]> www.infradead.org Git - users/jedix/linux-maple.git/log
users/jedix/linux-maple.git
4 years agomaple_tree: combine_separate reduction with parents and cp_left/cp_right
Liam R. Howlett [Fri, 17 Jul 2020 03:07:31 +0000 (23:07 -0400)]
maple_tree: combine_separate reduction with parents and cp_left/cp_right

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Cleaner check in combine_separate
Liam R. Howlett [Fri, 17 Jul 2020 02:51:43 +0000 (22:51 -0400)]
maple_tree: Cleaner check in combine_separate

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Cleaner check in combine_separate
Liam R. Howlett [Fri, 17 Jul 2020 02:49:01 +0000 (22:49 -0400)]
maple_tree: Cleaner check in combine_separate

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: reduce line count
Liam R. Howlett [Fri, 17 Jul 2020 02:46:16 +0000 (22:46 -0400)]
maple_tree: reduce line count

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: mas_mab_cp sets b_end now
Liam R. Howlett [Fri, 17 Jul 2020 02:32:32 +0000 (22:32 -0400)]
maple_tree: mas_mab_cp sets b_end now

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Attempt to reduce split function size
Liam R. Howlett [Fri, 17 Jul 2020 02:09:28 +0000 (22:09 -0400)]
maple_tree: Attempt to reduce split function size

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip, try the maple_subtree_state for smaller functions
Liam R. Howlett [Fri, 17 Jul 2020 01:54:36 +0000 (21:54 -0400)]
maple_tree: wip, try the maple_subtree_state for smaller functions

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip, add maple_subtree_state
Liam R. Howlett [Fri, 17 Jul 2020 01:11:03 +0000 (21:11 -0400)]
maple_tree: wip, add maple_subtree_state

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip, add maple_subtree_state
Liam R. Howlett [Fri, 17 Jul 2020 01:08:53 +0000 (21:08 -0400)]
maple_tree: wip, add maple_subtree_state

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Code cleanup, mostly in split
Liam R. Howlett [Thu, 16 Jul 2020 15:59:40 +0000 (11:59 -0400)]
maple_tree: Code cleanup, mostly in split

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Start using b_node->b_end in mas_split
Liam R. Howlett [Thu, 16 Jul 2020 15:50:27 +0000 (11:50 -0400)]
maple_tree: Start using b_node->b_end in mas_split

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Remove fixed array size from mas_split and use ma_topiary instead.
Liam R. Howlett [Thu, 16 Jul 2020 15:05:44 +0000 (11:05 -0400)]
maple_tree: Remove fixed array size from mas_split and use ma_topiary instead.

Also, move common smp_wmb() from split and combine_separate function into a helper.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Code reduction
Liam R. Howlett [Thu, 16 Jul 2020 02:08:51 +0000 (22:08 -0400)]
maple_tree: Code reduction

also move __mas_walk, mas_rewind_node, move mas_skip_node

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: mas_combine_separate changes.
Liam R. Howlett [Sat, 11 Jul 2020 00:48:44 +0000 (20:48 -0400)]
maple_tree: mas_combine_separate changes.

Fix the calculation of node splitting when needing 3 nodes.

Change allocation count, spanning store may cause more allocations if
rebalancing across nodes occurs.

Remove free/destroy lists and use a linked list.

Weight reduction of the function.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Change debug for maple tree
Liam R. Howlett [Sat, 11 Jul 2020 00:48:21 +0000 (20:48 -0400)]
mm/mmap: Change debug for maple tree

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Remove debug
Liam R. Howlett [Fri, 10 Jul 2020 19:07:50 +0000 (15:07 -0400)]
maple_tree: Remove debug

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Add freeing of nodes.
Liam R. Howlett [Tue, 7 Jul 2020 15:26:19 +0000 (11:26 -0400)]
maple_tree: Add freeing of nodes.

Node freeing is tracked by what was consumed in the previous step and
freed along with what was overwritten in the spanning store that must be
destroyed.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip, fix underrun on root expand when range starts at zero but ends elsewhere
Liam R. Howlett [Fri, 3 Jul 2020 01:34:43 +0000 (21:34 -0400)]
maple_tree: wip, fix underrun on root expand when range starts at zero but ends elsewhere

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip, fix splitting into 3, fix range of combine_separate, fix test for...
Liam R. Howlett [Fri, 3 Jul 2020 01:03:51 +0000 (21:03 -0400)]
maple_tree: wip, fix splitting into 3, fix range of combine_separate, fix test for gap_combining.

