radixtree: regulate tag get return value
Andrew, this small patch makes the radixtree tester program from
http://www.zip.com.au/~akpm/linux/patches/stuff/rtth.tar.gz
run OK, with the latest radix tree code in linux-2.6.17-rc5-mm2.
It regulates the return value to 0/1 for functions radix_tree_tag_get() and radix_tree_tagged().
---
lib/radix-tree.c +2 -2
@@ -156,7 +156,7 @@ static inline void tag_clear(struct radistatic inline int tag_get (struct radix_tree_node *node,unsigned int tag,int offset) { -return test_bit (offset, node->tags[tag]); +return !!test_bit (offset, node->tags[tag]); }static inline void root_tag_set (struct radix_tree_root *root,unsigned int tag)
@@ -177,7 +177,7 @@ static inline void root_tag_clear_all(ststatic inline int root_tag_get (struct radix_tree_root *root,unsigned int tag) { -return root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT)); +return !! (root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT))); }/*
Quoted 7 lines · On Sun, 4 Jun 2006 16:45:48 +0800 Wu Fengguang <[email protected]> wrote:
Andrew, this small patch makes the radixtree tester program from
http://www.zip.com.au/~akpm/linux/patches/stuff/rtth.tar.gz
run OK, with the latest radix tree code in linux-2.6.17-rc5-mm2.It regulates the return value to 0/1 for functions
radix_tree_tag_get()andradix_tree_tagged().
Well yes. But it slows down the kernel. It would be better to fix rtth.
lib/radix-tree.c static inline void tag_clear(struct radi · line 160
Quoted 11 lines
---
lib/radix-tree.c +1
-1@@ -156,7 +156,7 @@ static inline void tag_clear(struct radistatic inline int tag_get (struct radix_tree_node *node,unsigned int tag,int offset) { -return test_bit (offset, node->tags[tag]); +return !!test_bit (offset, node->tags[tag]); }
test_bit() is (sadly) defined to return 0 or 1. Did this really make a difference?
Quoted 9 lines
lib/radix-tree.c +1
-1static inline void root_tag_set (struct radix_tree_root *root,unsigned int tag)@@ -177,7 +177,7 @@ static inline void root_tag_clear_all(ststatic inline int root_tag_get (struct radix_tree_root *root,unsigned int tag) { -return root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT)); +return !! (root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT))); }
Quoted 7 lines · Andrew Morton
Quoted 4 lines · On Sun, 4 Jun 2006 16:45:48 +0800 Wu Fengguang <[email protected]> wrote:
Andrew, this small patch makes the radixtree tester program from
http://www.zip.com.au/~akpm/linux/patches/stuff/rtth.tar.gz
run OK, with the latest radix tree code in linux-2.6.17-rc5-mm2.
Quoted 5 lines
Quoted 3 lines
It regulates the return value to 0/1 for functions
radix_tree_tag_get()andradix_tree_tagged().Well yes. But it slows down the kernel. It would be better to fix rtth.
OK. I'll send an updated patch to directly fix the for-rtth radix_tree_tag_get().
Quoted 13 lines
lib/radix-tree.c static inline void tag_clear(struct radi · line 160
Quoted 11 lines
---
lib/radix-tree.c +1
-1@@ -156,7 +156,7 @@ static inline void tag_clear(struct radistatic inline int tag_get (struct radix_tree_node *node,unsigned int tag,int offset) { -return test_bit (offset, node->tags[tag]); +return !!test_bit (offset, node->tags[tag]); }
test_bit()is (sadly) defined to return 0 or 1. Did this really make a difference?
Interesting. I got the following gdb outputs. Note that tag_get() returns -1 and root_tag_get() returns 1048576.
(gdb) n
399 while (height > 0) {
(gdb) n
402 offset = (index >> shift) & RADIX_TREE_MAP_MASK;
(gdb)
403 if (!tag_get(slot, tag, offset))
(gdb)
404 tag_set(slot, tag, offset);
(gdb) p tag_get(slot, tag, offset)
$14 = 0
(gdb) n
405 slot = slot->slots[offset];
(gdb) p tag_get(slot, tag, offset)
$15 = -1
(gdb) n
406 BUG_ON(slot == NULL);
(gdb) n
407 shift -= RADIX_TREE_MAP_SHIFT;
(gdb) n
408 height--;
(gdb) n
399 while (height > 0) {
(gdb) n
412 if (slot && !root_tag_get(root, tag))
(gdb) p root_tag_get(root, tag)
$16 = 0
(gdb) n
413 root_tag_set(root, tag);
(gdb) n
415 return slot;
(gdb) p root_tag_get(root, tag)
$17 = 1048576
Quoted 10 lines
Quoted 9 lines
lib/radix-tree.c +1
-1static inline void root_tag_set (struct radix_tree_root *root,unsigned int tag)@@ -177,7 +177,7 @@ static inline void root_tag_clear_all(ststatic inline int root_tag_get (struct radix_tree_root *root,unsigned int tag) { -return root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT)); +return !! (root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT))); }
Quoted 5 lines · On Sun, 4 Jun 2006 19:17:54 +0800 Wu Fengguang <[email protected]> wrote:
Quoted 1 line
test_bit()is (sadly) defined to return 0 or 1. Did this really make a difference?Interesting. I got the following gdb outputs. Note that
tag_get()returns -1 androot_tag_get()returns 1048576.
Quoted 14 lines
(gdb) n
399 while (height > 0) {
(gdb) n
402 offset = (index >> shift) &RADIX_TREE_MAP_MASK;
(gdb)
403 if (!tag_get(slot, tag, offset))
(gdb)
404tag_set(slot, tag, offset);
(gdb) ptag_get(slot, tag, offset)
$14 = 0
(gdb) n
405 slot = slot->slots[offset];
(gdb) ptag_get(slot, tag, offset)
$15 = -1
You trust gdb more than I do ;) It's doing a pretty tricky thing there.
test_bit() returns (1 & (expr)) - it _has_ to return 0 or 1.
PATCH 1/1 radixtree: normalize radix_tree_tag_get() return value
In reply to Andrew MortonIn radix_tree_tag_get(), return normalized value of 0/1, as indicated by its comment.
---
lib/radix-tree.c +1 -1
@@ -531,7 +531,7 @@ int radix_tree_tag_get(struct radix_treeint ret =tag_get (slot, tag, offset);BUG_ON (ret && saw_unset_tag); -return ret; +return !! ret; } slot = slot->slots[offset]; shift -=RADIX_TREE_MAP_SHIFT ;