SCSI errors on powerpc with 2.6.24-rc6-mm1

SCSI errors on powerpc with 2.6.24-rc6-mm1

[Balbir Singh]

Hi,

I've just seen this on my dmesg, this is new, never seen this before on
this box and it happens only with this version of the kernel.

In this configuration, the page size is set to 64K and I've enabled fake
NUMA nodes on PowerPC.

tce_buildmulti_pSeriesLP: plpar_tce_put failed. rc=-4
        index   = 0x4000002
        npages  = 0x0
        tce[0] val = 0x15ad0001
Call Trace:
[c00000000ffe74f0] [c0000000000491a4]
.tce_buildmulti_pSeriesLP+0x26c/0x2ac (unreliable)
[c00000000ffe75c0] [c0000000000295e4] .iommu_map_sg+0x1d4/0x418
[c00000000ffe76d0] [c000000000028664] .dma_iommu_map_sg+0x3c/0x50
[c00000000ffe7750] [c0000000003b6c30] .scsi_dma_map+0x70/0x94
[c00000000ffe77d0] [c0000000003dedbc] .ipr_queuecommand+0x300/0x500
[c00000000ffe7880] [c0000000003ae964] .scsi_dispatch_cmd+0x21c/0x2b8
[c00000000ffe7920] [c0000000003b67a0] .scsi_request_fn+0x310/0x460
[c00000000ffe79d0] [c00000000024ab90] .blk_run_queue+0x94/0xec
[c00000000ffe7a70] [c0000000003b3b08] .scsi_run_queue+0x24c/0x27c
[c00000000ffe7b20] [c0000000003b4424] .scsi_next_command+0x48/0x70
[c00000000ffe7bc0] [c0000000003b4b48] .scsi_end_request+0xbc/0xe4
[c00000000ffe7c60] [c0000000003b5294] .scsi_io_completion+0x170/0x3e8
[c00000000ffe7d40] [c0000000003ae0e4] .scsi_finish_command+0xb4/0xd4
[c00000000ffe7dd0] [c0000000003b584c] .scsi_softirq_done+0x114/0x138
[c00000000ffe7e60] [c00000000024af70] .blk_done_softirq+0xa0/0xd0
[c00000000ffe7ef0] [c00000000007a2a0] .__do_softirq+0xa8/0x164
[c00000000ffe7f90] [c000000000027edc] .call_do_softirq+0x14/0x24
[c00000003e183950] [c00000000000bdcc] .do_softirq+0x74/0xc0
[c00000003e1839e0] [c00000000007a450] .irq_exit+0x5c/0xac
[c00000003e183a60] [c00000000000c414] .do_IRQ+0x17c/0x1f4
[c00000003e183b00] [c000000000004c24] hardware_interrupt_entry+0x24/0x28
--- Exception: 501 at .ppc64_runlatch_off+0x28/0x60
    LR = .pseries_dedicated_idle_sleep+0xd8/0x1a4
[c00000003e183df0] [c000000000048494]
.pseries_dedicated_idle_sleep+0x78/0x1a4 (unreliable)
[c00000003e183e80] [c00000000001110c] .cpu_idle+0x10c/0x1e8
[c00000003e183f00] [c00000000002b5b0] .start_secondary+0x1b4/0x1d8
[c00000003e183f90] [c0000000000083c4] .start_secondary_prolog+0xc/0x10
ipr: 0:0:3:0: 8150: PCI bus error
ipr: IOASA Dump:
ipr: 00000000: 04418000 009000D0 0006FFD0 00000000
ipr: 00000010: 00000000 00000000 00000000 00000000
ipr: 00000020: 01060000 D0001814 01062A00 000F531E
ipr: 00000030: 00040000 00000000 303800B6 00000000
ipr: 00000040: 00000000 04060000 00CCDDDD AABBCCCC
ipr: 00000050: 0006FFD0 6800FFD0 7FFF0030 B9000010
ipr: 00000060: 40000000 80000800 08200000 000013FE
ipr: 00000070: 00000000 00000000 8C00000E 00000000
ipr: 00000080: 00000000 00000000 00000000 00000000

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[FUJITA Tomonori]

On Mon, 24 Dec 2007 10:18:50 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> Hi,
> 
> I've just seen this on my dmesg, this is new, never seen this before on
> this box and it happens only with this version of the kernel.
> 
> In this configuration, the page size is set to 64K and I've enabled fake
> NUMA nodes on PowerPC.
> 
> tce_buildmulti_pSeriesLP: plpar_tce_put failed. rc=-4
>         index   = 0x4000002
>         npages  = 0x0
>         tce[0] val = 0x15ad0001
> Call Trace:
> [c00000000ffe74f0] [c0000000000491a4]
> .tce_buildmulti_pSeriesLP+0x26c/0x2ac (unreliable)
> [c00000000ffe75c0] [c0000000000295e4] .iommu_map_sg+0x1d4/0x418
> [c00000000ffe76d0] [c000000000028664] .dma_iommu_map_sg+0x3c/0x50
> [c00000000ffe7750] [c0000000003b6c30] .scsi_dma_map+0x70/0x94
> [c00000000ffe77d0] [c0000000003dedbc] .ipr_queuecommand+0x300/0x500
> [c00000000ffe7880] [c0000000003ae964] .scsi_dispatch_cmd+0x21c/0x2b8
> [c00000000ffe7920] [c0000000003b67a0] .scsi_request_fn+0x310/0x460
> [c00000000ffe79d0] [c00000000024ab90] .blk_run_queue+0x94/0xec
> [c00000000ffe7a70] [c0000000003b3b08] .scsi_run_queue+0x24c/0x27c
> [c00000000ffe7b20] [c0000000003b4424] .scsi_next_command+0x48/0x70
> [c00000000ffe7bc0] [c0000000003b4b48] .scsi_end_request+0xbc/0xe4
> [c00000000ffe7c60] [c0000000003b5294] .scsi_io_completion+0x170/0x3e8
> [c00000000ffe7d40] [c0000000003ae0e4] .scsi_finish_command+0xb4/0xd4
> [c00000000ffe7dd0] [c0000000003b584c] .scsi_softirq_done+0x114/0x138
> [c00000000ffe7e60] [c00000000024af70] .blk_done_softirq+0xa0/0xd0
> [c00000000ffe7ef0] [c00000000007a2a0] .__do_softirq+0xa8/0x164
> [c00000000ffe7f90] [c000000000027edc] .call_do_softirq+0x14/0x24
> [c00000003e183950] [c00000000000bdcc] .do_softirq+0x74/0xc0
> [c00000003e1839e0] [c00000000007a450] .irq_exit+0x5c/0xac
> [c00000003e183a60] [c00000000000c414] .do_IRQ+0x17c/0x1f4
> [c00000003e183b00] [c000000000004c24] hardware_interrupt_entry+0x24/0x28
> --- Exception: 501 at .ppc64_runlatch_off+0x28/0x60
>     LR = .pseries_dedicated_idle_sleep+0xd8/0x1a4
> [c00000003e183df0] [c000000000048494]
> .pseries_dedicated_idle_sleep+0x78/0x1a4 (unreliable)
> [c00000003e183e80] [c00000000001110c] .cpu_idle+0x10c/0x1e8
> [c00000003e183f00] [c00000000002b5b0] .start_secondary+0x1b4/0x1d8
> [c00000003e183f90] [c0000000000083c4] .start_secondary_prolog+0xc/0x10

I might break the IOMMU code. Can you reproduce it easily? If so,
reverting my IOMMU patches (I've attached a patch to revert them) fix
the problem?

Thanks,

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index ff2a62d..59899b2 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -244,9 +244,6 @@ config IOMMU_VMERGE
 
 	  Most drivers don't have this problem; it is safe to say Y here.
 
-config IOMMU_HELPER
-	def_bool PPC64
-
 config HOTPLUG_CPU
 	bool "Support for enabling/disabling CPUs"
 	depends on SMP && HOTPLUG && EXPERIMENTAL && (PPC_PSERIES || PPC_PMAC)
diff --git a/arch/powerpc/kernel/dma_64.c b/arch/powerpc/kernel/dma_64.c
index 6fcb7cb..1806d96 100644
--- a/arch/powerpc/kernel/dma_64.c
+++ b/arch/powerpc/kernel/dma_64.c
@@ -31,8 +31,8 @@ static inline unsigned long device_to_mask(struct device *dev)
 static void *dma_iommu_alloc_coherent(struct device *dev, size_t size,
 				      dma_addr_t *dma_handle, gfp_t flag)
 {
-	return iommu_alloc_coherent(dev, dev->archdata.dma_data, size,
-				    dma_handle, device_to_mask(dev), flag,
+	return iommu_alloc_coherent(dev->archdata.dma_data, size, dma_handle,
+				    device_to_mask(dev), flag,
 				    dev->archdata.numa_node);
 }
 
@@ -52,7 +52,7 @@ static dma_addr_t dma_iommu_map_single(struct device *dev, void *vaddr,
 				       size_t size,
 				       enum dma_data_direction direction)
 {
-	return iommu_map_single(dev, dev->archdata.dma_data, vaddr, size,
+	return iommu_map_single(dev->archdata.dma_data, vaddr, size,
 			        device_to_mask(dev), direction);
 }
 
diff --git a/arch/powerpc/kernel/iommu.c b/arch/powerpc/kernel/iommu.c
index 18e8860..050e9ac 100644
--- a/arch/powerpc/kernel/iommu.c
+++ b/arch/powerpc/kernel/iommu.c
@@ -31,7 +31,6 @@
 #include <linux/string.h>
 #include <linux/dma-mapping.h>
 #include <linux/bitops.h>
-#include <linux/iommu-helper.h>
 #include <asm/io.h>
 #include <asm/prom.h>
 #include <asm/iommu.h>
@@ -82,19 +81,17 @@ static int __init setup_iommu(char *str)
 __setup("protect4gb=", setup_protect4gb);
 __setup("iommu=", setup_iommu);
 
-static unsigned long iommu_range_alloc(struct device *dev,
-				       struct iommu_table *tbl,
+static unsigned long iommu_range_alloc(struct iommu_table *tbl,
                                        unsigned long npages,
                                        unsigned long *handle,
                                        unsigned long mask,
                                        unsigned int align_order)
 { 
-	unsigned long n, end, start;
+	unsigned long n, end, i, start;
 	unsigned long limit;
 	int largealloc = npages > 15;
 	int pass = 0;
 	unsigned long align_mask;
-	unsigned long boundary_size;
 
 	align_mask = 0xffffffffffffffffl >> (64 - align_order);
 
@@ -139,17 +136,14 @@ static unsigned long iommu_range_alloc(struct device *dev,
 			start &= mask;
 	}
 
-	if (dev)
-		boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
-				      1 << IOMMU_PAGE_SHIFT);
-	else
-		boundary_size = ALIGN(1UL << 32, 1 << IOMMU_PAGE_SHIFT);
-	/* 4GB boundary for iseries_hv_alloc and iseries_hv_map */
+	n = find_next_zero_bit(tbl->it_map, limit, start);
+
+	/* Align allocation */
+	n = (n + align_mask) & ~align_mask;
+
+	end = n + npages;
 
-	n = iommu_area_alloc(tbl->it_map, limit, start, npages,
-			     tbl->it_offset, boundary_size >> IOMMU_PAGE_SHIFT,
-			     align_mask);
-	if (n == -1) {
+	if (unlikely(end >= limit)) {
 		if (likely(pass < 2)) {
 			/* First failure, just rescan the half of the table.
 			 * Second failure, rescan the other half of the table.
@@ -164,7 +158,14 @@ static unsigned long iommu_range_alloc(struct device *dev,
 		}
 	}
 
-	end = n + npages;
+	for (i = n; i < end; i++)
+		if (test_bit(i, tbl->it_map)) {
+			start = i+1;
+			goto again;
+		}
+
+	for (i = n; i < end; i++)
+		__set_bit(i, tbl->it_map);
 
 	/* Bump the hint to a new block for small allocs. */
 	if (largealloc) {
@@ -183,17 +184,16 @@ static unsigned long iommu_range_alloc(struct device *dev,
 	return n;
 }
 
-static dma_addr_t iommu_alloc(struct device *dev, struct iommu_table *tbl,
-			      void *page, unsigned int npages,
-			      enum dma_data_direction direction,
-			      unsigned long mask, unsigned int align_order)
+static dma_addr_t iommu_alloc(struct iommu_table *tbl, void *page,
+		       unsigned int npages, enum dma_data_direction direction,
+		       unsigned long mask, unsigned int align_order)
 {
 	unsigned long entry, flags;
 	dma_addr_t ret = DMA_ERROR_CODE;
 
 	spin_lock_irqsave(&(tbl->it_lock), flags);
 
-	entry = iommu_range_alloc(dev, tbl, npages, NULL, mask, align_order);
+	entry = iommu_range_alloc(tbl, npages, NULL, mask, align_order);
 
 	if (unlikely(entry == DMA_ERROR_CODE)) {
 		spin_unlock_irqrestore(&(tbl->it_lock), flags);
@@ -224,6 +224,7 @@ static void __iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
 			 unsigned int npages)
 {
 	unsigned long entry, free_entry;
+	unsigned long i;
 
 	entry = dma_addr >> IOMMU_PAGE_SHIFT;
 	free_entry = entry - tbl->it_offset;
@@ -245,7 +246,9 @@ static void __iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
 	}
 
 	ppc_md.tce_free(tbl, entry, npages);
-	iommu_area_free(tbl->it_map, free_entry, npages);
+	
+	for (i = 0; i < npages; i++)
+		__clear_bit(free_entry+i, tbl->it_map);
 }
 
 static void iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
@@ -309,8 +312,7 @@ int iommu_map_sg(struct device *dev, struct scatterlist *sglist,
 		/* Allocate iommu entries for that segment */
 		vaddr = (unsigned long) sg_virt(s);
 		npages = iommu_num_pages(vaddr, slen);
-		entry = iommu_range_alloc(dev, tbl, npages, &handle,
-					  mask >> IOMMU_PAGE_SHIFT, 0);
+		entry = iommu_range_alloc(tbl, npages, &handle, mask >> IOMMU_PAGE_SHIFT, 0);
 
 		DBG("  - vaddr: %lx, size: %lx\n", vaddr, slen);
 
@@ -448,6 +450,9 @@ void iommu_unmap_sg(struct iommu_table *tbl, struct scatterlist *sglist,
 struct iommu_table *iommu_init_table(struct iommu_table *tbl, int nid)
 {
 	unsigned long sz;
+	unsigned long start_index, end_index;
+	unsigned long entries_per_4g;
+	unsigned long index;
 	static int welcomed = 0;
 	struct page *page;
 
@@ -469,7 +474,6 @@ struct iommu_table *iommu_init_table(struct iommu_table *tbl, int nid)
 
 #ifdef CONFIG_CRASH_DUMP
 	if (ppc_md.tce_get) {
-		unsigned long index;
 		unsigned long tceval;
 		unsigned long tcecount = 0;
 
@@ -500,6 +504,23 @@ struct iommu_table *iommu_init_table(struct iommu_table *tbl, int nid)
 	ppc_md.tce_free(tbl, tbl->it_offset, tbl->it_size);
 #endif
 
+	/*
+	 * DMA cannot cross 4 GB boundary.  Mark last entry of each 4
+	 * GB chunk as reserved.
+	 */
+	if (protect4gb) {
+		entries_per_4g = 0x100000000l >> IOMMU_PAGE_SHIFT;
+
+		/* Mark the last bit before a 4GB boundary as used */
+		start_index = tbl->it_offset | (entries_per_4g - 1);
+		start_index -= tbl->it_offset;
+
+		end_index = tbl->it_size;
+
+		for (index = start_index; index < end_index - 1; index += entries_per_4g)
+			__set_bit(index, tbl->it_map);
+	}
+
 	if (!welcomed) {
 		printk(KERN_INFO "IOMMU table initialized, virtual merging %s\n",
 		       novmerge ? "disabled" : "enabled");
@@ -547,9 +568,9 @@ void iommu_free_table(struct iommu_table *tbl, const char *node_name)
  * need not be page aligned, the dma_addr_t returned will point to the same
  * byte within the page as vaddr.
  */
-dma_addr_t iommu_map_single(struct device *dev, struct iommu_table *tbl,
-			    void *vaddr, size_t size, unsigned long mask,
-			    enum dma_data_direction direction)
+dma_addr_t iommu_map_single(struct iommu_table *tbl, void *vaddr,
+		size_t size, unsigned long mask,
+		enum dma_data_direction direction)
 {
 	dma_addr_t dma_handle = DMA_ERROR_CODE;
 	unsigned long uaddr;
@@ -561,7 +582,7 @@ dma_addr_t iommu_map_single(struct device *dev, struct iommu_table *tbl,
 	npages = iommu_num_pages(uaddr, size);
 