Splitting into 3 was not functioning due to an error in setting the parent and the split calc.

combine_separate was not copying the contents of the mas state over to orig_l_mas which became a problem with the limit checking the end of data.

gap_combining tests were rendered broken due to splitting changes with a NULL ending.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip. Combine mas_extend_null in mas_store and
Liam R. Howlett [Wed, 1 Jul 2020 01:32:52 +0000 (21:32 -0400)]
maple_tree: wip.  Combine mas_extend_null in mas_store and
mas_spanning_store.

Pass: 13130626 Run:13130627

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip, maple_tree: Pass 824089 - rebalance and spanning store using most of the same...
Liam R. Howlett [Tue, 30 Jun 2020 17:30:06 +0000 (13:30 -0400)]
wip, maple_tree: Pass 824089 - rebalance and spanning store using most of the same code now

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip, maple_tree: Pass 850785 - added check for min node size. need to rebalance...
Liam R. Howlett [Thu, 25 Jun 2020 23:13:24 +0000 (19:13 -0400)]
wip, maple_tree: Pass 850785 - added check for min node size.  need to rebalance spanning store

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: maple_tree: factor out descend and node adoption
Liam R. Howlett [Thu, 25 Jun 2020 01:41:15 +0000 (21:41 -0400)]
wip: maple_tree: factor out descend and node adoption

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: maple_tree: factor out descend and node adoption
Liam R. Howlett [Thu, 25 Jun 2020 01:40:53 +0000 (21:40 -0400)]
wip: maple_tree: factor out descend and node adoption

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: pass 747201. Rework some gap stuff.
Liam R. Howlett [Wed, 24 Jun 2020 14:13:14 +0000 (10:13 -0400)]
wip: pass 747201.  Rework some gap stuff.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: pass 747114 - fix r_mas.max when there is no r_mas.max for skipping data
Liam R. Howlett [Wed, 24 Jun 2020 00:35:09 +0000 (20:35 -0400)]
wip: pass 747114 - fix r_mas.max when there is no r_mas.max for skipping data

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: Add height reset to destroy and debug output.
Liam R. Howlett [Wed, 24 Jun 2020 00:10:56 +0000 (20:10 -0400)]
wip: Add height reset to destroy and debug output.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: pass 678642 tc
Liam R. Howlett [Wed, 24 Jun 2020 00:05:25 +0000 (20:05 -0400)]
wip: pass 678642 tc

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: wip, initial rebalance on regular commit, fixes for spanning store, adopt...
Liam R. Howlett [Thu, 18 Jun 2020 18:14:45 +0000 (14:14 -0400)]
maple_tree: wip, initial rebalance on regular commit, fixes for spanning store, adoption, gaps..

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip: spanning store & signed full_cnt
Liam R. Howlett [Fri, 12 Jun 2020 14:57:22 +0000 (10:57 -0400)]
wip: spanning store & signed full_cnt

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: WIP, start passing big data through for all stores
Liam R. Howlett [Thu, 11 Jun 2020 01:13:38 +0000 (21:13 -0400)]
maple_tree: WIP, start passing big data through for all stores

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: WIP, diet
Liam R. Howlett [Tue, 9 Jun 2020 20:23:21 +0000 (16:23 -0400)]
maple_tree: WIP, diet

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: WIP, partial spanning store code.
Liam R. Howlett [Tue, 9 Jun 2020 20:11:57 +0000 (16:11 -0400)]
maple_tree: WIP, partial spanning store code.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: 3 wip
Liam R. Howlett [Tue, 2 Jun 2020 20:15:28 +0000 (16:15 -0400)]
maple_tree: 3 wip

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agowip, broken and abandoning this plan for no retry
Liam R. Howlett [Wed, 20 May 2020 22:28:28 +0000 (18:28 -0400)]
wip, broken and abandoning this plan for no retry

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Remove mt_is_advanced in favor of xa_is_advanced.
Liam R. Howlett [Wed, 20 May 2020 17:22:16 +0000 (13:22 -0400)]
maple_tree: Remove mt_is_advanced in favor of xa_is_advanced.

