Re: [bug] __blk_mq_run_hw_queue suspicious rcu usage

[Christoph Hellwig]

On Wed, Nov 27, 2019 at 02:11:28PM -0800, David Rientjes wrote:
> So we're left with making dma_pool_alloc(GFP_ATOMIC) actually be atomic 
> even when the DMA needs to be unencrypted for SEV.  Christoph's suggestion 
> was to wire up dmapool in kernel/dma/remap.c for this.  Is that necessary 
> to be done for all devices that need to do dma_pool_alloc(GFP_ATOMIC) or 
> can we do it within the DMA API itself so it's transparent to the driver?

It needs to be transparent to the driver.  Lots of drivers use GFP_ATOMIC
dma allocations, and all of them are broken on SEV setups currently.