Re: [PATCH RFC bpf-next 01/20] trait: limited KV store for packet metadata

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Mar 5, 2025 at 6:33 AM <arthur@xxxxxxxxxxxxxxx> wrote:
>
> +struct __trait_hdr {
> +       /* Values are stored ordered by key, immediately after the header.
> +        *
> +        * The size of each value set is stored in the header as two bits:
> +        *  - 00: Not set.
> +        *  - 01: 2 bytes.
> +        *  - 10: 4 bytes.
> +        *  - 11: 8 bytes.

...

> +        *  - hweight(low) + hweight(high)<<1 is offset.

the comment doesn't match the code

> +        */
> +       u64 high;
> +       u64 low;

...

> +static __always_inline int __trait_total_length(struct __trait_hdr h)
> +{
> +       return (hweight64(h.low) << 1) + (hweight64(h.high) << 2)
> +               // For size 8, we only get 4+2=6. Add another 2 in.
> +               + (hweight64(h.high & h.low) << 1);
> +}

This is really cool idea, but 2 byte size doesn't feel that useful.
How about:
- 00: Not set.
- 01: 4 bytes.
- 10: 8 bytes.
- 11: 16 bytes.

4 byte may be useful for ipv4, 16 for ipv6, and 8 is just a good number.
And compute the same way with 3 popcount with extra +1 to shifts.





[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux