On 9/17/21 10:01 AM, Alexei Starovoitov wrote:

On Mon, Sep 13, 2021 at 09:04:30PM -0700, Joanne Koong wrote:

quoted

+
+/* For bloom filter maps, the next 4 bits represent how many hashes to use.
+ * The maximum number of hash functions supported is 15. If this is not set,
+ * the default number of hash functions used will be 5.
+ */
+	BPF_F_BLOOM_FILTER_HASH_BIT_1 = (1U << 13),
+	BPF_F_BLOOM_FILTER_HASH_BIT_2 = (1U << 14),
+	BPF_F_BLOOM_FILTER_HASH_BIT_3 = (1U << 15),
+	BPF_F_BLOOM_FILTER_HASH_BIT_4 = (1U << 16),

The bit selection is unintuitive.
Since key_size has to be zero may be used that instead to indicate the number of hash
functions in the rare case when 5 is not good enough?
Or use inner_map_fd since there is no possibility of having an inner map in bloomfilter.
It could be a union:
     __u32   max_entries;    /* max number of entries in a map */
     __u32   map_flags;      /* BPF_MAP_CREATE related
                              * flags defined above.
                              */
     union {
        __u32  inner_map_fd;   /* fd pointing to the inner map */
        __u32  nr_hash_funcs;  /* or number of hash functions */
     };
     __u32   numa_node;      /* numa node */

I really like the idea of union-ing inner_map_fd with the number of hash 
functions (my worry with
using key_size is that it might be a confusing / non-intuitive API quirk 
for users), but I think this
would later require us to add some bloom filter specific APIs to libbpf 
(such as bpf_map__set_nr_hashes).

To make the bit selection more intuitive, Andrii suggested defining some 
helper like

BPF_F_BLOOM_NR_HASH_OFF = 13

where the user could then do something like

struct {
     __uint(type, BPF_MAP_TYPE_BLOOM_FILTER),
     ...
     __uint(map_flags, 5 << BPF_F_BLOOM_NR_HASH_OFF),
};

to set the number of hash functions.

Would this approach address your concerns about the unintuitiveness of 
the bit selection?

quoted

+struct bpf_bloom_filter {
+	struct bpf_map map;
+	u32 bit_array_mask;
+	u32 hash_seed;
+	/* If the size of the values in the bloom filter is u32 aligned,
+	 * then it is more performant to use jhash2 as the underlying hash
+	 * function, else we use jhash. This tracks the number of u32s
+	 * in an u32-aligned value size. If the value size is not u32 aligned,
+	 * this will be 0.
+	 */
+	u32 aligned_u32_count;

what is the performance difference?

Using results from the hashmap benchmark tests, using jhash2 instead of 
jhash for 4-byte
aligned value sizes improved the performance by roughly 5% to 15%. For 
non-4-byte aligned
value sizes, there wasn't a noticeable difference between using jhash2 
(and truncating the
remainder bits) vs. using jhash.

May be we enforce 4-byte sized value for simplicity?

Sounds great! And if in the future this becomes too restrictive, we 
could always loosen this
as well