Repository navigation
DMA6: mask CHCR writes, start the OTC on bits 24 and 28 - #2475
Conversation
Only bits 24, 28 and 30 of D6_CHCR are writable and bit 1 reads 1. The OTC runs when 24 and 28 are both set, whatever the other bits; 24 alone stays busy. MADR and BCR are not written back. Measured on SCPH-1000, SCPH-5501 and SCPH-9002; tests in nugget tests/dma.
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Merge Risk: ⚪ Minimal · up to
|
| Check name | Status | Explanation | Resolution |
|---|---|---|---|
| Docstring Coverage |  |
Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 4 functions across 2 files. (2 skipped: 2 … | Write docstrings for the functions missing them to satisfy the coverage threshold. |
-
Check name Status Explanation Title check The title names the main change: DMA6 CHCR writes are masked, and the OTC (ordering table clear) starts on bits 24 and 28. The raw summary confirms that src/core/psxdma.cc, src/core/psxhw.h, and the n… Description check The description explains the DMA6 bug: dma6() ran only for chcr == 0x11000002. It names the writable CHCR bits (24, 28, 30) and the OTC start condition (bits 24 and 28). It also covers the removed MAD… Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Full details: Title check
-
Explanation
The title names the main change: DMA6 CHCR writes are masked, and the OTC (ordering table clear) starts on bits 24 and 28. The raw summary confirms that src/core/psxdma.cc, src/core/psxhw.h, and the nugget pin all match this title.
Full details: Description check
-
Explanation
The description explains the DMA6 bug: dma6() ran only for chcr == 0x11000002. It names the writable CHCR bits (24, 28, 30) and the OTC start condition (bits 24 and 28). It also covers the removed MADR/BCR writeback and the scope limit to DMA6. These match the raw summary's changes in psxdma.cc and psxhw.h and the nugget submodule pin.
-
Full details: Docstring Coverage
-
Explanation
Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 4 functions across 2 files. (2 skipped: 2 unsupported.)
Resolution
Write docstrings for the functions missing them to satisfy the coverage threshold.
✨ Finishing Touches
-
🧪 Generate unit tests (beta)
-
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
dma6()only ran onchcr == 0x11000002, so0x11000000(what psx-spx documents as enough) did nothing, and any other value with bit 24 set completed instantly with an IRQ and no OT written. On SCPH-1000, SCPH-5501 and SCPH-9002 the writable bits are 24, 28 and 30, bit 1 reads 1, the OTC runs when 24 and 28 are both set, and 24 alone stays busy forever. CHCR writes are now masked to that, and DMA6 skips the generic MADR/BCR writeback, which had been adding 0x200000 to MADR and zeroing BCR.DMA6 only on purpose; the other channels' CHCR masks and DREQ behaviour are a bigger rework of the DMA dispatch. Not modelled either: a transfer latched with ch6 disabled starts when DPCR enables the channel (on hardware it does; that test stays
CESTER_MAYBE_TEST).The pin points at nugget
dma6-otc-strict, which turns the newtests/dmacases on underPCSX_TESTS. pcsx-redux/nugget#272 merges them as hardware-only, since nugget CI runs against the current dev build.