powernv/iommu: disable IOMMU bypass with param iommu=nobypass

When IOMMU bypass is enabled, a PCI device can read and write memory
that was not mapped by the driver without causing an EEH. That might
cause memory corruption, for example.

When we disable bypass, DMA reads and writes to addresses not mapped by
the IOMMU will cause an EEH, allowing us to debug such issues.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 4df73da..7dedfe5 100644 (file)
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -1493,6 +1493,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
                forcesac
                soft
                pt              [x86, IA-64]
+               nobypass        [PPC/POWERNV]
+                       Disable IOMMU bypass, using IOMMU for PCI devices.
 
 
        io7=            [HW] IO7 for Marvel based alpha systems

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index 4f1e43c..85b4738 100644 (file)
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c
@@ -75,6 +75,28 @@ static void pe_level_printk(const struct pnv_ioda_pe *pe, const char *level,
 #define pe_info(pe, fmt, ...)                                  \
        pe_level_printk(pe, KERN_INFO, fmt, ##__VA_ARGS__)
 
+static bool pnv_iommu_bypass_disabled __read_mostly;
+
+static int __init iommu_setup(char *str)
+{
+       if (!str)
+               return -EINVAL;
+
+       while (*str) {
+               if (!strncmp(str, "nobypass", 8)) {
+                       pnv_iommu_bypass_disabled = true;
+                       pr_info("PowerNV: IOMMU bypass window disabled.\n");
+                       break;
+               }
+               str += strcspn(str, ",");
+               if (*str == ',')
+                       str++;
+       }
+
+       return 0;
+}
+early_param("iommu", iommu_setup);
+
 /*
  * stdcix is only supposed to be used in hypervisor real mode as per
  * the architecture spec
@@ -1351,7 +1373,9 @@ static void pnv_pci_ioda2_setup_dma_pe(struct pnv_phb *phb,
                pnv_ioda_setup_bus_dma(pe, pe->pbus, true);
 
        /* Also create a bypass window */
-       pnv_pci_ioda2_setup_bypass_pe(phb, pe);
+       if (!pnv_iommu_bypass_disabled)
+               pnv_pci_ioda2_setup_bypass_pe(phb, pe);
+
        return;
 fail:
        if (pe->tce32_seg >= 0)