Re: [PATCH 0/1] Dead stores in maple-tree

[Liam Howlett <liam.howlett@oracle.com>]

* Dan Carpenter <dan.carpenter@oracle.com> [221027 03:43]:
> On Wed, Oct 26, 2022 at 02:23:19PM +0000, Liam Howlett wrote:
> > * Lukas Bulwahn <lukas.bulwahn@gmail.com> [221026 08:01]:
> > > Dear maple-tree authors, dear Liam, dear Matthew,
> > > 
> > > there are some Dead Stores that clang-analyzer reports:
> > > 
> > > lib/maple_tree.c:2906:2: warning: Value stored to 'last' is never read [clang-analyzer-deadcode.DeadStores]
> > > lib/maple_tree.c:2907:2: warning: Value stored to 'prev_min' is never read [clang-analyzer-deadcode.DeadStores]
> > > 
> > > I addressed these two cases, which were most obvious and clear to fix;
> > > see patch of this one-element series.
> > > 
> > > Further, clang-analyzer reports more, which I did not address:
> > > 
> > > lib/maple_tree.c:332:2: warning: Value stored to 'node' is never read [clang-analyzer-deadcode.DeadStores]
> > > lib/maple_tree.c:337:2: warning: Value stored to 'node' is never read [clang-analyzer-deadcode.DeadStores]
> > > 
> > > Unclear to me if the tool is wrong or right in its analysis here for the two functions above.
> > 
> > The tool is correct but these aren't going anywhere.  They are compiled
> > out and are needed for the future.
> > 
> 
> lib/maple_tree.c

~line 302:
/* Bit 1 indicates the root is a node */
#define MAPLE_ROOT_NODE                 0x02
/* maple_type stored bit 3-6 */
#define MAPLE_ENODE_TYPE_SHIFT          0x03
/* Bit 2 means a NULL somewhere below */
#define MAPLE_ENODE_NULL                0x04


>    330  static inline void mte_set_full(const struct maple_enode *node)
>    331  {
>    332          node = (void *)((unsigned long)node & ~MAPLE_ENODE_NULL);
>    333  }
>    334  
>    335  static inline void mte_clear_full(const struct maple_enode *node)
>    336  {
>    337          node = (void *)((unsigned long)node | MAPLE_ENODE_NULL);
>    338  }

Looking at the code.... the analysis is correct and these need to be
fixed.  Thanks Dan & Lukas.

> 
> That code is really puzzling...  How far into the future before it starts
> making sense?

If you want to know details like this, you can look at the comments in
the header and c file - that's where the development information
resides.  Information about a node is encoded in the last bits of that
nodes pointer - since they are aligned we can use a mask to restore the
pointer.  Internally I refer to nodes with encoded information as
maple_enodes.  This part is to do with finding out if there is a free
index within the range the node holds.  Think about searching for the
next available index for a unique identifier.

Thanks,
Liam