Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 23 Dec 2019 20:41:46 -0800
Message-Id: <20191224044146.232713-1-saravanak@google.com>
Mime-Version: 1.0
Subject: [PATCH v2] efi: arm: defer probe of PCIe backed efifb on DT systems
From:   Saravana Kannan <saravanak@google.com>
To:     Ard Biesheuvel <ardb@kernel.org>
Cc:     linux-arm-kernel@lists.infradead.org, will@kernel.org,
        bhelgaas@google.com,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Saravana Kannan <saravanak@google.com>,
        kernel-team@android.com, linux-efi@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

From: Ard Biesheuvel <ardb@kernel.org>

The new of_devlink support breaks PCIe probing on ARM platforms booting
via UEFI if the firmware exposes a EFI framebuffer that is backed by a
PCI device. The reason is that the probing order gets reversed,
resulting in a resource conflict on the framebuffer memory window when
the PCIe probes last, causing it to give up entirely.

Given that we rely on PCI quirks to deal with EFI framebuffers that get
moved around in memory, we cannot simply drop the memory reservation, so
instead, let's use the device link infrastructure to register this
dependency, and force the probing to occur in the expected order.

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Co-developed-by: Saravana Kannan <saravanak@google.com>
Signed-off-by: Saravana Kannan <saravanak@google.com>
---

Hi Ard,

I compile tested it and I think it should work. If you can actually run
and test it, that'd be nice.

You can also optimize find_pci_overlap_node() by caching the result if
you think that's necessary.

Right now this code will run always just like your code did. But once I
rename of_devlink to fw_devlink, this code won't be run if fw_devlink is
disabled.

v1 -> v2:
- Rewrote the device linking part to not depend on initcall ordering

 drivers/firmware/efi/arm-init.c | 106 ++++++++++++++++++++++++++++++--
 1 file changed, 102 insertions(+), 4 deletions(-)

diff --git a/drivers/firmware/efi/arm-init.c b/drivers/firmware/efi/arm-init.c
index 904fa09e6a6b..8b789ff83af0 100644
--- a/drivers/firmware/efi/arm-init.c
+++ b/drivers/firmware/efi/arm-init.c
@@ -10,10 +10,12 @@
 #define pr_fmt(fmt)	"efi: " fmt
 
 #include <linux/efi.h>
+#include <linux/fwnode.h>
 #include <linux/init.h>
 #include <linux/memblock.h>
 #include <linux/mm_types.h>
 #include <linux/of.h>
+#include <linux/of_address.h>
 #include <linux/of_fdt.h>
 #include <linux/platform_device.h>
 #include <linux/screen_info.h>
@@ -276,15 +278,111 @@ void __init efi_init(void)
 		efi_memmap_unmap();
 }
 
+static bool efifb_overlaps_pci_range(const struct of_pci_range *range)
+{
+	u64 fb_base = screen_info.lfb_base;
+
+	if (screen_info.capabilities & VIDEO_CAPABILITY_64BIT_BASE)
+		fb_base |= (u64)(unsigned long)screen_info.ext_lfb_base << 32;
+
+	return fb_base >= range->cpu_addr &&
+	       fb_base < (range->cpu_addr + range->size);
+}
+
+static struct device_node *find_pci_overlap_node(void)
+{
+	struct device_node *np;
+
+	for_each_node_by_type(np, "pci") {
+		struct of_pci_range_parser parser;
+		struct of_pci_range range;
+		int err;
+
+		err = of_pci_range_parser_init(&parser, np);
+		if (err) {
+			pr_warn("of_pci_range_parser_init() failed: %d\n", err);
+			continue;
+		}
+
+		for_each_of_pci_range(&parser, &range)
+			if (efifb_overlaps_pci_range(&range))
+				return np;
+	}
+	return NULL;
+}
+
+/*
+ * If the efifb framebuffer is backed by a PCI graphics controller, we have
+ * to ensure that this relation is expressed using a device link when
+ * running in DT mode, or the probe order may be reversed, resulting in a
+ * resource reservation conflict on the memory window that the efifb
+ * framebuffer steals from the PCIe host bridge.
+ */
+static int efifb_add_links(const struct fwnode_handle *fwnode,
+			   struct device *dev)
+{
+	struct device_node *sup_np;
+	struct device *sup_dev;
+
+	sup_np = find_pci_overlap_node();
+
+	/*
+	 * If there's no PCI graphics controller backing the efifb, we are
+	 * done here.
+	 */
+	if (!sup_np)
+		return 0;
+
+	sup_dev = get_dev_from_fwnode(&sup_np->fwnode);
+	of_node_put(sup_np);
+
+	/*
+	 * Return -ENODEV if the PCI graphics controller device hasn't been
+	 * registered yet.  This ensures that efifb isn't allowed to probe
+	 * and this function is retried again when new devices are
+	 * registered.
+	 */
+	if (!sup_dev)
+		return -ENODEV;
+
+	/*
+	 * If this fails, retrying this function at a later point won't
+	 * change anything. So, don't return an error after this.
+	 */
+	if (!device_link_add(dev, sup_dev, 0))
+		dev_warn(dev, "device_link_add() failed\n");
+
+	put_device(sup_dev);
+
+	return 0;
+}
+
+static struct fwnode_operations efifb_fwnode_ops = {
+	.add_links = efifb_add_links,
+};
+
+static struct fwnode_handle efifb_fwnode = {
+	.ops = &efifb_fwnode_ops,
+};
+
 static int __init register_gop_device(void)
 {
-	void *pd;
+	struct platform_device *pd;
+	int err;
 
 	if (screen_info.orig_video_isVGA != VIDEO_TYPE_EFI)
 		return 0;
 
-	pd = platform_device_register_data(NULL, "efi-framebuffer", 0,
-					   &screen_info, sizeof(screen_info));
-	return PTR_ERR_OR_ZERO(pd);
+	pd = platform_device_alloc("efi-framebuffer", 0);
+	if (!pd)
+		return -ENOMEM;
+
+	pd->dev.fwnode = &efifb_fwnode;
+
+	err = platform_device_add_data(pd, &screen_info, sizeof(screen_info));
+	if (err)
+		return err;
+
+	return platform_device_add(pd);
 }
 subsys_initcall(register_gop_device);
-- 
2.24.1.735.g03f4e72817-goog