:52 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released 20 02 11 :48 PM7340 TM tem be r, S/UNI- ay ,1 9S ep IMA-8 yV inv ef uo fo liv ett io nT hu rsd S/UNI-IMA-8 Do wn loa de db Revision B Device Errata Released Issue 1: April 2002 Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released :52 Legal Information :48 Copyright 11 © 2002 PMC-Sierra, Inc. tem be r, 20 02 The information is proprietary and confidential to PMC-Sierra, Inc., and for its customers’ internal use. In any event, you cannot reproduce any part of this document, in any form, without the express written consent of PMC-Sierra, Inc. PMC-2020873 (R1) ep Disclaimer nT hu rsd ay ,1 9S None of the information contained in this document constitutes an express or implied warranty by PMC-Sierra, Inc. as to the sufficiency, fitness or suitability for a particular purpose of any such information or the fitness, or suitability for a particular purpose, merchantability, performance, compatibility with other parts or systems, of any of the products of PMC-Sierra, Inc., or any portion thereof, referred to in this document. PMC-Sierra, Inc. expressly disclaims all representations and warranties of any kind regarding the contents or use of the information, including, but not limited to, express and implied warranties of accuracy, completeness, merchantability, fitness for a particular use, or non-infringement. uo fo liv ett io In no event will PMC-Sierra, Inc. be liable for any direct, indirect, special, incidental or consequential damages, including, but not limited to, lost profits, lost business or lost data resulting from any use of or reliance upon the information, whether or not PMC-Sierra, Inc. has been advised of the possibility of such damage. ef Trademarks db Patents yV inv S/UNI is a registered trademark of PMC-Sierra, Inc. Other product and company names mentioned herein may be the trademarks of their respective owners. Do wn loa de The technology discussed is protected by one or more of the following Patents: U.S. Patent No. 5,875,192 Relevant patent applications and other patents may also exist. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 2 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released :52 Contacting PMC-Sierra 11 :48 PMC-Sierra 8555 Baxter Place Burnaby, BC Canada V5A 4V7 20 02 Tel: (604) 415-6000 Fax: (604) 415-6200 9S ep tem be r, Document Information: [email protected] Corporate Information: [email protected] Technical Support: [email protected] Web Site: http://www.pmc-sierra.com Issue Date Details of Change 1 April, 2002 Document created. Do wn loa de db yV inv ef uo fo liv ett io nT hu rsd Issue No. ay ,1 Revision History Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 3 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released :52 Table of Contents :48 Legal Information ................................................................................................................ 2 Contacting PMC-Sierra ....................................................................................................... 3 11 Table of Contents ................................................................................................................ 4 02 Device Identification............................................................................................ 6 1.2 References ......................................................................................................... 7 r, 20 1.1 tem be Device Functional Deficiency List ............................................................................... 8 S/UNI-IMA-8 Revision B Status .......................................................................... 8 2.2 IMA Group Feature RX_ADD_DELAY_EN Does Not Work ............................. 10 2.2.2 Workaround.......................................................................................... 11 2.2.3 Performance with Workaround ............................................................ 11 2.2.4 Performance without Workaround ....................................................... 11 ,1 ay rsd Description ........................................................................................... 11 2.3.2 Workarounds........................................................................................ 12 2.3.3 Performance with workaround ............................................................. 12 2.3.4 Performance without workaround ........................................................ 12 fo liv ett io 2.3.1 Receive Link Recovery Blocks LIF State Change Interrupt.............................. 12 2.4.1 Description ........................................................................................... 12 2.4.2 Workaround.......................................................................................... 13 2.4.3 Performance with Workaround ............................................................ 13 db Performance without Workaround ....................................................... 13 CTC Mode: After Many Changes of TX TRL on an Active IMA Group, TX Link FIFOs may Underrun or Overrun................................................................................. 14 Do wn loa de 2.6 nT hu LID Value must be Deleted For Those Links Rejected Due to an IMA_ID Mismatch .......................................................................................................................... 11 2.4.4 2.5 9S Description ........................................................................................... 10 uo 2.4 2.2.1 ef 2.3 ep 2.1 inv 2 Introduction.................................................................................................................. 6 yV 1 2.5.1 Description ........................................................................................... 14 2.5.2 Workaround ......................................................................................... 14 2.5.3 Performance with Workaround ............................................................ 14 2.5.4 Performance without Workaround ....................................................... 14 ITC Mode: After Many Changes of TX TRL on an Active IMA Group, TX Link FIFOs may Underrun or Overrun................................................................................. 14 2.6.1 Description ........................................................................................... 14 2.6.2 Workaround #1 .................................................................................... 15 Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 4 PM7340 S/UNI-IMA-8 Revision B Device Errata Released Performance with Workaround #1 ....................................................... 16 2.6.4 Performance without Workaround #1 .................................................. 16 2.6.5 Workaround #2 .................................................................................... 16 2.6.6 Workaround #2A .................................................................................. 16 2.6.7 Performance with Workaround #2A..................................................... 17 2.6.8 Performance without Workaround #2A................................................ 17 2.6.9 Workaround #2B .................................................................................. 17 20 02 11 :48 :52 PM 2.6.3 tem be r, 2.6.10 Performance with Workaround #2B..................................................... 17 2.6.11 Performance without Workaround #2B................................................ 17 Description ........................................................................................... 18 2.7.2 Workaround.......................................................................................... 18 2.7.3 Performance with Workaround ............................................................ 18 2.7.4 Performance without Workaround ....................................................... 19 ,1 9S ep 2.7.1 rsd Documentation Deficiency List .................................................................................. 20 Do wn loa de db yV inv ef uo fo liv ett io nT hu 3 When the RX_TRL_LID is Changed Repeatedly, LODS Can Occur ............... 18 ay 2.7 Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 5 Introduction :52 1 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released Section 2 lists the known functional errata for Revision B of the PM7340 S/UNI-IMA-8 Device. · Section 3 lists documentation deficiencies, if any, in Issue 4 of the PM7340 S/UNI-IMA-8 Data Sheet (PMC-2001723) and any other document that is covered by this errata. 20 02 11 · r, Device Identification tem be 1.1 :48 In this document: 9S ep The information contained in this section relates to Revision B of the PM7340 S/UNI-IMA-8 Device only. The device revision code is marked at the end of the Wafer Batch Code on the face of the device. PM7340 S/UNI-IMA-8 Branding Format ,1 Figure 1 ett io nT hu rsd ay Pin A1 Reference Mark TM fo liv S/UNI ef uo IMA-8 PMC Logo Do wn loa de db yV inv PM7340-PI C________B Myyww Philippines Part Number Wafer Batch Code Assembly Date Code Country of Manufacture SCALE = 3:1 (Approx.) Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 6 :48 :52 References Issue 4 of the S/UNI Inverse Multiplexing for ATM, 8 Links (PM7340) Data Sheet (PMC-2001723). · Issue 3 of the PM7340 S/UNI-IMA-8 Revision A Device Errata (PMC-2010509). 02 11 · Do wn loa de db yV inv ef uo fo liv ett io nT hu rsd ay ,1 9S ep tem be r, 20 1.2 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 7 Device Functional Deficiency List :52 2 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released 11 :48 This section lists the known functional deficiencies for Revision B of the PM7340 S/UNI-IMA-8 as of the publication date of this document. For each deficiency, the known workaround and the operating constraints, with and without the workaround, are also described. 02 S/UNI-IMA-8 Revision B Status 20 2.1 tem be r, The PM7340 S/UNI-IMA-8 Revision B device has been fully qualified, characterized and functionally tested to production-grade status. Table 1 summarizes all S/UNI-IMA-8 Revision B errata into two groupings: 9S ep 1. The first portion of Table 1 summarizes the S/UNI-IMA-8 Revision A deficiencies and the status of these deficiencies in the S/UNI-IMA-8 Revision B device. ay ,1 2. The second grouping summarizes any new deficiencies in the S/UNI-IMA-8 Revision B device. hu rsd Table 1. Summary of S/UNI-IMA-8 Revision B Deficiencies. S/UNI-IMA-8 Rev A Fixed in S/UNIIMA-8 Rev B Comment ett io nT Deficiencies a a a uo RX UTOPIA Bus Driven Only When Data is Available ef fo liv UTOPIA Level 2 Mode – Additional Pins Need to be Tied Low yV inv Any-PHY / UTOPIA Level 2 Interface Clock Rate May Need to be Reduced From 50 MHz a RIPP Interrupt Status Register Read Pulse Constraint a Simultaneous Support of IMA Groups and TC Links Not Functional a IMA Groups with Minimal Delay in DCB (e.g. RX_DELAY_GUARD_BAND=0) May Experience LODS Due to a DCB Underrun a IMA Groups with Minimal Delay in DCB (e.g. a Do wn loa de db Line Loopback Configuration Restrictions Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 8 Deficiencies Comment :52 Fixed in S/UNIIMA-8 Rev B :48 S/UNI-IMA-8 Rev A PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released 11 RX_DELAY_GUARD_BAND Less Than 3) May Reject Link Additions ay a ett io nT hu rsd a a fo liv Avoid Issuing Simultaneous Adjust_Delay Commands See Section 2.6. ,1 Change of TRL in ITC Mode Results in Improper Stuffing When Adding Links to Existing IMA Groups, Under Certain Conditions the Transmit Link FIFOs May Underrun or Overrun r, a tem be Lack of Valid IMA Transmit Reference Link (TRL) Condition Does not Trigger Interrupt During Group Startup ep a 9S RX IMA Groups with Certain ICP Offsets Will Cause Dropped Cells and LODS For TC-Only Links the ATM Congestion Counter is Inaccurate 02 See Section 2.2 20 IMA Group Feature RX_ADD_DELAY_EN Does Not Work a uo Idle Cells Are Forwarded to ATM Layer on IMA groups See Section 2.3 yV inv ef LID Value Must be Deleted For Those Links Rejected Due to an IMA_ID Mismatch a Do wn loa de db In the Absence of ICP Cells, LCD and LIF Not Detected During Link Addition or Group Startup When IMA Receiver is in LIF, DCB Overrun and Underrun Interrupts May Erroneously Occur a Restart Group Command – Expected Interrupts are not Present a Mixing IMA1.0 and IMA1.1 Groups May Generate Incorrect Filler Cell OAM_LABELs a Filler Cells Received on Inactive Links Erroneously Counted as Dropped Cells a Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 9 ICP_VIOLATIONS Counter Erroneously Increments During Link Startup a Receive ICP_VIOLATIONS Counter is Inconsistent with INVALID_ICP_INT Interrupt When Second ICP Cell of a Stuff Event is Invalid a :52 :48 a 20 02 During Group Startup and Group Restart, Links Within the Specified Differential Delay Tolerance May Still Be Erroneously Rejected Due to Error in Calculation r, Deficiencies Comment 11 Fixed in S/UNIIMA-8 