Re: [PATCH v3 1/5] hw/ssi: Add Synopsys DWC SSI standard PIO controller
Kangjie Huang <[email protected]>
| Newsgroups | org.nongnu.qemu-riscv,org.nongnu.qemu-devel |
|---|---|
| Message-ID | <CAAW0U4Z4hSpbC62xE78PJjJPQKNh+7XxF3-Q14QzymELedxhSg@mail.gmail.com> |
Portia Stephens <[email protected]> 于2026年8月21日周五 09:11写道: > > I don't think this will work for frame != 8-bit with the m25p80 flash > model used by k230. ssi_transfer() takes a 32-bit value but it then > calls m25p80_transfer8() which will just transfer the bottom 8-bits and > drop the rest. > > I don't know if this is an issue with your code or with the ssi_transfer > function though. It seems like ssi_transfer intends to accept 32-bit > values. > Hi Portia, Thanks for the review. On SPI_FRF: you are right, the explicit clearing in dw_ssi_write() is redundant because dw_ssi_write_masked() already drops the bits that are not in the writable mask. I will remove it and add a LOG_UNIMP warning when a non-zero SPI_FRF value is written. On the frame width: ssi_transfer() takes a uint32_t word. The issue is that the DWC model uses CTRLR0.DFS (4..32 bits) to define each transmitted frame, while SPI NOR is an 8-bit consumer, so the two are aligned only when DFS is 8. The K230 SDK SPI-NOR paths use 8-bit frames, so this case is not exposed by the current K230 use case. Maybe either add a DFS != 8 restriction with a LOG_UNIMP warning, or have the DWC model split byte-aligned 16/24/32-bit frames into multiple 8-bit ssi_transfer() calls? Thanks, Kangjie