netdev - Re: [PATCH] net: remove redundant checking for sock timer state

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]

Date:	Thu, 31 Jan 2013 23:21:04 -0800
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Xue Ying <ying.xue0@...il.com>
Cc:	David Miller <davem@...emloft.net>, ying.xue@...driver.com,
	netdev@...r.kernel.org
Subject: Re: [PATCH] net: remove redundant checking for sock timer state

On Fri, 2013-02-01 at 15:14 +0800, Xue Ying wrote:
> Eric Dumazet wrote:

> > I had the same reaction but maybe its not anymore a valid thing.
> >
> > Before commit 55c888d6d ([PATCH] timers fixes/improvements) there was
> > indeed a significant cost calling del_timer() because of unconditional
> > spinlock acquisition.
> >
> > But nowadays del_timer() doesn't blindly lock the spinlock.
> >
> > So I guess we could change all occurrences of :
> >
> > if (timer_pending(X))
> >     del_timer(X);
> >
> > It would save some bytes of code.
> >   
> Eric, thanks for your explanation and suggestion.
> 
> But I cannot understand why we should first call timer_pending() before 
> del_timer() in your proposal.
> By my understanding, we might get an unreal timer pending state out of 
> timer base lock (ie, lock_timer_base()),
> and the "unreal" is only for pending state, on the contrary, the value 
> is real for inactive sate.
> So calling timer_pending() out of timer base lock scope can make us 
> avoid some unnecessary grabbing spin lock operations.
> However, in del_timer() there already has placed a timer_pending() 
> before lock_timer_base() is called. So why do we need
> another before calling del_timer()?
> 
> Please you explain more.

I think you misunderstood me.

I said that the old construct :

if (timer_pending(X))
    del_timer(X);

could be changed to

del_timer(X);

Thats what your patch does.

But instead of changing sk_stop_timer(), I suggested you do a patch on
whole net tree.



--
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