The two functions are now the same so just use the xa_ version.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agoxarray: Change range of xa_is_advanced
Liam R. Howlett [Wed, 20 May 2020 17:14:47 +0000 (13:14 -0400)]
xarray: Change range of xa_is_advanced

XA_ZERO_ENTRY is the is largest advanced internal entry.
Return true if an internal entry is less than the XA_ZERO_ENTRY.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix gap issues and combining issue
Liam R. Howlett [Tue, 19 May 2020 18:57:46 +0000 (14:57 -0400)]
maple_tree: Fix gap issues and combining issue

Incorrect gap calculation caused by mas_prev not setting the ma_state
limits correctly.

Gap relocation failed due to fully empty node (deleted)

Gap relocation issues when hitting a retry entry

Missing data due to combining nodes incorrectly when the right node has
a retry at the start.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree and mm/mmap: Try to stop tracepoints on BUG_ON
Liam R. Howlett [Sat, 9 May 2020 01:02:27 +0000 (21:02 -0400)]
maple_tree and mm/mmap: Try to stop tracepoints on BUG_ON

Restore printk's until we know it happens.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agotrace/events/mm_mt.h: Fix prototype for __vma_mt_szero
Liam R. Howlett [Fri, 8 May 2020 15:29:55 +0000 (11:29 -0400)]
trace/events/mm_mt.h: Fix prototype for __vma_mt_szero

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agotrace/events/mm_mt: Add maple tree mm tracepoints.
Liam R. Howlett [Fri, 8 May 2020 14:14:25 +0000 (10:14 -0400)]
trace/events/mm_mt: Add maple tree mm tracepoints.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agotrace/events/mm rename to mm_mt
Liam R. Howlett [Fri, 8 May 2020 13:29:45 +0000 (09:29 -0400)]
trace/events/mm rename to mm_mt

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Fix unmapped_area_topdown with align_offset
Liam R. Howlett [Fri, 8 May 2020 02:12:58 +0000 (22:12 -0400)]
mm/mmap: Fix unmapped_area_topdown with align_offset

align_offset should be aligned to align_mask.. apparently.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Move maple tree logging to tracepoints
Liam R. Howlett [Thu, 7 May 2020 21:59:51 +0000 (17:59 -0400)]
mm/mmap: Move maple tree logging to tracepoints

Able to toggle at runtime? Count me in.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Fix maple tree calculation for topdown with alignment mask.
Liam R. Howlett [Thu, 7 May 2020 20:07:49 +0000 (16:07 -0400)]
mm/mmap: Fix maple tree calculation for topdown with alignment mask.

When using alignment mask, calculate the address correctly.

Found by LTP futex_wake04

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Add test for reverse search with mask.
Liam R. Howlett [Thu, 7 May 2020 20:07:17 +0000 (16:07 -0400)]
maple_tree: Add test for reverse search with mask.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Add call to maple_destroy
Liam R. Howlett [Thu, 7 May 2020 01:13:43 +0000 (21:13 -0400)]
mm/mmap: Add call to maple_destroy

When tearing down a vma, destroy the maple tree as well.

Also clean up some old print debugs.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix mas_append of extra null.
Liam R. Howlett [Thu, 7 May 2020 01:12:28 +0000 (21:12 -0400)]
maple_tree: Fix mas_append of extra null.

When attempting to append to a node, it is possible that the source
contains multiple NULL entries that will be coalesced together and thus
will not overflow even though it may appear that was on initial
inspection.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Skip multiple deletes in split calculation.
Liam R. Howlett [Wed, 6 May 2020 18:04:47 +0000 (14:04 -0400)]
maple_tree: Skip multiple deletes in split calculation.

When splitting, do not split within consecutive NULL entries.  Doing so
would cause separated gaps.

Also reset mas->node before walking in two other places which could
potentially cause issue.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Reset mas location before walking in mas_rebalance_gaps.
Liam R. Howlett [Wed, 6 May 2020 03:17:52 +0000 (23:17 -0400)]
maple_tree: Reset mas location before walking in mas_rebalance_gaps.

After rebalancing, the mas node may not be in the correct location to
find the mas->index, so reset to MAS_START prior to starting the walk.

Added test case for this issue as well.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Move VM_BUG_ON inside other check for maple tree
Liam R. Howlett [Wed, 6 May 2020 01:18:56 +0000 (21:18 -0400)]
mm/mmap: Move VM_BUG_ON inside other check for maple tree

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix skip entries on spanning adds and issue in split data
Liam R. Howlett [Wed, 6 May 2020 01:15:23 +0000 (21:15 -0400)]
maple_tree: Fix skip entries on spanning adds and issue in split data

When spanning adds hit a skip entry, they were not updated and thus
future events may use the incorrect minimum from the previous
not-updated pivot.  Fix these pivots when writing spanning adds.

There was also an issue with using the incorrect ranges in the split
code path in mas_append_split_data() which could cause data to be
skipped.  This is now fixed and there are testcases to test for the
discovered event.

The spanning add issue was tested with set25 in check_erase2_testset.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix off-by-one error on mas->min when calling mas_prev.
Liam R. Howlett [Wed, 29 Apr 2020 19:04:29 +0000 (15:04 -0400)]
maple_tree: Fix off-by-one error on mas->min when calling mas_prev.

When using a non-zero slot, the mas->min was set incorrectly for
mas_prev causing the gap calculation to be off-by-one and verification
to fail.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agotest_maple_tree: Add test which discovered the retry limit issues.
Liam R. Howlett [Tue, 21 Apr 2020 19:53:27 +0000 (15:53 -0400)]
test_maple_tree: Add test which discovered the retry limit issues.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix gap on skip entries caused by relocated data.
Liam R. Howlett [Tue, 21 Apr 2020 19:52:31 +0000 (15:52 -0400)]
maple_tree: Fix gap on skip entries caused by relocated data.

The gap should be zero for skip entries.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix overrunning end of node into RETRY entries.
Liam R. Howlett [Tue, 21 Apr 2020 19:50:51 +0000 (15:50 -0400)]
maple_tree: Fix overrunning end of node into RETRY entries.

When entries are moved and replaced with a RETRY, the node will end
prior to hitting a 0 pivot or the end of the node.  Fix this scenario by
detecting the pivot > mas->max and handle these cases correctly.

This also fixes the validation code to avoid checking the range on RETRY
entries.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix limits check with retry entries
Liam R. Howlett [Fri, 17 Apr 2020 19:19:47 +0000 (15:19 -0400)]
maple_tree: Fix limits check with retry entries

When entries are relocated, the retry entry may not fall within the
limits of this node as it has been relocated to the next node.

Also, add define U8_MAX UCHAR_MAX in test code.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agotest_maple_tree: Don't count retry entries during entry counting.
Liam R. Howlett [Fri, 6 Mar 2020 18:59:43 +0000 (13:59 -0500)]
test_maple_tree: Don't count retry entries during entry counting.

retries are skipped, so don't count them so the test framework will come
up with the correct count.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Support freeing empty nodes on gap relocation.
Liam R. Howlett [Fri, 6 Mar 2020 18:45:54 +0000 (13:45 -0500)]
maple_tree: Support freeing empty nodes on gap relocation.

When a gap is created during a low memory situation, coalesce may not be
able to group the gap together.  This can be handled by moving the right
node to the left & using a SKIP entry.  Note that this situation should
not arise unless in a low-memory situation, so a SKIP entry is useful as
it indicates the tree may need to be rebuilt after memory pressure
subsides.

When an empty node is found,

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Zero pivots of gaps that have been moved.
Liam R. Howlett [Wed, 4 Mar 2020 17:00:33 +0000 (12:00 -0500)]
maple_tree: Zero pivots of gaps that have been moved.

This doesn't seem totally correct and should really use a new node.
A solution for low mem is needed too.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Combine gaps with retry entries.
Liam R. Howlett [Tue, 3 Mar 2020 17:03:47 +0000 (12:03 -0500)]
maple_tree: Combine gaps with retry entries.

When a retry entry is hit, combine it into the gap as it was relocated
to where the gap exists now.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Clean up mas_data_end interface.
Liam R. Howlett [Tue, 3 Mar 2020 13:19:30 +0000 (08:19 -0500)]
maple_tree: Clean up mas_data_end interface.

