On 08/02/2012 12:32 PM, Josh Triplett wrote:

On Thu, Aug 02, 2012 at 12:00:33PM +0200, Sasha Levin wrote:

quoted

On 08/02/2012 12:45 AM, Tejun Heo wrote:

quoted

On Thu, Aug 02, 2012 at 12:41:56AM +0200, Sasha Levin wrote:

quoted

How would your DEFINE_HASHTABLE look like if we got for the simple
'struct hash_table' approach?

I think defining a different enclosing anonymous struct which the
requested number of array entries and then aliasing the actual
hash_table to that symbol should work.  It's rather horrible and I'm
not sure it's worth the trouble.

I agree that this is probably not worth the trouble.

At the moment I see two alternatives:

1. Dynamically allocate the hash buckets.

2. Use the first bucket to store size. Something like the follows:

	#define HASH_TABLE(name, bits)	\
        	struct hlist_head name[1 << bits + 1];

	#define HASH_TABLE_INIT (bits) ({name[0].next = bits});

And then have hash_{add,get} just skip the first bucket.


While it's not a pretty hack, I don't see a nice way to avoid having to dynamically allocate buckets for all cases.

What about using a C99 flexible array member?  Kernel style prohibits
variable-length arrays, but I don't think the same rationale applies to
flexible array members.

struct hash_table {
    size_t count;
    struct hlist_head buckets[];
};

#define DEFINE_HASH_TABLE(name, length) struct hash_table name = { .count = length, .buckets = { [0 ... (length - 1)] = HLIST_HEAD_INIT } }

The limitation of this approach is that the struct hash_table variable must be 'static', which is a bit limiting - see for example the use of hashtable in 'struct user_namespace'.

- Josh Triplett

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>