A method and system for updating page configuration

By switching to the active page transaction when the candidate transaction is stable, the page security boundary is valid, and the commit permission is valid, and generating a consistent physical page on the sending side, the problems of page tearing and inconsistent configuration versions in page configuration updates are solved, and the correctness of link partner capability matching and the accuracy of acknowledgment count are achieved.

CN123123152APending Publication Date: 2026-11-17CHENGDU YONGZHI TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611573000.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-12
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, page configuration updates may lead to page tearing and inconsistent configuration versions, resulting in incorrect matching of link partner capabilities, improper accumulation of confirmation counts, and loss of completion notifications.

Method used

By switching to the active page transaction when the candidate transaction is stable, the page security boundary is valid, and the commit permission is valid, and generating a consistent physical page on the sending side, the consistency and integrity of the page configuration are ensured by combining configuration version identifiers and protocol context management.

Benefits of technology

It avoids page tearing and mismatch between configuration updates and protocol states, ensures the correct matching of link partner capabilities, avoids improper accumulation of acknowledgment counts and loss of completion notifications, and achieves consistent capture of multi-bit transactions in the target domain and safe release of storage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123152A_ABST
    Figure CN123123152A_ABST
Patent Text Reader

Abstract

This application provides a method and system for updating page configuration. Page configuration data is written to a shadow configuration storage unit to form candidate page transactions and remain stable. The page security boundary is detected in the page transmission clock domain, and the submission permission is reviewed based on the current protocol state, the update category of the candidate page transaction, and the protocol context information. When the submission request is valid, the candidate page transaction meets the capture conditions, the page security boundary is valid, and the submission permission is valid, the candidate page transaction is switched to the active page transaction and a configuration version identifier is assigned. During the transmission of the complete physical page, a physical page is generated based on the same selected active page transaction, ensuring that the configuration data used by the page is consistent with the configuration version identifier. This application combines the page security boundary and the protocol submission permission as common conditions for atomic submission, eliminating the need to transmit internal version information on the line, thus avoiding page tearing, improper accumulation of acknowledgment counts, and loss of completion events.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of communication chip and digital integrated circuit design technology, specifically to a method and system for updating page configuration. Background Technology

[0002] Ethernet auto-negotiation is used to exchange capability information between the local end and the link partner during the link establishment phase, and to select a mutually supported operating mode accordingly. Taking the auto-negotiation defined in IEEE 802.3 Clause 73 as an example, auto-negotiation uses DME pages to carry 48-bit link codewords. The Base Page contains fields such as Selector, Echoed Nonce, Acknowledge, Next Page, Transmitted Nonce, Technology Ability, and FEC; the Next Page contains control fields such as Toggle, ACK2, and Message Page. Another example is the twisted-pair auto-negotiation defined in IEEE 802.3 Clause 28, where capability information is carried by FLP bursts. An FLP burst contains alternating clock and data pulse positions, and there is a protocol-defined silence period between adjacent FLP bursts. Both of these auto-negotiation mechanisms repeatedly transmit protocol content in units of defined physical pages or bursts, and complete capability matching and confirmation according to the field comparison rules, continuous reception conditions, and state transition conditions specified in the protocol.

[0003] In existing methods, the management clock domain configures page payloads and transmission control parameters through registers. The autonegotiation arbitration state machine updates fields such as ACK, ACK2, Echoed Nonce, and Toggle based on the received results and protocol status. The page transmission clock domain transmits page data position by position according to the physical page structure, and the transmission of a complete physical page typically lasts for multiple transmission clock cycles. Since the management configuration logic, arbitration state machine, and page transmission state machine may reside in different clock domains, or even if they are in the same clock domain, the page configuration may be updated before the page transmission is complete. If the transmission state machine reads the page bus bit by bit, which can change in real time, some positions on the same physical page may come from the old configuration, while others may come from the new configuration, resulting in page tearing. Updating the configuration only at the physical page boundary is insufficient to guarantee the correctness of the autonegotiation process. Link partners may have already entered the subsequent confirmation phase based on the old capability page, while the local internal configuration version identifier is not transmitted online. Link partners will not automatically revoke existing matching results due to changes in their local internal version. Furthermore, there is no one-to-one correspondence between configuration version, logical page identity, and acknowledgment counting phase. If the acknowledgment page count is only associated with the current active configuration version without specifying the counting start point, logical page identity, and session conditions, page errors that do not belong to the same acknowledgment process may be accumulated, or the count may be repeatedly reset to zero due to unrelated configuration updates. For multi-bit configuration buses, if only independent synchronization of each data bit is performed without transaction-level hold and capture mechanisms, the target clock domain may obtain a combined value composed of data bits from different source times. For implementations with pipelined or asynchronous event channels, old page completion events may only reach the arbitration state machine after the new configuration takes effect. If an event is discarded simply because its version differs from the latest active version, valid completion notifications may be lost. Summary of the Invention

[0004] This application provides a method and system for updating page configuration. Page data and related control parameters are transmitted from the configuration side to the sending side in the form of a complete transaction. The candidate transaction only takes effect as a whole when the candidate transaction is stable, the page security boundary is valid, the commit permission is valid, and the capture conditions are met.

[0005] In a first aspect, this application provides a method for updating page configuration, applied to a communication transmission circuit that includes a configuration clock domain and a page transmission clock domain.

[0006] The method includes: receiving page configuration data, writing the page configuration data into a shadow configuration storage unit to form a candidate page transaction, and maintaining the candidate page transaction stable until a submission result indicating the end of the submission is received; detecting the page security boundary in the page sending clock domain, wherein the page security boundary is the boundary time or valid window where the specified sending position of the current complete physical page and the tail timing specified by the protocol have been completed, and the next physical page has not yet referenced its sending configuration, and the boundary time is allowed to coincide with the starting edge of the next physical page, or the page security boundary is a security state where the page sender does not have any in-process pages that need protection and is allowed to accept configuration; reviewing the submission permission according to the current protocol state and the update category and protocol context information of the candidate page transaction; when the submission request is valid, the candidate page transaction meets the capture conditions, the page security boundary is valid, and the submission permission is valid, the candidate page transaction is switched to the active page transaction as a whole, and a configuration version identifier is assigned to the active page transaction; during the sending of a complete physical page, a physical page is generated according to the same selected active page transaction, so that the configuration data used by the physical page is consistent with the configuration version identifier.

[0007] In one possible implementation, the candidate page transaction includes a page payload and at least one of the following protocol-related fields: page type, confirmation field, next page field, flip field, random number field, and encoded auxiliary information generator configuration; wherein, the values ​​generated page by page in the encoded auxiliary information according to the protocol rules are managed separately from the generator configuration, and normal changes in the values ​​generated page by page are not submitted as new configurations, nor do they change the configuration version identifier.

[0008] In one possible implementation, the method further includes: recording the transaction identifier, update category, and applicable protocol context information of the candidate page transaction; updating the page completion statistics based on the sending completion event of a valid complete page, the configuration version identifier of the page corresponding to the sending completion event, and its protocol context; wherein, the page completion statistics include the confirmation page count, after entering the corresponding protocol confirmation stage, establishing a confirmation context at the count initialization point specified in the confirmation stage, and binding the confirmation context with the session identifier, logical page identifier, page qualification rules, and the effective configuration version actually used in the confirmation stage, and only adding the confirmation page count for pages that are validly and completely sent, meet the page qualification rules, whose session and logical page match the confirmation context, and whose configuration version conforms to the configuration version association relationship.

[0009] In one possible implementation, the shadow configuration storage unit and the storage unit for storing active page transactions consist of at least two storage units, and the overall switching is completed by changing the active storage unit selection signal; the candidate storage unit that becomes the active storage unit is prohibited from being written to by the configuration side while there are still valid read dependencies, and the storage unit that exits the active state is only allowed to be rewritten to by the configuration side after the read dependencies end and ownership is released.