Some places don't care about coalesce and the last pivot, so provide a
cleaner way to get just the end to clean up that code.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: whitespace fix.
Liam R. Howlett [Tue, 3 Mar 2020 13:18:51 +0000 (08:18 -0500)]
maple_tree: whitespace fix.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix rcu_dereference usages
Matthew Wilcox (Oracle) [Mon, 2 Mar 2020 21:16:00 +0000 (16:16 -0500)]
maple_tree: Fix rcu_dereference usages

Enabling CONFIG_PROVE_LOCKING reveals some compilation errors and
running it shows that loading the root pointer needs to be done with
rcu_dereference_protected() (which will permit the load if the lock is
held) instead of a plain rcu_dereference() (which permits the load if
the rcu read lock is held).

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agomaple_tree: no need to rcu_dereference in mtree_empty()
Matthew Wilcox (Oracle) [Mon, 2 Mar 2020 21:09:00 +0000 (16:09 -0500)]
maple_tree: no need to rcu_dereference in mtree_empty()

Since we're only comparing the value of the pointer to NULL, we
don't need to use rcu_dereference().  This is explicitly documented
in Documentation/RCU/rcu_dereference.rst.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agomm: Fix compilation warnings
Matthew Wilcox (Oracle) [Mon, 2 Mar 2020 21:24:19 +0000 (16:24 -0500)]
mm: Fix compilation warnings

Nothing terribly interesting here.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agomaple_tree: Fix lock_dep calls to use pointers.
Liam R. Howlett [Tue, 3 Mar 2020 03:33:28 +0000 (22:33 -0500)]
maple_tree: Fix lock_dep calls to use pointers.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Move variable declaring prior to code.
Liam R. Howlett [Tue, 3 Mar 2020 03:32:23 +0000 (22:32 -0500)]
mm/mmap: Move variable declaring prior to code.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix overwritten slot in mas during __mas_add
Liam R. Howlett [Tue, 3 Mar 2020 03:21:37 +0000 (22:21 -0500)]
maple_tree: Fix overwritten slot in mas during __mas_add

Overwriting the slot in the ma state caused issues later on.  Restore
the value and fix the code in mas_may_move_gap to also detect such
scenarios.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Small code cleanup, white spaces etc.
Liam R. Howlett [Mon, 2 Mar 2020 19:54:06 +0000 (14:54 -0500)]
maple_tree: Small code cleanup, white spaces etc.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agoMAINTAINERS: Add maple tree
Liam R. Howlett [Mon, 2 Mar 2020 18:45:42 +0000 (13:45 -0500)]
MAINTAINERS: Add maple tree

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Add support for combining gaps.
Liam R. Howlett [Mon, 2 Mar 2020 18:31:17 +0000 (13:31 -0500)]
maple_tree: Add support for combining gaps.

When gaps touch but are not in the same node, move them to the right
node.  This allows for searches to find the desired gap.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agoextra debug
Matthew Wilcox (Oracle) [Fri, 28 Feb 2020 20:46:27 +0000 (15:46 -0500)]
extra debug

4 years agomaple_tree: Fix RCU reads to use rcu_dereference_check and writes to use
Liam R. Howlett [Mon, 2 Mar 2020 16:22:53 +0000 (11:22 -0500)]
maple_tree: Fix RCU reads to use rcu_dereference_check and writes to use
rcu_assign_pointer

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix placement of data during add operation during certain
Liam R. Howlett [Mon, 2 Mar 2020 15:43:00 +0000 (10:43 -0500)]
maple_tree: Fix placement of data during add operation during certain
conditions.

When partially overwriting the sources end slot and coalescing data
during the copy operation, there was a potential for an overflow during
a store operation.  Avoid the overflow by using the new node end instead
of the slot passed in as the destination.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm: Get rid of maple tree whitespace changes
Matthew Wilcox (Oracle) [Thu, 27 Feb 2020 19:57:32 +0000 (14:57 -0500)]
mm: Get rid of maple tree whitespace changes

Nothing here makes any difference to the generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agomaple_tree: Clean up debug infrastructure
Matthew Wilcox (Oracle) [Thu, 27 Feb 2020 15:47:07 +0000 (10:47 -0500)]
maple_tree: Clean up debug infrastructure

making tests_run & tests_passed static means there's one copy of them in
every file that includes maple_tree.h.  Make them extern, prefix them
with maple_tree_, move them to maple_tree.c and export them so that
test_maple_tree can be built as a module.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agomaple_tree: Move xa_get_alloc_size
Matthew Wilcox (Oracle) [Thu, 27 Feb 2020 15:16:09 +0000 (10:16 -0500)]
maple_tree: Move xa_get_alloc_size

... into ma_xa_benchmark.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agomm/mmap: Remove maple tree erase call in __vma_unlink_common
Liam R. Howlett [Fri, 28 Feb 2020 20:33:24 +0000 (15:33 -0500)]
mm/mmap: Remove maple tree erase call in __vma_unlink_common

__vma_unlink_common is used when overwriting the next vma entry.  Maple
tree overwrites these entries on the store of the vma itself.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Maple tree update for validate_mm_mt and comments
Liam R. Howlett [Fri, 28 Feb 2020 20:19:07 +0000 (15:19 -0500)]
mm/mmap: Maple tree update for validate_mm_mt and comments

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agommap: Write changes to next to the maple tree.
Liam R. Howlett [Fri, 28 Feb 2020 17:09:32 +0000 (12:09 -0500)]
mmap: Write changes to next to the maple tree.

When mprotect case 4 is hit (expand next back over vma), you have to
write the changes to the maple tree or there will be a missing mapping.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Initial documentation straw man.
Liam R. Howlett [Thu, 27 Feb 2020 03:39:36 +0000 (22:39 -0500)]
maple_tree: Initial documentation straw man.

Just add the file and some headers.  Details need to come from Thoughts.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agoRenumber XA.*ENTRY constants
Matthew Wilcox (Oracle) [Wed, 26 Feb 2020 19:00:43 +0000 (14:00 -0500)]
Renumber XA.*ENTRY constants

Entries below 256 are reserved for siblings, so renumber the RETRY
entry back to 256.  Entries which are ZERO and higher are allowed to
be inserted into the XArray, so renumber ZERO to 259, and slot SKIP and
DELETED in at 257 and 258.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
4 years agoRevert "maple_tree: Rework node split calculation for better non-leaf node split"
Liam R. Howlett [Tue, 25 Feb 2020 16:45:28 +0000 (11:45 -0500)]
Revert "maple_tree: Rework node split calculation for better non-leaf node split"

Efficient splitting is causing issues in certain scenarios with regards
to allocation trees.  Revisit later.

This reverts commit 4a0a2dd623fae22f2823a0711193aafef3479e71.

4 years agomaple_tree: Fix off by one on mas_prev when going to previous node.
Liam R. Howlett [Tue, 25 Feb 2020 16:18:51 +0000 (11:18 -0500)]
maple_tree: Fix off by one on mas_prev when going to previous node.

mas_prev_nentry decrements the slot prior to checking for an entry, so
don't decrement the slot in mas_prev loop on new nodes.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap: Don't adjust next for maple tree operations.
Liam R. Howlett [Tue, 25 Feb 2020 02:36:03 +0000 (21:36 -0500)]
mm/mmap: Don't adjust next for maple tree operations.

As maple tree store operation will alter the start location of the next
node if it they overlap, there is no need to erase and store the same
VMA.  In fact, since the start location is now occupied by the current
VMA, the erase operation will remove the wrong entry.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Set slot in mas_prev_slot to get the pivot
Liam R. Howlett [Tue, 25 Feb 2020 02:35:05 +0000 (21:35 -0500)]
maple_tree: Set slot in mas_prev_slot to get the pivot

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Rework node split calculation for better non-leaf node split
Liam R. Howlett [Mon, 24 Feb 2020 15:14:18 +0000 (10:14 -0500)]
maple_tree: Rework node split calculation for better non-leaf node split
usage.

When calculating where to split non-leaf nodes, assume the minimum span
is 8 in each entry, so 8*8.  This still needs work as we could figure
out the level from a leaf and keep going up in the minimum span to
better utilize more levels of the tree.

There is also the issue of the right-most entry (ULONG_MAX) which ends
up wasting a slot in the linear allocation case.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Always split regardless of spanning node boundaries or not.
Liam R. Howlett [Mon, 24 Feb 2020 15:11:39 +0000 (10:11 -0500)]
maple_tree: Always split regardless of spanning node boundaries or not.

