feat(sbp2): complete SBP-2 login, command transport, and INQUIRY flow#14
Closed
gly11 wants to merge 49 commits into
Closed
feat(sbp2): complete SBP-2 login, command transport, and INQUIRY flow#14gly11 wants to merge 49 commits into
gly11 wants to merge 49 commits into
Conversation
…or controller compatibility
# Conflicts: # ASFW/ASFWDriverConnector.swift # ASFWDriver/Async/AsyncSubsystem.cpp # ASFWDriver/Controller/ControllerCore.cpp # ASFWDriver/Controller/ControllerCore.hpp # ASFWDriver/Service/DriverContext.cpp # ASFWDriver/UserClient/Core/ASFWDriverUserClient.cpp # ASFWDriver/UserClient/Core/ASFWDriverUserClient.iig # tests/CMakeLists.txt
Call DriverWiring::EnsureSbp2Deps() after FCPResponseRouter setup to initialize SBP2 AddressSpaceManager and register PacketRouter handlers for quadlet/block read/write requests (tCode 0x0/0x1/0x4/0x5). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix ARPacketView→BlockWriteRequestView conversion in block write handler (was passing raw ARPacketView to RouteBlockWrite) - Fix FCPResponseRouter constructor: use ControllerCore::Bus() instead of GetGenerationTracker() (wrong type, missing IFireWireBusInfo) - Add missing IFireWireBus.hpp include for derived-to-base conversion - Add DriverConnector+SBP2.swift with 4 Swift API methods for SBP2 address space management (allocate, deallocate, read, write) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Per Linux firewire_ohci and Apple IOFireWireController, the contender bit is always set during init. This is required for proper bus management (IRM election, cycle-start generation), especially in 2-node topologies. Delegation policy remains controlled by the experimental flag. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Restore explicit PHY bus reset in EnableInterruptsAndStartBus() that was previously removed — the auto reset from linkEnable alone may not reliably activate the Config ROM shadow on all controllers. Add SBP-2 protocol layer ported from Apple IOFireWireSBP2: - SBP2WireFormats.hpp: big-endian wire structures (LoginORB, LoginResponse, ReconnectORB, LogoutORB, NormalORB, PageTableEntry, StatusBlock, TaskManagementORB) plus constants and helper functions - SBP2LoginSession: full login/reconnect/logout state machine with retry logic (up to 32 retries), bus reset handling, and address space allocation for ORB/response/status buffers Add SBP2 logging category. Co-Authored-By: zcode <noreply@zcode.dev>
…r infrastructure - Fix Reconnect/Logout ORBs using their own meta instead of statusBlockMeta_ for the status FIFO address. Device would write status blocks to wrong addr. - Fix kExclusiveLogin bit: 0x2000 (bit 13 per SBP-2 Table 14), not 0x0020. - Add RemoteWriteCallback to AddressSpaceManager so SBP2LoginSession gets notified when device writes status blocks to its address space. - Fix login response read timing: OnLoginWriteComplete now only ACKs the management agent write and waits for the status block callback instead of immediately reading the login response (which is empty at that point). New OnStatusBlockRemoteWrite dispatches to state-specific handlers. - Wire SubmitDelayedCallback to IODispatchQueue (DispatchAsync + IOSleep pattern matching BusResetCoordinator), replacing the no-op placeholder.
…cture Add Normal Command ORB (SBP2CommandORB) with prepare-for-execution, page table binding, and chain management. Port SBP2PageTable from Apple IOFireWireSBP2ORB scatter-gather logic with direct-address optimization for single-PTE cases. Extend SBP2LoginSession with SubmitORB, fetch agent write, doorbell ring, and ORB completion via status block dispatch.
Delete .bak/.bak2 documentation files and fixlog.sh that were left over from bus reset development. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
IEEE 1394 generation values range 0–16383 (14 bits), which overflows uint8_t. Use the 16-bit accessor to avoid truncation after generation 128+. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…t short/long bus reset OHCI hardware auto-clears cycleMaster when cycleTooLong fires (OHCI §6.2.1). Per Linux irq_handler, re-assert it immediately to keep cycle-start packets flowing — devices like Nikon SAA7356HL depend on them for firmware download. Also re-assert in StepRearming (per Linux bus_reset_work) and implement HardwareInterface::InitiateBusReset with proper short (IEEE 1394a PHY reg 5 SBR) and long (PHY reg 1 IBR) reset paths. Add ControllerCore diagnostic accessors and BusResetCoordinator:: RequestUserReset for UserClient-initiated resets. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
… reset UserClient methods New handler exposes three methods (selectors 50–52): - GetBusStateDiagnostics: OHCI registers, topology, PHY, FSM state - ReadPhyRegister: direct PHY register read (addr 0–7) - InitiateBusReset: user-initiated short/long bus reset Includes BusStateWire packed struct (64 bytes) and full lifecycle wiring in UserClientRuntimeState. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Co-Authored-By: zcode <noreply@zcode.dev>
…, fix CBA offsets - Add SBP2ManagementORB for AbortTask/AbortTaskSet/LogicalUnitReset/TargetReset with per-ORB status FIFO, timeout via IODispatchQueue, BE wire conversion - Fix CBA register offsets to match Apple (AgentReset=+0x04, FetchAgent=+0x08, Doorbell=+0x10, UnsolicitedStatusEnable=+0x14) - Add ResetFetchAgent() for error recovery via CBA quadlet write - Add EnableUnsolicitedStatus() with deferred-enable after login/reconnect - Compute CBA-derived addresses once at login, not per-ORB - Unsolicited status auto-re-enables and is consumed after reconnect
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
SBP2CommandORB::inProgress_ and SBP2ManagementORB::inProgress_/timerActive_ were plain bools accessed from both the IODispatchQueue timer thread and caller/destructor threads. Use std::atomic<bool> with relaxed ordering to avoid undefined behavior while the existing generation counter handles logical synchronization. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Added SubmitSBP2Command and GetSBP2CommandResult methods to SBP2Handler for command submission and result retrieval. - Introduced SBP2CommandWireFormats.hpp to define wire formats for SBP-2 command requests and results. - Extended DeviceDiscoveryWireFormats to include management agent offset, LUN, unit characteristics, and fast start fields. - Created SBP2SessionRegistryTests to validate command submission and response handling. - Updated documentation to reflect the current state and goals of the SBP-2 scanner integration.
…r signature CompletionCallback was changed to void(int, uint8_t) but test lambdas still used the old void(int) signature, causing build failures. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…d bus reset timing SBP-2 Management_Agent_Offset now uses the combined IEEE 1212 CSR key 0x54 (keyType=1, keyId=0x14) matching real devices like Nikon, with 0x38 retained as a legacy fallback. AR DMA test helpers now correctly use little-endian byte order matching OHCI hardware. Bus reset discovery always applies a 100ms baseline delay matching Apple ScanBus behavior. Cycle master is disabled by default; explicit PHY bus reset removed in favor of linkEnable auto reset. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- fix OHCI AR DMA block data-length decoding and copy misaligned request payloads into aligned scratch buffers - encode local-bus 16-bit node IDs in SBP-2 login, management, command, and page-table ORB embedded addresses - release owner-bound SBP-2 address-space resources on user client Stop/free to avoid stale allocation conflicts - add host coverage for address-space writes, FCP parsing, and SBP-2 ORB layout/address encoding, and sync the roadmap with current verification status
…ry metrics and automated recovery resets
…r better diagnostics
Contributor
Author
|
I split this larger PR into smaller stacked PRs so each layer can be reviewed independently:
I also prepared a few follow-up slices locally, but have not pushed them yet pending review:
Local debug UI, diagnostic handlers, install helpers, scripts, and documentation experiments are intentionally excluded from the remote PR stack. |
Owner
Thanks! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This PR brings up the complete SBP-2 initiator stack — login, command transport, and end-to-end INQUIRY — together with the async TX/RX and bus robustness fixes that were developed alongside it.
SBP-2 Protocol
DriverConnector+SBP2,SBP2DebugView)Async TX/RX Foundations
Bus & Diagnostics
Tests
Fixes
Why a single PR
The async infrastructure, bus fixes, and SBP-2 protocol were developed iteratively on one branch — several commits touch both layers. The async response sender and bus robustness work were purpose-built for SBP-2 bring-up rather than standalone improvements. Splitting would require manually reconstructing intermediate states of ~15 shared files without adding value for the reviewer.
If you'd prefer the async/bus foundations and SBP-2 protocol in separate PRs, happy to split — just let me know.
Verification