Apply the same change as for if_iwi.c r1.114 here.

This driver sleeps during wpi_media_change(), and thus requires an adaptive
mutex for the media lock.

Suspend the getrfkill thread while we're suspended.

Otherwise it tries to futz with device registers, which doesn't work,
and who knows, maybe is bad.

branches: 1.21.12; 1.21.14;
wlan interfaces make interrupt routine running on softint context.

see http://mail-index.netbsd.org/tech-kern/2016/12/06/msg021281.html

tested device:
* ath at pci: AR5212, AR5424
* athn at pci: AR9287
* ipw at pci: 2100BG
* iwi at pci: 2915ABG
* iwm at pci: 3165, 7260, 8260
* iwn at pci: 4945, 6235
* ral at pci: RT2560
* rtwn at pci: RTL8192CE

wpi(4): use MSI if available.

branches: 1.19.2; 1.19.4;
As proposed on tech-net@, introduce a new switch type, PSWITCH_TYPE_RADIO,
to be used to report to userland hardware radio switch changes.
powerd(8) will call a "radio_button" script to handle the event.
This script can e.g. start or stop wpa_supplicant.
Update wpi(4) to report PSWITCH_TYPE_RADIO events to sysmon.

branches: 1.18.2; 1.18.4;
increase WPI_RBUF_COUNT

simplify wpi_rx_intr rbuf handling and add some more bus_dmamap_sync; fixes stalls when downloading large files for me

bus_dmamap_sync for rx descriptors too, derived from openbsd driver

branches: 1.15.22; 1.15.36;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

Forward declare struct bpf_if and use that as the type for bpf_if
instead of "void *". Buys us oo times the type-safety for 0 times
the price.
(no functional change)

Cache wpi firmware in memory, free it only on invalid content or when
the last user is detached. This stops wpi from accessing the disk on
resume.

branches: 1.12.2; 1.12.4; 1.12.6;
Protect private mbuf external data storage allocater with a mutex, as it
was done in nfe(4), sk(4) and msk(4). And maybe some others.
Confirmed to fix the usual issue of data corruption by Jared McNeill.

branches: 1.11.6; 1.11.10; 1.11.12; 1.11.14;
Add some sysctl to retrieve the radio state (and the debug level).

While here, fixe the return value in case where radio is off (suggested by
joerg@).

branches: 1.10.2;
Merge jmcneill-pm branch.

branches: 1.9.2; 1.9.4;
Use device_t instead of struct device* and use associated functions
Use callout_setfunc and callout_scheduler instead of callout_reset

No functionnal change expected

branches: 1.8.4; 1.8.6;
The service queue isn't used anymore so don't allocate it.
It saves a bit of memory and reduces diff with other BSD.

Fix scanning code for wpi based on the iwi code.

Keep track of the status of the scan.
On a transition IEEE80211_S_SCAN -> IEEE80211_S_SCAN, don't stop the previous
scan, finish the current scan.
When we receive some frames in the IEEE80211_S_SCAN state, set current_channel
based on the value advertised in beacons or probre reponse.

Moreover, it fixes WPA issues for me.

branches: 1.6.2; 1.6.6; 1.6.8;
Sync the driver with the OpenBSD one

1/ Update the driver to use the new firmware images from Intel (2.14.3.)
2/ Read the list of supported channels from the EEPROM instead of having
it hard-coded in the driver.
3/ Limit output power to what is specified in EEPROM.
4/ Decrease output power for highest OFDM rates to reduce distortion.
5/ Automatically adjust output power to temperature changes for increased
throughput and range.
6/ Attach the adapter's onboard thermal sensor to the sensor framework.
7/ Replace 'magic' fields in structures with their correct definitions.
8/ Rewrite the firmware load in order to reduce the diff with OpenBSD one

NOTE2: you must install sysutils/wpi-firmware2 in order to use the new
driver

NOTE2: if you are using a channel not allowed by the regulatory domain
of your adapter, you will no longer be able to associate.

Thanks a lot for his hard work to damien@OpenBSD.org

Add a workaround in the case where we have low number of rbuf.
It seems to fix problem of frozen network with wpi.

ok by @joerg.

branches: 1.4.2; 1.4.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.3.2;
Sync the wpi driver with the openbsd one. Thanks to Jean-Baptiste
Campesato for his work.
Fix PR/34463


ok dyoung@ joerg@

branches: 1.2.2; 1.2.4; 1.2.6;
Move AMRR code out of wpi(4) and ural(4) into net80211 itself.
From OpenBSD.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8;
Add a port of the OpenBSD Intel 3945ABG wpi(4) wireless driver, by
Jean-Baptiste Campesato.

