Hi Rik,
On Thu, 16 Nov 2000, Rik van Riel wrote:
> On 16 Nov 2000, Christoph Rohland wrote:
>> Also we have to make sure to derefence the swap entry if the
>> last reference is in the shm segmant table .
>
> Why is this?
Because you never get a page fault on the shm segmants. So you never
will
On 16 Nov 2000, Christoph Rohland wrote:
> Oh, I missed one point: we need to handle the swapout of
> nonattached pages: in shm you can detach the last user and the
> segment with content is still around. So we have to scan the shm
> objects themselves also. Should We could do this in the same
>
Oh, I missed one point: we need to handle the swapout of nonattached
pages: in shm you can detach the last user and the segment with
content is still around. So we have to scan the shm objects themselves
also. Should We could do this in the same loop as we scan the mm's?
Also we have to make sure
Hi Rik,
On Thu, 16 Nov 2000, Rik van Riel wrote:
> On 16 Nov 2000, Christoph Rohland wrote:
>> So in shm_swap_out I check if the page is already in the swap
>> cache. If not I put the page into it and note the swap entry in
>> the shadow pte of shm. Right?
>
> Exactly. And I'll change page_laund
On 16 Nov 2000, Christoph Rohland wrote:
> On Wed, 15 Nov 2000, Rik van Riel wrote:
> > On 15 Nov 2000, Christoph Rohland wrote:
> > You really want to have it in the swap cache, so we have
> > a place for it allocated in cache, etc...
> >
> > Basically, when we unmap it in try_to_swap_out(), we
Hi Rik,
On Wed, 15 Nov 2000, Rik van Riel wrote:
> On 15 Nov 2000, Christoph Rohland wrote:
> You really want to have it in the swap cache, so we have
> a place for it allocated in cache, etc...
>
> Basically, when we unmap it in try_to_swap_out(), we
> should add the page to the swap cache, and
On 15 Nov 2000, Christoph Rohland wrote:
> On Wed, 15 Nov 2000, Rik van Riel wrote:
> > On 15 Nov 2000, Christoph Rohland wrote:
> >> 2) Integrating it into the global lru lists and/or the page cache.
> >>
> >> I think the second approach is the way to go but I do not
> >> understand the global
Hi Rik,
On Wed, 15 Nov 2000, Rik van Riel wrote:
> On 15 Nov 2000, Christoph Rohland wrote:
>
>> - shm_swap is called from swap_out. Actually on my machine after a
>>while it only gets called without __GFP_IO set, which means it
>>will not do anything which again leads to deadlock.
>
>
On 15 Nov 2000, Christoph Rohland wrote:
> - shm_swap is called from swap_out. Actually on my machine after a
>while it only gets called without __GFP_IO set, which means it will
>not do anything which again leads to deadlock.
Only _without_ __GFP_IO ? That's not quite right since
that
Hi Rik et al,
here I am again. I investigated more into the shm swapping in 2.4 and
discovered the following things:
- shm_swap is called from swap_out. Actually on my machine after a
while it only gets called without __GFP_IO set, which means it will
not do anything which again leads to
Forgot to append the patch...
Greetings
Christoph
diff -uNr 4-11-3/ipc/shm.c c/ipc/shm.c
--- 4-11-3/ipc/shm.cWed Oct 4 15:58:02 2000
+++ c/ipc/shm.c Tue Nov 14 17:43:01 2000
@@ -80,6 +80,7 @@
unsigned long shm_npages; /* size of segment (pages) */
p
11 matches
Mail list logo
