On Sun, 2016-08-14 at 19:55 +0100, Peter Maydell wrote: > On 13 August 2016 at 12:52, Benjamin Herrenschmidt > <b...@kernel.crashing.org> wrote: > > > > This is very handy when debugging a guest, especially when it's > > stuck on accessing some HW and the only way to figure out what > > specific piece of HW is to translate the virtual address to > > a hardware address that can then be matched with the mtree > > > > Signed-off-by: Benjamin Herrenschmidt <b...@kernel.crashing.org> > > Should this wrap cpu_get_phys_page_attrs_debug() instead > (and report the attributes to the user)?
Ah good idea, I'll look into it. On a vaguely related note, do you know if anybody is planning (or trying to) overhaul that API to get the access type as an input as well ? We can have widely different translations for data and instructions and that has been a problem (for example if some OS uses a split MMU mode like that, currently -d in_asm will try to disassemble using data translation). Cheers, Ben.
