[parisc-linux] PA-RISC Linux VM

[parisc-linux] PA-RISC Linux VM

[CARSON,KEVIN (HP-Canada,ex1)]

The recent article in Byte
(http://www.byte.com/documents/s=1436/byt20011024s0002/1029_moshe.html)
makes me curious; is the current pa-risc Linux port using a Rik van Viel VM
or Andrea Arcangeli VM or some pa-risc port specific design?

Is there possibly a large changeover looming dependent upon what shakes out
in the Linus tree?

Kevin

Re: [parisc-linux] PA-RISC Linux VM

[Matthew Wilcox]

On Wed, Oct 31, 2001 at 01:29:14PM -0500, CARSON,KEVIN (HP-Canada,ex1) wrote:
> The recent article in Byte
> (http://www.byte.com/documents/s=1436/byt20011024s0002/1029_moshe.html)
> makes me curious; is the current pa-risc Linux port using a Rik van Viel VM
> or Andrea Arcangeli VM or some pa-risc port specific design?
> 
> Is there possibly a large changeover looming dependent upon what shakes out
> in the Linus tree?

We're sticking with 2.4.9 until we see a Linus kernel that's a little
more stable.  That's from slightly before Linus introduced Andrea's VM.

i just posted a small patch we need to linux-mm
http://mail.nl.linux.org/linux-mm/2001-10/msg00098.html
which is the equivalent of the only major vm change that we have
in our tree.  There's a couple of minor ones:

diff -u -p -r1.1.1.5 -r1.8
--- mm/shmem.c  2001/08/15 16:33:42     1.1.1.5
+++ mm/shmem.c  2001/10/01 10:19:21     1.8
@@ -377,6 +377,7 @@ repeat:
                if (!page)
                        return ERR_PTR(-ENOMEM);
                clear_highpage(page);
+               flush_dcache_page(page);
                inode->i_blocks += BLOCKS_PER_PAGE;
                add_to_page_cache (page, mapping, idx);
        }
diff -u -p -r1.1.1.13 -r1.15
--- mm/vmscan.c 2001/08/16 22:10:13     1.1.1.13
+++ mm/vmscan.c 2001/09/06 09:44:13     1.15
@@ -120,6 +120,7 @@ static void try_to_swap_out(struct mm_st
         * any IO - it's already up-to-date on disk.
         */
        if (PageSwapCache(page)) {
+               flush_cache_page(vma, address);
                entry.val = page->index;
                if (pte_dirty(pte))
                        set_page_dirty(page);

I don't anticipate either of these changes being major merge problems.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] PA-RISC Linux VM

[Jes Sorensen]

>>>>> "Kevin" == CARSON,KEVIN (HP-Canada,ex1) <kevin_carson@hp.com> writes:

Kevin> The recent article in Byte
Kevin> (http://www.byte.com/documents/s=1436/byt20011024s0002/1029_moshe.html)
Kevin> makes me curious; is the current pa-risc Linux port using a Rik
Kevin> van Viel VM or Andrea Arcangeli VM or some pa-risc port
Kevin> specific design?

Kevin> Is there possibly a large changeover looming dependent upon
Kevin> what shakes out in the Linus tree?

There'll be some shaking up depending on what happens in Linus' tree
and what level of large bats people bring out to get things
straightened out. It doesn't seem totally obvious right now where we
are going to end (we as in the 2.4.x kernels in general).

On the other hand I wouldn't take that Byte article too serious; his
'scientific' conclusions seems to be based on how much the mp3 player
skips when loading large files into an editor.

Cheers,
Jes