Re: Fiasco.oc & trustzone

On Thu Sep 20, 2012 at 20:25:32 +0400, Nikolay Golikov wrote:
> 2012/9/20 Nikolay Golikov <[email protected]>:
> > 2012/9/20 Adam Lackorzynski <[email protected]>:
> >> On Thu Sep 20, 2012 at 15:25:54 +0400, Nikolay Golikov wrote:
> >>> 2012/9/20 Adam Lackorzynski <[email protected]>:
> >>> > On Wed Sep 19, 2012 at 17:41:01 +0400, Nikolay Golikov wrote:
> >>> >> Now I have another problem: I map aipstz-1 registers with
> >>> >> l4io_request_iomem(base, AISPTZ_SIZE, L4IO_MEM_NONCACHED, &_vbase);
> >>> >> (base = 0x53F00000, AISPTZ_SIZE = 0x4000)
> >>> >> It return 0.
> >>> >>
> >>> >> But when I try to write to mapped registers my program stopping:
> >>> >> *((volatile l4_umword_t *)(_vbase + reg)) = val
> >>> >> (reg = 0x0, val = 0x77777777)
> >>> >
> >>> > Could you check whether it page-faults permanently? (Enable page-fault
> >>> > logging in jdb: P*, let it go, then check trace-buffer via shift-T.)
> >>> > Also try using l4_mem_write as it could be that writing from userspace
> >>> > does not work.
> >>>
> >>> Yes, It's page-faults permanently. In tracebuffer I see:
> >>>
> >>> pf:  0044 pfa=00003000 ip=01000294 (Wp) spc=0xf12e4690                     79598
> >>> pf:  0044 pfa=00003000 ip=01000294 (Wp) spc=0xf12e4690                     79597
> >>> pf:  0044 pfa=00003000 ip=01000294 (Wp) spc=0xf12e4690                     79596
> >>>
> >>> why it happening?
> >>
> >> Device may only be written/read from secure privileged mode, this is one
> >> of them (I think). Please try writing/reading via l4_mem_write and
> >> l4_mem_read.
> >>
> >>
> >>
> >> Adam
> >> --
> >> Adam                 [email protected]
> >>   Lackorzynski         http://os.inf.tu-dresden.de/~adam/
> >
> > I tried use l4_mem_write for write 0x77777777, it looks ok. But, when
> > I tried l4_mem_read for read from same register it returns 0x0.
> > But similar code injected to bootstrap startup function works fine.
> 
> I tried use l4_mem_write for write 0x77777777, it looks ok. But, when
> I tried l4_mem_read for read from same register it returns 0x0.
> But similar code injected to bootstrap startup function works fine.

Hmm, strange. Could you verify in Fiasco in kern/arm/mem_op.cpp which
value is actually written/read?


Adam
-- 
Adam                 [email protected]
  Lackorzynski         http://os.inf.tu-dresden.de/~adam/