linux-kernel - [PATCH AUTOSEL 6.6 12/19] ALSA: hda: intel: Fix Optimus when GPU has no sound

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: <20250331145601.1705784-12-sashal@kernel.org>
Date: Mon, 31 Mar 2025 10:55:53 -0400
From: Sasha Levin <sashal@...nel.org>
To: linux-kernel@...r.kernel.org,
	stable@...r.kernel.org
Cc: Maxim Mikityanskiy <maxtram95@...il.com>,
	Takashi Iwai <tiwai@...e.de>,
	Sasha Levin <sashal@...nel.org>,
	perex@...ex.cz,
	tiwai@...e.com,
	pierre-louis.bossart@...ux.dev,
	yung-chuan.liao@...ux.intel.com,
	peter.ujfalusi@...ux.intel.com,
	hkallweit1@...il.com,
	linux-sound@...r.kernel.org
Subject: [PATCH AUTOSEL 6.6 12/19] ALSA: hda: intel: Fix Optimus when GPU has no sound

From: Maxim Mikityanskiy <maxtram95@...il.com>

[ Upstream commit 2b360ba9a4936486380bc30d1eabceb40a714d98 ]

quirk_nvidia_hda() forcefully enables HDA controller on all NVIDIA GPUs,
because some buggy BIOSes leave it disabled. However, some dual-GPU
laptops do not have a functional HDA controller in DGPU, and BIOS
disables it on purpose. After quirk_nvidia_hda() reenables this dummy
HDA controller, attempting to probe it fails at azx_first_init(), which
is too late to cancel the probe, as it happens in azx_probe_continue().

The sna_hda_intel driver calls azx_free() and stops the chip, however,
it stays probed, and from the runtime PM point of view, the device
remains active (it was set as active by the PCI subsystem on probe). It
prevents vga_switcheroo from turning off the DGPU, because
pci_create_device_link() syncs power management for video and audio
devices.

Affected devices should be added to driver_denylist to prevent them from
probing early. This patch helps identify such devices by printing a
warning, and also forces the device to the suspended state to allow
vga_switcheroo turn off DGPU.

Signed-off-by: Maxim Mikityanskiy <maxtram95@...il.com>
Link: https://patch.msgid.link/20250208214602.39607-2-maxtram95@gmail.com
Signed-off-by: Takashi Iwai <tiwai@...e.de>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
 sound/pci/hda/hda_intel.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c
index d90151910bee9..da9b9dc0c96d4 100644
--- a/sound/pci/hda/hda_intel.c
+++ b/sound/pci/hda/hda_intel.c
@@ -1363,8 +1363,21 @@ static void azx_free(struct azx *chip)
 	if (use_vga_switcheroo(hda)) {
 		if (chip->disabled && hda->probe_continued)
 			snd_hda_unlock_devices(&chip->bus);
-		if (hda->vga_switcheroo_registered)
+		if (hda->vga_switcheroo_registered) {
 			vga_switcheroo_unregister_client(chip->pci);
+
+			/* Some GPUs don't have sound, and azx_first_init fails,
+			 * leaving the device probed but non-functional. As long
+			 * as it's probed, the PCI subsystem keeps its runtime
+			 * PM status as active. Force it to suspended (as we
+			 * actually stop the chip) to allow GPU to suspend via
+			 * vga_switcheroo, and print a warning.
+			 */
+			dev_warn(&pci->dev, "GPU sound probed, but not operational: please add a quirk to driver_denylist\n");
+			pm_runtime_disable(&pci->dev);
+			pm_runtime_set_suspended(&pci->dev);
+			pm_runtime_enable(&pci->dev);
+		}
 	}

 	if (bus->chip_init) {
-- 
2.39.5