[PATCH v19 00/12] ALSA: virtio: fix capture accounting

[Joonho Lee <mynameisjune111@gmail.com>]

Hi Takashi,

This cumulative resend keeps the twelve-patch series on the same base,
a4311e9447bbb219044db38897f95be112f46a8a. It covers PCM stop-threshold
sampling, locked partial-progress reporting, the optional reset operation,
VirtIO ACK error propagation and recovery, PCM and control DMA ownership,
signal-safe teardown, staged capture lifecycle and work prioritization, and
capture parameter documentation.

Relative to v18, patch 7 validates ordinary indirect capture appl_ptr
accounting before calling the indirect transfer helper. It bounds ACK
consumption by completed, unread capture bytes and rejects positive
advances before RX ring seeding. Active over-budget advances follow the
existing XRUN path; invalid PAUSED advances roll back without ending
PAUSED.

Validation:

- Cumulative ARM64 stages 04–12 passed W=1 builds with KASAN, lockdep,
  and OSS PCM enabled. Stage 04 built Image and modules; stages 05–12
  rebuilt modules incrementally.
- A bounded candidate-only raw SYNC_PTR audit rejected unseeded PREPARED
  and over-budget RUNNING advances before indirect accounting. The PCM
  pointer rolled back, and prepare/start/read recovery passed. Valid
  partial ACKs also passed. No vulnerable baseline sequence was rerun.
- Candidate state runs passed valid subperiod ACKs, PAUSED rollback and
  resume, DRAINING tail consumption and over-budget rollback, and raw ACK
  before and after PAUSED RESET.
- A slow-reader QEMU run kept the unread capture slot unchanged while
  appl_ptr was held, then recycled it once after one period was consumed.
- Four final-byte no-hook QEMU runs covered ordinary and staged capture
  through RW and MMAP, including reset and reopen. The 14-sample oracle
  passed across ring wrap with no KASAN or lockdep diagnostics.
- v18 MMIO rollback, timeout-width, and PCI remove/reprobe checks are
  inherited for the unchanged transport and control-message paths.

The bounded ownership trace did not reproduce an in-flight duplicate
virtqueue add or message-length overflow; neither is claimed as reproduced.
Numeric ALSA runtime->boundary wrap, physical Virtualization.framework
devices, and KCSAN were not tested.

Previous version:
https://lore.kernel.org/r/20261106000644.20460-1-mynameisjune111@gmail.com/

AI assistance was used for analysis, implementation, and test preparation.
Assisted-by: LLM

Thanks,
Joonho

base-commit: a4311e9447bbb219044db38897f95be112f46a8a