netdev - Re: [PATCH] [RFC] tcp: TSQ - do not always throttle.

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: <1342530654.2626.563.camel@edumazet-glaptop>
Date:	Tue, 17 Jul 2012 15:10:54 +0200
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Krishna Kumar <krkumar2@...ibm.com>
Cc:	davem@...emloft.net, netdev@...r.kernel.org
Subject: Re: [PATCH] [RFC] tcp: TSQ - do not always throttle.

On Tue, 2012-07-17 at 17:33 +0530, Krishna Kumar wrote:
> Do not throttle if sysctl_tcp_limit_output_bytes==0.
> 
> Maybe it is better to throttle earlier in the loop, after
> calling tcp_init_tso_segs().
> 

I wonder why, and why you put this question in a changelog instead of
outside of it...

Idea was to avoid setting TSQ_THROTTLED if we break out the loop.

About disabling TSQ, my initial intent was to instead use a negative
sysctl_tcp_limit_output_bytes value.

Thats why I have in tcp_transmit_skb() :

skb->destructor = (sysctl_tcp_limit_output_bytes > 0) ?
		  tcp_wfree : sock_wfree;

So I suggest you change the tcp_write_xmit(() test to a single unsigned
compare :

if (atomic_read(&sk->sk_wmem_alloc) >=
    (unsigned) sysctl_tcp_limit_output_bytes) {

Also use :

skb->destructor = (sysctl_tcp_limit_output_bytes >= 0) ?
  tcp_wfree : sock_wfree;

and document the 'negative value disables TSQ' in
Documentation/networking/ip-sysctl.txt

--
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