From: "Liam R. Howlett" <Liam.Howlett@oracle.com>
To: Linus Torvalds <torvalds@linux-foundation.org>,
Andrew Morton <akpm@linux-foundation.org>,
maple-tree@lists.infradead.org, linux-mm@kvack.org,
linux-kernel@vger.kernel.org,
Matthew Wilcox <willy@infradead.org>,
Suren Baghdasaryan <surenb@google.com>
Cc: "Liam R. Howlett" <Liam.Howlett@Oracle.com>,
stable@vger.kernel.org,
"Liam R . Howlett" <Liam.Howlett@oracle.com>
Subject: [PATCH 7/8] maple_tree: add RCU lock checking to rcu callback functions
Date: Mon, 27 Mar 2023 14:55:31 -0400 [thread overview]
Message-ID: <20230327185532.2354250-8-Liam.Howlett@oracle.com> (raw)
In-Reply-To: <20230327185532.2354250-1-Liam.Howlett@oracle.com>
From: "Liam R. Howlett" <Liam.Howlett@Oracle.com>
Dereferencing RCU objects within the RCU callback without the RCU check
has caused lockdep to complain. Fix the RCU dereferencing by using the
RCU callback lock to ensure the operation is safe.
Also stop creating a new lock to use for dereferencing during destruction
of the tree or subtree. Instead, pass through a pointer to the tree that
has the lock that is held for RCU dereferencing checking. It also does
not make sense to use the maple state in the freeing scenario as the tree
walk is a special case where the tree no longer has the normal encodings
and parent pointers.
Link: https://lkml.kernel.org/r/20230227173632.3292573-8-surenb@google.com
Fixes: 54a611b60590 ("Maple Tree: add new data structure")
Cc: stable@vger.kernel.org
Reported-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Liam R. Howlett <Liam.Howlett@oracle.com>
---
lib/maple_tree.c | 188 ++++++++++++++++++++++++-----------------------
1 file changed, 96 insertions(+), 92 deletions(-)
diff --git a/lib/maple_tree.c b/lib/maple_tree.c
index 8ad2d1669fad..2be86368237d 100644
--- a/lib/maple_tree.c
+++ b/lib/maple_tree.c
@@ -824,6 +824,11 @@ static inline void *mt_slot(const struct maple_tree *mt,
return rcu_dereference_check(slots[offset], mt_locked(mt));
}
+static inline void *mt_slot_locked(struct maple_tree *mt, void __rcu **slots,
+ unsigned char offset)
+{
+ return rcu_dereference_protected(slots[offset], mt_locked(mt));
+}
/*
* mas_slot_locked() - Get the slot value when holding the maple tree lock.
* @mas: The maple state
@@ -835,7 +840,7 @@ static inline void *mt_slot(const struct maple_tree *mt,
static inline void *mas_slot_locked(struct ma_state *mas, void __rcu **slots,
unsigned char offset)
{
- return rcu_dereference_protected(slots[offset], mt_locked(mas->tree));
+ return mt_slot_locked(mas->tree, slots, offset);
}
/*
@@ -907,34 +912,35 @@ static inline void ma_set_meta(struct maple_node *mn, enum maple_type mt,
}
/*
- * mas_clear_meta() - clear the metadata information of a node, if it exists
- * @mas: The maple state
+ * mt_clear_meta() - clear the metadata information of a node, if it exists
+ * @mt: The maple tree
* @mn: The maple node
- * @mt: The maple node type
+ * @type: The maple node type
* @offset: The offset of the highest sub-gap in this node.
* @end: The end of the data in this node.
