Re: [PATCH] powerpc/4xx/ocm: fix compiler error

[Segher Boessenkool <segher@kernel.crashing.org>]

On Sat, Dec 22, 2018 at 08:37:28PM +0100, christophe leroy wrote:
> Le 22/12/2018 à 18:16, Segher Boessenkool a écrit :
> >On Sat, Dec 22, 2018 at 02:08:02PM +0100, christophe leroy wrote:
> >>
> >>Usually, Guarded implies no exec (at least on 6xx and 8xx).
> >
> >Huh?  What do you mean here?
> 
> From the 885 Reference Manual:
> 
> Address translation: the EA is translated by using the MMU’s TLB 
> mechanism. Instructions are not fetched from no-execute or guarded 
> memory and data accesses are not executed speculatively to or from the 
> guarded memory.
> 
> 6.1.3.4 Instruction TLB Error Exception (0x01300)
> This type of exception occurs as a result of one of the following 
> conditions if MSR[IR] = 1:
> - The EA cannot be translated.
> - The fetch access violates memory protection
> - The fetch access is to guarded memory
> 
> 
> From e300core reference manual:
> 
> Translation Exception Conditions:
> Exception condition: Instruction fetch from guarded memory
> with MSR[IR] = 1 ==> ISI interrupt SRR1[3] = 1

Right, but you said 6xx as well, i.e. pure PowerPC.

If for example IR=0 you cannot have N=1, but you do have G=1.  There is
no case where G=1 implies N=1 afaik, or where fetch is prohibited some
other way (causes an ISI, say).


Segher