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:	Mon, 09 Jul 2007 19:43:40 +0300
From:	Ranko Zivojnovic <ranko@...dernet.net>
To:	Patrick McHardy <kaber@...sh.net>
Cc:	Jarek Poplawski <jarkao2@...pl>, akpm@...ux-foundation.org,
	netdev@...r.kernel.org
Subject: Re: + gen_estimator-fix-locking-and-timer-related-bugs.patch added
	to -mm tree

On Mon, 2007-07-09 at 15:52 +0200, Patrick McHardy wrote:
> Ranko Zivojnovic wrote:
> > Patrick, I've taken liberty to try and implement this myself. Attached
> > is the whole new gen_estimator-fix-locking-and-timer-related-bugs.patch
> > that is RCU lists based. Please be kind to review.
> 
> Thanks for taking care of this, I'm a bit behind.

Glad to be able to help.

> See below for comments ..
> 
> > --- a/net/core/gen_estimator.c	2007-06-25 02:21:48.000000000 +0300
> > +++ b/net/core/gen_estimator.c	2007-07-09 14:27:12.053336875 +0300
> > @@ -79,7 +79,7 @@
> >  
> >  struct gen_estimator
> >  {
> > -	struct gen_estimator	*next;
> > +	struct list_head	list;
> >  	struct gnet_stats_basic	*bstats;
> >  	struct gnet_stats_rate_est	*rate_est;
> >  	spinlock_t		*stats_lock;
> > @@ -89,26 +89,27 @@
> >  	u32			last_packets;
> >  	u32			avpps;
> >  	u32			avbps;
> > +	struct rcu_head		e_rcu;
> 
> 
> You could also use synchronize_rcu(), estimator destruction is
> not particulary performance critical.

I've tried synchronize_rcu(), but it went kaboom with the below, thus
call_rcu() must stay:

---cut---
BUG: sleeping function called from invalid context at kernel/sched.c:3928
in_atomic():1, irqs_disabled():0
3 locks held by tc/2933:
 #0:  (rtnl_mutex){--..}, at: [<c0296e0a>] rtnetlink_rcv+0x18/0x42
 #1:  (&dev->queue_lock){-+..}, at: [<c029c337>] qdisc_lock_tree+0xe/0x1c
 #2:  (&dev->ingress_lock){-...}, at: [<c029dfb8>] tc_get_qdisc+0x192/0x1e9
 [<c02eec9e>] wait_for_completion+0x1a/0xb7
 [<c01e3cd4>] __spin_lock_init+0x29/0x4b
 [<c012e410>] synchronize_rcu+0x2a/0x2f
 [<c012e0bc>] wakeme_after_rcu+0x0/0x8
 [<c028da00>] gen_kill_estimator+0x70/0x9b
 [<f8b7c5a6>] htb_destroy_class+0x27/0x12f [sch_htb]
 [<f8b7d037>] htb_destroy+0x34/0x70 [sch_htb]
 [<c029c554>] qdisc_destroy+0x44/0x69
 . 
 .
 .
---cut---

> >  static struct gen_estimator_head elist[EST_MAX_INTERVAL+1];
> >  
> >  /* Estimator array lock */
> > -static DEFINE_RWLOCK(est_lock);
> > +static DEFINE_SPINLOCK(est_lock);
> 
> 
> The lock isn't needed anymore since we can rely on the rtnl
> for creation/destruction.

The 'est_lock' now completely removed.

> >  /**
> > @@ -152,6 +153,7 @@
> >  {
> >  	struct gen_estimator *est;
> >  	struct gnet_estimator *parm = RTA_DATA(opt);
> > +	int idx;
> >  
> >  	if (RTA_PAYLOAD(opt) < sizeof(*parm))
> >  		return -EINVAL;
> > @@ -163,7 +165,8 @@
> >  	if (est == NULL)
> >  		return -ENOBUFS;
> >  
> > -	est->interval = parm->interval + 2;
> > +	INIT_LIST_HEAD(&est->list);
> 
> 
> Initializing the members' list_head isn't necessary.
> 
Removed.

All the changes above now included in the attached patch. I've clean
compiled it with 2.6.22 and now testing this version.

R.

View attachment "gen_estimator-fix-locking-and-timer-related-bugs.patch" of type "text/x-patch" (3636 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