Re: [PATCH v2] mm: Add MREMAP_DONTUNMAP to mremap().

[Brian Geffon]

Hi Kirill,

On Wed, Jan 29, 2020 at 11:46 AM Kirill A. Shutemov
<kirill@xxxxxxxxxxxxx> wrote:
> And based on the use case you probably don't really need 'fixed'
> semantics all the time. The user should be fine with moving the mapping
> *somewhere*, not neccessary to the given address

This is true and and it simplifies things a bit as for the outlined
use cases the user would not be required to mmap the destination
before hand. Part of the reason I chose to require MREMAP_FIXED was
because mremap need not move the mapping if it can shrink/grow in
place and it seemed a bit awkward to have "MUSTMOVE" behavior without
MAP_FIXED. I'll make this change to drop the requirement on
MREMAP_FIXED in my next patch.

> BTW, name of the flag is confusing. My initial reaction was that it is
> variant of MREMAP_FIXED that does't anything at the target address.
> Like MAP_FIXED vs. MAP_FIXED_NOREPLACE.
>
> Any better options for the flag name? (I have none)

I see your point. Perhaps MREMAP_MOVEPAGES or MREMAP_KEEP_SOURCE? I
struggle to come up with a single name that encapsulates this behavior
but I'll try to think of other ideas before I mail the next patch.
Given that we will drop the requirement on MREMAP_FIXED, perhaps
MOVEPAGES is the better option as it captures that the mapping WILL be
moved?

Thanks again for taking the time to look at this.

Best,
Brian