Re: [PATCH 2/2] sparc64: flush vmalloc ranges from the D-cache on map and unmap

[Imre Kaloz]

Hi Stian,

> Since the vmemmap has no aliases to flush, I suggest we skip
> anything outside vmalloc and module space, like this:

Thanks for chasing this down. Confirmed the path: dcache_size is
still 0 there, so the old code took the "exceeds cache size" branch
into flush_dcache_all() every time, and the mondo block PA is 0
until init_IRQ() runs, so the cross-call landed on physical address
0 instead of a real mondo queue. Skipping non-vmalloc/module
addresses is the right fix.

v2 with your check and your Tested-by follows.


Best,
Imre