Re: [PATCH v11 12/26] gunyah: vm_mgr: Add/remove user memory regions

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

 



On Wed, Apr 12, 2023 at 01:48:07PM -0700, Elliot Berman wrote:
> 
> 
> On 4/11/2023 2:19 PM, Will Deacon wrote:
> > On Tue, Apr 11, 2023 at 01:34:34PM -0700, Elliot Berman wrote:
> > > On 3/24/2023 11:37 AM, Will Deacon wrote:
> > > > On Fri, Mar 03, 2023 at 05:06:18PM -0800, Elliot Berman wrote:
> > > > > +
> > > > > +	pinned = pin_user_pages_fast(region->userspace_addr, mapping->npages,
> > > > > +					FOLL_WRITE | FOLL_LONGTERM, mapping->pages);
> > > > > +	if (pinned < 0) {
> > > > > +		ret = pinned;
> > > > > +		mapping->npages = 0; /* update npages for reclaim */
> > > > > +		goto reclaim;
> > > > > +	} else if (pinned != mapping->npages) {
> > > > > +		ret = -EFAULT;
> > > > > +		mapping->npages = pinned; /* update npages for reclaim */
> > > > > +		goto reclaim;
> > > > > +	}
> > > > 
> > > > I think Fuad mentioned this on an older version of these patches, but it
> > > > looks like you're failing to account for the pinned memory here which is
> > > > a security issue depending on who is able to issue the ioctl() calling
> > > > into here.
> > > > 
> > > > Specifically, I'm thinking that your kXalloc() calls should be using
> > > > GFP_KERNEL_ACCOUNT in this function and also that you should be calling
> > > > account_locked_vm() for the pages being pinned.
> > > > 
> > > 
> > > Added the accounting for the v12.
> > > 
> > > > Finally, what happens if userspace passes in a file mapping?
> > > 
> > > Userspace will get EBADADDR (-14) back when trying to launch the VM
> > > (pin_user_pages_fast returns this as you might have been expecting). We
> > > haven't yet had any need to support file-backed mappings.
> > 
> > Hmm, no, that's actually surprising to me. I'd have thought GUP would
> > happily pin page-cache pages for file mappings, so I'm intrigued as to
> > which FOLL_ flag is causing you to get an error code back. Can you
> > enlighten me on where the failure originates, please?
> 
> Ah this ended up being an error on my part. Userspace was opening the file
> as RO and Gunyah driver will unconditionally add FOLL_WRITE as part of the
> gup flags. I got the flags aligned and seemed to be able to boot the VM ok
> and it works as expected.

I suspect you can run into latent filesystem corruption issues in this case,
as the VM can dirty pages without the filesystem knowing. That's why we
restricted anonymous memory with pKVM for now.

Will



[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux