SDSTOR-23030 fix backward compatibility

[Hooper9973]

Background

Commit 4b25833 made CREATE_SHARD_MSG and SEAL_SHARD_MSG log-only: the leader now calls
async_alloc_write with empty data (HS_DATA_INLINED), so no data blocks are allocated for
shard creation. Correspondingly, HeapChunkSelector::select_chunk was changed to
RELEASE_ASSERT(false, ...) since it should never be called for shard messages in the new
design.

During a rolling upgrade, a pre-upgrade SM (old code) still creates shards in the old
format:
it allocates one block, serializes the BlkId into the Raft journal value, and stores the
entry
as HS_DATA_LINKED (value_size = sizeof(MultiBlkId) > 0, blk_count = 1). A
post-upgrade SM (new code) must handle these old-format entries when acting as a
follower.

Root Cause

The crash is a two-phase cascade caused by localize_journal_entry_prepare overwriting
the
Raft journal value with the follower's localized blkid.

Phase 1 — Post-upgrade follower processes old-format entry from pre-upgrade leader

Pre-upgrade leader log:
HS_DATA_LINKED | value_size > 0 | blkid = {X, nblks=1, Y}

Post-upgrade follower, localize_journal_entry_prepare:
entry_blkid.blk_count() = 1 → data_size = 1 × blk_size = 4096
→ get_blk_alloc_hints(data_size=4096)
data_size > 0 branch → returns tombstone_pbas ✓
→ alloc_local_blks: m_local_blkids = [tombstone_pbas{0,0,0}] (blk_count = 0)
→ memcpy(journal_value, local_blkid.serialize(), sizeof(MultiBlkId))
overwrites journal value with tombstone bytes (all zeros)

The post-upgrade follower's local logstore now holds:

HS_DATA_LINKED | value_size = sizeof(MultiBlkId) > 0 | value = tombstone bytes (blk_count =
0)

code and value_size are unchanged, but the value bytes now encode a zero blkid.

Phase 2 — Any subsequent post-upgrade follower receives the already-corrupted entry

Once the corrupted entry exists in any SM's local log, any new post-upgrade follower that
catches
up from that SM receives:

HS_DATA_LINKED | value_size > 0 | value = tombstone bytes

localize_journal_entry_prepare:
entry_blkid.deserialize(...) → tombstone_pbas, blk_count() = 0
data_size = 0 × blk_size = 0

Condition: (code == HS_DATA_LINKED) && (value_size > 0) → TRUE (both hold)
→ takes HS_DATA_LINKED branch
→ applier_create_req(op_code = HS_DATA_LINKED, data_size = 0)
→ init(): has_linked_data() = true, data_size = 0
→ alloc_local_blks(listener, data_size = 0)
→ get_blk_alloc_hints(data_size = 0)
data_size == 0 branch → returns empty blk_alloc_hints{} ← BUG
→ committed_blk_id.has_value() = false
→ falls through to data_service().alloc_blks()
→ VirtualDev::alloc_blks() → HeapChunkSelector::select_chunk()
→ RELEASE_ASSERT(false, "create shard is log only...") → SIGABRT

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 0% with 4 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (stable/v4.x@4b25833). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...ib/homestore_backend/replication_state_machine.cpp	0.00%	4 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@              Coverage Diff               @@
##             stable/v4.x     #450   +/-   ##
==============================================
  Coverage               ?   55.08%           
==============================================
  Files                  ?       36           
  Lines                  ?     5310           
  Branches               ?      671           
==============================================
  Hits                   ?     2925           
  Misses                 ?     2087           
  Partials               ?      298           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.