linux-kernel - Re: [PATCH v2 2/4] PCI/ASPM: Add pcie_aspm_remove

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: <alpine.DEB.2.21.2511130354500.25436@angie.orcam.me.uk>
Date: Thu, 13 Nov 2025 04:02:39 +0000 (GMT)
From: "Maciej W. Rozycki" <macro@...am.me.uk>
To: Bjorn Helgaas <helgaas@...nel.org>
cc: Manivannan Sadhasivam <mani@...nel.org>, linux-pci@...r.kernel.org, 
    Christian Zigotzky <chzigotzky@...osoft.de>, 
    mad skateman <madskateman@...il.com>, "R. T. Dickinson" <rtd2@...a.co.nz>, 
    Darren Stevens <darren@...vens-zone.net>, 
    John Paul Adrian Glaubitz <glaubitz@...sik.fu-berlin.de>, 
    Lukas Wunner <lukas@...ner.de>, luigi burdo <intermediadc@...mail.com>, 
    Al <al@...azap.net>, Hongxing Zhu <hongxing.zhu@....com>, 
    hypexed@...oo.com.au, linuxppc-dev@...ts.ozlabs.org, 
    debian-powerpc@...ts.debian.org, linux-kernel@...r.kernel.org, 
    Bjorn Helgaas <bhelgaas@...gle.com>
Subject: Re: [PATCH v2 2/4] PCI/ASPM: Add pcie_aspm_remove_cap() to override
 advertised link states

On Wed, 12 Nov 2025, Bjorn Helgaas wrote:

> > > +	pci_info(pdev, "ASPM:%s%s removed from Link Capabilities to avoid device defect\n",
> > > +		 lnkcap & PCI_EXP_LNKCAP_ASPM_L0S ? " L0s" : "",
> > > +		 lnkcap & PCI_EXP_LNKCAP_ASPM_L1 ? " L1" : "");
> > 
> > I think this gives a false impression that the ASPM CAPs are being
> > removed from the LnkCap register. This function is just removing it
> > from the internal cache and the LnkCap register is left unchanged.
> 
> Very true, this is confusing since we're not actually changing the
> LnkCap register, so lspci etc will still show these states as
> supported.  The quirk needs to work for arbitrary devices, and there's
> no generic way to change LnkCap, so the quirk can't do that.

 There's no way to poke at hw, but that is only relevant for x86 I believe 
and not the default access method for `lspci' anyway.  For sysfs we do it 
already for things such as fixing the device class; cf. `quirk_isa_bridge'
(arch/alpha/kernel/pci.c), so why is it a problem here?  Unless we want to 
keep it for `lspci' actually.

  Maciej