On Tuesday 28 April 2009 03:13:05 Wolfgang Denk wrote:
In message Mike Frysinger wrote:
Should we not rather use 0xFF for padding - given that the envrionment is frequently stored in NOR flash?
i would love to (and with the Blackfin stuff, i do just that), but the problem is that there is no way of forcing the embedded environment to 0xff pad that i know of ...
You mean in the code? Well, if you're using a separate section for it (like the ppcenv section) this should be straightforward using the `=FILLEXP' output section attribute for the linker.
hmm, a quick grep shows that almost no one has a dedicated section. the majority of people (Blackfin boards included) insert it into the middle of their .text section. this is the use case i was thinking of when i said "no way" because i hadnt seen anyone using a dedicated section before.
so, to keep the most common behavior from changing (the char[] env is embedded in the .data section padded with 0x00), the padding defaults to 0x00. this way `envcrc` outputs the right value (see common/Makefile).
for people (like me) who are using --binary to extract the env blob, we can do --binary=0xff and get what you're talking about.
i could extend the behavior so that using plain "--binary" is the same thing as "--binary=0xff" ...
OK.
I guess you want to suggest that I accept this patch as is, and we address the gap filling with 0xFF in a separate patch? ;-)
if you want the "--binary" thing, i can post a new patch. otherwise yes, i dont see any changes to be made for now. -mike