->tb_rate[0] = rate;
tb->tb_rate[1] = rate_fb;
sc->sc_tx_rate = rate;
/*
* TX radio tap
*/
if (ieee80211_radiotap_active_vap(vap)) {
sc->sc_tx_th.wt_flags = 0;
if (wh->i_fc[1] & IEEE80211_FC1_PROTECTED)
sc->sc_tx_th.wt_flags |= IEEE80211_RADIOTAP_F_WEP;
if (ieee80211_rate2phytype(sc->sc_rates, rate) == IEEE80211_T_DS &&
(ic->ic_flags & IEEE80211_F_SHPREAMBLE) &&
rate != (1 * 2)) {
sc->sc_tx_th.wt_flags |= IEEE80211_RADIOTAP_F_SHORTPRE;
}
sc->sc_tx_th.wt_rate = rate;
ieee80211_radiotap_tx(vap, m);
}
/*
* Setup the embedded TX header
*/
M_PREPEND(m, sizeof(*hdr), M_NOWAIT);
if (m == 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
device_printf(sc->sc_dev, "%s: prepend TX header failed\n",
__func__);
return ENOBUFS;
}
hdr = mtod(m, struct bwi_txbuf_hdr *);
bzero(hdr, sizeof(*hdr));
bcopy(wh->i_fc, hdr->txh_fc, sizeof(hdr->txh_fc));
bcopy(wh->i_addr1, hdr->txh_addr1, sizeof(hdr->txh_addr1));
if (!ismcast) {
uint16_t dur;
dur = ieee80211_ack_duration(sc->sc_rates, rate,
ic->ic_flags & ~IEEE80211_F_SHPREAMBLE);
hdr->txh_fb_duration = htole16(dur);
}
hdr->txh_id = __SHIFTIN(BWI_TX_DATA_RING, BWI_TXH_ID_RING_MASK) |
__SHIFTIN(idx, BWI_TXH_ID_IDX_MASK);
bwi_plcp_header(sc->sc_rates, hdr->txh_plcp, pkt_len, rate);
bwi_plcp_header(sc->sc_rates, hdr->txh_fb_plcp, pkt_len, rate_fb);
phy_ctrl = __SHIFTIN(mac->mac_rf.rf_ant_mode,
BWI_TXH_PHY_C_ANTMODE_MASK);
if (ieee80211_rate2phytype(sc->sc_rates, rate) == IEEE80211_T_OFDM)
phy_ctrl |= BWI_TXH_PHY_C_OFDM;
else if ((ic->ic_flags & IEEE80211_F_SHPREAMBLE) && rate != (2 * 1))
phy_ctrl |= BWI_TXH_PHY_C_SHPREAMBLE;
mac_ctrl = BWI_TXH_MAC_C_HWSEQ | BWI_TXH_MAC_C_FIRST_FRAG;
if (!ismcast)
mac_ctrl |= BWI_TXH_MAC_C_ACK;
if (ieee80211_rate2phytype(sc->sc_rates, rate_fb) == IEEE80211_T_OFDM)
mac_ctrl |= BWI_TXH_MAC_C_FB_OFDM;
hdr->txh_mac_ctrl = htole32(mac_ctrl);
hdr->txh_phy_ctrl = htole16(phy_ctrl);
/* Catch any further usage */
hdr = NULL;
wh = NULL;
/* DMA load */
error = bus_dmamap_load_mbuf(sc->sc_buf_dtag, tb->tb_dmap, m,
bwi_dma_buf_addr, &paddr, BUS_DMA_NOWAIT);
if (error && error != EFBIG) {
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
device_printf(sc->sc_dev, "%s: can't load TX buffer (1) %d\n",
__func__, error);
goto back;
}
if (error) { /* error == EFBIG */
struct mbuf *m_new;
m_new = m_defrag(m, M_NOWAIT);
if (m_new == 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
device_printf(sc->sc_dev,
"%s: can't defrag TX buffer\n", __func__);
error = ENOBUFS;
goto back;
} else {
m = m_new;
}
error = bus_dmamap_load_mbuf(sc->sc_buf_dtag, tb->tb_dmap, m,
bwi_dma_buf_addr, &paddr,
BUS_DMA_NOWAIT);
if (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
device_printf(sc->sc_dev,
"%s: can't load TX buffer (2) %d\n",
__func__, error);
goto back;
}
}
error = 0;
bus_dmamap_sync(sc->sc_buf_dtag, tb->tb_dmap, BUS_DMASYNC_PREWRITE);
tb->tb_mbuf = m;
tb->tb_ni = ni;
#if 0
p = mtod(m, const uint8_t *);
for (i = 0; i < m->m_pkthdr.len; ++i) {
if (i != 0 && i % 8 == 0)
printf("\n");
printf("%02x ", p[i]);
}
printf("\n");
#endif
DPRINTF(sc, BWI_DBG_TX, "idx %d, pkt_len %d, buflen %d\n",
idx, pkt_len, m->m_pkthdr.len);
/* Setup TX descriptor */
sc->sc_setup_txdesc(sc, rd, idx, paddr, m->m_pkthdr.len);
bus_dmamap_sync(sc->sc_txring_dtag, rd->rdata_dmap,
BUS_DMASYNC_PREWRITE);
/* Kick start */
