[Newlib] Bug with malloc_r()

[JamesM]

I noticed that in some cases ph_virtaddr wasn't on a page boundary

That sound strange, I never saw that. Did you put align for every section in the linker script? Can you confirm there is nothing on SHT_RELA?
I am interested to know why, since the same problem may also arise in mine kernel.

There's no reason why a user application should have its segments aligned to page boundaries.

[bluemoon]

I see, thanks.

To support that scenario I simply align/adjust ph_virtaddr/size before passing to alloc/mapping the memory. blanham may try this as well.