Date:	Tue, 6 Oct 2009 16:31:59 -0400
From:	Neil Horman <nhorman@...driver.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	Oliver Hartkopp <socketcan@...tkopp.net>,
	David Miller <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: [PATCH] af_packet: add interframe drop cmsg (v6)

On Tue, Oct 06, 2009 at 06:24:12PM +0200, Eric Dumazet wrote:
> Oliver Hartkopp a écrit :
> > Neil Horman wrote:
> >> On Tue, Oct 06, 2009 at 03:42:26PM +0200, Eric Dumazet wrote:
> >>> Neil Horman a écrit :
> >>>> Actually, no, I don't think this is sufficient.  Looking at how the
> >>>> implementation would work, we would query the sk_drop value in the recvmsg path,
> >>>> but we wouldn't record it during frame enqueue, which would skew the data I need
> >>>> to record.  Consider a queue of holding frames with sequence 1 2 3 4 101 102
> >>>> 110.  Assume that missing sequence numbers were dropped during enqueue to the
> >>>> sk_recieve_queue for lack of space.  Using the proposed above implementation,
> >>>> the frames the gap gets reported with is determined by the scheduling of the
> >>>> user space app, rather than the actual order in which they were lost.  I.e if
> >>>> frames 5-100 were received & dropped immediately after the user process called
> >>>> recvmsg for the first time, we would get the cmsg notification when we called
> >>>> recvmsg and read frame 2, but if we lost those frames after the process called
> >>>> recvmsg twice, we'd recognize the loss when we read frame 3.  With my current
> >>>> implementation we recognize the gap on the next enqueued frame after we
> >>>> encountered the loss, keeping everything in order, which is important.  We would
> >>>> need to do that with whatever socket level option we'd consider implementing.
> >>>> Not saying its not doable, but we need to take it into account, and its not a
> >>>> straightforward as simply reading sk_drops on recvmsg.  We'd have to store the
> >>>> sk_drop value in skb->mask universally or some such, and then teach sock_recvmsg
> >>>> to check that field.
> >>>>
> >>>>
> >>> Sorry cannot parse this, its too long to be true :)
> >>>
> >>> If you count every af_packet drops in sk_drops, and every time you _enqueue_
> >>> a frame in sk->receive_queue, you copy sk->sk_drops in skb->mark, you implement
> >>> the thing. All this is protected by a lock.
> >>>
> >>> sk->receive_queue is a FIFO, so there is no loss of information.
> >>>
> >>> You only need to increment sk->sk_drops when necessary, if not already done in af_packet code.
> >>>
> >>> If you dont trust me I can provide patch ;)
> >>>
> >> No I trust you.  The problem (as you just mentioned) is that you didn't parse my
> >> note.  We are saying the same thing.  If you want to implement this properly,
> >> you have to do exactly as you said, take a snapshot of the sk_drops value for
> >> the socket and store it in skb->mark, or some other unsued field, so that the
> >> drop information travels along the queue with the frames.  As I was reading
> >> Olivers note, it sounded to me as though he were proposing that we simply check
> >> sk_drops when calling sock_recvmsg, which doesn't implement the same sort of
> >> functionality.  I wanted to be sure that we were clear on what was trying to be
> >> done.
> > 
> > Puh - indeed much to read :-)
> > 
> > My intention was exactly your suggestion - as you also implemented the current
> > approach:
> > 
> > In socket_recvmsg the formerly stored (skb->mark) snapshot of sk_drops should
> > be used to create a cmsg, when the cmsg creation is enabled by a new setsockopt().
> > 
> > Did i get i right?
> > 
> 
> Yes, I could spend a moment this evening to cook a patch if you want.
> 
If you would like, I've got something half put together already though.  I'll
post it in the AM after I've tested it a bit.
Neil

--
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

Powered by blists - more mailing lists