Rev B 9S ep tem be S/UNI-IMA-8 Rev A ay ,1 Receive Link Recovery Blocks LIF State Change Interrupt io nT hu rsd S/UNI-IMA-8 Rev B Deficiency CTC Mode: After Many Changes of TX TRL on an Active IMA Group, TX Link FIFOs may Underrun or Overrun See Section 2.4 Comment See Section 2.5 See Section 2.6 fo liv ett ITC Mode: After Many Changes of TX TRL on an Active IMA Group, TX Link FIFOs may Underrun or Overrun PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released See Section 2.7 inv ef uo When the RX_TRL_LID is Changed Repeatedly, LODS Can Occur IMA Group Feature RX_ADD_DELAY_EN Does Not Work 2.2.1 Description Do wn loa de db yV 2.2 In the S/UNI-IMA-8 device, there are two methods to add delay to a group in order to accept new links having more network delay than the existing links. The first method uses the RX_ADD_DELAY_EN feature where delay is automatically added to other links in the group as necessary. The second method is to manually read the delay profile using the read_delay command and then use the adjust_delay command to add the necessary amount of group delay. This second method is described in detail in the S/UNI-IMA-8 datasheet. In the S/UNI-IMA-8 device, the RX_ADD_DELAY_EN feature, described in method one, does not work. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 10 PM7340 S/UNI-IMA-8 Revision B Device Errata Released Workaround PM 2.2.2 :48 :52 Do not set RX_ADD_DELAY_EN. The S/UNI-IMA Device Driver will also prevent the application layer from enabling this feature. tem be r, 20 02 11 If, after a link addition or LASR procedure the new link fails to be added due to a differential delay problem (DIFF_DELAY_INT), use the read_delay and adjust_delay commands to add the necessary amount of delay to the group, and then recover the link. This procedure is described in detail in section 13.5.14 Issue 4 of the S/UNI-IMA-8 data sheet. Also note that this approach is different than with the Revision A S/UNI-IMA-8 device where delete and add back commands were suggested in its errata. Performance with Workaround ,1 2.2.3 9S ep Note that the RX_DELAY_GUARDBAND should be configured to a value of 3 or greater in order to prevent rejection of link additions solely due to differential delay measurement jitter associated with ICP cells and stuff events. Performance without Workaround nT 2.2.4 hu rsd ay Functions normally. Instead of relying on the RX_ADD_DELAY_EN for automatically adjusting delay on link addition, the same functionality is achieved using the read_delay and adjust_delay commands as described in the workaround. fo liv ett io If RX_ADD_DELAY_EN is set, and upon receiving requests to add links at the near-end (having more network delay than the existing links), the existing links in the group with the least amount of network delay may erroneously detect LODS, rather than the expected behavior of rejecting the new links with a DIFF_DELAY_INT interrupt. LID Value must be Deleted For Those Links Rejected Due to an IMA_ID Mismatch 2.3.1 Description yV inv ef uo 2.3 db For each IMA group, matching of the IMA_ID during LID Validation (during IMA group startup) is performed in one of two modes. Do wn loa de 1. Mode 1: Plane Management (PM) configures the IMA group with the IMA_ID to be used for matching. 2. Mode 2: Plane Management (PM) configures the IMA group such that the IMA_ID is automatically captured from the first incoming ICP cell and then used for matching. In Revision A of the S/UNI-IMA-8 device, there are problems with Mode 1 and Mode 2 as described in the S/UNI-IMA-8 Revision A errata. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 11 PM7340 S/UNI-IMA-8 Revision B Device Errata Released :48 :52 PM Mode 1 has been fixed in the S/UNI-IMA-8 Revision B device. Therefore, in Mode 1, if a link is rejected due to an IMA_ID mismatch, the LID value is no longer erroneously reserved. Future link additions with a matching IMA_ID will be validated successfully. 