add management port page, grammar

README.md

@@ -20,4 +20,4 @@ A. When Broadcom introduced the BCM56820 switch ASIC, they released a reference
  
  Because of this, firmware written for one platform will have no issue running on the other - it's the same hardware.
  
-Open to PR's! If you know of any other switch pairs that are the same underlying hardware and would like to cross-flash, let us know! We're mainly seeking firmware candidates for the LB4M - so far we can't find any other switches with the same hardware configuration. We need to find a switch that uses a MPC8541 management CPU + BCM56514 ASIC)
+Open to PR's! If you know of any other switch pairs that are the same underlying hardware and would like to cross-flash, let us know! We're mainly seeking firmware candidates for the LB4M - so far we can't find any other switches with the same hardware configuration. We need to find a switch that uses a MPC8541 management CPU + BCM56514 ASIC

documentation/index.md

@@ -2,6 +2,7 @@
 
 
 
+
 # Flashing the LB6M to a Brocade TurboIron 24X
 
 ## Disclaimer & Caveats
@@ -11,8 +12,8 @@ Two things will also change due to hardware differences:
 * The SFP+ port activity/status LEDs stop doing anything. The copper ports and chassis LEDs continue to work as normal. The Quanta uses a CPLD to multiplex the LED signals, while the Brocade uses native CPU I/O. There's no way around this difference.
 
 * The Brocade only has one Out Of Band management port, so the code is only aware of the #1 OOB port. The #2 management port will no longer do anything. 
-* The Brocade code seems to also use a different clock or MII registers to configure the management port PHY (since it only expects one), which results in the management port not linking up at full speed for some.  
-For some it links up at 100mbps and works fine, for others, only 10mbps. In rare cases it will refuse to link up at all. You still have management access on all other ports as usual if you configure in-band management access. If your management port refuses to link up, and you need it to TFTP firmware (or other emergency), [it can be forced to link](http://brokeaid.com/mgmt/). To reiterate, this only effects the out of band management port.
+* Due to a software configuration difference, the out of band management port may link at slower speeds than it did under Fastpath (eg at 100mbit). In one case, we saw it stop linking altogether. For more information, please [click here](http://brokeaid.com/mgmt/). To reiterate, this only affects the management port for OOB telnet/ssh. It does not affect switched/routed traffic, and you can configure other switch ports to be used for isolated management via ACLs if you really need something other than in-band management. That said, 99% of peoples management port continues to work fine.
+* If you don't like the Brocade OS, or have other issues, you can always flash back to 100% stock Fastpath using the Revert guide to the left, so none of this is permanent.  
 
 If you're looking to purchase an LB6M, we recommend [UnixPlus](https://www.unixplus.com/products/quanta-lb6m-24-port-10gbe-sfp-4x-1gbe-l2-l3-switch) - their stock is all brand new and of known origin.
 ## Prerequisites

documentation/mgmt.md

@@ -1,13 +1,62 @@
 # Management Port Behavior
 
 ## What Happens
-coming soon(tm)
+The out of band management port is gigabit PHY - however for some users, after flashing the Brocade firmware it will instead start linking up at 100mbit. In even rarer cases, some users have seen it link at 10mbit. In all of these cases it continues to operate fine as it's only passing management traffic. In some extremely rare cases (observed on 2 chassis), it will refuse to link at all. You can obviously still use in-band management IP's to access the switch, or even designate another switch port in its own isolated VLAN as a management-only port via ACLs. 
 
 ## Why It Happens
+It's hard to say with 100% certainty without using an oscilloscope on the various clocks and data lines for the management PHY, and comparing scoped values to values seen while running Fastpath (when the port works properly). What is most likely happening is the Brocade firmware tells the management CPU to use slightly different clock speed values to drive the management PHY, as the traces from the management CPU to the management PHY are shorter in the Brocade.  
 
+It's also possible the pre-emphasis configuration registers on the SGMII data transmitter were tuned for the trace layout and length of the Brocade chassis, and are therefore slightly out of spec when that configuration is loaded via the Brocade software in the Quanta chassis.  
+
+Sadly neither of these are fixable without disassembling the bootloader (the Brocade bootloader is what's responsible for initializing the PHY, and also remains running at all times and acts as a scheduler for the main OS). Once you find these configuration values in the disassembled bootloader they can be changed to appropriate values for the Quanta chassis and re-compiled. I received an estimate from a firmware engineer of approximately $2000 USD for this amount of reverse engineering, which I'm not willing to spend for a management port I never use on a $300 switch.
 
 ## Emergency Workaround
+As it's entirely a software issue (the port goes right back to working perfectly when reverting to Fastpath software), it can be forced to link in emergency situations with a temporary software trick.
 
+If you find yourself in a situation where you must have the management port link up and it won't, follow the below. To reiterate, we've only seen it completely cease linking twice in a hundred flashed chassis.  
+
+Assuming you're in the Brocade bootloader, enter the ```t2``` test menu, by simply typing t2 into the bootloader prompt:
+
+```
+t2
+```
+
+You'll be dropped into a hardware test menu that looks like this:
+```
+     Host Peripherals Test Menu
+
+[1  ]  PCI Bus Test
+
+[2  ]  UART Loopback Test
+
+[3  ]  SDRAM Test
+
+[4  ]  FLASH and File CRC32 Checksum Test
+
+[5  ]  System Timer Test
+
+[6  ]  I2C Test
+
+[7  ]  FAN Test
+
+[8  ]  Ethernet Test
+
+[9  ]  RPSU Test
+
+[A  ]  Test All Items
+
+[ESC]  Exit
+
+
+ Enter your choice:
+```
+Enter in ```8``` for the ethernet test. On the following submenu, enter in ```6``` for "test all items" - it will run through a test of the ethernet PHY, and some of them will fail. Thankfully this test seems to have been "borrowed" from Fastpath, and it re-initializes the PHY with the MII registers/clock values/etc that Fastpath uses on the LB6M - so the port will begin working perfectly as it does in Fastpath.  
+
+Once it finishes, your management port should be succesfully linked up. Hit ```ESC``` a few times to get out of the test menus and back to the bootloader prompt. You can now succesfully TFTP firmware images in, or whatever else you needed the management port for.  
+
+**Note 1:** This only works until the PHY is reinitialized with the Brocade values, eg next time it reboots, or you pull power.  
+
+**Note 2:** Unless you know what you're doing, do not run any of the other tests in the t2 menu. For instance, the I2C test wipes the onboard EEPROM, which is where your MAC address and serial are stored. If you run that test, you'll need to redo the Customizing MAC guide.
 
 
 ### Thanks:

documentation/story.md

@@ -15,7 +15,7 @@ If you know of any other switch pairs that are the same underlying hardware and
 We're mainly seeking firmware candidates for the LB4M - so far we can't find any other switches with the same hardware configuration. We need to find a switch that uses a MPC8541 management CPU + BCM56514 ASIC.
 
 ## Why is Fastpath so quirky?
-Fastpath is a [Broadcom software product](https://www.broadcom.com/products/ethernet-connectivity/software/fastpath), and the Fastpath image that comes with our switches is closer to a demo version than a full build.  
+Fastpath is a [Broadcom software product](https://www.broadcom.com/products/ethernet-connectivity/software/fastpath), and the Fastpath image that comes with our switches is closer to a demo version than a full build. That's why simple things like 1gbE SFP's don't work, even though the ASIC supports them.  
 
 When Quanta built a BCM56820 reference box, they included a barebones Fastpath build. The intent was that large customers (Microsoft & Amazon) would replace it with their own OS, or their own Fastpath builds with the exact featureset needed for their environment -  which they did. This means there's custom LB6M firmware packages floating around internally, but the chances we'll ever see them is slim to none.