[0010] In one possible implementation, for a candidate page transaction that has already been received, if it meets the data capture conditions but the page security boundary or commit permission is not yet valid, the currently active page transaction remains unchanged, and the candidate page transaction is kept in a pending state; for subsequent requests, they are only accepted if there is an independent available transaction slot, otherwise a busy state is output or back pressure is applied, and candidate page transactions that have not yet ended are not covered.

[0011] In one possible implementation, when the completion of the old page and the commit of the new transaction occur at the same page security boundary, the old page completion event retains the configuration version identifier and protocol context of the old page, the data and configuration version identifier of the new physical page are determined from the same new transaction, and the overall switch is no later than the moment when the next physical page first references the configuration.

[0012] In one possible implementation, for page completion events that have already been generated, the transaction identifier or acknowledgment context association of the page completion event is maintained until the page completion event is reliably consumed or becomes invalid because its context is explicitly canceled, not because the page completion event is discarded because the current activity configuration version has changed.

[0013] In one possible implementation, the submission permission is reviewed based on the current protocol state, the update category of the candidate page transaction, and the protocol context information, including at least one of the following: reviewing whether the session to which the candidate page transaction belongs is still valid; reviewing whether the target logical page is compatible with the current exchange object; reviewing whether the updates to the acknowledgment field, acknowledgment field 2, or echo random number field meet the corresponding protocol conditions; reviewing whether the retention or flipping of the flip field is consistent with the logical page progression relationship; reviewing whether the next page indication and the status provided by subsequent pages meet the current protocol requirements; reviewing whether the base version or receiving context on which the candidate page transaction depends has expired; determining the order of submission based on the time of the sending side's receiving context expiration or cancellation event, and when cancellation and normal submission are established simultaneously at the same arbitration time, cancellation takes precedence.

[0014] In one possible implementation, the commit request and commit result are transmitted between the configuration clock domain and the page sending clock domain via a request confirmation handshake; the multi-bit candidate page transaction meets the establishment and maintenance requirements before being captured in the target domain by means of data path delay constraints, capture wait mechanisms, or a verified cross-clock structure; the returned transaction identifier, commit status, and configuration version identifier are transmitted in consistency with the confirmation event, and the associated returned data remains stable until reliable capture.

[0015] Secondly, this application provides a system for updating page configuration, including: a configuration receiving unit, a shadow configuration storage unit, a submission request unit, a page boundary detection unit, an atomic submission control unit, an active configuration storage unit, a version management unit, and a page sending unit. The configuration receiving unit receives page configurations and update categories. The shadow configuration storage unit stores complete candidate page transactions and keeps them stable while the request is in progress. The submission request unit generates and maintains submission requests with transaction identifiers. The page boundary detection unit detects page security boundaries based on the complete physical page sending status. The atomic submission control unit reviews the submission permission based on the current protocol status and the session and logical page context applicable to the candidate transaction, and switches the entire candidate page transaction to an active page transaction when the submission request, data capture conditions, page security boundaries, and submission permission are all satisfied. The active configuration storage unit stores or selects currently protected active page transactions. The version management unit assigns a configuration version identifier to successfully submitted active transactions. The page sending unit consistently determines page snapshots and configuration version identifiers from the same selected transaction and generates complete physical pages based on the page snapshots.

[0016] Beneficial effects:

[0017] The page configuration update method and system provided in this application combine page security boundaries and protocol submission permission as common conditions for atomic submission. The overall switch is completed under the conditions that the candidate transaction is stable, the page is completely sent, the protocol state allows it, and the context is still applicable. This can avoid page tearing and mismatch between configuration update and protocol state.

[0018] By distinguishing configuration versions, logical pages, and acknowledgment sequences, and retaining identified completion events, improper accumulation of acknowledgment counts and loss of completion notifications can be avoided; through cross-clock handshakes and storage ownership management, consistent capture of multi-bit transactions in the target domain and safe release of storage can be guaranteed. Attached Figure Description

[0019] Figure 1 A schematic diagram illustrating the conditions under which real-time configuration updates cause page tearing, provided for embodiments of this application;

[0020] Figure 2A structural block diagram of the automatic negotiation page sending system provided in the embodiments of this application;

[0021] Figure 3 A state diagram of atomic submission and cancellation processing provided for embodiments of this application;

[0022] Figure 4 This is a schematic diagram of a copying asynchronous request confirmation handshake process provided in an embodiment of this application. Detailed Implementation

[0023] The page configuration update method provided in this application embodiment can be applied in the Ethernet auto-negotiation process to perform cross-clock reception, legality verification, protocol context review, page boundary atomic commit, version management, acknowledgment statistics, and consistent transmission of the capability page to be transmitted and its control parameters. The following description uses the auto-negotiation defined by IEEE 802.3 protocol Clause 73 as an example.

[0024] It is understood that the embodiments of this application are also applicable to FLP bursts as defined in Clause 28 and other serial communication transmission circuits with defined physical transmission unit boundaries and protocol state constraints, and the embodiments of this application do not limit them.

[0025] Before introducing the page configuration update method provided in the embodiments of this application, the relevant technical issues will be explained first.

[0026] For example, please refer to Figure 1 , Figure 1 This is a schematic diagram illustrating the conditions under which a real-time configuration update causes page tearing, as provided in an embodiment of this application.

[0027] like Figure 1 As shown, position i and position j are two different payload positions that actually change in the link codeword, and position i is sent before position j. The old configuration A and the new configuration B differ at both positions.

[0028] When position i has already been sent according to the old configuration A, but position j has not yet been sent, if the management clock domain updates the real-time configuration bus to the new configuration B, then position i comes from the old configuration A and position j comes from the new configuration B. The resulting hybrid link codeword is neither equal to the old configuration A nor equal to the new configuration B. When the receiving side performs capability matching or confirmation judgment according to this hybrid codeword, it may get an incorrect result.

[0029] Figure 1 The example shown is merely a mixed risk illustration of multi-field or multi-bit updates. Updates that only change a single field or a single bit do not necessarily result in a third type of mixed content, but the field sampling and encoding timing must still meet the requirements. Figure 1 This does not constitute a limitation on the embodiments of this application.

[0030] Therefore, this application provides an automatic negotiation page sending circuit, for example, please refer to [link to example]. Figure 2 , Figure 2 This is a structural block diagram of the automatic negotiation page sending circuit provided in an embodiment of this application.

[0031] like Figure 2 As shown, the auto-negotiation page sending circuit includes a configuration receiving module 101, a shadow configuration storage module 102, a configuration validity check module 103, a submission request module 104, a page boundary detection module 105, an atomic submission control module 106, an active configuration storage module 107, a version management module 108, a page sending module 109, a context binding completion statistics module 110, a cross-clock handshake module 111, an exception handling module 112, a protocol context and submission permission module 113, and a completion event retention module 114.

[0032] The configuration receiving module 101 receives page configurations and update types submitted by software, management bus, arbitration state machine or other control sources. For multiple configuration sources, the configuration receiving module 101 arbitrates the write and commit to prevent fields from different sources from being unexpectedly merged into the same transaction.

[0033] The shadow configuration storage module 102 saves candidate page transactions, field validity information, and storage status, and can set the dirty bit vector, field validity vector, and transaction integrity flag.

[0034] The configuration validity check module 103 checks page fields, control parameters, field dependencies, and generator configuration.

[0035] The commit request module 104 generates a persistent commit request after the candidate transaction is complete and the initial check passes, and maintains the request identity.

[0036] The page boundary detection module 105 identifies security boundaries based on the complete physical page sending status and sending pipeline status.

