netdev - Re: TSO trimming question

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: <Pine.LNX.4.64.0712201417470.11007@kivilampi-30.cs.helsinki.fi>
Date:	Thu, 20 Dec 2007 14:35:19 +0200 (EET)
From:	"Ilpo Järvinen" <ilpo.jarvinen@...sinki.fi>
To:	David Miller <davem@...emloft.net>
cc:	Netdev <netdev@...r.kernel.org>,
	Herbert Xu <herbert@...dor.apana.org.au>
Subject: Re: TSO trimming question

On Thu, 20 Dec 2007, David Miller wrote:

> From: "Ilpo_Järvinen" <ilpo.jarvinen@...sinki.fi>
> Date: Thu, 20 Dec 2007 13:40:51 +0200 (EET)
> 
> > That's not the only case, IMHO if there's odd boundary due to 
> > snd_una+snd_wnd - skb->seq limit (done in tcp_window_allows()), we don't 
> > consider it as odd but break the skb at arbitary point resulting
> > two small segments to the network, and what's worse, when the later skb 
> > resulting from the first split is matching skb->len < limit check as well 
> > causing an unnecessary small skb to be created for nagle purpose too, 
> > solving it fully requires some thought in case the mss_now != mss_cache 
> > even if non-odd boundaries are honored in the middle of skb.
> 
> In the most ideal sense, tcp_window_allows() should probably
> be changed to only return MSS multiples.

That's what Herbert suggested already, I'll send a patch later
on... :-)

> Unfortunately this would add an expensive modulo operation,
> however I think it would elimiate this problem case.

Yes. Should we still call tcp_minshall_update() if split in the middle of 
wq results in smaller than MSS tail (occurs only if mss_now != mss_cache)?

-- 
 i.