linux-kernel - Re: [PATCH 1/2] PCI/portdrv: Add necessary delay for disabling hotplug events

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: <Z6HYuBDP6uvE1Sf4@wunner.de>
Date: Tue, 4 Feb 2025 10:07:04 +0100
From: Lukas Wunner <lukas@...ner.de>
To: Feng Tang <feng.tang@...ux.alibaba.com>
Cc: Bjorn Helgaas <bhelgaas@...gle.com>,
	Jonathan Cameron <Jonthan.Cameron@...wei.com>,
	ilpo.jarvinen@...ux.intel.com, linux-pci@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/2] PCI/portdrv: Add necessary delay for disabling
 hotplug events

On Tue, Feb 04, 2025 at 01:37:57PM +0800, Feng Tang wrote:
> According to PCIe 6.1 spec, section 6.7.3.2, software need to wait at
> least 1 second for the command-complete event, before resending the cmd
> or sending a new cmd.
> 
> Currently get_port_device_capability() sends slot control cmd to disable
> PCIe hotplug interrupts without waiting for its completion and there was
> real problem reported for the lack of waiting.
> 
> Add the necessary wait to comply with PCIe spec. The waiting logic refers
> existing pcie_poll_cmd().
[...]
> --- a/drivers/pci/pcie/portdrv.c
> +++ b/drivers/pci/pcie/portdrv.c
> @@ -230,8 +260,7 @@ static int get_port_device_capability(struct pci_dev *dev)
>  		 * Disable hot-plug interrupts in case they have been enabled
>  		 * by the BIOS and the hot-plug service driver is not loaded.
>  		 */
> -		pcie_capability_clear_word(dev, PCI_EXP_SLTCTL,
> -			  PCI_EXP_SLTCTL_CCIE | PCI_EXP_SLTCTL_HPIE);
> +		pcie_disable_hp_interrupts_early(dev);
>  	}

The language of the code comment is a bit confusing in that it
says the hot-plug driver may not be "loaded".  This sounds like
it could be modular.  But it can't.  It's always built-in.

So I think what is really meant here is that the driver may be
*disabled* in the config, i.e. CONFIG_HOTPLUG_PCI_PCIE=n.

Now if CONFIG_HOTPLUG_PCI_PCIE=n, you don't need to observe the
Command Completed delay because the hotplug driver won't touch
the Slot Control register afterwards.  It's not compiled in.

On the other hand if CONFIG_HOTPLUG_PCI_PCIE=y, you don't need
to disable the CCIE and HPIE interrupt because the hotplug driver
will handle them.

So I think the proper solution here is to make the write to the
Slot Control register conditional on CONFIG_HOTPLUG_PCI_PCIE,
like this:

	if (dev->is_hotplug_bridge &&
	    (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT ||
-	     pci_pcie_type(dev) == PCI_EXP_TYPE_DOWNSTREAM) &&
-	    (pcie_ports_native || host->native_pcie_hotplug)) {
-		services |= PCIE_PORT_SERVICE_HP;
+	     pci_pcie_type(dev) == PCI_EXP_TYPE_DOWNSTREAM)) {
+		if (pcie_ports_native || host->native_pcie_hotplug)
+			services |= PCIE_PORT_SERVICE_HP;

		/*
		 * Disable hot-plug interrupts in case they have been enabled
		 * by the BIOS and the hot-plug service driver is not loaded.
		 */
-		pcie_capability_clear_word(dev, PCI_EXP_SLTCTL,
-			  PCI_EXP_SLTCTL_CCIE | PCI_EXP_SLTCTL_HPIE);
+		if (!IS_ENABLED(CONFIG_HOTPLUG_PCI_PCIE))
+			pcie_capability_clear_word(dev, PCI_EXP_SLTCTL,
+				  PCI_EXP_SLTCTL_CCIE | PCI_EXP_SLTCTL_HPIE);
	}

The above patch also makes sure the interrupts are quiesced if the
platform didn't grant hotplug control to OSPM.

Thanks,

Lukas