bus-scan-delay-5.10-rc4.patch

From ceae51b1cc0e5a5b42999274657bd55606193661 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jari=20H=C3=A4m=C3=A4l=C3=A4inen?= <[email protected]>
Date: Sun, 22 Nov 2020 15:24:49 +0200
Subject: [PATCH] nuumio: pcie: Reimplement rockchip PCIe bus scan delay

Reimplementation of my old Rockchip PCIe bus scan delay patch for
kernels >= 5.9.

Delay may fix panix with some PCIe devices, like LSI SAS 9201-8i with
SAS2008 chipset in my case.

Crash dump (customized Manjaro kernel before this patch):
[    1.229856] SError Interrupt on CPU4, code 0xbf000002 -- SError
[    1.229860] CPU: 4 PID: 1 Comm: swapper/0 Not tainted 5.9.9-2.0-MANJARO-ARM #1
[    1.229862] Hardware name: Pine64 RockPro64 v2.1 (DT)
[    1.229864] pstate: 60000085 (nZCv daIf -PAN -UAO BTYPE=--)
[    1.229866] pc : rockchip_pcie_rd_conf+0xb4/0x270
[    1.229868] lr : rockchip_pcie_rd_conf+0x1b4/0x270
[    1.229870] sp : ffff80001004b850
[    1.229872] x29: ffff80001004b850 x28: 0000000000000001
[    1.229877] x27: 0000000000000000 x26: ffff00007a795000
[    1.229882] x25: ffff00007a7910b0 x24: 0000000000000000
[    1.229887] x23: 0000000000000000 x22: ffff00007b3a4380
[    1.229891] x21: ffff80001004b8c4 x20: 0000000000000004
[    1.229895] x19: 0000000000100000 x18: 0000000000000020
[    1.229900] x17: 0000000000000001 x16: 0000000000000019
[    1.229904] x15: ffff00007b222fd8 x14: ffffffffffffffff
[    1.229908] x13: ffff00007a79ba1c x12: ffff00007a79b290
[    1.229912] x11: 0101010101010101 x10: 7f7f7f7f7f7f7f7f
[    1.229917] x9 : ff72646268756463 x8 : 0000000000000391
[    1.229921] x7 : ffff80001004b880 x6 : 0000000000000001
[    1.229925] x5 : 0000000000000000 x4 : 0000000000000000
[    1.229930] x3 : 0000000000c00008 x2 : 000000000080000a
[    1.229934] x1 : 0000000000000000 x0 : ffff800014000000
[    1.229939] Kernel panic - not syncing: Asynchronous SError Interrupt
[    1.229942] CPU: 4 PID: 1 Comm: swapper/0 Not tainted 5.9.9-2.0-MANJARO-ARM #1
[    1.229944] Hardware name: Pine64 RockPro64 v2.1 (DT)
[    1.229946] Call trace:
[    1.229948]  dump_backtrace+0x0/0x1d0
[    1.229949]  show_stack+0x18/0x24
[    1.229951]  dump_stack+0xc0/0x118
[    1.229953]  panic+0x148/0x320
[    1.229955]  nmi_panic+0x8c/0x90
[    1.229956]  arm64_serror_panic+0x78/0x84
[    1.229958]  do_serror+0x15c/0x160
[    1.229960]  el1_error+0x84/0x100
[    1.229962]  rockchip_pcie_rd_conf+0xb4/0x270
[    1.229964]  pci_bus_read_config_dword+0x6c/0xd0
[    1.229966]  pci_bus_generic_read_dev_vendor_id+0x34/0x1b0
[    1.229968]  pci_scan_single_device+0xa4/0x144
[    1.229970]  pci_scan_slot+0x40/0x12c
[    1.229972]  pci_scan_child_bus_extend+0x58/0x34c
[    1.229974]  pci_scan_bridge_extend+0x310/0x590
[    1.229976]  pci_scan_child_bus_extend+0x210/0x34c
[    1.229978]  pci_scan_root_bus_bridge+0x68/0xdc
[    1.229980]  pci_host_probe+0x18/0xc4
[    1.229981]  rockchip_pcie_probe+0x204/0x330
[    1.229984]  platform_drv_probe+0x54/0xb0
[    1.229985]  really_probe+0xe8/0x500
[    1.229987]  driver_probe_device+0xd8/0xf0
[    1.229989]  device_driver_attach+0xc0/0xcc
[    1.229991]  __driver_attach+0xa4/0x170
[    1.229993]  bus_for_each_dev+0x70/0xc0
[    1.229994]  driver_attach+0x24/0x30
[    1.229996]  bus_add_driver+0x140/0x234
[    1.229998]  driver_register+0x78/0x130
[    1.230000]  __platform_driver_register+0x4c/0x60
[    1.230002]  rockchip_pcie_driver_init+0x1c/0x28
[    1.230004]  do_one_initcall+0x54/0x1c0
[    1.230005]  do_initcalls+0xf4/0x130
[    1.230007]  kernel_init_freeable+0x144/0x19c
[    1.230009]  kernel_init+0x14/0x11c
[    1.230011]  ret_from_fork+0x10/0x34
[    1.230035] SMP: stopping secondary CPUs
[    1.230037] Kernel Offset: disabled
[    1.230039] CPU features: 0x0240022,2100200c
[    1.230041] Memory Limit: none
---
 .../admin-guide/kernel-parameters.txt         |  8 ++++++
 .../boot/dts/rockchip/rk3399-rockpro64.dtsi   |  1 +
 drivers/pci/controller/pcie-rockchip-host.c   | 25 +++++++++++++++++++
 drivers/pci/controller/pcie-rockchip.c        |  6 +++++
 drivers/pci/controller/pcie-rockchip.h        |  2 ++
 5 files changed, 42 insertions(+)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 44fde25bb221..941dc943eb84 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3816,6 +3816,14 @@
 		nomsi	Do not use MSI for native PCIe PME signaling (this makes
 			all PCIe root ports use INTx for all services).
 
