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

bootutil: Add support for devices without erase #2114

Open
wants to merge 5 commits into
base: main
Choose a base branch
from

Conversation

de-nordic
Copy link
Collaborator

@de-nordic de-nordic commented Nov 11, 2024

Commits:

  • bootutil: add support for not calling erase when device does not require it and MCUBOOT_DEV_WITHOUT_ERASE_THEN_NO_ERASE option that controls this; add MCUBOOT_MINIMAL_SCRAMBLE that allows to reduce amount of removed data, where applicable;
  • Zephyr: Kconfig options and supporting code
  • Zephyr: switch to flash_area_flatten in bs_custom_storage_erase
    - Zephyr: switch to boot_set_next, from flash_area_erase, for Zephyr builds of bootutil_public.

The first MCUboot option allows to improve life time of devices for which erase is emulated by write, and is not really required by the device design and the second MCUboot option allows to just remove enough of data, where needed, by either reduced range of erase or reduced write; for example by scrambling only image header, rather than removing the entire image.

Tested on nrf52840dk and nrf54l15. In case of nrf54l it reduces swap time by half, in comparison to code using erase.

Note: Even though, finally, swap_erase_trailer_sectors is no longer used, I have left it there because it may be used in the future.
I have left a Note around swap_status_init function, that partially corresponds to this; it seems that in many places call to swap_scramble_trailer_sectors, whee followed by swap_status_init, should in reality be swap_erase_trailer_sectors fulfilling hardware requirement for preparing device for write; the problem is that swap_status_init initializes only part of swap information, which means that if the region of entire swap info is not scrambled first, leftovers may lead to improper behaviour, for example test will not revert.
There is re-design of swap_status_init required to mitigate this, as in some paths the whole "init" may just mean removal of everything except information that swap_status_init would put there anyway, or combined init operation that would first remove no longer needed information than overwrite what should be re-initialized - but this requires further analysis.

So far this PR works fine on devices that do not require erase as well as these that require erase before write.

@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch 15 times, most recently from cf8def0 to 03ac085 Compare November 15, 2024 12:22
@de-nordic de-nordic marked this pull request as ready for review November 15, 2024 13:48
@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch from 03ac085 to b96fe4a Compare November 15, 2024 14:42
int boot_erase_region(const struct flash_area *fap, uint32_t off, uint32_t sz);
/* SImilar to boot_erase_region but will alwasy remove data */
Copy link
Collaborator

Choose a reason for hiding this comment

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

-I+i, *always

Comment on lines 1228 to 1235
uint8_t buf[BOOT_MAX_ALIGN];
size_t size_done = 0;

memset(buf, flash_area_erased_val(fa), sizeof(buf));

while (size_done < size) {
ret = flash_area_write(fa, size_done + off, buf, sizeof(buf));
if (ret != 0) {
break;
}
size_done += sizeof(buf);
}
Copy link
Collaborator

Choose a reason for hiding this comment

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

this just assumes max align will not overflow the size but makes no checks to ensure that, it should reduces buffer on final write to ensure it does not overflow

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Yes, that was intentional as I have assumes that MCUboot should write at BOOT_MAX_ALIGN pace anyway, but maybe that should be fixed.

} else {
size = MAX(sizeof(((struct image_header *)0)->ih_magic),
BOOT_MAX_ALIGN);
size = (size + BOOT_MAX_ALIGN - 1) & ~(BOOT_MAX_ALIGN - 1);
Copy link
Collaborator

Choose a reason for hiding this comment

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

This should do the same for footer data because it might contain swap status or mcuboot flags for swapping images and those should always be cleared

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Ah right, header and slot trailer are verified separately.

Comment on lines +856 to +873
The same can be achieved with just removal of header, leaving the
rest of image untouched, as without header MCUboot will not be able
to recognize image in slot as bootable.
Copy link
Collaborator

Choose a reason for hiding this comment

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

include footer

Comment on lines 551 to 553
#if !defined(CONFIG_MCUBOOT) && defined(__ZEPHYR__)
flash_area_flatten(fa, 0, flash_area_get_size(fa));
#else
Copy link
Collaborator

Choose a reason for hiding this comment

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

think it was mentioned in a previous PR that OS specific code should not be present in here

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

I can fix that although MCUboot has significant problem where we are coupled with Flash Map API which means that this has to be replaced with a function that is implemented for every supported system. I will figure way around.

@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch from b96fe4a to 269f9e2 Compare December 10, 2024 19:23
@de-nordic de-nordic marked this pull request as draft December 10, 2024 19:23
@de-nordic
Copy link
Collaborator Author

Converting to draft; after significant changes I need to re-test this on devices.

@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch 4 times, most recently from 75494c5 to 5a405e7 Compare December 18, 2024 13:47
@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch 7 times, most recently from d2fecc0 to 693327f Compare December 19, 2024 13:36
The commit adds two MCUboot configuration options:
  - MCUBOOT_SUPPORT_DEV_WITHOUT_ERASE
  - MCUBOOT_SUPPORT_DEV_WITH_ERASE
  - MCUBOOT_MINIMAL_SCRAMBLE

The first one should be enabled to support devices that do not require erase.
When such devices are used in system then MCUboot will avoid erasing such
device, which is not needed by hardware, and will just write data to it.
This allows to both improve device lifetime and reduce time of operations
like swap.

The second option is just bringing a configuration option for already existing
support for deviceses with erase.

The third option allows to reduce amount of removed data. When enabled,
MCUboot will remove enough of data, depending on the purpose of the removal,
to just fulfill the purpose; for example if removal of data is done to
make image unrecognizable for MCUboot, with this option, it will only
remove header.

Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch from 693327f to cec59f0 Compare December 19, 2024 14:56
@de-nordic de-nordic marked this pull request as ready for review December 19, 2024 15:05
Add Kconfig options:
  - CONFIG_MCUBOOT_STORAGE_WITHOUT_ERASE that enables MCUboot
    configuration MCUBOOT_SUPPORT_DEV_WITHOUT_ERASE
  - CONFIG_MCUBOOT_STORAGE_WITH_ERASE that enables MCUboot
    configuration MCUBOOT_SUPPORT_DEV_WITH_ERASE
  - CONFIG_MCUBOOT_STORAGE_MINIMAL_SCRAMBLE that enables MCUboot
    configuration MCUBOOT_MINIMAL_SCRAMBLE

Adds implementation of flash_area_erase_required, which is required when
MCUBOOT_STORAGE_DEV_WITHOUT_ERASE is enabled.

Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
The intention of bs_custom_storage_erase is to remove data from device;
to support devices that do not require erase, without calling erase,
so that devices that do not implement such functions could work,
the flash_area_erase has been replaced with flash_area_flatten.

Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
By default enable all other systems to work with devices that
require erase prior to write.

Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
Getter for sector size stored in flash_sector object.

Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
@de-nordic de-nordic force-pushed the improve-no-erase-device-time branch from cec59f0 to 8b23727 Compare December 19, 2024 15:20
@de-nordic de-nordic self-assigned this Dec 19, 2024
@de-nordic de-nordic added area: zephyr Affects the Zephyr port area: core Affects core functionality labels Dec 19, 2024
@nvlsianpu nvlsianpu self-requested a review December 20, 2024 13:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area: core Affects core functionality area: zephyr Affects the Zephyr port
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants