[POWERPC] Use 4kB iommu pages even on 64kB-page systems

The 10Gigabit ethernet device drivers appear to be able to chew
up all 256MB of TCE mappings on pSeries systems, as evidenced by
numerous error messages:

 iommu_alloc failed, tbl c0000000010d5c48 vaddr c0000000d875eff0 npages 1

Some experimentation indicates that this is essentially because
one 1500 byte ethernet MTU gets mapped as a 64K DMA region when
the large 64K pages are enabled. Thus, it doesn't take much to
exhaust all of the available DMA mappings for a high-speed card.

This patch changes the iommu allocator to work with its own
unique, distinct page size. Although the patch is long, its
actually quite simple: it just #defines a distinct IOMMU_PAGE_SIZE
and then uses this in all the places that matter.

As a side effect, it also dramatically improves network performance
on platforms with H-calls on iommu translation inserts/removes (since
we no longer call it 16 times for a 1500 bytes packet when the iommu HW
is still 4k).

In the future, we might want to make the IOMMU_PAGE_SIZE a variable
in the iommu_table instance, thus allowing support for different HW
page sizes in the iommu itself.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>

1 files changed, 45 insertions, 32 deletions

diff --git a/arch/powerpc/kernel/iommu.c b/arch/powerpc/kernel/iommu.c
index f88a2a675d90..ba6b7256084b 100644
--- a/arch/powerpc/kernel/iommu.c
+++ b/arch/powerpc/kernel/iommu.c
@@ -47,6 +47,17 @@ static int novmerge = 0;
 static int novmerge = 1;
 #endif
 
+static inline unsigned long iommu_num_pages(unsigned long vaddr,
+                                            unsigned long slen)
+{
+        unsigned long npages;
+
+        npages = IOMMU_PAGE_ALIGN(vaddr + slen) - (vaddr & IOMMU_PAGE_MASK);
+        npages >>= IOMMU_PAGE_SHIFT;
+
+        return npages;
+}
+
 static int __init setup_iommu(char *str)
 {
         if (!strcmp(str, "novmerge"))
@@ -178,10 +189,10 @@ static dma_addr_t iommu_alloc(struct iommu_table *tbl, void *page,
         }
 
         entry += tbl->it_offset;        /* Offset into real TCE table */
-        ret = entry << PAGE_SHIFT;      /* Set the return dma address */
+        ret = entry << IOMMU_PAGE_SHIFT;        /* Set the return dma address */
 
         /* Put the TCEs in the HW table */
-        ppc_md.tce_build(tbl, entry, npages, (unsigned long)page & PAGE_MASK,
+        ppc_md.tce_build(tbl, entry, npages, (unsigned long)page & IOMMU_PAGE_MASK,
                          direction);
 
 
@@ -203,7 +214,7 @@ static void __iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
         unsigned long entry, free_entry;
         unsigned long i;
 
-        entry = dma_addr >> PAGE_SHIFT;
+        entry = dma_addr >> IOMMU_PAGE_SHIFT;
         free_entry = entry - tbl->it_offset;
 
         if (((free_entry + npages) > tbl->it_size) ||
@@ -270,7 +281,7 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
         /* Init first segment length for backout at failure */
         outs->dma_length = 0;
 
-        DBG("mapping %d elements:\n", nelems);
+        DBG("sg mapping %d elements:\n", nelems);
 
         spin_lock_irqsave(&(tbl->it_lock), flags);
 
@@ -285,9 +296,8 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
                 }
                 /* Allocate iommu entries for that segment */
                 vaddr = (unsigned long)page_address(s->page) + s->offset;
-                npages = PAGE_ALIGN(vaddr + slen) - (vaddr & PAGE_MASK);
-                npages >>= PAGE_SHIFT;
-                entry = iommu_range_alloc(tbl, npages, &handle, mask >> PAGE_SHIFT, 0);
+                npages = iommu_num_pages(vaddr, slen);
+                entry = iommu_range_alloc(tbl, npages, &handle, mask >> IOMMU_PAGE_SHIFT, 0);
 
                 DBG("  - vaddr: %lx, size: %lx\n", vaddr, slen);
 
@@ -301,14 +311,14 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
 
                 /* Convert entry to a dma_addr_t */
                 entry += tbl->it_offset;
-                dma_addr = entry << PAGE_SHIFT;
-                dma_addr |= s->offset;
+                dma_addr = entry << IOMMU_PAGE_SHIFT;
+                dma_addr |= (s->offset & ~IOMMU_PAGE_MASK);
 
-                DBG("  - %lx pages, entry: %lx, dma_addr: %lx\n",
+                DBG("  - %lu pages, entry: %lx, dma_addr: %lx\n",
                             npages, entry, dma_addr);
 
                 /* Insert into HW table */
-                ppc_md.tce_build(tbl, entry, npages, vaddr & PAGE_MASK, direction);
+                ppc_md.tce_build(tbl, entry, npages, vaddr & IOMMU_PAGE_MASK, direction);
 
                 /* If we are in an open segment, try merging */
                 if (segstart != s) {
@@ -323,7 +333,7 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
                                 DBG("    can't merge, new segment.\n");
                         } else {
                                 outs->dma_length += s->length;
-                                DBG("    merged, new len: %lx\n", outs->dma_length);
+                                DBG("    merged, new len: %ux\n", outs->dma_length);
                         }
                 }
 
