[PATCH v2 07/33] mm: Enable maple tree RCU mode by default.

[PATCH v2 00/33] Per-VMA locks · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 02/33] maple_tree: Detect dead nodes in mas_start() · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 05/33] maple_tree: Fix write memory barrier of nodes once dead for RCU mode · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 04/33] maple_tree: remove extra smp_wmb() from mas_dead_leaves() · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 06/33] maple_tree: Add smp_rmb() to dead node detection · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 07/33] mm: Enable maple tree RCU mode by default. · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 09/33] mm: rcu safe VMA freeing · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 10/33] mm: move mmap_lock assert function definitions · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 11/33] mm: add per-VMA lock and helper functions to control it · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 12/33] mm: mark VMA as being written when changing vm_flags · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 13/33] mm/mmap: move VMA locking before vma_adjust_trans_huge call · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 14/33] mm/khugepaged: write-lock VMA while collapsing a huge page · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 08/33] mm: introduce CONFIG_PER_VMA_LOCK · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 15/33] mm/mmap: write-lock VMAs before merging, splitting or expanding them · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 19/33] mm: conditionally write-lock VMA in free_pgtables · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 17/33] mm/mremap: write-lock VMA while remapping it to a new address range · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 22/33] mm/mmap: prevent pagefault handler from racing with mmu_notifier registration · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 20/33] mm/mmap: write-lock adjacent VMAs if they can grow into unmapped area · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 21/33] kernel/fork: assert no VMA readers during its destruction · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 23/33] mm: introduce lock_vma_under_rcu to be used from arch-specific code · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 27/33] mm: prevent userfaults to be handled under per-vma lock · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 28/33] mm: introduce per-VMA lock statistics · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 25/33] mm: add FAULT_FLAG_VMA_LOCK flag · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 29/33] x86/mm: try VMA lock-based page fault handling first · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 24/33] mm: fall back to mmap_lock if vma->anon_vma is not yet set · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 16/33] mm/mmap: write-lock VMA before shrinking or expanding it · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 31/33] powerc/mm: try VMA lock-based page fault handling first · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 32/33] mm/mmap: free vm_area_struct without call_rcu in exit_mmap · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 26/33] mm: prevent do_swap_page from handling page faults under VMA lock · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 30/33] arm64/mm: try VMA lock-based page fault handling first · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 33/33] mm: separate vma->lock from vm_area_struct · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 03/33] maple_tree: Fix freeing of nodes in rcu mode · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 18/33] mm: write-lock VMAs before removing them from VMA tree · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
[PATCH v2 01/33] maple_tree: Be more cautious about dead nodes · Suren Baghdasaryan <surenb@google.com> · 2023-01-27
Re: [PATCH v2 00/33] Per-VMA locks · Andrew Morton <akpm@linux-foundation.org> · 2023-01-27
Re: [PATCH v2 00/33] Per-VMA locks · Matthew Wilcox <willy@infradead.org> · 2023-01-27
Re: [PATCH v2 00/33] Per-VMA locks · Suren Baghdasaryan <surenb@google.com> · 2023-01-28
Re: [PATCH v2 00/33] Per-VMA locks · Suren Baghdasaryan <surenb@google.com> · 2023-02-14

STALE1201d

Revisions (4)

2023-01-09 v1 [diff vs current]
2023-01-27 v2 current
2023-02-16 v3 [diff vs current]
2023-02-27 v4 [diff vs current]

From: Suren Baghdasaryan <surenb@google.com>
Date: 2023-01-27 19:43:27
Also in: linux-arm-kernel, linux-mm, lkml
Subsystem: exec & binfmt api, elf, memory management, memory management - core, memory mapping, scheduler, the rest · Maintainers: Kees Cook, Andrew Morton, David Hildenbrand, Liam R. Howlett, Lorenzo Stoakes, Ingo Molnar, Peter Zijlstra, Juri Lelli, Vincent Guittot, Linus Torvalds

From: "Liam R. Howlett" <redacted>

Use the maple tree in RCU mode for VMA tracking.  This is necessary for
the use of per-VMA locking.  RCU mode is enabled by default but disabled
when exiting an mm and for the new tree during a fork.

Also enable RCU for the tree used in munmap operations to ensure the
nodes remain valid for readers.

Signed-off-by: Liam R. Howlett <redacted>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
---
 include/linux/mm_types.h | 3 ++-
 kernel/fork.c            | 3 +++
 mm/mmap.c                | 4 +++-
 3 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index da983aedb741..8410c3052148 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h

@@ -768,7 +768,8 @@ struct mm_struct {
 	unsigned long cpu_bitmap[];
 };
 
-#define MM_MT_FLAGS	(MT_FLAGS_ALLOC_RANGE | MT_FLAGS_LOCK_EXTERN)
+#define MM_MT_FLAGS	(MT_FLAGS_ALLOC_RANGE | MT_FLAGS_LOCK_EXTERN | \
+			 MT_FLAGS_USE_RCU)
 extern struct mm_struct init_mm;
 
 /* Pointer magic because the dynamic array size confuses some compilers. */

diff --git a/kernel/fork.c b/kernel/fork.c
index 5e3029ea8e1e..5f23d5e03362 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c

@@ -617,6 +617,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
 	if (retval)
 		goto out;
 
+	mt_clear_in_rcu(vmi.mas.tree);
 	for_each_vma(old_vmi, mpnt) {
 		struct file *file;

@@ -700,6 +701,8 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
 	retval = arch_dup_mmap(oldmm, mm);
 loop_out:
 	vma_iter_free(&vmi);
+	if (!retval)
+		mt_set_in_rcu(vmi.mas.tree);
 out:
 	mmap_write_unlock(mm);
 	flush_tlb_mm(oldmm);

diff --git a/mm/mmap.c b/mm/mmap.c
index 9d0dadb55e9a..9efe13d36df7 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c

@@ -2277,7 +2277,8 @@ do_vmi_align_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
 	int count = 0;
 	int error = -ENOMEM;
 	MA_STATE(mas_detach, &mt_detach, 0, 0);
-	mt_init_flags(&mt_detach, MT_FLAGS_LOCK_EXTERN);
+	mt_init_flags(&mt_detach, vmi->mas.tree->ma_flags &
+		      (MT_FLAGS_LOCK_MASK | MT_FLAGS_USE_RCU));
 	mt_set_external_lock(&mt_detach, &mm->mmap_lock);
 
 	/*

@@ -3042,6 +3043,7 @@ void exit_mmap(struct mm_struct *mm)
 	 */
 	set_bit(MMF_OOM_SKIP, &mm->flags);
 	mmap_write_lock(mm);
+	mt_clear_in_rcu(&mm->mm_mt);
 	free_pgtables(&tlb, &mm->mm_mt, vma, FIRST_USER_ADDRESS,
 		      USER_PGTABLES_CEILING);
 	tlb_finish_mmu(&tlb);

-- 
2.39.1

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help