apritzel has quit [Ping timeout: 480 seconds]
jernej has joined #linux-sunxi
Mangy_Dog has quit [Ping timeout: 480 seconds]
Luke-Jr has quit [Ping timeout: 480 seconds]
Luke-Jr has joined #linux-sunxi
machinehum has joined #linux-sunxi
<
MoeIcenowy>
jernej: I hacked your rtw88-sdio branch to add 8723ds support, and it work.
<
MoeIcenowy>
but frequently deassoicate with AP
ftg has quit [Ping timeout: 480 seconds]
<
jernej>
MoeIcenowy: What hacks did you add? anything to SDIO?
<
jernej>
note that just yesterday I fixed few bugs
<
jernej>
but there might be still things missing...
<
anarsoul>
MoeIcenowy: what about 8723bs/cs?
<
jernej>
anarsoul: rtw88 has already support for D variant, but not for B or C
<
jernej>
and at least between 8822 B and C, there is pretty big difference regarding configuration
<
anarsoul>
yeah, even B and C differ a lot
<
anarsoul>
I mean 8723
<
jernej>
so adding different variants would take some time...
<
jernej>
but at least rtw88 maintainer is super responsive and can be asked about details...
<
jernej>
MoeIcenowy: which board has 8723ds?
<
jernej>
MoeIcenowy: I added few registers reads/writes to match vendor driver more closely, maybe it will change something for you
<
jernej>
additionally, I also found two differences:
<
jernej>
1. vendor driver aggregates packets, but that would only affect max. speed
<
jernej>
2. vendor driver checks if there is enough space in TX queue, but that should not affect stability under light load
<
anarsoul>
jernej: it shouldn't really affect sdio
<
anarsoul>
packet aggregation is USB thing
<
anarsoul>
basically min transfer size is 512 on USB 2.0 (i.e. USB packet size)
<
jernej>
well, SDIO works best when ops are aligned to block size too
<
MoeIcenowy>
jernej: I just adapted your changes to 8822cs to 8723ds
<
jernej>
and most of the time, it's 512 too
<
anarsoul>
jernej: but it doesn't waste the cycles, does it?
<
MoeIcenowy>
(a stub driver + efuse parser
<
anarsoul>
i.e. transferring 1 byte and 512 bytes over USB takes the same time
<
jernej>
SDIO operations block (sleep), so they are pretty slow
<
anarsoul>
OK, anyway it's usually easy to add packet aggregation later
<
jernej>
queue free space check is more pressing need
<
MoeIcenowy>
well what I met is, in fact, realtek stock driver hanging on 5.14
<
anarsoul>
MoeIcenowy: no wonder if it's based off 8723cs :)
<
MoeIcenowy>
Is there a standard DT binding that indicates a peripheral has a different base of memory with the CPU?
<
MoeIcenowy>
(or should this kind of thing being added as a platform glue in the driver, and then a dedicated compatible
<
anarsoul>
MoeIcenowy: reg?
<
MoeIcenowy>
anarsoul: nope
<
MoeIcenowy>
it's a offset which should be applied when accessing system RAM
<
MoeIcenowy>
e.g. to pass an address to sun4i-a10-display-backend, you need to subtract it with 0x40000000
<
jernej>
MoeIcenowy: I believe you're looking for dma-ranges, check sun50i-a64.dtsi for example, there is mbus node for that
<
jernej>
other nodes have interconnects property with phandle to mbus
<
jernej>
note that you don't need actual mbus driver for that
apritzel has joined #linux-sunxi
<
MoeIcenowy>
jernej: thanks for your tips ;-)
hlauer has joined #linux-sunxi
cmeerw has joined #linux-sunxi
apritzel has quit [Ping timeout: 480 seconds]
<
MoeIcenowy>
jernej: for 8723ds driver, I think I met some issues in kernel rtnl lock
<
jernej>
yeah, that's completely possible
<
jernej>
rtw88 code is not prepared for sdio, where all operations block
<
jernej>
MoeIcenowy: can you pinpoint which lock is bothering you? stack trace would help...
<
jernej>
it's entirely possible we missed some lock
<
MoeIcenowy>
rtnl_lock
<
MoeIcenowy>
jernej: I mean BSP driver, not rtw88
<
MoeIcenowy>
rtw88 just meets continous disassociate
<
MoeIcenowy>
(I am doubting my kernel'
<
MoeIcenowy>
's regulatory stack is broken
<
jernej>
how often does it disassociate with rtw88? now and then or often (range minutes or less)?
<
MoeIcenowy>
first time ~1 minutes
<
MoeIcenowy>
then ~10s
<
jernej>
oh, ok, then I guess it's something 8723ds specific, it's 10 min or more for me
<
MoeIcenowy>
jernej: for stock driver, I think the problem is calling rtw_regd_init inside rtw_cfg80211_init_wiphy
<
MoeIcenowy>
your 8822bs one has it called in rtw_wiphy_register instead
<
jernej>
you can test that easily :)
<
MoeIcenowy>
jernej: I got BSP driver working now, but it's also behaving bad on this AP and this
<
jernej>
also disassociates frequently?
<
MoeIcenowy>
the AP even rejected me
<
MoeIcenowy>
lwfinger/rtl8723ds is just way toooooooooooooooooooooooooooooooooooooooo old to use.
<
jernej>
I'm not sure if newer vendor driver is available
Mangy_Dog has joined #linux-sunxi
apritzel has joined #linux-sunxi
<
MoeIcenowy>
finally the most working driver for me is the rtw88-sdio hack
<
gamiee>
jernej: there was a lot of latest drivers on Realtek FTP,but sadly, the leaked credentials doesn't work anymore :(
warpme_ has quit [Quit: Connection closed for inactivity]
<
jernej>
I believe you can get vendor driver if you ask Realtek main rtw88 developer
<
jernej>
at least he offered 8822bs driver to me, when I asked about SDIO support
apritzel has quit [Ping timeout: 480 seconds]
ftg has joined #linux-sunxi
hlauer has quit [Ping timeout: 480 seconds]
apritzel has joined #linux-sunxi
apritzel has quit [Ping timeout: 480 seconds]
sunshavi has joined #linux-sunxi
vagrantc has joined #linux-sunxi
apritzel has joined #linux-sunxi
cmeerw has quit [Ping timeout: 480 seconds]