rbtree: undo augmented damage -v2
in [Kernel]
|
Prev: sched: adjust when cpu_active and cpuset configurations are updated during cpu on/offlining
Next: [PATCH][1/1] fs: wrong type for 'magic' argument in 'simple_fill_super()', fs/libfs.c
From: Venkatesh Pallipadi on 1 Jun 2010 13:10 On Sat, May 29, 2010 at 6:31 AM, Peter Zijlstra <peterz(a)infradead.org> wrote: > A better changelog, and I guess we can keep the explanation of > the augmented rb-tree since it doesn't actually mentions the > interface, merely the idea of adding a heap to the binary tree. > > The patch simply reverts the rbtree.c hunk, and keeps the non-conforming > style it has.. > > --- > Subject: rbtree: Undo augmented damage > From: Peter Zijlstra <a.p.zijlstra(a)chello.nl> > Date: Sat May 29 15:14:02 CEST 2010 > > Reimplement augmented RB-trees without sprinkling extra branches > all over the RB-tree code (which lives in the scheduler hot path). > > This approach is 'borrowed' from Fabio's BFQ implementation and > relies on traversing the rebalance path after the RB-tree-op to > correct the heap property for insertion/removal and make up for > the damage done by the tree rotations. > > For insertion the rebalance path is trivially that from the new > node upwards to the root, for removal it is that from the deepest > node in the path from the to be removed node that will still > be around after the removal. Yes. I like avoiding the sprinkled if's. But, I don't think rb_augment_erase_begin() and rb_augment_insert() is covering all the callbacks needed to maintain the augmented tree correctly. More comments below. > > Cc: Fabio Checconi <fabio(a)gandalf.sssup.it> > Cc: suresh.b.siddha(a)intel.com > Cc: venki(a)google.com > Signed-off-by: Peter Zijlstra <a.p.zijlstra(a)chello.nl> > --- > �arch/x86/mm/pat_rbtree.c | � �9 ++-- > �include/linux/rbtree.h � | � 13 ++++-- > �lib/rbtree.c � � � � � � | � 97 +++++++++++++++++++++++++---------------------- > �3 files changed, 68 insertions(+), 51 deletions(-) > > Index: linux-2.6/arch/x86/mm/pat_rbtree.c > =================================================================== > --- linux-2.6.orig/arch/x86/mm/pat_rbtree.c > +++ linux-2.6/arch/x86/mm/pat_rbtree.c > @@ -34,8 +34,7 @@ > �* memtype_lock protects the rbtree. > �*/ > > -static void memtype_rb_augment_cb(struct rb_node *node); > -static struct rb_root memtype_rbroot = RB_AUGMENT_ROOT(&memtype_rb_augment_cb); > +static struct rb_root memtype_rbroot = RB_ROOT; > > �static int is_node_overlap(struct memtype *node, u64 start, u64 end) > �{ > @@ -190,7 +189,7 @@ failure: > � � � �return -EBUSY; > �} > > -static void memtype_rb_augment_cb(struct rb_node *node) > +static void memtype_rb_augment_cb(struct rb_node *node, void *data) > �{ > � � � �if (node) > � � � � � � � �update_path_max_end(node); There is a optimization in memtype_rb_augment_cb, where in it does not do the fixup all the way to the root, when a nodes max_end does not change. That has to be removed for this change to work. > @@ -213,6 +212,7 @@ static void memtype_rb_insert(struct rb_ > > � � � �rb_link_node(&newdata->rb, parent, node); > � � � �rb_insert_color(&newdata->rb, root); > + � � � rb_augment_insert(&newdata->rb, memtype_rb_augment_cb, NULL); > �} > <snip> > @@ -324,6 +284,55 @@ void rb_erase(struct rb_node *node, stru > �EXPORT_SYMBOL(rb_erase); > > �/* > + * after inserting @node into the tree, update the tree to account for > + * both the new entry and any damage done by rebalance > + */ > +void rb_augment_insert(struct rb_node *node, rb_augment_f func, void *data) > +{ > + � � � if (node->rb_left) > + � � � � � � � node = node->rb_left; > + � � � else if (node->rb_right) > + � � � � � � � node = node->rb_right; > + > + � � � func(node, data); > +} I am not seeing how this can cover all the callbacks we will need to maintain the augmented tree property. May be I am missing something. For example, during insertion there can be a rotate at the grand parent. So, (bad ascii art alert) G / \ P U / N P / \ N G \ U Where N is the new node. Here G has new children and has to recalculate the nodes augmented data. I don't see that happening with rb_augment_insert() > + > +/* > + * before removing the node, find the deepest node on the rebalance path > + * that will still be there after @node gets removed > + */ > +struct rb_node *rb_augment_erase_begin(struct rb_node *node) > +{ > + � � � struct rb_node *deepest; > + > + � � � if (!node->rb_right && !node->rb_left) > + � � � � � � � deepest = rb_parent(node); > + � � � else if (!node->rb_right) > + � � � � � � � deepest = node->rb_left; > + � � � else if (!node->rb_left) > + � � � � � � � deepest = node->rb_right; > + � � � else { > + � � � � � � � deepest = rb_next(node); > + � � � � � � � if (deepest->rb_right) > + � � � � � � � � � � � deepest = deepest->rb_right; > + � � � � � � � else if (rb_parent(deepest) != node) > + � � � � � � � � � � � deepest = rb_parent(deepest); > + � � � } > + > + � � � return deepest; > +} > + I have similar concern with rb_augment_erase_begin() returning the deepest node. There is a case in __rb_erase_color where we do a rotate_left of 'other' (sibling) node. That changes the children of some nodes that are not in deepest to root path. No? Thanks, Venki -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 1 Jun 2010 13:20 On Tue, 2010-06-01 at 10:00 -0700, Venkatesh Pallipadi wrote: > I am not seeing how this can cover all the callbacks we will need to > maintain the augmented tree property. May be I am missing something. Yes, indeed, how about the below delta, which my eevdf code did do but I overlooked on the conversion to the PAT code. That removes the break as you said, but also adds code to update the child nodes when walking up the path. So in your rotation case: G P / \ --> / \ P U N G / <-- \ N U Say we take the right rotation, then the traversal up from N to P will find that since N was the left child of P and it has a right child (G) it will also update G. Signed-off-by: Peter Zijlstra <a.p.zijlstra(a)chello.nl> --- arch/x86/mm/pat_rbtree.c | 23 ++++++++++++----------- 1 files changed, 12 insertions(+), 11 deletions(-) diff --git a/arch/x86/mm/pat_rbtree.c b/arch/x86/mm/pat_rbtree.c index f537087..ca19aae 100644 --- a/arch/x86/mm/pat_rbtree.c +++ b/arch/x86/mm/pat_rbtree.c @@ -81,20 +81,21 @@ static void update_node_max_end(struct rb_node *node) /* Update 'subtree_max_end' for a node and all its ancestors */ static void update_path_max_end(struct rb_node *node) { - u64 old_max_end, new_max_end; + struct rb_node *parent; - while (node) { - struct memtype *data = container_of(node, struct memtype, rb); - - old_max_end = data->subtree_max_end; - update_node_max_end(node); - new_max_end = data->subtree_max_end; +up: + update_node_max_end(node); + parent = rb_parent(node); + if (!parent) + return; - if (new_max_end == old_max_end) - break; + if (node == parent->rb_left && parent->rb_right) + update_node_max_end(parent->rb_right); + else if (parent->rb_left) + update_node_max_end(parent->rb_left); - node = rb_parent(node); - } + node = parent; + goto up; } /* Find the first (lowest start addr) overlapping range from rb tree */ -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 1 Jun 2010 13:40 On Tue, 2010-06-01 at 19:19 +0200, Peter Zijlstra wrote: > On Tue, 2010-06-01 at 10:00 -0700, Venkatesh Pallipadi wrote: > > I am not seeing how this can cover all the callbacks we will need to > > maintain the augmented tree property. May be I am missing something. > > Yes, indeed, how about the below delta, which my eevdf code did do but I > overlooked on the conversion to the PAT code. > > That removes the break as you said, but also adds code to update the > child nodes when walking up the path. > > So in your rotation case: > > G P > / \ --> / \ > P U N G > / <-- \ > N U > > Say we take the right rotation, then the traversal up from N to P will > find that since N was the left child of P and it has a right child (G) > it will also update G. Hmm, that looks like it could well be folded in to the generic code.. let me spin a new patch. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Venkatesh Pallipadi on 1 Jun 2010 13:40 On Tue, Jun 1, 2010 at 10:19 AM, Peter Zijlstra <peterz(a)infradead.org> wrote: > On Tue, 2010-06-01 at 10:00 -0700, Venkatesh Pallipadi wrote: >> I am not seeing how this can cover all the callbacks we will need to >> maintain the augmented tree property. May be I am missing something. > > Yes, indeed, how about the below delta, which my eevdf code did do but I > overlooked on the conversion to the PAT code. > > That removes the break as you said, but also adds code to update the > child nodes when walking up the path. > > So in your rotation case: > > � �G � � � � � � P > � / \ � �--> � �/ \ > �P � U � � � � N � G > �/ � � � �<-- � � � �\ > N � � � � � � � � � � U > > Say we take the right rotation, then the traversal up from N to P will > find that since N was the left child of P and it has a right child (G) > it will also update G. Yes. This will cover all the cases on insert. But on erase, there is still a case where a rotate of sibling node is done during the re-coloration process. There we have a child change on sibling's child. I am not able to think of any easy way to handle that case. Thanks, Venki > > Signed-off-by: Peter Zijlstra <a.p.zijlstra(a)chello.nl> > --- > �arch/x86/mm/pat_rbtree.c | � 23 ++++++++++++----------- > �1 files changed, 12 insertions(+), 11 deletions(-) > > diff --git a/arch/x86/mm/pat_rbtree.c b/arch/x86/mm/pat_rbtree.c > index f537087..ca19aae 100644 > --- a/arch/x86/mm/pat_rbtree.c > +++ b/arch/x86/mm/pat_rbtree.c > @@ -81,20 +81,21 @@ static void update_node_max_end(struct rb_node *node) > �/* Update 'subtree_max_end' for a node and all its ancestors */ > �static void update_path_max_end(struct rb_node *node) > �{ > - � � � u64 old_max_end, new_max_end; > + � � � struct rb_node *parent; > > - � � � while (node) { > - � � � � � � � struct memtype *data = container_of(node, struct memtype, rb); > - > - � � � � � � � old_max_end = data->subtree_max_end; > - � � � � � � � update_node_max_end(node); > - � � � � � � � new_max_end = data->subtree_max_end; > +up: > + � � � update_node_max_end(node); > + � � � parent = rb_parent(node); > + � � � if (!parent) > + � � � � � � � return; > > - � � � � � � � if (new_max_end == old_max_end) > - � � � � � � � � � � � break; > + � � � if (node == parent->rb_left && parent->rb_right) > + � � � � � � � update_node_max_end(parent->rb_right); > + � � � else if (parent->rb_left) > + � � � � � � � update_node_max_end(parent->rb_left); > > - � � � � � � � node = rb_parent(node); > - � � � } > + � � � node = parent; > + � � � goto up; > �} > > �/* Find the first (lowest start addr) overlapping range from rb tree */ > > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 1 Jun 2010 13:50
On Tue, 2010-06-01 at 10:34 -0700, Venkatesh Pallipadi wrote: > > Yes. This will cover all the cases on insert. But on erase, there is > still a case where a rotate of sibling node is done during the > re-coloration process. There we have a child change on sibling's > child. I am not able to think of any easy way to handle that case. Let me go draw some figures with pen and paper to match up the erase path with the rb_augment_erase_begin() code, because I can't quite spot the case we're missing. If you have it handy, ascii art might help.. Below a new patch with the missing bit folded into the generic code. --- Subject: rbtree: Undo augmented damage From: Peter Zijlstra <peterz(a)infradead.org> Date: Sat, 29 May 2010 15:31:43 +0200 Reimplement augmented RB-trees without sprinkling extra branches all over the RB-tree code (which lives in the scheduler hot path). This approach is 'borrowed' from Fabio's BFQ implementation and relies on traversing the rebalance path after the RB-tree-op to correct the heap property for insertion/removal and make up for the damage done by the tree rotations. For insertion the rebalance path is trivially that from the new node upwards to the root, for removal it is that from the deepest node in the path from the to be removed node that will still be around after the removal. Cc: Fabio Checconi <fabio(a)gandalf.sssup.it> Cc: suresh.b.siddha(a)intel.com Cc: venki(a)google.com Signed-off-by: Peter Zijlstra <a.p.zijlstra(a)chello.nl> --- arch/x86/mm/pat_rbtree.c | 34 ++----------- include/linux/rbtree.h | 13 +++-- lib/rbtree.c | 116 +++++++++++++++++++++++++++++------------------ 3 files changed, 87 insertions(+), 76 deletions(-) Index: linux-2.6/arch/x86/mm/pat_rbtree.c =================================================================== --- linux-2.6.orig/arch/x86/mm/pat_rbtree.c +++ linux-2.6/arch/x86/mm/pat_rbtree.c @@ -34,8 +34,7 @@ * memtype_lock protects the rbtree. */ -static void memtype_rb_augment_cb(struct rb_node *node); -static struct rb_root memtype_rbroot = RB_AUGMENT_ROOT(&memtype_rb_augment_cb); +static struct rb_root memtype_rbroot = RB_ROOT; static int is_node_overlap(struct memtype *node, u64 start, u64 end) { @@ -56,7 +55,7 @@ static u64 get_subtree_max_end(struct rb } /* Update 'subtree_max_end' for a node, based on node and its children */ -static void update_node_max_end(struct rb_node *node) +static void memtype_rb_augment_cb(struct rb_node *node, void *__unused) { struct memtype *data; u64 max_end, child_max_end; @@ -78,25 +77,6 @@ static void update_node_max_end(struct r data->subtree_max_end = max_end; } -/* Update 'subtree_max_end' for a node and all its ancestors */ -static void update_path_max_end(struct rb_node *node) -{ - u64 old_max_end, new_max_end; - - while (node) { - struct memtype *data = container_of(node, struct memtype, rb); - - old_max_end = data->subtree_max_end; - update_node_max_end(node); - new_max_end = data->subtree_max_end; - - if (new_max_end == old_max_end) - break; - - node = rb_parent(node); - } -} - /* Find the first (lowest start addr) overlapping range from rb tree */ static struct memtype *memtype_rb_lowest_match(struct rb_root *root, u64 start, u64 end) @@ -190,12 +170,6 @@ failure: return -EBUSY; } -static void memtype_rb_augment_cb(struct rb_node *node) -{ - if (node) - update_path_max_end(node); -} - static void memtype_rb_insert(struct rb_root *root, struct memtype *newdata) { struct rb_node **node = &(root->rb_node); @@ -213,6 +187,7 @@ static void memtype_rb_insert(struct rb_ rb_link_node(&newdata->rb, parent, node); rb_insert_color(&newdata->rb, root); + rb_augment_insert(&newdata->rb, memtype_rb_augment_cb, NULL); } int rbt_memtype_check_insert(struct memtype *new, unsigned long *ret_type) @@ -233,13 +208,16 @@ int rbt_memtype_check_insert(struct memt struct memtype *rbt_memtype_erase(u64 start, u64 end) { + struct rb_node *deepest; struct memtype *data; data = memtype_rb_exact_match(&memtype_rbroot, start, end); if (!data) goto out; + deepest = rb_augment_erase_begin(&data->rb); rb_erase(&data->rb, &memtype_rbroot); + rb_augment_erase_end(deepest, memtype_rb_augment_cb, NULL); out: return data; } Index: linux-2.6/include/linux/rbtree.h =================================================================== --- linux-2.6.orig/include/linux/rbtree.h +++ linux-2.6/include/linux/rbtree.h @@ -110,7 +110,6 @@ struct rb_node struct rb_root { struct rb_node *rb_node; - void (*augment_cb)(struct rb_node *node); }; @@ -130,9 +129,7 @@ static inline void rb_set_color(struct r rb->rb_parent_color = (rb->rb_parent_color & ~1) | color; } -#define RB_ROOT (struct rb_root) { NULL, NULL, } -#define RB_AUGMENT_ROOT(x) (struct rb_root) { NULL, x} - +#define RB_ROOT (struct rb_root) { NULL, } #define rb_entry(ptr, type, member) container_of(ptr, type, member) #define RB_EMPTY_ROOT(root) ((root)->rb_node == NULL) @@ -142,6 +139,14 @@ static inline void rb_set_color(struct r extern void rb_insert_color(struct rb_node *, struct rb_root *); extern void rb_erase(struct rb_node *, struct rb_root *); +typedef void (*rb_augment_f)(struct rb_node *node, void *data); + +extern void rb_augment_insert(struct rb_node *node, + rb_augment_f func, void *data); +extern struct rb_node *rb_augment_erase_begin(struct rb_node *node); +extern void rb_augment_erase_end(struct rb_node *node, + rb_augment_f func, void *data); + /* Find logical next and previous nodes in a tree */ extern struct rb_node *rb_next(const struct rb_node *); extern struct rb_node *rb_prev(const struct rb_node *); Index: linux-2.6/lib/rbtree.c =================================================================== --- linux-2.6.orig/lib/rbtree.c +++ linux-2.6/lib/rbtree.c @@ -44,11 +44,6 @@ static void __rb_rotate_left(struct rb_n else root->rb_node = right; rb_set_parent(node, right); - - if (root->augment_cb) { - root->augment_cb(node); - root->augment_cb(right); - } } static void __rb_rotate_right(struct rb_node *node, struct rb_root *root) @@ -72,20 +67,12 @@ static void __rb_rotate_right(struct rb_ else root->rb_node = left; rb_set_parent(node, left); - - if (root->augment_cb) { - root->augment_cb(node); - root->augment_cb(left); - } } void rb_insert_color(struct rb_node *node, struct rb_root *root) { struct rb_node *parent, *gparent; - if (root->augment_cb) - root->augment_cb(node); - while ((parent = rb_parent(node)) && rb_is_red(parent)) { gparent = rb_parent(parent); @@ -240,15 +227,12 @@ void rb_erase(struct rb_node *node, stru else { struct rb_node *old = node, *left; - int old_parent_cb = 0; - int successor_parent_cb = 0; node = node->rb_right; while ((left = node->rb_left) != NULL) node = left; if (rb_parent(old)) { - old_parent_cb = 1; if (rb_parent(old)->rb_left == old) rb_parent(old)->rb_left = node; else @@ -263,10 +247,8 @@ void rb_erase(struct rb_node *node, stru if (parent == old) { parent = node; } else { - successor_parent_cb = 1; if (child) rb_set_parent(child, parent); - parent->rb_left = child; node->rb_right = old->rb_right; @@ -277,24 +259,6 @@ void rb_erase(struct rb_node *node, stru node->rb_left = old->rb_left; rb_set_parent(old->rb_left, node); - if (root->augment_cb) { - /* - * Here, three different nodes can have new children. - * The parent of the successor node that was selected - * to replace the node to be erased. - * The node that is getting erased and is now replaced - * by its successor. - * The parent of the node getting erased-replaced. - */ - if (successor_parent_cb) - root->augment_cb(parent); - - root->augment_cb(node); - - if (old_parent_cb) - root->augment_cb(rb_parent(old)); - } - goto color; } @@ -303,19 +267,15 @@ void rb_erase(struct rb_node *node, stru if (child) rb_set_parent(child, parent); - - if (parent) { + if (parent) + { if (parent->rb_left == node) parent->rb_left = child; else parent->rb_right = child; - - if (root->augment_cb) - root->augment_cb(parent); - - } else { - root->rb_node = child; } + else + root->rb_node = child; color: if (color == RB_BLACK) @@ -323,6 +283,74 @@ void rb_erase(struct rb_node *node, stru } EXPORT_SYMBOL(rb_erase); +static void rb_augment_path(struct rb_node *node, rb_augment_f func, void *data) +{ + struct rb_node *parent; + +up: + func(node, data); + parent = rb_parent(node); + if (!parent) + return; + + if (node == parent->rb_left && parent->rb_right) + func(parent->rb_right, data); + else if (parent->rb_left) + func(parent->rb_left, data); + + node = parent; + goto up; +} + +/* + * after inserting @node into the tree, update the tree to account for + * both the new entry and any damage done by rebalance + */ +void rb_augment_insert(struct rb_node *node, rb_augment_f func, void *data) +{ + if (node->rb_left) + node = node->rb_left; + else if (node->rb_right) + node = node->rb_right; + + rb_augment_path(node, func, data); +} + +/* + * before removing the node, find the deepest node on the rebalance path + * that will still be there after @node gets removed + */ +struct rb_node *rb_augment_erase_begin(struct rb_node *node) +{ + struct rb_node *deepest; + + if (!node->rb_right && !node->rb_left) + deepest = rb_parent(node); + else if (!node->rb_right) + deepest = node->rb_left; + else if (!node->rb_left) + deepest = node->rb_right; + else { + deepest = rb_next(node); + if (deepest->rb_right) + deepest = deepest->rb_right; + else if (rb_parent(deepest) != node) + deepest = rb_parent(deepest); + } + + return deepest; +} + +/* + * after removal, update the tree to account for the removed entry + * and any rebalance damage. + */ +void rb_augment_erase_end(struct rb_node *node, rb_augment_f func, void *data) +{ + if (node) + rb_augment_path(node, func, data); +} + /* * This function returns the first node (in sort order) of the tree. */ -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |