[PATCH 0/5] powerpc: Implement masked user access

[Segher Boessenkool]

On Thu, Jun 26, 2025 at 07:56:10AM +0200, Christophe Leroy wrote:
> Le 24/06/2025 à 23:08, David Laight a écrit :
> >On Tue, 24 Jun 2025 13:25:05 -0500
> >Segher Boessenkool <segher at kernel.crashing.org> wrote:
> >>>>isel (which is base PowerPC, not something "e500" only) is a
> >>>>computational instruction, it copies one of two registers to a third,
> >>>>which of the two is decided by any bit in the condition register.
> >>>
> >>>Does that mean it could be used for all the ppc cpu variants?
> >>
> >>No, only things that implement architecture version of 2.03 or later.
> >>That is from 2006, so essentially everything that is still made
> >>implements it :-)
> >>
> >>But ancient things do not.  Both 970 (Apple G5) and Cell BE do not yet
> >>have it (they are ISA 2.01 and 2.02 respectively).  And the older p5's
> >>do not have it yet either, but the newer ones do.
> 
> For book3s64, GCC only use isel with -mcpu=power9 or -mcpu=power10

I have no idea what "book3s64" means.

Some ancient Power architecture versions had something called
"Book III-S", which was juxtaposed to "Book III-E", which essentially
corresponds to the old aborted BookE stuff.

I guess you mean almost all non-FSL implementations?  Most of those
support the isel insns.  Like, Power5+ (GS).  And everything after that.

I have no idea why you think power9 has it while older CPUS do not.  In
the GCC source code we have this comment:
  /* For ISA 2.06, don't add ISEL, since in general it isn't a win, but
     altivec is a win so enable it.  */
and in fact we do not enable it for ISA 2.06 (p8) either, probably for
a similar reason.

> >>And all classic PowerPC is ISA 1.xx of course.  Medieval CPUs :-)
> >
> >That make more sense than the list in patch 5/5.
> 
> Sorry for the ambiguity. In patch 5/5 I was addressing only powerpc/32, 
> and as far as I know the only powerpc/32 supported by Linux that has 
> isel is the 85xx which has an e500 core.

What is "powerpc/32"?  It does not help if you use different names from
what everyone else does.

The name "powerpc32" is sometimes used colloquially to mean PowerPC code
running in SF=0 mode (MSR[SF]=0), but perhaps more often it is used for
32-bit only implementations (so, those that do not even have that bit:
it's bit 0 in the 64-bit MSR, so all implementations that have an only
32-bit MSR, for example).

> For powerpc/64 we have less constraint than on powerpc32:
> - Kernel memory starts at 0xc000000000000000
> - User memory stops at 0x0010000000000000

That isn't true, not even if you mean some existing name.  Usually
userspace code is mapped at 256MB (0x10000000).  On powerpc64-linux
anyway, different default on different ABIs of course :-)

> >And for access_ok() avoiding the conditional is a good enough reason
> >to use a 'conditional move' instruction.
> >Avoiding speculation is actually free.
> 
> And on CPUs that are not affected by Spectre and Meltdown like powerpc 
> 8xx or powerpc 603,

Erm.


Segher