Hi, just lazy backup of a patch[0] likely to end up in linux soon.
i haven't tested the diff below on openbsd yet, but it's clear that sxiahci is affected by the less-than-expected performance, and i hope this fixes part of it :] -Artturi [0] https://lkml.org/lkml/2019/5/12/161 diff --git a/sys/arch/armv7/sunxi/sxiahci.c b/sys/arch/armv7/sunxi/sxiahci.c index 194005ab642..03557db2dc2 100644 --- a/sys/arch/armv7/sunxi/sxiahci.c +++ b/sys/arch/armv7/sunxi/sxiahci.c @@ -49,8 +49,8 @@ #define SXIAHCI_PWRPIN 40 #define SXIAHCI_PREG_DMA 0x70 -#define SXIAHCI_PREG_DMA_MASK (0xff<<8) -#define SXIAHCI_PREG_DMA_INIT (0x44<<8) +#define SXIAHCI_PREG_DMA_MASK (0xffff) +#define SXIAHCI_PREG_DMA_INIT (0x4433) int sxiahci_match(struct device *, void *, void *); void sxiahci_attach(struct device *, struct device *, void *);