On Friday 04 March 2016 15:43:56 Alexandre Courbot wrote: > > > > Yeah, I'm not too sure what is the point of setting the fake mask to > > be honest, but you are definitely right that it is a contradiction to > > call a DMA function on a device that is not DMA-capable. > > Ah, I finally got it - we are just setting it to the *address* of > host->dma_mask so the device's DMA mask does not end up being a NULL > pointer. > > That actually changes things a bit. DMA-capable devices are clearly > expected to set the mask themselves, but the only one to do it is > host/mtk-sd.c. And dma_set_mask() is only called in dw_mmc and > sdhci-acpi's enable_dma callback. > > This means most DMA-capable devices (including Tegra, but not only) > are simply left with no DMA setup at all. > > Probably we can detect when the host did not do any DMA setup in the > probe function and attempt some sane defaults depending on what the > hardware says it is capable of? When the host leaves an empty DMA mask, the intended meaning is that the device is not on a DMA capable bus, so if we run into that case, we should instead fix the creation of the device rather than the driver that looks at the data. Arnd -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
