[parisc-linux] flush_tlb_all_local:

[parisc-linux] flush_tlb_all_local:

[Joel Soete]

Hi all,

into parisc/kernel/pacache.S, i find this:
[...]
flush_tlb_all_local:
[...]
         rsm     PSW_SM_Q,%r0       /* Turn off Q bit to load iia queue */
         ldil    L%REAL_MODE_PSW, %r1
         ldo     R%REAL_MODE_PSW(%r1), %r1
         mtctl   %r1, %cr22
         mtctl   %r0, %cr17
         mtctl   %r0, %cr17
[...]

Why twice 'mtctl   %r0, %cr17'?

and also:
[...]
fitmanymiddle:                                  /* Loop if LOOP >= 2 */
         ADDIB>          -1,%r31,fitmanymiddle   /* Adjusted inner loop 
decr */
         pitlbe          0(%sr1,%r28)
         pitlbe,m        %arg1(%sr1,%r28)        /* Last pitlbe and addr 
adjust */

Typicaly the function name is flush_..., so why do we 'purge' here?

Thanks in advance for help,
	Joel

Re: [parisc-linux] flush_tlb_all_local:

[Grant Grundler]

On Sun, Oct 12, 2003 at 04:07:50PM +0000, Joel Soete wrote:
>         ldo     R%REAL_MODE_PSW(%r1), %r1
>         mtctl   %r1, %cr22
>         mtctl   %r0, %cr17
>         mtctl   %r0, %cr17
> [...]
> 
> Why twice 'mtctl   %r0, %cr17'?

Good question. No clue. Looks like a brain fart since clearing it
 once should work unless there's a weird bug. I'll remove it.

> and also:
> [...]
> fitmanymiddle:                                  /* Loop if LOOP >= 2 */
>         ADDIB>          -1,%r31,fitmanymiddle   /* Adjusted inner loop 
> decr */
>         pitlbe          0(%sr1,%r28)
>         pitlbe,m        %arg1(%sr1,%r28)        /* Last pitlbe and addr 
> adjust */
> 
> Typicaly the function name is flush_..., so why do we 'purge' here?

AFAICT, because purge is sufficient.
It doesn't look like parisc even provides a FITBLE instruction.

thanks,
grant

Re: [parisc-linux] flush_tlb_all_local:

[Grant Grundler]

On Sun, Oct 12, 2003 at 09:12:58PM -0600, Grant Grundler wrote:
> On Sun, Oct 12, 2003 at 04:07:50PM +0000, Joel Soete wrote:
> > Why twice 'mtctl   %r0, %cr17'?
> 
> Good question. No clue. Looks like a brain fart since clearing it
>  once should work unless there's a weird bug. I'll remove it.

definitely not a brainfart. Read PA2.0 arch book (section 2) where
it talks about cr17/cr18. CR17 is two registers and writing it twice
clear both head/tail of the IIAS. Just like cr18 is also written twice.

sorry for not catching that the first time around.
grant

Re: [parisc-linux] flush_tlb_all_local:

[John David Anglin]

> On Sun, Oct 12, 2003 at 09:12:58PM -0600, Grant Grundler wrote:
> > On Sun, Oct 12, 2003 at 04:07:50PM +0000, Joel Soete wrote:
> > > Why twice 'mtctl   %r0, %cr17'?
> > 
> > Good question. No clue. Looks like a brain fart since clearing it
> >  once should work unless there's a weird bug. I'll remove it.
> 
> definitely not a brainfart. Read PA2.0 arch book (section 2) where
> it talks about cr17/cr18. CR17 is two registers and writing it twice
> clear both head/tail of the IIAS. Just like cr18 is also written twice.

Probably, there should be a comment in the code noting this special
behavior.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: [parisc-linux] flush_tlb_all_local:

[Matthew Wilcox]

On Mon, Oct 13, 2003 at 02:57:19PM -0400, John David Anglin wrote:
> Probably, there should be a comment in the code noting this special
> behavior.

Done.  They're deliberately a little bit cryptic to ensure that people
actually go and look at the arch book, but they are different so people
don't think this might be a merge error.

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk