02:21 <bashintosh> Hi, does someone know why Rockchip has the DWC3 (USB controller) driver in their tree even if they don't use it? http://crewrktablets.arctablet.com:8080/opendev/kernel_rockchip/tree/7f63aac7046422b5e1e5aec96af1eccde0e753ca/drivers/usb/dwc3?ModPagespeed=noscript

02:23 <bashintosh> They use the Synopsys driver instead (not part of the mainline kernel) http://crewrktablets.arctablet.com:8080/opendev/kernel_rockchip/tree/7f63aac7046422b5e1e5aec96af1eccde0e753ca/drivers/usb/dwc_otg_310?ModPagespeed=noscript whereas the Raspberry Pi kernel uses DWC2, for exactly the same controller https://github.com/raspberrypi/linux/tree/rpi-4.0.y/drivers/usb/dwc2

03:22 <naobsd> as far as I know, crewrktablets.arctablet.com is not "Rockchip".

03:24 <naobsd> I don't surprise if kernel tree for RPi has thousands of unused drivers...

04:30 <bashintosh> naobsd: basically I am investigating some (apparently) serious issues with the dwc_otg_310 driver shipped with the Rockchip v3.10 kernel, for RK3288 - the CPU is flooded with IRQs when using a USB hub especially and that is something that the RPi community has taken onboard with a complete USB controller driver re-write: https://www.raspberrypi.org/forums/viewtopic.php?t=39175

04:34 <bashintosh> The link from the RPi community shows the same issues that I am experiencing with all RK3288 platforms - keyboards repeating keys, unresponsive USB touch controllers, slow isochronous data transfers etc. The manufacturer of the OTG HS controller (embedded in the RK3288) which is Synopsys, suggests that Rockchip should use DWC2 driver rather than dwc_otg_310. Was thinking to port the RPi driver (or

04:34 <bashintosh> attempt) onto RK3288 and see what happens.

04:34 <amstan> bashintosh: woah? keyboards repeating keys?

04:35 <amstan> bashintosh: that happens on dwc2 as well still?

04:36 <amstan> bashintosh: we use dwc2 on the chromebooks, it has worked pretty well, besides the interrupt storm bothering us battery-life wise

04:36 <bashintosh> amstan: baaaadly... With USB hubs mostly (tried VIA based hubs and Cypress) and input devices

04:36 <bashintosh> amstan: how bad does it affect battery life, overall?

04:36 <amstan> ok, i can try this tomorrow: usb hub+keyboard

04:37 <amstan> well.. it's lots of interrupts, can't be good for the battery

04:37 <amstan> haven't really done power load test with usb plugged in

04:39 <bashintosh> if it doesn't happen with one keyboard, try having 2 USB keyboards - I have seen the problem on Firefly and basically any RK3288 based board - I have been told not to even think about using the OTG controller (even in host mode only) for anything more than a USB stick. The RK3288 has an EHCI controller too but I can't find a board that uses it except for the PopMetal where they use it for SATA.

04:39 <amstan> we use the ehci for the webcam since it has different issues

04:39 <amstan> for keyboards and such

04:40 <amstan> ehci with hub is ideal

04:40 <amstan> though.. i noticed it's a little slower to tftp through ehci+hub

04:41 <bashintosh> Mmmh very interesting! I was reading about issues with the EHCI (driver) too, here http://code.metager.de/source/history/linux/stable/drivers/usb/core/hub.c

04:41 <bashintosh> See Julius Werner comment/patch

04:41 <amstan> ah yes, that was for the chromebooks

04:42 <amstan> julius works in my team

04:42 <bashintosh> amstan: oh! No wonder you're across all that eheh :)

04:43 <amstan> i looked a little at the dwc interrupt storm back in october, but i couldn't find a good fix for it, short of changing the irq to an fiq similar to how the rpi does it

04:43 <amstan> but even then it won't help with interrupts, it'll just make them less cpu hungry

04:44 <amstan> but i wasn't aware that the keyboard issues still remain

04:44 <amstan> i hated that when i first got my rpi

04:44 <bashintosh> Yeah I got quite impressed by what the RPi guys have done https://www.raspberrypi.org/forums/viewtopic.php?t=70437

04:45 <amstan> yep, except none of that was upstream

04:47 <bashintosh> amstan: I tried to flag the problem with Rockchip as well but they basically say that they rely on Synopsys for driver patches etc. - is there a newer kernel version after 3.10 from Rockchip that you know?

04:47 <amstan> the stuff i did in october was with rockchip, and yeah

04:48 <amstan> so.. in 3.10 they used whatever the rpi uses

04:49 <amstan> after that they worked with us(google) to get the chromebooks working, we did this on 3.14

04:49 <amstan> https://chromium.googlesource.com/chromiumos%2Fthird_party%2Fkernel/+/chromeos-3.14

04:51 <amstan> but that's mostly exactly the same as upstream, we've been cherry-picking dwc patches

04:51 <amstan> or at least dianders was doing that last

04:52 <bashintosh> amstan: oooh, thanks for the 3.14 tree!

04:53 <amstan> bashintosh: it doesn't have any dts files for non-chromebook rk3288 devices

04:53 <amstan> since none existed at the time when we forked

04:53 <amstan> except the evb

04:59 <bashintosh> amstan: I might try a quick port unless someone has done it already for Firefly or similar RK3288 based hardware! I do thank you VERY much for pointing me to 3.14... I see the DWC2!

05:00 <amstan> bashintosh: but again, it's the same dwc2 as upstream, i wouldn't bother porting it just to see if it works better

05:00 <amstan> i can tell you that tomorrow when i test a chromebook with a hub and a keyboard

05:03 <bashintosh> OK - the hubs we're using are VIA VL812 based, Cypress HX2 (USB 2.0) and HX3 (USB 3.0) - cascading hubs make things a lot worse, same with multiple HID devices (e.g. keyboard + USB touch controller). All with relation to the OTG controller (host only or dual role), not EHCI AFAIK

05:16 <bashintosh> amstan: would you know how big is the battery on the Chromebook? In mAh?

05:17 <amstan> bashintosh: asus c201 looks like 38WHrs

05:17 <amstan> but that's rated for 13 hours, the others are about 8-10

05:17 <bashintosh> amstan: thanks!!

05:52 <naobsd> bashintosh: then why you asked "why Rockchip has the DWC3 (USB controller) driver in their tree even if they don't use it?"? I think it's totally unrelated to dwc2 issue.

05:54 <naobsd> (I cannot find any reason to remove unused drivers from kernel tree...)

05:55 <naobsd> rock2 uses ehci for external port via hub

05:56 <naobsd> (dwc2 is for OTG and USB-SATA)

07:11 <sjoerd> Is the current state of nand on RK3188 still that it can only be used with a bit of rockchip closed goo ?

07:11 <sjoerd> or well unfortunate

07:26 <naobsd> no nand support on mainline

07:29 <sjoerd> naobsd: and no open solution in sight i guess from my googling around?

07:34 <AstralixNB> naobsd, are you aware of a current git of u-boot, where Simon's mailing list patches are included?

07:35 <naobsd> sjoerd: well, writing new code from scratch might be possible...?

07:35 <naobsd> AstralixNB: no idea

07:38 <AstralixNB> Hmmm... about NAND, the interface and the code is closed. And the documentations leaked on 3188 are not very readable to write a driver from scratch...

07:38 <AstralixNB> That is way more complicated than to scope out some things on a failing rock1 :)

07:39 <sjoerd> naobsd: improving the state for mainline a bit for this project would be fine, doing a nand driver would be too much though :)

07:39 <sjoerd> There is always the option to just use an SD card ofcourse

07:40 <sjoerd> AstralixNB: i didn't see sjg's rockchip patches land in u-boot master just yet

07:41 <naobsd> PX2 TRM should have some info

07:42 <AstralixNB> sjoerd, for NAND / SD, if you would like to use mainline, use a board that provides eMMC.

07:42 <bashintosh> naobsd: basically Synopsys asked me that question - they were expecting DWC2 in the Rockchip default kernel, instead there is DWC3 which I did not know what that was until I went in and read the sources :)

07:43 <sjoerd> AstralixNB: Not my choice of board unfortunately

07:43 <naobsd> bashintosh: as you said, dwc3 is unused. nothing is wrong.

07:44 <AstralixNB> Yeah, unfortunately there was a decision to use NAND instead of eMMC... But you can use SD or hack the RK driver to work with any kernel

07:45 <AstralixNB> And for u-boot, I want to jump in early as I did some work on u-boot. I started with latest HEAD of denx git and tried patching in Simons 54 patches but I have so much missing links...

07:45 <sjoerd> AstralixNB: https//github.com/radxa/u-boot-rockchip.git not good ? (I wanted to start from something known working)

07:46 <bashintosh> naobsd: yes true, the problem (left) is with the dwc_otg_310 driver which I'll attempt to replace with DWC2 for the OTG controller, hoping to solve at least some of the issues with USB performance etc.

07:46 <sjoerd> AstralixNB: iirc sjgs patches were for 3288, not sure if they would work for 3188 as well

07:46 <sjoerd> and for this specific board

07:46 <AstralixNB> I am not talking of good / no-good, just want to go completely mainline.

07:46 <bashintosh> naobsd: ah thanks for the Rock2 tip! Will grab one to test EHCI!

07:46 <AstralixNB> I do have several 3288 devices and I am working on these

07:46 <sjoerd> AstralixNB: ah sorry, i misread you

07:47 <sjoerd> coffee still nees to take effect

07:47 <AstralixNB> Are you from germany? If so, you might think about spending a few euros on postage and send me the rock1 for a health check and possible repair

07:50 <AstralixNB> however, if you like, write me a note on astralix ..at.. crewrktablets .. de

08:07 <sjoerd> AstralixNB: .nl

08:08 <sjoerd> AstralixNB: but i already promised my client to send the dead board back to him :)

08:08 <AstralixNB> Ok

08:08 <sjoerd> AstralixNB: thanks though

08:08 <AstralixNB> welcomed

15:06 <AstralixNB> someone able to help me on u-boot?

23:57 <amstan> bashintosh: keyboard works just fine for me

23:58 <amstan> Bus 002 Device 002: ID 0424:9514 Standard Microsystems Corp. SMC9514 Hub

23:58 <amstan> Bus 002 Device 005: ID 1a40:0101 Terminus Technology Inc. Hub

23:58 <amstan> Bus 002 Device 006: ID 413c:2106 Dell Computer Corp. Dell QuietKey Keyboard