[v3,00/18] Folio conversion of memcg

Message

Matthew Wilcox June 30, 2021, 4 a.m. UTC

After Michel's comments on "Folio-enabling the page cache", I thought it
best to split out the memcg patches from the rest of the page cache folio
patches and redo them to focus on correctness (ie always passing a folio).

This is fundamentally for review rather than application.  I've rebased
on Linus' current tree, which includes the recent patchbomb from akpm.
That upstream version won't boot on any system I have available, and I'm
not comfortable asking for patches to be applied unless I can actually
try them.  That said, these patches were fine on top of 5.13.

There are still a few functions which take pages, but they rely on other
conversions happening first, which in turn rely on this set of patches,
so I think this is a good place to stop, with the understanding that
there will be more patches later.

Some of the commit logs may be garbled ... I haven't fully taken into
account all of Muchun's recent rework.  All of this work is visible in
context in a git tree here:
https://git.infradead.org/users/willy/pagecache.git/shortlog/refs/heads/folio

Matthew Wilcox (Oracle) (18):
  mm: Add folio_nid()
  mm/memcg: Remove 'page' parameter to mem_cgroup_charge_statistics()
  mm/memcg: Use the node id in mem_cgroup_update_tree()
  mm/memcg: Remove soft_limit_tree_node()
  mm/memcg: Convert memcg_check_events to take a node ID
  mm/memcg: Add folio_memcg() and related functions
  mm/memcg: Convert commit_charge() to take a folio
  mm/memcg: Convert mem_cgroup_charge() to take a folio
  mm/memcg: Convert uncharge_page() to uncharge_folio()
  mm/memcg: Convert mem_cgroup_uncharge() to take a folio
  mm/memcg: Convert mem_cgroup_migrate() to take folios
  mm/memcg: Convert mem_cgroup_track_foreign_dirty_slowpath() to folio
  mm/memcg: Add folio_memcg_lock() and folio_memcg_unlock()
  mm/memcg: Convert mem_cgroup_move_account() to use a folio
  mm/memcg: Add mem_cgroup_folio_lruvec()
  mm/memcg: Add folio_lruvec_lock() and similar functions
  mm/memcg: Add folio_lruvec_relock_irq() and
    folio_lruvec_relock_irqsave()
  mm/workingset: Convert workingset_activation to take a folio

 include/linux/memcontrol.h       | 249 +++++++++++++++----------
 include/linux/mm.h               |   5 +
 include/linux/swap.h             |   2 +-
 include/trace/events/writeback.h |   8 +-
 kernel/events/uprobes.c          |   3 +-
 mm/compaction.c                  |   2 +-
 mm/filemap.c                     |   8 +-
 mm/huge_memory.c                 |   2 +-
 mm/khugepaged.c                  |   8 +-
 mm/ksm.c                         |   3 +-
 mm/memcontrol.c                  | 311 +++++++++++++++----------------
 mm/memory-failure.c              |   2 +-
 mm/memory.c                      |   9 +-
 mm/memremap.c                    |   2 +-
 mm/migrate.c                     |   6 +-
 mm/page_alloc.c                  |   2 +-
 mm/shmem.c                       |   7 +-
 mm/swap.c                        |   4 +-
 mm/userfaultfd.c                 |   2 +-
 mm/vmscan.c                      |   2 +-
 mm/workingset.c                  |  10 +-
 21 files changed, 355 insertions(+), 292 deletions(-)

Comments

Michal Hocko June 30, 2021, 8:44 a.m. UTC | #1

On Wed 30-06-21 05:00:16, Matthew Wilcox wrote:
> After Michel's comments on "Folio-enabling the page cache", I thought it
> best to split out the memcg patches from the rest of the page cache folio
> patches and redo them to focus on correctness (ie always passing a folio).
> 
> This is fundamentally for review rather than application.  I've rebased
> on Linus' current tree, which includes the recent patchbomb from akpm.
> That upstream version won't boot on any system I have available, and I'm
> not comfortable asking for patches to be applied unless I can actually
> try them.  That said, these patches were fine on top of 5.13.
> 
> There are still a few functions which take pages, but they rely on other
> conversions happening first, which in turn rely on this set of patches,
> so I think this is a good place to stop, with the understanding that
> there will be more patches later.

Thanks for considering the last review feedback. After looking at this
pile the conversion is really straightforward and reasonably easy to
review. There are still some cases where you opted to go with
compatibility wrappers and I would still like to avoid those as it
somehow obfuscates the conversion IMHO. As I've said previously, though,
I do understand that this a tedious work so I do not want to make it
more annoying unnecessarily for you. The replacement with page_folio for
the replaced API looks better to me. If you have strong reasons (e.g.
due to dependencies etc.) please make it explicit in the respective
changelogs.

Thanks!