09:09 <jsolano> Hi guys, anybody knows of a reliable usb wifi dongle that works well with the BBB? I am trying since yesterday with a Tplink WN725N, using driver 8188eu that I got in github (lwfinger) but it is unable to get an ip, wpa_supplicant fails. This is with kernel 6.1 on a buildroot system

09:09 <jsolano> The dongles are cheap enought that I might as well try with another one that works better

13:28 <zmatt> jsolano: wifi dongles tend to me kinda hit-and-miss... sometimes manufacturers will even produce the exact same model of usb wifi dongle with different wifi chips inside

14:32 <jsolano> zmatt: right, I'd appreciate if somebody can point me at a dongle model that has worked well recently, I would take the risk

15:47 <bradfa> jsolano: it's been a while since I worked on RTL8188EU devices, but it looks like there was a change in which driver was used for this chip around kernel v5.14. The older driver was the rtl8188eu staging driver, but that was removed in favor of the r8188eu driver.

15:48 <bradfa> jsolano: The new driver seems to have thrown away everything from the old one, and is based on a different Realtek code dump than the older rtl8188eu driver was (it was also based on a Realtek code dump but from much further in the past).

15:49 <bradfa> jsolano: the old rtl8188eu driver worked decently for me when I was last using it, but that was on a Linux 4.19 kernel

15:51 <bradfa> jsolano: I suggest you try using the r8188eu driver which is in the mainline linux kernel, in the staging tree, rather than something from github. It seems lots of recent development has been going on in mainline on it

15:53 <jsolano> bradfa: thanks, I forgot to mention that I have also used the driver in the mainline kernel (6.1) but same issues there, it also does not work... also it seems to be a staging driver with reservations about quality

15:54 <bradfa> jsolano: rtl8188eu drivers, of any vintage, in mainline have always been staging drivers, because they're based on horrible code quality code dumps from realtek. rtl8188eu drivers have been in staging for like a decade already

15:55 <bradfa> jsolano: if it makes you feel any better, I had to integrate an rtl8188eu device into a product before. We got like 14 code dumps from realtek, only 1 of them actually worked. So it's not like the staging driver is really any worse than getting code directly from the vendor :)

15:56 <bradfa> for the most part, Realtek is a pain in the butt to work with

15:56 <bradfa> Realtek parts are cheap, though, so you get what you pay for :)

16:04 <jsolano> right, I paid around 6 euro for these dongles, but I am willing to pay up to 30 or 40, one day playing with this is not worth the time

17:43 <zmatt> interesting... while the cortex-A8 will never issue read/write transfers that cross an aligned 64-byte boundary, and EDMA will likewise split large transfers at 16/32/64/128-byte boundaries (depending on configured max burst length), PRU will happily issue transfers with any amount of misalignment/boundary-crossing

17:44 <zmatt> e.g. PRU performing a 16-byte read at address 0xfffffffc will read the last 4 bytes and first 12 bytes of ddr memory

17:45 <zmatt> (instead of wrapping around to 0x00000000, which would be GPMC address space rather than ddr memory)

17:48 <zmatt> which also tells us the L3 interconnect routes transfers based on their start address and doesn't pay any attention to whether the transfer extends beyond the bounds of the targeted region