On Tue, Aug 26, 2025 at 8:11 AM Liam R. Howlett <liam.howl...@oracle.com> wrote:
>
> * Suren Baghdasaryan <sur...@google.com> [250822 16:25]:
> > On Wed, Jul 23, 2025 at 6:35 AM Vlastimil Babka <vba...@suse.cz> wrote:
> > >
> > > From: "Liam R. Howlett" <liam.howl...@oracle.com>
> > >
> > > The fast path through a write will require replacing a single node in
> > > the tree. Using a sheaf (32 nodes) is too heavy for the fast path, so
> > > special case the node store operation by just allocating one node in the
> > > maple state.
> > >
> > > Signed-off-by: Liam R. Howlett <liam.howl...@oracle.com>
> > > Signed-off-by: Vlastimil Babka <vba...@suse.cz>
> > > ---
> > > include/linux/maple_tree.h | 4 +++-
> > > lib/maple_tree.c | 47
> > > ++++++++++++++++++++++++++++++++++++++++------
> > > 2 files changed, 44 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/include/linux/maple_tree.h b/include/linux/maple_tree.h
> > > index
> > > 3cf1ae9dde7ce43fa20ae400c01fefad048c302e..61eb5e7d09ad0133978e3ac4b2af66710421e769
> > > 100644
> > > --- a/include/linux/maple_tree.h
> > > +++ b/include/linux/maple_tree.h
> > > @@ -443,6 +443,7 @@ struct ma_state {
> > > unsigned long min; /* The minimum index of this node
> > > - implied pivot min */
> > > unsigned long max; /* The maximum index of this node
> > > - implied pivot max */
> > > struct slab_sheaf *sheaf; /* Allocated nodes for this
> > > operation */
> > > + struct maple_node *alloc; /* allocated nodes */
> > > unsigned long node_request;
> > > enum maple_status status; /* The status of the state
> > > (active, start, none, etc) */
> > > unsigned char depth; /* depth of tree descent during
> > > write */
> > > @@ -491,8 +492,9 @@ struct ma_wr_state {
> > > .status = ma_start, \
> > > .min = 0, \
> > > .max = ULONG_MAX, \
> > > - .node_request= 0, \
> > > .sheaf = NULL, \
> > > + .alloc = NULL, \
> > > + .node_request= 0, \
> > > .mas_flags = 0, \
> > > .store_type = wr_invalid, \
> > > }
> > > diff --git a/lib/maple_tree.c b/lib/maple_tree.c
> > > index
> > > 3c3c14a76d98ded3b619c178d64099b464a2ca23..9aa782b1497f224e7366ebbd65f997523ee0c8ab
> > > 100644
> > > --- a/lib/maple_tree.c
> > > +++ b/lib/maple_tree.c
> > > @@ -1101,16 +1101,23 @@ static int mas_ascend(struct ma_state *mas)
> > > *
> > > * Return: A pointer to a maple node.
> > > */
> > > -static inline struct maple_node *mas_pop_node(struct ma_state *mas)
> > > +static __always_inline struct maple_node *mas_pop_node(struct ma_state
> > > *mas)
> > > {
> > > struct maple_node *ret;
> > >
> > > + if (mas->alloc) {
> > > + ret = mas->alloc;
> > > + mas->alloc = NULL;
> > > + goto out;
> > > + }
> > > +
> > > if (WARN_ON_ONCE(!mas->sheaf))
> > > return NULL;
> > >
> > > ret = kmem_cache_alloc_from_sheaf(maple_node_cache, GFP_NOWAIT,
> > > mas->sheaf);
> > > - memset(ret, 0, sizeof(*ret));
> > >
> > > +out:
> > > + memset(ret, 0, sizeof(*ret));
> > > return ret;
> > > }
> > >
> > > @@ -1121,9 +1128,34 @@ static inline struct maple_node
> > > *mas_pop_node(struct ma_state *mas)
> > > */
> > > static inline void mas_alloc_nodes(struct ma_state *mas, gfp_t gfp)
> > > {
> > > - if (unlikely(mas->sheaf)) {
> > > - unsigned long refill = mas->node_request;
> > > + if (!mas->node_request)
> > > + return;
> > > +
> > > + if (mas->node_request == 1) {
> > > + if (mas->sheaf)
> > > + goto use_sheaf;
> > > +
> > > + if (mas->alloc)
> > > + return;
> > >
> > > + mas->alloc = mt_alloc_one(gfp);
> > > + if (!mas->alloc)
> > > + goto error;
> > > +
> > > + mas->node_request = 0;
> > > + return;
> > > + }
> > > +
> > > +use_sheaf:
> > > + if (unlikely(mas->alloc)) {
> >
> > When would this condition happen?
>
>
> This would be the case if we have one node allocated and requested more
> than one node. That is, a chained request for nodes that ends up having
> the alloc set and requesting a sheaf.
Ah, ok. So this is also a recovery case when we thought we need only
one node and then the situation changed and we need more than one?
>
> > Do we really need to free mas->alloc
> > here or it can be reused for the next 1-node allocation?
>
> Most calls end in mas_destroy() so that won't happen today.
>
> We could reduce the number of allocations requested to the sheaf and let
> the code find the mas->alloc first and use that.
>
> But remember, we are getting into this situation where code did a
> mas_preallocate() then figured they needed to do something else (error
> recovery, or changed the vma flags and now it can merge..) and will now
> need additional nodes. So this is a rare case, so I figured just free
> it was the safest thing.
Ok, got it. Both situations would be part of the unusual recovery
case. Makes sense then. Thanks!
>
>
> > > + mt_free_one(mas->alloc);
> > > + mas->alloc = NULL;
> > > + }
> > > +
> > > + if (mas->sheaf) {
> > > + unsigned long refill;
> > > +
> > > + refill = mas->node_request;
> > > if(kmem_cache_sheaf_size(mas->sheaf) >= refill) {
> > > mas->node_request = 0;
> > > return;
> > > @@ -5386,8 +5418,11 @@ void mas_destroy(struct ma_state *mas)
> > > mas->node_request = 0;
> > > if (mas->sheaf)
> > > mt_return_sheaf(mas->sheaf);
> > > -
> > > mas->sheaf = NULL;
> > > +
> > > + if (mas->alloc)
> > > + mt_free_one(mas->alloc);
> > > + mas->alloc = NULL;
> > > }
> > > EXPORT_SYMBOL_GPL(mas_destroy);
> > >
> > > @@ -6074,7 +6109,7 @@ bool mas_nomem(struct ma_state *mas, gfp_t gfp)
> > > mas_alloc_nodes(mas, gfp);
> > > }
> > >
> > > - if (!mas->sheaf)
> > > + if (!mas->sheaf && !mas->alloc)
> > > return false;
> > >
> > > mas->status = ma_start;
> > >
> > > --
> > > 2.50.1
> > >
