iwx: migrate to using net80211 crypto key methods

[adrian]

Summary

In preparation for further MFP key work, migrate iwx away from using
IEEE80211_KEY_GROUP and pointer arithmetic. This is a hold over from
ye olde days.

Locally tested:

• (STA mode, CCMP/CCMP network)
• iwx0: <Wi-Fi 6 AX210> mem 0x84c00000-0x84c03fff at device 0.0 on pci4

[bz]

Logic seems fine to me.

The details are probably worth doing to avoid further copy and paste ...

sys/dev/iwx/if_iwx.c

10951

If you use net80211_vap_printf() you neither need ic or sc and we are not using device_printf in a wireless driver anymore directly (bonus point)

10953

return (0)

10988

Here too.

10989

Here () too.

11005

Here too.

11007

Here () too

[adrian]

sys/dev/iwx/if_iwx.c

10951

If you use net80211_vap_printf() you neither need ic or sc and we are not using device_printf in a wireless driver anymore directly (bonus point)

Yeah, i really should use those routines I added. :-)

Bonus points is a subsequent commit to convert the rest of device_printf() here to ic_printf / vap_printf where appropriate.

[adrian]

updates from bz

[bz]

Sorry, nothing new from you but the original driver already seems to have gotten the return code wrong; can you please double-check?

sys/dev/iwx/if_iwx.c

10970

Can/MUST this even happen? Are we not checking the announced suites anywhere upfront in net80211? I know this was promised but did it happen?
iwx only sets:

ic->ic_cryptocaps |= IEEE80211_CRYPTO_AES_CCM;

for (*iv_key_set) [ieee80211_crypto_setkey]:
0 means error; != 0 means OK according to ieee80211_ioctl.c (no one else checks the return code it seems)

I think all errors in this function are wrong. Question is "what should be an error"?

11038

Are we leaking keys here always telling net80211 we deleted it? XXX-TODO?

[adrian]

sys/dev/iwx/if_iwx.c

10970

cryptocaps just says "what we support in hardware", remember. That's why I added the methods to make it clearer what's supported in hardware and in software.

net80211 expects all NICs to support WEP, TKIP, AES-CCM-128 in software unless you explicitly set the software cryptocaps via the calls I added earlier in the year.

As for the errors - yeah, I don't like it at all either; and I'm currently working on adding comments to net80211 key lookup / type routines so it's clearer what the API is and what the driver expectations are.

(The more I poke at cleaning up the crypto / cipher management stuff as part of MFP, the more I find gems like these hiding away :-) )

[adrian]

sys/dev/iwx/if_iwx.c

11038

oh god this would be hilarious wouldn't it

[adrian]

ok, I'd like to land this and then follow it up with some more iwx cleanups. Thanks for pointing them out though, I'm not shocked that the original upstream code was messy with key management too. ;-)

(This is why I'm trying to document the net80211 stuff, so we can better reason about all of this and then fix drivers as needed!)

[bz]

In D52523#1211903, @adrian wrote:

ok, I'd like to land this and then follow it up with some more iwx cleanups. Thanks for pointing them out though, I'm not shocked that the original upstream code was messy with key management too. ;-)

Well then fix the error returns.