netdev - RE: [PATCH v1 net-next] net: pkt

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Mon, 30 Sep 2013 10:19:32 +0100
From:	"David Laight" <David.Laight@...LAB.COM>
To:	"Vijay Subramanian" <subramanian.vijay@...il.com>,
	<netdev@...r.kernel.org>
Cc:	<davem@...emloft.net>, <shemminger@...tta.com>,
	<eric.dumazet@...il.com>, "Mythili Prabhu" <mysuryan@...co.com>,
	"Dave Taht" <dave.taht@...ferbloat.net>
Subject: RE: [PATCH v1 net-next] net: pkt_sched: PIE AQM scheme

> +#define PIE_DEFAULT_QUEUE_LIMIT 200	/* in packets */
> +#define QUEUE_THRESHOLD (5000)
> +#define DQCOUNT_INVALID -1
> +#define THRESHOLD_PKT_SIZE	1500
> +#define MAX_INT_VALUE  0xffffffff
> +#define MAX_INT_VALUE_CAP  (0xffffffff >> 8)

The above 2 constants are both unsigned (unless 'int' is larger than 32 bits).
If nothing else this means that they are very badly named.

> +
> +typedef u32 pie_time_t;
> +typedef s32 pie_tdiff_t;
> +#define PIE_SHIFT 10
> +#define MS2PIETIME(a) ((a * NSEC_PER_MSEC) >> PIE_SHIFT)
> +#define PIE_TIME_PER_SEC  ((NSEC_PER_SEC >> PIE_SHIFT))
> +
> +static inline pie_time_t pie_get_time(void)
> +{
> +	u64 ns = ktime_to_ns(ktime_get());
> +	return ns >> PIE_SHIFT;
> +}
> +
> +static inline u32 pie_time_to_ms(pie_time_t val)
> +{
> +	u64 valms = ((u64) val << PIE_SHIFT);
> +
> +	do_div(valms, NSEC_PER_MSEC);
> +	return (u32) valms;
> +}

There seems to be a lot of conversions between 1/1000 ms and 1/1024 ms.
On 32 bit systems they are horrid.
Not only that the conversions are open coded a lot of times as well.
It might also be better replacing '(val * 1000u) >> 10' with
'(val * (u64)(1000 << (32 - 10))) >> 32' since I'm not sure the compiler
will perform that substitution.

What happens if pie_time_to_ms() overflows 32 bits?
If it is only ever called for small values, there may be no need for
64bit maths at all.

...
> +	if (delta > (s32) (MAX_INT_VALUE * 2 / 100)
> +	    && q->vars.prob >= MAX_INT_VALUE / 10) {
> +		delta = MAX_INT_VALUE * 2 / 100;

Calculating 'MAX_INT_VALUE * 2' isn't a good idea!
Whatever value it is supposed to be, it should be a named constant.

	David

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html