Attachment 312916 Details for Bug 298811 – RHEL4 fix for this issue

[patch] RHEL4 fix for this issue

298811-v3.patch (text/plain), 8.41 KB, created by Prarit Bhargava on 2008-07-29 18:19:57 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Prarit Bhargava

Created: 2008-07-29 18:19:57 UTC

Size: 8.41 KB

patch

obsolete

>pci_alloc_consistent/dma_alloc_coherent is supposed to return size aligned
>addresses.
>
>From Documentation/DMA-mapping.txt:
>
>"pci_alloc_consistent returns two values: the virtual address which you
>can use to access it from the CPU and dma_handle which you pass to the
>card.
>
>The cpu return address and the DMA bus master address are both
>guaranteed to be aligned to the smallest PAGE_SIZE order which
>is greater than or equal to the requested size.  This invariant
>exists (for example) to guarantee that if you allocate a chunk
>which is smaller than or equal to 64 kilobytes, the extent of the
>buffer you receive will not cross a 64K boundary."
>
>1.  Backport upstream iommu-helper code into RHEL5.
>
>While #1 was being done, two bugs were noted:
>
>2.  It is possible that alloc_iommu()'s boundary_size overflows as
>dma_get_seg_boundary can return 0xffffffff.  In that case, further usage of
>boundary_size triggers a BUG_ON() in the iommu code.
>
>3.  Fix the GART's alloc_iommu code to return a size aligned address.  Also fix
>an incorrect alignment calculation in the iommu-helper code.
>
>Upstream patch was sent to LKML & the pci maintainer.  It is currently in
>his the pci-2.6 tree, on branch for-linus which will be merged shortly.
>
>(The patch is in Jesse Barnes' tree as 074ec1727196c4a2ba2f909902d55ec5014e50c3
> -- I will update the BZ when it lands in Linus' tree)
>
>Patch was tested with test module provided in BZ 298811, and by customer
>who reported BZ 298811.
>
>Resolves BZ 298811.
>
>
>diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
>index b71be6c..e06e5d5 100644
>--- a/arch/x86_64/kernel/pci-gart.c
>+++ b/arch/x86_64/kernel/pci-gart.c
>@@ -22,6 +22,7 @@
> #include <linux/module.h>
> #include <linux/topology.h>
> #include <linux/interrupt.h>
>+#include <linux/iommu-helper.h>
> #include <asm/atomic.h>
> #include <asm/io.h>
> #include <asm/mtrr.h>
>@@ -101,7 +102,8 @@ AGPEXTERN __u32 *agp_gatt_table;
> static unsigned long next_bit;  /* protected by iommu_bitmap_lock */
> static int need_flush; 		/* global flush state. set for each gart wrap */
> static dma_addr_t dma_map_area(struct device *dev, unsigned long phys_mem,
>-			       size_t size, int dir, int do_panic);
>+			       size_t size, int dir, int do_panic,
>+			       u64 align_mask);
> 
> /* Dummy device used for NULL arguments (normally ISA). Better would
>    be probably a smaller DMA mask, but this is bug-to-bug compatible to i386. */
>@@ -111,37 +113,48 @@ static struct device fallback_dev = {
> 	.dma_mask = &fallback_dev.coherent_dma_mask,
> };
> 
>-static unsigned long alloc_iommu(int size) 
>-{ 	
>+static unsigned long alloc_iommu(struct device *dev, int size,
>+				 unsigned long mask)
>+{
> 	unsigned long offset, flags;
>+	unsigned long boundary_size;
>+	unsigned long base_index;
>+
>+	base_index = ALIGN(iommu_bus_base & 0xffffffff,
>+			   PAGE_SIZE) >> PAGE_SHIFT;
>+	boundary_size = ALIGN(0x100000000ULL, PAGE_SIZE) >> PAGE_SHIFT;
> 
>-	spin_lock_irqsave(&iommu_bitmap_lock, flags);	
>-	offset = find_next_zero_string(iommu_gart_bitmap,next_bit,iommu_pages,size);
>+	spin_lock_irqsave(&iommu_bitmap_lock, flags);
>+	offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, next_bit,
>+				  size, base_index, boundary_size, mask);
> 	if (offset == -1) {
> 		need_flush = 1;
>-	       	offset = find_next_zero_string(iommu_gart_bitmap,0,next_bit,size);
>+		offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, 0,
>+					  size, base_index, boundary_size,
>+					  mask);
> 	}
>-	if (offset != -1) { 
>-		set_bit_string(iommu_gart_bitmap, offset, size); 
>-		next_bit = offset+size; 
>-		if (next_bit >= iommu_pages) { 
>+	if (offset != -1) {
>+		next_bit = offset+size;
>+		if (next_bit >= iommu_pages) {
> 			next_bit = 0;
> 			need_flush = 1;
>-		} 
>-	} 
>+		}
>+	}
> 	if (iommu_fullflush)
> 		need_flush = 1;
>-	spin_unlock_irqrestore(&iommu_bitmap_lock, flags);      
>+	spin_unlock_irqrestore(&iommu_bitmap_lock, flags);
>+
> 	return offset;
>-} 
>+}
> 
> static void free_iommu(unsigned long offset, int size)
>-{ 
>+{
> 	unsigned long flags;
>+
> 	spin_lock_irqsave(&iommu_bitmap_lock, flags);
>-	__clear_bit_string(iommu_gart_bitmap, offset, size);
>+	iommu_area_free(iommu_gart_bitmap, offset, size);
> 	spin_unlock_irqrestore(&iommu_bitmap_lock, flags);
>-} 
>+}
> 
> /* 
>  * Use global flush state to avoid races with multiple flushers.
>@@ -252,7 +265,8 @@ dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle,
> 		}
> 	} 
> 
>-	*dma_handle = dma_map_area(dev, bus, size, PCI_DMA_BIDIRECTIONAL, 0);
>+	*dma_handle = dma_map_area(dev, bus, size, PCI_DMA_BIDIRECTIONAL, 0,
>+				   size - 1);
> 	if (*dma_handle == bad_dma_address)
> 		goto error; 
> 	flush_gart(dev);
>@@ -370,10 +384,12 @@ static inline int nonforced_iommu(struct device *dev, unsigned long addr, size_t
>  * Caller needs to check if the iommu is needed and flush.
>  */
> static dma_addr_t dma_map_area(struct device *dev, unsigned long phys_mem,
>-				size_t size, int dir, int do_panic)
>+			       size_t size, int dir, int do_panic,
>+			       u64 align_mask)
> { 
> 	unsigned long npages = to_pages(phys_mem, size);
>-	unsigned long iommu_page = alloc_iommu(npages);
>+	unsigned long palign_mask = align_mask >> PAGE_SHIFT;
>+	unsigned long iommu_page = alloc_iommu(dev, npages, palign_mask);
> 	int i;
> 	if (iommu_page == -1) {
> 		if (!nonforced_iommu(dev, phys_mem, size))
>@@ -408,7 +424,7 @@ dma_addr_t dma_map_single(struct device *dev, void *addr, size_t size, int dir)
> 	if (!need_iommu(dev, phys_mem, size))
> 		return phys_mem; 
> 
>-	bus = dma_map_area(dev, phys_mem, size, dir, 1);
>+	bus = dma_map_area(dev, phys_mem, size, dir, 1, 0);
> 	flush_gart(dev); 
> 	return bus; 
> } 
>@@ -427,7 +443,7 @@ static int dma_map_sg_nonforce(struct device *dev, struct scatterlist *sg,
> 		struct scatterlist *s = &sg[i];
> 		unsigned long addr = page_to_phys(s->page) + s->offset; 
> 		if (nonforced_iommu(dev, addr, s->length)) { 
>-			addr = dma_map_area(dev, addr, s->length, dir, 0);
>+			addr = dma_map_area(dev, addr, s->length, dir, 0, 0);
> 			if (addr == bad_dma_address) { 
> 				if (i > 0) 
> 					dma_unmap_sg(dev, sg, i, dir);
>@@ -444,10 +460,11 @@ static int dma_map_sg_nonforce(struct device *dev, struct scatterlist *sg,
> }
> 
> /* Map multiple scatterlist entries continuous into the first. */
>-static int __dma_map_cont(struct scatterlist *sg, int start, int stopat,
>-		      struct scatterlist *sout, unsigned long pages)
>+static int __dma_map_cont(struct device *dev, struct scatterlist *sg,
>+			  int start, int stopat, struct scatterlist *sout,
>+			  unsigned long pages)
> {
>-	unsigned long iommu_start = alloc_iommu(pages);
>+	unsigned long iommu_start = alloc_iommu(dev, pages, 0);
> 	unsigned long iommu_page = iommu_start; 
> 	int i;
> 
>@@ -482,9 +499,9 @@ static int __dma_map_cont(struct scatterlist *sg, int start, int stopat,
> 	return 0;
> }
> 
>-static inline int dma_map_cont(struct scatterlist *sg, int start, int stopat,
>-		      struct scatterlist *sout,
>-		      unsigned long pages, int need)
>+static inline int dma_map_cont(struct device *dev, struct scatterlist *sg,
>+			       int start, int stopat, struct scatterlist *sout,
>+		      	       unsigned long pages, int need)
> {
> 	if (!need) { 
> 		BUG_ON(stopat - start != 1);
>@@ -492,7 +509,7 @@ static inline int dma_map_cont(struct scatterlist *sg, int start, int stopat,
> 		sout->dma_length = sg[start].length; 
> 		return 0;
> 	} 
>-	return __dma_map_cont(sg, start, stopat, sout, pages);
>+	return __dma_map_cont(dev, sg, start, stopat, sout, pages);
> }
> 		
> /*
>@@ -533,8 +550,8 @@ int dma_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
> 			   boundary and the new one doesn't have an offset. */
> 			if (!iommu_merge || !nextneed || !need || s->offset ||
> 			    (ps->offset + ps->length) % PAGE_SIZE) { 
>-				if (dma_map_cont(sg, start, i, sg+out, pages,
>-						 need) < 0)
>+				if (dma_map_cont(dev, sg, start, i, sg+out,
>+						 pages, need) < 0)
> 					goto error;
> 				out++;
> 				pages = 0;
>@@ -545,7 +562,7 @@ int dma_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
> 		need = nextneed;
> 		pages += to_pages(s->offset, s->length);
> 	}
>-	if (dma_map_cont(sg, start, i, sg+out, pages, need) < 0)
>+	if (dma_map_cont(dev, sg, start, i, sg+out, pages, need) < 0)
> 		goto error;
> 	out++;
> 	flush_gart(dev);
>diff --git a/lib/Makefile b/lib/Makefile
>index 6b0defc..c01f7f8 100644
>--- a/lib/Makefile
>+++ b/lib/Makefile
>@@ -26,6 +26,8 @@ obj-$(CONFIG_GENERIC_IOMAP) += iomap.o
> obj-$(CONFIG_ZLIB_INFLATE) += zlib_inflate/
> obj-$(CONFIG_ZLIB_DEFLATE) += zlib_deflate/
> 
>+obj-$(CONFIG_GART_IOMMU) += iommu-helper.o
>+
> hostprogs-y	:= gen_crc32table
> clean-files	:= crc32table.h
>

Actions: View | Diff

Attachments on bug 298811: 216601 | 217641 | 217701 | 217711 | 246651 | 296071 | 296072 | 302376 | 310803 | 310807 | 310860 | 312159 | 312162 | 312163 | 312462 | 312916