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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <e60d2cf59666b6f670996bac80cb948acb1d7b5c.camel@web.de>
Date: Mon, 06 Oct 2025 18:22:31 +0200
From: Bert Karwatzki <spasswolf@....de>
To: Christian König <christian.koenig@....com>, 
	linux-kernel@...r.kernel.org
Cc: linux-next@...r.kernel.org, linux-stable@...r.kernel.org, 
	regressions@...ts.linux.dev, linux-pci@...r.kernel.org, 
	linux-acpi@...r.kernel.org, Mario Limonciello <superm1@...nel.org>, "Rafael
 J . Wysocki" <rafael.j.wysocki@...el.com>, spasswolf@....de
Subject: Re: [REGRESSION 00/04] Crash during resume of pcie bridge

Am Montag, dem 06.10.2025 um 14:39 +0200 schrieb Christian König:
> On 06.10.25 14:09, Bert Karwatzki wrote:
> > Since linux version v6.15 I experience random crashes on my MSI Alpha 15 Laptop
> > running debian trixie (amd64). The first such crash happened about in the midth
> > of june, and as there were no useful log messages and even using netconsole
> > gave no useful message I suspected faulty hardware. So I ran memtest86+ and
> > found a faulty address line and replaced the memory (unfortunately 64G to 16G).
> > But the crashes occured again and so I did a thorough investigation.
> > 
> > The crashes occur after 30min to 33h (yes, hours) of uptime and consist of a
> > sudden reboot after which the PCI bridge at 00:02.4 and the nvme device 
> > connected to it are missing. If there's sound running during the crash then the
> > first sign of the crash is the sound looping like a broken record for about 2s,
> > after which the reboot happens. With the missing nvme device the reboot drops to
> > a rescue shell. Using "shutdown -h now" from that shell and starting the laptop
> > with the power button restores the missing PCI bridge and nvme device.
> 
> Oh well, it sounds like some PCIe device is dropping of the bus and taking it's upstream bridge with it.
> 
> > As the bisections were not succesfull I tried to monitor the crash using
> > netconsole and CONFIG_ACPI_DEBUG and "acpi.debug_layer=0xf acpi.debug_level=0x107"
> > as command line parameters. With this the last message on netconsole before
> > the crash is usually:
> > 
> > [21465.639279] [    T251]    evmisc-0132 ev_queue_notify_reques: Dispatching Notify on [GPP0] (Device) Value 0x00 (Bus Check) Node 00000000f81f36b8
> 
> A full dump of that might be helpful. That sounds like the dGPU is powering up/down.

Yes, that's what's happening.

> 
> > 
> > GPP0 is the ACPI name of this PCI bridge (at least that's my best guess):
> > 
> > 00:01.1 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] Renoir PCIe GPP Bridge [1022:1633]
> > 
> > to which the discrete GPU is connected
> > 
> > 03:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 23 [Radeon RX 6600/6600 XT/6600M] [1002:73ff] (rev c3)
> > 
> > via the pci express switch
> > 
> > 01:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Upstream Port of PCI Express Switch [1002:1478] (rev c3)
> > 02:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Downstream Port of PCI Express Switch [1002:1479]
> > 
> > While the GUI (xfce on xorg) on my laptop runs on the built-in GPU the discrete 
> > GPU usually wakes up quite often, e.g. when a window is opened or when scrolling down on youtube.
> 
> Yeah, that is a known issue and we are working on it.

Until linux v6.15 this didn't cause any harm.

> 
> Basically an application enumerates the possible render or video decode devices in the system and that wakes up the dGPU even when it isn't actually used.
> 
> > A somewhat reliable method to generate GPP0 notifies is putting on a youtube
> > video and the periodically starting evolution with this script:
> > 
> > #!/bin/bash
> > for i in {0..1000}
> > do
> > 	echo $i
> > 	evolution &
> > 	sleep 5
> > 	killall evolution
> > 	sleep 55
> > done
> > 
> > This is also the method I used to test the debug kernel in the following mails.
> 
> To further narrow down the issue please run your laptop with amdgpu.runpm=0 on the kernel command line for a while and see if that is stable or not.
> 
Even versions that did crash can be stable for 24h of uptime so I think this 
will take too long.
I think I've already chased down the crash to this part of rpm_resume()
(I'm currently doing a testrun with more dev_info()s in this part):

 skip_parent:

	if (!strcmp(dev_name(dev), "0000:00:01.1"))
		dev_info(dev, "%s %d\n", __func__, __LINE__); // this is the last reported line in netconsole
	if (dev->power.no_callbacks)
		goto no_callback;	/* Assume success. */

	__update_runtime_status(dev, RPM_RESUMING);

	callback = RPM_GET_CALLBACK(dev, runtime_resume);

	dev_pm_disable_wake_irq_check(dev, false);
	retval = rpm_callback(callback, dev);
	if (retval) {
		__update_runtime_status(dev, RPM_SUSPENDED);
		pm_runtime_cancel_pending(dev);
		dev_pm_enable_wake_irq_check(dev, false);
	} else {
 no_callback:


Bert Karwatzki

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