[PATCH] ppc64: support 64k pages

Adds a new CONFIG_PPC_64K_PAGES which, when enabled, changes the kernel
base page size to 64K.  The resulting kernel still boots on any
hardware.  On current machines with 4K pages support only, the kernel
will maintain 16 "subpages" for each 64K page transparently.

Note that while real 64K capable HW has been tested, the current patch
will not enable it yet as such hardware is not released yet, and I'm
still verifying with the firmware architects the proper to get the
information from the newer hypervisors.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

1 files changed, 42 insertions, 32 deletions

diff --git a/arch/powerpc/platforms/iseries/iommu.c b/arch/powerpc/platforms/iseries/iommu.c
index 1a6845b5c5a..bf081b34582 100644
--- a/arch/powerpc/platforms/iseries/iommu.c
+++ b/arch/powerpc/platforms/iseries/iommu.c
@@ -43,9 +43,12 @@ static void tce_build_iSeries(struct iommu_table *tbl, long index, long npages,
         u64 rc;
         union tce_entry tce;
 
+        index <<= TCE_PAGE_FACTOR;
+        npages <<= TCE_PAGE_FACTOR;
+
         while (npages--) {
                 tce.te_word = 0;
-                tce.te_bits.tb_rpn = virt_to_abs(uaddr) >> PAGE_SHIFT;
+                tce.te_bits.tb_rpn = virt_to_abs(uaddr) >> TCE_SHIFT;
 
                 if (tbl->it_type == TCE_VB) {
                         /* Virtual Bus */
@@ -66,7 +69,7 @@ static void tce_build_iSeries(struct iommu_table *tbl, long index, long npages,
                         panic("PCI_DMA: HvCallXm_setTce failed, Rc: 0x%lx\n",
                                         rc);
                 index++;
-                uaddr += PAGE_SIZE;
+                uaddr += TCE_PAGE_SIZE;
         }
 }
 
@@ -74,6 +77,9 @@ static void tce_free_iSeries(struct iommu_table *tbl, long index, long npages)
 {
         u64 rc;
 
+        npages <<= TCE_PAGE_FACTOR;
+        index <<= TCE_PAGE_FACTOR;
+
         while (npages--) {
                 rc = HvCallXm_setTce((u64)tbl->it_index, (u64)index, 0);
                 if (rc)
@@ -83,27 +89,6 @@ static void tce_free_iSeries(struct iommu_table *tbl, long index, long npages)
         }
 }
 
-#ifdef CONFIG_PCI
-/*
- * This function compares the known tables to find an iommu_table
- * that has already been built for hardware TCEs.
- */
-static struct iommu_table *iommu_table_find(struct iommu_table * tbl)
-{
-        struct pci_dn *pdn;
-
-        list_for_each_entry(pdn, &iSeries_Global_Device_List, Device_List) {
-                struct iommu_table *it = pdn->iommu_table;
-                if ((it != NULL) &&
-                    (it->it_type == TCE_PCI) &&
-                    (it->it_offset == tbl->it_offset) &&
-                    (it->it_index == tbl->it_index) &&
-                    (it->it_size == tbl->it_size))
-                        return it;
-        }
-        return NULL;
-}
-
 /*
  * Call Hv with the architected data structure to get TCE table info.
  * info. Put the returned data into the Linux representation of the
@@ -113,8 +98,10 @@ static struct iommu_table *iommu_table_find(struct iommu_table * tbl)
  * 2. TCE table per Bus.
  * 3. TCE Table per IOA.
  */
-static void iommu_table_getparms(struct pci_dn *pdn,
-                                 struct iommu_table* tbl)
+void iommu_table_getparms_iSeries(unsigned long busno,
+                                  unsigned char slotno,
+                                  unsigned char virtbus,
+                                  struct iommu_table* tbl)
 {
         struct iommu_table_cb *parms;
 
@@ -124,9 +111,9 @@ static void iommu_table_getparms(struct pci_dn *pdn,
 
         memset(parms, 0, sizeof(*parms));
 
-        parms->itc_busno = pdn->busno;
-        parms->itc_slotno = pdn->LogicalSlot;
-        parms->itc_virtbus = 0;
+        parms->itc_busno = busno;
+        parms->itc_slotno = slotno;
+        parms->itc_virtbus = virtbus;
 
         HvCallXm_getTceTableParms(iseries_hv_addr(parms));
 
@@ -134,17 +121,40 @@ static void iommu_table_getparms(struct pci_dn *pdn,
                 panic("PCI_DMA: parms->size is zero, parms is 0x%p", parms);
 
         /* itc_size is in pages worth of table, it_size is in # of entries */
-        tbl->it_size = (parms->itc_size * PAGE_SIZE) / sizeof(union tce_entry);
+        tbl->it_size = ((parms->itc_size * TCE_PAGE_SIZE) /
+                        sizeof(union tce_entry)) >> TCE_PAGE_FACTOR;
         tbl->it_busno = parms->itc_busno;
-        tbl->it_offset = parms->itc_offset;
+        tbl->it_offset = parms->itc_offset >> TCE_PAGE_FACTOR;
         tbl->it_index = parms->itc_index;
         tbl->it_blocksize = 1;
-        tbl->it_type = TCE_PCI;
+        tbl->it_type = virtbus ? TCE_VB : TCE_PCI;
 
         kfree(parms);
 }
 
 
+#ifdef CONFIG_PCI
+/*
+ * This function compares the known tables to find an iommu_table
+ * that has already been built for hardware TCEs.
+ */
+static struct iommu_table *iommu_table_find(struct iommu_table * tbl)
+{
+        struct pci_dn *pdn;
+
+        list_for_each_entry(pdn, &iSeries_Global_Device_List, Device_List) {
+                struct iommu_table *it = pdn->iommu_table;
+                if ((it != NULL) &&
+                    (it->it_type == TCE_PCI) &&
+                    (it->it_offset == tbl->it_offset) &&
+                    (it->it_index == tbl->it_index) &&
+                    (it->it_size == tbl->it_size))
+                        return it;
+        }
+        return NULL;
+}
+
+
 void iommu_devnode_init_iSeries(struct device_node *dn)
 {
         struct iommu_table *tbl;
@@ -152,7 +162,7 @@ void iommu_devnode_init_iSeries(struct device_node *dn)
 
         tbl = kmalloc(sizeof(struct iommu_table), GFP_KERNEL);
 
-        iommu_table_getparms(pdn, tbl);
+        iommu_table_getparms_iSeries(pdn->busno, pdn->LogicalSlot, 0, tbl);
 
         /* Look for existing tce table */
         pdn->iommu_table = iommu_table_find(tbl);