sc->sc_start_tx(sc, rd->rdata_txrx_ctrl, idx);
back:
if (error)
m_freem(m);
return error;
}
static int
bwi_encap_raw(struct bwi_softc *sc, int idx, struct mbuf *m,
struct ieee80211_node *ni, const struct ieee80211_bpf_params *params)
{
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211com *ic = ni->ni_ic;
struct bwi_ring_data *rd = &sc->sc_tx_rdata[BWI_TX_DATA_RING];
struct bwi_txbuf_data *tbd = &sc->sc_tx_bdata[BWI_TX_DATA_RING];
struct bwi_txbuf *tb = &tbd->tbd_buf[idx];
struct bwi_mac *mac;
struct bwi_txbuf_hdr *hdr;
struct ieee80211_frame *wh;
uint8_t rate, rate_fb;
uint32_t mac_ctrl;
uint16_t phy_ctrl;
bus_addr_t paddr;
int ismcast, pkt_len, error;
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
wh = mtod(m, struct ieee80211_frame *);
ismcast = IEEE80211_IS_MULTICAST(wh->i_addr1);
/* Get 802.11 frame len before prepending TX header */
pkt_len = m->m_pkthdr.len + IEEE80211_CRC_LEN;
/*
* Find TX rate
*/
rate = params->ibp_rate0;
if (!ieee80211_isratevalid(ic->ic_rt, rate)) {
/* XXX fall back to mcast/mgmt rate? */
2009-05-30 01:33:05 +00:00
m_freem(m);
return EINVAL;
}
if (params->ibp_try1 != 0) {
rate_fb = params->ibp_rate1;
if (!ieee80211_isratevalid(ic->ic_rt, rate_fb)) {
/* XXX fall back to rate0? */
2009-05-30 01:33:05 +00:00
m_freem(m);
return EINVAL;
}
} else
rate_fb = rate;
tb->tb_rate[0] = rate;
tb->tb_rate[1] = rate_fb;
sc->sc_tx_rate = rate;
/*
* TX radio tap
*/
if (ieee80211_radiotap_active_vap(vap)) {
sc->sc_tx_th.wt_flags = 0;
/* XXX IEEE80211_BPF_CRYPTO */
if (wh->i_fc[1] & IEEE80211_FC1_PROTECTED)
sc->sc_tx_th.wt_flags |= IEEE80211_RADIOTAP_F_WEP;
if (params->ibp_flags & IEEE80211_BPF_SHORTPRE)
sc->sc_tx_th.wt_flags |= IEEE80211_RADIOTAP_F_SHORTPRE;
sc->sc_tx_th.wt_rate = rate;
ieee80211_radiotap_tx(vap, m);
}
/*
* Setup the embedded TX header
*/
M_PREPEND(m, sizeof(*hdr), M_NOWAIT);
if (m == 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
device_printf(sc->sc_dev, "%s: prepend TX header failed\n",
__func__);
return ENOBUFS;
}
hdr = mtod(m, struct bwi_txbuf_hdr *);
bzero(hdr, sizeof(*hdr));
bcopy(wh->i_fc, hdr->txh_fc, sizeof(hdr->txh_fc));
bcopy(wh->i_addr1, hdr->txh_addr1, sizeof(hdr->txh_addr1));
mac_ctrl = BWI_TXH_MAC_C_HWSEQ | BWI_TXH_MAC_C_FIRST_FRAG;
if (!ismcast && (params->ibp_flags & IEEE80211_BPF_NOACK) == 0) {
uint16_t dur;
dur = ieee80211_ack_duration(sc->sc_rates, rate_fb, 0);
hdr->txh_fb_duration = htole16(dur);
mac_ctrl |= BWI_TXH_MAC_C_ACK;
}
hdr->txh_id = __SHIFTIN(BWI_TX_DATA_RING, BWI_TXH_ID_RING_MASK) |
__SHIFTIN(idx, BWI_TXH_ID_IDX_MASK);
bwi_plcp_header(sc->sc_rates, hdr->txh_plcp, pkt_len, rate);
bwi_plcp_header(sc->sc_rates, hdr->txh_fb_plcp, pkt_len, rate_fb);
phy_ctrl = __SHIFTIN(mac->mac_rf.rf_ant_mode,
BWI_TXH_PHY_C_ANTMODE_MASK);
if (ieee80211_rate2phytype(sc->sc_rates, rate) == IEEE80211_T_OFDM) {
phy_ctrl |= BWI_TXH_PHY_C_OFDM;
mac_ctrl |= BWI_TXH_MAC_C_FB_OFDM;
} else if (params->ibp_flags & IEEE80211_BPF_SHORTPRE)
phy_ctrl |= BWI_TXH_PHY_C_SHPREAMBLE;
hdr->txh_mac_ctrl = htole32(mac_ctrl);
hdr->txh_phy_ctrl = htole16(phy_ctrl);
/* Catch any further usage */
hdr = NULL;
wh = NULL;
/* DMA load */
error = bus_dmamap_load_mbuf(sc->sc_buf_dtag, tb->tb_dmap, m,
bwi_dma_buf_addr, &paddr, BUS_DMA_NOWAIT);
if (error != 0) {
struct mbuf *m_new;
if (error != EFBIG) {
