2022-11-13 01:25:25

by Wei Yang

Subject: [PATCH] maple_tree: not necessary to filter MAPLE_PARENT_ROOT since it is not a root

Root node is return at the beginning, so we are sure bit 0 is not set.

Signed-off-by: Wei Yang <[email protected]>
CC: Liam R. Howlett <[email protected]>
---
lib/maple_tree.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib/maple_tree.c b/lib/maple_tree.c
index 9aad98c24f3e..f8c4755e7c75 100644
--- a/lib/maple_tree.c
+++ b/lib/maple_tree.c
@@ -436,7 +436,7 @@ enum maple_type mte_parent_enum(struct maple_enode *p_enode,
return 0; /* Validated in the caller. */

p_type &= MAPLE_NODE_MASK;
- p_type = p_type & ~(MAPLE_PARENT_ROOT | mte_parent_slot_mask(p_type));
+ p_type = p_type & ~mte_parent_slot_mask(p_type);

switch (p_type) {
case MAPLE_PARENT_RANGE64: /* or MAPLE_PARENT_ARANGE64 */
--
2.33.1

2022-11-15 14:41:51

by Liam R. Howlett

Subject: Re: [PATCH] maple_tree: not necessary to filter MAPLE_PARENT_ROOT since it is not a root

* Wei Yang <[email protected]> [221112 19:56]:
> Root node is return at the beginning, so we are sure bit 0 is not set.
>
> Signed-off-by: Wei Yang <[email protected]>
> CC: Liam R. Howlett <[email protected]>
> ---
> lib/maple_tree.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/lib/maple_tree.c b/lib/maple_tree.c
> index 9aad98c24f3e..f8c4755e7c75 100644
> --- a/lib/maple_tree.c
> +++ b/lib/maple_tree.c
> @@ -436,7 +436,7 @@ enum maple_type mte_parent_enum(struct maple_enode *p_enode,
> return 0; /* Validated in the caller. */
>
> p_type &= MAPLE_NODE_MASK;
> - p_type = p_type & ~(MAPLE_PARENT_ROOT | mte_parent_slot_mask(p_type));
> + p_type = p_type & ~mte_parent_slot_mask(p_type);

I think there is a larger cleanup that can be done here. It looks like
mte_parent_enum() is called from one location and that location is a
wrapper.

The check for the root bit should also probably trigger a WARN_ON() and
still return 0. I don't think the callers are doing enough to validate
it - although they should never reach this function with a root node.
And, in fact, I am not doing enough in the test code since I didn't
guard this correctly in the verification of the parent slot before
calling this function.

Thanks for pointing this out. I will send out a patch to clean this up
shortly.

>
> switch (p_type) {
> case MAPLE_PARENT_RANGE64: /* or MAPLE_PARENT_ARANGE64 */
> --
> 2.33.1
>

2022-11-15 15:27:18

by Liam R. Howlett

Subject: Re: [PATCH] maple_tree: not necessary to filter MAPLE_PARENT_ROOT since it is not a root

* Liam R. Howlett <[email protected]> [221115 09:29]:
> * Wei Yang <[email protected]> [221112 19:56]:
> > Root node is return at the beginning, so we are sure bit 0 is not set.
> >
> > Signed-off-by: Wei Yang <[email protected]>
> > CC: Liam R. Howlett <[email protected]>
> > ---
> > lib/maple_tree.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/lib/maple_tree.c b/lib/maple_tree.c
> > index 9aad98c24f3e..f8c4755e7c75 100644
> > --- a/lib/maple_tree.c
> > +++ b/lib/maple_tree.c
> > @@ -436,7 +436,7 @@ enum maple_type mte_parent_enum(struct maple_enode *p_enode,
> > return 0; /* Validated in the caller. */
> >
> > p_type &= MAPLE_NODE_MASK;
> > - p_type = p_type & ~(MAPLE_PARENT_ROOT | mte_parent_slot_mask(p_type));
> > + p_type = p_type & ~mte_parent_slot_mask(p_type);
>
> I think there is a larger cleanup that can be done here. It looks like
> mte_parent_enum() is called from one location and that location is a
> wrapper.
>
> The check for the root bit should also probably trigger a WARN_ON() and
> still return 0. I don't think the callers are doing enough to validate
> it - although they should never reach this function with a root node.
> And, in fact, I am not doing enough in the test code since I didn't
> guard this correctly in the verification of the parent slot before
> calling this function.
>
> Thanks for pointing this out. I will send out a patch to clean this up
> shortly.

On second thought, I will hold off for the 6.2 merge window for this to
go upstream.

Thanks,
Liam

2022-11-17 04:44:38

by Wei Yang

Subject: Re: [PATCH] maple_tree: not necessary to filter MAPLE_PARENT_ROOT since it is not a root

On Tue, Nov 15, 2022 at 02:31:15PM +0000, Liam Howlett wrote:
>* Wei Yang <[email protected]> [221112 19:56]:
>> Root node is return at the beginning, so we are sure bit 0 is not set.
>>
>> Signed-off-by: Wei Yang <[email protected]>
>> CC: Liam R. Howlett <[email protected]>
>> ---
>> lib/maple_tree.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/lib/maple_tree.c b/lib/maple_tree.c
>> index 9aad98c24f3e..f8c4755e7c75 100644
>> --- a/lib/maple_tree.c
>> +++ b/lib/maple_tree.c
>> @@ -436,7 +436,7 @@ enum maple_type mte_parent_enum(struct maple_enode *p_enode,
>> return 0; /* Validated in the caller. */
>>
>> p_type &= MAPLE_NODE_MASK;
>> - p_type = p_type & ~(MAPLE_PARENT_ROOT | mte_parent_slot_mask(p_type));
>> + p_type = p_type & ~mte_parent_slot_mask(p_type);
>
>I think there is a larger cleanup that can be done here. It looks like
>mte_parent_enum() is called from one location and that location is a
>wrapper.
>
>The check for the root bit should also probably trigger a WARN_ON() and
>still return 0. I don't think the callers are doing enough to validate
>it - although they should never reach this function with a root node.
>And, in fact, I am not doing enough in the test code since I didn't
>guard this correctly in the verification of the parent slot before
>calling this function.
>
>Thanks for pointing this out. I will send out a patch to clean this up
>shortly.
>

Yep, look forward your cleanup.

While I have a question here. We get 4 types in maple_type, here we just
return two of them. This means the other two is not possible to be parent
node, right?

>>
>> switch (p_type) {
>> case MAPLE_PARENT_RANGE64: /* or MAPLE_PARENT_ARANGE64 */
>> --
>> 2.33.1
>>

--
Wei Yang
Help you, Help me