Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add Wi-Fi firmware partition support for Pico 2 W #1969

Open
wants to merge 19 commits into
base: develop
Choose a base branch
from

Conversation

will-v-pi
Copy link
Contributor

This adds the ability to store and load the Wi-Fi firmware for Pico 2 W in a partition. It can be enabled by adding pico_use_partition_firmware(<exe_name>) to your CMakeLists.txt, which will embed a compatible partition table in the binary, and output lots of firmware UF2s to use (all called <exe_name>_firmware_..._.uf2, eg ..._firmware_w for Wi-Fi only, ..._firmware_wb for Wi-Fi and Bluetooth, ..._firmware_w_tbyb for TBYB). You can also create your own partition table and use that.

A Wi-Fi firmware partition is detected as having the ID 0x123456789abcdef0, and the UF2 family_is for Wi-Fi firmware blobs is 0x12345678 - these should probably both be changed to something else before merging? The default firmware partition starts at 3500K into the flash - should this be changed to depend on PICO_FLASH_SIZE_BYTES? The default firmware partition is also duplicated with A/B partitions in the same location in flash - this is required to ensure a signature check is performed before loading the Wi-Fi firmware, as there's no way to call the bootrom to check the signature of a single partition (unless chaining into it), you can only call pick_ab_partition.

The Wi-Fi firmware blob is marked in it's image_def as an RP2350 Risc-V executable, and the partition is marked as ignored_during_riscv_boot - this ensures that it can work with TBYB (as TBYB only works for executable image_defs), and that signature checks are performed before loading the firmware when Secure Boot is enabled (because signature checks are performed for all executable image_defs in a partition that is not marked ignored_during_arm_boot). This is slightly clunky, but seems to work robustly.

Supercedes #1850, as it now includes that function in this PR, and shows a use for it.

@lurch
Copy link
Contributor

lurch commented Oct 3, 2024

pinging @peterharperuk as he did a lot of the Wifi-related stuff for Pico 1.

@will-v-pi will-v-pi marked this pull request as ready for review October 17, 2024 09:01
@will-v-pi will-v-pi added this to the 2.0.1 milestone Oct 30, 2024
@peterharperuk
Copy link
Contributor

Works nicely. I think we need to fix the Pico W (rp2040) build errors in src/rp2_common/pico_cyw43_driver/cyw43_driver.c
It would be nice if it gave an error if you're a numpty like me and put pico_use_partition_firmware after pico_add_extra_outputs, but that could be improved later?

@will-v-pi
Copy link
Contributor Author

It would be nice if it gave an error if you're a numpty like me and put pico_use_partition_firmware after pico_add_extra_outputs, but that could be improved later?

Have added in separate PR #2054

@will-v-pi
Copy link
Contributor Author

I think we need to fix the Pico W (rp2040) build errors in src/rp2_common/pico_cyw43_driver/cyw43_driver.c

I've added a check that it's not RP2040 in the CMake function, so it'll throw a fatal error at that point if you try to build a binary with pico_use_partition_firmware for RP2040

@armandomontanez
Copy link
Contributor

armandomontanez commented Nov 19, 2024

If the Bazel checks are getting in your way, you can add the two new files to a filegroup (which will silence the error) and file an issue+leave a TODO to support this in Bazel and assign it to me.

Copy link
Contributor

@peterharperuk peterharperuk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Tested this and it seems good.

peterharperuk
peterharperuk previously approved these changes Nov 19, 2024
@will-v-pi
Copy link
Contributor Author

If the Bazel checks are getting in your way, you can add the two new files to a filegroup (which will silence the error) and file an issue+leave a TODO to support this in Bazel and assign it to me.

Thanks, have done - I couldn't actually assign the issue to you, but have tagged you in it

@lurch
Copy link
Contributor

lurch commented Nov 19, 2024

A Wi-Fi firmware partition is detected as having the ID 0x123456789abcdef0, and the UF2 family_is for Wi-Fi firmware blobs is 0x12345678 - these should probably both be changed to something else before merging?

Is that still going to be happening? I guess choosing randomly-generated IDs is much less likely to produce an accidental collision than a "nice" ID like 0x123456789abcdef0 ? And should those IDs be added to a header-file, to allow other UF2s to make use of this Wi-Fi firmware partition?

@will-v-pi
Copy link
Contributor Author

A Wi-Fi firmware partition is detected as having the ID 0x123456789abcdef0, and the UF2 family_is for Wi-Fi firmware blobs is 0x12345678 - these should probably both be changed to something else before merging?

Is that still going to be happening? I guess choosing randomly-generated IDs is much less likely to produce an accidental collision than a "nice" ID like 0x123456789abcdef0 ? And should those IDs be added to a header-file, to allow other UF2s to make use of this Wi-Fi firmware partition?

Yes - I was looking for ideas. We could go with 0x776966696669726d (hex for wififirm) for the partition ID, and just use the data family_id we already have?

Comment on lines 1078 to 1080
* It requires the same minimum workarea size as `rom_pick_ab_partition`.
*
* Also checks that the chosen partition contained a valid image
* For example, if an `explicit_buy` is pending then calling `pick_ab_partition` would normally clear the saved `flash_erase_addr` so the required erase would not
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This doxygen mentions both rom_pick_ab_partition and pick_ab_partition - are they referring to the same function or are they different?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

rom_… is the SDK function wrapper for the pick_ab_partition function - so I was using that to refer to it in the SDK (ie the function this calls internally, and where more docs are), but using pick_ab_partition to refer to the behaviour of the underlying bootrom function

@lurch
Copy link
Contributor

lurch commented Nov 20, 2024

Yes - I was looking for ideas. We could go with 0x776966696669726d (hex for wififirm) for the partition ID, and just use the data family_id we already have?

I've got no opinions on this so it's probably something that you and @peterharperuk ought to decide. Would MicroPython be able to use the same Wifi firmware partition as C-SDK code? (or is that another one of my stupidly naive questions? 😆 )

// printf("IMAGE_DEF_TBYB_FLAGGED %08x\n", (uint32_t)&(scan_workarea->parsed_block_loops[0].image_def.core.tbyb_flagged) - (uint32_t)scan_workarea);
// printf("IMAGE_DEF_BASE %08x\n", (uint32_t)&(scan_workarea->parsed_block_loops[0].image_def.core.enclosing_window.base) - (uint32_t)scan_workarea);
// printf("IMAGE_DEF_REL_BLOCK_OFFSET %08x\n", (uint32_t)&(scan_workarea->parsed_block_loops[0].image_def.core.window_rel_block_offset) - (uint32_t)scan_workarea);
#define VERSION_DOWNGRADE_ERASE_ADDR *(uint32_t*)0x400e0338
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

arguably these are not addresses. i would leave the leading * off and use in code references

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Technically they are addresses, eg the version downgrade erase address (ie the address in flash to erase) is stored at 0x400e0338

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Page 31 of the RP2350 datasheet says BOOTRAM_BASE 0x400e0000

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yeah, more of a stylistic thing I guess; it's weird to see a CONSTANT as a lhs

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

but i'm ok

@@ -1071,6 +1071,29 @@ static inline int rom_get_last_boot_type(void) {
*/
int rom_add_flash_runtime_partition(uint32_t start_offset, uint32_t size, uint32_t permissions);

/*! \brief Pick A/B partition with TBYB guards
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I still find the function description a bit confusing:

  • maybe something along the lines of "Pick A/B partition for an arbitrary pair of partitions without disturbing any in process update A/B or TBYB boot" (needs improvement still)...
  • not sure people know what a flash_erase_addr is maybe describe it in words (e.g. the erase of xxx because yyy"
  • I'd say "using the flash_update_boot_window_base from the current boot"

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think "Pick A/B partition without disturbing any in progress update or TBYB boot" works for the description - pick_ab_partition is always for an arbitrary pair of partitions, so that probably doesn't need including in the description?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, I was (wrongly) thinking this new API existed because pick_ab_partition was for the regular boot partitions, rather than because that API is somewhat broken. Perhaps just add a warning to the other API referencing this one

Enabled using the pico_use_partition_firmware(exe_name) cmake function
RP2040 does not support partitions, so throw fatal_error at the CMake stage
… family

Adds CYW43_WIFI_FW_PARTITION_ID define to override the partition ID
Add copyright headers

Put ptinfo buffer on the stack, and break when partition is found

Improve function description
RCP mask constants still need to be accessed from Risc-V
@will-v-pi
Copy link
Contributor Author

I've noticed an issue with the Risc-V build of the firmware UF2 - for some reason the compilation is putting a programme header in the ELF with offset of 0x0000 in the file loading to 0x1ffff000, when the Arm compiler correctly puts this at offset 0x1000 in the file loading to 0x20000000. @Wren6991 or @kilograham any ideas why the Arm and Risc-V compilers might be compiling the wifi_firmware.S file differently?

… CYW43_WIFI_FW_PARTITION_ID->CYW43_FIRMWARE_PARTITION_ID
The Risc-V compiler doesn't link the firmware blobs as expected, so use Arm ELFs instead
@will-v-pi
Copy link
Contributor Author

I've now got it working for Risc-V by storing copies of the firmware ELFs in the SDK, and then copying those into the build when necessary. It also needs increased bootrom stack size. This is tested and working fine.

However, I'm not sure if this is actually a solution that we want to put in the 2.1.0 SDK release, so it might be best if we don't merge this for 2.1.0 and I can investigate further to fix the Risc-V firmware build?

@will-v-pi will-v-pi modified the milestones: 2.1.0, 2.2.0 Nov 22, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants