linux-kernel - Re: [PATCH v3 6/6] hisi_acc_vfio_pci: Add support for VFIO live migration

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 27 Sep 2021 15:32:06 -0300
From:   Jason Gunthorpe <jgg@...dia.com>
To:     Leon Romanovsky <leonro@...dia.com>
Cc:     Shameerali Kolothum Thodi <shameerali.kolothum.thodi@...wei.com>,
        "kvm@...r.kernel.org" <kvm@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-crypto@...r.kernel.org" <linux-crypto@...r.kernel.org>,
        "alex.williamson@...hat.com" <alex.williamson@...hat.com>,
        "mgurtovoy@...dia.com" <mgurtovoy@...dia.com>,
        liulongfang <liulongfang@...wei.com>,
        "Zengtao (B)" <prime.zeng@...ilicon.com>,
        Jonathan Cameron <jonathan.cameron@...wei.com>,
        "Wangzhou (B)" <wangzhou1@...ilicon.com>
Subject: Re: [PATCH v3 6/6] hisi_acc_vfio_pci: Add support for VFIO live
 migration

On Mon, Sep 27, 2021 at 09:30:23PM +0300, Leon Romanovsky wrote:
> On Mon, Sep 27, 2021 at 03:22:24PM -0300, Jason Gunthorpe wrote:
> > On Mon, Sep 27, 2021 at 09:17:19PM +0300, Leon Romanovsky wrote:
> > 
> > > > The point is to all out a different locking regime that relies on the
> > > > sriov enable/disable removing the VF struct devices
> > > 
> > > You can't avoid trylock, because this pci_get_sriov_pf_devdata() will be
> > > called in VF where it already holds lock, so attempt to take PF lock
> > > will cause to deadlock.
> > 
> > My whole point is we cannot use the device_lock *at all* and a
> > pci_get_sriov_pf_devdata() would not have it.
> 
> Right
> 
> > 
> > Instead it would have some test to confirm that the 'current' struct
> > device is a VF of the 'target' struct device and thus the drvdata
> > must be valid so long as the 'current' struct device hasn't completed
> > remove.
> 
> I'm curious to see how can you implement it without holding VF lock.

The VF lock is fine, it is the PF lock you can't take

And you don't need the VF lock or the PF lock if it is called in a
context that blocks VF remove() from completing - which describes an
entire VFIO driver.

Jason