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
device_printf(sc->sc_dev,
"%s: can't load TX buffer (1) %d\n",
__func__, error);
goto back;
}
m_new = m_defrag(m, M_NOWAIT);
if (m_new == 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
device_printf(sc->sc_dev,
"%s: can't defrag TX buffer\n", __func__);
error = ENOBUFS;
goto back;
}
m = m_new;
error = bus_dmamap_load_mbuf(sc->sc_buf_dtag, tb->tb_dmap, m,
bwi_dma_buf_addr, &paddr,
BUS_DMA_NOWAIT);
if (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
device_printf(sc->sc_dev,
"%s: can't load TX buffer (2) %d\n",
__func__, error);
goto back;
}
}
bus_dmamap_sync(sc->sc_buf_dtag, tb->tb_dmap, BUS_DMASYNC_PREWRITE);
tb->tb_mbuf = m;
tb->tb_ni = ni;
DPRINTF(sc, BWI_DBG_TX, "idx %d, pkt_len %d, buflen %d\n",
idx, pkt_len, m->m_pkthdr.len);
/* Setup TX descriptor */
sc->sc_setup_txdesc(sc, rd, idx, paddr, m->m_pkthdr.len);
bus_dmamap_sync(sc->sc_txring_dtag, rd->rdata_dmap,
BUS_DMASYNC_PREWRITE);
/* Kick start */
sc->sc_start_tx(sc, rd->rdata_txrx_ctrl, idx);
back:
if (error)
m_freem(m);
return error;
}
static void
bwi_start_tx32(struct bwi_softc *sc, uint32_t tx_ctrl, int idx)
{
idx = (idx + 1) % BWI_TX_NDESC;
CSR_WRITE_4(sc, tx_ctrl + BWI_TX32_INDEX,
idx * sizeof(struct bwi_desc32));
}
static void
bwi_start_tx64(struct bwi_softc *sc, uint32_t tx_ctrl, int idx)
{
/* TODO:64 */
}
static void
bwi_txeof_status32(struct bwi_softc *sc)
{
uint32_t val, ctrl_base;
int end_idx;
ctrl_base = sc->sc_txstats->stats_ctrl_base;
val = CSR_READ_4(sc, ctrl_base + BWI_RX32_STATUS);
end_idx = __SHIFTOUT(val, BWI_RX32_STATUS_INDEX_MASK) /
sizeof(struct bwi_desc32);
bwi_txeof_status(sc, end_idx);
CSR_WRITE_4(sc, ctrl_base + BWI_RX32_INDEX,
end_idx * sizeof(struct bwi_desc32));
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
bwi_start_locked(sc);
}
static void
bwi_txeof_status64(struct bwi_softc *sc)
{
/* TODO:64 */
}
static void
_bwi_txeof(struct bwi_softc *sc, uint16_t tx_id, int acked, int data_txcnt)
{
struct bwi_txbuf_data *tbd;
struct bwi_txbuf *tb;
int ring_idx, buf_idx;
struct ieee80211_node *ni;
if (tx_id == 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
device_printf(sc->sc_dev, "%s: zero tx id\n", __func__);
return;
}
ring_idx = __SHIFTOUT(tx_id, BWI_TXH_ID_RING_MASK);
buf_idx = __SHIFTOUT(tx_id, BWI_TXH_ID_IDX_MASK);
KASSERT(ring_idx == BWI_TX_DATA_RING, ("ring_idx %d", ring_idx));
KASSERT(buf_idx < BWI_TX_NDESC, ("buf_idx %d", buf_idx));
tbd = &sc->sc_tx_bdata[ring_idx];
KASSERT(tbd->tbd_used > 0, ("tbd_used %d", tbd->tbd_used));
tbd->tbd_used--;
tb = &tbd->tbd_buf[buf_idx];
DPRINTF(sc, BWI_DBG_TXEOF, "txeof idx %d, "
"acked %d, data_txcnt %d, ni %p\n",
buf_idx, acked, data_txcnt, tb->tb_ni);
bus_dmamap_unload(sc->sc_buf_dtag, tb->tb_dmap);
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 ((ni = tb->tb_ni) != NULL) {
const struct bwi_txbuf_hdr *hdr =
mtod(tb->tb_mbuf, const struct bwi_txbuf_hdr *);
struct ieee80211_ratectl_tx_status txs;
/* NB: update rate control only for unicast frames */
if (hdr->txh_mac_ctrl & htole32(BWI_TXH_MAC_C_ACK)) {
/*
* Feed back 'acked and data_txcnt'. Note that the
* generic AMRR code only understands one tx rate
* and the estimator doesn't handle real retry counts
* well so to avoid over-aggressive downshifting we
* treat any number of retries as "1".
