On Fri, 9 Mar 2018 14:47:31 -0800 Mike Kravetz <mike.krav...@oracle.com> wrote:
> start_isolate_page_range() is used to set the migrate type of a
> set of pageblocks to MIGRATE_ISOLATE while attempting to start
> a migration operation. It assumes that only one thread is
> calling it for the specified range. This routine is used by
> CMA, memory hotplug and gigantic huge pages. Each of these users
> synchronize access to the range within their subsystem. However,
> two subsystems (CMA and gigantic huge pages for example) could
> attempt operations on the same range. If this happens, one thread
> may 'undo' the work another thread is doing. This can result in
> pageblocks being incorrectly left marked as MIGRATE_ISOLATE and
> therefore not available for page allocation.
> What is ideally needed is a way to synchronize access to a set
> of pageblocks that are undergoing isolation and migration. The
> only thing we know about these pageblocks is that they are all
> in the same zone. A per-node mutex is too coarse as we want to
> allow multiple operations on different ranges within the same zone
> concurrently. Instead, we will use the migration type of the
> pageblocks themselves as a form of synchronization.
> start_isolate_page_range sets the migration type on a set of page-
> blocks going in order from the one associated with the smallest
> pfn to the largest pfn. The zone lock is acquired to check and
> set the migration type. When going through the list of pageblocks
> check if MIGRATE_ISOLATE is already set. If so, this indicates
> another thread is working on this pageblock. We know exactly
> which pageblocks we set, so clean up by undo those and return
> This allows start_isolate_page_range to serve as a synchronization
> mechanism and will allow for more general use of callers making
> use of these interfaces. Update comments in alloc_contig_range
> to reflect this new functionality.
> + * There is no high level synchronization mechanism that prevents two threads
> + * from trying to isolate overlapping ranges. If this happens, one thread
> + * will notice pageblocks in the overlapping range already set to isolate.
> + * This happens in set_migratetype_isolate, and set_migratetype_isolate
> + * returns an error. We then clean up by restoring the migration type on
> + * pageblocks we may have modified and return -EBUSY to caller. This
> + * prevents two threads from simultaneously working on overlapping ranges.
Well I can kinda visualize how this works, with two CPUs chewing away
at two overlapping blocks of pfns, possibly with different starting
pfns. And I can't immediately see any holes in it, apart from possible
memory ordering issues. What guarantee is there that CPU1 will see
CPU2's writes in the order in which CPU2 performed them? And what
guarantee is there that CPU1 will see CPU2's writes in a sequential
manner? If four of CPU2's writes get written back in a single atomic
flush, what will CPU1 make of that?