2.3.2 tem be r, 20 02 11 There are still issues with Mode 2 in the S/UNI-IMA-8 Revision B device. In Mode 2, if the far end FE_GSM (connected to S/UNI-IMA-8) happens to be in STARTUP state, it is possible that the near end Rx LID is validated as correct even when the IMA_ID is incorrect (i.e. current IMA_ID is different from first IMA_ID octet captured). In this case the LID value is erroneously reserved in the S/UNI-IMA-8, which causes future link additions with a properly matching IMA_ID to be erroneously rejected. This problem is not fixed in Revision B of the S/UNI-IMA8. Workarounds ,1 Performance with workaround ay 2.3.3 9S ep Ensure that all links that have been rejected due to an IMA_ID mismatch in Mode 2 are deleted using the delete link command. Performance without workaround hu 2.3.4 rsd The device operates normally. io nT If an IMA group has had a link rejected due to an IMA_ID mismatch in Mode 2, then future add link requests to that group may be erroneously rejected. Receive Link Recovery Blocks LIF State Change Interrupt 2.4.1 Description uo fo liv ett 2.4 db yV inv ef The S/UNI-IMA-8 device provides an interrupt (LIF_INT) that is set when the internal LIF (Loss of IMA Frame) status of a link changes state (from no LIF to LIF and from LIF to no LIF). This interrupt allows the application layer to integrate LIF defects over time before declaring a LIF failure. During group startup and link addition, the S/UNI-IMA-8 will appropriately prevent LIF change interrupts from occurring; otherwise, all links would experience these interrupts during link addition, creating excessive and unnecessary events to the application layer. Do wn loa de This interrupt is also blocked during a Receive Link Recovery operation, which is either initiated via the same command as link addition (Start LASR), or by issuing S/UNI-IMA Device Driver API function call “ima84ImaRecoverLinks”. If a link was in LIF prior to the Link Recovery operation, and the link exited LIF during the recovery, upon completion of the Link Recovery operation, the LIF_INT interrupt would erroneously not be set and hence not indicate the change of state. The application layer would then not properly terminate the LIF failure integration due to the missed interrupt. The following workaround describes a mechanism to allow the application layer to detect a change of LIF state if one were to occur during a Receive Link Recovery operation. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 12 PM7340 S/UNI-IMA-8 Revision B Device Errata Released Workaround PM 2.4.2 :48 :52 Depending on the operation mode of the IMA group, the application layer can implement the following workaround(s). 11 Symmetric Operation ep tem be r, 20 02 In Symmetric Operation, which involves Link Recovery of RX and TX links simultaneously, the Start LASR associated with the Receive Link Recovery operation is complete when the TX links become active or time-out. When the TX links become active or time-out, the corresponding interrupt is triggered. Upon receiving the interrupt, a “read event” operation should be executed as described in the S/UNI-IMA-8 datasheet. As part of the “read event” operation, the LIF status is also retrieved. Note that the S/UNI-IMA Device Driver Production Release (Rel-1.0) returns the LIF status as part of the structure, which is used by the application layer for LIF integration. 9S Asymmetric Operation ay ,1 In Asymmetric Operation, which involves Link Recovery of RX Only links, the application layer is required to perform the following tasks in order to detect a change in LIF status. Receive the RX_ACTIVE or RX_TIMEOUT interrupt events for all the RX links involved. · Upon receiving the interrupt, wait for 100 ms. This delay corresponds to approximately one IMA frame; ((53x8)bits/(1.544)MHz)x256 >= 70.3 ms. M=256 & T1 Mode are used to calculate the worst-case delay required. This wait time can be used for all other combinations of M and line mode. · Get the LIF status for each link from "Ripp RX Link Context Memory" using the S/UNI-IMA Device Driver API function "ima84StatsGetImaLinkStatus". Performance with Workaround ef 2.4.3 uo fo liv ett io nT hu rsd · Performance without Workaround db 2.4.4 yV inv The device operates normally – changes in LIF state during Receive Link Recovery operations can be detected in both Symmetric and Asymmetric Operation. Do wn loa de During Receive Link Recovery operations, LIF state changes cannot be detected which inhibits the application layer’s ability to accurately integrate LIF failures. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 13 PM7340 S/UNI-IMA-8 Revision B Device Errata Released CTC Mode: After Many Changes of TX TRL on an Active IMA Group, TX Link FIFOs may Underrun or Overrun 2.5.1 Description :48 :52 PM 2.5 2.5.2 tem be r, 20 02 11 Once a Transmit Reference Link (TRL) is assigned to an operational IMA group, this assignment typically remains fixed. In some cases the TRL assignment needs to change - for example under conditions of a T1/E1 link failure where the link happens to be the TRL. Lab testing has shown that after approximately 1500 changes of TRL, there is a high probability that one or more of the TX Link FIFOs in the IMA group associated with the TRL may underrun or overrun. Workaround ep A viable option is to do no workaround at all. See Section 2.5.4. Performance with Workaround rsd 2.5.3 ay ,1 9S The other option is, instead of waiting for a Tx Link FIFO to actually underrun or overrun, after a certain number of TRL changes have occurred (i.e. a number below 1500 such as 1000), proactively re-provision the IMA group (deleting and adding it back). io Performance without Workaround ett 2.5.4 nT hu Each time an IMA group is re-provisioned, the probability of links in the group underrunning is reset to approximately 1500 changes of TRL. ef uo fo liv Without the workaround implemented, for TX Link FIFO underruns or overruns on CTC Mode IMA groups, the effect is the insertion of an Idle cell or a dropped cell, respectively. After the insertion of an Idle cell or a dropped cell the IMA group returns to functioning normally without any external intervention. For both cases (Idle cell or dropped cell), the far end will likely detect an OIF anomaly resulting in temporary data corruption (typically two to three IMA frames). ITC Mode: After Many Changes of TX TRL on an Active IMA Group, TX Link FIFOs may Underrun or Overrun 2.6.1 Description Do wn loa de db yV inv 2.6 Once a Transmit Reference Link (TRL) is assigned to an operational IMA group, this assignment typically remains fixed. In some cases the TRL assignment needs to change - for example under conditions of a T1/E1 link failure where the link happens to be the TRL. Lab testing has shown that after approximately 100 changes of TRL, there is a possibility that any of the TX Link FIFOs in the ITC Mode IMA group associated with the TRL may underrun or overrun. The impact of an underrun or overrun depends on whether the affected link is the TRL or not, as summarized in Table 2. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 14 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released Impact of Underrun Impact of Overrun ITC Mode: TRL Link Single Inserted IDLE Cell Single Dropped Cell ITC Mode: Non-TRL Links Errant IMA Stuff Behavior, Link Unusable Single Dropped Cell 20 02 11 :48 :52 Group / Link Type tem be r, Table 2. ITC Mode: Impact of a TX Link FIFO Underrun or Overrun. Workaround #1 ,1 2.6.2 9S ep There are two workarounds described below. Workaround #1 reduces the likelihood of a TX Link FIFO underrun or overrun for ITC Mode groups. Although the likelihood of an underrun or overrun is reduced, it is not eliminated. Workaround #2 addresses this situation. rsd ay This workaround reduces the likelihood of a TX Link FIFO underrun or overrun for ITC Mode groups. The workaround is as follows: When the transmitter (near-end) must change the TRL, it should poll the Transmit Stuff Event counter in the TIMA Physical Link Context Record for the TRL candidate at 100ms intervals. When a change in value is detected indicating that a link stuff event has just occurred, issue the “update_tx_trl” command. After issuing this command, polling the link Transmit Stuff Event counter is no longer required. Note that the production release (Rel1.0) of the S/UNI-IMA Device Driver provides API function calls to read the TX stuff event counter and to change the TRL. The API function calls are “ima84StatsGetImaLinkCnts” and “ima84ImaChngGrpTxTrl”. · After following the procedure above, delete the old TRL. When or if the old TRL recovers, it can be added back to the IMA group. Note: in order to reduce the number of TRL changes, it is recommended that the TRL assignment remain as is (i.e. avoid re-assigning the old TRL as the new TRL). Do wn loa de db yV inv ef uo fo liv ett io nT hu · Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 15 PM7340 S/UNI-IMA-8 Revision B Device Errata Released Performance with Workaround #1 PM 2.6.3 S/UNI-IMA TYPICAL NUMBER OF TRL CHANGES BEFORE AN UNDERRUN OR OVERRUN