*/
txs.flags = IEEE80211_RATECTL_STATUS_LONG_RETRY;
txs.long_retries = acked;
if (data_txcnt > 1)
txs.status = IEEE80211_RATECTL_TX_SUCCESS;
else {
txs.status =
IEEE80211_RATECTL_TX_FAIL_UNSPECIFIED;
}
ieee80211_ratectl_tx_complete(ni, &txs);
}
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_tx_complete(ni, tb->tb_mbuf, !acked);
tb->tb_ni = 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
m_freem(tb->tb_mbuf);
tb->tb_mbuf = NULL;
if (tbd->tbd_used == 0)
sc->sc_tx_timer = 0;
}
static void
bwi_txeof_status(struct bwi_softc *sc, int end_idx)
{
struct bwi_txstats_data *st = sc->sc_txstats;
int idx;
bus_dmamap_sync(st->stats_dtag, st->stats_dmap, BUS_DMASYNC_POSTREAD);
idx = st->stats_idx;
while (idx != end_idx) {
const struct bwi_txstats *stats = &st->stats[idx];
if ((stats->txs_flags & BWI_TXS_F_PENDING) == 0) {
int data_txcnt;
data_txcnt = __SHIFTOUT(stats->txs_txcnt,
BWI_TXS_TXCNT_DATA);
_bwi_txeof(sc, le16toh(stats->txs_id),
stats->txs_flags & BWI_TXS_F_ACKED,
data_txcnt);
}
idx = (idx + 1) % BWI_TXSTATS_NDESC;
}
st->stats_idx = idx;
}
static void
bwi_txeof(struct bwi_softc *sc)
{
for (;;) {
uint32_t tx_status0, tx_status1;
uint16_t tx_id;
int data_txcnt;
tx_status0 = CSR_READ_4(sc, BWI_TXSTATUS0);
if ((tx_status0 & BWI_TXSTATUS0_VALID) == 0)
break;
tx_status1 = CSR_READ_4(sc, BWI_TXSTATUS1);
tx_id = __SHIFTOUT(tx_status0, BWI_TXSTATUS0_TXID_MASK);
data_txcnt = __SHIFTOUT(tx_status0,
BWI_TXSTATUS0_DATA_TXCNT_MASK);
if (tx_status0 & (BWI_TXSTATUS0_AMPDU | BWI_TXSTATUS0_PENDING))
continue;
_bwi_txeof(sc, le16toh(tx_id), tx_status0 & BWI_TXSTATUS0_ACKED,
data_txcnt);
}
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
bwi_start_locked(sc);
}
static int
bwi_bbp_power_on(struct bwi_softc *sc, enum bwi_clock_mode clk_mode)
{
bwi_power_on(sc, 1);
return bwi_set_clock_mode(sc, clk_mode);
}
static void
bwi_bbp_power_off(struct bwi_softc *sc)
{
bwi_set_clock_mode(sc, BWI_CLOCK_MODE_SLOW);
bwi_power_off(sc, 1);
}
static int
bwi_get_pwron_delay(struct bwi_softc *sc)
{
struct bwi_regwin *com, *old;
struct bwi_clock_freq freq;
uint32_t val;
int error;
com = &sc->sc_com_regwin;
KASSERT(BWI_REGWIN_EXIST(com), ("no regwin"));
if ((sc->sc_cap & BWI_CAP_CLKMODE) == 0)
return 0;
error = bwi_regwin_switch(sc, com, &old);
if (error)
return error;
bwi_get_clock_freq(sc, &freq);
val = CSR_READ_4(sc, BWI_PLL_ON_DELAY);
sc->sc_pwron_delay = howmany((val + 2) * 1000000, freq.clkfreq_min);
DPRINTF(sc, BWI_DBG_ATTACH, "power on delay %u\n", sc->sc_pwron_delay);
return bwi_regwin_switch(sc, old, NULL);
}
static int
bwi_bus_attach(struct bwi_softc *sc)
{
struct bwi_regwin *bus, *old;
int error;
bus = &sc->sc_bus_regwin;
error = bwi_regwin_switch(sc, bus, &old);
if (error)
return error;
if (!bwi_regwin_is_enabled(sc, bus))
bwi_regwin_enable(sc, bus, 0);
/* Disable interripts */
CSR_WRITE_4(sc, BWI_INTRVEC, 0);
return bwi_regwin_switch(sc, old, NULL);
}
static const char *