*/
-static inline void mas_clear_meta(struct ma_state *mas, struct maple_node *mn,
- enum maple_type mt)
+static inline void mt_clear_meta(struct maple_tree *mt, struct maple_node *mn,
+ enum maple_type type)
{
struct maple_metadata *meta;
unsigned long *pivots;
void __rcu **slots;
void *next;
- switch (mt) {
+ switch (type) {
case maple_range_64:
pivots = mn->mr64.pivot;
if (unlikely(pivots[MAPLE_RANGE64_SLOTS - 2])) {
slots = mn->mr64.slot;
- next = mas_slot_locked(mas, slots,
- MAPLE_RANGE64_SLOTS - 1);
- if (unlikely((mte_to_node(next) && mte_node_type(next))))
- return; /* The last slot is a node, no metadata */
+ next = mt_slot_locked(mt, slots,
+ MAPLE_RANGE64_SLOTS - 1);
+ if (unlikely((mte_to_node(next) &&
+ mte_node_type(next))))
+ return; /* no metadata, could be node */
}
fallthrough;
case maple_arange_64:
- meta = ma_meta(mn, mt);
+ meta = ma_meta(mn, type);
break;
default:
return;
@@ -5497,7 +5503,7 @@ static inline int mas_rev_alloc(struct ma_state *mas, unsigned long min,
}
/*
- * mas_dead_leaves() - Mark all leaves of a node as dead.
+ * mte_dead_leaves() - Mark all leaves of a node as dead.
* @mas: The maple state
* @slots: Pointer to the slot array
* @type: The maple node type
@@ -5507,16 +5513,16 @@ static inline int mas_rev_alloc(struct ma_state *mas, unsigned long min,
* Return: The number of leaves marked as dead.
*/
static inline
-unsigned char mas_dead_leaves(struct ma_state *mas, void __rcu **slots,
- enum maple_type mt)
+unsigned char mte_dead_leaves(struct maple_enode *enode, struct maple_tree *mt,
+ void __rcu **slots)
{
struct maple_node *node;
enum maple_type type;
void *entry;
int offset;
- for (offset = 0; offset < mt_slots[mt]; offset++) {
- entry = mas_slot_locked(mas, slots, offset);
+ for (offset = 0; offset < mt_slot_count(enode); offset++) {
+ entry = mt_slot(mt, slots, offset);
type = mte_node_type(entry);
node = mte_to_node(entry);
/* Use both node and type to catch LE & BE metadata */
@@ -5531,162 +5537,160 @@ unsigned char mas_dead_leaves(struct ma_state *mas, void __rcu **slots,
return offset;
}
-static void __rcu **mas_dead_walk(struct ma_state *mas, unsigned char offset)
+/**
+ * mte_dead_walk() - Walk down a dead tree to just before the leaves
+ * @enode: The maple encoded node
+ * @offset: The starting offset
+ *
+ * Note: This can only be used from the RCU callback context.
+ */
+static void __rcu **mte_dead_walk(struct maple_enode **enode, unsigned char offset)
{
- struct maple_node *next;
+ struct maple_node *node, *next;
void __rcu **slots = NULL;
- next = mas_mn(mas);
+ next = mte_to_node(*enode);
do {
- mas->node = mt_mk_node(next, next->type);
- slots = ma_slots(next, next->type);
- next = mas_slot_locked(mas, slots, offset);
+ *enode = ma_enode_ptr(next);
+ node = mte_to_node(*enode);
+ slots = ma_slots(node, node->type);
+ next = rcu_dereference_protected(slots[offset],
+ lock_is_held(&rcu_callback_map));
offset = 0;
} while (!ma_is_leaf(next->type));
return slots;
}
+/**
+ * mt_free_walk() - Walk & free a tree in the RCU callback context
+ * @head: The RCU head that's within the node.
+ *
+ * Note: This can only be used from the RCU callback context.