From PR kern/33778.

Suspend the getrfkill thread while we're suspended.

Otherwise it tries to futz with device registers, which doesn't work,
and who knows, maybe is bad.

branches: 1.21.12;
wlan interfaces make interrupt routine running on softint context.

see http://mail-index.netbsd.org/tech-kern/2016/12/06/msg021281.html

tested device:
* ath at pci: AR5212, AR5424
* athn at pci: AR9287
* ipw at pci: 2100BG
* iwi at pci: 2915ABG
* iwm at pci: 3165, 7260, 8260
* iwn at pci: 4945, 6235
* ral at pci: RT2560
* rtwn at pci: RTL8192CE

wpi(4): use MSI if available.

branches: 1.19.2; 1.19.4;
As proposed on tech-net@, introduce a new switch type, PSWITCH_TYPE_RADIO,
to be used to report to userland hardware radio switch changes.
powerd(8) will call a "radio_button" script to handle the event.
This script can e.g. start or stop wpa_supplicant.
Update wpi(4) to report PSWITCH_TYPE_RADIO events to sysmon.

branches: 1.18.2; 1.18.4;
increase WPI_RBUF_COUNT

simplify wpi_rx_intr rbuf handling and add some more bus_dmamap_sync; fixes stalls when downloading large files for me

bus_dmamap_sync for rx descriptors too, derived from openbsd driver

branches: 1.15.22; 1.15.36;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

Forward declare struct bpf_if and use that as the type for bpf_if
instead of "void *". Buys us oo times the type-safety for 0 times
the price.
(no functional change)

Cache wpi firmware in memory, free it only on invalid content or when
the last user is detached. This stops wpi from accessing the disk on
resume.

branches: 1.12.2; 1.12.4; 1.12.6;
Protect private mbuf external data storage allocater with a mutex, as it
was done in nfe(4), sk(4) and msk(4). And maybe some others.
Confirmed to fix the usual issue of data corruption by Jared McNeill.

branches: 1.11.6; 1.11.10; 1.11.12; 1.11.14;
Add some sysctl to retrieve the radio state (and the debug level).

While here, fixe the return value in case where radio is off (suggested by
joerg@).

branches: 1.10.2;
Merge jmcneill-pm branch.

branches: 1.9.2; 1.9.4;
Use device_t instead of struct device* and use associated functions
Use callout_setfunc and callout_scheduler instead of callout_reset

No functionnal change expected

branches: 1.8.4; 1.8.6;
The service queue isn't used anymore so don't allocate it.
It saves a bit of memory and reduces diff with other BSD.

Fix scanning code for wpi based on the iwi code.

Keep track of the status of the scan.
On a transition IEEE80211_S_SCAN -> IEEE80211_S_SCAN, don't stop the previous
scan, finish the current scan.
When we receive some frames in the IEEE80211_S_SCAN state, set current_channel
based on the value advertised in beacons or probre reponse.

Moreover, it fixes WPA issues for me.

branches: 1.6.2; 1.6.6; 1.6.8;
Sync the driver with the OpenBSD one

1/ Update the driver to use the new firmware images from Intel (2.14.3.)
2/ Read the list of supported channels from the EEPROM instead of having
it hard-coded in the driver.
3/ Limit output power to what is specified in EEPROM.
4/ Decrease output power for highest OFDM rates to reduce distortion.
5/ Automatically adjust output power to temperature changes for increased
throughput and range.
6/ Attach the adapter's onboard thermal sensor to the sensor framework.
7/ Replace 'magic' fields in structures with their correct definitions.
8/ Rewrite the firmware load in order to reduce the diff with OpenBSD one

NOTE2: you must install sysutils/wpi-firmware2 in order to use the new
driver

NOTE2: if you are using a channel not allowed by the regulatory domain
of your adapter, you will no longer be able to associate.

Thanks a lot for his hard work to damien@OpenBSD.org

Add a workaround in the case where we have low number of rbuf.
It seems to fix problem of frozen network with wpi.

ok by @joerg.

branches: 1.4.2; 1.4.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.3.2;
Sync the wpi driver with the openbsd one. Thanks to Jean-Baptiste
Campesato for his work.
Fix PR/34463


ok dyoung@ joerg@

branches: 1.2.2; 1.2.4; 1.2.6;
Move AMRR code out of wpi(4) and ural(4) into net80211 itself.
From OpenBSD.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8;
Add a port of the OpenBSD Intel 3945ABG wpi(4) wireless driver, by
Jean-Baptiste Campesato.

From PR kern/33778.