On 31/10/2011 17:23, Simon Schwarz wrote:
Adds direct Linux boot to SPL. It implements a spl export command to save ATAGS or FDT to NAND flash. The kernel image has to be in place for this!
Changes in V5:
- Rebased on u-boot-ti
- fixed MAKEALL warnings and errors
- adapted to general gpio interface
Changes in V6:
- Change old commit message
Changes in V7:
- Correct style and format errors
based on:
- The new SPL layout
- OMAP3 new SPL layout (http://article.gmane.org/gmane.comp.boot-loaders.u-boot/105260)
- CONFIG_MACH_TYPE fix (http://article.gmane.org/gmane.comp.boot-loaders.u-boot/105809)
- Prep subcommand patch for arm (http://article.gmane.org/gmane.comp.boot-loaders.u-boot/106725)
Related to:
Hi Simon,
I am trying your patches on the Beagleboard - really I want to get a feeling how it is possible to introduce this mechanism for other architectures, too.
I have the major issue when the SPL gives the control to the kernel ( Linux-3.2.0-rc4-00154-g8e8da02). The code is started, however I get the error "uncompression error" in the gunzip() function of the kernel - really at the beginning of the code.
I have already checked the boot parameters area at address 0x80000100 created by spl export, and it is the same as the parameters generated by bootm under u-boot (where, of course, everything is working fine).
In jump_to_image_linux() all parameters are correct and image_entry points to the load address read from the uImage header (0x80008000 for TI kernel).
The most obvious thing could be that the Kernel image is not copied correctly from NAND to RAM, but checking some parts of memory I have not found any discrepancy. Of course, I must check better this point, but it seems to me not so probable that there is an issue with the NAND.
I have also added the cache functions to invalidate I-Cache to the SPL, calling the cleanup_before_linux() function - but no success. Do you have any hints ?
Best regards, Stefano Babic