fdt: increase supported image number to MAX_NUM_IMGS, catch overflows#4
fdt: increase supported image number to MAX_NUM_IMGS, catch overflows#4beku-epitome wants to merge 1 commit intonxp-imx-support:masterfrom
Conversation
…errors The previous g_images buffer would be overflown if e.g. TEE was included in the FIT image, causing other globals to be overwritten, and eventually segfaulting the program. 1. Increase the g_images buffer size 2. Pass the buffer size to the parser, and limit the parsing to that limit Signed-off-by: Benedek Kupper <benedek.kupper@epitome.inc>
|
Hello, thank you for the PR and apologies for the delay. We are reviewing the fix and will get back to you soon. |
|
@beku-epitome Could you explain in which case we need more than 5 images? typically we can have uboot proper, uboot dtb, atf and tee. Anything else that you have included in your FIT image that needs this extension? |
|
Sorry for the late reply, the company I worked at went under. I don't remember all the details anymore, but as a recap, we had a Variscite DART-MX8M-PLUS SoM, and we used optee. Variscite's Wiki suggests the old approach to HAB, possibly because they also couldn't get this tool to work correctly. I implemented HAB for the i.MX 8M Plus EVK first, where this tool works out of the box. But when targeting the SoM, there was a segmentation error in This bug, especially the lack of bounds checking is the likely reason why the many different yocto BSPs didn't converge to the optimal solution of using |
3 participants
The previous g_images buffer would be overflown if e.g. TEE was included in the FIT image, causing other globals to be overwritten, and eventually segfaulting the program.