Guy Harris (guy@auspex.com) pointed out that I had boo-boo'ed
in a summary of my request for info on how the filesystem buffer
cache works. Specifically:
I had said that a *file open* causes a file to get memory-mapped
into virtual address space. What I meant to say was, *reading*
or *writing* the file causes the parts of it from which the data
is being read/written, to get mapped in. I.e. open files don't
use kernel or user address space.
For those who might be doing performance testing, Guy pointed out
that it follows from the implementation, that writing or reading
data serve equally well to load up the cache, when one needs to
flush other data out of it.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:11 CDT