bcm53xx: use upstream accepted fix for ARM core aborts

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
owl
Rafał Miłecki 2016-11-28 13:07:29 +01:00
parent cb7ab730c7
commit 32875a2d79
2 changed files with 75 additions and 42 deletions

View File

@ -0,0 +1,75 @@
From 09f3510fb70a46c8921f2cf4a90dbcae460a6820 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rafa=C5=82=20Mi=C5=82ecki?= <rafal@milecki.pl>
Date: Sat, 29 Oct 2016 13:12:29 +0200
Subject: [PATCH] ARM: BCM5301X: Add back handler ignoring external imprecise
aborts
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Since early BCM5301X days we got abort handler that was removed by
commit 937b12306ea79 ("ARM: BCM5301X: remove workaround imprecise abort
fault handler"). It assumed we need to deal only with pending aborts
left by the bootloader. Unfortunately this isn't true for BCM5301X.
When probing PCI config space (device enumeration) it is expected to
have master aborts on the PCI bus. Most bridges don't forward (or they
allow disabling it) these errors onto the AXI/AMBA bus but not the
Northstar (BCM5301X) one.
iProc PCIe controller on Northstar seems to be some older one, without
a control register for errors forwarding. It means we need to workaround
this at platform level. All newer platforms are not affected by this
issue.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
arch/arm/mach-bcm/bcm_5301x.c | 28 ++++++++++++++++++++++++++++
1 file changed, 28 insertions(+)
--- a/arch/arm/mach-bcm/bcm_5301x.c
+++ b/arch/arm/mach-bcm/bcm_5301x.c
@@ -9,14 +9,42 @@
#include <asm/hardware/cache-l2x0.h>
#include <asm/mach/arch.h>
+#include <asm/siginfo.h>
+#include <asm/signal.h>
+
+#define FSR_EXTERNAL (1 << 12)
+#define FSR_READ (0 << 10)
+#define FSR_IMPRECISE 0x0406
static const char *const bcm5301x_dt_compat[] __initconst = {
"brcm,bcm4708",
NULL,
};
+static int bcm5301x_abort_handler(unsigned long addr, unsigned int fsr,
+ struct pt_regs *regs)
+{
+ /*
+ * We want to ignore aborts forwarded from the PCIe bus that are
+ * expected and shouldn't really be passed by the PCIe controller.
+ * The biggest disadvantage is the same FSR code may be reported when
+ * reading non-existing APB register and we shouldn't ignore that.
+ */
+ if (fsr == (FSR_EXTERNAL | FSR_READ | FSR_IMPRECISE))
+ return 0;
+
+ return 1;
+}
+
+static void __init bcm5301x_init_early(void)
+{
+ hook_fault_code(16 + 6, bcm5301x_abort_handler, SIGBUS, BUS_OBJERR,
+ "imprecise external abort");
+}
+
DT_MACHINE_START(BCM5301X, "BCM5301X")
.l2c_aux_val = 0,
.l2c_aux_mask = ~0,
.dt_compat = bcm5301x_dt_compat,
+ .init_early = bcm5301x_init_early,
MACHINE_END

View File

@ -1,42 +0,0 @@
From: =?UTF-8?q?Rafa=C5=82=20Mi=C5=82ecki?= <zajec5@gmail.com>
Subject: [PATCH] PCI: iproc: Ignore ARM core aborts when running on bcma bus
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Northstar devices have host bridges that forward errors to the ARM core.
These errors happen during PCI bus probing (device enumeration) and are
expected. There is no reason for them to cause a die() so this patch
adds a handler silencing them.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
---
--- a/drivers/pci/host/pcie-iproc-bcma.c
+++ b/drivers/pci/host/pcie-iproc-bcma.c
@@ -40,6 +40,15 @@ static int iproc_pcie_bcma_map_irq(const
return bcma_core_irq(bdev, 5);
}
+static int iproc_pcie_bcma_abort_handler(unsigned long addr, unsigned int fsr,
+ struct pt_regs *regs)
+{
+ if (fsr == 0x1406)
+ return 0;
+
+ return 1;
+}
+
static int iproc_pcie_bcma_probe(struct bcma_device *bdev)
{
struct iproc_pcie *pcie;
@@ -65,6 +74,9 @@ static int iproc_pcie_bcma_probe(struct
pcie->map_irq = iproc_pcie_bcma_map_irq;
+ hook_fault_code(16 + 6, iproc_pcie_bcma_abort_handler, SIGBUS, BUS_OBJERR,
+ "imprecise external abort");
+
ret = iproc_pcie_setup(pcie, &res);
if (ret)
dev_err(pcie->dev, "PCIe controller setup failed\n");