OCCURS 100 300 r, 20 With Workaround #1 tem be ITC Mode (TRL and non-TRL links) Without Workaround 02 11 STUFFING MODE :48 :52 As summarized in Table 3, implementing workaround #1 increases the typical number of TRL changes to 300 before an underrun or overrun is likely to occur. Performance without Workaround #1 9S 2.6.4 ep Table 3. ITC Mode: Typical Number of TRL Changes Before an Underrun or Overrun. rsd Workaround #2 hu 2.6.5 ay ,1 As summarized in Table 3, without implementing this workaround the typical number of TRL changes possible before an underrun or overrun occurs remains at 100. nT There are two options: Detect a Tx Link FIFO underrun at the near end then take action. This requires software polling as described in Workaround #2A. Overruns autorecover to normal operation as described in “Performance Without Workaround #2A” below. B. After a certain number of TRL changes have occurred (i.e. less than the typical numbers in Table 3), proactively re-provision the IMA group (deleting and adding it back). fo uo ef Workaround #2A inv 2.6.6 liv ett io A. Do wn loa de db yV If a TX Link FIFO underrun occurs on an ITC mode non-TRL link, errant IMA stuff behavior is experienced on that link making it unusable. Deleting this faulty link and then adding it back to the group has the effect of re-centering that Tx Link FIFO which will allow the IMA group to resume normal operation. To detect TX Link FIFO underruns, read the Transmit Link FIFO Depth in the TIMA Transmit Physical Link Context Record for each link in the group. The value read represents the current depth, in cells, of the FIFO associated with that link (the S/UNI-IMA Device Driver provides an API function call “ima84StatsGetImaGrpTxLinkFifoDepths” to return the current FIFO depths for a group). The production release (Rel-1.0) of the S/UNI-IMA Device Driver also provides example code of how to use this API function to detect the underrun condition. These FIFO depth values must be polled continuously at 100 ms intervals. If four consecutive values of 3cells-or-less are retrieved, then this link has underrun. In order to resume normal IMA group operation, the link must be deleted and re-added which has the effect of re-centering that Tx Link FIFO. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 16 PM7340 S/UNI-IMA-8 Revision B Device Errata Released Performance with Workaround #2A PM 2.6.7 Performance without Workaround #2A 20 2.6.8 02 11 :48 :52 With Workaround #2A implemented, ITC Mode non-TRL TX Link FIFO underruns at the near end can be detected and subsequently the misbehaving link can be deleted and re-added which has the effect of re-centering the Tx Link FIFO. Once this is done, normal IMA group operation resumes. Note that in an underrun condition, until the link has been deleted, there will be errant IMA stuff behavior exhibited on that link making it unusable. ep tem be r, Without implementing this workaround, the far end application layer IMA recovery mechanism is relied upon to detect OIF anomalies (as it should). Upon detecting OIF anomalies, the far end will then indicate the unreliable link status to the near end. The near end can then take necessary action (e.g. delete and re-add the link to resume normal operation). Workaround #2B hu 2.6.9 rsd ay ,1 9S If a Tx Link FIFO overrun occurs (as indicated by the S/UNI-IMA’s Transmit Link FIFO Overflow interrupt LINK_FIFO_OVERFLOW_STAT), there is no workaround required since the IMA group automatically returns to normal operation without any external intervention. A single cell will be dropped and the far end will likely detect an OIF anomaly resulting in temporary data corruption (typically two to three IMA frames). Performance with Workaround #2B liv 2.6.10 ett io nT Instead of waiting for a Tx Link FIFO to underrun or overrun, after the number of TRL changes has reached a threshold, (below the typical numbers in Table 3), proactively re-provision the IMA group (deleting and adding it back). Performance without Workaround #2B yV 2.6.11 inv ef uo fo Each time an IMA group is re-provisioned, the probability of links in the group underrunning is reset to approximately those values listed in Table 3. Note that this workaround reduces the probability of an underrun but does not completely eliminate it - the lower the threshold used, the lower the underrun probability. In the unlikely event that an underrun occurs, the impact is exactly as described in “Performance Without Workaround #2A.” Do wn loa de db Without this workaround implemented, the situation is exactly as described in “Performance without Workaround #2A.” Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 17 PM7340 S/UNI-IMA-8 Revision B Device Errata Released When the RX_TRL_LID is Changed Repeatedly, LODS Can Occur 2.7.1 Description :48 :52 PM 2.7 tem be r, 20 02 11 Each time the far end transmitter changes the group Transmit Reference Link (TRL) in either CTC or ITC mode, it is detected at the near end receiver as a change in RX_TRL_LID as expected. However, a small group rate inaccuracy may occur at the S/UNI-IMA-8 receiver due to the change of receiver TRL operation. This rate inaccuracy can slightly increase or decrease the current depths of all Delay Compensation Buffers (DCB) for the group. Also, if the physical link, which is currently designated as the RX_TRL_LID becomes unreliable (no longer frequency locked to the far end), these same inaccuracies can occur as they would at any IMA receiver. Workaround ,1 2.7.2 9S ep After many of these inaccuracies accumulate, DCBs may underrun or overrun resulting in an LODS interrupt. LODS underrun/overrun can cause loss of user cells until this condition is exited either by deleting the link or adjusting the delay to re-center the DCBs. nT hu rsd ay Option 1: Allow LODS to occur: Once LODS occurs, perform a “read_delay” operation for the group, and use the “adjust_delay” command which automatically adds or removes delay as needed to allow the link to recover. Once the delay profile has been adjusted, the link will automatically recover from LODS without further intervention. Performance with Workaround yV 2.7.3 inv ef uo fo liv ett io Option 2: Prevent LODS from occurring: If it is necessary to prevent LODS and the ensuing data loss after many changes of RX_TRL_LID, then it is required to periodically poll, once per second, the DCB FIFO depths (representing link transit delays) using the “read_delay”(API function “ima84StatsGetImaGrpDelayInfo”) command. If the depth of any link in the group is less than or equal to min_lower_guardband (Reg 0x208) or greater than or equal to (RX_DELAY_TOL – min_upper_guardband Reg 0x208) then the “adjust_delay” (API function “ima84ImaAdjustGrpDelay”) command can be used to “re-center” the DCB depths by adding or removing group delay as needed. This procedure is also described in detail in section 13.5.14 Issue 4 of the S/UNI-IMA-8 data sheet. db The performance with each workaround is as follows: Do wn loa de Option 1: If LODS occurs after many changes of RX_TRL_LID, user data may be lost until the workaround has been implemented (i.e. the application layer adjusts delay on the group to recover from LODS). Once the workaround is implemented (i.e. group recovers from LODS) normal operation resumes. Option 2: Even after many changes of RX_TRL_LID, the IMA group continues normal operation (i.e. LODS is prevented by implementing this workaround option). Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 18 PM7340 S/UNI-IMA-8 Revision B Device Errata Released Performance without Workaround PM 2.7.4 Do wn loa de db yV inv ef uo fo liv ett io nT hu rsd ay ,1 9S ep tem be r, 20 02 11 :48 :52 LODS may erroneously occur on any link of an IMA group that has had many changes of RX_TRL_LID. As a part of normal IMA operation, when a link is detected to be in LODS, it will be deleted after integrating the LODS failure. Note that the application layer may add the link back into the group using normal add link procedures. Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 19 Documentation Deficiency List :52 3 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released Do wn loa de db yV inv ef uo fo liv ett io nT hu rsd ay ,1 9S ep tem be r, 20 02 11 :48 As of the publication date of this document there are no known documentation deficiencies for Issue 4 of the PM7340 S/UNI-IMA-8 Data Sheet (PMC-2001723). Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 20 PM PM7340 S/UNI-IMA-8 Revision B Device Errata Released Do wn loa de db yV inv ef uo fo liv ett io nT hu rsd ay ,1 9S ep tem be r, 20 02 11 :48 :52 Notes Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use Document ID: PMC-2020873, Issue 1 21