+	pcie_rockchip_host.bus_scan_delay=	[PCIE] Delay in ms before
+			scanning PCIe bus in Rockchip PCIe host driver. Some PCIe
+			cards seem to need delays that can be several hundred ms.
+			If set to greater than or equal to 0 this parameter will
+			override delay that can be set in device tree.
+			Values less than 0 mean that this parameter is ignored.
+			default=-1
+
 	pcmv=		[HW,PCMCIA] BadgePAD 4
 
 	pd_ignore_unused
diff --git a/arch/arm64/boot/dts/rockchip/rk3399-rockpro64.dtsi b/arch/arm64/boot/dts/rockchip/rk3399-rockpro64.dtsi
index 6e553ff47534..9ec248d1c4c3 100644
--- a/arch/arm64/boot/dts/rockchip/rk3399-rockpro64.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3399-rockpro64.dtsi
@@ -548,6 +548,7 @@ &pcie0 {
 	pinctrl-0 = <&pcie_perst>;
 	vpcie12v-supply = <&vcc12v_dcin>;
 	vpcie3v3-supply = <&vcc3v3_pcie>;
+	bus-scan-delay-ms = <1000>;
 	status = "okay";
 };
 
diff --git a/drivers/pci/controller/pcie-rockchip-host.c b/drivers/pci/controller/pcie-rockchip-host.c
index 9705059523a6..632dac43a037 100644
--- a/drivers/pci/controller/pcie-rockchip-host.c
+++ b/drivers/pci/controller/pcie-rockchip-host.c
@@ -24,6 +24,7 @@
 #include <linux/kernel.h>
 #include <linux/mfd/syscon.h>
 #include <linux/module.h>
+#include <linux/moduleparam.h>
 #include <linux/of_address.h>
 #include <linux/of_device.h>
 #include <linux/of_pci.h>
@@ -39,6 +40,9 @@
 #include "../pci.h"
 #include "pcie-rockchip.h"
 
+static int bus_scan_delay = -1;
+module_param_named(bus_scan_delay, bus_scan_delay, int, S_IRUGO);
+
 static void rockchip_pcie_enable_bw_int(struct rockchip_pcie *rockchip)
 {
 	u32 status;
@@ -942,6 +946,7 @@ static int rockchip_pcie_probe(struct platform_device *pdev)
 	struct device *dev = &pdev->dev;
 	struct pci_host_bridge *bridge;
 	int err;
+	u32 delay = 0;
 
 	if (!dev->of_node)
 		return -ENODEV;
@@ -993,6 +998,26 @@ static int rockchip_pcie_probe(struct platform_device *pdev)
 	bridge->sysdata = rockchip;
 	bridge->ops = &rockchip_pcie_ops;
 
+	/* Checking if bus scan delay was given from command line and prefer
+	 * that over the value in device tree (which defaults to 0 if not set).
+	 */
+	if (bus_scan_delay >= 0) {
+		delay = bus_scan_delay;
+		dev_info(dev, "wait %u ms (from command-line) before bus scan\n", delay);
+	} else {
+		delay = rockchip->bus_scan_delay;
+		dev_info(dev, "wait %u ms (from device tree) before bus scan\n", delay);
+	}
+	/* Workaround for some devices crashing on pci_host_probe / pci_scan_root_bus_bridge
+	 * calls: sleep a bit before bus scan. Call trace gets to rockchip_pcie_rd_conf when
+	 * trying to read vendor id (pci_bus_generic_read_dev_vendor_id is in call stack)
+	 * before panicing. I have no idea why this works or what causes the panic. I just
+	 * found this hack by luck when trying to "make it break differently if possible".
+	 */
+	if (delay > 0) {
+		msleep(delay);
+	}
+
 	err = pci_host_probe(bridge);
 	if (err < 0)
 		goto err_remove_irq_domain;
diff --git a/drivers/pci/controller/pcie-rockchip.c b/drivers/pci/controller/pcie-rockchip.c
index 904dec0d3a88..e6c97f9944ba 100644
--- a/drivers/pci/controller/pcie-rockchip.c
+++ b/drivers/pci/controller/pcie-rockchip.c
@@ -149,6 +149,12 @@ int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
 		return PTR_ERR(rockchip->clk_pcie_pm);
 	}
 
+	err = of_property_read_u32(node, "bus-scan-delay-ms", &rockchip->bus_scan_delay);
+	if (err) {
+		dev_info(dev, "no bus scan delay, default to 0 ms\n");
+		rockchip->bus_scan_delay = 0;
+	}
+
 	return 0;
 }
 EXPORT_SYMBOL_GPL(rockchip_pcie_parse_dt);
diff --git a/drivers/pci/controller/pcie-rockchip.h b/drivers/pci/controller/pcie-rockchip.h
index c7d0178fc8c2..52fd2108b704 100644
--- a/drivers/pci/controller/pcie-rockchip.h
+++ b/drivers/pci/controller/pcie-rockchip.h
@@ -306,6 +306,8 @@ struct rockchip_pcie {
 	phys_addr_t msg_bus_addr;
 	bool is_rc;
 	struct resource *mem_res;
+	/* Bus scan delay is a workaround for some pcie devices causing crashes */
+	u32 bus_scan_delay;
 };
 
 static u32 rockchip_pcie_read(struct rockchip_pcie *rockchip, u32 reg)
-- 
2.29.2