[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20171115095856.591ee0a9@xeon-e3>
Date: Wed, 15 Nov 2017 09:58:56 -0800
From: Stephen Hemminger <stephen@...workplumber.org>
To: Nishanth Devarajan <ndev2021@...il.com>
Cc: netdev@...r.kernel.org, doucette@...edu, michel.machado@...il.com
Subject: Re: [PATCH iproute2/net-next v2]tc: B.W limits can now be specified
in %.
On Wed, 15 Nov 2017 07:06:21 +0530
Nishanth Devarajan <ndev2021@...il.com> wrote:
> int parse_percent_rate(char *rate, const char *str, char *dev)
You aren't modifyin dev so it should be const char *
> +{
> + long max_rate_bits;
> + int ret, saved_errno;
> + double perc, rate_bits;
> + char *str_perc, *p;
> +
> + if (!dev[0]) {
> + fprintf(stderr, "No device specified; specify device to rate limit by percentage\n");
> + return -1;
> + }
> +
> + if (read_prop(dev, "speed", &max_rate_bits))
> + return -1;
If speed is unknown, then many will devices will return -1.
You need to handle that.
Since speed reported by kernel is in mbits per second, it
would make sense to rename max_rate_bits to mbit.
> +
> + ret = sscanf(str, "%m[0-9.%]", &str_perc);
> + if (ret != 1)
> + goto malf;
> +
> + /* Make sure there's only one percent sign and it's at the end */
> + perc = strtod(str_perc, &p);
> + if (*p != '%' || *(p++) != '\0')
> + goto malf;
There already is parse_percent in tc/q_netem.c.
Please move that to util and use that instead of coding here.
> +
> + saved_errno = errno;
> + free(str_perc);
> +
> + if (perc > 100.0 || perc < 0.0 || saved_errno == ERANGE) {
> + fprintf(stderr, "Invalid rate specified; should be between [0,100]%% but is %s\n", str);
> + return -1;
> + }
> +
> + rate_bits = (perc * max_rate_bits) / 100.0;
> +
> + ret = snprintf(rate, 20, "%lf", rate_bits);
> + if (ret <= 0 || ret >= 20) {
> + fprintf(stderr, "Unable to parse calculated rate\n");
> + return -1;
> + }
> +
> + return 0;
> +
> +malf:
> + fprintf(stderr, "Specified rate value could not be read or is malformed\n");
> + return -1;
> +}
Powered by blists - more mailing lists