bwi_regwin_name(const struct bwi_regwin *rw)
{
switch (rw->rw_type) {
case BWI_REGWIN_T_COM:
return "COM";
case BWI_REGWIN_T_BUSPCI:
return "PCI";
case BWI_REGWIN_T_MAC:
return "MAC";
case BWI_REGWIN_T_BUSPCIE:
return "PCIE";
}
panic("unknown regwin type 0x%04x\n", rw->rw_type);
return NULL;
}
static uint32_t
bwi_regwin_disable_bits(struct bwi_softc *sc)
{
uint32_t busrev;
/* XXX cache this */
busrev = __SHIFTOUT(CSR_READ_4(sc, BWI_ID_LO), BWI_ID_LO_BUSREV_MASK);
DPRINTF(sc, BWI_DBG_ATTACH | BWI_DBG_INIT | BWI_DBG_MISC,
"bus rev %u\n", busrev);
if (busrev == BWI_BUSREV_0)
return BWI_STATE_LO_DISABLE1;
else if (busrev == BWI_BUSREV_1)
return BWI_STATE_LO_DISABLE2;
else
return (BWI_STATE_LO_DISABLE1 | BWI_STATE_LO_DISABLE2);
}
int
bwi_regwin_is_enabled(struct bwi_softc *sc, struct bwi_regwin *rw)
{
uint32_t val, disable_bits;
disable_bits = bwi_regwin_disable_bits(sc);
val = CSR_READ_4(sc, BWI_STATE_LO);
if ((val & (BWI_STATE_LO_CLOCK |
BWI_STATE_LO_RESET |
disable_bits)) == BWI_STATE_LO_CLOCK) {
DPRINTF(sc, BWI_DBG_ATTACH | BWI_DBG_INIT, "%s is enabled\n",
bwi_regwin_name(rw));
return 1;
} else {
DPRINTF(sc, BWI_DBG_ATTACH | BWI_DBG_INIT, "%s is disabled\n",
bwi_regwin_name(rw));
return 0;
}
}
void
bwi_regwin_disable(struct bwi_softc *sc, struct bwi_regwin *rw, uint32_t flags)
{
uint32_t state_lo, disable_bits;
int i;
state_lo = CSR_READ_4(sc, BWI_STATE_LO);
/*
* If current regwin is in 'reset' state, it was already disabled.
*/
if (state_lo & BWI_STATE_LO_RESET) {
DPRINTF(sc, BWI_DBG_ATTACH | BWI_DBG_INIT,
"%s was already disabled\n", bwi_regwin_name(rw));
return;
}
disable_bits = bwi_regwin_disable_bits(sc);
/*
* Disable normal clock
*/
state_lo = BWI_STATE_LO_CLOCK | disable_bits;
CSR_WRITE_4(sc, BWI_STATE_LO, state_lo);
/*
* Wait until normal clock is disabled
*/
#define NRETRY 1000
for (i = 0; i < NRETRY; ++i) {
state_lo = CSR_READ_4(sc, BWI_STATE_LO);
if (state_lo & disable_bits)
break;
DELAY(10);
}
if (i == NRETRY) {
device_printf(sc->sc_dev, "%s disable clock timeout\n",
bwi_regwin_name(rw));
}
for (i = 0; i < NRETRY; ++i) {
uint32_t state_hi;
state_hi = CSR_READ_4(sc, BWI_STATE_HI);
if ((state_hi & BWI_STATE_HI_BUSY) == 0)
break;
DELAY(10);
}
if (i == NRETRY) {
device_printf(sc->sc_dev, "%s wait BUSY unset timeout\n",
bwi_regwin_name(rw));
}
#undef NRETRY
/*
* Reset and disable regwin with gated clock
*/
state_lo = BWI_STATE_LO_RESET | disable_bits |
BWI_STATE_LO_CLOCK | BWI_STATE_LO_GATED_CLOCK |
__SHIFTIN(flags, BWI_STATE_LO_FLAGS_MASK);
CSR_WRITE_4(sc, BWI_STATE_LO, state_lo);
/* Flush pending bus write */
CSR_READ_4(sc, BWI_STATE_LO);
DELAY(1);
/* Reset and disable regwin */
state_lo = BWI_STATE_LO_RESET | disable_bits |
__SHIFTIN(flags, BWI_STATE_LO_FLAGS_MASK);
CSR_WRITE_4(sc, BWI_STATE_LO, state_lo);
/* Flush pending bus write */
CSR_READ_4(sc, BWI_STATE_LO);
DELAY(1);
}
void
bwi_regwin_enable(struct bwi_softc *sc, struct bwi_regwin *rw, uint32_t flags)
