Ticket #584 (closed defect: fixed)

Opened 8 years ago

Last modified 5 years ago

WPA/WPA2 not working with Marvell Libertas

Reported by: marcelo Owned by: marcelo
Priority: high Milestone: Trial-2
Component: distro Version:
Keywords: relnote Cc: dcbw@…
Action Needed: never set Verified: no
Deployments affected: Blocked By:
Blocking:

Description

wpa_supplicant fails to authenticate via WAP-PSK with the generic wext driver:

# cat /etc/wpa_supplicant/wpa_supplicant.conf
ctrl_interface=/var/run/wpa_supplicant
ctrl_interface_group=wheel
network={
        ssid="default"
        key_mgmt=WPA-PSK
        proto=WPA
        pairwise=TKIP
        group=TKIP
        psk="1234567890"
}

# strace wpa_supplicant -ieth0 -Dwext -c/etc/wpa_supplicant/wpa_supplicant.conf
...
ioctl(3, 0x8b32, 0xbfc6e2fc)            = -1 EOPNOTSUPP (Operation not supported)
write(2, "ioctl[SIOCSIWAUTH]: Operation no"..., 44ioctl[SIOCSIWAUTH]: Operation
not supported
) = 44
write(2, "WEXT auth param 10 value 0x1 - ", 31WEXT auth param 10 value 0x1 - ) =
 31
write(1, "wpa_driver_wext_associate: assoc"..., 87) = 87
ioctl(3, 0x8b32, 0xbfc6e2fc)            = -1 EOPNOTSUPP (Operation not supported
)
write(2, "ioctl[SIOCSIWAUTH]: Operation no"..., 44ioctl[SIOCSIWAUTH]: Operation
not supported
) = 44
write(2, "WEXT auth param 8 value 0x0 - ", 30WEXT auth param 8 value 0x0 - ) = 3
0
write(1, "wpa_driver_wext_associate: assoc"..., 92) = 92
ioctl(3, SIOCSIWFREQ, 0xbfc6e2fc)       = -1 EOPNOTSUPP (Operation not supported
)
B
write(2, "ioctl[SIOCSIWFREQ]: Operation no"..., 44ioctl[SIOCSIWFREQ]: Operation
not supported

pprpplague mentions a marvell wpa_supplicant driver patch at http://svn.gumstix.com/gumstix-buildroot/trunk/package/wpa_supplicant/patch_wpa_supplicant-0.4.7-MSI-2-GUM

Change History

  Changed 8 years ago by dcbw

Can you just run wpa_supplicant with "-ddd" and grab the output rather than stracing it? That should make for more readable debug info.

  Changed 8 years ago by jg

  • owner changed from marcelo to dcbw

dcbw, what's the state a month later with lots of driver and firmware goodies in the meanwhile?

  Changed 8 years ago by dcbw

Nothing specific to WPA TKIP or AES/CCMP has been fixed since then AFAIK. Ticket should remain open.

  Changed 8 years ago by dcbw

  • milestone changed from BTest-2 to BTest-3

  Changed 8 years ago by marcelo

  • owner changed from dcbw to marcelo

Reassigning it to me so I remember about it easily, from what I gather dcbw is busy with other things, but I won't mind if he fixes it before I get a chance to ;)

  Changed 7 years ago by cjb

  • keywords relnote added

follow-up: ↓ 8   Changed 7 years ago by dcbw

Looked into this somewhat tonight. It appears we need to bring some sanity to scan result handling internally in the driver. There are a few problems:

1) wpa_supplicant can't get scan results out of the driver most of the time when using WPA; which is a showstopper for wpa support because wpa_supplicant then can't find the AP to connect to. It's somewhat odd because it works with NetworkManager, but I think some internal wpa_supplicant scan handling when trying to associate immediately after startup without a first scan is screwing it up. But that doesn't preclude the need for the real solution.

2) We should use 'struct ieee80211_network' from include/net/ieee80211.h rather than the internal 'struct bss_descriptor'. We could then potentially use the ieee80211 SIOCGIWSCAN handler rather than our own cocked-up one, plus use the normal kernel linked list logic rather than the really big 'scantable' list from dev.h

3) The driver should keep the scan list around (not replace it entirely every scan), cull old results based on last beacon time, update existing entries that are in the scan, and add new entries. This way, the GIWSCAN handler can always return the scan table regardless of the internal scan progress, which is the _real_ fix for current the wpa_supplicant scan problems. Essentially, do what ipw* and mac80211 do. I converted airo to do this at one point as well, and it's a somewhat simpler example.

It's not really complex work but it would take some time and is certainly not possible within the Trial 1 timeframe. BTest-3 is appropriate.

in reply to: ↑ 7   Changed 7 years ago by jg

I agree with your analysis; important to fix, but not for Trial-1. BTest-3 is indeed the right milestone.

Replying to dcbw:

It's not really complex work but it would take some time and is certainly not possible within the Trial 1 timeframe. BTest-3 is appropriate.

  Changed 7 years ago by jg

  • priority changed from normal to high
  • verified unset
  • milestone changed from BTest-4 to Trial-2

  Changed 7 years ago by dcbw

  • status changed from new to closed
  • resolution set to fixed

WPA support fixed in libertas-2.6 git head. UI bits to accept a WPA pre-shared key (or, god forbid, IEEE 802.1x auth parameters) should get another ticket.

  Changed 5 years ago by alivenk

  • next_action set to never set

If you’re a dedicated follower of tiffany jewellery like me. Don't miss the tiffany co & co. on sale including pendants, necklace, earrings, bracelets on line. tiffany jewelry is the one thing that outlasts the cake, champagne and music. links of london jewelry discount , famous for its sweetie and friendship bracelets.

Christian Audigier designer, ed hardy clothing , is a brand of ed hardy, and now are favored in the moderning world as a mark of its nice tatoo.

Note: See TracTickets for help on using tickets.