Spanning node boundaries are now considered when calculating the number
of elements in a node, so if the count is larger than what the node will
hold, the node needs to be split.

set20 in test_maple_tree caught this issue.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix mas_prev issue with parent slot & loop at root.
Liam R. Howlett [Sat, 22 Feb 2020 04:00:32 +0000 (23:00 -0500)]
maple_tree: Fix mas_prev issue with parent slot & loop at root.

parent slots were not being set correctly for the first ascend.  This
caused data to be skipped.

There was also an infinite loop if this was the first entry in a deep
tree.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix finding the end of a node when the node ends in a NULL.
Liam R. Howlett [Sat, 22 Feb 2020 02:49:55 +0000 (21:49 -0500)]
maple_tree: Fix finding the end of a node when the node ends in a NULL.
Also fix the limit calculation on gaps when searches backwards goes into
adjacent nodes during the search.

Ignoring the end null may cause gaps to be skipped when searching
backwards.

Adjacent nodes caused the max limit to be reset to the parent but never
set if the very next entry also had a potential gap that would work.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix limit updating when rewinding nodes during reverse
Liam R. Howlett [Fri, 21 Feb 2020 18:19:21 +0000 (13:19 -0500)]
maple_tree: Fix limit updating when rewinding nodes during reverse
walks.

When walking backwards in a node that sat in slot 0 of the parent, the
limit would not be set correctly.  The fix is not to set the limits
until descending into the node.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap and mm/mprotect: Alter mmap erase method.
Liam R. Howlett [Thu, 20 Feb 2020 21:39:36 +0000 (16:39 -0500)]
mm/mmap and mm/mprotect: Alter mmap erase method.

Can't call erase every time the rbtree erases as the maple tree doesn't
allow overlaps, so it's not necessary.

mprotect was too noisy.

mmap also had debug altered and verify changed in this commit.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix off-by-one on append_entry
Liam R. Howlett [Thu, 20 Feb 2020 21:36:16 +0000 (16:36 -0500)]
maple_tree: Fix off-by-one on append_entry

When appending an entry, ensure that the previous pivot is one less than
the desired range.

Don't subtraction coalesce from the slot calculation as that is now
taken into account within  _mas_add_slot_cnt itself.

Fix testing to kick out if mas_for_each is in a retry loop.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agotest_maple_tree: Standardize tests on using what was stored into the
Liam R. Howlett [Thu, 20 Feb 2020 20:08:30 +0000 (15:08 -0500)]
test_maple_tree: Standardize tests on using what was stored into the
maple tree

Use [x,y], not [x,y).

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix u64 => unsigned long
Liam R. Howlett [Thu, 20 Feb 2020 20:06:48 +0000 (15:06 -0500)]
maple_tree: Fix u64 => unsigned long

Incorrect types fixed.  This will need to be revisited for different
arch and also u16/u32 one day.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Fix append_entry for NULL appending, fix append for retry
Liam R. Howlett [Thu, 20 Feb 2020 16:28:50 +0000 (11:28 -0500)]
maple_tree: Fix append_entry for NULL appending, fix append for retry
entries, fix split appending on coalesce.

Add some testcases for the above fixes.

Also fix an off-by-one error in spanning_add.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomm/mmap and maple_tree: Fix mt_for_each interface, get_unmapped_area_rev
Liam R. Howlett [Tue, 18 Feb 2020 15:56:41 +0000 (10:56 -0500)]
mm/mmap and maple_tree: Fix mt_for_each interface, get_unmapped_area_rev

mt_for_each interface simplification for external users.

get_unampped_area_rev was not getting the right answer.  Fix it and
clean up the condensed and unused interface internal to the maple tree.

Fix associated tests as well.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agommap/fork/mm.h: Add maple tree support.
Liam R. Howlett [Fri, 14 Feb 2020 17:04:34 +0000 (12:04 -0500)]
mmap/fork/mm.h: Add maple tree support.

When forking, init the maple tree and add entries.

mmap: Clean up prints.

mm: Export vma_store for adding things to mt in fork.

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
4 years agomaple_tree: Thoughts, fix whitespace, fix compile issue
Liam R. Howlett [Fri, 14 Feb 2020 16:39:56 +0000 (11:39 -0500)]
maple_tree: Thoughts, fix whitespace, fix compile issue

Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>