{
uint32_t state_lo, state_hi, imstate;
bwi_regwin_disable(sc, rw, flags);
/* Reset regwin with gated clock */
state_lo = BWI_STATE_LO_RESET |
BWI_STATE_LO_CLOCK |
BWI_STATE_LO_GATED_CLOCK |
__SHIFTIN(flags, BWI_STATE_LO_FLAGS_MASK);
CSR_WRITE_4(sc, BWI_STATE_LO, state_lo);
/* Flush pending bus write */
CSR_READ_4(sc, BWI_STATE_LO);
DELAY(1);
state_hi = CSR_READ_4(sc, BWI_STATE_HI);
if (state_hi & BWI_STATE_HI_SERROR)
CSR_WRITE_4(sc, BWI_STATE_HI, 0);
imstate = CSR_READ_4(sc, BWI_IMSTATE);
if (imstate & (BWI_IMSTATE_INBAND_ERR | BWI_IMSTATE_TIMEOUT)) {
imstate &= ~(BWI_IMSTATE_INBAND_ERR | BWI_IMSTATE_TIMEOUT);
CSR_WRITE_4(sc, BWI_IMSTATE, imstate);
}
/* Enable regwin with gated clock */
state_lo = BWI_STATE_LO_CLOCK |
BWI_STATE_LO_GATED_CLOCK |
__SHIFTIN(flags, BWI_STATE_LO_FLAGS_MASK);
CSR_WRITE_4(sc, BWI_STATE_LO, state_lo);
/* Flush pending bus write */
CSR_READ_4(sc, BWI_STATE_LO);
DELAY(1);
/* Enable regwin with normal clock */
state_lo = BWI_STATE_LO_CLOCK |
__SHIFTIN(flags, BWI_STATE_LO_FLAGS_MASK);
CSR_WRITE_4(sc, BWI_STATE_LO, state_lo);
/* Flush pending bus write */
CSR_READ_4(sc, BWI_STATE_LO);
DELAY(1);
}
static void
bwi_set_bssid(struct bwi_softc *sc, const uint8_t *bssid)
{
struct bwi_mac *mac;
struct bwi_myaddr_bssid buf;
const uint8_t *p;
uint32_t val;
int n, i;
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
bwi_set_addr_filter(sc, BWI_ADDR_FILTER_BSSID, bssid);
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
bcopy(sc->sc_ic.ic_macaddr, buf.myaddr, sizeof(buf.myaddr));
bcopy(bssid, buf.bssid, sizeof(buf.bssid));
n = sizeof(buf) / sizeof(val);
p = (const uint8_t *)&buf;
for (i = 0; i < n; ++i) {
int j;
val = 0;
for (j = 0; j < sizeof(val); ++j)
val |= ((uint32_t)(*p++)) << (j * 8);
TMPLT_WRITE_4(mac, 0x20 + (i * sizeof(val)), val);
}
}
static void
bwi_updateslot(struct ieee80211com *ic)
{
struct bwi_softc *sc = ic->ic_softc;
struct bwi_mac *mac;
BWI_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
if (sc->sc_flags & BWI_F_RUNNING) {
DPRINTF(sc, BWI_DBG_80211, "%s\n", __func__);
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
bwi_mac_updateslot(mac, (ic->ic_flags & IEEE80211_F_SHSLOT));
}
BWI_UNLOCK(sc);
}
static void
bwi_calibrate(void *xsc)
{
struct bwi_softc *sc = xsc;
struct bwi_mac *mac;
BWI_ASSERT_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
KASSERT(sc->sc_ic.ic_opmode != IEEE80211_M_MONITOR,
("opmode %d", sc->sc_ic.ic_opmode));
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
bwi_mac_calibrate_txpower(mac, sc->sc_txpwrcb_type);
sc->sc_txpwrcb_type = BWI_TXPWR_CALIB;
/* XXX 15 seconds */
callout_reset(&sc->sc_calib_ch, hz * 15, bwi_calibrate, sc);
}
static int
bwi_calc_rssi(struct bwi_softc *sc, const struct bwi_rxbuf_hdr *hdr)
{
struct bwi_mac *mac;
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
return bwi_rf_calc_rssi(mac, hdr);
}
static int
