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] [thread-next>] [day] [month] [year] [list]
Message-ID: <52A52A5D.5060601@huawei.com>
Date:	Mon, 9 Dec 2013 10:26:37 +0800
From:	Yang Yingliang <yangyingliang@...wei.com>
To:	David Laight <David.Laight@...LAB.COM>, <davem@...emloft.net>,
	<netdev@...r.kernel.org>
CC:	<eric.dumazet@...il.com>, <brouer@...hat.com>, <jpirko@...hat.com>,
	<jbrouer@...hat.com>
Subject: Re: [PATCH net v6 1/2] net: sched: tbf: fix the calculation of max_size

On 2013/12/6 18:56, David Laight wrote:
>> From: Yang Yingliang
> ...
>>
>> +/* Time to Length, convert time in ns to length in bytes
>> + * to determinate how many bytes can be sent in given time.
>> + */
>> +static u64 psched_ns_t2l(const struct psched_ratecfg *r,
>> +			 u64 time_in_ns)
>> +{
>> +	/* The formula is :
>> +	 * len = (time_in_ns * r->rate_bytes_ps) / NSEC_PER_SEC
>> +	 */
>> +	u64 len = time_in_ns * r->rate_bytes_ps;
>> +
>> +	do_div(len, NSEC_PER_SEC);
> 
> You are multiplying two values then dividing by 10**9
> I'd guess that the intermediate value might exceed 2**64.

I thought the max value of len is burst which is a type of u32 sent
userland, so the max value of (time_in_ns * r->rate_bytes_ps)
should be 64K*(10**9).
Hmm, maybe I should do something in this helper to avoid overflow.

> 
>> +	if (unlikely(r->linklayer == TC_LINKLAYER_ATM))
>> +		len = (len / 53) * 48;
> 
> You probably want to do the multiply first.
> But why not scale rate_bytes_ps instead.
> 
>> +	if (len > r->overhead)
>> +		len -= r->overhead;
>> +	else
>> +		len = 0;
>> +
>> +	return len;
>> +}
> 
> Personally I'd work out how much time you have to send each byte.
> So if you want a rate of 1MB/sec you have a 'time cost' per byte of 1000ns.
> The cost of sending a packet is simply the length multiplied by this cost.
> To work out whether a packet can be sent you have a credit variable that
> tracks current time.
> If 'credit > now' the packet can't be sent, queue and schedule a wakeup.
> if 'credit + backlog < now' set credit = now - backlog.
> if 'credit <= now' send the packet and add the packet's 'cost' to 'credit'.
> 
> In the non-ratelimited case this is almost no work.
> 
> You'd probably need to work in 1/1024ns time units and/or blocks of 16 bytes.
> 
> 	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

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