VM STATS Question

[Zachary Frenette]

When we zero-fill part of the page only (so say, part of the page is in
the elf file, but the rest has to be zero-filled), do we count this as a
VMSTAT_PAGE_FAULT_ZERO, or a VMSTAT_PAGE_FAULT_DISK?

Right now I have: VM_PAGE_FAULT_ZERO is increase whenever part of the
page is zero-filled, and VM_PAGE_FAULT_DISK whenever the ENTIRE page is
in the elf file. Is this correct?

[cs350]

I would do it differently: if the disk is involved at all, I would count
that as a FAULT_DISK even if part of the page is padded with zeroes.

However, it is expected that different implementations will have slightly
different vm statistics. So, I believe that counting your way is also
acceptable.

-Alex

When we zero-fill part of the page only (so say, part of the page is in the
elf file, but the rest has to be zero-filled), do we count this as a
VMSTAT_PAGE_FAULT_ZERO, or a VMSTAT_PAGE_FAULT_DISK?
Right now I have: VM_PAGE_FAULT_ZERO is increase whenever part of the page is
zero-filled, and VM_PAGE_FAULT_DISK whenever the ENTIRE page is in the elf
file. Is this correct?

[mammukht]

Another VM Stats question: Do we count when we enter vm_fault because of
a read only fault? I'm assuming not but just wanted to make sure.

Post by cs350
I would do it differently: if the disk is involved at all, I would count
that as a FAULT_DISK even if part of the page is padded with zeroes.
However, it is expected that different implementations will have
slightly different vm statistics. So, I believe that counting your way
is also acceptable.
-Alex

Post by Zachary Frenette
When we zero-fill part of the page only (so say, part of the page is
in the elf file, but the rest has to be zero-filled), do we count this
as a VMSTAT_PAGE_FAULT_ZERO, or a VMSTAT_PAGE_FAULT_DISK?
Right now I have: VM_PAGE_FAULT_ZERO is increase whenever part of the
page is zero-filled, and VM_PAGE_FAULT_DISK whenever the ENTIRE page
is in the elf file. Is this correct?