 	if (tbl) {
-		dma_handle = iommu_alloc(dev, tbl, vaddr, npages, direction,
+		dma_handle = iommu_alloc(tbl, vaddr, npages, direction,
 					 mask >> IOMMU_PAGE_SHIFT, 0);
 		if (dma_handle == DMA_ERROR_CODE) {
 			if (printk_ratelimit())  {
@@ -593,9 +614,8 @@ void iommu_unmap_single(struct iommu_table *tbl, dma_addr_t dma_handle,
  * Returns the virtual address of the buffer and sets dma_handle
  * to the dma address (mapping) of the first page.
  */
-void *iommu_alloc_coherent(struct device *dev, struct iommu_table *tbl,
-			   size_t size,	dma_addr_t *dma_handle,
-			   unsigned long mask, gfp_t flag, int node)
+void *iommu_alloc_coherent(struct iommu_table *tbl, size_t size,
+		dma_addr_t *dma_handle, unsigned long mask, gfp_t flag, int node)
 {
 	void *ret = NULL;
 	dma_addr_t mapping;
@@ -629,7 +649,7 @@ void *iommu_alloc_coherent(struct device *dev, struct iommu_table *tbl,
 	/* Set up tces to cover the allocated range */
 	nio_pages = size >> IOMMU_PAGE_SHIFT;
 	io_order = get_iommu_order(size);
-	mapping = iommu_alloc(dev, tbl, ret, nio_pages, DMA_BIDIRECTIONAL,
+	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);
diff --git a/arch/powerpc/platforms/iseries/iommu.c b/arch/powerpc/platforms/iseries/iommu.c
index 11fa3c7..6a0c6f6 100644
--- a/arch/powerpc/platforms/iseries/iommu.c
+++ b/arch/powerpc/platforms/iseries/iommu.c
@@ -199,7 +199,7 @@ static struct iommu_table vio_iommu_table;
 
 void *iseries_hv_alloc(size_t size, dma_addr_t *dma_handle, gfp_t flag)
 {
-	return iommu_alloc_coherent(NULL, &vio_iommu_table, size, dma_handle,
+	return iommu_alloc_coherent(&vio_iommu_table, size, dma_handle,
 				DMA_32BIT_MASK, flag, -1);
 }
 EXPORT_SYMBOL_GPL(iseries_hv_alloc);
@@ -213,7 +213,7 @@ EXPORT_SYMBOL_GPL(iseries_hv_free);
 dma_addr_t iseries_hv_map(void *vaddr, size_t size,
 			enum dma_data_direction direction)
 {
-	return iommu_map_single(NULL, &vio_iommu_table, vaddr, size,
+	return iommu_map_single(&vio_iommu_table, vaddr, size,
 				DMA_32BIT_MASK, direction);
 }
 
diff --git a/include/asm-powerpc/iommu.h b/include/asm-powerpc/iommu.h
index 852e15f..a07a67c 100644
--- a/include/asm-powerpc/iommu.h
+++ b/include/asm-powerpc/iommu.h
@@ -85,13 +85,13 @@ extern int iommu_map_sg(struct device *dev, struct scatterlist *sglist,
 extern void iommu_unmap_sg(struct iommu_table *tbl, struct scatterlist *sglist,
 			   int nelems, enum dma_data_direction direction);
 
-extern void *iommu_alloc_coherent(struct device *dev, struct iommu_table *tbl,
-				  size_t size, dma_addr_t *dma_handle,
-				  unsigned long mask, gfp_t flag, int node);
+extern void *iommu_alloc_coherent(struct iommu_table *tbl, size_t size,
+				  dma_addr_t *dma_handle, unsigned long mask,
+				  gfp_t flag, int node);
 extern void iommu_free_coherent(struct iommu_table *tbl, size_t size,
 				void *vaddr, dma_addr_t dma_handle);
-extern dma_addr_t iommu_map_single(struct device *dev, struct iommu_table *tbl,
-				   void *vaddr, size_t size, unsigned long mask,
+extern dma_addr_t iommu_map_single(struct iommu_table *tbl, void *vaddr,
+				   size_t size, unsigned long mask,
 				   enum dma_data_direction direction);
 extern void iommu_unmap_single(struct iommu_table *tbl, dma_addr_t dma_handle,
 			       size_t size, enum dma_data_direction direction);

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[Balbir Singh]

FUJITA Tomonori wrote:
> On Mon, 24 Dec 2007 10:18:50 +0530
> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
[snip]

> 
> I might break the IOMMU code. Can you reproduce it easily? If so,
> reverting my IOMMU patches (I've attached a patch to revert them) fix
> the problem?

