On Thu, Feb 23, 2023 at 09:59:35AM +1100, Alistair Popple wrote:
> The idea was every driver already needs to allocate a pages array to
> pass to pin_user_pages(), and by necessity drivers have to keep a
> reference to the contents of that in one form or another. So
> conceptually the equivalent of:
>
> struct vm_account {
> struct list_head possible_pinners;
> struct mem_cgroup *memcg;
> struct pages **pages;
> [...]
> };
>
> Unpinnig involves finding a new owner by traversing the list of
> page->memcg_data->possible_pinners and iterating over *pages[] to figure
> out if that vm_account actually has this page pinned or not and could
> own it.
>
> Agree this is costly though. And I don't think all drivers keep the
> array around so "iterating over *pages[]" may need to be a callback.
Is pinning in this context referring to FOLL_LONGTERM pins or any
FOLL_PIN? In the latter case block based direct I/O does not keep
the pages array around, and also is absolutely not willing to pay
for the overhead.
For FOLL_LONGTERM the schemes sounds vaguely reasonable to me.
