dtaht / tc-adv Goto Github PK

15cbec5

I have observed overflowing deficit values and wondering if the commit above neglected to change the printf format from unsigned to signed?

# tc -s class show dev eth0
class cake 8005:42b parent 8005:
 (dropped 44, overlimits 0 requeues 0)
 backlog 10598b 7p requeues 0
  deficit 4294967060 count 16 dropping drop_next -4295.0s blue_prob 0
class cake 8005:4e5 parent 8005:
 (dropped 48, overlimits 0 requeues 0)
 backlog 6056b 4p requeues 0
  deficit 4294967226 count 17 dropping drop_next -4294.9s blue_prob 0
class cake 8005:7b2 parent 8005:
 (dropped 55, overlimits 0 requeues 0)
 backlog 4542b 3p requeues 0
  deficit 4294965931 count 21 dropping drop_next -4294.9s blue_prob 0

# tc -s class show dev ifb4eth0
class cake 8006:a1 parent 8006:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
  deficit 4294966517 count 1 dropping drop_next 1.1s blue_prob 0
class cake 8006:fc parent 8006:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
  deficit 1449 count 0 blue_prob 0

Maybe we could change:

	if (!atm && overhead == ethernet) {
		fprintf(f, "raw ");
	} else {
		if (atm == 1)
			fprintf(f, "atm ");
		else if (atm == 2)
			fprintf(f, "ptm ");
		else
			fprintf(f, "noatm ");

		fprintf(f, "overhead %d ", overhead);

		// This is actually the *amount* of automatic compensation, so
		// also report the actual number... keep it short
		if (ethernet) {
			 fprintf(f, "via-ethernet(%d) ", ethernet);
	}

to:

	if (!atm) {
		fprintf(f, "raw ");
	} else {
		if (atm == 1)
			fprintf(f, "atm ");
		else if (atm == 2)
			fprintf(f, "ptm ");
		else
			fprintf(f, "noatm ");
    }
    
    // unconditionally report the total accounted overhead
    // as well as the auto compensated for amount
    fprintf(f, "overhead %d ", overhead);
    fprintf(f, "(auto-compensated overhead: %d) ", ethernet);
    
    // This is actually the *amount* of automatic compensation
    if (ethernet) {
        fprintf(f, "via-ethernet ");
	}

So that total compensated overhead and automatically compensated overhead are reported unconditionally. For debugging this would help quite a lot...

Note via-ethernet then is redundant, but it might be oaky to keep it for backward compatibility with scripts parsing "tc -s" output...

@kdarbyshirebryant @rmounce

I just pulled this to see this

   TCA_CAKE_OVERHEAD,
    TCA_CAKE_MPU, nonononono
    TCA_CAKE_RTT,

When it needed to be at the end (after wash), so as to not break userspace. Is this new MPU param, in this order, in any form, anywhere we can't fix it? (e.g. lede?).

Also the pull I just merged had no wash in it in the first place for tc, and it's inserted in the wrong place on cake, also. So what I want to do is pull wash for tc-adv (where is it?), then move MPU to after wash, in both the sch_cake and tc cases.

Hi, I'm tryin to build on one machine, and transfer to another, so I'm using the DESTDIR= parameter.

The problem is that none of the executables are being installed into the DESTDIR, only all the supporting files. The executables are still being installed into their "production" locations.

I'm just not familiar enough with editing the Makefile to spot the necessary changes needed.

Or am I missing something?

As written elsewhere, the following section of tc/q_cake.c seems to require more explanation:

} else if (strcmp(*argv, "pppoe-ptm") == 0) {
atm = 2;
overhead += 27;

This is somewhat unexpected. Could you explain how you reached 27? I believe that the following to be true:

PPPoE: 2 Byte PPP + 6 Byte PPPoE = 8 Byte
VDSL2 (see IEEE 802.3-2012 61.3 and Rec. ITU-T G.992.3 (04/2009) Annex N relevant for VDSL2):
VDSL2-PTM: 1 Byte Start of Frame (S), 1 Byte End of Frame (Ck), 2 Byte TC-CRC (PTM-FCS), = 1 + 1 + 2 = 4 Byte
included ethernet header parts: 4 Byte Frame Check Sequence (FCS) + 6 Byte (dest MAC) + 6 Byte (src MAC) + 2 Byte (ethertype) = 4 + 6 + 6 + 2 = 18 Byte

Total: 8 + 4 + 18 = 30 Byte

Now, all of this could be wrong, but could you please describe why 27 is a better match for what happens on a PPPoE-PTM link?