[PATCH v9 00/18] mtd: rawnand: sunxi: fix transfers and support randomized OOB
James Hilliard
james.hilliard1 at gmail.com
Sat Oct 17 12:45:55 PDT 2026
Patches 1-16 address the existing driver independently of the Allwinner
randomized-OOB format in patches 17-18. Please review and apply that prefix
regardless of whether the format extension is accepted. Supported geometries
keep their existing on-flash ECC layout, bad-block-marker representation
and raw callbacks.
Patch 1 centralizes OOB handling without changing behavior. Patches 2-12
fix the controller's ECC step limit, logical-step lengths, read and write
error handling, completion reuse, pattern handling, duplicate program
confirms, small-page read commands and the H6/H616 user-data register limit.
Patches 13-16 then reduce repeated status reads, column changes, chip setup
and length-register accesses.
The pattern fixes precede status caching so correctness can be reviewed
before the register-access optimization.
Here, "user data" means ECC-protected metadata in OOB, not the page's main
data. PIO reuses hardware slot zero for each logical ECC step; DMA can
transfer several steps at once. The patches distinguish logical steps,
hardware slots and packed register offsets, and explain their triggering
conditions independently of the format work that prompted the review.
The IRQ fix retains timeout-only disable_irq() and explains both the
forced-threaded PREEMPT_RT race and the unbounded cleanup/no-priority-
inheritance tradeoff. The small-page fix is limited to read-side
repositioning within existing geometry restrictions. DMA batching preserves
the installed OOB layout rather than capping its allocation.
For devices provisioned with the Allwinner normal-page format, support is
required compatibility with existing flash contents, not a preference for
another layout. Its bad-block marker is randomized: applying the existing
plain-marker compensation can make normal OOB reads and bad-block scans
interpret good blocks as bad. Switching formats is not a transparent
kernel upgrade.
Patch 17 documents the opt-in binding; patch 18 implements the format
through separate OOB callbacks and an attach helper. Its geometry,
page-wide classification and contiguous parity reads stay together.
Without the property, existing format selection and callbacks are unchanged.
Assisted-by: Codex:gpt-6-astra
Signed-off-by: James Hilliard <james.hilliard1 at gmail.com>
Changes in v9:
- Add an independent fix rejecting pages that exceed the controller's ECC
step limit, as reported by Sashiko. Keep it after the initial refactor and
retain the final ECC step size when validating explicit or maximized ECC.
- Add missing Reported-by and report links for the IRQ lifetime,
small-page read and DMA user-data batching fixes.
- Make the existing PIO and DMA OOB-read branches explicit in the initial
refactor, preserving their transfer ordering and error handling.
- Add a cached pattern-ID accessor with status caching, so the Allwinner
format can request pattern IDs without duplicating the shared reader's
register-read conditions.
- Clarify OOB callback indices, buffer and ECC-cache lifetimes, and why
an empty pending-event mask completes the IRQ timeout recheck.
- Explain packed parity placement and byte-based erased-page classification
as Allwinner format compatibility rules, distinct from hardware limits.
- Link to v8: https://patch.msgid.link/20261015-submit-sunxi-nand-vendor-oob-layout-v1-v8-0-8ba4881fbee8@gmail.com
Changes in v8:
- Add a separate fix for the pre-existing zero-pattern overwrite reported
by Sashiko. Preserve synthesized data and OOB in the existing-format
read path without treating zero chunks as erased.
- Place the fix after pattern-ID selection and update the later status
caching conversion to use the same pattern check. Keep the initial
refactor behavior-preserving and the Allwinner format callbacks unchanged.
- Recheck the remaining command/DMA events after draining a timed-out
handler, as suggested by Sebastian Andrzej Siewior. Accept completion
during cleanup and balance IRQ disabling even when the recheck succeeds.
- Link to v7: https://patch.msgid.link/20261014-submit-sunxi-nand-vendor-oob-layout-v1-v7-0-5d3fc6ad2da1@gmail.com
Changes in v7:
- Put a behavior-preserving OOB refactor first, the correctness fixes next,
register-access optimizations after them, and the binding and Allwinner
format last. Keep 16 patches with no external prerequisites.
- Keep logical-step length, read recovery, IRQ lifetime, write failures,
pattern IDs and DMA limits independently reviewable. Correct pattern-ID
selection before introducing per-transfer status caching.
- Update both DMA callers with the protected-OOB helper contract in the
logical-step length fix, before the later batching change.
- Isolate format policy behind its own callbacks and one opt-in attach
path. Fold packed H6/H616 geometry and contiguous parity reads into the
final patch, leaving existing format callbacks unchanged.
- Use scrambled_plain_bbm_oob for the existing scrambled profile and
allwinner_oob for the firmware profile. Keep shared randomizer helpers
generic and name the H6-generation spare predicate after H6.
- Clarify protected-OOB terminology, logical/hardware indices, transfer
failures and the limited small-page read scope. Explain the RT timeout
mechanism and disable_irq() tradeoff discussed with Miquel Raynal.
- Explain installed-format compatibility and that the spare check re-encodes
ECC-decoded protected OOB rather than rereading physical bytes.
- Drop the standalone OOB-helper renaming cleanup and explicit stable Cc
trailers; retain Fixes tags and Conor Dooley's binding Acked-by.
- Rebase onto current MTD nand/next.
- Link to v6: https://patch.msgid.link/20260914-submit-sunxi-nand-vendor-oob-layout-v1-v6-0-2fb084fe0751@gmail.com
Changes in v6:
- retain the bank-sized first DMA batch, but transfer later logical steps
individually to avoid PAGE_OP's absolute internally generated main columns
- keep ECC exception mode enabled for randomized reads on all supported
controllers, using the shared physical-pattern classifier
- classify uniform-pattern steps even when hardware suppresses their ECC
errors; skip their correction counters and preserve decoded buffers unless
the page-wide zero override or vendor spare predicate accepts the page
- Link to v5: https://patch.msgid.link/20260913-submit-sunxi-nand-vendor-oob-layout-v1-v5-0-7d711076a6f7@gmail.com
Changes in v5:
- add a separate small-page command-handling fix: use READ0/READ1/READOOB
with the page address for rereads and normal OOB access, and select PIO
instead of the large-page DMA sequencer
- keep physical reread errors visible, preserve the OOB cursor and retain
the existing ECC geometry requirements, including rejection of 512+16
with controller ECC
- group the small-page fix and the DMA register-bank fix with the opening
fixes, ahead of cleanups, bindings, format support and optimizations
- make the DMA register-bank fix independent of randomized-OOB support
and the later register-access optimizations
- Link to v4: https://patch.msgid.link/20260912-submit-sunxi-nand-vendor-oob-layout-v1-v4-0-4a64bed94229@gmail.com
Changes in v4:
- prepend an independent fix for completion reuse and IRQ timeout races:
initialize the completion once, finish IRQ register updates before
signalling success, and drain timed-out handlers before clearing their
interrupt state
- add a separate fix for aggregate protected-user-data register overflow
by splitting DMA transfers into bounded batches, retaining the existing
OOB layout, ECC offsets and DMA support
- distinguish logical page steps from batch-local hardware slots, retain
page-wide ECC accounting and avoid retrying partially transferred writes
- Link to v3: https://patch.msgid.link/20260909-submit-sunxi-nand-vendor-oob-layout-v1-v3-0-838cb0ba1547@gmail.com
Changes in v3:
- add a prerequisite fix for the logical OOB length used by PIO transfers
- clarify logical ECC steps versus hardware slots and share protected-OOB
register indexing
- select the controller-specific vendor spare-byte erased-page check from
the SoC capabilities, only in randomized-OOB mode and without rereading
the main data
- retain hardware-decoded data and protected OOB on other ECC failures
for bad-block and BBT pattern scans (reported by Miquel Raynal)
- retain a bad marker and ECC failure for all-zero physical pages, and
disable the ECC exception for the vendor format
- share page classification between PIO and DMA, reading complete pages
for randomized-format subpage requests
- propagate OOB read errors and defer randomized-format ECC accounting
until those reads have succeeded
- propagate read/program setup, column-change and buffer-transfer errors,
including extra OOB; stop failed writes and disable ECC and randomization
- discard partial DMA ECC statistics before retrying in PIO, and keep
correction counts separate from successful OOB-transfer status
- select the current hardware step's pattern ID instead of slot zero
- avoid a second program confirm after an OOB-only write
- avoid redundant column changes before writing extra OOB bytes
- reject oversized ECC steps in randomized-OOB mode before the core can
fall back to software ECC
- combine adjacent parity and trailing OOB reads in randomized-OOB mode
- remove duplicate chip setup immediately before core page commands
- program each packed DMA user-data length register once per operation,
and write PIO slot zero directly without read-modify-write
- reuse pattern IDs and packed error counters within a DMA read, while
refreshing the snapshot after every PIO ECC operation
- Link to v2: https://patch.msgid.link/20260904-submit-sunxi-nand-vendor-oob-layout-v1-v2-0-b12074f4aca7@gmail.com
Changes in v2:
- rebase on the current MTD nand/next branch
- retain the merged protected-OOB allocation, BBM reservation and
stack-buffer fixes
- clarify that randomization is part of the normal hardware-ECC page
format while MTD_OPS_RAW continues to expose physical bytes
- explain why a BSP-compatible BBM remains randomized in physical raw data
- document the decoded bad-block and flash-BBT access paths
- reject the firmware OOB format with software or disabled ECC
- document the BSP page-format compatibility contract and the
older-controller format audit
- Link to v1: https://patch.msgid.link/20260810-submit-sunxi-nand-vendor-oob-layout-v1-v1-0-463853a14ad9@gmail.com
---
James Hilliard (18):
mtd: rawnand: sunxi: centralize OOB handling
mtd: rawnand: sunxi: reject pages exceeding the ECC step limit
mtd: rawnand: sunxi: use the logical step's OOB length in PIO
mtd: rawnand: sunxi: propagate page-setup and erased-check errors
mtd: rawnand: sunxi: drain interrupts before reusing the completion
mtd: rawnand: sunxi: stop failed program operations and disable ECC
mtd: rawnand: sunxi: select the pattern ID for the current ECC step
mtd: rawnand: sunxi: preserve synthesized zero-pattern data and OOB
mtd: rawnand: sunxi: propagate buffer and column transfer errors
mtd: rawnand: sunxi: avoid a second program confirm for OOB writes
mtd: rawnand: sunxi: use page reads to reposition small-page NAND
mtd: rawnand: sunxi: bound DMA batches by the user-data register bank
mtd: rawnand: sunxi: reuse ECC status within each transfer
mtd: rawnand: sunxi: avoid redundant column changes for extra OOB
mtd: rawnand: sunxi: avoid duplicate chip setup before page commands
mtd: rawnand: sunxi: reduce user-data length register accesses
dt-bindings: mtd: sunxi: Add randomized OOB flag
mtd: rawnand: sunxi: support the Allwinner randomized OOB format
.../bindings/mtd/allwinner,sun4i-a10-nand.yaml | 10 +
drivers/mtd/nand/raw/sunxi_nand.c | 1578 ++++++++++++++------
2 files changed, 1109 insertions(+), 479 deletions(-)
---
base-commit: 30daaf3e7897b4a8d9b72e5b0c86dd6e2a4d7d85
change-id: 20260810-submit-sunxi-nand-vendor-oob-layout-v1-e3114d10cc9c
Best regards,
--
James Hilliard <james.hilliard1 at gmail.com>
More information about the linux-mtd
mailing list