[snip]


Yes, this patch fixes the problem for me.
Tested-by: Balbir Singh <balbir@linux.vnet.ibm.com>

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[FUJITA Tomonori]

On Thu, 27 Dec 2007 10:08:25 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> FUJITA Tomonori wrote:
> > On Mon, 24 Dec 2007 10:18:50 +0530
> > Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> > 
> [snip]
> 
> > 
> > I might break the IOMMU code. Can you reproduce it easily? If so,
> > reverting my IOMMU patches (I've attached a patch to revert them) fix
> > the problem?
> 
> [snip]
> 
> 
> Yes, this patch fixes the problem for me.

Thanks, so you can reproduce it easily, right?

The problem is that I don't want to revert these changes. I'll see
how these changes cause the problem shortly.

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[Balbir Singh]

FUJITA Tomonori wrote:
> On Thu, 27 Dec 2007 10:08:25 +0530
> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
>> FUJITA Tomonori wrote:
>>> On Mon, 24 Dec 2007 10:18:50 +0530
>>> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
>>>
>> [snip]
>>
>>> I might break the IOMMU code. Can you reproduce it easily? If so,
>>> reverting my IOMMU patches (I've attached a patch to revert them) fix
>>> the problem?
>> [snip]
>>
>>
>> Yes, this patch fixes the problem for me.
> 
> Thanks, so you can reproduce it easily, right?
> 

Yes, quite easily

> The problem is that I don't want to revert these changes. I'll see
> how these changes cause the problem shortly.

I'll try and find some bandwidth to review/test the patches and help you
figure out the right solution.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[FUJITA Tomonori]

On Thu, 27 Dec 2007 11:15:37 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> FUJITA Tomonori wrote:
> > On Thu, 27 Dec 2007 10:08:25 +0530
> > Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> > 
> >> FUJITA Tomonori wrote:
> >>> On Mon, 24 Dec 2007 10:18:50 +0530
> >>> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> >>>
> >> [snip]
> >>
> >>> I might break the IOMMU code. Can you reproduce it easily? If so,
> >>> reverting my IOMMU patches (I've attached a patch to revert them) fix
> >>> the problem?
> >> [snip]
> >>
> >>
> >> Yes, this patch fixes the problem for me.
> > 
> > Thanks, so you can reproduce it easily, right?
> > 
> 
> Yes, quite easily
> 
> > The problem is that I don't want to revert these changes. I'll see
> > how these changes cause the problem shortly.
> 
> I'll try and find some bandwidth to review/test the patches and help you
> figure out the right solution.

Thanks,

Can you try this?


diff --git a/lib/iommu-helper.c b/lib/iommu-helper.c
index e7d8544..495575a 100644
--- a/lib/iommu-helper.c
+++ b/lib/iommu-helper.c
@@ -8,15 +8,20 @@
 static unsigned long find_next_zero_area(unsigned long *map,
 					 unsigned long size,
 					 unsigned long start,
-					 unsigned int nr)
+					 unsigned int nr,
+					 unsigned long align_mask)
 {
 	unsigned long index, end, i;
 again:
 	index = find_next_zero_bit(map, size, start);
+
+	/* Align allocation */
+	index = (index + align_mask) & ~align_mask;
+
 	end = index + nr;
-	if (end > size)
+	if (end >= size)
 		return -1;
-	for (i = index + 1; i < end; i++) {
+	for (i = index; i < end; i++) {
 		if (test_bit(i, map)) {
 			start = i+1;
 			goto again;
@@ -50,9 +55,8 @@ unsigned long iommu_area_alloc(unsigned long *map, unsigned long size,
 {
 	unsigned long index;
 again:
-	index = find_next_zero_area(map, size, start, nr);
+	index = find_next_zero_area(map, size, start, nr, align_mask);
 	if (index != -1) {
-		index = (index + align_mask) & ~align_mask;
 		if (is_span_boundary(index, nr, shift, boundary_size)) {
 			/* we could do more effectively */
 			start = index + 1;

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[Balbir Singh]

