Discovery sessions have nil Target and nil ITNexus. The cleanup
path crashed with nil pointer dereference when trying to remove
ITNexus or access Target.Sessions for discovery sessions.
Guard against nil Target (just release TSIH) and nil ITNexus.
Also add nil safety to clearHostIP helper.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Restore CurrentHostIP cleanup on connection close (needed for
blockMultipleHostLogin feature). Extract to clearHostIP helper
to reduce duplication.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When a new login arrives while the previous connection's async cleanup
is still running, LookupISCSISession finds the stale session but
LookupConnection returns nil (old connection already closed). This
caused a nil pointer dereference when accessing existConn.session
during session reinstatement.
Fix by checking existConn != nil before reinstatement. If the old
connection is already gone, unbind the stale session and create
a fresh one instead.
Also add sync.Once to removeConnectionFromSession to prevent
concurrent goroutine and main-path cleanup from racing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move session cleanup out of iscsiExecLogout() and keep it only in
the CONN_STATE_CLOSE handler. The logout response must be fully sent
before the session is removed; cleaning up during logout causes the
daemon to hang because subsequent operations reference a nil session.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When an iSCSI connection closes or a logout occurs, the associated
session was never removed from the target's Sessions map, and the
TSIH was never released back to the bitmap allocator. This caused
session and TSIH leaks over repeated connect/disconnect cycles.
Changes:
- Add removeConnectionFromSession() to properly clean up session
when its last connection is closed
- Call session cleanup from handler() on CONN_STATE_CLOSE
- Convert iscsiExecLogout to a method and add session cleanup on logout
- Release TSIH in UnBindISCSISession to prevent TSIH bitmap exhaustion
Fixes#42
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a new backend store that enables iSCSI targets backed by
S3-compatible object storage (AWS S3, MinIO, Ceph RGW, etc.).
The implementation uses a chunked storage strategy where the virtual
block device is divided into fixed-size chunks (default 4 MiB), each
stored as an independent S3 object. This enables efficient random
read/write access on top of object storage.
Key features:
- Chunked storage with configurable chunk size
- Sparse device support (unwritten chunks treated as zeros)
- Concurrent multi-chunk I/O via errgroup
- Per-chunk locking for safe read-modify-write
- AWS SDK v2 with default credential chain
- In-process gofakes3 test server (no Docker needed)
- 12 unit tests + 2 integration tests
Also updates CI workflow to run S3 backend tests and updates
README with S3 backend documentation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix target delete URL path mismatch (/targets/ -> /target/)
- Implement target create/delete server handlers with proper validation
- Add DeleteTarget method with force flag and mutex locking to SCSITargetService
- Implement full LU management: create/list/delete through CLI, client, and server
- Add TPGT list command to show target portal group tags
- Add unit tests for target/LU router handlers and SCSI service
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add comprehensive benchmark suite (io_bench_test.go):
- BenchmarkEndToEndRead/Write: full SCSI stack (512B to 256KB)
- BenchmarkEndToEndReadParallel/WriteParallel: concurrent IO
- BenchmarkFileBackingStoreRead/Write: isolated backing store
pprof-guided optimizations:
- Guard hot-path log.Debugf with log.GetLevel() check in scsi.go,
sbc.go, backingstore.go — eliminates 22% CPU overhead from logrus
Entry allocation even when debug logging is disabled
- Add FileBackingStore.ReadAt for zero-copy reads directly into
caller's buffer, bypassing Read()'s per-call make([]byte, tl)
- Use ReadAt via interface assertion in bsPerformCommand to read
directly into InSDBBuffer, eliminating allocation + copy
Results (256KB reads): +42% throughput, allocs reduced from 10 to 5
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Read path: eliminate redundant allocation in bsPerformCommand - remove
the pre-allocation before bs.Read() and the append loop for zero-fill,
use direct copy and in-place zero-fill instead
- parseHeader: use command pool (getCommand) instead of direct allocation,
reducing GC pressure on the hot path
- Unmap: use a shared 1MB zero buffer instead of allocating per-descriptor,
dramatically reducing allocations for large unmap operations
- Network I/O: add 256KB bufio.Writer to iSCSI connections, batching
small PDU writes into fewer syscalls. Flush after txHandler completes
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
gorilla/mux explicitly rejects capturing groups () in route regexps,
only non-capturing groups (?:) are allowed. The original regex was
missing the ? to make -dirty optional.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix versionMatcher regex: gorilla/mux does not support (?:) syntax,
and -dirty suffix was required instead of optional
- Replace unsafe.Pointer LUN casts with binary.LittleEndian.Uint64
in sbc.go, scsi.go, and target.go
- Implement graceful HTTP server shutdown with 5s timeout using
srv.Shutdown() instead of raw listener close
- Replace golang.org/x/net/context with standard library context
- Respect existing req.Cancel value in canceler to avoid overwriting
- Add early context cancellation check in Do() to fail fast
Based on review of PR #120 by @orzhang, with fixes applied.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The UNMAP command was a no-op in all backing stores, causing unmapped
blocks to retain stale data instead of returning zeros per SCSI spec.
- Implement Unmap in FileBackingStore to zero out unmapped blocks
- Implement Unmap in IOUringBackingStore to zero out unmapped blocks
- Enable Unmap in RemBackingStore (was commented out)
- Change UnmapBlockDescriptor.TL from uint32 to uint64 to prevent
integer overflow when converting block count to byte length with
large block shifts
Fixes#119
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix nil pointer dereference in BindISCSISession when existSess is nil
- Fix reversed logic in SPCLuOffline/SPCLuOnline (Online flag was swapped)
- Use negotiated MaxXmitDataSegmentLength for response PDU segmentation (issue #41)
- Fix debug log incorrectly using Warn level in SBCGetLbaStatus
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Convert constant to var so that it can be configured from backend
- Add options to disable persistent reservation and ORWrite16 commands
Signed-off-by: Utkarsh Mani Tripathi <utkarsh.tripathi@mayadata.io>
- Respond with failure for unsupported task management commands
- Increase MaxRecvDataSegmentLength to 65536, to match FirstBurstLength
Signed-off-by: Utkarsh Mani Tripathi <utkarsh.tripathi@mayadata.io>
- Gotgt was crashing for few opcodes (sg_inq (sg3-utils)) because InSDBBuffer
was not initialized for such opcodes (Need help)
Signed-off-by: Utkarsh Mani Tripathi <utkarsh.tripathi@mayadata.io>
