[PATCH] IEEE1394 updates

This is against 2.5.25. Lots of fixes. Brings things inline with 2.5.25.
Probably makes the subsystem actually work now (well, it will work, but
I bet without it, it doesn't). Merged in changes from the current 2.5.x
source.

21 files changed, 1023 insertions, 513 deletions

diff --git a/drivers/ieee1394/Config.in b/drivers/ieee1394/Config.in
index 02c6ccb388af..f6d555d0080d 100644
--- a/drivers/ieee1394/Config.in
+++ b/drivers/ieee1394/Config.in
@@ -9,22 +9,33 @@ if [ "$CONFIG_PCI" = "y" -a "$CONFIG_EXPERIMENTAL" = "y" ]; then
     if [ "$CONFIG_IEEE1394" != "n" ]; then
 
 	comment "Device Drivers"
-	dep_tristate '  Texas Instruments PCILynx support' CONFIG_IEEE1394_PCILYNX $CONFIG_IEEE1394
-	if [ "$CONFIG_IEEE1394_PCILYNX" != "n" ]; then
-	    bool '    Use PCILynx local RAM' CONFIG_IEEE1394_PCILYNX_LOCALRAM
-	    bool '    Support for non-IEEE1394 local ports' CONFIG_IEEE1394_PCILYNX_PORTS
+
+	if [ "$CONFIG_I2C" = "n" -o "$CONFIG_I2C_ALGOBIT" = "n" ]; then
+	    comment '  Texas Instruments PCILynx requires I2C bit-banging'
+	else
+	    dep_tristate '  Texas Instruments PCILynx support' CONFIG_IEEE1394_PCILYNX $CONFIG_IEEE1394 $CONFIG_I2C $CONFIG_I2C_ALGOBIT
 	fi
+
+	# Non-maintained pcilynx options
+	# if [ "$CONFIG_IEEE1394_PCILYNX" != "n" ]; then
+	#     bool '    Use PCILynx local RAM' CONFIG_IEEE1394_PCILYNX_LOCALRAM
+	#     bool '    Support for non-IEEE1394 local ports' CONFIG_IEEE1394_PCILYNX_PORTS
+	# fi
+
 	dep_tristate '  OHCI-1394 support' CONFIG_IEEE1394_OHCI1394 $CONFIG_IEEE1394
 
 	comment "Protocol Drivers"
 	dep_tristate '  OHCI-1394 Video support' CONFIG_IEEE1394_VIDEO1394 $CONFIG_IEEE1394_OHCI1394
 	dep_tristate '  SBP-2 support (Harddisks etc.)' CONFIG_IEEE1394_SBP2 $CONFIG_SCSI $CONFIG_IEEE1394
+	if [ "$CONFIG_IEEE1394_SBP2" != "n" ]; then
+	    bool '    Enable Phys DMA support for SBP2 (Debug)' CONFIG_IEEE1394_SBP2_PHYS_DMA
+	fi
 	dep_tristate '  Ethernet over 1394' CONFIG_IEEE1394_ETH1394 $CONFIG_IEEE1394
 	dep_tristate '  OHCI-DV I/O support' CONFIG_IEEE1394_DV1394 $CONFIG_IEEE1394_OHCI1394
 	dep_tristate '  Raw IEEE1394 I/O support' CONFIG_IEEE1394_RAWIO $CONFIG_IEEE1394
 	dep_tristate '  IEC61883-1 Plug support' CONFIG_IEEE1394_CMP $CONFIG_IEEE1394
 
-	if [ "$CONFIG_IEEE1394_PCILYNX" != "n" ]; then
+	if [ "$CONFIG_IEEE1394_CMP" != "n" ]; then
 	    dep_tristate '    IEC61883-6 (Audio transmission) support' CONFIG_IEEE1394_AMDTP $CONFIG_IEEE1394_OHCI1394 $CONFIG_IEEE1394_CMP
 	fi
 
diff --git a/drivers/ieee1394/Makefile b/drivers/ieee1394/Makefile
index 59d51d5e518e..9237b30b450a 100644
--- a/drivers/ieee1394/Makefile
+++ b/drivers/ieee1394/Makefile
@@ -2,8 +2,11 @@
 # Makefile for the Linux IEEE 1394 implementation
 #
 
+O_TARGET := ieee1394drv.o
+
 export-objs := ieee1394_core.o ohci1394.o cmp.o
 
+list-multi := ieee1394.o
 ieee1394-objs := ieee1394_core.o ieee1394_transactions.o hosts.o \
 		 highlevel.o csr.o nodemgr.o
 
@@ -19,3 +22,6 @@ obj-$(CONFIG_IEEE1394_AMDTP) += amdtp.o
 obj-$(CONFIG_IEEE1394_CMP) += cmp.o
 
 include $(TOPDIR)/Rules.make
+
+ieee1394.o: $(ieee1394-objs)
+	$(LD) -r -o $@ $(ieee1394-objs)
diff --git a/drivers/ieee1394/amdtp.c b/drivers/ieee1394/amdtp.c
index 1f4a8eaef372..4b86e7f2e311 100644
--- a/drivers/ieee1394/amdtp.c
+++ b/drivers/ieee1394/amdtp.c
@@ -1110,7 +1110,7 @@ MODULE_LICENSE("GPL");
 
 static int __init amdtp_init_module (void)
 {
-	if (ieee1394_register_chardev(IEEE1394_MINOR_BLOCK_EXPERIMENTAL,
+	if (ieee1394_register_chardev(IEEE1394_MINOR_BLOCK_AMDTP,
 				      THIS_MODULE, &amdtp_fops)) {
 		HPSB_ERR("amdtp: unable to get minor device block");
  		return -EIO;
@@ -1120,7 +1120,7 @@ static int __init amdtp_init_module (void)
 						   &amdtp_highlevel_ops);
 	if (amdtp_highlevel == NULL) {
 		HPSB_ERR("amdtp: unable to register highlevel ops");
-		ieee1394_unregister_chardev(IEEE1394_MINOR_BLOCK_EXPERIMENTAL);
+		ieee1394_unregister_chardev(IEEE1394_MINOR_BLOCK_AMDTP);
 		return -EIO;
 	}
 
@@ -1132,7 +1132,7 @@ static int __init amdtp_init_module (void)
 static void __exit amdtp_exit_module (void)
 {
         hpsb_unregister_highlevel(amdtp_highlevel);
-        ieee1394_unregister_chardev(IEEE1394_MINOR_BLOCK_EXPERIMENTAL);
+        ieee1394_unregister_chardev(IEEE1394_MINOR_BLOCK_AMDTP);
 
 	HPSB_INFO("Unloaded AMDTP driver");
 }
diff --git a/drivers/ieee1394/dv1394-private.h b/drivers/ieee1394/dv1394-private.h
index fb30b90a46dc..3e6fff77c5cb 100644
--- a/drivers/ieee1394/dv1394-private.h
+++ b/drivers/ieee1394/dv1394-private.h
@@ -460,7 +460,7 @@ struct video_card {
 	   This is a regular int, but use test_and_set_bit() (on bit zero) 
 	   for atomicity.
 	 */
-	int open;
+	unsigned long open;
 
 	/* 
 	   2) the spinlock - this provides mutual exclusion between the interrupt
diff --git a/drivers/ieee1394/dv1394.c b/drivers/ieee1394/dv1394.c
index 355f274b46e0..35ae739c81a4 100644
--- a/drivers/ieee1394/dv1394.c
+++ b/drivers/ieee1394/dv1394.c
@@ -53,6 +53,7 @@
     via pci_alloc_consistent()
     
   DONE:
+  - restart IT DMA after a bus reset
   - safely obtain and release ISO Tx channels in cooperation with OHCI driver
   - map received DIF blocks to their proper location in DV frame (ensure
     recovery if dropped packet)
@@ -192,71 +193,19 @@ static inline struct video_card* file_to_video_card(struct file *file)
 /* Memory management functions */
 /*******************************/
 
-#define MDEBUG(x)	do { } while(0)		/* Debug memory management */
-
-/* [DaveM] I've recoded most of this so that:
- * 1) It's easier to tell what is happening
- * 2) It's more portable, especially for translating things
- *    out of vmalloc mapped areas in the kernel.
- * 3) Less unnecessary translations happen.
- *
- * The code used to assume that the kernel vmalloc mappings
- * existed in the page tables of every process, this is simply
- * not guarenteed.  We now use pgd_offset_k which is the
- * defined way to get at the kernel page tables.
- */
-
-/* Given PGD from the address space's page table, return the kernel
- * virtual mapping of the physical memory mapped at ADR.
- */
-static inline struct page *uvirt_to_page(pgd_t *pgd, unsigned long adr)
-{
-	pmd_t *pmd;
-	pte_t *ptep, pte;
-	struct page *ret = NULL;
-	
-	if (!pgd_none(*pgd)) {
-                pmd = pmd_offset(pgd, adr);
-                if (!pmd_none(*pmd)) {
-                        ptep = pte_offset_kernel(pmd, adr);
-                        pte = *ptep;
-                        if(pte_present(pte))
-				ret = pte_page(pte);
-                }
-        }
-	return ret;
-}
-
-/* Here we want the physical address of the memory.
- * This is used when initializing the contents of the
- * area and marking the pages as reserved, and for
- * handling page faults on the rvmalloc()ed buffer
- */
-static inline unsigned long kvirt_to_pa(unsigned long adr) 
-{
-        unsigned long va, kva, ret;
-
-        va = VMALLOC_VMADDR(adr);
-	kva = (unsigned long) page_address(uvirt_to_page(pgd_offset_k(va), va));
-	kva |= adr & (PAGE_SIZE-1); /* restore the offset */
-	ret = __pa(kva);
-        MDEBUG(printk("kv2pa(%lx-->%lx)", adr, ret));
-        return ret;
-}
-
 static void * rvmalloc(unsigned long size)
 {
 	void * mem;
-	unsigned long adr, page;
-        
+	unsigned long adr;
+
+	size = PAGE_ALIGN(size);
 	mem=vmalloc_32(size);
 	if (mem) {
 		memset(mem, 0, size); /* Clear the ram out, 
 					 no junk to the user */
 	        adr=(unsigned long) mem;
 		while (size > 0) {
-	                page = kvirt_to_pa(adr);
-			mem_map_reserve(virt_to_page(__va(page)));
+			mem_map_reserve(vmalloc_to_page((void *)adr));
 			adr+=PAGE_SIZE;
 			size-=PAGE_SIZE;
 		}
@@ -266,13 +215,12 @@ static void * rvmalloc(unsigned long size)
 
 static void rvfree(void * mem, unsigned long size)
 {
-        unsigned long adr, page;
-        
+        unsigned long adr;
+
 	if (mem) {
 	        adr=(unsigned long) mem;
 		while (size > 0) {
-	                page = kvirt_to_pa(adr);
-			mem_map_unreserve(virt_to_page(__va(page)));
+			mem_map_unreserve(vmalloc_to_page((void *)adr));
 			adr+=PAGE_SIZE;
 			size-=PAGE_SIZE;
 		}
@@ -1166,9 +1114,9 @@ static int do_dv1394_init(struct video_card *video, struct dv1394_init *init)
 		
 		/* fill the sglist with the kernel addresses of pages in the non-contiguous buffer */
 		for(i = 0; i < video->user_dma.n_pages; i++) {
-			unsigned long va = VMALLOC_VMADDR( (unsigned long) video->user_buf + i * PAGE_SIZE );
+			unsigned long va = (unsigned long) video->user_buf + i * PAGE_SIZE;
 			
-			video->user_dma.sglist[i].page = uvirt_to_page(pgd_offset_k(va), va);
+			video->user_dma.sglist[i].page = vmalloc_to_page((void *)va);
 			video->user_dma.sglist[i].length = PAGE_SIZE;
 		}
 		
@@ -1492,7 +1440,7 @@ static int do_dv1394_shutdown(struct video_card *video, int free_user_buf)
 static struct page * dv1394_nopage(struct vm_area_struct * area, unsigned long address, int write_access)
 {
 	unsigned long offset;
-	unsigned long page, kernel_virt_addr;
+	unsigned long kernel_virt_addr;
 	struct page *ret = NOPAGE_SIGBUS;
 
 	struct video_card *video = (struct video_card*) area->vm_private_data;
@@ -1510,10 +1458,7 @@ static struct page * dv1394_nopage(struct vm_area_struct * area, unsigned long a
 
 	offset = address - area->vm_start;
 	kernel_virt_addr = (unsigned long) video->user_buf + offset;
-
-	page = kvirt_to_pa(kernel_virt_addr);
-	
-	ret = virt_to_page(__va(page));
+	ret = vmalloc_to_page((void *)kernel_virt_addr);
 	get_page(ret);
 
  out:
@@ -2936,9 +2881,129 @@ static void dv1394_add_host (struct hpsb_host *host)
 	dv1394_init(ohci, DV1394_PAL, MODE_TRANSMIT);
 }
 
+
+/* Bus reset handler. In the event of a bus reset, we may need to
+   re-start the DMA contexts - otherwise the user program would
+   end up waiting forever.
+*/
+
+static void dv1394_host_reset(struct hpsb_host *host)
+{
+	struct ti_ohci *ohci;
+	struct video_card *video = NULL;
+	unsigned long flags;
+	struct list_head *lh;
+	
+	/* We only work with the OHCI-1394 driver */
+	if (strcmp(host->driver->name, OHCI1394_DRIVER_NAME))
+		return;
+
+	ohci = (struct ti_ohci *)host->hostdata;
+
+
+	/* find the corresponding video_cards */
+	spin_lock_irqsave(&dv1394_cards_lock, flags);
+	if(!list_empty(&dv1394_cards)) {
+		list_for_each(lh, &dv1394_cards) {
+			video = list_entry(lh, struct video_card, list);
+			if((video->id >> 2) == ohci->id)
+				break;
+		}
+	}
+	spin_unlock_irqrestore(&dv1394_cards_lock, flags);
+
+	if(!video)
+		return;
+
+	/* check IT context */
+	if(video->ohci_it_ctx != -1) {
+		u32 ctx;
+		
+		spin_lock_irqsave(&video->spinlock, flags);
+		
+		ctx = reg_read(video->ohci, video->ohci_IsoXmitContextControlSet);
+
+		/* if(RUN but not ACTIVE) */
+		if( (ctx & (1<<15)) &&
+		    !(ctx & (1<<10)) ) {
+
+			debug_printk("dv1394: IT context stopped due to bus reset; waking it up\n");
+
+			/* to be safe, assume a frame has been dropped. User-space programs
+			   should handle this condition like an underflow. */
+			video->dropped_frames++;
+			
+			/* for some reason you must clear, then re-set the RUN bit to restart DMA */
+			
+			/* clear RUN */
+			reg_write(video->ohci, video->ohci_IsoXmitContextControlClear, (1 << 15));
+			flush_pci_write(video->ohci);
+			
+			/* set RUN */
+			reg_write(video->ohci, video->ohci_IsoXmitContextControlSet, (1 << 15));
+			flush_pci_write(video->ohci);
+			
+			/* set the WAKE bit (just in case; this isn't strictly necessary) */
+			reg_write(video->ohci, video->ohci_IsoXmitContextControlSet, (1 << 12));
+			flush_pci_write(video->ohci);
+
+			irq_printk("dv1394: AFTER IT restart ctx 0x%08x ptr 0x%08x\n",
+				   reg_read(video->ohci, video->ohci_IsoXmitContextControlSet),
+				   reg_read(video->ohci, video->ohci_IsoXmitCommandPtr));
+		}
+
+		spin_unlock_irqrestore(&video->spinlock, flags);
+	}
+	
+	/* check IR context */
+	if(video->ohci_ir_ctx != -1) {
+		u32 ctx;
+		
+		spin_lock_irqsave(&video->spinlock, flags);
+		
+		ctx = reg_read(video->ohci, video->ohci_IsoRcvContextControlSet);
+
+		/* if(RUN but not ACTIVE) */
+		if( (ctx & (1<<15)) &&
+		    !(ctx & (1<<10)) ) {
+
+			debug_printk("dv1394: IR context stopped due to bus reset; waking it up\n");
+
+			/* to be safe, assume a frame has been dropped. User-space programs
+			   should handle this condition like an overflow. */
+			video->dropped_frames++;
+
+			/* for some reason you must clear, then re-set the RUN bit to restart DMA */
+			/* XXX this doesn't work for me, I can't get IR DMA to restart :[ */
+			
+			/* clear RUN */
+			reg_write(video->ohci, video->ohci_IsoRcvContextControlClear, (1 << 15));
+			flush_pci_write(video->ohci);
+			
+			/* set RUN */
+			reg_write(video->ohci, video->ohci_IsoRcvContextControlSet, (1 << 15));
+			flush_pci_write(video->ohci);
+			
+			/* set the WAKE bit (just in case; this isn't strictly necessary) */
+			reg_write(video->ohci, video->ohci_IsoRcvContextControlSet, (1 << 12));
+			flush_pci_write(video->ohci);
+
+			irq_printk("dv1394: AFTER IR restart ctx 0x%08x ptr 0x%08x\n",
+				   reg_read(video->ohci, video->ohci_IsoRcvContextControlSet),
+				   reg_read(video->ohci, video->ohci_IsoRcvCommandPtr));
+		}
+
+		spin_unlock_irqrestore(&video->spinlock, flags);
+	}
+
+	/* wake readers/writers/ioctl'ers */
+	wake_up_interruptible(&video->waitq);
+}
+
 static struct hpsb_highlevel_ops hl_ops = {
 	add_host:	dv1394_add_host,
 	remove_host:	dv1394_remove_host,
+	host_reset:     dv1394_host_reset,
 };
 
 
diff --git a/drivers/ieee1394/dv1394.h b/drivers/ieee1394/dv1394.h
index d427cfde6eb7..089851076c2e 100644
--- a/drivers/ieee1394/dv1394.h
+++ b/drivers/ieee1394/dv1394.h
@@ -123,6 +123,67 @@
    frame 0. Then call DV1394_SUBMIT_FRAMES to inform the device that
    it may transmit the new frames.
 
+   ERROR HANDLING
+
+   An error (buffer underflow/overflow or a break in the DV stream due
+   to a 1394 bus reset) can be detected by checking the dropped_frames
+   field of struct dv1394_status (obtained through the
+   DV1394_GET_STATUS ioctl).
+
+   The best way to recover from such an error is to re-initialize
+   dv1394, either by using the DV1394_INIT ioctl call, or closing the
+   file descriptor and opening it again. (note that you must unmap all
+   ringbuffer mappings when closing the file descriptor, or else
+   dv1394 will still be considered 'in use').
+
+   MAIN LOOP
+
+   For maximum efficiency and robustness against bus errors, you are
+   advised to model the main loop of your application after the
+   following pseudo-code example:
+
+   (checks of system call return values omitted for brevity; always
+   check return values in your code!)
+   
+   while( frames left ) {
+   
+    struct pollfd *pfd = ...;
+
+    pfd->fd = dv1394_fd;
+    pfd->revents = 0;
+    pfd->events = POLLOUT | POLLIN; (OUT for transmit, IN for receive)
+
+    (add other sources of I/O here)
+    
+    poll(pfd, 1, -1); (or select(); add a timeout if you want)
+
+    if(pfd->revents) {
+         struct dv1394_status status;
+	 
+         ioctl(dv1394_fd, DV1394_GET_STATUS, &status);
+
+	 if(status.dropped_frames > 0) {
+	      reset_dv1394();
+         } else {
+              for(int i = 0; i < status.n_clear_frames; i++) {
+	          copy_DV_frame();
+              }
+         }
+    }
+   }
+
+   where copy_DV_frame() reads or writes on the dv1394 file descriptor
+   (read/write mode) or copies data to/from the mmap ringbuffer and
+   then calls ioctl(DV1394_SUBMIT_FRAMES) to notify dv1394 that new
+   frames are availble (mmap mode).
+
+   reset_dv1394() is called in the event of a buffer
+   underflow/overflow or a halt in the DV stream (e.g. due to a 1394
+   bus reset). To guarantee recovery from the error, this function
+   should close the dv1394 file descriptor (and munmap() all
+   ringbuffer mappings, if you are using them), then re-open the
+   dv1394 device (and re-map the ringbuffer).
+   
 */
 
 
@@ -218,6 +279,13 @@ struct dv1394_init {
 	unsigned int syt_offset;
 };
 
+/* NOTE: you may only allocate the DV frame ringbuffer once each time
+   you open the dv1394 device. DV1394_INIT will fail if you call it a
+   second time with different 'n_frames' or 'format' arguments (which
+   would imply a different size for the ringbuffer). If you need a
+   different buffer size, simply close and re-open the device, then
+   initialize it with your new settings. */
+   
 /* Q: What are cip_n and cip_d? */
 
 /*
@@ -262,8 +330,9 @@ struct dv1394_status {
 	   ready to be filled with data */
 	unsigned int n_clear_frames;
 
-	/* how many times the DV output has underflowed
-	   since the last call to DV1394_GET_STATUS */
+	/* how many times the DV stream has underflowed, overflowed,
+	   or otherwise encountered an error, since the previous call
+	   to DV1394_GET_STATUS */
 	unsigned int dropped_frames;
 
 	/* N.B. The dropped_frames counter is only a lower bound on the actual
diff --git a/drivers/ieee1394/hosts.c b/drivers/ieee1394/hosts.c
index 5701f53f32a6..0bd840be7df9 100644
--- a/drivers/ieee1394/hosts.c
+++ b/drivers/ieee1394/hosts.c
@@ -124,7 +124,7 @@ struct hpsb_host *hpsb_alloc_host(struct hpsb_host_driver *drv, size_t extra)
 
         h = kmalloc(sizeof(struct hpsb_host) + extra, SLAB_KERNEL);
         if (!h) return NULL;
-        memset(h, 0, sizeof(struct hpsb_host));
+        memset(h, 0, sizeof(struct hpsb_host) + extra);
 
 	h->hostdata = h + 1;
         h->driver = drv;
diff --git a/drivers/ieee1394/hosts.h b/drivers/ieee1394/hosts.h
index 5488e8ecba43..58cf427609db 100644
--- a/drivers/ieee1394/hosts.h
+++ b/drivers/ieee1394/hosts.h
@@ -31,6 +31,7 @@ struct hpsb_host {
         u32 tlabel_pool[2];
         struct semaphore tlabel_count;
         spinlock_t tlabel_lock;
+	u32 tlabel_current;
 
         unsigned char iso_listen_count[64];
 
diff --git a/drivers/ieee1394/ieee1394_core.c b/drivers/ieee1394/ieee1394_core.c
index f72a0b9641af..347346599671 100644
--- a/drivers/ieee1394/ieee1394_core.c
+++ b/drivers/ieee1394/ieee1394_core.c
@@ -740,7 +740,8 @@ void abort_requests(struct hpsb_host *host)
         host->ops->devctl(host, CANCEL_REQUESTS, 0);
 
         spin_lock_irqsave(&host->pending_pkt_lock, flags);
-        list_splice_init(&host->pending_packets, &llist);
+        list_splice(&host->pending_packets, &llist);
+        INIT_LIST_HEAD(&host->pending_packets);
         spin_unlock_irqrestore(&host->pending_pkt_lock, flags);
 
         list_for_each(lh, &llist) {
@@ -905,17 +906,16 @@ static int ieee1394_dispatch_open(struct inode *inode, struct file *file)
 
 	/* printk("ieee1394_dispatch_open(%d)", blocknum); */
 
-	/* lock the whole kernel here, to prevent a driver from
-	   being unloaded between the file_ops lookup and the open */
-
-	lock_kernel();
-
 	read_lock(&ieee1394_chardevs_lock);
-	file_ops = ieee1394_chardevs[blocknum].file_ops;
 	module = ieee1394_chardevs[blocknum].module;
+	/* bump the reference count of the driver that
+	   will receive the open() */
+	INCREF(module);
+	file_ops = ieee1394_chardevs[blocknum].file_ops;
 	read_unlock(&ieee1394_chardevs_lock);
 
 	if(file_ops == NULL) {
+		DECREF(module);
 		goto out_fail;
 	}
 
@@ -923,10 +923,6 @@ static int ieee1394_dispatch_open(struct inode *inode, struct file *file)
 	   own file_operations */
 	file->f_op = file_ops;
 
-	/* bump the reference count of the driver that
-	   will receive the open() */
-	INCREF(module);
-	
 	/* at this point BOTH ieee1394 and the task-specific driver have
 	   an extra reference */
 
@@ -955,7 +951,6 @@ static int ieee1394_dispatch_open(struct inode *inode, struct file *file)
 		   and will be dropped by the VFS when the file is
 		   released. */
 		
-		unlock_kernel();
 		return 0;
 	}
 	       
@@ -965,7 +960,6 @@ out_fail:
 	   function returns. */
 	
 	file->f_op = &ieee1394_chardev_ops;
-	unlock_kernel();
 	return retval;
 
 #undef INCREF
diff --git a/drivers/ieee1394/ieee1394_core.h b/drivers/ieee1394/ieee1394_core.h
index e3d1975d977d..693c4d3b816b 100644
--- a/drivers/ieee1394/ieee1394_core.h
+++ b/drivers/ieee1394/ieee1394_core.h
@@ -182,6 +182,7 @@ void hpsb_packet_received(struct hpsb_host *host, quadlet_t *data, size_t size,
 #define IEEE1394_MINOR_BLOCK_RAW1394       0
 #define IEEE1394_MINOR_BLOCK_VIDEO1394     1
 #define IEEE1394_MINOR_BLOCK_DV1394        2
+#define IEEE1394_MINOR_BLOCK_AMDTP         3
 #define IEEE1394_MINOR_BLOCK_EXPERIMENTAL 15
 
 /* return the index (within a minor number block) of a file */
diff --git a/drivers/ieee1394/ieee1394_transactions.c b/drivers/ieee1394/ieee1394_transactions.c
index 5a63fe3571d1..c153cb724463 100644
--- a/drivers/ieee1394/ieee1394_transactions.c
+++ b/drivers/ieee1394/ieee1394_transactions.c
@@ -169,8 +169,9 @@ void fill_phy_packet(struct hpsb_packet *packet, quadlet_t data)
  */
 int get_tlabel(struct hpsb_host *host, nodeid_t nodeid, int wait)
 {
-	int tlabel;
+	int tlabel = 0;
 	unsigned long flags;
+	int found_tlabel = 0;
 
 	if (wait) {
 		down(&host->tlabel_count);
@@ -180,15 +181,18 @@ int get_tlabel(struct hpsb_host *host, nodeid_t nodeid, int wait)
 
 	spin_lock_irqsave(&host->tlabel_lock, flags);
 
-	if (host->tlabel_pool[0] != ~0) {
-		tlabel = ffz(host->tlabel_pool[0]);
-		host->tlabel_pool[0] |= 1 << tlabel;
-	} else {
-		tlabel = ffz(host->tlabel_pool[1]);
-		host->tlabel_pool[1] |= 1 << tlabel;
-		tlabel += 32;
+	while (!found_tlabel) {
+		tlabel = host->tlabel_current;
+		if (tlabel < 32 && !(host->tlabel_pool[0] & 1 << tlabel)) {
+			host->tlabel_pool[0] |= 1 << tlabel;
+			found_tlabel = 1;
+		} else if (!(host->tlabel_pool[1] & 1 << (tlabel - 32))) {
+			host->tlabel_pool[1] |= 1 << (tlabel - 32);
+			found_tlabel = 1;
+		}
+		host->tlabel_current = (host->tlabel_current + 1) % 64;
 	}
-
+	
 	spin_unlock_irqrestore(&host->tlabel_lock, flags);
 
 	return tlabel;
diff --git a/drivers/ieee1394/ieee1394_types.h b/drivers/ieee1394/ieee1394_types.h
index 9f348d0cbf2b..d5ce71d9dcfa 100644
--- a/drivers/ieee1394/ieee1394_types.h
+++ b/drivers/ieee1394/ieee1394_types.h
@@ -7,7 +7,6 @@
 #include <linux/version.h>
 #include <linux/list.h>
 #include <linux/init.h>
-#include <linux/string.h>
 #include <asm/byteorder.h>
 
 
diff --git a/drivers/ieee1394/nodemgr.c b/drivers/ieee1394/nodemgr.c
index 5f769564f33c..9960ca86c266 100644
--- a/drivers/ieee1394/nodemgr.c
+++ b/drivers/ieee1394/nodemgr.c
@@ -177,7 +177,7 @@ static int nodemgr_read_quadlet(struct hpsb_host *host,
 
 	for (i = 0; i < 3; i++) {
 		ret = hpsb_read(host, nodeid, generation, address, quad, 4);
-		if (ret != -EAGAIN)
+		if (!ret)
 			break;
 	}
 	*quad = be32_to_cpu(*quad);
@@ -985,12 +985,26 @@ static int read_businfo_block(struct hpsb_host *host, nodeid_t nodeid, unsigned
 	HPSB_INFO("Initiating ConfigROM request for node " NODE_BUS_FMT,
 		  NODE_BUS_ARGS(nodeid));
 #endif
+	/* 
+	 * Must retry a few times if config rom read returns zero (how long?). Will
+	 * not normally occur, but we should do the right thing. For example, with
+	 * some sbp2 devices, the bridge chipset cannot return valid config rom reads
+	 * immediately after power-on, since they need to detect the type of 
+	 * device attached (disk or CD-ROM).
+	 */
+	for (i = 0; i < 4; i++) {
+		if (nodemgr_read_quadlet(host, nodeid, generation,
+					 addr, &buffer[0]) < 0) {
+			HPSB_ERR("ConfigROM quadlet transaction error for node "
+				 NODE_BUS_FMT, NODE_BUS_ARGS(nodeid));
+			return -1;
+		}
+		if (buffer[0])
+			break;
 
-	if (nodemgr_read_quadlet(host, nodeid, generation,
-				 addr, &buffer[0]) < 0) {
-		HPSB_ERR("ConfigROM quadlet transaction error for node "
-			 NODE_BUS_FMT, NODE_BUS_ARGS(nodeid));
-		return -1;
+		set_current_state(TASK_INTERRUPTIBLE);
+		if (schedule_timeout (HZ/4))
+			return -1;
 	}
 
 	header_size = buffer[0] >> 24;
@@ -1051,10 +1065,15 @@ static void nodemgr_node_probe_one(struct hpsb_host *host,
 		return;
 
 	if (buffer[1] != IEEE1394_BUSID_MAGIC) {
-		/* This isn't a 1394 device */
-		HPSB_ERR("Node " NODE_BUS_FMT " isn't an IEEE 1394 device",
-			 NODE_BUS_ARGS(nodeid));
-		return;
+		/* This isn't a 1394 device, but we let it slide. There
+		 * was a report of a device with broken firmware which
+		 * reported '2394' instead of '1394', which is obviously a
+		 * mistake. One would hope that a non-1394 device never
+		 * gets connected to Firewire bus. If someone does, we
+		 * shouldn't be held responsible, so we'll allow it with a
+		 * warning.  */
+		HPSB_WARN("Node " NODE_BUS_FMT " has invalid busID magic [0x%08x]",
+			 NODE_BUS_ARGS(nodeid), buffer[1]);
 	}
 
 	guid = ((u64)buffer[3] << 32) | buffer[4];
@@ -1103,11 +1122,11 @@ static void nodemgr_node_probe(struct hpsb_host *host)
 	nodeid_t nodeid = LOCAL_BUS;
 	unsigned int generation;
 