[0037] The atomic commit control module 106 enables the candidate transaction to take effect as a whole when the request, data capture conditions, page security boundaries, and protocol commit permission are all satisfied.

[0038] The activity configuration storage module 107 saves the current active page transaction, or determines the current protected transaction through the activity storage selection signal.

[0039] Version management module 108 assigns a version identifier to successfully submitted activity transactions.

[0040] The page sending module 109 generates a physical page based on a consistent page snapshot and generates a completion event with context after complete sending.

[0041] The context binding completion statistics module 110 updates and completes the statistics based on the relationship between the session, logical page, confirmation stage, page qualification, and configuration version.

[0042] The cross-clock handshake module 111 consistently transmits commit requests, candidate transactions, commit results, memory release information, and completion events that need to cross domains.

[0043] The exception handling module 112 handles illegal configurations, context invalidations, wait timeouts, independent resets, gating exceptions, and page aborts.

[0044] The protocol context and commit permission module 113 maintains or receives the current protocol context, performs dynamic review of candidate transactions, and generates commit permissions.

[0045] The completion event retention module 114 stores unconsumed completion events and releases them based on explicit consumption or cancellation conditions.

[0046] It should be noted that the protocol context and submission permission module 113 can be implemented in combination with the automatic negotiation arbitration state machine or the atomic submission control module 106, and the completion event retention module 114 can be implemented in combination with the context binding completion statistics module 110, the cross-clock handshake module 111 or the exception handling module 112. The method of dividing the modules is not limited in this application embodiment.

[0047] In a preferred embodiment, automatic negotiation arbitration, permission review, cancellation of arbitration, atomic commit, and page-by-page completion statistics reside in the same transmission clock domain. The authoritative control logic in the transmission clock domain determines the effective order of commits and cancellations. The management side commits configurations through a complete transaction handshake. Ordinary management candidates and control candidates used to advance the current exchange use independent storage. In this document, "module" and "unit" refer to hardware functional units implemented by digital logic circuits. Module division does not limit their physical layout or require them to be implemented using independent circuits.

[0048] The embodiments of this application distinguish the following identifiers:

[0049] The transaction ID (transaction_id) identifies a candidate commit, including successful, rejected, or canceled requests.

[0050] The negotiation session identifier (session_id) distinguishes different negotiation sessions and the context before and after a restart;

[0051] The active_epoch indicator identifies an active configuration that has been successfully applied in a single instance.

[0052] The logical page identifier logical_page_id identifies either the current swapped Base Page or the specific logical NextPage;

[0053] The ack_sequence_id identifier confirms the confirmation phase, which defines the starting point for counting and the eligibility rules.

[0054] The physical page sequence number (page_sequence) identifies a specific physical page transmission. This identifier does not require separate registers and can be expressed using combined encoding, state machine states, table indexes, or other equivalent mechanisms.

[0055] Different configuration versions do not necessarily mean different confirmation sequences, and the same configuration version does not necessarily mean that the two pages belong to the same confirmation stage.

[0056] Candidate page transactions include at least the page payload, as well as one or more of the protocol-related fields and transaction control and management fields.

[0057] Protocol-related fields include: a link codeword of a predetermined width, such as a 48-bit link codeword for Clause 73 or a 16-bit link codeword for Clause 28; page types such as Base Page, Message Next Page, or Unformatted Next Page; control information applicable to the protocol and page type, including acknowledgment fields, acknowledgment field 2, next page field, flip field, and random number field; and generator configuration, seed, mode, or loading control for generating protocol-specified encoded supplementary information.

[0058] Transaction control and management fields include: commit transaction identifier; update category and field update mask; target session identifier, logical page identifier, and expected protocol stage; necessary prior version or context dependency information; acknowledgment counting rules or references to corresponding protocol configuration items; send enable, repeat policy, abort policy, and error handling methods; configuration source, write sequence number, and debug metadata. If the link codeword and a separate control field describe the same protocol bit, a unique assembly rule is specified, and their consistency is checked to avoid conflicting sources for the same send field.

[0059] When the configuration side performs one or more field writes, the written data first enters the writable shadow storage area. During the writing process, the page sending module 109 continues to use the original active transaction or its page snapshot.

[0060] Partial field updates form complete candidate transactions by: constructing complete candidate values ​​based on a specified baseline configuration, overwriting the updated fields according to the field mask, recording the baseline version or context on which they depend, and confirming that the dependencies are still valid before committing.

[0061] The submit request module 104 generates a request after all of the following conditions are met: the fields required for this transaction are complete; the initial validity check is passed; the software is explicitly committed or the predefined hardware commit conditions are met; there are available transaction slots; and there are no high-priority reset, disable, or error handling conditions that would prevent the request.

[0062] Submission requests are represented using hold-up levels, request toggles, or transaction queue entries, and instantaneous notifications that may be missed by the pause clock are not used as the sole basis. For a single-slot management handshake channel, at most one incomplete management request is allowed at any given time. Before the request is completed, the original candidate transaction remains stable, and new management writes are blocked, return to a busy state, or enter another free slot with independent ownership.

[0063] Control updates for advancing the current exchange, such as the acknowledgment field, acknowledgment field 2, and echo random number, are performed on the sending side using independent candidate storage, request status, and result status consumed by local protocol logic. Control candidates are assembled from the currently active logical page and the valid receiving context, and still need to undergo integrity checks, protocol review, and the same page boundary submission.

[0064] Therefore, the back pressure of pending management transactions and management results will not consume all resources of the necessary control path, and management-side pauses will not block necessary protocol actions. The configuration side must not directly overwrite candidate transactions that may still be captured by the sending side due to local timeouts; cancellation operations are completed according to a mutually agreed cancellation or reset protocol.

[0065] The checks performed by the configuration validity check module 103 include one or more of the following: field width, reserved bits, and selection field validity checks; page type and payload format matching checks; consistency checks between repeated description link codeword bits and independent control fields; checks that declared capabilities do not exceed hardware support capabilities; register range checks for transmission parameters and counting parameters; validity checks for the seed and state of the selected pseudo-random generator; and configuration data integrity checks, such as parity checks or cyclic redundancy checks.

[0066] For generators with an all-zero locked state, a seed non-all-zero check is performed; for other generators, their prohibited state is checked according to the selected architecture, and non-all-zero is not a uniform requirement for all generators. The number of confirmation pages, the counting starting point, and the allowed range are determined by the protocol version adopted and the specific confirmation stage. A register can represent a certain value, but this does not mean that the value is valid in the protocol. Therefore, the implementation checks both the register range and the protocol's allowed range simultaneously.

[0067] For conditions that depend on the current session, received results, previous logical page, or arbitration status, the protocol context and submission permission module 113 performs a dynamic review before actual submission, including one or more of the following:

[0068] Is the session to which the candidate transaction belongs still valid? Is the target logical page compatible with the current exchange object?

[0069] Confirm whether the update of the confirmation field, confirmation field 2, or the echo random number field meets the corresponding protocol conditions;

[0070] Whether the retention or flipping of a field is consistent with the logical page progression;

[0071] Whether the next page indication and the status provided on subsequent pages comply with the requirements of the current agreement;

[0072] Has the baseline version or receiving context that the candidate transaction depends on become invalid?

[0073] Does this type of configuration change require a delay to a subsequent session, or renegotiation as specified in the protocol?

[0074] Repeated sending of the same logical Next Page or updates of controlled fields will not automatically flip the Toggle due to internal transaction commits. The initial value of the Toggle and its progression relationship are determined by the corresponding rules of the adopted protocol.

[0075] The cross-domain context used for dynamic review is obtained through a consistency interface, and the permission can be guaranteed to correspond to the current transaction through context generation identifiers, authorization tokens, or equivalent mechanisms. In this embodiment, the context invalidation and cancellation take effect at the moment when the authoritative control logic on the sending side accepts the corresponding event, and permission review, context invalidation, and submission are uniformly arbitrated in this logic.

[0076] After the cancellation takes effect, the old license may not continue to make uncommitted transactions effective; when cancellation and ordinary commit are established at the same arbitration time, cancellation takes precedence, and previously committed transactions are returned to the actual committed results, and the release of candidate data complies with existing handshake and ownership rules.

[0077] The update categories and corresponding effective rules are as follows:

[0078] The updates to the confirmation field, confirmation field 2, and echo random number field generated by the arbitration state machine are submitted at the security boundary of the valid page after the corresponding protocol conditions are met.

[0079] The next logic, Next Page, and its flipped field updates are committed at the valid page security boundary after the current exchange progress conditions are met; if the current protocol state does not allow direct updates, the software modifications to the capabilities adopted in the current negotiation, forward error correction, and other information are delayed until a subsequent session or after a restart as specified.

[0080] Sending enable, disable, and abort policies are handled according to different rules for normal stop and emergency exception; debug metadata that does not affect sending, timing, matching, and confirmation eligibility is managed independently and does not change the sending configuration version or confirmation sequence.

[0081] The current exchange progress conditions include the local sending conditions, receiving matching conditions, acknowledgment conditions, and state transition conditions required by the corresponding protocol, and do not use the completion of the local acknowledgment field count as the sole condition.

[0082] If the check fails, the candidate transaction will not take effect and an error result with the committed transaction identifier will be output; for static illegal configurations, the original active transaction will continue to run; for cases where the session has expired, the recovery process specified in the protocol will be executed, but there is no guarantee that the original page will continue to be sent.

[0083] Page security boundaries are used to ensure the integrity of physical pages and transmitted codes, but they do not constitute a sufficient condition for a transaction to be committed on its own.

[0084] The page boundary detection module 105 identifies one or more of the following states based on the encoder, the sending state machine, and the necessary pipeline states:

[0085] There are currently no physical pages that need to be protected, and the sender is in an idle or disabled state that allows receiving configurations;

[0086] The current physical page has completed all specified sending positions and protocol-defined tail timings, and the next physical page has not yet started and has not yet referenced its sending configuration;

[0087] The current FLP burst has completely ended, and there is an effective silent window before the next burst begins;

[0088] The current page has been aborted according to the exception policy, the output has entered the specified state, and the recovery conditions allow for loading the new configuration; after the session is restarted, the relevant page state, context and sending pipeline have been initialized according to the recovery process.

[0089] Security boundaries must not be generated solely based on the fact that the last payload bit has been read by upstream logic. If there are protocol-specified encoding positions or tail timings after the link codeword, they should be included within the entire page scope.

[0090] If a First Input First Output (FIFO) buffer, encoding pipeline, or parallel-to-serial conversion stage exists, the page completion event reflects that the corresponding page has actually completed sending, rather than merely indicating that the payload has entered the downstream buffer.

[0091] During the full page transmission, regular configuration requests do not change the active transaction and snapshot used by the page. Candidate transactions remain pending commit until the data capture conditions, security boundaries, and protocol permissions are simultaneously met. The security boundary in the continuous transmission protocol can be a logical clock edge, and no additional idle periods are required. The online page interval must not be arbitrarily increased in order to wait for the commit status.

[0092] For example, please refer to the table below, which illustrates the page boundary submission timing relationship bound by the license agreement provided in the embodiments of this application:

[0093] Table 1. Page Boundary Submission Sequence Table Subject to License Agreement

[0094] Logical Moment Page being sent Candidate Transaction B Safety boundary License Agreement action T0 Version A page internal Not ready yet none No restrictions Continue sending A T1 Version A page internal Freezing and request issued none This can effectively keep B pending submission. Continue sending A T2 A complete physical page and tail timing end. Meet capture conditions efficient none Do not submit; continue using A according to the agreement. T3 The full page boundaries allowed for subsequent submissions Still valid and context verification passed efficient efficient Generate a commit event, select B and assign a new version. T3 Same edge New page loading The next page begins referencing configuration B at this edge. Selected at this edge Valid at the same boundary Consistently establish data and version snapshots from the selected B. T4 Version B page internal The original request has been marked for processing. none No restrictions Use only page snapshots of B T5 Return path reaches the configuration side The corresponding result has been captured. No restrictions No restrictions Slot release is handled by the storage method on the configuration side.

[0095] As shown in the table above, the edge at the end of the last specified time interval of the current physical page is also the starting edge of the next physical page. At this edge, the old page end determination, active transaction selection, and new page snapshot creation are completed in logical order. The content that depends on the configuration in the next page is only obtained from the selected transaction. The above logical actions do not require an additional clock cycle. The commit and new snapshot selection are completed at this edge, no later than the moment when the configuration is first referenced on the next page.

[0096] The logical timings in Table 1 are not limited to equal-length periods. For a wider security window, execution can be carried out in steps within the window, but it must not be later than the moment when the configuration is first referenced on the next physical page.

[0097] The atomic commit control module 106 generates a commit event when all of the following conditions are met: there is an unprocessed valid request; the candidate transaction is complete and meets the target domain capture condition; the page security boundary is valid; the current protocol state allows this type of update; the candidate transaction is applicable to the current session and logical page context; and there are no reset or exception conditions to cancel the transaction.

[0098] If the boundary has been reached but the protocol does not yet allow updates, the transaction remains in a pending state; if the suitability of the candidate transaction has expired, the transaction is rejected, canceled, or regenerated, rather than waiting for any subsequent boundary to take effect.

[0099] For example, please refer to Figure 3 , Figure 3 A state diagram of atomic submission and cancellation processing provided for embodiments of this application.

[0100] like Figure 3 As shown, the commit event `commit_fire` indicates that the request is valid, the data capture conditions are met, the security boundary is valid, the protocol permission is valid, and there are no cancellation conditions. The actual configuration switch occurs at the valid boundary that generated `commit_fire`, not after the boundary fails. The diagram does not specify commit states that require an additional clock cycle to execute. When a cancellation request is received after a commit has been completed, the actual commit result is returned; a valid transaction must not be disguised as uncommitted.

[0101] The following steps are completed within a single logical atomic commit event: Selecting a complete candidate transaction as the new active transaction; synchronously switching page payload, type, and related send control parameters; assigning a new active configuration version identifier; establishing the association between the transaction identifier, session, and active version; processing statistical status according to the confirmation context rules; marking the request as processed to prevent duplicate commits; and generating a return record containing the transaction identifier, commit result, and effective version. Configuration-side dirty bit cleanup, request reversal, and writable slot release are performed after the configuration side receives the corresponding result or release notification, and are not required to occur at the same physical moment as the sender's commit.

[0102] When the old page completes, the new transaction commits, the old confirmation count reaches the threshold, and the new page snapshot is created at the same boundary, the old page completion event always carries the configuration version and context of the old page. The statistical eligibility of the old page is determined based on its corresponding confirmation context. The existing valid old completion events are saved first and are not overwritten by the new version. The new configuration only takes effect when the commit permission is valid. The data, control parameters, and version of the new page are determined from the same selected transaction.

[0103] For session cancellation or emergency termination accepted at the same arbitration time on the sending side, cancellation or termination takes precedence over normal submission; events in expired contexts shall not continue to advance the protocol state; pages that have been fully sent may retain diagnostic records, but whether they still have protocol completion significance depends on whether the context is valid.

[0104] In one implementation, after the sending side copies the complete shadow transaction to the independent active register at the valid page security boundary and completes the commit, it returns a commit confirmation. The source side shadow storage is no longer the source of reading the current page thereafter. After receiving the commit confirmation, the configuration side releases the original shadow slot according to the channel rules. In this implementation, the source slot is not released with a separate capture confirmation before the commit.

