linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v3 00/30] maple_tree: Replace big node with maple copy
@ 2026-01-30 20:59 Liam R. Howlett
  2026-01-30 20:59 ` [PATCH v3 01/30] maple_tree: Fix mas_dup_alloc() sparse warning Liam R. Howlett
                   ` (30 more replies)
  0 siblings, 31 replies; 41+ messages in thread
From: Liam R. Howlett @ 2026-01-30 20:59 UTC (permalink / raw)
  To: Andrew Morton
  Cc: maple-tree, linux-mm, linux-kernel, Suren Baghdasaryan,
	Matthew Wilcox, Sidhartha Kumar, Vlastimil Babka, Alice Ryhl,
	Kuninori Morimoto, Geert Uytterhoeven, Arnd Bergmann,
	Christian Kujau, SeongJae Park, Liam R. Howlett

The big node struct was created for simplicity of splitting,
rebalancing, and spanning store operations by using a copy buffer to
create the data necessary prior to breaking it up into 256B nodes.
Certain operations were rather tricky due to the restriction of keeping
NULL entries together and never at the end of a node (except the
right-most node).

The big node struct is incompatible with future features that are
currently in development.  Specifically different node types and
different data type sizes for pivots.

The big node struct was also a stack variable, which caused issues with
certain configurations of kernel build.

This series removes big node by introducing another node type which will
never be written to the tree: maple_copy.  The maple copy node operates
more like a scatter/gather operation with a number of sources and
destinations of allocated nodes.

The sources are copied to the destinations, in turn, until the sources
are exhausted.  The destination is changed if it is filled or the split
location is reached prior to the source data end.

New data is inserted by using the maple copy node itself as a source
with up to 3 slots and pivots.  The data in the maple copy node is the
data being written to the tree along with any fragment of the range(s)
being overwritten.

As with all nodes, the maple copy node is of size 256B.  Using a node
type allows for the copy operation to treat the new data stored in the
maple copy node the same as any other source node.

Analysis of the runtime shows no regression or benefit of removing the
larger stack structure.  The motivation is the ground work to use new
node types and to help those with odd configurations that have had
issues.

The change was tested by myself using mm_tests on amd64 and by Suren on
android (arm64).  Limited testing on s390 qemu was also performed using
stress-ng on the virtual memory, which should cover many corner cases.

v1: https://lore.kernel.org/all/20260115193647.1695937-1-Liam.Howlett@oracle.com/
v2: https://lore.kernel.org/all/20260121164526.2093265-1-Liam.Howlett@oracle.com/

Changes since v2:
 - Whitespace fix in rebalance_data()
 - Added ma_init_slot() for cleaner casting during RCU_INIT_POINTER().
   Thanks test robot & SK [1]
 - Fixed off-by-one in rebalance_data() which caused node overuse,
   reported by linux-next. Thanks Mark [2]
 - Added new patch to reproducer to test cases in userspace testing for
   the rebalance_data() off-by-one

[1]. https://lore.kernel.org/all/20260122055516.69335-1-sj@kernel.org/
[2]. https://lore.kernel.org/all/bd1c3356-11a1-4d0b-bf58-47eb21bfd24d@sirena.org.uk/

Liam R. Howlett (30):
  maple_tree: Fix mas_dup_alloc() sparse warning
  maple_tree: Move mas_spanning_rebalance loop to function
  maple_tree: Extract use of big node from mas_wr_spanning_store()
  maple_tree: Remove unnecessary assignment of orig_l index
  maple_tree: inline mas_spanning_rebalance() into
    mas_wr_spanning_rebalance()
  maple_tree: Make ma_wr_states reliable for reuse in spanning store
  maple_tree: Remove l_wr_mas from mas_wr_spanning_rebalance
  maple_tree: Don't pass through height in mas_wr_spanning_store
  maple_tree: Move maple_subtree_state from mas_wr_spanning_store to
    mas_wr_spanning_rebalance
  maple_tree: Correct right ma_wr_state end pivot in
    mas_wr_spanning_store()
  maple_tree: Introduce maple_copy node and use it in
    mas_spanning_rebalance()
  maple_tree: Testing update for spanning store
  maple_tree: Inline mas_spanning_rebalance_loop() into
    mas_wr_spanning_rebalance()
  maple_tree: Change initial big node setup in
    mas_wr_spanning_rebalance()
  maple_tree: Introduce ma_leaf_max_gap()
  maple_tree: Add gap support, slot and pivot sizes for maple copy
  maple_tree: Start using maple copy node for destination
  maple_tree: inline mas_wr_spanning_rebalance()
  maple_tree: Remove unnecessary return statements
  maple_tree: Separate wr_split_store and wr_rebalance store type code
    path
  maple_tree: Add cp_is_new_root() helper
  maple_tree: Use maple copy node for mas_wr_rebalance() operation
  maple_tree: Add test for rebalance calculation off-by-one
  maple_tree: Add copy_tree_location() helper
  maple_tree: Add cp_converged() helper
  maple_tree: Use maple copy node for mas_wr_split()
  maple_tree: Remove maple big node and subtree structs
  maple_tree: Pass maple copy node to mas_wmb_replace()
  maple_tree: Don't pass end to mas_wr_append()
  maple_tree: Clean up mas_wr_node_store()

