|
|
fibre_channel - Re: [fibre_channel] Thoughts on FC-FS-5 Open Comments Cisco-001 & Cisco-
|
Message Thread: Previous | Next
|
- To: Anil Mehta <anil.mehta@xxxxxxxxxxxx>
- From: "Rajesh L G (rajeshlg)" <rajeshlg@xxxxxxxxx>
- Date: Mon, 12 Mar 2018 07:59:20 +0000
- Cc: Adrian Butter <adrian.butter@xxxxxxxxxxxxxxxxxxx>, "Mike Blair (mblair)" <mblair@xxxxxxxxx>, "fibre_channel@xxxxxxxxxxxxxxxxxxxx" <fibre_channel@xxxxxxxxxxxxxxxxxxxx>, "Abhinav Chander (achander)" <achander@xxxxxxxxx>, "Vasuki H A (avasuki)" <avasuki@xxxxxxxxx>
|
Hi Anil,
Thanks for the confirmation.
Would be good to have this documented lest somebody designs a receiver assuming 5.3.7.2 will not be violated.
Regards, Rajesh
From: Anil Mehta <anil.mehta@xxxxxxxxxxxx>
Date: Thursday, 8 March 2018 at 1:23 AM
To: "Rajesh L G (rajeshlg)" <rajeshlg@xxxxxxxxx>
Cc: Adrian Butter <adrian.butter@xxxxxxxxxxxxxxxxxxx>, "Mike Blair (mblair)" <mblair@xxxxxxxxx>, "fibre_channel@xxxxxxxxxxxxxxxxxxxx" <fibre_channel@xxxxxxxxxxxxxxxxxxxx>, "Abhinav Chander (achander)" <achander@xxxxxxxxx>, "Vasuki H A (avasuki)" <avasuki@xxxxxxxxx>
Subject: Re: [fibre_channel] Thoughts on FC-FS-5 Open Comments Cisco-001 & Cisco-002
Hi Rajesh,
Adrian is out of office. So I thought I'll respond to this.
When an FC port is generating streams for transmission it should follow all the rules mentioned in all the sections e.g. preceding and following the primitive signal with two fill words. Then
AM logic can insert AM and delete fill words to compensate for it according to the AM insertion and fill word deletion rules.
If this approach is followed then to make room for AM, fill words would be deleted from the original stream generated by the FC Port and 1 and 2 would be legal to see on the wire.
On Fri, Mar 2, 2018 at 3:46 AM, Rajesh L G (rajeshlg) <rajeshlg@xxxxxxxxx>
wrote:
Hi Adrian,
Quoting the section relevant to my question:
“If an FC_Port transmitter is not repeating the Transmission Word stream of a receiver, the FC_Port shall
transmit at least six Primitive Signals between each EOF and the next SOF, unless Alignment Markers are
being inserted or rate adaptation requires Primitive Signal deletion. Of the six or more Primitive Signals
transmitted, at least four shall be Fill Words. If Alignment Marker insertion or rate adaptation require
deletion of Primitive Signals, then the FC_Port shall retain at least two Fill Words consecutively following
each EOF.”
This text seems to indicate that in streams that have alignment marker elsewhere, the following are legal.
-
EOF-IDLE-IDLE-RRDY-SOF
-
EOF-IDLE-IDLE-RRDY-IDLE-IDLE-RRDY-IDLE-SOF
But this violates the rules mentioned in other sections, viz.:
5.3.7.2
To assure a sufficient number of Fill Words between frames, the originator of any Primitive Signal other than Idle
shall precede and follow the Primitive Signal by a minimum of two Fill Words.
Are the examples I gave above legal?
Regards, Rajesh
Hi Rajesh,
I think the relevant rules to mention in this regard are contained in 11.3.3:
=======================================================
11.3.3 Frame Transmission
Frame transmission shall be performed by inserting a frame immediately following a series of Fill Words
(see 11.3.2). Fill Words shall be transmitted immediately upon completion of the frame. A minimum of two
Fill Words shall be transmitted consecutively following the EOF of each frame transmitted by any FC_Port
If an FC_Port transmitter is repeating the Transmission Word stream of a receiver that is not capable of
exercising buffer-to-buffer flow control (e.g., L_Ports with REPEAT true), the transmitter may insert or
remove Fill Words from the Transmission Word stream in order to adjust for timing skew between the
receiver and transmitter; however, it shall retain at least two Fill Words consecutively following each EOF.
If an FC_Port transmitter is repeating the Transmission Word stream of a receiver that is capable of
exercising buffer-to-buffer flow control, the FC_Port shall transmit at least six Primitive Signals between
each EOF and the next SOF. Of the six or more Primitive Signals transmitted, at least four shall be Fill
If an FC_Port transmitter is not repeating the Transmission Word stream of a receiver, the FC_Port shall
transmit at least six Primitive Signals between each EOF and the next SOF, unless Alignment Markers are
being inserted or rate adaptation requires Primitive Signal deletion. Of the six or more Primitive Signals
transmitted, at least four shall be Fill Words. If Alignment Marker insertion or rate adaptation require
deletion of Primitive Signals, then the FC_Port shall retain at least two Fill Words consecutively following
See FC-AL-2 for Arbitrated Loop specific frame transmission requirements. See 11.3.9 for frame reception
=======================================================
Does that help address the question/concern?
---------- Forwarded message ----------
From: Rajesh L G (rajeshlg) <rajeshlg@xxxxxxxxx>
Date: Tue, Feb 27, 2018 at 11:00 AM
Subject: Re: [fibre_channel] Thoughts on FC-FS-5 Open Comments Cisco-001 & Cisco-002
To: Anil Mehta <anil.mehta@xxxxxxxxxxxx>,
"Mike Blair (mblair)" <mblair@xxxxxxxxx>
Cc: Adrian Butter <adrian.butter@xxxxxxxxxxxxxxxxxxx>,
"fibre_channel@xxxxxxxxxxxxxxxxxxxx"
<fibre_channel@xxxxxxxxxxxxxxxxxxxx>,
"Abhinav Chander (achander)" <achander@xxxxxxxxx>,
"Vasuki H A (avasuki)" <avasuki@xxxxxxxxx>
Hi Anil, Adrian,
We have had some people interpreting that the 2 IDLE rule is applicable only for slower speeds. Hence seeking the
clarification.
Depending on implementation, IDLEs may need to be deleted to make space for alignment marker. Say, if one IDLE each
is deleted in 4 inter-packet gaps and this resulted in violating the 2 IDLE rule, is that compliant?
Regards, Rajesh
On Mon, Feb 26, 2018 at 12:10 PM, Mike Blair <mblair@xxxxxxxxx>
wrote:
Hi Adrian,Anil,
We had a discussion in Cisco today and the real question we are wondering on has not really been addressed (or at least I dont think so). Our question relates to the difference in the length of an idle (32 bits) and the alignment marker (257 bits). It is going
to take the removal of many idles to "make room" for the insertion of a single alignment marker at the transmitter. This means that there are going to need to be a number of places that we violate the 2 fill words in order to find enough idles to remove such
that a alignment marker can be inserted. A similar issue is on the receiver side.
I am adding to the email thread a few of our hardware folks to help with the discussion.
Thanks ... Mike
On Mon, 26 Feb 2018, Anil Mehta wrote:
Hi Adrian/Raul,I agree with your comments on this issue. For Cisco-001, I think we can just simply delete the following sentence:
The alignment marker has the form of a specially defined 66-bit block with a control transmission word
synchronization header.
Just continue the first paragraph in 5.5.3 with: These markers interrupt ....
For Cisco-002 I agree with what Adrian stated. Furthermore, we have been working on a proposal to allow the remote link partner to query what the parameters were set to when
a degrade indication is received and take
administrative action to change them if desired.
Anil
On Fri, Feb 23, 2018 at 2:43 PM, Adrian Butter <adrian.butter@xxxxxxxxxxxxxxxxxxx>
wrote:
Hi Raul,
Thanks for the feedback on Cisco-001. And I do agree with your point about the AM format being a specially defined 257-bit pattern (rather than 66 bit). I hadn't
considered that when I reused the 802.3 Clause 82 text. I'll update the proposed text accordingly...
Thanks,
Adrian
---------- Forwarded message ----------
From: Mr. Raul Oteyza <raul.oteyza@xxxxxxxxxx>
Date: Fri, Feb 23, 2018 at 4:52 PM
Subject: Re: Fwd: [fibre_channel] Thoughts on FC-FS-5 Open Comments Cisco-001 & Cisco-002
To:
fibre_channel@xxxxxxxxxxxxxxxxxxxx
Hi Adrian,
My input for Cisco-001:
I agree with much of what you have written –
1. That Alignment Markers can’t be considered Fill Words and can’t follow the transmission and reception requirements of Fill Words.
2. A more apt description is that Alignment Markers replace Fill Words at the FC-1 transmitter, and conversely Fill Words replace Alignment Markers at the FC-1
receiver. Note that the manner in which this replacement occurs is vendor specific and therefore beyond the scope of the FC-FS-5 standard
The proposal for the added text for 64GFC, 128GFC, and 256GFC is spot on and appears clear and easy to understand.
The only part of the sentence that I disagree with is the first sentence that says “The alignment marker has the form of a specially defined 66-bit block with a control
transmission word
synchronization header.”
a. Reason: The Alignment Marker has no relationship to a 66B formatted word nor does it have a synchronization header. Its format is simply a 257-bit pattern
containing 2-bits of Remote Degrade values (one which is inverse to another), and a fixed data pattern. This is true for 64GFC, 128GFC, and 256GFC.
b. Proposal:
i. Replace the sentence with
1. The alignment marker has the form of a specially defined 257-bit pattern consisting of two Remote Degrade bits and a fixed data pattern.
2. Or, The alignment marker has the form of a specially defined 257-bit pattern consisting of two Remote Degrade bits (one which is an inverse to another) and a fixed
data pattern.
Raul
InterNational Committee for Information Technology Standards (INCITS)
Secretariat, Information Technology Industry Council (ITI)
1101 K Street NW, Suite 610, Washington, DC 20005
Notice:
This message is intended exclusively for the individual or entity to which it is addressed. This mailing list may not be used for unlawful purposes. All postings should be
relevant, but ITI accepts no responsibility for any posting and may terminate access to any subscriber violating any policies of the Association. Please review the INCITS
Anti-Trust Guidelines at
http://www.incits.org/standards-information/legal-info.
Please be advised that replying to this message goes to the sender. If you wish to send a reply to all on the list, please respond with "Reply All". If you have received
this message in error, please notify the sender immediately by e-mail and delete all copies of the message.
|
|
|