[PATCH 1/2] srcu: Prevent sdp->srcu_gp_seq_needed_exp counter wrap

[PATCH 1/2] srcu: Prevent sdp->srcu_gp_seq_needed_exp counter wrap

[Ildar Ismagilov]

Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
---
 kernel/rcu/srcutree.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 1798d1aa60fa..c54617280245 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -552,6 +552,9 @@ static void srcu_gp_end(struct srcu_struct *sp)
 				if (ULONG_CMP_GE(gpseq,
 						 sdp->srcu_gp_seq_needed + 100))
 					sdp->srcu_gp_seq_needed = gpseq;
+				if (ULONG_CMP_GE(gpseq,
+						 sdp->srcu_gp_seq_needed_exp + 100))
+					sdp->srcu_gp_seq_needed_exp = gpseq;
 				raw_spin_unlock_irqrestore_rcu_node(sdp, flags);
 			}
 	}
-- 
2.11.0

Re: [PATCH 1/2] srcu: Prevent sdp->srcu_gp_seq_needed_exp counter wrap

[Paul E. McKenney]

On Wed, Jan 31, 2018 at 10:39:53PM +0300, Ildar Ismagilov wrote:
> Signed-off-by: Ildar Ismagilov <devix84@gmail.com>

I did apply this for testing and review, thank you and good eyes!

However, I had to:

1.	hand-edit it to make it work against the rcu/dev
	branch of my -rcu tree:
	git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git

2.	fill out the commit logs.

Please see below for the updated patch, and please let me know if I
messed something up.

In the future, could you please apply your patches against the rcu/dev
branch of -rcu and provide more complete commit logs?  The commit log
should describe the problem and how that problem is fixed.

Also, are you able to test these using rcutorture?  If not, let's get you
set up for that, as I am prone to testing backlogs just now.  Plus I am
writing this while 34,000 feet over Western Australia.  So it would mean
fewer delays for you if you could test your own patches.  ;-)

							Thanx, Paul

------------------------------------------------------------------------

commit 753abb606ba61012c8838c73e2c888aea20efe13
Author: Ildar Ismagilov <devix84@gmail.com>
Date:   Wed Jan 31 22:39:53 2018 +0300

    srcu: Prevent sdp->srcu_gp_seq_needed_exp counter wrap
    
    SRCU checks each srcu_data structure's grace-period number for counter
    wrap four times per cycle by default.  This frequency guarantees that
    normal comparisons will detect potential wrap.  However, the expedited
    grace-period number is not checked.  The consquences are not too horrible
    (a failure to expedite a grace period when requested), but it would be
    good to avoid such things.  This commit therefore adds this check to
    the expedited grace-period number.
    
    Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
    Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 2878387d4189..ad3e1aa5e6ea 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -579,6 +579,9 @@ static void srcu_gp_end(struct srcu_struct *sp)
 				if (ULONG_CMP_GE(gpseq,
 						 sdp->srcu_gp_seq_needed + 100))
 					sdp->srcu_gp_seq_needed = gpseq;
+				if (ULONG_CMP_GE(gpseq,
+						 sdp->srcu_gp_seq_needed_exp + 100))
+					sdp->srcu_gp_seq_needed_exp = gpseq;
 				spin_unlock_irqrestore_rcu_node(sdp, flags);
 			}
 	}

Re: [PATCH 1/2] srcu: Prevent sdp->srcu_gp_seq_needed_exp counter wrap

[Paul E. McKenney]

On Tue, Feb 06, 2018 at 01:44:10PM +0300, Ildar Ismagilov wrote:
> The results of rcutorture tests in attachment. I only tested: SRCU-N,
> SRCU-P, SRCU-t and SRCU-u configs.

That is a good set for an SRUC change, thank you!

							Thanx, Paul

> 2018-02-03 12:16 GMT+03:00 Paul E. McKenney <paulmck@linux.vnet.ibm.com>:
> > On Wed, Jan 31, 2018 at 10:39:53PM +0300, Ildar Ismagilov wrote:
> >> Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
> >
> > I did apply this for testing and review, thank you and good eyes!
> >
> > However, I had to:
> >
> > 1.      hand-edit it to make it work against the rcu/dev
> >         branch of my -rcu tree:
> >         git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git
> >
> > 2.      fill out the commit logs.
> >
> > Please see below for the updated patch, and please let me know if I
> > messed something up.
> >
> > In the future, could you please apply your patches against the rcu/dev
> > branch of -rcu and provide more complete commit logs?  The commit log
> > should describe the problem and how that problem is fixed.
> >
> > Also, are you able to test these using rcutorture?  If not, let's get you
> > set up for that, as I am prone to testing backlogs just now.  Plus I am
> > writing this while 34,000 feet over Western Australia.  So it would mean
> > fewer delays for you if you could test your own patches.  ;-)
> >
> >                                                         Thanx, Paul
> >
> > ------------------------------------------------------------------------
> >
> > commit 753abb606ba61012c8838c73e2c888aea20efe13
> > Author: Ildar Ismagilov <devix84@gmail.com>
> > Date:   Wed Jan 31 22:39:53 2018 +0300
> >
> >     srcu: Prevent sdp->srcu_gp_seq_needed_exp counter wrap
> >
> >     SRCU checks each srcu_data structure's grace-period number for counter
> >     wrap four times per cycle by default.  This frequency guarantees that
> >     normal comparisons will detect potential wrap.  However, the expedited
> >     grace-period number is not checked.  The consquences are not too horrible
> >     (a failure to expedite a grace period when requested), but it would be
> >     good to avoid such things.  This commit therefore adds this check to
> >     the expedited grace-period number.
> >
> >     Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
> >     Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> >
> > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> > index 2878387d4189..ad3e1aa5e6ea 100644
> > --- a/kernel/rcu/srcutree.c
> > +++ b/kernel/rcu/srcutree.c
> > @@ -579,6 +579,9 @@ static void srcu_gp_end(struct srcu_struct *sp)
> >                                 if (ULONG_CMP_GE(gpseq,
> >                                                  sdp->srcu_gp_seq_needed + 100))
> >                                         sdp->srcu_gp_seq_needed = gpseq;
> > +                               if (ULONG_CMP_GE(gpseq,
> > +                                                sdp->srcu_gp_seq_needed_exp + 100))
> > +                                       sdp->srcu_gp_seq_needed_exp = gpseq;
> >                                 spin_unlock_irqrestore_rcu_node(sdp, flags);
> >                         }
> >         }
> >
> 
> 
> 
> -- 
> Best regards
> Ildar