iwn: fix ht rate stuck at mcs 0 when using the new ra #548
Conversation
The original function take value directly from ic->ic_bss, which may not always consist with other places..
|
Thank you for the testing. which commit introduced this bug? |
| @@ -4102,7 +4102,7 @@ iwn_set_link_quality(struct iwn_softc *sc, struct ieee80211_node *ni) | |||
| } | |||
| } | |||
|
|
|||
| if (tab == 0) | |||
| if (tab == 0 && rflags == 0) | |||
There was a problem hiding this comment.
This line is fixing the bug.
When rflags contains IWN_RFLAG_MCS, 0 is a good value for tab means mcs 0. In this case it should not break the loop. The result of this bug that the initial link quality retry table contain only legacy rates and no ht rate. In this condition the wifi card will not send any packets using ht rate, and will not trigger the iwn_ra_choose to update the table again. For some reason it was working with mira, but not the new ra. The fix is to check and break the loop when both tab and rflags are 0.
This bug was introduced in the df94da8 when port the logic from iwm to support tx mimo in iwn. The struct is one uint16 in iwm instead of two uint8 in iwn. The additional check is missed.
There is a bug in the
iwndriver for Centrino Wireless-N devices when set firmware rate table.It was masked by the original
miraalgorithm. When the iwn driver is switched to the newraalgorithm, the tx rate is stuck at mcs 0 (6mpbs).This also include a miscellaneous fix the rxon flags in a corner case.