On Tue, Dec 11, 2012 at 09:42:21PM -0600, Joe Hershberger wrote:
Hi Thomas,
On Tue, Dec 11, 2012 at 5:47 PM, Langer Thomas (LQDE RD ST PON SW) thomas.langer@lantiq.com wrote:
Hello Joe,
These patches for static initialization of the struct eth_ops will break the drivers on any architecture which needs manual relocation!
I've send a comment regarding this, please see here: http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/146092/focus=147607
I missed that message before. Thanks for keeping an eye out!
The tests with QEMU are insufficient, as probably the memory locations before relocation were still available and used via the uncorrected pointers.
I just ran into another issue on MIPS where the is the case... serial is busted (serial_initialize() isn't called, which manually relocates), but it doesn't appear as a crash because of the pre-relocation addresses still being valid.
So please consider to revoke the pull request until an update with fixes for the "eth_ops" series is available.
Indeed. Tom, please disregard this pull request.
Disregarded.