Skip to contentSkip to conversation

Hashtable implementation breaks on flow statements

DISCUSSION Schrober 2 participants 2 messages Jan 14, 2013

lore
Share & tools
lore

Schrober

author
#0 lore raw

Hi,

I think the new hash implementation in include/linux/hashtable.h is slightly unintuitive when using things like hash_for_each, hash_for_each_rcu, hash_for_each_safe and hash_for_each_possible.

Users may want to get out of the loop through the hashtable iteration when some condition is met. This works well on hlist, list and other "abstract" data types. But the hash table would just jump to the next bucket and continue.

Signature · 1 line
Franz Schrober
#1 lore raw

Hi Franz,

The hashtable should behave correctly (that is, not continue iteration from the next
bucket) with regards to a 'break' inside the loop, this issue was raised previously
during one of the RFCs and has been dealt with before the merge of the hashtable into the kernel.

Do you see a case where it behaves incorrectly? can you share your code which does that?

Thanks,
Sasha

Quoted 12 lines · Schrober

Hi,

I think the new hash implementation in include/linux/hashtable.h is slightly unintuitive when using things like hash_for_each, hash_for_each_rcu, hash_for_each_safe and hash_for_each_possible.

Users may want to get out of the loop through the hashtable iteration when some condition is met. This works well on hlist, list and other "abstract" data types. But the hash table would just jump to the next bucket and continue.
--
Franz Schrober