Fix Bootloader not Flashing

[BogdanTheGeek]

Issue:

Found B003Fun Bootloader
Halting Boot Countdown
Interface Setup
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
ioctl (SFEATURE): Protocol error
Warning: Issue with hid_send_feature_report. Retrying
Error: Flash operation error
Error: Failed to erase sector at 08000000
Error writing block at memory 08000000 (error = -9)
Error: Fault writing image.
make: *** [ch32v003fun/ch32v003fun/ch32v003fun.mk:188: cv_flash] Error 243

[xsrf]

This fix adds 20 bytes to the bootloader, rendering even more of the boot configurations unusable. Can you explain when you encountered the issue and why this fix fixes it? I've never had issues with the bootloader not flashing on windows and linux.

[BogdanTheGeek]

Can you explain when you encountered the issue and why this fix fixes it?

@cnlohr troubleshooted the issue on discord last Saturday and wrote the fix. From what I understood and remember, because flash writes are blocking, the device sent the response at the wrong time. He mentioned that most usb hubs are not as fussy, but that this fix is the "correct" way of handling the blocking writes. I hope I am not misrepresenting his view. I can share the usb traces if that helps.

This fix adds 20 bytes to the bootloader

My compiler seems to be a bit more aggressive with optimisations. Its only 16 bytes larger maste:

riscv64-unknown-elf-size bootloader.elf
   text    data     bss     dec     hex filename
   1884       0      92    1976     7b8 bootloader.elf

fix:

riscv64-unknown-elf-size bootloader.elf
   text    data     bss     dec     hex filename
   1896       0      96    1992     7c8 bootloader.elf

[xsrf]

Ah okay, thx. Guess that's a pill we have to take then. Yours is only +12 Bytes btw ;) For me (on windows) master actually compiles 8 Bytes less but with the PR it's also 1896.

> make build
riscv64-unknown-elf-gcc -o bootloader.elf  bootloader.c ../rv003usb/rv003usb.S  -g -Os -flto -ffunction-sections -fdata-sections -fmessage-length=0 -msmall-data-limit=8 -march=rv32ec 
-mabi=ilp32e -DCH32V003=1 -static-libgcc -I/usr/include/newlib -I../ch32v003fun/ch32v003fun/../extralibs -I../ch32v003fun/ch32v003fun -nostdlib -I. -Wall -I. -I../lib -DUSE_TINY_BOOT 
-I../rv003usb -L../ch32v003fun/ch32v003fun/../misc -lgcc -T ch32v003fun-usb-bootloader.ld -Wl,--gc-sections
riscv64-unknown-elf-size bootloader.elf
   text    data     bss     dec     hex filename
   1876       0      92    1968     7b0 bootloader.elf

[BogdanTheGeek]

Yours is only +12 Bytes btw ;)

I was looking at dec not text. bss also increase by 4 bytes because of the new global.

EDIT: an no, making it a u8 or u16 didn't reduce it, i guess its pretty well optimised.

[xsrf]

The bss is not relevant. The text needs to be 1920 or less.

[cnlohr]

Circling back, @xsrf is this something that should be merged?

[cnlohr]

Tested, good now! Just was weird because of some slightly incorrect macro logic.