There's a bug in the current implementation of dma_get_required_mask() where it ands the returned mask with the current device mask. This rather defeats the purpose if you're using the call to determine what your mask should be (since you will at that time have the default DMA_32BIT_MASK). This bug results in any driver that uses this function *always* getting a 32 bit mask, which is wrong. Fix by removing the and with dev->dma_mask. Signed-off-by: James Bottomley <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx> --- This is a pretty nasty bug which can cause mysterious slow downs and panics on >4GB machines (we've just had one on an aic79xx production system). It probably also needs to be backported as far as it will go. James diff --git a/drivers/base/platform.c b/drivers/base/platform.c index efaf282..911ec60 100644 --- a/drivers/base/platform.c +++ b/drivers/base/platform.c @@ -648,7 +648,7 @@ u64 dma_get_required_mask(struct device *dev) high_totalram += high_totalram - 1; mask = (((u64)high_totalram) << 32) + 0xffffffff; } - return mask & *dev->dma_mask; + return mask; } EXPORT_SYMBOL_GPL(dma_get_required_mask); #endif -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
