Re: DMA-API mapping errors in vmw_pvscsi

[Arvind Kumar <arvindkumar@vmware.com>]

>It's missing checking in two places: from pci_map_single() and on
>scsi_dma_map().  In the latter case, pvscsi_map_buffers needs to return
>an error code and then pvscsi_queue_ring can return -1.

Agreed! And in addition to that pvscsi_queue_ring() also calls pci_map_single()
to map the sense buffer, which should also have the check. And if 
pvscsi_map_buffers() fails then we need to unmap the sense buffer too.

We already have an internal bug filed and we are working on this.

Thanks!
Arvind

----- Original Message -----
From: "James Bottomley" Bottomley@HansenPartnership.com>
To: "Josh Boyer" <jwboyer@fedoraproject.org>
Cc: "Arvind Kumar" <arvindkumar@vmware.com>, "VMware PV-Drivers" , linux-scsi@vger.kernel.org, "Linux-Kernel@Vger. Kernel. Org" vger.kernel.org>
Sent: Monday, March 17, 2014 10:08:10 AM
Subject: Re: DMA-API mapping errors in vmw_pvscsi

On Mon, 2014-03-17 at 12:52 -0400, Josh Boyer wrote:
> Hi All,
> 
> We've had a few reports[1][2] now on the vmw_pvscsi driver throwing
> DMA-API mapping errors when the DMA_API_DEBUG option is enabled.  I've
> included one from a recent 3.14-rc6 kernel below.  Looking at
> vmw_pvscsi.c, I can see pvscsi_map_buffers is missing the calls to
> pci_dma_mapping_error, which is what causes the warnings to be thrown.
> 
> However, I'm not familiar with this driver and I can't see what the
> proper error path should be in this case.  pvscsi_map_buffers is a
> void function and doesn't currently have the ability to return an
> error to the caller.  Even if it did, I'm not sure what the proper
> response to an error should be.  Thoughts?

It's missing checking in two places: from pci_map_single() and on
scsi_dma_map().  In the latter case, pvscsi_map_buffers needs to return
an error code and then pvscsi_queue_ring can return -1.

James