 include/linux/maple_tree.h       |   42 +
 lib/maple_tree.c                 | 2134 +++++++++++++-----------------
 lib/test_maple_tree.c            |   55 +-
 tools/testing/radix-tree/maple.c |  308 ++++-
 4 files changed, 1340 insertions(+), 1199 deletions(-)

-- 
2.47.3



^ permalink raw reply	[flat|nested] 41+ messages in thread

end of thread, other threads:[~2026-02-04  6:36 UTC | newest]

Thread overview: 41+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-01-30 20:59 [PATCH v3 00/30] maple_tree: Replace big node with maple copy Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 01/30] maple_tree: Fix mas_dup_alloc() sparse warning Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 02/30] maple_tree: Move mas_spanning_rebalance loop to function Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 03/30] maple_tree: Extract use of big node from mas_wr_spanning_store() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 04/30] maple_tree: Remove unnecessary assignment of orig_l index Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 05/30] maple_tree: inline mas_spanning_rebalance() into mas_wr_spanning_rebalance() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 06/30] maple_tree: Make ma_wr_states reliable for reuse in spanning store Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 07/30] maple_tree: Remove l_wr_mas from mas_wr_spanning_rebalance Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 08/30] maple_tree: Don't pass through height in mas_wr_spanning_store Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 09/30] maple_tree: Move maple_subtree_state from mas_wr_spanning_store to mas_wr_spanning_rebalance Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 10/30] maple_tree: Correct right ma_wr_state end pivot in mas_wr_spanning_store() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 11/30] maple_tree: Introduce maple_copy node and use it in mas_spanning_rebalance() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 12/30] maple_tree: Testing update for spanning store Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 13/30] maple_tree: Inline mas_spanning_rebalance_loop() into mas_wr_spanning_rebalance() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 14/30] maple_tree: Change initial big node setup in mas_wr_spanning_rebalance() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 15/30] maple_tree: Introduce ma_leaf_max_gap() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 16/30] maple_tree: Add gap support, slot and pivot sizes for maple copy Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 17/30] maple_tree: Start using maple copy node for destination Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 18/30] maple_tree: inline mas_wr_spanning_rebalance() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 19/30] maple_tree: Remove unnecessary return statements Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 20/30] maple_tree: Separate wr_split_store and wr_rebalance store type code path Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 21/30] maple_tree: Add cp_is_new_root() helper Liam R. Howlett
2026-02-01  0:10   ` SeongJae Park
2026-02-02 14:58     ` Liam R. Howlett
2026-02-02 15:56       ` SeongJae Park
2026-02-02 17:01         ` Liam R. Howlett
2026-02-02 17:53           ` SeongJae Park
2026-02-03 17:26   ` Liam R. Howlett
2026-02-04  6:36     ` SeongJae Park
2026-01-30 20:59 ` [PATCH v3 22/30] maple_tree: Use maple copy node for mas_wr_rebalance() operation Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 23/30] maple_tree: Add test for rebalance calculation off-by-one Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 24/30] maple_tree: Add copy_tree_location() helper Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 25/30] maple_tree: Add cp_converged() helper Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 26/30] maple_tree: Use maple copy node for mas_wr_split() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 27/30] maple_tree: Remove maple big node and subtree structs Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 28/30] maple_tree: Pass maple copy node to mas_wmb_replace() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 29/30] maple_tree: Don't pass end to mas_wr_append() Liam R. Howlett
2026-01-30 20:59 ` [PATCH v3 30/30] maple_tree: Clean up mas_wr_node_store() Liam R. Howlett
2026-01-31 20:27 ` [PATCH v3 00/30] maple_tree: Replace big node with maple copy Andrew Morton
2026-02-02 15:40   ` Liam R. Howlett
2026-02-02 18:31     ` Andrew Morton

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox