Add support for Allwinner F1C100s family #1220

t123yh · 2021-12-01T08:34:14Z

Allwinner F1Cx00s family is a series of SoCs with Mentor MUSB controller and HS phy. It comes with a slightly different register space layout, and some quirks, so it's not multiplexed with the existing musb support library.

This porting currently requires to be compiled with https://github.com/hongxuyao/F1C100s_with_Keil_RTX4_emWin5 . Support for Keil RTX4 is also added in this pull request.

There is a working CDC loopback demo project at https://github.com/t123yh/F1C100s_RTX4_USB .

ARMCC also sets __GNU__ macro, but doesn't support GCC diagnostic pragmas.

Allwinner F1Cx00s family is a series of SoCs with Mentor MUSB controller and HS phy. It comes with a slightly different register space layout, and some quirks, so it's not multiplexed with the existing musb support library. This library currently requires to be compiled with https://github.com/hongxuyao/F1C100s_with_Keil_RTX4_emWin5

hathach · 2021-12-01T10:42:33Z

Superb !! Thank you very muhc, this is great PR. This would be great option to run TinyUSB on another popular SOC besides rpi broadcom. As required for maintaining the driver, could you please:

Tell which board you are testing with, I am planning to order and Sipeed Lichee Nano to test with your PR.
Add mcu and bsp files so that existing example could run on the actual hardware. This is actually very important, it not only provides proof of concept but also ease the maintaining/developing in the future. You don't have to organize things as other ports/boards. You could just bundle all of the files into your board. Once it is compiled and tested, I will move them into separated mcu submodule later on.
Some short instruction to deploy the firmware on the board would be very helpful since I am not familiar with the chip.

PS: I am ok with Allwinner F1C100s with is own musb driver for now. The existing musb is also newly added. We could later on merge them together in the future (much like synopsys dwc2 driver).

t123yh · 2021-12-01T10:57:38Z

Hi hathach,

Thanks very much for your response!

I'm indeed testing with a Lichee Pi Nano, but it's more convenient to use a Widora Tiny200 v2 (also called MangoPi-R3c). MangoPi-R3c have two USB ports for OTG and UART and comes with dedicated RESET and BOOT button, which brings a lot convenience to development.

Now a basic example is provided here: https://github.com/t123yh/F1C100s_RTX4_USB, which requires Keil to be compiled. I'll try to port it to the BSPs in TinyMCU later.

I forgot to mention that there's currently a bug in the current code: TX and RX cannot share an Endpoint Number. I don't know whether it's a hardware limitation or software bug, though from existing examples the hardware likely supports sharing Endpoint Numbers. This needs to be digged deeper.

hathach · 2021-12-01T11:48:11Z

@t123yh thank you for the tip, I just placed an order for the MangoPi R3 from aliexpress instead of Lichee Pi nano. It probably takes a couple of weeks for it to deliver. Meanwhile we could try to get familiar with the chip and the PR finalized 👍 👍

t123yh · 2021-12-01T12:00:09Z

BTW, currently I have only tried Keil RTX on F1C100s, but I believe RT-Thread will also work. I'll try later.

t123yh · 2021-12-04T10:04:08Z

BSP support is now added. No RTOS is required.

Please see README.md for details.

Currently tested with audio_test and hid_generic_inout and cdc_msc. audio_test is working. cdc_msc is not working, only echoing the first character.

src/portable/sunxi/dcd_sunxi_musb.c

The datasheet says 2KB FIFO, but accroding to many code examples, the F1C100s has at least 4KB of FIFO memory. This is working with cdc_msc example, but I'm not sure, this should be checked.

t123yh · 2021-12-05T14:40:46Z

After fixing a stupid mistake, the cdc_msc example is now working.

t123yh · 2021-12-05T14:51:39Z

Just played with the msc. It seems that there're a few STALLs before the disk is properly detected. Seems that the cause is the code below. I'm not familiar with msc, so I don't know what's happening.

if ( p_msc->stage == MSC_STAGE_STATUS )
  {
    // skip status if epin is currently stalled, will do it when received Clear Stall request
    if ( !usbd_edpt_stalled(rhport,  p_msc->ep_in) )
    {
      if ( (p_cbw->total_bytes > p_msc->xferred_len) && is_data_in(p_cbw->dir) )
      {
        // 6.7 The 13 Cases: case 5 (Hi > Di): STALL before status
        // TU_LOG(MSC_DEBUG, "  SCSI case 5 (Hi > Di): %lu > %lu\r\n", p_cbw->total_bytes, p_msc->xferred_len);
        usbd_edpt_stall(rhport, p_msc->ep_in);
      }else
      {
        TU_ASSERT( send_csw(rhport, p_msc) );
      }
    }

Also I tried to do a cdc benchmark with dd (loopback code commented out). The result is not very impressive (as a High Speed device). Also, after repeating a few times of benchmark, the device just dies. No further benchmark can be made.

[I] root@t123yh-XPS15-9570 ~/t/e/d/cdc_msc (master)# dd if=/dev/zero of=/dev/ttyACM0 count=10000
10000+0 records in
10000+0 records out
5120000 bytes (5.1 MB, 4.9 MiB) copied, 2.37053 s, 2.2 MB/s

hathach · 2021-12-06T13:16:01Z

BSP support is now added. No RTOS is required.

Please see README.md for details.

Currently tested with audio_test and hid_generic_inout and cdc_msc. audio_test is working. cdc_msc is not working, only echoing the first character.

Superb !! That is great, too bad, my tiny200 haven't shipped to me just yet for trying this out. This will make things super easy to support and maintain F1C in the future.

Just played with the msc. It seems that there're a few STALLs before the disk is properly detected. Seems that the cause is the code below. I'm not familiar with msc, so I don't know what's happening.

MSC device may need to stall unsupported command and request from host. It is fairly typical, and required to pass USB compliance test for MSC device #1059 . We may have issue if we haven't fully implement stall/clear stall properly for this port, which is very typical for new chip.

Also I tried to do a cdc benchmark with dd (loopback code commented out). The result is not very impressive (as a High Speed device). Also, after repeating a few times of benchmark, the device just dies. No further benchmark can be made.

It is new port and may need some tweak, I have tested fullspeed device with cdc that could reach 80-90% of the bandwidth. I am pretty sure it is not the stack issue, we just need to tune thing up E.g you could try to increase the CFG_TUD_MSC_EP_BUFSIZE to 4096 to see if that help

https://github.com/hathach/tinyusb/blob/master/examples/device/cdc_msc/src/tusb_config.h#L113

It is a great start, I am looking forward to trying this out.

t123yh · 2021-12-06T13:46:44Z

Now there's one thing that's to be solved: if we want to build a flashable bin file to run on SPI Flash, we must use mksunxi tool to process the output bin file. I'm not sure where can I insert this into our build flow, could you please help me?

hathach · 2021-12-07T05:58:18Z

@t123yh in general, tool should be installed separately by user and added to PATH similarly to gcc toolchain or openocd. You could also make it as makefile variable to pass through command line e.g we can put it in the board/family readme doc.

# default to mksunxi is in PATH
MKSUNXI ?= mksunxi

if user need to, it can be passed to makefile as make MKSUNXI=path-to-tool. Let me know if that make sense to you.

t123yh · 2021-12-07T15:00:27Z

I've written a Python version to avoid the need of installing mksunxi. Now it will build a flashable image that you can write to SPI. README.md is updated to reflect the changes.

hathach · 2021-12-08T03:25:50Z

I've written a Python version to avoid the need of installing mksunxi. Now it will build a flashable image that you can write to SPI. README.md is updated to reflect the changes.

great works, do you have a need to merge it asap ? my board from aliexpress could take 2 more weeks to arrive. I would like to have hand-on testing and could make some cosmetic changes if needed. But we could merge it earlier if this blocks your work.

t123yh · 2021-12-08T05:49:04Z

great works, do you have a need to merge it asap ?

No need, it's better to merge after you have tested the whole thing. I don't think it's currently stable enough to be merged. I can just work on my own repo before merging.

BTW, after fixing a bug (commit 527036b) that causes MMU and cache not initialized when executing from FEL, serial performance test reaches 15.7MB/s, which is a lot better than previous tests.

kkitayam · 2021-12-11T15:10:37Z

Make some adjustments to process_ep0 so that if _dcd.status_out is 1 when entering process_ep0 -> USB_CSRL0_RXRDY, send a dcd_event_xfer_complete immediately to notify the stack;

I guess this way is good. I would consider whether the way works for other conditions.

If this adjustments are applied, after the step 24 of C) The next SETUP request comes before TUD calls dcd_xfer() for current STATUS OUT , DCD will enqueue DCD_EVENT_XFER_COMPLETE to event queue. Therefore DCD generates event before TUD task calls dcd_xfer(). It is strange behavior, however, it might not be a problem as TUD task will call dcd_xfer() for STATUS OUT later.

@hathach
Is it acceptable that DCD generates DCD_EVENT_XFER_COMPLETE before TUD task calls dcd_edpts_xfer() for STATUS OUT?

If this is not acceptable, I would consider whether this condition can be avoided by adding a new flag.
Or, if this is acceptable, I would modify that the step 20 of Condition B) and the step 20 of Condition C) generate DCD_EVENT_XFER_COMPLETE for consistency.

hathach · 2021-12-14T09:20:54Z

@kkitayam the diagram is super useful, though I haven't tried to look closely at musb so far. Please give me a bit of time to catch up with this.

t123yh · 2021-12-27T00:51:48Z

@kkitayam the diagram is super useful, though I haven't tried to look closely at musb so far. Please give me a bit of time to catch up with this.

Hi hathach,

Just a friendly reminder on this issue. Could you please take some time to figure this out? This has been a blocking bug for me for a while. BTW, have you received your Tiny200 board?

Thanks!

hathach · 2021-12-27T09:16:20Z

@t123yh thanks for the reminder, I did indeed receive tiny200 board a few days ago. I have been busy with other works. Please give me a bit more time to get it setup and trying this out.

…h-master

hathach

PR looks great, thank you very much for your effort. I only have one question regarding the keil warning suppression of "statement is unreachable"

hathach · 2021-12-29T12:33:21Z

src/device/usbd.c

@@ -44,6 +44,11 @@
  #define CFG_TUD_TASK_QUEUE_SZ   16
 #endif

+#ifdef __ARMCC_VERSION
+// Supress "statement is unreachable" warning
+#pragma diag_suppress 111


I don't have keil to try this out. Maybe it catches an actual valid warnings that worth fixing. Would you tell more on this, which specific line that this

The warnings are for the switch(desc_type) inside process_get_descriptor in which there're several break; that is unreachable because are return statements in both if and else block. Line 1016, 1033, 1059, 1072, 1086.

thank you I see, we could just remove the break; they are mostly are there for case/break pair.

PS: just push an update to fix the warning. Let me know if that would fix your compiling with keil

hathach

I have tried it on my tiny200, and it works great. Thank you very much for the excellent PR. I did a bit of cosmetic change e.g adding license term to dcd file, move hw/mcu/allwinner to its own submodule etc.. Once the ci is passed, we could merge this

t123yh · 2021-12-30T14:46:09Z

Hi @hathach ,

Thank you for testing. The bug mentioned above still exists, and I believe it needs to be resolved before merging. I have an example to reproduce this issue. I'll upload it to a branch soon.

hathach · 2021-12-30T14:48:48Z

@t123yh no problem at all, I am downloading keil trial version on my VM windows, will try it out if you have any ready-to-build example

t123yh · 2021-12-30T15:00:23Z

@t123yh no problem at all, I am downloading keil trial version on my VM windows, will try it out if you have any ready-to-build example

No you don't have to download keil. You can run my example totally within the tinyusb example framework. The code to reproduce the problem is based on gud-pico which is part of the gud project. You need to have RPi with the latest RPi OS (which backported gud to 5.10 kernel) or PC with Arch Linux to run my example.

t123yh · 2021-12-30T15:22:16Z

@hathach The example is ready. You can downloaded it from gud-test branch. This is a dirty quick version based on earlier version of my port, but this shouldn't matter. It has my workaround commit for the problem discussed above with kkitayam applied. You have to prepare a Linux host running kernel with gud kernel module. Latest RPi OS have this module backported.

To test the example, you can connect the USB device port to the host Linux device, and the USB-UART port to your serial monitor on PC (note that the serial port of F1C100s is changed to UART1 to adapt to Tiny200v2), then compile and flash the gud code just like other examples. If you run my code directly, it should be recognized immediately on RPi as a USB screen without out any errors (although it's sorry that you probably doesn't have a screen on your Tiny200v2, but it shouldn't matter), with the following logs on your serial monitor:

Boot to SPI mode
Timer INIT done
tud_mount_cb:
gud_req_get: request=0x1 index=0 size=30
XFER 30-30!!
gud_req_get: request=0x40 index=0 size=32
XFER 32-1!!
gud_req_get: request=0x41 index=0 size=128
gud_req_get: request=0x50 index=0 size=128
DONE5
XFER 128-5!!
Drop CONTROL_STAGE_ACK
gud_req_get: request=0x51 index=0 size=128
XFER 128-10!!
gud_req_get: request=0x54 index=0 size=1
XFER 1-1!!
gud_req_get: request=0x56 index=0 size=128
XFER 128-128!!
gud_req_get: request=0x55 index=0 size=128
XFER 128-24!!
Drop CONTROL_STAGE_ACK
controller_enable: enable=1
state_commit: mode=320x480 format=0x80 connector=0 num_properties=1
  prop=12 val=100
display_enable: enable=1
Set backlight: 100

Note the CONTROL_STAGE_ACK is where the problem occurs, but workarounded. If you revert my workaround commit (8656b3b), the whole thing will stop on the first Drop CONTROL_STAGE_ACK, and RPi will report an error in its dmesg.

Please tell me if you can reproduce this problem.

hathach · 2021-12-30T15:47:46Z

Thank you for the detail info, I will grab an Pi4 and try to reproduce and look into this issue. To be honest, I haven't read the musb specs/code yet, thankfully, kkitayam already does an great job illustrated it.

t123yh · 2021-12-30T16:06:42Z

musb doc can be found by Google mentor usb sunxi. Also you may refer to the Linux musb driver. Specifically, musb quirks for sunxi devices can be found on sunxi.c.

hathach · 2021-12-30T18:21:22Z

yeah, I mean I haven't got time to really look at musb yet. So far all the great work is done by @kkitayam :)

hathach · 2022-01-25T15:37:42Z

@t123yh sorry I am lacking behind PRs/issues. I still haven't managed time to look at musb and test the above problem just yet. This PR is already pending for a while and is relatively large. I think we should merge this as it is for now, since it has good running example already. And then do another PR to fix the problem, we could create an github ticket/issue for discussion. That would also minimize effort for other contributors to fix the issue as well. What do you say ?

t123yh · 2022-01-26T01:02:09Z

@t123yh sorry I am lacking behind PRs/issues. I still haven't managed time to look at musb and test the above problem just yet. This PR is already pending for a while and is relatively large. I think we should merge this as it is for now, since it has good running example already. And then do another PR to fix the problem, we could create an github ticket/issue for discussion. That would also minimize effort for other contributors to fix the issue as well. What do you say ?

Yes, I agree with you. Thanks for your time and effort.

hathach · 2022-01-26T05:52:19Z

Yes, I agree with you. Thanks for your time and effort.

Thanks, I wish I could be more involved with this port, but the Lunar New Year is coming and there is lots of other works/project. I will look at this later on when I have time. Thank you @t123yh for your contribution. I will make an new issue as reminder for us after the merge

t123yh · 2022-01-26T06:08:39Z

the Lunar New Year is coming and there is lots of other works/project

I totally understand this as a Chinese. Happy New Year!

hathach · 2022-01-28T05:02:18Z

thanks, Happy New Year to you as well !!! 🎉

t123yh added 4 commits November 30, 2021 17:19

usbd: supress "statement is unreachable" warning on ARMCC

dba5921

Ignore certain compiler options on ARMCC

ac20605

ARMCC also sets __GNU__ macro, but doesn't support GCC diagnostic pragmas.

Add support for Keil RTX4 OS abstraction layer

a6929cf

t123yh added 2 commits December 4, 2021 17:18

Modify sunxi_musb code

dff54d8

Add BSP support for F1C100s

68ca62d

t123yh added 4 commits December 4, 2021 18:08

Add TODO to README

28fb51c

Save current EP before querying other EPs

fa0e4d9

Close all EPs upon reset

64b81fd

Add FIFO size check

702698e

kkitayam reviewed Dec 5, 2021

View reviewed changes

src/portable/sunxi/dcd_sunxi_musb.c Show resolved Hide resolved

t123yh added 2 commits December 5, 2021 22:31

Fix a typo that leads to incorrect RX handling

f308603

Change FIFO size to 4KB (not sure)

1ffc366

The datasheet says 2KB FIFO, but accroding to many code examples, the F1C100s has at least 4KB of FIFO memory. This is working with cdc_msc example, but I'm not sure, this should be checked.

Fix handling of RXRDY bit

96979a2

t123yh added 2 commits December 7, 2021 22:53

Add mksunxi tool to make flashable image

75f7fb3

Update README

43621ab

Initialize MMU if we are executing from DDR

527036b

hathach added 5 commits December 29, 2021 18:27

Merge branch 'master' of https://github.com/t123yh/tinyusb into t123y…

30aba24

…h-master

add allwinnner driver as submodule

2d13ecd

update allwinner submodule

b943a53

skip freertos example for f1c100s

84e2f6e

add license to dcd file, slightly update readme

5f280b3

hathach changed the base branch from master to port-embos December 29, 2021 12:27

hathach changed the base branch from port-embos to master December 29, 2021 12:28

update docs

adb848f

hathach reviewed Dec 30, 2021

View reviewed changes

fix unreachable warnings with keil

c5d2c82

hathach approved these changes Dec 30, 2021

View reviewed changes

t123yh requested a review from kkitayam January 2, 2022 01:04

Merge branch 'master' into master

7de1663

hathach merged commit eea19da into hathach:master Jan 26, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add support for Allwinner F1C100s family #1220

Add support for Allwinner F1C100s family #1220

t123yh commented Dec 1, 2021

hathach commented Dec 1, 2021 •

edited

Loading

t123yh commented Dec 1, 2021 •

edited

Loading

hathach commented Dec 1, 2021

t123yh commented Dec 1, 2021

t123yh commented Dec 4, 2021 •

edited

Loading

t123yh commented Dec 5, 2021

t123yh commented Dec 5, 2021

hathach commented Dec 6, 2021 •

edited

Loading

t123yh commented Dec 6, 2021

hathach commented Dec 7, 2021

t123yh commented Dec 7, 2021

hathach commented Dec 8, 2021

t123yh commented Dec 8, 2021

kkitayam commented Dec 11, 2021

hathach commented Dec 14, 2021

t123yh commented Dec 27, 2021

hathach commented Dec 27, 2021

hathach left a comment

hathach Dec 29, 2021

t123yh Dec 30, 2021

hathach Dec 30, 2021 •

edited

Loading

hathach left a comment

t123yh commented Dec 30, 2021

hathach commented Dec 30, 2021

t123yh commented Dec 30, 2021

t123yh commented Dec 30, 2021 •

edited

Loading

hathach commented Dec 30, 2021

t123yh commented Dec 30, 2021 •

edited

Loading

hathach commented Dec 30, 2021

hathach commented Jan 25, 2022

t123yh commented Jan 26, 2022

hathach commented Jan 26, 2022

t123yh commented Jan 26, 2022

hathach commented Jan 28, 2022

Add support for Allwinner F1C100s family #1220

Add support for Allwinner F1C100s family #1220

Conversation

t123yh commented Dec 1, 2021

hathach commented Dec 1, 2021 • edited Loading

t123yh commented Dec 1, 2021 • edited Loading

hathach commented Dec 1, 2021

t123yh commented Dec 1, 2021

t123yh commented Dec 4, 2021 • edited Loading

t123yh commented Dec 5, 2021

t123yh commented Dec 5, 2021

hathach commented Dec 6, 2021 • edited Loading

t123yh commented Dec 6, 2021

hathach commented Dec 7, 2021

t123yh commented Dec 7, 2021

hathach commented Dec 8, 2021

t123yh commented Dec 8, 2021

kkitayam commented Dec 11, 2021

hathach commented Dec 14, 2021

t123yh commented Dec 27, 2021

hathach commented Dec 27, 2021

hathach left a comment

Choose a reason for hiding this comment

hathach Dec 29, 2021

Choose a reason for hiding this comment

t123yh Dec 30, 2021

Choose a reason for hiding this comment

hathach Dec 30, 2021 • edited Loading

Choose a reason for hiding this comment

hathach left a comment

Choose a reason for hiding this comment

t123yh commented Dec 30, 2021

hathach commented Dec 30, 2021

t123yh commented Dec 30, 2021

t123yh commented Dec 30, 2021 • edited Loading

hathach commented Dec 30, 2021

t123yh commented Dec 30, 2021 • edited Loading

hathach commented Dec 30, 2021

hathach commented Jan 25, 2022

t123yh commented Jan 26, 2022

hathach commented Jan 26, 2022

t123yh commented Jan 26, 2022

hathach commented Jan 28, 2022

hathach commented Dec 1, 2021 •

edited

Loading

t123yh commented Dec 1, 2021 •

edited

Loading

t123yh commented Dec 4, 2021 •

edited

Loading

hathach commented Dec 6, 2021 •

edited

Loading

hathach Dec 30, 2021 •

edited

Loading

t123yh commented Dec 30, 2021 •

edited

Loading

t123yh commented Dec 30, 2021 •

edited

Loading