httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron Bannert <>
Subject Re: chunking of content in mod_include?
Date Tue, 11 Sep 2001 21:02:21 GMT
On Mon, Sep 10, 2001 at 09:10:05PM -0400, Cliff Woolley wrote:
> On Mon, 10 Sep 2001, dean gaudet wrote:
> > > I don't care if mod_include buffers 200 Megs, as long as it is
> > > constantly doing something with the data.  If we have a 200 Meg file
> > > that has no SSI tags in it, but we can get all 200 Meg at one time,
> > > then we shouldn't have any problem just scanning through the entire
> > > 200 Megs very quickly.  Worst case, we do what Brian suggested, and
> > > just check the bucket length once we have finished processing all of
> > > the data in that bucket.  The buffering only becomes a real problem
> > > when we sit waiting for data from a CGI or some other slow content
> > > generator.
> >
> > when you say "buffers 200 Megs" up there do you mean it's mmap()d?
> >
> > (i think that some of the page freeing decisions in linux are based on
> > whether a page is currently mapped or not, and a 200MB file could cause a
> > bit of memory shortage...)
> It is possible for us to be talking about 200MB MMAP bucket here, yes.
> But that won't ever happen in the typical case.  Some module (eg
> mod_file_cache) would have to have explicitly set up that 200MB MMAP
> bucket.
> When you read a file bucket and it automatically MMAPs the file, it places
> an upper bound on file size for mmaping at 16MB.  Files bigger than that
> will be read onto the heap 8KB at a time (one 8KB hunk per heap bucket).
> In that case, mod_include will look at those 8KB heap buckets until it
> reaches one that puts it over its buffering threshold, then it will flush
> the data it's already examined down the filter chain.

I'm sorry if I'm completely uninformed here, but I've been thinking
of this for the last day or so and I don't quite understand why we wouldn't
incrementally mmap() and then munmap() different segments of the file
as we walk through it (at least for mod_include). Wouldn't that save
at least one copy of the data, not to mention the free-list management
overhead we incur from malloc()ing/free()ing all those 8K buckets?


View raw message