[Bug 300064] iwlwifi AX210 (fw 89): firmware assert 0x2010350B on RX BA start (ADDBA from AP) with the default firmware-package tunables

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=300064

--- Comment #2 from Michal Kilijanek <michal+freebsd@kilijanek.pl> ---
Correction to the throughput section of the description, plus a working
configuration.

I wrote that without RX aggregation the download is capped at about
1400 frames/s regardless of the PHY rate, and that only a working RX BA
session could raise it.  That is wrong.  Both of my 5 GHz measurements
were made with the 160 MHz channel width, which on its own is close to
unusable here.  With 80 MHz the same setup is about ten times faster,
still without any RX BA session:

Config: compat.linuxkpi.80211.hw_crypto=1, and `ifconfig wlan0 -ampdurx
-vht160 -vht80p80` set BEFORE association (ifconfig_wlan0 in rc.conf).
Changing vht160 on an already associated vap and running
`wpa_cli reassociate` did not change the negotiated width.  Region
PL/ETSI; AP on DFS channel 100.  All measurements were taken over wlan0
only (host routes via the AP, checked with the wlan0 byte counters).

  5 GHz, VHT80:   100 MiB at 12.0-13.8 MB/s (25 MB tests at 14-17 MB/s);
                  upload about 24 MB/s.  About 11,500 RX packets/s at
                  about 8 packets per interrupt.
  5 GHz, VHT160:  first ~45 s about 76 kB/s, then about 1.4 MB/s
                  (hw_crypto=0: stalls at 11-44 kB/s)
  2.4 GHz, HT20:  100 MiB at 2.06 MB/s; about 1 packet per interrupt
  Firmware crashes in all of these runs: 0 (RX BA declined via -ampdurx).

So with VHT80 the AP delivers several MSDUs per interrupt even without an
RX BA agreement (presumably A-MSDU), and the RX BA assert in this bug
affects throughput far less than I claimed.  It still matters: it crashes
the firmware with the package-default tunables, and full A-MPDU would
likely help further.  The VHT160 slowdown looks like a separate problem.
I am happy to file it separately if that is preferred.

-- 
You are receiving this mail because:
You are the assignee for the bug.