Re: [PATCH] sctp: fix the check for path failure detection

[Luo Chunbo]

On Mon, 2009-08-24 at 15:54 +0800, Wei Yongjun wrote:
> Luo Chunbo åé:
> > On Fri, 2009-08-21 at 17:47 -0400, Vlad Yasevich wrote:
> >   
> >> Chunbo Luo wrote:
> >>     
> >>> The transport is marked DOWN immediately after sending the max+1 HB,
> >>> which is equal to not sending the max+1 HB at all. We should wait
> >>> a next period and make sure the last HB is not acknowledged.
> >>>
> >>>       
> >> I don't think this code does what you want either...
> >>
> >> Let's say path_max_rxt = 2.  What we'll get is:
> >> 	timeout:
> >> 		err++ (1)
> >> 		if (err > 2) false
> >> 		send HB
> >> 		reset timer
> >> 	timeout:
> >> 		err++ (2)
> >> 		if (err > 2) false
> >> 		send HB
> >> 		reset timer
> >> 	timeout:
> >> 		err++ (3)
> >> 		if (err > 2)
> >> 			set transport DOWN
> >> 		send HB
> >> 		reset timer.
> >>
> >> We only had 2 unacknowledged HB when we should have had 3.
> >>     
> >
> > The error count is increment after the HB was sent, and the error count
> > check is before sending HB.
> >
> > Let's say path_max_rxt =2 . What we really get is:
> >
> > timeout:
> > 	if( err > 2) false
> > 	send HB
> > 	err++ (1)
> > 	reset timer
> > timeout:
> > 	if( err > 2) false
> > 	send HB
> > 	err++ (2)
> > 	reset timer
> > timeout:
> > 	if( err > 2) false
> > 	send HB
> > 	err++ (3)
> > 	reset timer
> > timeout:
> > 	if( err > 2)
> > 		set transport DOWN
> > 	send HB
> > 	reset timer
> >
> > Here We had 3 unacknowledged HBs
> 
> 
> But with Vlad's advice, you just need to do little change to the kernel
> codes, may be two lines's patch, and it can do the same thing.


OK, I'll send a new patch later.


> 
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/