On 5 January 2011 02:02, Albert ARIBAUD albert.aribaud@free.fr wrote:
(sending again, this time to the list)
Le 04/01/2011 12:02, Minkyu Kang a écrit :
Some bss variables are initialized to 0xffffffff in my test case. In this case we should skip the fixup too. So, I make v2 patch for check it.
0xffffffff? What variables? Feels like a some other bug to me.
It's DRAM initial variable.
Yes, seems like an unrelated issue to me, or at least an unexplained one so far.
For instance, there is no fixup to BSS variables (they are to be initialized to 0 whatever the load address is, and this initialization is not in the binary, it is done in the startup code). So which fixup should be skipped exactly and how is is related to the BSS variable?
Yes, you right. Seems to another bug. It sometimes happens at arrays on BSS area. OK, I'll investigate this problem.
How about v1 patch for NULL pointer. Do you agreed with this patch? http://patchwork.ozlabs.org/patch/76784/
If so, I'm going to make a patch for all of ARM CPUs.
Thanks Minkyu Kang