Re: [PATCH nft] mnl: continue monitor if errno is ESRCH

[Pablo Neira Ayuso <pablo@netfilter.org>]

On Sun, Feb 26, 2017 at 05:30:58PM +0100, Alexander Alemayhu wrote:
> Running the test cases in the shell directory while running nft monitor results
> in nft exiting with '# ERROR: No such process'. The minimal steps where I could
> reproduce is:
> 
> nft monitor # shell 1
> run-tests.sh testcases/sets/0011add_many_elements_0 # shell 2
> 
> Signed-off-by: Alexander Alemayhu <alexander@alemayhu.com>
> ---
> 
> Not sure if this is considered a fix or desired behaviour, but I was not
> expecting monitor to exit like this.
> 
>  src/mnl.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/src/mnl.c b/src/mnl.c
> index 295dd84a5840..a0066a28b44f 100644
> --- a/src/mnl.c
> +++ b/src/mnl.c
> @@ -1129,7 +1129,10 @@ int mnl_nft_event_listener(struct mnl_socket *nf_sock,
>  				printf("# ERROR: We lost some netlink events!\n");
>  				continue;
>  			}
> +
>  			fprintf(stdout, "# ERROR: %s\n", strerror(errno));
> +			if (errno == ESRCH)
> +				continue;

It seems netlink is returning ESRCH when the number of events is high,
however, it should hit ENOBUFS instead, this is strange.