[802.3_MAINT] Corrigenda? Timing considerations for PAUSE issue
While working on the max delay for a link partner to react to priority based flow control, I ran into an inconsistancy in 802.3. The PAUSE reaction delay in 31B.3.7 is not sufficient for all PHYs. A corrigenda should be considered for fixing this because it is fairly common for people allocating the buffers needed for non-drop operation with PAUSE to rely on the value they read in 31B.3.7
31B.3.7 Timing considerations for PAUSE operation allows 60 pause_quantum bit times for the reaction to the reception of PAUSE:
"At operating speeds of 10 Gb/s and above, a station shall not begin to transmit a (new) frame more than sixty
pause_quantum bit times after the reception of a valid PAUSE frame that contains a non-zero value of
pause_time, as measured at the MDI."
That was the maximum when we completed the first 10 Gig addition to 802.3. That reflected the delays in Table 44-2 for a MAC plus the longest delay PHY at that time: a pair of XGXS sublayers, 10GBASE-R PCS, WIS and PMD.
But since then we have added other PHYs or PMDs and some have a higher delay.
When we added the later PHYs we didn't change the value in Clause 31B but the value isn't sufficient.
I'm particularly concerned about 10BASE-T - The 10GBASE-T PHY is allowed a delay of 50. Add 16 for the MAC, RS and MAC Control and 8 for XGXS and the delay is 74.
10GBASE-KR without FEC also slightly breaks the 60 pause quanta - its PMD allows 2 pause_quantum bit times of round trip delay. Given how short the medium delay is for this PHY, the extra pause_quantum isn't going to break any upper layer implementation that is built for 100 m links over other PHYs.
FEC adds another 12 pause quanta (6144 bit times) so 10GBASE-KR with FEC can have 73 pause_quantum bit times of delay - almost the same as 10GBASE-T.
Another instance occurs if one runs WIS over LRM. In that case one has a PMD with 18 pause quanta of delay replacing one with 1 so the totoal delay would be 77, but I think this can be ignored as outside the standard. There isn't a WIS PMD type for LRM and it specifies only the signaling rate for 10GBASE-R without a WIS.
-------------------------------------------
The original calculation
Delay sublayer
16 MAC, RS and MAC Control
8 an XGXS pair plus XAUI
7 10GBASE-R PCS
28 WIS
1 clause 52, 53 or 54 PMD not LRM
---
60