On Wed, 27 Jul 2022 at 12:55, BALATON Zoltan <bala...@eik.bme.hu> wrote: > > On Wed, 27 Jul 2022, Cédric Le Goater wrote: > > On 7/26/22 20:23, Peter Maydell wrote: > >> This patchset is mainly trying to fix a problem that Coverity spotted > >> in the dcr_write_dma() function in hw/ppc/ppc440_uc.c, where the code > >> is not correctly using the cpu_physical_memory_map() function. > > Likely I did not know how this function works when implementing it and may > have used it wrong but none of the reviews spotted it either. (I may have > used some other DMA device model as an inspiration but don't remember > which.) > > >> While I was fixing that I noticed a second problem in this code, > >> where it doesn't have a fallback for when cpu_physical_memory_map() > >> says "I couldn't map that for you". > > When can that happen? If only in cases when guest gives invalid parameters > then maybe we don't have to bother with that and can let it fail but > having a fallback does not hurt.
Mostly it happens when the thing being DMA'd to or from is not RAM. Ordinarily I wouldn't expect that to be likely, but the DMA device here has a "don't advance the src/destination" option which I assume would be used for things like DMA'ing to or from a device FIFO. Perhaps AmigaOS doesn't in practice do that. > >> and partly because > >> I don't have any documentation of the hardware to tell me how it > >> should behave, so patch 2 in particular has some FIXMEs. I also > > I remember I've found some info on this in some similar SoC that had docs > on-line but don't remember which. Maybe 440EPx/GPx or something like that. > It may not be the same but seems similar enough for AmigaOS to work. I > think the two main sources were PPC440EPx/GRx Embedded Processor User's > Manual and NXP/Freescale Application Note AN2661 Software Migration from > the IBM (AMCC) 440GP to the MPC8540 which seem to be similar to 460EX and > have some info on the DMA controller registers. Thanks, I'll see if I can track those down. -- PMM