+ */
static void mt_free_walk(struct rcu_head *head)
{
void __rcu **slots;
struct maple_node *node, *start;
- struct maple_tree mt;
+ struct maple_enode *enode;
unsigned char offset;
enum maple_type type;
- MA_STATE(mas, &mt, 0, 0);
node = container_of(head, struct maple_node, rcu);
if (ma_is_leaf(node->type))
goto free_leaf;
- mt_init_flags(&mt, node->ma_flags);
- mas_lock(&mas);
start = node;
- mas.node = mt_mk_node(node, node->type);
- slots = mas_dead_walk(&mas, 0);
- node = mas_mn(&mas);
+ enode = mt_mk_node(node, node->type);
+ slots = mte_dead_walk(&enode, 0);
+ node = mte_to_node(enode);
do {
mt_free_bulk(node->slot_len, slots);
offset = node->parent_slot + 1;
- mas.node = node->piv_parent;
- if (mas_mn(&mas) == node)
- goto start_slots_free;
-
- type = mte_node_type(mas.node);
- slots = ma_slots(mte_to_node(mas.node), type);
- if ((offset < mt_slots[type]) && (slots[offset]))
- slots = mas_dead_walk(&mas, offset);
-
- node = mas_mn(&mas);
+ enode = node->piv_parent;
+ if (mte_to_node(enode) == node)
+ goto free_leaf;
+
+ type = mte_node_type(enode);
+ slots = ma_slots(mte_to_node(enode), type);
+ if ((offset < mt_slots[type]) &&
+ rcu_dereference_protected(slots[offset],
+ lock_is_held(&rcu_callback_map)))
+ slots = mte_dead_walk(&enode, offset);
+ node = mte_to_node(enode);
} while ((node != start) || (node->slot_len < offset));
slots = ma_slots(node, node->type);
mt_free_bulk(node->slot_len, slots);
-start_slots_free:
- mas_unlock(&mas);
free_leaf:
mt_free_rcu(&node->rcu);
}
-static inline void __rcu **mas_destroy_descend(struct ma_state *mas,
- struct maple_enode *prev, unsigned char offset)
+static inline void __rcu **mte_destroy_descend(struct maple_enode **enode,
+ struct maple_tree *mt, struct maple_enode *prev, unsigned char offset)
{
struct maple_node *node;
- struct maple_enode *next = mas->node;
+ struct maple_enode *next = *enode;
void __rcu **slots = NULL;
+ enum maple_type type;
+ unsigned char next_offset = 0;
do {
- mas->node = next;
- node = mas_mn(mas);
- slots = ma_slots(node, mte_node_type(mas->node));
- next = mas_slot_locked(mas, slots, 0);
- if ((mte_dead_node(next))) {
- mte_to_node(next)->type = mte_node_type(next);
- next = mas_slot_locked(mas, slots, 1);
- }
+ *enode = next;
+ node = mte_to_node(*enode);
+ type = mte_node_type(*enode);
+ slots = ma_slots(node, type);
+ next = mt_slot_locked(mt, slots, next_offset);
+ if ((mte_dead_node(next)))
+ next = mt_slot_locked(mt, slots, ++next_offset);
- mte_set_node_dead(mas->node);
- node->type = mte_node_type(mas->node);
- mas_clear_meta(mas, node, node->type);
+ mte_set_node_dead(*enode);
+ node->type = type;
node->piv_parent = prev;
node->parent_slot = offset;
- offset = 0;
- prev = mas->node;
+ offset = next_offset;
+ next_offset = 0;
+ prev = *enode;
} while (!mte_is_leaf(next));
return slots;
}
-static void mt_destroy_walk(struct maple_enode *enode, unsigned char ma_flags,
+static void mt_destroy_walk(struct maple_enode *enode, struct maple_tree *mt,
bool free)
{
void __rcu **slots;
struct maple_node *node = mte_to_node(enode);
struct maple_enode *start;
- struct maple_tree mt;
-
- MA_STATE(mas, &mt, 0, 0);
- mas.node = enode;
if (mte_is_leaf(enode)) {
node->type = mte_node_type(enode);
goto free_leaf;
}
- ma_flags &= ~MT_FLAGS_LOCK_MASK;
- mt_init_flags(&mt, ma_flags);
- mas_lock(&mas);
-
- mte_to_node(enode)->ma_flags = ma_flags;
start = enode;
- slots = mas_destroy_descend(&mas, start, 0);
- node = mas_mn(&mas);
+ slots = mte_destroy_descend(&enode, mt, start, 0);
+ node = mte_to_node(enode); // Updated in the above call.
