Might as well join this thread as the other one (here) is getting awfully long. (If the moderator prefers otherwise, please let me know.)
I took a mango with 3.025 and a mango with the firmware @wellnw provided to a local electronics store with a selection of Chromebooks for sale.
I was unable to find any model which absolutely refused to connect. However, I found a number of models which would connect rather erratically. Sometimes they would just time out with an unspecified “error”, other times they connected in a couple seconds and some times it took over 10 seconds to connect.
I’m still going through the pcap files I gathered. First a qualifier:
- I’m not sure that all the traffic was captured. I seem to have many Response messages without a matching Request message. This would make it useless for recreating the problem, though I can still do things like count the number of requests or measure the time between Authentication and the Key exchange.
- Also, being an electronics store, there is a huge amount of other WiFi activity going on. Most of the captured packets were for other equipment and have to be filtered out.
Observations:
- Problematic models Acer Spin 11 (CP311-HN, 48:f1:7f:49:cd:d9), PIxelbooks GA00122-US (f8:94:c2:c1:67:90) and GA00345-US (40:74:e0:e4:41:84), Dell P54G, and ASUS C433 (90:78:41:d5:b3:94).
- Not surprisingly, all these mac addresses are associated with Intel corp. Probably all have the same driver. However, there are units with Intel mac addresses which DO work. I don’t have any way to map a mac address to a chip set, however.
- We do get into the case reported (3.025 firmware) where the Association Requests eventually stop getting sent and a Deauthentification is sent with a 4-way handshake timeout code.
- Once a model connects, it’s more likey to connect again, but timing is still wildly variable. Even after a successful connection, future connection attempts can fail.
- The Acer R751T (a4:c3:f0:bb:b9:17) connected pretty quickly, however, it still sent many Association Requests after the Authentication and before the key exchange. The HP was much cleaner - not tons of repeat Association Requests.
- The HP x360a (0c:dd:24:c7:2e:be) connected quickly and reliably. Its mac address is also assigned to Intel.
Now for the bad new… I didn’t see any real difference between connecting to a mango with 3.025 or the test firmware @wellnw sent me.
There is still enormous variation in connect times and some of these chromebooks do fail to connect at any given time.
I will take a another look into the pcap files when I get a chance. The “best” circumstance for me, however, is to wait until I get home and can test in a quiet environment with a chromebook which has failed to connect consistently.