bwi_calc_noise(struct bwi_softc *sc)
{
struct bwi_mac *mac;
KASSERT(sc->sc_cur_regwin->rw_type == BWI_REGWIN_T_MAC,
("current regwin type %d", sc->sc_cur_regwin->rw_type));
mac = (struct bwi_mac *)sc->sc_cur_regwin;
return bwi_rf_calc_noise(mac);
}
static __inline uint8_t
bwi_plcp2rate(const uint32_t plcp0, enum ieee80211_phytype type)
{
uint32_t plcp = le32toh(plcp0) & IEEE80211_OFDM_PLCP_RATE_MASK;
return (ieee80211_plcp2rate(plcp, type));
}
static void
bwi_rx_radiotap(struct bwi_softc *sc, struct mbuf *m,
struct bwi_rxbuf_hdr *hdr, const void *plcp, int rate, int rssi, int noise)
{
const struct ieee80211_frame_min *wh;
sc->sc_rx_th.wr_flags = IEEE80211_RADIOTAP_F_FCS;
if (htole16(hdr->rxh_flags1) & BWI_RXH_F1_SHPREAMBLE)
sc->sc_rx_th.wr_flags |= IEEE80211_RADIOTAP_F_SHORTPRE;
wh = mtod(m, const struct ieee80211_frame_min *);
if (wh->i_fc[1] & IEEE80211_FC1_PROTECTED)
sc->sc_rx_th.wr_flags |= IEEE80211_RADIOTAP_F_WEP;
sc->sc_rx_th.wr_tsf = hdr->rxh_tsf; /* No endian conversion */
sc->sc_rx_th.wr_rate = rate;
sc->sc_rx_th.wr_antsignal = rssi;
sc->sc_rx_th.wr_antnoise = noise;
}
static void
bwi_led_attach(struct bwi_softc *sc)
{
const uint8_t *led_act = NULL;
uint16_t gpio, val[BWI_LED_MAX];
int i;
for (i = 0; i < nitems(bwi_vendor_led_act); ++i) {
if (sc->sc_pci_subvid == bwi_vendor_led_act[i].vid) {
led_act = bwi_vendor_led_act[i].led_act;
break;
}
}
if (led_act == NULL)
led_act = bwi_default_led_act;
gpio = bwi_read_sprom(sc, BWI_SPROM_GPIO01);
val[0] = __SHIFTOUT(gpio, BWI_SPROM_GPIO_0);
val[1] = __SHIFTOUT(gpio, BWI_SPROM_GPIO_1);
gpio = bwi_read_sprom(sc, BWI_SPROM_GPIO23);
val[2] = __SHIFTOUT(gpio, BWI_SPROM_GPIO_2);
val[3] = __SHIFTOUT(gpio, BWI_SPROM_GPIO_3);
for (i = 0; i < BWI_LED_MAX; ++i) {
struct bwi_led *led = &sc->sc_leds[i];
if (val[i] == 0xff) {
led->l_act = led_act[i];
} else {
if (val[i] & BWI_LED_ACT_LOW)
led->l_flags |= BWI_LED_F_ACTLOW;
led->l_act = __SHIFTOUT(val[i], BWI_LED_ACT_MASK);
}
led->l_mask = (1 << i);
if (led->l_act == BWI_LED_ACT_BLINK_SLOW ||
led->l_act == BWI_LED_ACT_BLINK_POLL ||
led->l_act == BWI_LED_ACT_BLINK) {
led->l_flags |= BWI_LED_F_BLINK;
if (led->l_act == BWI_LED_ACT_BLINK_POLL)
led->l_flags |= BWI_LED_F_POLLABLE;
else if (led->l_act == BWI_LED_ACT_BLINK_SLOW)
led->l_flags |= BWI_LED_F_SLOW;
if (sc->sc_blink_led == NULL) {
sc->sc_blink_led = led;
if (led->l_flags & BWI_LED_F_SLOW)
BWI_LED_SLOWDOWN(sc->sc_led_idle);
}
}
DPRINTF(sc, BWI_DBG_LED | BWI_DBG_ATTACH,
"%dth led, act %d, lowact %d\n", i,
led->l_act, led->l_flags & BWI_LED_F_ACTLOW);
}
callout_init_mtx(&sc->sc_led_blink_ch, &sc->sc_mtx, 0);
}
static __inline uint16_t
bwi_led_onoff(const struct bwi_led *led, uint16_t val, int on)
{
if (led->l_flags & BWI_LED_F_ACTLOW)
on = !on;
if (on)
val |= led->l_mask;
else
val &= ~led->l_mask;
return val;
}
static void
bwi_led_newstate(struct bwi_softc *sc, enum ieee80211_state nstate)
{
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;
uint16_t val;
int i;
if (nstate == IEEE80211_S_INIT) {
callout_stop(&sc->sc_led_blink_ch);
sc->sc_led_blinking = 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
if ((sc->sc_flags & BWI_F_RUNNING) == 0)
return;
val = CSR_READ_2(sc, BWI_MAC_GPIO_CTRL);
for (i = 0; i < BWI_LED_MAX; ++i) {
struct bwi_led *led = &sc->sc_leds[i];
int on;
if (led->l_act == BWI_LED_ACT_UNKN ||
led->l_act == BWI_LED_ACT_NULL)
continue;
if ((led->l_flags & BWI_LED_F_BLINK) &&
nstate != IEEE80211_S_INIT)
continue;
switch (led->l_act) {
case BWI_LED_ACT_ON: /* Always on */
on = 1;
break;
case BWI_LED_ACT_OFF: /* Always off */
case BWI_LED_ACT_5GHZ: /* TODO: 11A */
on = 0;
break;
default:
on = 1;
switch (nstate) {
case IEEE80211_S_INIT:
on = 0;
break;
case IEEE80211_S_RUN:
if (led->l_act == BWI_LED_ACT_11G &&
ic->ic_curmode != IEEE80211_MODE_11G)
on = 0;
break;
default:
if (led->l_act == BWI_LED_ACT_ASSOC)
on = 0;
break;
}
break;
}
val = bwi_led_onoff(led, val, on);
}
CSR_WRITE_2(sc, BWI_MAC_GPIO_CTRL, val);
}
static void
bwi_led_event(struct bwi_softc *sc, int event)
{
struct bwi_led *led = sc->sc_blink_led;
int rate;
if (event == BWI_LED_EVENT_POLL) {
if ((led->l_flags & BWI_LED_F_POLLABLE) == 0)
return;
if (ticks - sc->sc_led_ticks < sc->sc_led_idle)
return;
}
sc->sc_led_ticks = ticks;
if (sc->sc_led_blinking)
return;
switch (event) {
case BWI_LED_EVENT_RX:
rate = sc->sc_rx_rate;
break;
case BWI_LED_EVENT_TX:
rate = sc->sc_tx_rate;
break;
case BWI_LED_EVENT_POLL:
rate = 0;
break;
default:
panic("unknown LED event %d\n", event);
break;
}
bwi_led_blink_start(sc, bwi_led_duration[rate].on_dur,
bwi_led_duration[rate].off_dur);
}
static void
bwi_led_blink_start(struct bwi_softc *sc, int on_dur, int off_dur)
{
struct bwi_led *led = sc->sc_blink_led;
uint16_t val;
val = CSR_READ_2(sc, BWI_MAC_GPIO_CTRL);
val = bwi_led_onoff(led, val, 1);
CSR_WRITE_2(sc, BWI_MAC_GPIO_CTRL, val);
if (led->l_flags & BWI_LED_F_SLOW) {
BWI_LED_SLOWDOWN(on_dur);
BWI_LED_SLOWDOWN(off_dur);
}
sc->sc_led_blinking = 1;
sc->sc_led_blink_offdur = off_dur;
callout_reset(&sc->sc_led_blink_ch, on_dur, bwi_led_blink_next, sc);
}
static void
bwi_led_blink_next(void *xsc)
{
struct bwi_softc *sc = xsc;
uint16_t val;
val = CSR_READ_2(sc, BWI_MAC_GPIO_CTRL);
val = bwi_led_onoff(sc->sc_blink_led, val, 0);
CSR_WRITE_2(sc, BWI_MAC_GPIO_CTRL, val);
callout_reset(&sc->sc_led_blink_ch, sc->sc_led_blink_offdur,
bwi_led_blink_end, sc);
}
static void
bwi_led_blink_end(void *xsc)
{
struct bwi_softc *sc = xsc;
sc->sc_led_blinking = 0;
}
static void
bwi_restart(void *xsc, int pending)
{
struct bwi_softc *sc = xsc;
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
device_printf(sc->sc_dev, "%s begin, help!\n", __func__);
BWI_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
bwi_init_statechg(sc, 0);
#if 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
bwi_start_locked(sc);
#endif
BWI_UNLOCK(sc);
}