[0105] In another implementation, at least two storage units respectively assume the storage roles of active and candidate transactions, and the commit is completed by changing the active storage unit selection signal. A storage unit has at least four logical states: writable, frozen in commit, actively occupied, and released. Once a candidate storage unit becomes the active storage unit after its pointer is toggled, it cannot be immediately overwritten simply because a commit confirmation has been returned. An old storage unit that has exited the active state is only allowed to be rewritten after a page is sent, the pipeline no longer reads its contents, and the configuration side is aware of the release result.

[0106] If the page has been fully copied to the independent snapshot register, the release of the storage is determined based on the actual read dependencies, and the release does not affect the completion events and context records that still need to be retained.

[0107] The page sending module 109 obtains the following consistently from the same selected active transaction at the beginning edge of the page or within the previous valid security window: page payload; page type and protocol control fields; configuration parameters affecting the page encoding; page configuration version; session, logical page, and necessary confirmation context; and page-by-page information actually used by this page.

[0108] During the transmission of a complete page, the already determined page content is not modified by the management register, arbitration combo output, or cross-domain real-time bus. For page-by-page random information or other coded supplementary information specified in the protocol, their position, quantity, and online timing comply with the requirements of the adopted protocol version and are not arbitrarily added simply because they are not part of the link codeword payload.

[0109] Page-by-page information can be generated and latched before the page begins, or it can be generated by a local generator with fixed configuration and initial conditions according to defined rules. The working state within the page proceeds normally according to the encoding algorithm, but no new seeds are loaded or the generation mode is changed midway through the page.

[0110] The generator seed loading and page value retrieval are controlled separately. Unless explicitly required by the protocol or the selected algorithm, the same seed will not be reloaded at the beginning of each page due to repeated sending of the same configuration version.

[0111] For example, please refer to the table below, which is a page transaction and page snapshot structure table provided in the embodiments of this application:

[0112] Table 2. Page Transaction and Page Snapshot Structure Table

[0113] Field Categories Field content Store or associate objects illustrate Page load Clause 73: 48-bit link codeword or Clause 28: 16-bit link codeword for ordinary FLP. Candidate and event affairs Explained according to specific page format Page Type Base Page, Message Next Page, Unformatted Next Page Candidate and event affairs Matching with target protocol and phase Protocol control Applicable fields for ACK, ACK2, NP, Toggle, Nonce, etc. Candidate and event affairs Follow the update rules of the corresponding protocol. Generator Configuration Seed, mode, load control and legal status parameters Candidate and event affairs Not equal to the random value on this page Submit Identity transaction_id, source, update category Candidate transactions and results Refusal to engage in affairs also carries an identity. Protocol Dependency Session, logical page, expected stage, baseline version Candidate Transaction and License Context Dynamic review upon submission Confirmation of control Counting start point, eligibility, threshold and configuration reference Confirm Context Determined at the specific agreement stage Configuration version active_epoch, page_epoch Events and Page Snapshots The actual configuration used is indicated. Page-by-page information Random information, page number, and valid identifiers on this page. Physical page snapshot Normal page-by-page changes do not constitute a submission. Sending strategy Enable, repeat, normal stop, and abort strategies Activity or controlled management status Online timing must not be violated Debug metadata Debug tags, record indexes, etc. Independent management area or affairs Independent updates are possible without affecting protocol behavior.

[0114] The page sending module 109 determines the page snapshot (page_snapshot) to be sent at the beginning edge of the page or within a previous valid security window, no later than the first time the configuration is referenced on that page. The page snapshot includes or is associated with: the selected active page transaction, the page configuration version identifier, the session identifier, the logical page identifier, applicable confirmation context information, the page-by-page random information actually used on this page or other coded ancillary information, and optional physical page sequence number and valid flags. Normal changes in the page-by-page random information do not constitute a configuration commit, do not increment the active configuration version, and do not reset the confirmation count.

[0115] The random number protocol field and the physical encoding-related random information are managed separately and do not share the same update rules because both involve random generation. Page data, configuration version, and session identity remain unchanged once determined. If the protocol initializes the acknowledgment count at a specified location within the page, the current page can be associated with a count qualification at that location. This association does not rewrite the page payload or configuration version, nor does it retroactively include other pages sent before initialization into the new sequence.

[0116] Each successful submission generates a new activity configuration version. At the start of each page, the actual page configuration version used is recorded and associated with its session and logical page identity. After a page is fully sent, a page completion event (page_complete) is generated, carrying a valid page completion flag, the page configuration version, session identifier, logical page identifier, the stage or sequence information required for qualification confirmation, and an optional physical page sequence number. Pages that are aborted, have expired timings, or cannot be confirmed as fully sent do not generate a valid completion event, but can generate independent exception diagnostic events.

[0117] When the protocol enters a stage that requires counting the number of local confirmation pages, an acknowledgment sequence identifier is established and the count is initialized at the send count initialization point specified in that stage, and the following are bound: the current session; the current logical page; the confirmation stage and the starting point of the count; the page type and control field eligibility; the applicable configuration version or verified version association; the completion quantity specified in that stage; and other conditions.

[0118] The confirmation page count is increased only if all of the following conditions are met:

[0119] The page has been sent successfully and completely.

[0120] The page belongs to a session that has not yet been canceled and is a logical page;

[0121] The page meets the counting eligibility requirements specified in this confirmation phase;

[0122] The page version is associated with the version of the confirmation context according to the same rules;

[0123] This page has not been counted repeatedly.

[0124] If a page starts before a phase switch but completes after a phase switch, its inclusion in the count is determined according to the corresponding protocol phase's counting definition, not solely based on the current state or version at the time of completion. The submission of the confirmation field update and the initialization of the confirmation sequence are separate events. The sequence is bound to the actual effective confirmation configuration used in that phase. Pages that have already sent confirmation field updates are not automatically included in later-established counting sequences simply because their confirmation bits are the same.

[0125] After entering the corresponding local acknowledgment sending phase, when the count initialization point specified by the protocol is reached, an acknowledgment context is established: the session identifier is S1, the logical page identifier is L1, the acknowledgment sequence identifier is Q1, the acknowledgment version identifier is A, and the acknowledgment count is cleared to zero.

[0126] Where A represents the configuration that was already in effect at the time of count initialization and is eligible for confirmation in this phase, rather than any old configuration before the confirmation field was updated; B represents another allowed update subsequently; and N represents the number of completions specified in the current protocol phase.

[0127] The statistics module determines whether a page belongs to the confirmation sequence according to the protocol. Only valid, complete pages that simultaneously meet both the context and version conditions are counted. When the confirmation count reaches the completion conditions specified for this stage, a confirmation sequence completion event is generated and saved.

[0128] If a normal configuration B would change the current confirmation semantics, then B remains pending submission, or the old confirmation context is canceled and a new confirmation context is established by the protocol recovery process. Simply clearing the count is not equivalent to completing a protocol restart.

[0129] If B takes effect as permitted by the agreement after confirmation sequence Q1 is completed, but the completion event of Q1 has not yet reached the arbitration side across domains, the event is still retained as Q1. The arbitration side consumes the event according to the waiting Q1 and does not discard it because the current active version is B.

[0130] B's submission still satisfies the full agreement license, and Q1's local count reaching N does not alone constitute a sufficient condition for advancing to the next logical page. The table below uses a strict single-version count as an example to illustrate the relationship between the confirmation context count and the retention of old completion events, and does not constitute a limitation on the embodiments of this application.