@@ -367,9 +377,8 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
                 if (s->dma_length != 0) {
                         unsigned long vaddr, npages;
 
-                        vaddr = s->dma_address & PAGE_MASK;
-                        npages = (PAGE_ALIGN(s->dma_address + s->dma_length) - vaddr)
-                                >> PAGE_SHIFT;
+                        vaddr = s->dma_address & IOMMU_PAGE_MASK;
+                        npages = iommu_num_pages(s->dma_address, s->dma_length);
                         __iommu_free(tbl, vaddr, npages);
                         s->dma_address = DMA_ERROR_CODE;
                         s->dma_length = 0;
@@ -398,8 +407,7 @@ void iommu_unmap_sg(struct iommu_table *tbl, struct scatterlist *sglist,
 
                 if (sglist->dma_length == 0)
                         break;
-                npages = (PAGE_ALIGN(dma_handle + sglist->dma_length)
-                          - (dma_handle & PAGE_MASK)) >> PAGE_SHIFT;
+                npages = iommu_num_pages(dma_handle,sglist->dma_length);
                 __iommu_free(tbl, dma_handle, npages);
                 sglist++;
         }
@@ -532,12 +540,11 @@ dma_addr_t iommu_map_single(struct iommu_table *tbl, void *vaddr,
         BUG_ON(direction == DMA_NONE);
 
         uaddr = (unsigned long)vaddr;
-        npages = PAGE_ALIGN(uaddr + size) - (uaddr & PAGE_MASK);
-        npages >>= PAGE_SHIFT;
+        npages = iommu_num_pages(uaddr, size);
 
         if (tbl) {
                 dma_handle = iommu_alloc(tbl, vaddr, npages, direction,
-                                         mask >> PAGE_SHIFT, 0);
+                                         mask >> IOMMU_PAGE_SHIFT, 0);
                 if (dma_handle == DMA_ERROR_CODE) {
                         if (printk_ratelimit())  {
                                 printk(KERN_INFO "iommu_alloc failed, "
@@ -545,7 +552,7 @@ dma_addr_t iommu_map_single(struct iommu_table *tbl, void *vaddr,
                                                 tbl, vaddr, npages);
                         }
                 } else
-                        dma_handle |= (uaddr & ~PAGE_MASK);
+                        dma_handle |= (uaddr & ~IOMMU_PAGE_MASK);
         }
 
         return dma_handle;
@@ -554,11 +561,14 @@ dma_addr_t iommu_map_single(struct iommu_table *tbl, void *vaddr,
 void iommu_unmap_single(struct iommu_table *tbl, dma_addr_t dma_handle,
                 size_t size, enum dma_data_direction direction)
 {
+        unsigned int npages;
+
         BUG_ON(direction == DMA_NONE);
 
-        if (tbl)
-                iommu_free(tbl, dma_handle, (PAGE_ALIGN(dma_handle + size) -
-                                        (dma_handle & PAGE_MASK)) >> PAGE_SHIFT);
+        if (tbl) {
+                npages = iommu_num_pages(dma_handle, size);
+                iommu_free(tbl, dma_handle, npages);
+        }
 }
 
 /* Allocates a contiguous real buffer and creates mappings over it.
@@ -570,11 +580,11 @@ void *iommu_alloc_coherent(struct iommu_table *tbl, size_t size,
 {
         void *ret = NULL;
         dma_addr_t mapping;
-        unsigned int npages, order;
+        unsigned int order;
+        unsigned int nio_pages, io_order;
         struct page *page;
 
         size = PAGE_ALIGN(size);
-        npages = size >> PAGE_SHIFT;
         order = get_order(size);
 
         /*
@@ -598,8 +608,10 @@ void *iommu_alloc_coherent(struct iommu_table *tbl, size_t size,
         memset(ret, 0, size);
 
         /* Set up tces to cover the allocated range */
-        mapping = iommu_alloc(tbl, ret, npages, DMA_BIDIRECTIONAL,
-                              mask >> PAGE_SHIFT, order);
+        nio_pages = size >> IOMMU_PAGE_SHIFT;
+        io_order = get_iommu_order(size);
+        mapping = iommu_alloc(tbl, ret, nio_pages, DMA_BIDIRECTIONAL,
+                              mask >> IOMMU_PAGE_SHIFT, io_order);
         if (mapping == DMA_ERROR_CODE) {
                 free_pages((unsigned long)ret, order);
                 return NULL;
@@ -611,12 +623,13 @@ void *iommu_alloc_coherent(struct iommu_table *tbl, size_t size,
 void iommu_free_coherent(struct iommu_table *tbl, size_t size,
                          void *vaddr, dma_addr_t dma_handle)
 {
-        unsigned int npages;
-
         if (tbl) {
+                unsigned int nio_pages;
+
+                size = PAGE_ALIGN(size);
+                nio_pages = size >> IOMMU_PAGE_SHIFT;
+                iommu_free(tbl, dma_handle, nio_pages);
                 size = PAGE_ALIGN(size);
-                npages = size >> PAGE_SHIFT;
-                iommu_free(tbl, dma_handle, npages);
                 free_pages((unsigned long)vaddr, get_order(size));
         }
 }