netdev - RE: [PATCH net v6 1/2] net: sched: tbf: fix the calculation of max

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 for Android: free password hash cracker in your pocket

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

Message-ID: <AE90C24D6B3A694183C094C60CF0A2F6026B7473@saturn3.aculab.com>
Date:	Mon, 9 Dec 2013 13:11:04 -0000
From:	"David Laight" <David.Laight@...LAB.COM>
To:	"Yang Yingliang" <yangyingliang@...wei.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

> From: Yang Yingliang [mailto:yangyingliang@...wei.com]
> On 2013/12/9 18:07, David Laight wrote:
> >> From: Yang Yingliang
> >> On 2013/12/6 18:56, David Laight wrote:
> >>>> From: Yang Yingliang
> >>> ...
> >>>
> >>> You are multiplying two values then dividing by 10**9
> >>> I'd guess that the intermediate value might exceed 2**64.
> >>>
> >>>> +	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.
> >>>
> >>
> >> When the linklayer is ATM, the formula to calculate tokens is:
> >>
> >> ((u64)(DIV_ROUND_UP(len,48)*53) * r->mult) >> r->shift
> >>
> >> So, I scale len here.
> >
> > Seems to me that there are a lot of unnecessary multiplies and
> > divides going on here.
> > Looks to me like the code was originally written to require one
> > multiply and one shift for each packet.
> 
> The way to calc tokens in psched_l2t_ns() means:
> we got a payload which length is 'len'. To get the actual size we
> need send, round len up to a multiple of 53. After getting the
> size, we do multiply and shift to calculate tokens that we need.
> 
> >
> > In any case the latter code is allowing for more of the ATM cell
> > overhead. I'm not at all sure the intent is to remove that when
> > setting up the constants.
> >
> > OTOH should this code be worrying about the packet overheads at all?
> > Does it add in the ethernet pre-emable, CRC and inter packet gap?
> 
> This overhead is sent from userspace by tc.
> 
> >
> > I'd guess that the most the ATM code should do it round the length
> > up to a multiple of 48 (user payload in a cell).
> 
> 53 is ATM cell size include header, sending a header needs tokens as
> well. I'd guess round the length up to a multiple of 53 in psched_l2t_ns()
> is necessary.

The point I'm trying to make is that in one place you multiply by 53/48
and in the other you multiply be 48/53.
These two operations almost certainly cancel each other out.
Or would modulo small rounding errors if you did the multiplies first.
So why do either of them?

Whether you should be rounding up ATM data to a whole number of cells
is a different question entirely.

	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