[0131] Table 3. Relationship between Context Count and Retention of Old Completed Events

[0132] event Page version Confirm Context Count changes Complete event handling Reaching the specified count initialization point A Session S1, Logical Page L1, Sequence Q1 Initialize to 0 none Qualified page completed A Q1 Increased to 1 This page has been recorded. Subsequent qualified pages completed A Q1 Accumulated up to N-1 none The last page is now complete. A Q1 N-1 increases to N save ack_sequence_complete(Q1, A) Submit B after the agreement is approved B Reservation, termination, or establishment of other contexts in accordance with the agreement Do not overwrite Q1 completion records The Q1 issue remains unresolved. The new page has started sending B. B New page context Process according to the relevant rules It does not affect the identity of the Q1 incident. The arbitration side will receive the Q1 case later. At this point, active can be B. The arbitration side is awaiting Q1. Non-repeated accumulation Consumption based on Q1, not discarded even if active is B.

[0133] Furthermore, the sequence of the above process is illustrated using a normal exchange with both sides having a next page field of 0 in the IEEE 802.3 protocol Clause 73 Base Page.

[0134] During the confirmation detection phase, the local end forms a complete control candidate with the confirmation field set and the echoed random number equal to the valid random number sent by the peer. It then makes it effective as configuration B within the legal boundaries. In this example, B represents the confirmation configuration for this specific example.

[0135] After configuration B takes effect, the peer page meets the conditions for acknowledgment matching, consistency matching, and random number matching required by the Base Page, and the arbitration enters the completion acknowledgment state. According to the protocol state diagram, when the separator completion condition is met and the remaining acknowledgment sending state is entered, the remaining acknowledgment count is initialized, an acknowledgment sequence is established and bound to the actually effective configuration B, without binding to the configuration before the acknowledgment field update.

[0136] Only complete and valid pages that belong to the acknowledgment sequence according to the initialization point and subsequent transmission status are counted. If the delimiter has already started and the subsequent link codewords belong to the new initialization sequence, the page is not excluded just because the delimiter was established earlier than the acknowledgment sequence. Pages that were completed before initialization or do not belong to the transmission sequence are not retroactively counted.

[0137] In this example, six qualified sends are selected as the condition for the completion of the remaining confirmations. The local logic formation stage completes the process at the specified completion time, and enters the subsequent state in combination with arbitration conditions such as both parties' next page fields being 0. This number corresponds to the remaining confirmation count in this stage, not the total upper limit of the total number of confirmation pages in the entire negotiation.

[0138] In another implementation, the verification context is allowed to associate multiple verified configuration versions and retain existing counts.

[0139] The conditions that allow the count to be retained include:

[0140] The session, logical pages, and confirmation phase remain unchanged;

[0141] The starting point for counting, eligibility rules, and completion conditions remain unchanged.

[0142] The page content, control fields, and sending behavior on which peer matching and local statistics depend remain unchanged; the changing fields are explicitly limited to those that do not affect the current confirmation semantics.

[0143] For example, when only managing changes to debug tags, while all fields related to sending, timing, and confirmation remain unchanged, the new version can be associated with the same confirmation sequence, retaining existing counts and the actual version of each page for traceability. This method retains the count of the same valid confirmation context, rather than merging arbitrary versions simply because confirmation field values ​​are equal. The actual page configuration version in the event is not rewritten, and the version association is retained until the old event is consumed.

[0144] In this application, the embodiments distinguish the following events: commit_ack indicates that the local candidate transaction has been successfully activated; page_complete indicates that a valid physical page has been sent locally; ack_sequence_complete indicates that the local side has completed the sending statistics required for the current confirmation stage; and protocol exchange completion indicates that the conditions for receiving matching, peer interaction, local sending, and state transition are all met.

[0145] A successful local commit or reaching a threshold number of sends cannot independently prove that the link partner has confirmed the new configuration. The completion event retention module 114 stores unconsumed events. The arbitration state machine receives events based on the transaction or acknowledgment context it is waiting for, rather than simply comparing them to the latest active configuration version.

[0146] The fact that the activity configuration has been switched does not constitute a sufficient reason to discard old, legitimate completion events. Events can only be released after they have been reliably consumed or their context has been explicitly cancelled. When event storage resources are insufficient, control operations that would cause unconsumed events to be overwritten should be prevented, or a retention mechanism that does not disrupt online timing should be adopted to avoid silently discarding critical completion notifications.

[0147] In a preferred embodiment, page-by-page completion events are consumed promptly by the local statistical logic on the sending side. Confirmation counts and necessary protocol advancements dependent on these counts are executed locally by the logic at the state or sending boundary specified in the corresponding protocol, without waiting for notification from the management side. If cross-domain stage recording or page-by-page event queues are used, a maximum number of pending events or a maximum consumption pause are specified, and limited storage capacity is configured based on the event rate.

[0148] When the configuration clock domain is asynchronous with the page sending clock domain, a bundled-data request confirmation scheme can be used.

[0149] For example, please refer to Figure 4 , Figure 4 This is a schematic diagram of a copying asynchronous request confirmation handshake process provided in an embodiment of this application.

[0150] like Figure 4 As shown, the configuration side completes the candidate transaction and freezes the corresponding slot's back bit holding request or flip request flip bit. The sending side performs multi-level synchronization on the request control signal, waits for the page security boundary and valid protocol permission after confirming that the data capture conditions are met, captures or switches the complete transaction and generates the commit result.

[0151] The candidate transaction remains stable from the time the request is issued until release is permitted. Before the target domain is captured, all data bits have arrived and met the setup and hold times. This condition is guaranteed through maximum data path delay constraints, control and data relative delay analysis, target domain waiting margin, or a verified cross-clock structure. Multi-level synchronization requests only reduce the probability of metastable propagation of control signals and do not automatically replace the settling time requirements of multi-bit data paths.

[0152] The sending side will submit the transaction identifier, commit status, effective version, and necessary error information to form a stable return record, and then notify the configuration side through a handshake or asynchronous queue. The return record will not be overwritten by subsequent results before it is reliably captured by the receiving side. The receiving side can read the continuously changing real-time version bus without constraints after synchronizing an acknowledgment bit.

[0153] Figure 4 For copying to an independent active register, for the memory pointer flipping method, the commit confirmation does not mean that the new active memory can be overwritten, and memory release is handled separately based on read dependencies and release notifications.

[0154] In another implementation, a complete management transaction asynchronous FIFO is used. Each FIFO item stores transaction data, transaction identifier, update category, and context dependencies. Necessary control candidates for advancing the current exchange are submitted through an independent control store and channel. They are not queued after the head of the management queue that still needs to wait for the current exchange to complete. The sending side can read the head of the queue in advance for inspection, but it only takes effect and completes the corresponding dequeue when the capture condition, security boundary, and protocol permission are all met. The dequeue of the next logical Next Page also meets the page exchange advancement condition and is not unconditionally advanced at each physical page boundary.

[0155] The submitted request remains pending a successful, rejected, or canceled result, or until both parties complete the prescribed reset and recovery process.

[0156] Optionally, the submission request is detected by normally open logic or an ungated parent clock domain, controlling the glitch-free clock gating unit to perform controlled wake-up. The wake-up request does not completely depend on the generation of an already stopped clock. Normal gating satisfies: the current protocol state allows stopping the corresponding transmission function; it does not disrupt the physical timing of the ongoing page; it does not violate page intervals and protocol timing requirements; pending requests and return results are in a state that allows pause or recovery. For transmission engines that must continuously generate online timing, clocking cannot be arbitrarily stopped simply because the current page has ended.

[0157] If an abnormal clock stop has caused the current page's transition interval, pulse interval, or other physical timing sequence to fail, the page will be marked as invalid and will not be counted in the valid completion statistics. After recovery, the page will be resent from the page starting point specified in the protocol, and a session recovery or restart process will be executed if necessary.

[0158] If only configuration or control logic is suspended, and the independent sending engine can still continuously meet online timing requirements, then the control state can be retained. This situation does not fall under the category of physical page pause followed by transmission. Hold-alive requests guarantee that the request will not disappear solely due to a lack of sampling clock; the request's eventual completion also requires achievable wake-up or recovery conditions. If these conditions are not met for an extended period, a timeout status and controlled cancellation mechanism are provided.

[0159] Page-stopping strategies for resets, hardware failures, or emergency disablements include: immediately stopping output and marking incomplete pages as invalid; stopping after fully sending the current page if permitted by the protocol and without violating urgent requirements; and terminating transmission at a stop position explicitly permitted by the target protocol. Pages that are truly aborted do not update valid completion statistics. If stopping after full transmission is selected, whether the complete page is included in the current acknowledgment statistics is checked, and whether its acknowledgment context is still valid is also verified. When resuming transmission, a valid page start point and required state are re-established; transmission does not continue from the middle of an invalid page.

[0160] When a session restarts, update or re-establish the session identity; clean up the received matching results and confirmation status according to the protocol; cancel inapplicable candidate transactions of the old session; re-form transactions suitable for the new session for candidate configurations that need to be retained; make the old session completion event lose its protocol advancement qualification; and initialize the sending page and necessary timing status as required.

[0161] Changing only the session identifier without processing residual requests, events, and receive states does not constitute a complete recovery. Request toggle bits, acknowledgment toggle bits, processed request states, and slot ownership have consistent initialization and recovery rules. When a sender-side reset takes effect, or when the sender-side authoritative control logic receives a reset notification or recovery request from another clock domain, subsequent new transactions are first prevented from taking effect. Then, through reset generation, channel initialization handshake, or equivalent mechanisms, the mutual understanding of requests and ownership is re-established, avoiding the identification of residual levels before the reset as new requests.

[0162] A timeout on the source side does not necessarily mean that the target side has not submitted. The cancellation request takes effect when it is accepted by the authoritative control logic on the sending side. It is explicitly arbitrated with the target submission action. At the same arbitration moment, cancellation takes precedence over normal submission: if the submission has been completed, the result of the submission is returned; if the cancellation is accepted before the submission, its subsequent effect is prevented and the cancellation result is returned; the candidate slot is only released after both parties confirm that the cancellation is completed or the completion channel is restored.

[0163] Version identifier bit width and reuse rules consider: the number of incomplete commits, pages still being sent or in the pipeline, unconsumed completion events, unreleased version association records, and cross-domain delay and reset residual information. When an old identifier may still be validly referenced, the same identifier will not be reassigned to indistinguishable new objects. Mechanisms such as sufficient bit width, delayed reuse, session generation, and event purging prevent confusion between old events and new versions.

[0164] In one specific implementation, the configuration side sets up two storage banks, page_bank_0 and page_bank_1, each containing a 48-bit payload and associated control fields, for general management configuration, and separate control candidate registers and transmitting side control activity registers for control updates of the current exchange.

[0165] The active storage selection signal indicates the active storage device; another storage device is only allowed to be written to by the configuration side after writable ownership has been acquired. The configuration side assigns a commit transaction identifier to the candidate Base Page, fills in the update category, session, and context dependencies, freezes the storage device, and submits a request after completing the checks.

[0166] The sending side flips the active storage when the candidate data is stable, the current complete DME physical page and the tail timing specified by the protocol are completed, the next page has not yet started and has not referenced any configuration, the protocol state allows the update, and the session and dynamic dependency review are passed. This implementation does not reference the next page configuration in advance; the end of the complete page and the start of the next page occur at the same edge. The submission selection is determined logically before referencing a new page at that edge, without inserting additional idle cycles.

[0167] The same commit event assigns an active configuration version to the new active transaction, and the data and configuration version of the new page snapshot are consistently obtained from the same selected storage. Once the original candidate storage becomes the active storage, configuration writes remain disabled. The old active storage becomes writable after all read dependencies have ended and the configuration side has received a release notification.

[0168] If the software requests to modify capability information that is not allowed to be changed directly during the current confirmation process, the request will not take effect immediately upon reaching the page boundary. The configuration will be delayed to a subsequent session by the arbitration side, or resubmitted after the specified restart process.

[0169] When a regular candidate occupies an inactive storage unit, necessary updates such as confirmation fields can still be received by the independent control candidate register. At legal boundaries, the complete control candidate is copied to the control activity register, and the source of the active transaction is switched to that register. For regular management commits, the corresponding storage unit is selected. The active source, transaction identity, and configuration version are determined in the same commit. The control activity register and control candidate register are separate. The current page must not be overwritten in advance if there are still read dependencies. Control results are consumed locally and do not wait for management results to be released.

[0170] In one specific implementation, the shadow configuration storage module 102 sets up a logical Next Page queue, each item including Next Page payload, page type, logical page identifier, subsequent page indication, applicable session, and control information for the flip field expectation relationship or the flip field generated by hardware.

[0171] Advancement to the next queue item only occurs when the current page exchange advancement conditions are met and the valid page security boundary is reached. The exchange advancement conditions include the receive matching, acknowledgment, and local transmission conditions required by the adopted protocol. If the current logical page has not yet been exchanged, the logical page continues to be repeatedly transmitted.

[0172] To avoid blocking necessary control updates in the next logical page queue, updates to the current logical page, such as confirmation fields and second confirmation fields, are submitted through independent control candidate storage and request slots, or through priority channels with reserved data, request, and local result resources. Priority does not replace the aforementioned available resources, and management result backpressure does not block necessary control updates.

[0173] Control updates and the current logical page payload are assembled into a complete candidate transaction. After context review and page boundary commit, control updates of the same logical page do not cause the logical page queue to be dequeued, nor do they automatically flip the Toggle due to internal commits.

[0174] In one specific implementation, the management configuration uses a single shadow slot and a send domain independent activity register. Necessary control updates have separate candidate storage, requests, and local result states, which are not limited by the occupancy of the management slot.

[0175] After the management side completes the write, it flips the request flip bit and prevents further writing to that slot. The sending side identifies the request through a multi-level synchronizer and ensures the capture condition through data latency constraints or waiting margin. When the effective security boundary and protocol permission are both satisfied, the complete candidate transaction is copied to the active register, and a return record containing the transaction identifier, result, and version number is saved.

[0176] The management side captures the handshake consensus result upon returning a record, subsequently allowing the original shadow slot to be rewritten. The returned record remains unchanged until consumed by the management side. If the management side commits again, it returns to a busy state if there are no free slots; otherwise, it receives the data as a separate transaction if other independent slots are available. Overwriting the original request by flipping the request bits of the incomplete channel again is not allowed. Independent resets or cancellations are handled through channel recovery rules, and locally clearing the request register does not replace the termination of a mutually agreed-upon transaction.

[0177] In one specific implementation, the configuration check module rejects a transaction if it finds that a reserved bit is invalid, the page type is mismatched, the capability exceeds hardware support, the confirmation parameters do not conform to the current protocol stage, or the generator seed belongs to the prohibited state of the selected structure. The exception handling module records the error type, committed transaction identifier, target and current session identifier, current active version, current logical page, and arbitration status, and optionally records the timestamp and physical page sequence number.

[0178] If the initial check of a candidate transaction passes, but the sending-side authoritative control logic accepts a peer restart, session cancellation, or changes in related received results before committing, and determines that the candidate dependency has become invalid, then dynamic review fails, and a context invalidation result is returned. Transactions that have already been committed are processed according to the actual commit result and subsequent recovery procedures. For cases where only the candidate configuration is invalid but the current session is still valid, the current activity page continues sending according to the original process; for cases where the session has become invalid, protocol recovery is executed, and the original confirmation process continues unconditionally.

[0179] In one specific implementation, the above method is applied to Clause 28 FLP burst mapping. A complete FLP burst is considered a physical page, with the page payload being a 16-bit link codeword. The page type and control fields are interpreted as either Base Page or Next Page. Inapplicable Clause 73-specific fields are not involved in the transmission or validity judgment in this implementation. An FLP burst includes a specified clock pulse position and optional data pulse positions. The completion of the last data position is not considered the sole basis for the end of the complete burst, but rather covers all specified pulse positions and related timing.

[0180] The page security boundary lies within the effective silent window between the complete end of the current burst and the start of the next burst. Configuration commits should not extend or shorten this window beyond the limits allowed by the protocol. The number of acknowledgments, the starting point, cross-stage page eligibility, and completion conditions should be set according to the corresponding acknowledgment stage specifications of the adopted Clause 28 version, and the parameters of Clause 73 should not be used as default equivalent conditions.

[0181] Shadow storage, transaction cross-clock delivery, version association, event retention, and storage ownership management reuse the above mechanisms, while field validation, transmission timing, confirmation rules, and arbitration status mapping are configured separately according to Clause 28.

[0182] The aforementioned auto-negotiation page transmission circuitry can be integrated into a serial deserializer (SerDes), Ethernet physical layer transceiver (PHY), switching chip, network interface card (NIC) chip, or on-chip system, working in conjunction with the data transmission path, clock generation circuitry, and auto-negotiation arbitration logic. The circuitry can be delivered in the form of Register Transfer Level (RTL) code, gate-level netlist, hard-core macro, or layout data, and implemented by the integrator as a hardware unit within the chip, or implemented as an independent circuit module. Different delivery methods do not alter the functional relationships of transaction persistence, protocol licensing, boundary commit, context statistics, and event retention.

[0183] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for updating page configuration, applied to a communication transmission circuit including a configuration clock domain and a page transmission clock domain, characterized in that, The method includes: Receive page configuration data, write the page configuration data into the shadow configuration storage unit to form a candidate page transaction, and keep the candidate page transaction stable until a submission result indicating the end of the submission is received; The page security boundary is detected in the page transmission clock domain. The page security boundary is the boundary time or effective window where the specified transmission position of the current complete physical page and the tail timing specified by the protocol have been completed, and the next physical page has not yet referenced its transmission configuration. The boundary time is allowed to coincide with the start edge of the next physical page. Alternatively, the page security boundary is the page transmitter not having any ongoing pages that need to be protected and is allowed to accept the security configuration. The submission permission is reviewed based on the current protocol status, the update category of the candidate page transaction, and the protocol context information. When the submission request is valid, the candidate page transaction meets the capture conditions, the page security boundary is valid, and the submission permission is valid, the candidate page transaction is switched to the active page transaction as a whole, and a configuration version identifier is assigned to the active page transaction. During the sending of a complete physical page, the physical page is generated based on the same selected active page transaction, ensuring that the configuration data used by the physical page is consistent with the configuration version identifier.

2. The method according to claim 1, characterized in that, The candidate page transaction includes a page payload and at least one of the following protocol-related fields: page type, confirmation field, next page field, flip field, random number field, and encoded auxiliary information generator configuration; The values ​​generated page by page in the encoded ancillary information according to the protocol rules are managed separately from the generator configuration. Normal changes in the values ​​generated page by page are not submitted as new configurations, and do not change the configuration version identifier.

3. The method according to claim 1, characterized in that, The method further includes: Record the transaction identifier, update category, and applicable protocol context information of the candidate page transactions; The page completion statistics are updated based on the sending completion event of a valid complete page, the configuration version identifier of the page corresponding to the sending completion event, and its protocol context. The page completion statistics include confirming page counts. After entering the corresponding protocol confirmation stage, a confirmation context is established at the count initialization point specified in the confirmation stage. The confirmation context is then bound to the session identifier, logical page identifier, page qualification rules, and the effective configuration version actually used in the confirmation stage. The confirmation page count is only added to pages that are validly and completely sent, meet the page qualification rules, whose session and logical page match the confirmation context, and whose configuration version conforms to the configuration version association relationship.

4. The method according to claim 1, characterized in that, The shadow configuration storage unit and the storage unit for storing active page transactions consist of at least two storage units, and the overall switching is accomplished by changing the active storage unit selection signal; Candidate storage units that become active storage units are prohibited from being written to by the configuration side while there are still valid read / send dependencies. Storage units that have exited the active state are only allowed to be rewritten to by the configuration side after the read dependencies have ended and ownership has been released.

5. The method according to claim 1, characterized in that, For a candidate page transaction that has already been received, if it meets the data capture conditions but the page security boundary or the commit permission is not yet valid, the currently active page transaction remains unchanged, and the candidate page transaction remains in the pending commit state. Subsequent requests are only accepted if there is an independent available transaction slot; otherwise, a busy status is output or backpressure is applied, and candidate page transactions that have not yet ended are not covered.

6. The method according to claim 1, characterized in that, When the completion of the old page and the commit of the new transaction occur at the same page security boundary, the old page completion event retains the configuration version identifier and protocol context of the old page, the data and configuration version identifier of the new physical page are determined from the same new transaction, and the overall switch is no later than the moment when the next physical page first references the configuration.

7. The method according to claim 1, characterized in that, For page completion events that have already been generated, maintain the transaction identifier or acknowledgment context association of the page completion event until the page completion event is reliably consumed or becomes invalid due to the explicit cancellation of its context, not because the page completion event has changed due to a change in the current activity configuration version.

8. The method according to claim 1, characterized in that, The step of reviewing the submission permission based on the current protocol status and the update category and protocol context information of the candidate page transaction includes at least one of the following: Verify whether the session to which the candidate page transaction belongs is still valid; Verify that the target logic page is compatible with the current exchange object; Verify whether the updates to the confirmation field, confirmation field 2, or the echoed random number field meet the corresponding protocol conditions; Verify whether the retention or flipping of the flipped field is consistent with the logical page progression. Verify that the next page instructions and the status provided on subsequent pages comply with the current agreement requirements; Verify whether the base version or receiving context on which the candidate page transaction depends has expired; The order of submission is determined by the moment when the sending side receives the context invalidation or cancellation event. If cancellation and ordinary submission are established at the same arbitration moment, cancellation takes precedence.

9. The method according to claim 1, characterized in that, The submission request and the submission result are transmitted between the configuration clock domain and the page sending clock domain via a request confirmation handshake. By using data path delay constraints, capture-wait mechanisms, or proven cross-clock structures, multi-bit candidate page transactions can meet establishment and hold requirements before the target domain is captured. The returned transaction identifier, commit status, and configuration version identifier are transmitted in the same manner as the acknowledgment event, and the associated returned data remains stable until reliably captured.

10. A system for updating page configuration, characterized in that, include: The system includes a configuration receiving unit, a shadow configuration storage unit, a submission request unit, a page boundary detection unit, an atomic submission control unit, an active configuration storage unit, a version management unit, and a page sending unit. The configuration receiving unit is used to receive page configuration and update category; The shadow configuration storage unit is used to save complete candidate page transactions and keep the candidate page transactions stable while the request is not finished. The commit request unit is used to generate and maintain commit requests with transaction identifiers; The page boundary detection unit is used to detect page security boundaries based on the status sent by the complete physical page. The atomic commit control unit is used to review the commit permission according to the current protocol state and the session and logical page context applicable to the candidate transaction, and to switch the candidate page transaction as a whole to the active page transaction when the commit request, data capture conditions, page security boundary and commit permission are all satisfied. The activity configuration storage unit is used to save or select currently protected active page transactions; The version management unit is used to assign a configuration version identifier to successfully submitted activity transactions; The page sending unit is used to consistently determine the page snapshot and configuration version identifier from the same selected transaction, and generate a complete physical page based on the page snapshot.