linux-kernel - Re: [RESEND PATCH] vfio/pci: Skip hot reset on Link-Down

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] [day] [month] [year] [list]

Message-Id: <20251224025023.715-1-guojinhui.liam@bytedance.com>
Date: Wed, 24 Dec 2025 10:50:23 +0800
From: "Jinhui Guo" <guojinhui.liam@...edance.com>
To: <alex@...zbot.org>
Cc: <guojinhui.liam@...edance.com>, <kvm@...r.kernel.org>, 
	<linux-kernel@...r.kernel.org>, <linux-pci@...r.kernel.org>
Subject: Re: [RESEND PATCH] vfio/pci: Skip hot reset on Link-Down

On Tue, 23 Dec 2025 15:35:34 -0700, Alex Williamson wrote:
> On Mon, 15 Dec 2025 20:30:29 +0800
> "Jinhui Guo" <guojinhui.liam@...edance.com> wrote:
> 
> > On hot-pluggable ports, simultaneous surprise removal of multiple
> > PCIe endpoints whether by pulling the card, powering it off, or
> > dropping the link can trigger a system deadlock.
> 
> I think this only identifies one small aspect of the problems with
> surprise removal and vfio-pci.  It's not just the release path of the
> device that can trigger a reset, there are various user accessible
> paths as well, ex. the vfio reset and hot-reset ioctls.  I think those
> can trigger this same deadlock.
> 
> Beyond reset, CPU and DMA mappings to the device are still present after
> a surprise removal.  The latter can really only be revoked using the
> new dma-buf support for MMIO regions.
> 
> I think we should take a more comprehensive look at enabling vfio-pci to
> support surprise removal beyond this one case where a cooperative guest
> promptly released the device and encountered a deadlock.
> 
> In doing so, I think we're going to see several more cases where we
> should test for a disconnected device before reset, some of those may
> suggest that PCI-core is actually the better place for the test rather
> than the leaf caller.  Thanks,
> 
> Alex

Hi, Alex

Thank you for your time and helpful suggestions. I will follow up with
deeper research on adding surprise-removal support to vfio-pci.

Best Regards,
Jinhui