Hi again.
When I began work on swapfile support, I looked for an efficient method
to store all the information on which blocks were used. The process led
me to develop something I called ranges, which Pavel later looked at and
said something like 'Oh. Extents.'
Throughout the code, I still call them ranges (I have, for example
struct range and struct rangechain). In preparation for merging, should
I go through an rename ranges to extents, or will they be okay as it is?
Regards,
Nigel
--
My work on Software Suspend is graciously brought to you by
LinuxFund.org.
Nigel Cunningham <[email protected]> wrote:
>
> Hi again.
>
> When I began work on swapfile support, I looked for an efficient method
> to store all the information on which blocks were used. The process led
> me to develop something I called ranges, which Pavel later looked at and
> said something like 'Oh. Extents.'
>
> Throughout the code, I still call them ranges (I have, for example
> struct range and struct rangechain). In preparation for merging, should
> I go through an rename ranges to extents, or will they be okay as it is?
Are you aware of the current `struct swap_extent' and its supporting
infrastructure?
Hi.
Thanks for the pointer.
I was aware of swap extents, but didn't look at the supporting
infrastructure because I originally wrote the code for 2.4, before I
looked at porting it to 2.6.
Having said that, I've looked at them now, but think that they probably
wouldn't help me.
This is partly because I've ended up using extents to store all the
meta-data: swap addresses, which blocks on the swap devices are used,
and which pages are used for what. (Sorry, only mentioning blocks in the
previous message did mislead you).
More importantly that that, though, I'm using whole pages for storing
the extents, and store the pages in the header of the image (after
making them relocatable). They replace the 'pagedirs' in Patrick's and
Pavel's implementations.
Regards,
Nigel
On Fri, 2004-01-23 at 19:27, Andrew Morton wrote:
> Nigel Cunningham <[email protected]> wrote:
> >
> > Hi again.
> >
> > When I began work on swapfile support, I looked for an efficient method
> > to store all the information on which blocks were used. The process led
> > me to develop something I called ranges, which Pavel later looked at and
> > said something like 'Oh. Extents.'
> >
> > Throughout the code, I still call them ranges (I have, for example
> > struct range and struct rangechain). In preparation for merging, should
> > I go through an rename ranges to extents, or will they be okay as it is?
>
> Are you aware of the current `struct swap_extent' and its supporting
> infrastructure?
--
My work on Software Suspend is graciously brought to you by
LinuxFund.org.