linux-kernel - Re: [PATCH] PCI: Mark LSI FW643 to avoid bus reset

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]

Message-ID: <20240226210228.GA199890@bhelgaas>
Date: Mon, 26 Feb 2024 15:02:28 -0600
From: Bjorn Helgaas <helgaas@...nel.org>
To: Edmund Raile <edmund.raile@...ton.me>
Cc: bhelgaas@...gle.com, linux-pci@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	Alex Williamson <alex.williamson@...hat.com>
Subject: Re: [PATCH] PCI: Mark LSI FW643 to avoid bus reset

[+cc Alex]

On Mon, Feb 26, 2024 at 10:25:12AM +0000, Edmund Raile wrote:
> Using LSI / Agere FW643 with vfio-pci will issue an FLreset, causing
> a broken link only recoverable by removing power (power-off /
> suspend + rescan). Prevent this bus reset.
> With this change, the device can be assigned to VMs with VFIO.

I don't quite understand the commit log.  It says Function Level
Resets cause a problem, but the patch sets PCI_DEV_FLAGS_NO_BUS_RESET,
which prevents Secondary Bus Resets, not Function Level Resets.

By default, I think we try reset methods in order of
pci_reset_fn_methods[], so the fact that we get to the end and 
try pci_reset_bus_function() (where PCI_DEV_FLAGS_NO_BUS_RESET is
tested) suggests that we weren't able to use FLR or PM resets.

IIUC, vfio-pci resets devices to prevent leaking state from one VM to
another.  We might be willing to accept the downside of allowing that
leakage, but I think it's worth mentioning that in the commit log.

Add blank lines between paragraphs (also in the comment below), or
rewrap into a single paragraph.

> Signed-off-by: Edmund Raile <edmund.raile@...ton.me>
> ---
> Usefulness:
> The LSI FW643 PCIe->FireWire 800 interface may be EOL but it is
> the only one that does not use a PCIe->PCI bridge.
> It was used in the following Apple machines:
> MacBookPro10,1
> MacBookPro9,2
> MacBookPro6,2
> MacBookPro5,1
> Macmini6,1
> Macmini3,1
> iMac12,2
> iMac9,1
> iMac8,1
> It is reliable and enables FireWire audio interfaces to be used
> on modern machines.
> Virtualization allows for flexible access to professional audio
> software.
> 
> Implementation:
> PCI_VENDOR_ID_ATT was reused as they are identical and I am
> uncertain it is correct to add another ID for LSI to pci_ids.h.
>  drivers/pci/quirks.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index d797df6e5f3e..a6747e1b86da 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -3765,6 +3765,15 @@ DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATHEROS, 0x003e, quirk_no_bus_reset);
>   */
>  DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_CAVIUM, 0xa100, quirk_no_bus_reset);
>  
> +/*
> + * Using LSI / Agere FW643 with vfio-pci will issue an FLreset, causing
> + * a broken link only recoverable by removing power (power-off /
> + * suspend + rescan). Prevent this bus reset.
> + * With this change, the device can be assigned to VMs with VFIO.
> + */
> +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATT, 0x5900, quirk_no_bus_reset);
> +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATT, 0x5901, quirk_no_bus_reset);
> +
>  /*
>   * Some TI KeyStone C667X devices do not support bus/hot reset.  The PCIESS
>   * automatically disables LTSSM when Secondary Bus Reset is received and
> -- 
> 2.43.0
> 
>