aboutsummaryrefslogtreecommitdiff
path: root/arch/powerpc/sysdev/dart_iommu.c
diff options
context:
space:
mode:
Diffstat (limited to 'arch/powerpc/sysdev/dart_iommu.c')
-rw-r--r--arch/powerpc/sysdev/dart_iommu.c265
1 files changed, 202 insertions, 63 deletions
diff --git a/arch/powerpc/sysdev/dart_iommu.c b/arch/powerpc/sysdev/dart_iommu.c
index 61d31742861..9e5353ff6d1 100644
--- a/arch/powerpc/sysdev/dart_iommu.c
+++ b/arch/powerpc/sysdev/dart_iommu.c
@@ -27,36 +27,36 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
-#include <linux/config.h>
#include <linux/init.h>
#include <linux/types.h>
-#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/spinlock.h>
#include <linux/string.h>
#include <linux/pci.h>
#include <linux/dma-mapping.h>
#include <linux/vmalloc.h>
+#include <linux/suspend.h>
+#include <linux/memblock.h>
+#include <linux/gfp.h>
#include <asm/io.h>
#include <asm/prom.h>
#include <asm/iommu.h>
#include <asm/pci-bridge.h>
#include <asm/machdep.h>
-#include <asm/abs_addr.h>
#include <asm/cacheflush.h>
-#include <asm/lmb.h>
#include <asm/ppc-pci.h>
#include "dart.h"
-extern int iommu_force_on;
-
/* Physical base address and size of the DART table */
unsigned long dart_tablebase; /* exported to htab_initialize */
static unsigned long dart_tablesize;
/* Virtual base address of the DART table */
static u32 *dart_vbase;
+#ifdef CONFIG_PM
+static u32 *dart_copy;
+#endif
/* Mapped base address for the dart */
static unsigned int __iomem *dart;
@@ -69,13 +69,20 @@ static int iommu_table_dart_inited;
static int dart_dirty;
static int dart_is_u4;
+#define DART_U4_BYPASS_BASE 0x8000000000ull
+
#define DBG(...)
+static DEFINE_SPINLOCK(invalidate_lock);
+
static inline void dart_tlb_invalidate_all(void)
{
unsigned long l = 0;
unsigned int reg, inv_bit;
unsigned long limit;
+ unsigned long flags;
+
+ spin_lock_irqsave(&invalidate_lock, flags);
DBG("dart: flush\n");
@@ -100,49 +107,97 @@ retry:
if (l == (1L << limit)) {
if (limit < 4) {
limit++;
- reg = DART_IN(DART_CNTL);
- reg &= ~inv_bit;
+ reg = DART_IN(DART_CNTL);
+ reg &= ~inv_bit;
DART_OUT(DART_CNTL, reg);
goto retry;
} else
panic("DART: TLB did not flush after waiting a long "
"time. Buggy U3 ?");
}
+
+ spin_unlock_irqrestore(&invalidate_lock, flags);
+}
+
+static inline void dart_tlb_invalidate_one(unsigned long bus_rpn)
+{
+ unsigned int reg;
+ unsigned int l, limit;
+ unsigned long flags;
+
+ spin_lock_irqsave(&invalidate_lock, flags);
+
+ reg = DART_CNTL_U4_ENABLE | DART_CNTL_U4_IONE |
+ (bus_rpn & DART_CNTL_U4_IONE_MASK);
+ DART_OUT(DART_CNTL, reg);
+
+ limit = 0;
+wait_more:
+ l = 0;
+ while ((DART_IN(DART_CNTL) & DART_CNTL_U4_IONE) && l < (1L << limit)) {
+ rmb();
+ l++;
+ }
+
+ if (l == (1L << limit)) {
+ if (limit < 4) {
+ limit++;
+ goto wait_more;
+ } else
+ panic("DART: TLB did not flush after waiting a long "
+ "time. Buggy U4 ?");
+ }
+
+ spin_unlock_irqrestore(&invalidate_lock, flags);
}
static void dart_flush(struct iommu_table *tbl)
{
- if (dart_dirty)
+ mb();
+ if (dart_dirty) {
dart_tlb_invalidate_all();
- dart_dirty = 0;
+ dart_dirty = 0;
+ }
}
-static void dart_build(struct iommu_table *tbl, long index,
+static int dart_build(struct iommu_table *tbl, long index,
long npages, unsigned long uaddr,
- enum dma_data_direction direction)
+ enum dma_data_direction direction,
+ struct dma_attrs *attrs)
{
unsigned int *dp;
unsigned int rpn;
+ long l;
DBG("dart: build at: %lx, %lx, addr: %x\n", index, npages, uaddr);
- index <<= DART_PAGE_FACTOR;
- npages <<= DART_PAGE_FACTOR;
-
dp = ((unsigned int*)tbl->it_base) + index;
- /* On U3, all memory is contigous, so we can move this
+ /* On U3, all memory is contiguous, so we can move this
* out of the loop.
*/
- while (npages--) {
- rpn = virt_to_abs(uaddr) >> DART_PAGE_SHIFT;
+ l = npages;
+ while (l--) {
+ rpn = __pa(uaddr) >> DART_PAGE_SHIFT;
*(dp++) = DARTMAP_VALID | (rpn & DARTMAP_RPNMASK);
uaddr += DART_PAGE_SIZE;
}
- dart_dirty = 1;
+ /* make sure all updates have reached memory */
+ mb();
+ in_be32((unsigned __iomem *)dp);
+ mb();
+
+ if (dart_is_u4) {
+ rpn = index;
+ while (npages--)
+ dart_tlb_invalidate_one(rpn++);
+ } else {
+ dart_dirty = 1;
+ }
+ return 0;
}
@@ -157,9 +212,6 @@ static void dart_free(struct iommu_table *tbl, long index, long npages)
DBG("dart: free at: %lx, %lx\n", index, npages);
- index <<= DART_PAGE_FACTOR;
- npages <<= DART_PAGE_FACTOR;
-
dp = ((unsigned int *)tbl->it_base) + index;
while (npages--)
@@ -167,7 +219,7 @@ static void dart_free(struct iommu_table *tbl, long index, long npages)
}
-static int dart_init(struct device_node *dart_node)
+static int __init dart_init(struct device_node *dart_node)
{
unsigned int i;
unsigned long tmp, base, size;
@@ -193,17 +245,17 @@ static int dart_init(struct device_node *dart_node)
* that to work around what looks like a problem with the HT bridge
* prefetching into invalid pages and corrupting data
*/
- tmp = lmb_alloc(DART_PAGE_SIZE, DART_PAGE_SIZE);
+ tmp = memblock_alloc(DART_PAGE_SIZE, DART_PAGE_SIZE);
dart_emptyval = DARTMAP_VALID | ((tmp >> DART_PAGE_SHIFT) &
DARTMAP_RPNMASK);
/* Map in DART registers */
- dart = ioremap(r.start, r.end - r.start + 1);
+ dart = ioremap(r.start, resource_size(&r));
if (dart == NULL)
panic("DART: Cannot map registers!");
/* Map in DART table */
- dart_vbase = ioremap(virt_to_abs(dart_tablebase), dart_tablesize);
+ dart_vbase = ioremap(__pa(dart_tablebase), dart_tablesize);
/* Fill initial table */
for (i = 0; i < dart_tablesize/4; i++)
@@ -239,13 +291,14 @@ static void iommu_table_dart_setup(void)
iommu_table_dart.it_busno = 0;
iommu_table_dart.it_offset = 0;
/* it_size is in number of entries */
- iommu_table_dart.it_size = (dart_tablesize / sizeof(u32)) >> DART_PAGE_FACTOR;
+ iommu_table_dart.it_size = dart_tablesize / sizeof(u32);
+ iommu_table_dart.it_page_shift = IOMMU_PAGE_SHIFT_4K;
/* Initialize the common IOMMU code */
iommu_table_dart.it_base = (unsigned long)dart_vbase;
iommu_table_dart.it_index = 0;
iommu_table_dart.it_blocksize = 1;
- iommu_init_table(&iommu_table_dart);
+ iommu_init_table(&iommu_table_dart, -1);
/* Reserve the last page of the DART to avoid possible prefetch
* past the DART mapped area
@@ -253,42 +306,70 @@ static void iommu_table_dart_setup(void)
set_bit(iommu_table_dart.it_size - 1, iommu_table_dart.it_map);
}
-static void iommu_dev_setup_dart(struct pci_dev *dev)
+static void dma_dev_setup_dart(struct device *dev)
{
- struct device_node *dn;
-
/* We only have one iommu table on the mac for now, which makes
* things simple. Setup all PCI devices to point to this table
- *
- * We must use pci_device_to_OF_node() to make sure that
- * we get the real "final" pointer to the device in the
- * pci_dev sysdata and not the temporary PHB one
*/
- dn = pci_device_to_OF_node(dev);
-
- if (dn)
- PCI_DN(dn)->iommu_table = &iommu_table_dart;
+ if (get_dma_ops(dev) == &dma_direct_ops)
+ set_dma_offset(dev, DART_U4_BYPASS_BASE);
+ else
+ set_iommu_table_base(dev, &iommu_table_dart);
}
-static void iommu_bus_setup_dart(struct pci_bus *bus)
+static void pci_dma_dev_setup_dart(struct pci_dev *dev)
{
- struct device_node *dn;
+ dma_dev_setup_dart(&dev->dev);
+}
+static void pci_dma_bus_setup_dart(struct pci_bus *bus)
+{
if (!iommu_table_dart_inited) {
iommu_table_dart_inited = 1;
iommu_table_dart_setup();
}
+}
- dn = pci_bus_to_OF_node(bus);
-
- if (dn)
- PCI_DN(dn)->iommu_table = &iommu_table_dart;
+static bool dart_device_on_pcie(struct device *dev)
+{
+ struct device_node *np = of_node_get(dev->of_node);
+
+ while(np) {
+ if (of_device_is_compatible(np, "U4-pcie") ||
+ of_device_is_compatible(np, "u4-pcie")) {
+ of_node_put(np);
+ return true;
+ }
+ np = of_get_next_parent(np);
+ }
+ return false;
}
-static void iommu_dev_setup_null(struct pci_dev *dev) { }
-static void iommu_bus_setup_null(struct pci_bus *bus) { }
+static int dart_dma_set_mask(struct device *dev, u64 dma_mask)
+{
+ if (!dev->dma_mask || !dma_supported(dev, dma_mask))
+ return -EIO;
+
+ /* U4 supports a DART bypass, we use it for 64-bit capable
+ * devices to improve performances. However, that only works
+ * for devices connected to U4 own PCIe interface, not bridged
+ * through hypertransport. We need the device to support at
+ * least 40 bits of addresses.
+ */
+ if (dart_device_on_pcie(dev) && dma_mask >= DMA_BIT_MASK(40)) {
+ dev_info(dev, "Using 64-bit DMA iommu bypass\n");
+ set_dma_ops(dev, &dma_direct_ops);
+ } else {
+ dev_info(dev, "Using 32-bit DMA via iommu\n");
+ set_dma_ops(dev, &dma_iommu_ops);
+ }
+ dma_dev_setup_dart(dev);
-void iommu_init_early_dart(void)
+ *dev->dma_mask = dma_mask;
+ return 0;
+}
+
+void __init iommu_init_early_dart(void)
{
struct device_node *dn;
@@ -297,42 +378,95 @@ void iommu_init_early_dart(void)
if (dn == NULL) {
dn = of_find_compatible_node(NULL, "dart", "u4-dart");
if (dn == NULL)
- goto bail;
+ return; /* use default direct_dma_ops */
dart_is_u4 = 1;
}
+ /* Initialize the DART HW */
+ if (dart_init(dn) != 0)
+ goto bail;
+
/* Setup low level TCE operations for the core IOMMU code */
ppc_md.tce_build = dart_build;
ppc_md.tce_free = dart_free;
ppc_md.tce_flush = dart_flush;
- /* Initialize the DART HW */
- if (dart_init(dn) == 0) {
- ppc_md.iommu_dev_setup = iommu_dev_setup_dart;
- ppc_md.iommu_bus_setup = iommu_bus_setup_dart;
+ /* Setup bypass if supported */
+ if (dart_is_u4)
+ ppc_md.dma_set_mask = dart_dma_set_mask;
- /* Setup pci_dma ops */
- pci_iommu_init();
+ ppc_md.pci_dma_dev_setup = pci_dma_dev_setup_dart;
+ ppc_md.pci_dma_bus_setup = pci_dma_bus_setup_dart;
- return;
- }
+ /* Setup pci_dma ops */
+ set_pci_dma_ops(&dma_iommu_ops);
+ return;
bail:
/* If init failed, use direct iommu and null setup functions */
- ppc_md.iommu_dev_setup = iommu_dev_setup_null;
- ppc_md.iommu_bus_setup = iommu_bus_setup_null;
+ ppc_md.pci_dma_dev_setup = NULL;
+ ppc_md.pci_dma_bus_setup = NULL;
/* Setup pci_dma ops */
- pci_direct_iommu_init();
+ set_pci_dma_ops(&dma_direct_ops);
+}
+
+#ifdef CONFIG_PM
+static void iommu_dart_save(void)
+{
+ memcpy(dart_copy, dart_vbase, 2*1024*1024);
+}
+
+static void iommu_dart_restore(void)
+{
+ memcpy(dart_vbase, dart_copy, 2*1024*1024);
+ dart_tlb_invalidate_all();
}
+static int __init iommu_init_late_dart(void)
+{
+ unsigned long tbasepfn;
+ struct page *p;
+
+ /* if no dart table exists then we won't need to save it
+ * and the area has also not been reserved */
+ if (!dart_tablebase)
+ return 0;
+
+ tbasepfn = __pa(dart_tablebase) >> PAGE_SHIFT;
+ register_nosave_region_late(tbasepfn,
+ tbasepfn + ((1<<24) >> PAGE_SHIFT));
+
+ /* For suspend we need to copy the dart contents because
+ * it is not part of the regular mapping (see above) and
+ * thus not saved automatically. The memory for this copy
+ * must be allocated early because we need 2 MB. */
+ p = alloc_pages(GFP_KERNEL, 21 - PAGE_SHIFT);
+ BUG_ON(!p);
+ dart_copy = page_address(p);
+
+ ppc_md.iommu_save = iommu_dart_save;
+ ppc_md.iommu_restore = iommu_dart_restore;
+
+ return 0;
+}
+
+late_initcall(iommu_init_late_dart);
+#endif
void __init alloc_dart_table(void)
{
- /* Only reserve DART space if machine has more than 2GB of RAM
+ /* Only reserve DART space if machine has more than 1GB of RAM
* or if requested with iommu=on on cmdline.
+ *
+ * 1GB of RAM is picked as limit because some default devices
+ * (i.e. Airport Extreme) have 30 bit address range limits.
*/
- if (lmb_end_of_DRAM() <= 0x80000000ull && !iommu_force_on)
+
+ if (iommu_is_off)
+ return;
+
+ if (!iommu_force_on && memblock_end_of_DRAM() <= 0x40000000ull)
return;
/* 512 pages (2MB) is max DART tablesize. */
@@ -341,7 +475,12 @@ void __init alloc_dart_table(void)
* will blow up an entire large page anyway in the kernel mapping
*/
dart_tablebase = (unsigned long)
- abs_to_virt(lmb_alloc_base(1UL<<24, 1UL<<24, 0x80000000L));
+ __va(memblock_alloc_base(1UL<<24, 1UL<<24, 0x80000000L));
+ /*
+ * The DART space is later unmapped from the kernel linear mapping and
+ * accessing dart_tablebase during kmemleak scanning will fault.
+ */
+ kmemleak_no_scan((void *)dart_tablebase);
printk(KERN_INFO "DART table allocated at: %lx\n", dart_tablebase);
}