FUJITA Tomonori wrote:
[snip]
> Thanks,
> 
> Can you try this?
> 
> 
> diff --git a/lib/iommu-helper.c b/lib/iommu-helper.c
> index e7d8544..495575a 100644
> --- a/lib/iommu-helper.c
> +++ b/lib/iommu-helper.c
> @@ -8,15 +8,20 @@
>  static unsigned long find_next_zero_area(unsigned long *map,
>  					 unsigned long size,
>  					 unsigned long start,
> -					 unsigned int nr)
> +					 unsigned int nr,
> +					 unsigned long align_mask)
>  {
>  	unsigned long index, end, i;
>  again:
>  	index = find_next_zero_bit(map, size, start);
> +
> +	/* Align allocation */
> +	index = (index + align_mask) & ~align_mask;
> +
>  	end = index + nr;
> -	if (end > size)
> +	if (end >= size)
>  		return -1;
> -	for (i = index + 1; i < end; i++) {
> +	for (i = index; i < end; i++) {
>  		if (test_bit(i, map)) {
>  			start = i+1;
>  			goto again;
> @@ -50,9 +55,8 @@ unsigned long iommu_area_alloc(unsigned long *map, unsigned long size,
>  {
>  	unsigned long index;
>  again:
> -	index = find_next_zero_area(map, size, start, nr);
> +	index = find_next_zero_area(map, size, start, nr, align_mask);
>  	if (index != -1) {
> -		index = (index + align_mask) & ~align_mask;
>  		if (is_span_boundary(index, nr, shift, boundary_size)) {
>  			/* we could do more effectively */
>  			start = index + 1;

This on top of -mm? Or on top of the reverted iommu patch.


-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[FUJITA Tomonori]

On Fri, 28 Dec 2007 13:05:39 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> FUJITA Tomonori wrote:
> [snip]
> > Thanks,
> > 
> > Can you try this?
> > 
> > 
> > diff --git a/lib/iommu-helper.c b/lib/iommu-helper.c
> > index e7d8544..495575a 100644
> > --- a/lib/iommu-helper.c
> > +++ b/lib/iommu-helper.c
> > @@ -8,15 +8,20 @@
> >  static unsigned long find_next_zero_area(unsigned long *map,
> >  					 unsigned long size,
> >  					 unsigned long start,
> > -					 unsigned int nr)
> > +					 unsigned int nr,
> > +					 unsigned long align_mask)
> >  {
> >  	unsigned long index, end, i;
> >  again:
> >  	index = find_next_zero_bit(map, size, start);
> > +
> > +	/* Align allocation */
> > +	index = (index + align_mask) & ~align_mask;
> > +
> >  	end = index + nr;
> > -	if (end > size)
> > +	if (end >= size)
> >  		return -1;
> > -	for (i = index + 1; i < end; i++) {
> > +	for (i = index; i < end; i++) {
> >  		if (test_bit(i, map)) {
> >  			start = i+1;
> >  			goto again;
> > @@ -50,9 +55,8 @@ unsigned long iommu_area_alloc(unsigned long *map, unsigned long size,
> >  {
> >  	unsigned long index;
> >  again:
> > -	index = find_next_zero_area(map, size, start, nr);
> > +	index = find_next_zero_area(map, size, start, nr, align_mask);
> >  	if (index != -1) {
> > -		index = (index + align_mask) & ~align_mask;
> >  		if (is_span_boundary(index, nr, shift, boundary_size)) {
> >  			/* we could do more effectively */
> >  			start = index + 1;
> 
> This on top of -mm? Or on top of the reverted iommu patch.

Oops, it's for -mm.

Thanks,

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[Balbir Singh]

FUJITA Tomonori wrote:
> 
> Oops, it's for -mm.
> 

Hi,

I just tested this patch and it works fine for me so far.

Tested-by: Balbir Singh <balbir@linux.vnet.ibm.com>

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

Re: SCSI errors on powerpc with 2.6.24-rc6-mm1

[FUJITA Tomonori]

On Mon, 31 Dec 2007 17:57:35 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> FUJITA Tomonori wrote:
> > 
> > Oops, it's for -mm.
> > 
> 
> Hi,
> 
> I just tested this patch and it works fine for me so far.
> 
> Tested-by: Balbir Singh <balbir@linux.vnet.ibm.com>

Thanks!

I sent an updated patch including this fix several hours ago:

http://ozlabs.org/pipermail/linuxppc-dev/2007-December/048997.html