Re: [PATCH 3/6] RDMA/core: remove use of dma_virt_ops

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Nov 05, 2020 at 01:52:53PM -0400, Jason Gunthorpe wrote:
> On Thu, Nov 05, 2020 at 08:42:02AM +0100, Christoph Hellwig wrote:
> > @@ -1341,7 +1322,14 @@ int ib_register_device(struct ib_device *device, const char *name,
> >  	if (ret)
> >  		return ret;
> >  
> > -	setup_dma_device(device, dma_device);
> > +	/*
> > +	 * If the caller does not provide a DMA capable device then the IB core
> > +	 * will set up ib_sge and scatterlist structures that stash the kernel
> > +	 * virtual address into the address field.
> > +	 */
> > +	device->dma_device = dma_device;
> > +	WARN_ON(dma_device && !dma_device->dma_parms);
> 
> I noticed there were a couple of places expecting dma_device to be set
> to !NULL:
> 
> drivers/infiniband/core/umem.c:                 dma_get_max_seg_size(device->dma_device), sg, npages,

This needs to use ib_dma_max_seg_size.

> drivers/nvme/host/rdma.c:       ctrl->ctrl.numa_node = dev_to_node(ctrl->device->dev->dma_device);

> Don't know much about NUMA, but do you think the ib device setup
> should autocopy the numa node from the dma_device to the ib_device and
> this usage should just refer to the ib_device?

IMHO we could add a ib_device_get_numa_node API or something like that,
which uses dev_to_node on the DMA device is present and otherwise returns
-1.  If needed we can refine that later.

> net/rds/ib.c:                                              device->dma_device,
> 
> No sure what to do about RDS..

Yikes, this is completely broken.  We either need a wrapper for the
dma_pool API, or get rid of it.  Let me dig into that.



[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux