On Thu, May 30, 2019 at 12:55:27PM +0200, Pablo Neira Ayuso wrote:
> Set a receiver buffer size based on the number of commands and the
> average message size, this is useful for the --echo option in order to
> avoid ENOBUFS errors.
>
> Double the estimated size is used to ensure enough receiver buffer
> space.
>
> Skip buffer receiver logic if estimation is smaller than current buffer.
>
> Reported-by: Phil Sutter <[email protected]>
> Signed-off-by: Pablo Neira Ayuso <[email protected]>
> ---
[..]
> diff --git a/src/libnftables.c b/src/libnftables.c
> index 199dbc97b801..a58b8ca9dcf6 100644
> --- a/src/libnftables.c
> +++ b/src/libnftables.c
[..]
> @@ -308,14 +310,17 @@ int mnl_batch_talk(struct netlink_ctx *ctx, struct
> list_head *err_list)
> .tv_sec = 0,
> .tv_usec = 0
> };
> - fd_set readfds;
> struct iovec iov[iov_len];
> struct msghdr msg = {};
> + fd_set readfds;
> int err = 0;
>
> mnl_set_sndbuffer(ctx->nft->nf_sock, ctx->batch);
>
> - mnl_nft_batch_to_msg(ctx, &msg, &snl, iov, iov_len);
> + batch_size = mnl_nft_batch_to_msg(ctx, &msg, &snl, iov, iov_len);
> + avg_msg_size = div_round_up(batch_size, num_cmds);
> +
> + mnl_set_rcvbuffer(ctx->nft->nf_sock, num_cmds * avg_msg_size * 2);
I think this calculation is incorrect. I'm still getting ENOBUFS with
Phil's testcase and firewalld's testsuite (large json blob). I changed
the multiplier from 2 to 6 and it worked.
-->8--
# ./run-tests.sh ./testcases/transactions/0049huge_0
I: using nft binary ./../../src/nft
W: [FAILED] ./testcases/transactions/0049huge_0: got 1
netlink: Error: Could not process rule: No buffer space available
I: results: [OK] 0 [FAILED] 1 [TOTAL] 1