do {
enum maple_type type;
unsigned char offset;
struct maple_enode *parent, *tmp;
- node->type = mte_node_type(mas.node);
- node->slot_len = mas_dead_leaves(&mas, slots, node->type);
+ node->slot_len = mte_dead_leaves(enode, mt, slots);
if (free)
mt_free_bulk(node->slot_len, slots);
offset = node->parent_slot + 1;
- mas.node = node->piv_parent;
- if (mas_mn(&mas) == node)
- goto start_slots_free;
+ enode = node->piv_parent;
+ if (mte_to_node(enode) == node)
+ goto free_leaf;
- type = mte_node_type(mas.node);
- slots = ma_slots(mte_to_node(mas.node), type);
+ type = mte_node_type(enode);
+ slots = ma_slots(mte_to_node(enode), type);
if (offset >= mt_slots[type])
goto next;
- tmp = mas_slot_locked(&mas, slots, offset);
+ tmp = mt_slot_locked(mt, slots, offset);
if (mte_node_type(tmp) && mte_to_node(tmp)) {
- parent = mas.node;
- mas.node = tmp;
- slots = mas_destroy_descend(&mas, parent, offset);
+ parent = enode;
+ enode = tmp;
+ slots = mte_destroy_descend(&enode, mt, parent, offset);
}
next:
- node = mas_mn(&mas);
- } while (start != mas.node);
+ node = mte_to_node(enode);
+ } while (start != enode);
- node = mas_mn(&mas);
- node->type = mte_node_type(mas.node);
- node->slot_len = mas_dead_leaves(&mas, slots, node->type);
+ node = mte_to_node(enode);
+ node->slot_len = mte_dead_leaves(enode, mt, slots);
if (free)
mt_free_bulk(node->slot_len, slots);
-start_slots_free:
- mas_unlock(&mas);
-
free_leaf:
if (free)
mt_free_rcu(&node->rcu);
else
- mas_clear_meta(&mas, node, node->type);
+ mt_clear_meta(mt, node, node->type);
}
/*
@@ -5702,10 +5706,10 @@ static inline void mte_destroy_walk(struct maple_enode *enode,
struct maple_node *node = mte_to_node(enode);
if (mt_in_rcu(mt)) {
- mt_destroy_walk(enode, mt->ma_flags, false);
+ mt_destroy_walk(enode, mt, false);
call_rcu(&node->rcu, mt_free_walk);
} else {
- mt_destroy_walk(enode, mt->ma_flags, true);
+ mt_destroy_walk(enode, mt, true);
}
}
--
2.39.2
next prev parent reply other threads:[~2023-03-27 19:07 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-27 18:55 [PATCH 0/8] Fix VMA tree modification under mmap read lock Liam R. Howlett
2023-03-27 18:55 ` [PATCH 1/8] maple_tree: be more cautious about dead nodes Liam R. Howlett
2023-03-27 18:55 ` [PATCH 2/8] maple_tree: detect dead nodes in mas_start() Liam R. Howlett
2023-03-27 18:55 ` [PATCH 3/8] maple_tree: fix freeing of nodes in rcu mode Liam R. Howlett
2023-03-27 18:55 ` [PATCH 4/8] maple_tree: remove extra smp_wmb() from mas_dead_leaves() Liam R. Howlett
2023-03-27 18:55 ` [PATCH 5/8] maple_tree: fix write memory barrier of nodes once dead for RCU mode Liam R. Howlett
2023-03-27 19:05 ` Liam R. Howlett
2023-03-27 19:45 ` Andrew Morton
2023-03-27 18:55 ` [PATCH 6/8] maple_tree: add smp_rmb() to dead node detection Liam R. Howlett
2023-03-27 18:55 ` Liam R. Howlett [this message]
2023-03-27 18:55 ` [PATCH 8/8] mm: enable maple tree RCU mode by default Liam R. Howlett
2023-03-27 19:38 ` Andrew Morton
2023-03-27 19:43 ` Liam R. Howlett
2023-04-11 1:25 ` kernel test robot
2023-04-11 2:25 ` Matthew Wilcox
2023-03-27 19:35 ` [PATCH 0/8] Fix VMA tree modification under mmap read lock Andrew Morton
2023-03-27 19:48 ` Liam R. Howlett
2023-03-28 9:10 ` Vlastimil Babka
2023-03-28 13:02 ` Liam R. Howlett
2023-04-03 19:44 ` Liam R. Howlett
2023-04-03 20:19 ` Andrew Morton
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230327185532.2354250-8-Liam.Howlett@oracle.com \
--to=liam.howlett@oracle.com \
--cc=akpm@linux-foundation.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=maple-tree@lists.infradead.org \
--cc=stable@vger.kernel.org \
--cc=surenb@google.com \
--cc=torvalds@linux-foundation.org \
--cc=willy@infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox