From: Yazen Ghannam <Yazen.Ghannam@amd.com>
To: linux-acpi@vger.kernel.org
Cc: linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org,
        rjw@rjwysocki.net, lenb@kernel.org, bhelgaas@google.com,
        bp@suse.de, Yazen Ghannam <yazen.ghannam@amd.com>
Subject: [PATCH v2] PCI/ACPI: Disable AER when _OSC control bit is clear.
Date: Mon, 15 Jan 2018 10:20:06 -0600
Message-Id: <20180115162006.45553-1-Yazen.Ghannam@amd.com>
MIME-Version: 1.0
Content-Type: text/plain
SpamDiagnosticOutput: 1:99
SpamDiagnosticMetadata: NSPM
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 15 Jan 2018 16:20:23.1428 (UTC)
X-MS-Exchange-CrossTenant-Network-Message-Id: c9d8537a-29ff-4ecf-1bfc-08d55c33e17f
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: 3dd8961f-e488-4e60-8e11-a82d994e183d
X-MS-Exchange-Transport-CrossTenantHeadersStamped: DM5PR12MB1916
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Return-Path: <linux-kernel-owner@vger.kernel.org>

From: Yazen Ghannam <yazen.ghannam@amd.com>

Currently, aer_service_init() checks if AER is available and that
Firmware First handling is not enabled. The _OSC request for AER is not
taken into account when deciding to enable AER in Linux.

>From ACPI 6.2 Section 6.2.11.3, "If any bits in the Control Field are
returned cleared (masked to zero) by the _OSC control method, the
respective feature is designated unsupported by the platform and must
not be enabled by the OS."

The OS and the Platform should agree that the OS can have control of AER
otherwise we should disable AER in the OS.

Mark AER as disabled if the _OSC request was not made or accepted.

This covers two cases where the OS and Platform disagree:
1) The OS requests AER control and Platform denies the request.
2) The OS does not request AER control but the Platform returns the AER
   control bit set, possibly due to a Firmware bug.

The _OSC control for AER is not requested when APEI Firmware First is
used, so the same condition applies from case 2 above.

Remove redundant check for aer_acpi_firmware_first() when calling
aer_service_init(), since this check is already included when checking
the _OSC control.

Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
---
Link:
https://lkml.kernel.org/r/20180111150316.19951-1-Yazen.Ghannam@amd.com

v1->v2:
* Expand commit message.
* Add Spec reference to commit message.
* Fix spelling error in commit message.
* Add comment for 3-way bitwise AND.

 drivers/acpi/pci_root.c       | 7 +++++++
 drivers/pci/pcie/aer/aerdrv.c | 2 +-
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index 6fc204a52493..ab0192fd24c7 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -512,6 +512,13 @@ static void negotiate_os_control(struct acpi_pci_root *root, int *no_aspm)
 		 */
 		*no_aspm = 1;
 	}
+
+	/*
+	 * We can use a 3-way bitwise AND to check that the AER control bit is
+	 * both requested by the OS and granted by the Platform.
+	 */
+	if (!(requested & control & OSC_PCI_EXPRESS_AER_CONTROL))
+		pci_no_aer();
 }
 
 static int acpi_pci_root_add(struct acpi_device *device,
diff --git a/drivers/pci/pcie/aer/aerdrv.c b/drivers/pci/pcie/aer/aerdrv.c
index 6ff5f5b4f5e6..39bb059777d0 100644
--- a/drivers/pci/pcie/aer/aerdrv.c
+++ b/drivers/pci/pcie/aer/aerdrv.c
@@ -374,7 +374,7 @@ static void aer_error_resume(struct pci_dev *dev)
  */
 static int __init aer_service_init(void)
 {
-	if (!pci_aer_available() || aer_acpi_firmware_first())
+	if (!pci_aer_available())
 		return -ENXIO;
 	return pcie_port_service_register(&aerdriver);
 }
-- 
2.14.1