Re: [PATCH 1/2] perf tools: Use postorder rbtree iteration when removing symbols

[Ingo Molnar <mingo@kernel.org>]

* Alex Snast <asnast@gmail.com> wrote:

> Avoid using rb_erase when removing symbols as it requires rbtree
> rebalancing, instead preform a post order iteration when deleting tree
> symbols.
> 
> Signed-off-by: Alex Snast <asnast@gmail.com>
> ---
>  tools/include/linux/rbtree.h | 14 ++++++++++++++
>  tools/perf/util/symbol.c     | 11 ++++-------
>  2 files changed, 18 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/include/linux/rbtree.h b/tools/include/linux/rbtree.h
> index 1125822..14c646d 100644
> --- a/tools/include/linux/rbtree.h
> +++ b/tools/include/linux/rbtree.h
> @@ -90,6 +90,20 @@ static inline void rb_link_node(struct rb_node *node, struct rb_node *parent,
>  	   ____ptr ? rb_entry(____ptr, type, member) : NULL; \
>  	})
>  
> +/**
> + * rbtree_postorder_for_each_entry_safe - iterate over rb_root in post order of
> + * given type safe against removal of rb_node entry
> + *
> + * @pos:	the 'type *' to use as a loop cursor.
> + * @n:		another 'type *' to use as temporary storage
> + * @root:	'rb_root *' of the rbtree.
> + * @field:	the name of the rb_node field within 'type'.
> + */
> +#define rbtree_postorder_for_each_entry_safe(pos, n, root, field) \
> +	for (pos = rb_entry_safe(rb_first_postorder(root), typeof(*pos), field); \
> +	     pos && ({ n = rb_entry_safe(rb_next_postorder(&pos->field), \
> +			typeof(*pos), field); 1; }); \
> +	     pos = n)

So looks like this is something that include/linux/rbtree.h already has, right?

I think we should strive to match the two implementations and generate a build 
time warning (but not a build time failure) if the two diverge.

There were checking facilities added recently for another kernel source code file, 
to make sure tools/perf/util/intel-pt-decoder/insn.c matches arch/x86/lib/insn.c.

Btw., side note, regarding insn.c, the remaining delta between the two files:

-#include <asm/inat.h>
-#include <asm/insn.h>
+#include "inat.h"
+#include "insn.h"

should be eliminated too I think, to make it more obvious when the two versions 
match.

Thanks,

	Ingo