-	/* Pause for 1 second, to make sure things settle down. If
+	/* Pause for 1/4 second, to make sure things settle down. If
 	 * schedule_timeout returns non-zero, it means we caught a signal
 	 * and need to return. */
 	set_current_state(TASK_INTERRUPTIBLE);
-	if (schedule_timeout (HZ))
+	if (schedule_timeout (HZ/4))
 		return;
 
 	/* Now get the generation in which the node ID's we collect
diff --git a/drivers/ieee1394/ohci1394.c b/drivers/ieee1394/ohci1394.c
index e0d1172fc3e4..e25f4ecdcf8a 100644
--- a/drivers/ieee1394/ohci1394.c
+++ b/drivers/ieee1394/ohci1394.c
@@ -37,7 +37,7 @@
  * . DMA error recovery
  *
  * Known bugs:
- * . Apple PowerBook detected but not working yet (still true?)
+ * . devctl BUS_RESET arg confusion (reset type or root holdoff?)
  */
 
 /* 
@@ -77,12 +77,6 @@
  *  . Config ROM generation
  */
 
-/* Issues:
- *
- * - devctl BUS_RESET should treat arg as reset type
- *
- */
-
 #include <linux/config.h>
 #include <linux/kernel.h>
 #include <linux/list.h>
@@ -160,17 +154,21 @@ printk(level "%s: " fmt "\n" , OHCI1394_DRIVER_NAME , ## args)
 printk(level "%s_%d: " fmt "\n" , OHCI1394_DRIVER_NAME, card , ## args)
 
 static char version[] __devinitdata =
-	"$Revision: 1.101 $ Ben Collins <bcollins@debian.org>";
+	"$Rev: 504 $ Ben Collins <bcollins@debian.org>";
 
 /* Module Parameters */
 MODULE_PARM(attempt_root,"i");
-MODULE_PARM_DESC(attempt_root, "Attempt to make the host root.");
+MODULE_PARM_DESC(attempt_root, "Attempt to make the host root (default = 0).");
 static int attempt_root = 0;
 
+MODULE_PARM(phys_dma,"i");
+MODULE_PARM_DESC(phys_dma, "Enable physical dma (default = 1).");
+static int phys_dma = 1;
+
 static void dma_trm_tasklet(unsigned long data);
 static void dma_trm_reset(struct dma_trm_ctx *d);
 
-static void __devexit ohci1394_pci_remove(struct pci_dev *pdev);
+static void ohci1394_pci_remove(struct pci_dev *pdev);
 
 static inline void ohci1394_run_irq_hooks(struct ti_ohci *ohci,
 					  quadlet_t isoRecvEvent, 
@@ -224,7 +222,7 @@ static u8 get_phy_reg(struct ti_ohci *ohci, u8 addr)
 
 	spin_lock_irqsave (&ohci->phy_reg_lock, flags);
 
-	reg_write(ohci, OHCI1394_PhyControl, (((u16)addr << 8) & 0x00000f00) | 0x00008000);
+	reg_write(ohci, OHCI1394_PhyControl, (addr << 8) | 0x00008000);
 
 	for (i = 0; i < OHCI_LOOP_COUNT; i++) {
 		if (reg_read(ohci, OHCI1394_PhyControl) & 0x80000000)
@@ -252,7 +250,7 @@ static void set_phy_reg(struct ti_ohci *ohci, u8 addr, u8 data)
 
 	spin_lock_irqsave (&ohci->phy_reg_lock, flags);
 
-	reg_write(ohci, OHCI1394_PhyControl, 0x00004000 | (((u16)addr << 8) & 0x00000f00) | data);
+	reg_write(ohci, OHCI1394_PhyControl, (addr << 8) | data | 0x00004000);
 
 	for (i = 0; i < OHCI_LOOP_COUNT; i++) {
 		r = reg_read(ohci, OHCI1394_PhyControl);
@@ -291,8 +289,6 @@ static void handle_selfid(struct ti_ohci *ohci, struct hpsb_host *host,
 	size_t size;
 	quadlet_t q0, q1;
 
-	mdelay(10);
-
 	/* Check status of self-id reception */
 
 	if (ohci->selfid_swap)
@@ -363,13 +359,6 @@ static void ohci_soft_reset(struct ti_ohci *ohci) {
 			break;
 		mdelay(1);
 	}
-
-	/* Now reenable LPS, since that's usually what we want after a
-	 * softreset anyway. Wait 50msec to make sure we have full link
-	 * enabled.  */
-	reg_write(ohci, OHCI1394_HCControlSet, 0x00080000);
-	mdelay(50);
-
 	DBGMSG (ohci->id, "Soft reset finished");
 }
 
@@ -488,14 +477,42 @@ static void ohci_initialize(struct ti_ohci *ohci)
 {
 	quadlet_t buf;
 
+	/* Start off with a soft reset, to clear everything to a sane
+	 * state. */
+	ohci_soft_reset(ohci);
+
+	/* Now enable LPS, which we need in order to start accessing
+	 * most of the registers.  In fact, on some cards (ALI M5251),
+	 * accessing registers in the SClk domain without LPS enabled
+	 * will lock up the machine.  Wait 50msec to make sure we have
+	 * full link enabled.  */
+	reg_write(ohci, OHCI1394_HCControlSet, 0x00080000);
+	mdelay(50);
+
+	/* Determine the number of available IR and IT contexts. */
+	ohci->nb_iso_rcv_ctx =
+		get_nb_iso_ctx(ohci, OHCI1394_IsoRecvIntMaskSet);
+	DBGMSG(ohci->id, "%d iso receive contexts available",
+	       ohci->nb_iso_rcv_ctx);
+
+	ohci->nb_iso_xmit_ctx = 
+		get_nb_iso_ctx(ohci, OHCI1394_IsoXmitIntMaskSet);
+	DBGMSG(ohci->id, "%d iso transmit contexts available",
+	       ohci->nb_iso_xmit_ctx);
+
+	/* Set the usage bits for non-existent contexts so they can't
+	 * be allocated */
+	ohci->ir_ctx_usage |= ~0 << ohci->nb_iso_rcv_ctx;
+	ohci->it_ctx_usage |= ~0 << ohci->nb_iso_xmit_ctx;
+	
 	spin_lock_init(&ohci->phy_reg_lock);
 	spin_lock_init(&ohci->event_lock);
   
 	/* Put some defaults to these undefined bus options */
 	buf = reg_read(ohci, OHCI1394_BusOptions);
-	buf |=  0x60000000; /* Enable CMC and ISC */
+	buf |=  0xE0000000; /* Enable IRMC, CMC and ISC */
 	buf &= ~0x00ff0000; /* XXX: Set cyc_clk_acc to zero for now */
-	buf &= ~0x98000000; /* Disable PMC, IRMC and BMC */
+	buf &= ~0x18000000; /* Disable PMC and BMC */
 	reg_write(ohci, OHCI1394_BusOptions, buf);
 
 	/* Set the bus number */
@@ -507,8 +524,10 @@ static void ohci_initialize(struct ti_ohci *ohci)
 	/* Clear link control register */
 	reg_write(ohci, OHCI1394_LinkControlClear, 0xffffffff);
   
-	/* Enable cycle timer and cycle master */
+	/* Enable cycle timer and cycle master and set the IRM
+	 * contender bit in our self ID packets. */
 	reg_write(ohci, OHCI1394_LinkControlSet, 0x00300000);
+	set_phy_reg_mask(ohci, 4, 0xc0);
 
 	/* Clear interrupt registers */
 	reg_write(ohci, OHCI1394_IntMaskClear, 0xffffffff);
@@ -610,7 +629,7 @@ static void ohci_initialize(struct ti_ohci *ohci)
 	      ((((buf) >> 16) & 0xf) + (((buf) >> 20) & 0xf) * 10),
 	      ((((buf) >> 4) & 0xf) + ((buf) & 0xf) * 10), ohci->dev->irq,
 	      pci_resource_start(ohci->dev, 0),
-	      pci_resource_start(ohci->dev, 0) + pci_resource_len(ohci->dev, 0),
+	      pci_resource_start(ohci->dev, 0) + OHCI1394_REGISTER_SIZE,
 	      ohci->max_packet_size);
 }
 
@@ -1242,15 +1261,6 @@ static void ohci_irq_handler(int irq, void *dev_id,
 				  0xffffffff);
 			reg_write(ohci,OHCI1394_AsReqFilterLoSet, 
 				  0xffffffff);
-			/* Turn on phys dma reception. We should
-			 * probably manage the filtering somehow, 
-			 * instead of blindly turning it on.  */
-			reg_write(ohci,OHCI1394_PhyReqFilterHiSet,
-				  0xffffffff);
-			reg_write(ohci,OHCI1394_PhyReqFilterLoSet,
-				  0xffffffff);
-                       	reg_write(ohci,OHCI1394_PhyUpperBound,
-				  0xffff0000);
 		} else
 			PRINT(KERN_ERR, ohci->id, 
 			      "SelfID received outside of bus reset sequence");
@@ -1262,6 +1272,31 @@ static void ohci_irq_handler(int irq, void *dev_id,
 		reg_write(ohci, OHCI1394_IntMaskSet, OHCI1394_busReset); 
 		spin_unlock_irqrestore(&ohci->event_lock, flags);
 		event &= ~OHCI1394_selfIDComplete;	
+
+		/* Turn on phys dma reception. We should
+		 * probably manage the filtering somehow, 
+		 * instead of blindly turning it on.  */
+
+		/*
+		 * CAUTION!
+		 * Some chips (TI TSB43AB22) won't take a value in
+		 * the PhyReqFilter register until after the IntEvent
+		 * is cleared for bus reset, and even then a short
+		 * delay is required.
+		 */
+		if (phys_dma) {
+			mdelay(1);
+			reg_write(ohci,OHCI1394_PhyReqFilterHiSet,
+				  0xffffffff);
+			reg_write(ohci,OHCI1394_PhyReqFilterLoSet,
+				  0xffffffff);
+			reg_write(ohci,OHCI1394_PhyUpperBound,
+				  0xffff0000);
+		}
+
+		DBGMSG(ohci->id, "PhyReqFilter=%08x%08x\n",
+		       reg_read(ohci,OHCI1394_PhyReqFilterHiSet),
+		       reg_read(ohci,OHCI1394_PhyReqFilterLoSet));
 	}
 
 	/* Make sure we handle everything, just in case we accidentally
@@ -1831,7 +1866,7 @@ static u16 ohci_crc16 (u32 *ptr, int length)
 
 	crc = 0;
 	for (; length > 0; length--) {
-		data = *ptr++;
+		data = be32_to_cpu(*ptr++);
 		for (shift = 28; shift >= 0; shift -= 4) {
 			sum = ((crc >> 12) ^ (data >> shift)) & 0x000f;
 			crc = (crc << 4) ^ (sum << 12) ^ (sum << 5) ^ sum;
@@ -2023,7 +2058,7 @@ static int __devinit ohci1394_pci_probe(struct pci_dev *dev,
 
 	struct hpsb_host *host;
 	struct ti_ohci *ohci;	/* shortcut to currently handled device */
-	unsigned long ohci_base, ohci_len;
+	unsigned long ohci_base;
 	int i;
 	
 	if (version_printed++ == 0)
@@ -2041,6 +2076,7 @@ static int __devinit ohci1394_pci_probe(struct pci_dev *dev,
 	ohci->id = card_id_counter++;
 	ohci->dev = dev;
 	ohci->host = host;
+	ohci->init_state = OHCI_INIT_ALLOC_HOST;
 	host->pdev = dev;
 	pci_set_drvdata(dev, ohci);
 
@@ -2056,76 +2092,54 @@ static int __devinit ohci1394_pci_probe(struct pci_dev *dev,
 	 * zero. Should this work? Obviously it's not defined what these
 	 * registers will read when they aren't supported. Bleh! */
 	if (dev->vendor == PCI_VENDOR_ID_APPLE && 
-		dev->device == PCI_DEVICE_ID_APPLE_UNI_N_FW) {
-			ohci->no_swap_incoming = 1;
-			ohci->selfid_swap = 0;
+	    dev->device == PCI_DEVICE_ID_APPLE_UNI_N_FW) {
+		ohci->no_swap_incoming = 1;
+		ohci->selfid_swap = 0;
 	} else
 		ohci->selfid_swap = 1;
 #endif
 
-	/* csr_config rom allocation */
-	ohci->csr_config_rom_cpu = 
-		pci_alloc_consistent(ohci->dev, OHCI_CONFIG_ROM_LEN,
-				     &ohci->csr_config_rom_bus);
-	OHCI_DMA_ALLOC("consistent csr_config_rom");
-	if (ohci->csr_config_rom_cpu == NULL)
-		FAIL(-ENOMEM, "Failed to allocate buffer config rom");
-
-
+	/* We hardwire the MMIO length, since some CardBus adaptors
+	 * fail to report the right length.  Anyway, the ohci spec
+	 * clearly says it's 2kb, so this shouldn't be a problem. */ 
 	ohci_base = pci_resource_start(dev, 0);
-	ohci_len = pci_resource_len(dev, 0);
-
-	if (!request_mem_region (ohci_base, ohci_len, OHCI1394_DRIVER_NAME))
-		FAIL(-ENOMEM, "MMIO resource (0x%lx@0x%lx) unavailable, aborting.",
-		     ohci_base, ohci_len);
-
-	ohci->registers = ioremap(ohci_base, ohci_len);
+	if (pci_resource_len(dev, 0) != OHCI1394_REGISTER_SIZE)
+		PRINT(KERN_WARNING, ohci->id, "Unexpected PCI resource length of %lx!",
+		      pci_resource_len(dev, 0));
+
+	/* Seems PCMCIA handles this internally. Not sure why. Seems
+	 * pretty bogus to force a driver to special case this.  */
+#ifndef PCMCIA
+	if (!request_mem_region (ohci_base, OHCI1394_REGISTER_SIZE, OHCI1394_DRIVER_NAME))
+		FAIL(-ENOMEM, "MMIO resource (0x%lx - 0x%lx) unavailable",
+		     ohci_base, ohci_base + OHCI1394_REGISTER_SIZE);
+#endif
+	ohci->init_state = OHCI_INIT_HAVE_MEM_REGION;
 
+	ohci->registers = ioremap(ohci_base, OHCI1394_REGISTER_SIZE);
 	if (ohci->registers == NULL)
 		FAIL(-ENXIO, "Failed to remap registers - card not accessible");
-
+	ohci->init_state = OHCI_INIT_HAVE_IOMAPPING;
 	DBGMSG(ohci->id, "Remapped memory spaces reg 0x%p", ohci->registers);
 
+	/* csr_config rom allocation */
+	ohci->csr_config_rom_cpu =
+		pci_alloc_consistent(ohci->dev, OHCI_CONFIG_ROM_LEN,
+				     &ohci->csr_config_rom_bus);
+	OHCI_DMA_ALLOC("consistent csr_config_rom");
+	if (ohci->csr_config_rom_cpu == NULL)
+		FAIL(-ENOMEM, "Failed to allocate buffer config rom");
+	ohci->init_state = OHCI_INIT_HAVE_CONFIG_ROM_BUFFER;
 
-	/* Start off with a softreset, to clear everything to a sane
-	 * state.  This will also set Link Power State (LPS), which we
-	 * need in order to start accessing most of the registers.  */
-	ohci_soft_reset(ohci);
-
-	/* determinte the number of available IR and IT contexts right away,
-	   because they need to be known for alloc_dma_*_ctx() */
-	ohci->nb_iso_rcv_ctx = 
-		get_nb_iso_ctx(ohci, OHCI1394_IsoRecvIntMaskSet);
-	DBGMSG(ohci->id, "%d iso receive contexts available",
-	       ohci->nb_iso_rcv_ctx);
-
-	ohci->ir_ctx_usage = 0;
-	
-	/* set the usage bits for non-existent contexts so they can't be allocated */
-	for(i = ohci->nb_iso_rcv_ctx; i < sizeof(ohci->ir_ctx_usage)*8; i++)
-		__set_bit(i, &ohci->ir_ctx_usage);
-	
-	ohci->nb_iso_xmit_ctx = 
-		get_nb_iso_ctx(ohci, OHCI1394_IsoXmitIntMaskSet);
-	DBGMSG(ohci->id, "%d iso transmit contexts available",
-	       ohci->nb_iso_xmit_ctx);
-
-	ohci->it_ctx_usage = 0;
-	
-	/* set the usage bits for non-existent contexts so they can't be allocated */
-	for(i = ohci->nb_iso_xmit_ctx; i < sizeof(ohci->it_ctx_usage)*8; i++)
-		__set_bit(i, &ohci->it_ctx_usage);
-
-	
-	/* 
-	 * self-id dma buffer allocation
-	 */
+	/* self-id dma buffer allocation */
 	ohci->selfid_buf_cpu = 
 		pci_alloc_consistent(ohci->dev, OHCI1394_SI_DMA_BUF_SIZE,
                       &ohci->selfid_buf_bus);
 	OHCI_DMA_ALLOC("consistent selfid_buf");
+	
 	if (ohci->selfid_buf_cpu == NULL)
 		FAIL(-ENOMEM, "Failed to allocate DMA buffer for self-id packets");
+	ohci->init_state = OHCI_INIT_HAVE_SELFID_BUFFER;
 
 	if ((unsigned long)ohci->selfid_buf_cpu & 0x1fff)
 		PRINT(KERN_INFO, ohci->id, "SelfID buffer %p is not aligned on "
@@ -2135,6 +2149,7 @@ static int __devinit ohci1394_pci_probe(struct pci_dev *dev,
 	/* No self-id errors at startup */
 	ohci->self_id_errors = 0;
 
+	ohci->init_state = OHCI_INIT_HAVE_TXRX_BUFFERS__MAYBE;
 	/* AR DMA request context allocation */
 	ohci->ar_req_context = 
 		alloc_dma_rcv_ctx(ohci, DMA_CTX_ASYNC_REQ, 0, AR_REQ_NUM_DESC,
@@ -2169,6 +2184,9 @@ static int __devinit ohci1394_pci_probe(struct pci_dev *dev,
 	if (ohci->at_resp_context == NULL)
 		FAIL(-ENOMEM, "Failed to allocate AT Resp context");
 
+	ohci->ir_ctx_usage = 0;
+	ohci->it_ctx_usage = 0;
+	
 	/* IR DMA context */
 	ohci->ir_context =
 		alloc_dma_rcv_ctx(ohci, DMA_CTX_ISO, 0, IR_NUM_DESC,
@@ -2199,85 +2217,76 @@ static int __devinit ohci1394_pci_probe(struct pci_dev *dev,
 			 OHCI1394_DRIVER_NAME, ohci))
 		FAIL(-ENOMEM, "Failed to allocate shared interrupt %d", dev->irq);
 
+	ohci->init_state = OHCI_INIT_HAVE_IRQ;
 	ohci_initialize(ohci);
 
 	/* Tell the highlevel this host is ready */
 	hpsb_add_host(host);
+	ohci->init_state = OHCI_INIT_DONE;
 
 	return 0;
 #undef FAIL
 }
 
-static void __devexit ohci1394_pci_remove(struct pci_dev *pdev)
+static void ohci1394_pci_remove(struct pci_dev *pdev)
 {
 	struct ti_ohci *ohci;
-	quadlet_t buf;
 
 	ohci = pci_get_drvdata(pdev);
 	if (!ohci)
 		return;
 
-	if (ohci->host)
+	switch (ohci->init_state) {
+	case OHCI_INIT_DONE:
 		hpsb_remove_host(ohci->host);
 
-	/* Soft reset before we start */
-	ohci_soft_reset(ohci);
-
-	/* Free AR dma */
-	free_dma_rcv_ctx(&ohci->ar_req_context);
-	free_dma_rcv_ctx(&ohci->ar_resp_context);
+	case OHCI_INIT_HAVE_IRQ:
+		/* Soft reset before we start - this disables
+		 * interrupts and clears linkEnable and LPS. */
+		ohci_soft_reset(ohci);
+		free_irq(ohci->dev->irq, ohci);
 
-	/* Free AT dma */
-	free_dma_trm_ctx(&ohci->at_req_context);
-	free_dma_trm_ctx(&ohci->at_resp_context);
+	case OHCI_INIT_HAVE_TXRX_BUFFERS__MAYBE:
+		/* Free AR dma */
+		free_dma_rcv_ctx(&ohci->ar_req_context);
+		free_dma_rcv_ctx(&ohci->ar_resp_context);
 
-	/* Free IR dma */
-	free_dma_rcv_ctx(&ohci->ir_context);
+		/* Free AT dma */
+		free_dma_trm_ctx(&ohci->at_req_context);
+		free_dma_trm_ctx(&ohci->at_resp_context);
 
-        /* Free IT dma */
-        free_dma_trm_ctx(&ohci->it_context);
-
-	/* Disable all interrupts */
-	reg_write(ohci, OHCI1394_IntMaskClear, 0x80000000);
-	free_irq(ohci->dev->irq, ohci);
-
-	/* Free self-id buffer */
-	if (ohci->selfid_buf_cpu) {
+		/* Free IR dma */
+		free_dma_rcv_ctx(&ohci->ir_context);
+		
+		/* Free IT dma */
+		free_dma_trm_ctx(&ohci->it_context);
+	
+	case OHCI_INIT_HAVE_SELFID_BUFFER:
 		pci_free_consistent(ohci->dev, OHCI1394_SI_DMA_BUF_SIZE, 
 				    ohci->selfid_buf_cpu,
 				    ohci->selfid_buf_bus);
 		OHCI_DMA_FREE("consistent selfid_buf");
-	}
-	
-	/* Free config rom */
-	if (ohci->csr_config_rom_cpu) {
+		
+	case OHCI_INIT_HAVE_CONFIG_ROM_BUFFER:
 		pci_free_consistent(ohci->dev, OHCI_CONFIG_ROM_LEN,
-				    ohci->csr_config_rom_cpu, 
+				    ohci->csr_config_rom_cpu,
 				    ohci->csr_config_rom_bus);
 		OHCI_DMA_FREE("consistent csr_config_rom");
-	}
-
-	/* Disable our bus options */
-	buf = reg_read(ohci, OHCI1394_BusOptions);
-	buf &= ~0xf8000000;
-	buf |=  0x00ff0000;
-	reg_write(ohci, OHCI1394_BusOptions, buf);
-
-	/* Clear LinkEnable and LPS */
-	reg_write(ohci, OHCI1394_HCControlClear, 0x000a0000);
 
-	if (ohci->registers)
+	case OHCI_INIT_HAVE_IOMAPPING:
 		iounmap(ohci->registers);
 
-	release_mem_region (pci_resource_start(ohci->dev, 0),
-			    pci_resource_len(ohci->dev, 0));
+	case OHCI_INIT_HAVE_MEM_REGION:
+#ifndef PCMCIA
+		release_mem_region(pci_resource_start(ohci->dev, 0),
+				   OHCI1394_REGISTER_SIZE);
+#endif
 
 #ifdef CONFIG_ALL_PPC
-	/* On UniNorth, power down the cable and turn off the
-	 * chip clock when the module is removed to save power
-	 * on laptops. Turning it back ON is done by the arch
-	 * code when pci_enable_device() is called
-	 */
+	/* On UniNorth, power down the cable and turn off the chip
+	 * clock when the module is removed to save power on
+	 * laptops. Turning it back ON is done by the arch code when
+	 * pci_enable_device() is called */
 	{
 		struct device_node* of_node;
 
@@ -2289,8 +2298,10 @@ static void __devexit ohci1394_pci_remove(struct pci_dev *pdev)
 	}
 #endif /* CONFIG_ALL_PPC */
 
-	pci_set_drvdata(ohci->dev, NULL);
-	hpsb_unref_host(ohci->host);
+	case OHCI_INIT_ALLOC_HOST:
+		pci_set_drvdata(ohci->dev, NULL);
+		hpsb_unref_host(ohci->host);
+	}
 }
 
 #define PCI_CLASS_FIREWIRE_OHCI     ((PCI_CLASS_SERIAL_FIREWIRE << 8) | 0x10)
@@ -2313,7 +2324,7 @@ static struct pci_driver ohci1394_pci_driver = {
 	name:		OHCI1394_DRIVER_NAME,
 	id_table:	ohci1394_pci_tbl,
 	probe:		ohci1394_pci_probe,
-	remove:		__devexit_p(ohci1394_pci_remove),
+	remove:		ohci1394_pci_remove,
 };
 
 
diff --git a/drivers/ieee1394/ohci1394.h b/drivers/ieee1394/ohci1394.h
index 7d2bb3396bef..61e4988055b6 100644
--- a/drivers/ieee1394/ohci1394.h
+++ b/drivers/ieee1394/ohci1394.h
@@ -21,6 +21,8 @@
 #ifndef _OHCI1394_H
 #define _OHCI1394_H
 
+#include <asm/io.h>
+
 #include "ieee1394_types.h"
 #include <asm/io.h>
 
@@ -144,7 +146,16 @@ struct ti_ohci {
 
         struct pci_dev *dev;
 
-        u32 state;
+	enum { 
+		OHCI_INIT_ALLOC_HOST,
+		OHCI_INIT_HAVE_MEM_REGION,
+		OHCI_INIT_HAVE_IOMAPPING,
+		OHCI_INIT_HAVE_CONFIG_ROM_BUFFER,
+		OHCI_INIT_HAVE_SELFID_BUFFER,
+		OHCI_INIT_HAVE_TXRX_BUFFERS__MAYBE,
+		OHCI_INIT_HAVE_IRQ,
+		OHCI_INIT_DONE,
+	} init_state;
         
         /* remapped memory spaces */
         void *registers; 
diff --git a/drivers/ieee1394/pcilynx.c b/drivers/ieee1394/pcilynx.c
index f04d4ba13c25..17d105d443e7 100644
--- a/drivers/ieee1394/pcilynx.c
+++ b/drivers/ieee1394/pcilynx.c
@@ -2,6 +2,7 @@
  * ti_pcilynx.c - Texas Instruments PCILynx driver
  * Copyright (C) 1999,2000 Andreas Bombe <andreas.bombe@munich.netsurf.de>,
  *                         Stephan Linz <linz@mazet.de>
+ *                         Manfred Weihs <weihs@ict.tuwien.ac.at>
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -41,6 +42,8 @@
 #include "highlevel.h"
 #include "pcilynx.h"
 
+#include <linux/i2c.h>
+#include <linux/i2c-algo-bit.h>
 
 /* print general (card independent) information */
 #define PRINT_G(level, fmt, args...) printk(level "pcilynx: " fmt "\n" , ## args)
@@ -56,9 +59,85 @@
 #endif
 
 
+/* Module Parameters */
+MODULE_PARM(skip_eeprom,"i");
+MODULE_PARM_DESC(skip_eeprom, "Do not try to read bus info block from serial eeprom, but user generic one (default = 0).");
+static int skip_eeprom = 0;
+
+
 static struct hpsb_host_driver *lynx_driver;
 static unsigned int card_id;
 
+
+
+/*
+ * I2C stuff
+ */
+
+/* the i2c stuff was inspired by i2c-philips-par.c */
+
+static void bit_setscl(void *data, int state)
+{
+	if (state) {
+		  ((struct ti_lynx *) data)->i2c_driven_state |= 0x00000040;
+	} else {
+		  ((struct ti_lynx *) data)->i2c_driven_state &= ~0x00000040;
+	}
+	reg_write((struct ti_lynx *) data, SERIAL_EEPROM_CONTROL, ((struct ti_lynx *) data)->i2c_driven_state);
+}
+
+static void bit_setsda(void *data, int state)
+{
+	if (state) {
+		  ((struct ti_lynx *) data)->i2c_driven_state |= 0x00000010;
+	} else {
+		  ((struct ti_lynx *) data)->i2c_driven_state &= ~0x00000010;
+	}
+	reg_write((struct ti_lynx *) data, SERIAL_EEPROM_CONTROL, ((struct ti_lynx *) data)->i2c_driven_state);
+}
+
+static int bit_getscl(void *data)
+{
+	return reg_read((struct ti_lynx *) data, SERIAL_EEPROM_CONTROL) & 0x00000040;
+}
+
+static int bit_getsda(void *data)
+{
+	return reg_read((struct ti_lynx *) data, SERIAL_EEPROM_CONTROL) & 0x00000010;
+}
+
+static int bit_reg(struct i2c_client *client)
+{
+	return 0;
+}
+
+static int bit_unreg(struct i2c_client *client)
+{
+	return 0;
+}
+
+static struct i2c_algo_bit_data bit_data = {
+	NULL,
+	bit_setsda,
+	bit_setscl,
+	bit_getsda,
+	bit_getscl,
+	5, 5, 100,		/*	waits, timeout */
+}; 
+
+static struct i2c_adapter bit_ops = {
+	"PCILynx I2C adapter",
+	0xAA, //FIXME: probably we should get an id in i2c-id.h
+	NULL,
+	NULL,
+	NULL,
+	NULL,
+	bit_reg,
+	bit_unreg,
+};
+
+
+
 /*
  * PCL handling functions.
  */
@@ -880,7 +959,7 @@ static ssize_t mem_read(struct file *file, char *buffer, size_t count,
         retval = copy_to_user(buffer, md->lynx->mem_dma_buffer, count);
         up(&md->lynx->mem_dma_mutex);
 
-        if (retval) return -EFAULT;
+        if (retval < 0) return retval;
         *offset += count;
         return count;
 }
@@ -1232,6 +1311,11 @@ static int __devinit add_card(struct pci_dev *dev,
         int i;
         int error;
 
+        /* needed for i2c communication with serial eeprom */
+        struct i2c_adapter i2c_adapter;
+        struct i2c_algo_bit_data i2c_adapter_data;
+
+        int got_valid_bus_info_block = 0; /* set to 1, if we were able to get a valid bus info block from serial eeprom */
 
         error = -ENXIO;
 
@@ -1492,6 +1576,94 @@ static int __devinit add_card(struct pci_dev *dev,
                 if (i != -1) set_phy_reg(lynx, 4, i | 0x40);
         }
 
+
+        if (!skip_eeprom)
+        {
+                i2c_adapter = bit_ops;
+                i2c_adapter_data = bit_data;
+                i2c_adapter.algo_data = &i2c_adapter_data;
+                i2c_adapter_data.data = lynx;
+
+#ifdef CONFIG_IEEE1394_VERBOSEDEBUG
+                PRINT(KERN_DEBUG, lynx->id,"original eeprom control: %d",reg_read(lynx,SERIAL_EEPROM_CONTROL));
+#endif
+
+        	/* reset hardware to sane state */
+        	lynx->i2c_driven_state = 0x00000070;
+        	reg_write(lynx, SERIAL_EEPROM_CONTROL, lynx->i2c_driven_state);
+
+        	if (i2c_bit_add_bus(&i2c_adapter) < 0)
+        	{
+	        	PRINT(KERN_ERR, lynx->id,  "unable to register i2c");
+        	}
+        	else
+        	{
+                        /* do i2c stuff */
+                        unsigned char i2c_cmd = 0x10;
+                        struct i2c_msg msg[2] = { { 0x50, 0, 1, &i2c_cmd }, 
+                                                  { 0x50, I2C_M_RD, 20, (unsigned char*) lynx->config_rom }
+                                                };
+
+
+#ifdef CONFIG_IEEE1394_VERBOSEDEBUG
+                        union i2c_smbus_data data;
+
+                        if (i2c_smbus_xfer(&i2c_adapter, 80, 0, I2C_SMBUS_WRITE, 0, I2C_SMBUS_BYTE,NULL))
+                                PRINT(KERN_ERR, lynx->id,"eeprom read start has failed");
+                        else
+                        {
+                                u16 addr;
+                                for (addr=0x00; addr < 0x100; addr++) {
+                                        if (i2c_smbus_xfer(&i2c_adapter, 80, 0, I2C_SMBUS_READ, 0, I2C_SMBUS_BYTE,& data)) {
+                                                PRINT(KERN_ERR, lynx->id, "unable to read i2c %x", addr);
+                                                break;
+                                        }
+                                        else
+                                                PRINT(KERN_DEBUG, lynx->id,"got serial eeprom data at %x: %x",addr, data.byte);
+                                }
+                        }
+#endif
+
+                        /* we use i2c_transfer, because i2c_smbus_read_block_data does not work properly and we
+                           do it more efficiently in one transaction rather then using several reads */
+                        if (i2c_transfer(&i2c_adapter, msg, 2) < 0) {
+                                PRINT(KERN_ERR, lynx->id, "unable to read bus info block from i2c");
+                        } else {
+#ifdef CONFIG_IEEE1394_VERBOSEDEBUG
+                                int i;
+#endif
+                                PRINT(KERN_INFO, lynx->id, "got bus info block from serial eeprom");
+                                /* FIXME: probably we shoud rewrite the max_rec, max_ROM(1394a), generation(1394a) and link_spd(1394a) field
+                                   and recalculate the CRC */
+
+#ifdef CONFIG_IEEE1394_VERBOSEDEBUG
+                                for (i=0; i < 5 ; i++)
+                                        PRINT(KERN_DEBUG, lynx->id, "Businfo block quadlet %i: %08x",i, be32_to_cpu(lynx->config_rom[i]));
+#endif
+
+                                /* info_length, crc_length and 1394 magic number to check, if it is really a bus info block */
+                                if (((be32_to_cpu(lynx->config_rom[0]) & 0xffff0000) == 0x04040000) &&
+                                    (lynx->config_rom[1] == __constant_cpu_to_be32(0x31333934)))
+                                {
+                                        PRINT(KERN_DEBUG, lynx->id, "read a valid bus info block from");
+                                        got_valid_bus_info_block = 1;
+                                } else {
+                                        PRINT(KERN_WARNING, lynx->id, "read something from serial eeprom, but it does not seem to be a valid bus info block");
+                                }
+
+                        }
+
+                        i2c_bit_del_bus(&i2c_adapter);
+                }
+        }
+
+        if (got_valid_bus_info_block) {
+                memcpy(lynx->config_rom+5,lynx_csr_rom+5,sizeof(lynx_csr_rom)-20);
+        } else {
+                PRINT(KERN_INFO, lynx->id, "since we did not get a bus info block from serial eeprom, we use a generic one with a hard coded GUID");
+                memcpy(lynx->config_rom,lynx_csr_rom,sizeof(lynx_csr_rom));
+        }
+
         hpsb_add_host(host);
         lynx->state = is_host;
 
@@ -1503,7 +1675,8 @@ static int __devinit add_card(struct pci_dev *dev,
 
 static size_t get_lynx_rom(struct hpsb_host *host, const quadlet_t **ptr)
 {
-        *ptr = lynx_csr_rom;
+        struct ti_lynx *lynx = host->hostdata;
+        *ptr = lynx->config_rom;
         return sizeof(lynx_csr_rom);
 }
 
diff --git a/drivers/ieee1394/pcilynx.h b/drivers/ieee1394/pcilynx.h
index 879774645348..d1cc73c2b14b 100644
--- a/drivers/ieee1394/pcilynx.h
+++ b/drivers/ieee1394/pcilynx.h
@@ -25,6 +25,8 @@
 #define CHANNEL_ASYNC_SEND       3
 #define CHANNEL_ISO_SEND         4
 
+#define PCILYNX_CONFIG_ROM_LENGTH   1024
+
 typedef int pcl_t;
 
 struct ti_lynx {
@@ -48,7 +50,7 @@ struct ti_lynx {
         void *local_rom;
         void *local_ram;
         void *aux_port;
-
+        quadlet_t config_rom[PCILYNX_CONFIG_ROM_LENGTH/4];
 
 #ifdef CONFIG_IEEE1394_PCILYNX_PORTS
         atomic_t aux_intr_seen;
@@ -109,6 +111,8 @@ struct ti_lynx {
                 struct tasklet_struct tq;
                 spinlock_t lock;
         } iso_rcv;
+
+	u32 i2c_driven_state; /* the state we currently drive the Serial EEPROM Control register */
 };
 
 /* the per-file data structure for mem space access */
@@ -156,6 +160,8 @@ static inline void reg_clear_bits(const struct ti_lynx *lynx, int offset,
 #define MISC_CONTROL                      0x40
 #define MISC_CONTROL_SWRESET              (1<<0)
 
+#define SERIAL_EEPROM_CONTROL             0x44
+
 #define PCI_INT_STATUS                    0x48
 #define PCI_INT_ENABLE                    0x4c               
 /* status and enable have identical bit numbers */
diff --git a/drivers/ieee1394/raw1394.c b/drivers/ieee1394/raw1394.c
index 4a0f83221049..ee52d81f5264 100644
--- a/drivers/ieee1394/raw1394.c
+++ b/drivers/ieee1394/raw1394.c
@@ -189,6 +189,8 @@ static void remove_host(struct hpsb_host *host)
         }
 
         kfree(hi);
+
+        atomic_inc(&internal_generation);
 }
 
 static void host_reset(struct hpsb_host *host)
@@ -223,8 +225,6 @@ static void host_reset(struct hpsb_host *host)
                 }
         }
         spin_unlock_irqrestore(&host_info_lock, flags);
-
-        atomic_inc(&internal_generation);
 }
 
 static void iso_receive(struct hpsb_host *host, int channel, quadlet_t *data,
diff --git a/drivers/ieee1394/sbp2.c b/drivers/ieee1394/sbp2.c
index fe14a1e573c2..3338ed620e69 100644
--- a/drivers/ieee1394/sbp2.c
+++ b/drivers/ieee1394/sbp2.c
@@ -86,6 +86,9 @@
  * sbp2_max_sectors, 		- Change max sectors per I/O supported (default = 255)
  * sbp2_max_outstanding_cmds	- Change max outstanding concurrent commands (default = 8)
  * sbp2_max_cmds_per_lun	- Change max concurrent commands per sbp2 device (default = 1)
+ * sbp2_exclusive_login		- Set to zero if you'd like to allow multiple hosts the ability
+ *				  to log in at the same time. Sbp2 device must support this,
+ *				  and you must know what you're doing (default = 1)
  *
  * (e.g. insmod sbp2 sbp2_serialize_io = 1)
  *
@@ -135,11 +138,14 @@
  *	- Error Handling: SCSI aborts and bus reset requests are handled somewhat
  *	  but the code needs additional debugging.
  *
- *	- The SBP-2 driver is currently only supported as a module. It would not take
+ *	- Module: The SBP-2 driver is currently only supported as a module. It would not take
  *	  much work to allow it to be compiled into the kernel, but you'd have to 
  *	  add some init code to the kernel to support this... and modules are much
  *	  more flexible anyway.   ;-)
  *
+ *	- Hot-plugging: Interaction with the SCSI stack and support for hot-plugging could
+ *	  stand some improvement.
+ *
  *
  * History:
  *
@@ -265,7 +271,7 @@
  *		     which do not support requests of 128KB or greater. Now use 
  *		     max_sectors scsi host entry to limit transfer sizes.
  *		   * Change status fifo address from a single address to a set of addresses,
- *		     with each sbp2 device having it's own status fifo address. This makes
+ *		     with each sbp2 device having its own status fifo address. This makes
  *		     it easier to match the status write to the sbp2 device instance.
  *		   * Minor change to use lun when logging into sbp2 devices. First step in
  *		     supporting multi-lun devices such as CD/DVD changer devices.
@@ -278,7 +284,20 @@
  *	02/20/02 - Added a couple additional module load options. 
  *		   Needed to bump down max commands per lun because of the !%@&*^# QPS CDRW 
  *		   drive I have, which doesn't seem to get along with other sbp2 devices 
- *		   (or handle linked commands well).  
+ *		   (or handle linked commands well).
+ *	04/21/02 - Added some additional debug capabilities:
+ *		   * Able to handle phys dma requests directly, if host controller has phys
+ *		     dma disabled (e.g. insmod ohci1394 phys_dma=0). Undefine CONFIG_IEEE1394_SBP2_PHYS_DMA
+ *		     if you'd like to disable sbp2 driver from registering for phys address range. 
+ *		   * New packet dump debug define (CONFIG_IEEE1394_SBP2_PACKET_DUMP) which allows
+ *		     dumping of all sbp2 related packets sent and received. Especially effective
+ *		     when phys dma is disabled on ohci controller (e.g. insmod ohci1394 phys_dma=0).
+ *		   * Added new sbp2 module load option (sbp2_exclusive_login) for allowing
+ *		     non-exclusive login to sbp2 device, for special multi-host applications.
+ *	04/23/02 - Fix for Sony CD-ROM drives. Only send fetch agent reset to sbp2 device if it
+ *		   returns the dead bit in status. Thanks to Chandan (chandan@toad.net) for this one.
+ *	04/27/02 - Fix sbp2 login problem on SMP systems, enable real spinlocks by default. (JSG)
+ *	06/09/02 - Don't force 36-bute SCSI inquiry, but leave in a define for badly behaved devices. (JSG)   
  */
 
 
@@ -305,6 +324,7 @@
 #include <asm/uaccess.h>
 #include <asm/io.h>
 #include <asm/byteorder.h>
+#include <asm/atomic.h>
 #include <asm/system.h>
 #include <asm/io.h>
 #include <asm/scatterlist.h>
@@ -329,6 +349,9 @@
 #include "ieee1394_hotplug.h"
 #include "sbp2.h"
 
+static char version[] __devinitdata =
+	"$Rev: 507 $ James Goodwin <jamesg@filanet.com>";
+
 /*
  * Module load parameter definitions
  */
@@ -387,6 +410,18 @@ MODULE_PARM(sbp2_max_cmds_per_lun,"i");
 MODULE_PARM_DESC(sbp2_max_cmds_per_lun, "Change max concurrent commands per sbp2 device (default = 1)");
 static int sbp2_max_cmds_per_lun = SBP2SCSI_MAX_CMDS_PER_LUN;
 
+/*
+ * Exclusive login to sbp2 device? In most cases, the sbp2 driver should do an exclusive login, as it's
+ * generally unsafe to have two hosts talking to a single sbp2 device at the same time (filesystem
+ * coherency, etc.). If you're running an sbp2 device that supports multiple logins, and you're either
+ * running read-only filesystems or some sort of special filesystem supporting multiple hosts, then
+ * set sbp2_exclusive_login to zero. Note: The Oxsemi OXFW911 sbp2 chipset supports up to four
+ * concurrent logins.
+ */
+MODULE_PARM(sbp2_exclusive_login,"i");
+MODULE_PARM_DESC(sbp2_exclusive_login, "Exclusive login to sbp2 device (default = 1)");
+static int sbp2_exclusive_login = 1;
+
 
 /*
  * Export information about protocols/devices supported by this driver.
@@ -411,6 +446,7 @@ MODULE_DEVICE_TABLE(ieee1394, sbp2_id_table);
 /* #define CONFIG_IEEE1394_SBP2_DEBUG_DMA */
 /* #define CONFIG_IEEE1394_SBP2_DEBUG 1 */
 /* #define CONFIG_IEEE1394_SBP2_DEBUG 2 */
+/* #define CONFIG_IEEE1394_SBP2_PACKET_DUMP */
 
 #ifdef CONFIG_IEEE1394_SBP2_DEBUG_ORBS
 #define SBP2_ORB_DEBUG(fmt, args...)	HPSB_ERR("sbp2(%s): "fmt, __FUNCTION__, ## args)
@@ -456,10 +492,10 @@ static u32 global_outstanding_dmas = 0;
 #define SBP2_ERR(fmt, args...)		HPSB_ERR("sbp2: "fmt, ## args)
 
 /*
- * Spinlock debugging stuff. I'm playing it safe until the driver has been
- * debugged on SMP. (JSG)
+ * Spinlock debugging stuff.
  */
-/* #define SBP2_USE_REAL_SPINLOCKS */
+#define SBP2_USE_REAL_SPINLOCKS
+
 #ifdef SBP2_USE_REAL_SPINLOCKS
 #define sbp2_spin_lock(lock, flags)	spin_lock_irqsave(lock, flags)	
 #define sbp2_spin_unlock(lock, flags)	spin_unlock_irqrestore(lock, flags);
@@ -470,6 +506,14 @@ static spinlock_t sbp2_host_info_lock = SPIN_LOCK_UNLOCKED;
 #endif
 
 /*
+ * SCSI inquiry hack for really badly behaved sbp2 devices. Turn this on if your sbp2 device
+ * is not properly handling the SCSI inquiry command. This hack makes the inquiry look more 
+ * like a typical MS Windows inquiry.
+ */
+
+/* #define SBP2_FORCE_36_BYTE_INQUIRY */
+
+/*
  * Globals
  */
 
@@ -490,6 +534,13 @@ static struct hpsb_address_ops sbp2_ops = {
 	write: sbp2_handle_status_write
 };
 
+#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
+static struct hpsb_address_ops sbp2_physdma_ops = {
+        read: sbp2_handle_physdma_read,
+        write: sbp2_handle_physdma_write,
+};
+#endif
+
 static struct hpsb_protocol_driver sbp2_driver = {
 	name:		"SBP2 Driver",
 	id_table: 	sbp2_id_table,
@@ -498,7 +549,6 @@ static struct hpsb_protocol_driver sbp2_driver = {
 	update: 	sbp2_update
 };
 
-
 
 /**************************************
  * General utility functions
@@ -537,6 +587,56 @@ static __inline__ void sbp2util_cpu_to_be32_buffer(void *buffer, int length)
 #define sbp2util_cpu_to_be32_buffer(x,y)
 #endif
 
+#ifdef CONFIG_IEEE1394_SBP2_PACKET_DUMP
+/*
+ * Debug packet dump routine. Length is in bytes.
+ */
+static void sbp2util_packet_dump(void *buffer, int length, char *dump_name, u32 dump_phys_addr)
+{
+	int i;
+	unsigned char *dump = buffer;
+
+	if (!dump || !length || !dump_name)
+		return;
+
+	if (dump_phys_addr)
+		printk("[%s, 0x%x]", dump_name, dump_phys_addr);
+	else
+		printk("[%s]", dump_name);
+	for (i = 0; i < length; i++) {
+		if (i > 0x3f) {
+			printk("\n   ...");
+			break;
+		}
+		if ((i & 0x3) == 0)
+			printk("  ");
+		if ((i & 0xf) == 0)
+			printk("\n   ");
+		printk("%02x ", (int) dump[i]);
+	}
+	printk("\n");
+
+	return;
+}
+#else
+#define sbp2util_packet_dump(w,x,y,z)
+#endif
+
+/*
+ * Goofy routine that basically does a down_timeout function.
+ */
+static int sbp2util_down_timeout(atomic_t *done, int timeout)
+{
+	int i;
+
+	for (i = timeout; (i > 0 && atomic_read(done) == 0); i-= HZ/10) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		if (schedule_timeout(HZ/10))	/* 100ms */
+			return(1);
+	}
+	return ((i > 0) ? 0:1);
+}
+
 /*
  * This function is called to initially create a packet pool for use in
  * sbp2 I/O requests. This packet pool is used when sending out sbp2
@@ -665,7 +765,7 @@ sbp2util_allocate_write_request_packet(struct sbp2scsi_host_info *hi,
 		
 		hpsb_node_fill_packet(ne, packet);
 
-		packet->tlabel = get_tlabel(hi->host, packet->node_id, 1);
+		packet->tlabel = get_tlabel(hi->host, packet->node_id, 0);
 
 		if (!data_size) {
 			fill_async_writequad(packet, addr, data);
@@ -950,6 +1050,13 @@ int sbp2_init(void)
 				SBP2_STATUS_FIFO_ADDRESS + 
 				SBP2_STATUS_FIFO_ENTRY_TO_OFFSET(SBP2SCSI_MAX_SCSI_IDS+1));
 
+	/*
+	 * Handle data movement if physical dma is not enabled/supported on host controller
+	 */
+#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
+	hpsb_register_addrspace(sbp2_hl_handle, &sbp2_physdma_ops, 0x0ULL, 0xfffffffcULL);
+#endif
+
 	hpsb_register_protocol(&sbp2_driver);
 
 	return 0;
@@ -1241,7 +1348,7 @@ alloc_fail_first:
 	scsi_id->max_payload_size = sbp2_speedto_maxrec[SPEED_100];
 	ud->driver_data = scsi_id;
 
-	init_waitqueue_head(&scsi_id->sbp2_login_wait);
+	atomic_set(&scsi_id->sbp2_login_complete, 0);
 
 	/* 
 	 * Initialize structures needed for the command orb pool.
@@ -1379,20 +1486,62 @@ static void sbp2_remove_device(struct sbp2scsi_host_info *hi,
 	kfree(scsi_id);
 }
 
-
+#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
+/*
+ * This function deals with physical dma write requests (for adapters that do not support
+ * physical dma in hardware). Mostly just here for debugging...
+ */
+static int sbp2_handle_physdma_write(struct hpsb_host *host, int nodeid, int destid, quadlet_t *data,
+                                     u64 addr, unsigned int length)
+{
+
+        /*
+         * Manually put the data in the right place.
+         */
+        memcpy(bus_to_virt((u32)addr), data, length);
+	sbp2util_packet_dump(data, length, "sbp2 phys dma write by device", (u32)addr);
+        return(RCODE_COMPLETE);
+}
+
+/*
+ * This function deals with physical dma read requests (for adapters that do not support
+ * physical dma in hardware). Mostly just here for debugging...
+ */
+static int sbp2_handle_physdma_read(struct hpsb_host *host, int nodeid, quadlet_t *data,
+                                    u64 addr, unsigned int length)
+{
+
+        /*
+         * Grab data from memory and send a read response.
+         */
+        memcpy(data, bus_to_virt((u32)addr), length);
+	sbp2util_packet_dump(data, length, "sbp2 phys dma read by device", (u32)addr);
+        return(RCODE_COMPLETE);
+}
+#endif
+
 
 /**************************************
  * SBP-2 protocol related section
  **************************************/
 
 /*
+ * This function determines if we should convert scsi commands for a particular sbp2 device type
+ */
+static __inline__ int sbp2_command_conversion_device_type(u8 device_type)
+{
+	return (((device_type == TYPE_DISK) ||
+		 (device_type == TYPE_SDAD) ||
+		 (device_type == TYPE_ROM)) ? 1:0);
+}
+
+/*
  * This function is called in order to login to a particular SBP-2 device,
  * after a bus reset.
  */
 static int sbp2_login_device(struct sbp2scsi_host_info *hi, struct scsi_id_instance_data *scsi_id) 
 {
 	quadlet_t data[2];
-	unsigned long flags;
 
 	SBP2_DEBUG("sbp2_login_device");
 
@@ -1412,7 +1561,7 @@ static int sbp2_login_device(struct sbp2scsi_host_info *hi, struct scsi_id_insta
 
 	scsi_id->login_orb->lun_misc = ORB_SET_FUNCTION(LOGIN_REQUEST);
 	scsi_id->login_orb->lun_misc |= ORB_SET_RECONNECT(0);	/* One second reconnect time */
-	scsi_id->login_orb->lun_misc |= ORB_SET_EXCLUSIVE(1);	/* Exclusive access to device */
+	scsi_id->login_orb->lun_misc |= ORB_SET_EXCLUSIVE(sbp2_exclusive_login);	/* Exclusive access to device */
 	scsi_id->login_orb->lun_misc |= ORB_SET_NOTIFY(1);	/* Notify us of login complete */
 	/* Set the lun if we were able to pull it from the device's unit directory */
 	if (scsi_id->sbp2_device_type_and_lun != SBP2_DEVICE_TYPE_LUN_UNINITIALIZED) {
@@ -1437,6 +1586,9 @@ static int sbp2_login_device(struct sbp2scsi_host_info *hi, struct scsi_id_insta
 
 	SBP2_DEBUG("sbp2_login_device: orb byte-swapped");
 
+	sbp2util_packet_dump(scsi_id->login_orb, sizeof(struct sbp2_login_orb), 
+			     "sbp2 login orb", scsi_id->login_orb_dma);
+
 	/*
 	 * Initialize login response and status fifo
 	 */
@@ -1452,35 +1604,28 @@ static int sbp2_login_device(struct sbp2scsi_host_info *hi, struct scsi_id_insta
 	data[1] = scsi_id->login_orb_dma;
 	sbp2util_cpu_to_be32_buffer(data, 8);
 
+	atomic_set(&scsi_id->sbp2_login_complete, 0);
+
 	SBP2_DEBUG("sbp2_login_device: prepared to write");
 	hpsb_node_write(scsi_id->ne, scsi_id->sbp2_management_agent_addr, data, 8);
 	SBP2_DEBUG("sbp2_login_device: written");
 
 	/*
-	 * Wait for login status... but, only if the device has not
-	 * already logged-in (some devices are fast)
+	 * Wait for login status (up to 20 seconds)... 
 	 */
-
-	save_flags(flags);
-	cli();
-	/* 20 second timeout */
-	if (scsi_id->status_block.ORB_offset_lo != scsi_id->login_orb_dma)
-		sleep_on_timeout(&scsi_id->sbp2_login_wait, 20*HZ);
-	restore_flags(flags);
-
-	SBP2_DEBUG("sbp2_login_device: initial check");
+	if (sbp2util_down_timeout(&scsi_id->sbp2_login_complete, 20*HZ)) {
+		SBP2_ERR("Error logging into SBP-2 device - login timed-out");
+		return(-EIO);
+	}
 
 	/*
-	 * Match status to the login orb. If they do not match, it's
-	 * probably because the login timed-out.
+	 * Sanity. Make sure status returned matches login orb.
 	 */
 	if (scsi_id->status_block.ORB_offset_lo != scsi_id->login_orb_dma) {
 		SBP2_ERR("Error logging into SBP-2 device - login timed-out");
 		return(-EIO);
 	}
 
-	SBP2_DEBUG("sbp2_login_device: second check");
-
 	/*
 	 * Check status
 	 */
@@ -1552,6 +1697,9 @@ static int sbp2_logout_device(struct sbp2scsi_host_info *hi, struct scsi_id_inst
 	 */
 	sbp2util_cpu_to_be32_buffer(scsi_id->logout_orb, sizeof(struct sbp2_logout_orb));
 
+	sbp2util_packet_dump(scsi_id->logout_orb, sizeof(struct sbp2_logout_orb), 
+			     "sbp2 logout orb", scsi_id->logout_orb_dma);
+
 	/*
 	 * Ok, let's write to the target's management agent register
 	 */
@@ -1559,10 +1707,12 @@ static int sbp2_logout_device(struct sbp2scsi_host_info *hi, struct scsi_id_inst
 	data[1] = scsi_id->logout_orb_dma;
 	sbp2util_cpu_to_be32_buffer(data, 8);
 
+	atomic_set(&scsi_id->sbp2_login_complete, 0);
+
 	hpsb_node_write(scsi_id->ne, scsi_id->sbp2_management_agent_addr, data, 8);
 
 	/* Wait for device to logout...1 second. */
-	sleep_on_timeout(&scsi_id->sbp2_login_wait, HZ);
+	sbp2util_down_timeout(&scsi_id->sbp2_login_complete, HZ);
 
 	SBP2_INFO("Logged out of SBP-2 device");
 
@@ -1577,7 +1727,6 @@ static int sbp2_logout_device(struct sbp2scsi_host_info *hi, struct scsi_id_inst
 static int sbp2_reconnect_device(struct sbp2scsi_host_info *hi, struct scsi_id_instance_data *scsi_id) 
 {
 	quadlet_t data[2];
-	unsigned long flags;
 
 	SBP2_DEBUG("sbp2_reconnect_device");
 
@@ -1607,6 +1756,9 @@ static int sbp2_reconnect_device(struct sbp2scsi_host_info *hi, struct scsi_id_i
 	 */
 	sbp2util_cpu_to_be32_buffer(scsi_id->reconnect_orb, sizeof(struct sbp2_reconnect_orb));
 
+	sbp2util_packet_dump(scsi_id->reconnect_orb, sizeof(struct sbp2_reconnect_orb), 
+			     "sbp2 reconnect orb", scsi_id->reconnect_orb_dma);
+
 	/*
 	 * Initialize status fifo
 	 */
@@ -1619,22 +1771,20 @@ static int sbp2_reconnect_device(struct sbp2scsi_host_info *hi, struct scsi_id_i
 	data[1] = scsi_id->reconnect_orb_dma;
 	sbp2util_cpu_to_be32_buffer(data, 8);
 
+	atomic_set(&scsi_id->sbp2_login_complete, 0);
+
 	hpsb_node_write(scsi_id->ne, scsi_id->sbp2_management_agent_addr, data, 8);
 
 	/*
-	 * Wait for reconnect status... but, only if the device has not
-	 * already reconnected (some devices are fast).
+	 * Wait for reconnect status (up to 1 second)...
 	 */
-	save_flags(flags);
-	cli();
-	/* One second timout */
-	if (scsi_id->status_block.ORB_offset_lo != scsi_id->reconnect_orb_dma)
-		sleep_on_timeout(&scsi_id->sbp2_login_wait, HZ);
-	restore_flags(flags);
+	if (sbp2util_down_timeout(&scsi_id->sbp2_login_complete, HZ)) {
+		SBP2_ERR("Error reconnecting to SBP-2 device - reconnect timed-out");
+		return(-EIO);
+	}
 
 	/*
-	 * Match status to the reconnect orb. If they do not match, it's
-	 * probably because the reconnect timed-out.
+	 * Sanity. Make sure status returned matches reconnect orb.
 	 */
 	if (scsi_id->status_block.ORB_offset_lo != scsi_id->reconnect_orb_dma) {
 		SBP2_ERR("Error reconnecting to SBP-2 device - reconnect timed-out");
@@ -2002,6 +2152,10 @@ static int sbp2_create_command_orb(struct sbp2scsi_host_info *hi,
 			/* Number of page table (s/g) elements */
 			command_orb->misc |= ORB_SET_DATA_SIZE(sg_count);
 
+			sbp2util_packet_dump(scatter_gather_element, 
+					     (sizeof(struct sbp2_unrestricted_page_table)) * sg_count, 
+					     "sbp2 s/g list", command->sge_dma);
+
 			/*
 			 * Byte swap page tables if necessary
 			 */
@@ -2081,6 +2235,10 @@ static int sbp2_create_command_orb(struct sbp2scsi_host_info *hi,
 			/* Number of page table (s/g) elements */
 			command_orb->misc |= ORB_SET_DATA_SIZE(sg_count);
 
+			sbp2util_packet_dump(scatter_gather_element, 
+					     (sizeof(struct sbp2_unrestricted_page_table)) * sg_count, 
+					     "sbp2 s/g list", command->sge_dma);
+
 			/*
 			 * Byte swap page tables if necessary
 			 */
@@ -2219,13 +2377,11 @@ static int sbp2_send_command(struct sbp2scsi_host_info *hi, struct scsi_id_insta
 {
 	unchar *cmd = (unchar *) SCpnt->cmnd;
 	unsigned int request_bufflen = SCpnt->request_bufflen;
-	u8 device_type
-		= SBP2_DEVICE_TYPE (scsi_id->sbp2_device_type_and_lun);
 	struct sbp2_command_info *command;
 
 	SBP2_DEBUG("sbp2_send_command");
-	SBP2_DEBUG("SCSI command:");
-#if CONFIG_IEEE1394_SBP2_DEBUG >= 2
+#if (CONFIG_IEEE1394_SBP2_DEBUG >= 2) || defined(CONFIG_IEEE1394_SBP2_PACKET_DUMP)
+	printk("[scsi command]\n   ");
 	print_command (cmd);
 #endif
 	SBP2_DEBUG("SCSI transfer size = %x", request_bufflen);
@@ -2242,11 +2398,14 @@ static int sbp2_send_command(struct sbp2scsi_host_info *hi, struct scsi_id_insta
 	/*
 	 * The scsi stack sends down a request_bufflen which does not match the
 	 * length field in the scsi cdb. This causes some sbp2 devices to 
-	 * reject this inquiry command. Hack fix is to set both buff length and
-	 * length field in cdb to 36. This gives best compatibility.
+	 * reject this inquiry command. Fix the request_bufflen. 
 	 */
 	if (*cmd == INQUIRY) {
+#ifdef SBP2_FORCE_36_BYTE_INQUIRY
 		request_bufflen = cmd[4] = 0x24;
+#else
+		request_bufflen = cmd[4];
+#endif
 	}
 
 	/*
@@ -2259,11 +2418,10 @@ static int sbp2_send_command(struct sbp2scsi_host_info *hi, struct scsi_id_insta
 	 * Update our cdb if necessary (to handle sbp2 RBC command set
 	 * differences). This is where the command set hacks go!   =)
 	 */
-	if ((device_type == TYPE_DISK) ||
-	    (device_type == TYPE_SDAD) ||
-	    (device_type == TYPE_ROM)) {
-		sbp2_check_sbp2_command(command->command_orb.cdb);
-	}
+	sbp2_check_sbp2_command(scsi_id, command->command_orb.cdb);
+
+	sbp2util_packet_dump(&command->command_orb, sizeof(struct sbp2_command_orb), 
+			     "sbp2 command orb", command->command_orb_dma);
 
 	/*
 	 * Initialize status fifo
@@ -2283,9 +2441,10 @@ static int sbp2_send_command(struct sbp2scsi_host_info *hi, struct scsi_id_insta
  * This function deals with command set differences between Linux scsi
  * command set and sbp2 RBC command set.
  */
-static void sbp2_check_sbp2_command(unchar *cmd)
+static void sbp2_check_sbp2_command(struct scsi_id_instance_data *scsi_id, unchar *cmd)
 {
 	unchar new_cmd[16];
+	u8 device_type = SBP2_DEVICE_TYPE (scsi_id->sbp2_device_type_and_lun);
 
 	SBP2_DEBUG("sbp2_check_sbp2_command");
 
@@ -2293,67 +2452,79 @@ static void sbp2_check_sbp2_command(unchar *cmd)
 		
 		case READ_6:
 
-			SBP2_DEBUG("Convert READ_6 to READ_10");
+			if (sbp2_command_conversion_device_type(device_type)) {
 
-			/*
-			 * Need to turn read_6 into read_10
-			 */
-			new_cmd[0] = 0x28;
-			new_cmd[1] = (cmd[1] & 0xe0);
-			new_cmd[2] = 0x0;
-			new_cmd[3] = (cmd[1] & 0x1f);
-			new_cmd[4] = cmd[2];
-			new_cmd[5] = cmd[3];
-			new_cmd[6] = 0x0;
-			new_cmd[7] = 0x0;
-			new_cmd[8] = cmd[4];
-			new_cmd[9] = cmd[5];
-
-			memcpy(cmd, new_cmd, 10);
+				SBP2_DEBUG("Convert READ_6 to READ_10");
+					    
+				/*
+				 * Need to turn read_6 into read_10
+				 */
+				new_cmd[0] = 0x28;
+				new_cmd[1] = (cmd[1] & 0xe0);
+				new_cmd[2] = 0x0;
+				new_cmd[3] = (cmd[1] & 0x1f);
+				new_cmd[4] = cmd[2];
+				new_cmd[5] = cmd[3];
+				new_cmd[6] = 0x0;
+				new_cmd[7] = 0x0;
+				new_cmd[8] = cmd[4];
+				new_cmd[9] = cmd[5];
+	
+				memcpy(cmd, new_cmd, 10);
+
+			}
 
 			break;
 
 		case WRITE_6:
 
-			SBP2_DEBUG("Convert WRITE_6 to WRITE_10");
+			if (sbp2_command_conversion_device_type(device_type)) {
 
-			/*
-			 * Need to turn write_6 into write_10
-			 */
-			new_cmd[0] = 0x2a;
-			new_cmd[1] = (cmd[1] & 0xe0);
-			new_cmd[2] = 0x0;
-			new_cmd[3] = (cmd[1] & 0x1f);
-			new_cmd[4] = cmd[2];
-			new_cmd[5] = cmd[3];
-			new_cmd[6] = 0x0;
-			new_cmd[7] = 0x0;
-			new_cmd[8] = cmd[4];
-			new_cmd[9] = cmd[5];
-
-			memcpy(cmd, new_cmd, 10);
+				SBP2_DEBUG("Convert WRITE_6 to WRITE_10");
+	
+				/*
+				 * Need to turn write_6 into write_10
+				 */
+				new_cmd[0] = 0x2a;
+				new_cmd[1] = (cmd[1] & 0xe0);
+				new_cmd[2] = 0x0;
+				new_cmd[3] = (cmd[1] & 0x1f);
+				new_cmd[4] = cmd[2];
+				new_cmd[5] = cmd[3];
+				new_cmd[6] = 0x0;
+				new_cmd[7] = 0x0;
+				new_cmd[8] = cmd[4];
+				new_cmd[9] = cmd[5];
+	
+				memcpy(cmd, new_cmd, 10);
+
+			}
 
 			break;
 
 		case MODE_SENSE:
 
-			SBP2_DEBUG("Convert MODE_SENSE_6 to MOSE_SENSE_10");
+			if (sbp2_command_conversion_device_type(device_type)) {
 
-			/*
-			 * Need to turn mode_sense_6 into mode_sense_10
-			 */
-			new_cmd[0] = 0x5a;
-			new_cmd[1] = cmd[1];
-			new_cmd[2] = cmd[2];
-			new_cmd[3] = 0x0;
-			new_cmd[4] = 0x0;
-			new_cmd[5] = 0x0;
-			new_cmd[6] = 0x0;
-			new_cmd[7] = 0x0;
-			new_cmd[8] = cmd[4];
-			new_cmd[9] = cmd[5];
-
-			memcpy(cmd, new_cmd, 10);
+				SBP2_DEBUG("Convert MODE_SENSE_6 to MODE_SENSE_10");
+
+				/*
+				 * Need to turn mode_sense_6 into mode_sense_10
+				 */
+				new_cmd[0] = 0x5a;
+				new_cmd[1] = cmd[1];
+				new_cmd[2] = cmd[2];
+				new_cmd[3] = 0x0;
+				new_cmd[4] = 0x0;
+				new_cmd[5] = 0x0;
+				new_cmd[6] = 0x0;
+				new_cmd[7] = 0x0;
+				new_cmd[8] = cmd[4];
+				new_cmd[9] = cmd[5];
+	
+				memcpy(cmd, new_cmd, 10);
+
+			}
 
 			break;
 
@@ -2404,8 +2575,7 @@ static unsigned int sbp2_status_to_sense_data(unchar *sbp2_status, unchar *sense
  * This function is called after a command is completed, in order to do any necessary SBP-2
  * response data translations for the SCSI stack
  */
-static void sbp2_check_sbp2_response(struct sbp2scsi_host_info *hi,
-				     struct scsi_id_instance_data *scsi_id, 
+static void sbp2_check_sbp2_response(struct scsi_id_instance_data *scsi_id, 
 				     Scsi_Cmnd *SCpnt)
 {
 	u8 *scsi_buf = SCpnt->request_buffer;
@@ -2451,18 +2621,16 @@ static void sbp2_check_sbp2_response(struct sbp2scsi_host_info *hi,
 
 		case MODE_SENSE:
 
-			if ((device_type == TYPE_DISK) ||
-			    (device_type == TYPE_SDAD) ||
-			    (device_type == TYPE_ROM)) {
-
+			if (sbp2_command_conversion_device_type(device_type)) {
+			
 				SBP2_DEBUG("Modify mode sense response (10 byte version)");
-	
+
 				scsi_buf[0] = scsi_buf[1];	/* Mode data length */
 				scsi_buf[1] = scsi_buf[2];	/* Medium type */
 				scsi_buf[2] = scsi_buf[3];	/* Device specific parameter */
 				scsi_buf[3] = scsi_buf[7];	/* Block descriptor length */
 				memcpy(scsi_buf + 4, scsi_buf + 8, scsi_buf[0]);
-
+	
 			}
 
 			break;
@@ -2495,6 +2663,8 @@ static int sbp2_handle_status_write(struct hpsb_host *host, int nodeid, int dest
 
 	SBP2_DEBUG("sbp2_handle_status_write");
 
+	sbp2util_packet_dump(data, length, "sbp2 status write by device", (u32)addr);
+
 	if (!host) {
 		SBP2_ERR("host is NULL - this is bad!");
 		return(RCODE_ADDRESS_ERROR);
@@ -2566,21 +2736,21 @@ static int sbp2_handle_status_write(struct hpsb_host *host, int nodeid, int dest
 				/*
 				 * Translate SBP-2 status to SCSI sense data
 				 */
+				SBP2_DEBUG("CHECK CONDITION");
 				scsi_status = sbp2_status_to_sense_data((unchar *)&scsi_id->status_block, SCpnt->sense_buffer);
 			}
 
 			/*
-			 * Handle check conditions. If there is either SBP status or SCSI status
-			 * then we'll do a fetch agent reset and note that a check condition
-			 * occured.
+			 * Check to see if the dead bit is set. If so, we'll have to initiate
+			 * a fetch agent reset.
 			 */
-			if (STATUS_GET_SBP_STATUS(scsi_id->status_block.ORB_offset_hi_misc) ||
-			    scsi_status) {
+			if (STATUS_GET_DEAD_BIT(scsi_id->status_block.ORB_offset_hi_misc)) {
+
 				/*
 				 * Initiate a fetch agent reset. 
 				 */
-				SBP2_DEBUG("CHECK CONDITION");
-				sbp2_agent_reset(hi, scsi_id, SBP2_SEND_NO_WAIT);
+				SBP2_DEBUG("Dead bit set - initiating fetch agent reset");
+                                sbp2_agent_reset(hi, scsi_id, SBP2_SEND_NO_WAIT);
 			}
 
 			SBP2_ORB_DEBUG("completing command orb %p", &command->command_orb);
@@ -2602,10 +2772,19 @@ static int sbp2_handle_status_write(struct hpsb_host *host, int nodeid, int dest
 			scsi_id->last_orb = NULL;
 		}
 
+	} else {
+		
+		/* 
+		 * It's probably a login/logout/reconnect status.
+		 */
+		if ((scsi_id->login_orb_dma == scsi_id->status_block.ORB_offset_lo) ||
+		    (scsi_id->reconnect_orb_dma == scsi_id->status_block.ORB_offset_lo) ||
+		    (scsi_id->logout_orb_dma == scsi_id->status_block.ORB_offset_lo)) {
+			atomic_set(&scsi_id->sbp2_login_complete, 1);
+		}
 	}
 
 	sbp2_spin_unlock(&hi->sbp2_command_lock, flags);
-	wake_up(&scsi_id->sbp2_login_wait);
 	return(RCODE_COMPLETE);
 }
 
@@ -2805,7 +2984,7 @@ static void sbp2scsi_complete_command(struct sbp2scsi_host_info *hi, struct scsi
 	 * Take care of any sbp2 response data mucking here (RBC stuff, etc.)
 	 */
 	if (SCpnt->result == DID_OK) {
-		sbp2_check_sbp2_response(hi, scsi_id, SCpnt);
+		sbp2_check_sbp2_response(scsi_id, SCpnt);
 	}
 
 	/*
@@ -2822,10 +3001,13 @@ static void sbp2scsi_complete_command(struct sbp2scsi_host_info *hi, struct scsi
 	 * retried... it could have happened because of a 1394 bus reset
 	 * or hot-plug...
 	 */
-	if ((scsi_status == SBP2_SCSI_STATUS_CHECK_CONDITION) && (SCpnt->sense_buffer[2] == UNIT_ATTENTION)) {
+#if 0
+	if ((scsi_status == SBP2_SCSI_STATUS_CHECK_CONDITION) && 
+	    (SCpnt->sense_buffer[2] == UNIT_ATTENTION)) {
 		SBP2_DEBUG("UNIT ATTENTION - return busy");
 		SCpnt->result = DID_BUS_BUSY << 16;
 	}
+#endif
 
 	/*
 	 * Tell scsi stack that we're done with this command
@@ -2835,9 +3017,9 @@ static void sbp2scsi_complete_command(struct sbp2scsi_host_info *hi, struct scsi
 	done (SCpnt);
 	spin_unlock_irq(&io_request_lock);
 #else
-	spin_lock_irq(&hi->scsi_host->host_lock);
+	spin_lock_irq(hi->scsi_host->host_lock);
 	done (SCpnt);
-	spin_unlock_irq(&hi->scsi_host->host_lock);
+	spin_unlock_irq(hi->scsi_host->host_lock);
 #endif
 
 	return;
@@ -2905,7 +3087,7 @@ static int sbp2scsi_reset (Scsi_Cmnd *SCpnt)
 	SBP2_ERR("reset requested");
 
 	if (scsi_id) {
-		SBP2_ERR("Generating IEEE-1394 bus reset");
+		SBP2_ERR("Generating sbp2 fetch agent reset");
 		sbp2_agent_reset(hi, scsi_id, SBP2_SEND_NO_WAIT);
 	}
 
@@ -2969,19 +3151,22 @@ static const char *sbp2scsi_info (struct Scsi_Host *host)
 	if (!hi) /* shouldn't happen, but... */
         	return "IEEE-1394 SBP-2 protocol driver";
 
-	sprintf(info, "IEEE-1394 SBP-2 protocol driver (host: %s)\n"
+	sprintf(info, "IEEE-1394 SBP-2 protocol driver (host: %s)\n%s\n"
 		"SBP-2 module load options:\n"
 		"- Max speed supported: %s\n"
 		"- Max sectors per I/O supported: %d\n"
 		"- Max outstanding commands supported: %d\n"
 		"- Max outstanding commands per lun supported: %d\n"
-                "- Serialized I/O (debug): %s",
-		hi->host->driver->name, 
+                "- Serialized I/O (debug): %s\n"
+		"- Exclusive login: %s",
+		hi->host->driver->name,
+		version,
 		hpsb_speedto_str[sbp2_max_speed],
 		sbp2_max_sectors,
 		sbp2_max_outstanding_cmds,
 		sbp2_max_cmds_per_lun,
-		sbp2_serialize_io ? "yes" : "no");
+		sbp2_serialize_io ? "yes" : "no",
+		sbp2_exclusive_login ? "yes" : "no");
 
 	return info;
 }
diff --git a/drivers/ieee1394/sbp2.h b/drivers/ieee1394/sbp2.h
index 4ebe8c34a61b..7d6da4a61dd5 100644
--- a/drivers/ieee1394/sbp2.h
+++ b/drivers/ieee1394/sbp2.h
@@ -398,9 +398,9 @@ struct scsi_id_instance_data {
 	u32 sbp2_firmware_revision;
 
 	/* 
-	 * Wait queue used for logins, reconnects, logouts 
+	 * Variable used for logins, reconnects, logouts 
 	 */
-	wait_queue_head_t sbp2_login_wait;
+	atomic_t sbp2_login_complete;
 
 	/* 
 	 * Pool of command orbs, so we can have more than overlapped command per id
@@ -501,6 +501,13 @@ static int sbp2_start_device(struct sbp2scsi_host_info *hi,
 static void sbp2_remove_device(struct sbp2scsi_host_info *hi, 
 			       struct scsi_id_instance_data *scsi_id);
 
+#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
+static int sbp2_handle_physdma_write(struct hpsb_host *host, int nodeid, int destid, quadlet_t *data,
+                                     u64 addr, unsigned int length);
+static int sbp2_handle_physdma_read(struct hpsb_host *host, int nodeid, quadlet_t *data,
+                                    u64 addr, unsigned int length);
+#endif
+
 /*
  * SBP-2 protocol related prototypes
  */
@@ -523,9 +530,8 @@ static int sbp2_link_orb_command(struct sbp2scsi_host_info *hi, struct scsi_id_i
 static int sbp2_send_command(struct sbp2scsi_host_info *hi, struct scsi_id_instance_data *scsi_id,
 			     Scsi_Cmnd *SCpnt, void (*done)(Scsi_Cmnd *));
 static unsigned int sbp2_status_to_sense_data(unchar *sbp2_status, unchar *sense_data);
-static void sbp2_check_sbp2_command(unchar *cmd);
-static void sbp2_check_sbp2_response(struct sbp2scsi_host_info *hi, struct scsi_id_instance_data *scsi_id, 
-				     Scsi_Cmnd *SCpnt);
+static void sbp2_check_sbp2_command(struct scsi_id_instance_data *scsi_id, unchar *cmd);
+static void sbp2_check_sbp2_response(struct scsi_id_instance_data *scsi_id, Scsi_Cmnd *SCpnt);
 static void sbp2_parse_unit_directory(struct scsi_id_instance_data *scsi_id);
 static int sbp2_set_busy_timeout(struct sbp2scsi_host_info *hi, struct scsi_id_instance_data *scsi_id);
 static int sbp2_max_speed_and_size(struct sbp2scsi_host_info *hi, struct scsi_id_instance_data *scsi_id);
diff --git a/drivers/ieee1394/video1394.c b/drivers/ieee1394/video1394.c
index 730878ecab79..edab382285e3 100644
--- a/drivers/ieee1394/video1394.c
+++ b/drivers/ieee1394/video1394.c
@@ -168,86 +168,35 @@ static struct hpsb_highlevel *hl_handle = NULL;
 /* Memory management functions */
 /*******************************/
 
-#define MDEBUG(x)	do { } while(0)		/* Debug memory management */
-
-/* [DaveM] I've recoded most of this so that:
- * 1) It's easier to tell what is happening
- * 2) It's more portable, especially for translating things
- *    out of vmalloc mapped areas in the kernel.
- * 3) Less unnecessary translations happen.
- *
- * The code used to assume that the kernel vmalloc mappings
- * existed in the page tables of every process, this is simply
- * not guaranteed.  We now use pgd_offset_k which is the
- * defined way to get at the kernel page tables.
- */
-
-/* Given PGD from the address space's page table, return the kernel
- * virtual mapping of the physical memory mapped at ADR.
- */
-static inline unsigned long uvirt_to_kva(pgd_t *pgd, unsigned long adr)
-{
-        unsigned long ret = 0UL;
-	pmd_t *pmd;
-	pte_t *ptep, pte;
-  
-	if (!pgd_none(*pgd)) {
-                pmd = pmd_offset(pgd, adr);
-                if (!pmd_none(*pmd)) {
-                        ptep = pte_offset_kernel(pmd, adr);
-                        pte = *ptep;
-                        if(pte_present(pte)) {
-				ret = (unsigned long) 
-					page_address(pte_page(pte));
-                                ret |= (adr & (PAGE_SIZE - 1));
-			}
-                }
-        }
-        MDEBUG(printk("uv2kva(%lx-->%lx)", adr, ret));
-	return ret;
-}
-
-static inline unsigned long uvirt_to_bus(unsigned long adr) 
-{
-        unsigned long kva, ret;
-
-        kva = uvirt_to_kva(pgd_offset(current->mm, adr), adr);
-	ret = virt_to_bus((void *)kva);
-        MDEBUG(printk("uv2b(%lx-->%lx)", adr, ret));
-        return ret;
-}
-
 static inline unsigned long kvirt_to_bus(unsigned long adr) 
 {
-        unsigned long va, kva, ret;
+	unsigned long kva, ret;
 
-        va = VMALLOC_VMADDR(adr);
-        kva = uvirt_to_kva(pgd_offset_k(va), va);
+	kva = (unsigned long) page_address(vmalloc_to_page((void *)adr));
+	kva |= adr & (PAGE_SIZE-1); /* restore the offset */
 	ret = virt_to_bus((void *)kva);
-        MDEBUG(printk("kv2b(%lx-->%lx)", adr, ret));
-        return ret;
+	return ret;
 }
 
 /* Here we want the physical address of the memory.
- * This is used when initializing the contents of the
- * area and marking the pages as reserved.
+ * This is used when initializing the contents of the area.
  */
 static inline unsigned long kvirt_to_pa(unsigned long adr) 
 {
-        unsigned long va, kva, ret;
+        unsigned long kva, ret;
 
-        va = VMALLOC_VMADDR(adr);
-        kva = uvirt_to_kva(pgd_offset_k(va), va);
+	kva = (unsigned long) page_address(vmalloc_to_page((void *)adr));
+	kva |= adr & (PAGE_SIZE-1); /* restore the offset */
 	ret = __pa(kva);
-        MDEBUG(printk("kv2pa(%lx-->%lx)", adr, ret));
         return ret;
 }
 
 static void * rvmalloc(unsigned long size)
 {
 	void * mem;
-	unsigned long adr, page;
-        
+	unsigned long adr;
+
+	size=PAGE_ALIGN(size);
 	mem=vmalloc_32(size);
 	if (mem) 
 	{
@@ -256,8 +205,7 @@ static void * rvmalloc(unsigned long size)
 	        adr=(unsigned long) mem;
 		while (size > 0) 
                 {
-	                page = kvirt_to_pa(adr);
-			mem_map_reserve(virt_to_page(__va(page)));
+			mem_map_reserve(vmalloc_to_page((void *)adr));
 			adr+=PAGE_SIZE;
 			size-=PAGE_SIZE;
 		}
@@ -267,15 +215,14 @@ static void * rvmalloc(unsigned long size)
 
 static void rvfree(void * mem, unsigned long size)
 {
-        unsigned long adr, page;
-        
+        unsigned long adr;
+
 	if (mem) 
 	{
 	        adr=(unsigned long) mem;
-		while (size > 0) 
+		while ((long) size > 0) 
                 {
-	                page = kvirt_to_pa(adr);
-			mem_map_unreserve(virt_to_page(__va(page)));
+			mem_map_unreserve(vmalloc_to_page((void *)adr));
 			adr+=PAGE_SIZE;
 			size-=PAGE_SIZE;
 		}
@@ -520,11 +467,11 @@ alloc_dma_iso_ctx(struct ti_ohci *ohci, int type, int num_desc,
 static void reset_ir_status(struct dma_iso_ctx *d, int n)
 {
 	int i;
-	d->ir_prg[n][0].status = 4;
-	d->ir_prg[n][1].status = PAGE_SIZE-4;
+	d->ir_prg[n][0].status = cpu_to_le32(4);
+	d->ir_prg[n][1].status = cpu_to_le32(PAGE_SIZE-4);
 	for (i=2;i<d->nb_cmd-1;i++)
-		d->ir_prg[n][i].status = PAGE_SIZE;
-	d->ir_prg[n][i].status = d->left_size;
+		d->ir_prg[n][i].status = cpu_to_le32(PAGE_SIZE);
+	d->ir_prg[n][i].status = cpu_to_le32(d->left_size);
 }
 
 static void initialize_dma_ir_prg(struct dma_iso_ctx *d, int n, int flags)
@@ -534,38 +481,38 @@ static void initialize_dma_ir_prg(struct dma_iso_ctx *d, int n, int flags)
 	int i;
 
 	/* the first descriptor will read only 4 bytes */
-	ir_prg[0].control = DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE |
-		DMA_CTL_BRANCH | 4;
+	ir_prg[0].control = cpu_to_le32(DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE |
+		DMA_CTL_BRANCH | 4);
 
 	/* set the sync flag */
 	if (flags & VIDEO1394_SYNC_FRAMES)
-		ir_prg[0].control |= DMA_CTL_WAIT;
+		ir_prg[0].control |= cpu_to_le32(DMA_CTL_WAIT);
 
-	ir_prg[0].address = kvirt_to_bus(buf);
-	ir_prg[0].branchAddress =  (virt_to_bus(&(ir_prg[1].control)) 
-				    & 0xfffffff0) | 0x1;
+	ir_prg[0].address = cpu_to_le32(kvirt_to_bus(buf));
+	ir_prg[0].branchAddress =  cpu_to_le32((virt_to_bus(&(ir_prg[1].control)) 
+				    & 0xfffffff0) | 0x1);
 
 	/* the second descriptor will read PAGE_SIZE-4 bytes */
-	ir_prg[1].control = DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE |
-		DMA_CTL_BRANCH | (PAGE_SIZE-4);
-	ir_prg[1].address = kvirt_to_bus(buf+4);
-	ir_prg[1].branchAddress =  (virt_to_bus(&(ir_prg[2].control)) 
-				    & 0xfffffff0) | 0x1;
+	ir_prg[1].control = cpu_to_le32(DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE |
+		DMA_CTL_BRANCH | (PAGE_SIZE-4));
+	ir_prg[1].address = cpu_to_le32(kvirt_to_bus(buf+4));
+	ir_prg[1].branchAddress =  cpu_to_le32((virt_to_bus(&(ir_prg[2].control)) 
+				    & 0xfffffff0) | 0x1);
 	
 	for (i=2;i<d->nb_cmd-1;i++) {
-		ir_prg[i].control = DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE | 
-			DMA_CTL_BRANCH | PAGE_SIZE;
-		ir_prg[i].address = kvirt_to_bus(buf+(i-1)*PAGE_SIZE);
+		ir_prg[i].control = cpu_to_le32(DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE | 
+			DMA_CTL_BRANCH | PAGE_SIZE);
+		ir_prg[i].address = cpu_to_le32(kvirt_to_bus(buf+(i-1)*PAGE_SIZE));
 
 		ir_prg[i].branchAddress =  
-			(virt_to_bus(&(ir_prg[i+1].control)) 
-			 & 0xfffffff0) | 0x1;
+			cpu_to_le32((virt_to_bus(&(ir_prg[i+1].control)) 
+			 & 0xfffffff0) | 0x1);
 	}
 
 	/* the last descriptor will generate an interrupt */
-	ir_prg[i].control = DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE | 
-		DMA_CTL_IRQ | DMA_CTL_BRANCH | d->left_size;
-	ir_prg[i].address = kvirt_to_bus(buf+(i-1)*PAGE_SIZE);
+	ir_prg[i].control = cpu_to_le32(DMA_CTL_INPUT_MORE | DMA_CTL_UPDATE | 
+		DMA_CTL_IRQ | DMA_CTL_BRANCH | d->left_size);
+	ir_prg[i].address = cpu_to_le32(kvirt_to_bus(buf+(i-1)*PAGE_SIZE));
 }
 	
 static void initialize_dma_ir_ctx(struct dma_iso_ctx *d, int tag, int flags)
@@ -628,7 +575,7 @@ int wakeup_dma_ir_ctx(struct ti_ohci *ohci, struct dma_iso_ctx *d)
 
 	spin_lock(&d->lock);
 	for (i=0;i<d->num_desc;i++) {
-		if (d->ir_prg[i][d->nb_cmd-1].status & 0xFFFF0000) {
+		if (d->ir_prg[i][d->nb_cmd-1].status & cpu_to_le32(0xFFFF0000)) {
 			reset_ir_status(d, i);
 			d->buffer_status[i] = VIDEO1394_BUFFER_READY;
 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,18)
@@ -664,7 +611,7 @@ static inline void put_timestamp(struct ti_ohci *ohci, struct dma_iso_ctx * d,
 	buf[7] = timeStamp & 0xff; 
 
     /* if first packet is empty packet, then put timestamp into the next full one too */
-    if ( (d->it_prg[n][0].data[1] >>16) == 0x008) {
+    if ( (le32_to_cpu(d->it_prg[n][0].data[1]) >>16) == 0x008) {
    	    buf += d->packet_size;
     	buf[6] = timeStamp >> 8;
 	    buf[7] = timeStamp & 0xff;
@@ -683,7 +630,7 @@ static inline void put_timestamp(struct ti_ohci *ohci, struct dma_iso_ctx * d,
 	buf[7] = timeStamp & 0xff;
 
     /* if first packet is empty packet, then put timestamp into the next full one too */
-    if ( (d->it_prg[n][0].data[1] >>16) == 0x008) {
+    if ( (le32_to_cpu(d->it_prg[n][0].data[1]) >>16) == 0x008) {
    	    buf += d->packet_size;
     	buf[6] = timeStamp >> 8;
 	    buf[7] = timeStamp & 0xff;
@@ -707,7 +654,8 @@ int wakeup_dma_it_ctx(struct ti_ohci *ohci, struct dma_iso_ctx *d)
 
 	spin_lock(&d->lock);
 	for (i=0;i<d->num_desc;i++) {
-		if (d->it_prg[i][d->last_used_cmd[i]].end.status& 0xFFFF0000) {
+		if (d->it_prg[i][d->last_used_cmd[i]].end.status& 
+			cpu_to_le32(0xFFFF0000)) {
 			int next = d->next_buffer[i];
 			put_timestamp(ohci, d, next);
 			d->it_prg[i][d->last_used_cmd[i]].end.status = 0;
@@ -727,39 +675,40 @@ static void initialize_dma_it_prg(struct dma_iso_ctx *d, int n, int sync_tag)
 	d->last_used_cmd[n] = d->nb_cmd - 1;
 	for (i=0;i<d->nb_cmd;i++) {
 				 
-		it_prg[i].begin.control = DMA_CTL_OUTPUT_MORE |
-			DMA_CTL_IMMEDIATE | 8 ;
+		it_prg[i].begin.control = cpu_to_le32(DMA_CTL_OUTPUT_MORE |
+			DMA_CTL_IMMEDIATE | 8) ;
 		it_prg[i].begin.address = 0;
 		
 		it_prg[i].begin.status = 0;
 		
-		it_prg[i].data[0] = 
+		it_prg[i].data[0] = cpu_to_le32(
 			(SPEED_100 << 16) 
 			| (/* tag */ 1 << 14)
 			| (d->channel << 8) 
-			| (TCODE_ISO_DATA << 4);
-		if (i==0) it_prg[i].data[0] |= sync_tag;
-		it_prg[i].data[1] = d->packet_size << 16;
+			| (TCODE_ISO_DATA << 4));
+		if (i==0) it_prg[i].data[0] |= cpu_to_le32(sync_tag);
+		it_prg[i].data[1] = cpu_to_le32(d->packet_size << 16);
 		it_prg[i].data[2] = 0;
 		it_prg[i].data[3] = 0;
 		
-		it_prg[i].end.control = DMA_CTL_OUTPUT_LAST | DMA_CTL_BRANCH;
+		it_prg[i].end.control = cpu_to_le32(DMA_CTL_OUTPUT_LAST |
+			    	    	     DMA_CTL_BRANCH);
 		it_prg[i].end.address =
-			kvirt_to_bus(buf+i*d->packet_size);
+			cpu_to_le32(kvirt_to_bus(buf+i*d->packet_size));
 
 		if (i<d->nb_cmd-1) {
-			it_prg[i].end.control |= d->packet_size;
+			it_prg[i].end.control |= cpu_to_le32(d->packet_size);
 			it_prg[i].begin.branchAddress = 
-				(virt_to_bus(&(it_prg[i+1].begin.control)) 
-				 & 0xfffffff0) | 0x3;
+				cpu_to_le32((virt_to_bus(&(it_prg[i+1].begin.control)) 
+				 & 0xfffffff0) | 0x3);
 			it_prg[i].end.branchAddress = 
-				(virt_to_bus(&(it_prg[i+1].begin.control)) 
-				 & 0xfffffff0) | 0x3;
+				cpu_to_le32((virt_to_bus(&(it_prg[i+1].begin.control)) 
+				 & 0xfffffff0) | 0x3);
 		}
 		else {
 			/* the last prg generates an interrupt */
-			it_prg[i].end.control |= DMA_CTL_UPDATE | 
-				DMA_CTL_IRQ | d->left_size;
+			it_prg[i].end.control |= cpu_to_le32(DMA_CTL_UPDATE | 
+				DMA_CTL_IRQ | d->left_size);
 			/* the last prg doesn't branch */
 			it_prg[i].begin.branchAddress = 0;
 			it_prg[i].end.branchAddress = 0;
@@ -798,21 +747,21 @@ static void initialize_dma_it_prg_var_packet_queue(
 		} else {
 			size = packet_sizes[i];
 		}
-		it_prg[i].data[1] = size << 16; 
-		it_prg[i].end.control = DMA_CTL_OUTPUT_LAST | DMA_CTL_BRANCH;
+		it_prg[i].data[1] = cpu_to_le32(size << 16); 
+		it_prg[i].end.control = cpu_to_le32(DMA_CTL_OUTPUT_LAST | DMA_CTL_BRANCH);
 
 		if (i < d->nb_cmd-1 && packet_sizes[i+1] != 0) {
-			it_prg[i].end.control |= size;
+			it_prg[i].end.control |= cpu_to_le32(size);
 			it_prg[i].begin.branchAddress = 
-				(virt_to_bus(&(it_prg[i+1].begin.control)) 
-				 & 0xfffffff0) | 0x3;
+				cpu_to_le32((virt_to_bus(&(it_prg[i+1].begin.control)) 
+				 & 0xfffffff0) | 0x3);
 			it_prg[i].end.branchAddress = 
-				(virt_to_bus(&(it_prg[i+1].begin.control)) 
-				 & 0xfffffff0) | 0x3;
+				cpu_to_le32((virt_to_bus(&(it_prg[i+1].begin.control)) 
+				 & 0xfffffff0) | 0x3);
 		} else {
 			/* the last prg generates an interrupt */
-			it_prg[i].end.control |= DMA_CTL_UPDATE | 
-				DMA_CTL_IRQ | size;
+			it_prg[i].end.control |= cpu_to_le32(DMA_CTL_UPDATE | 
+				DMA_CTL_IRQ | size);
 			/* the last prg doesn't branch */
 			it_prg[i].begin.branchAddress = 0;
 			it_prg[i].end.branchAddress = 0;
@@ -1057,8 +1006,8 @@ static int video1394_ioctl(struct inode *inode, struct file *file,
 
 		if (d->last_buffer>=0) 
 			d->ir_prg[d->last_buffer][d->nb_cmd-1].branchAddress = 
-				(virt_to_bus(&(d->ir_prg[v.buffer][0].control)) 
-				 & 0xfffffff0) | 0x1;
+				cpu_to_le32((virt_to_bus(&(d->ir_prg[v.buffer][0].control)) 
+				 & 0xfffffff0) | 0x1);
 
 		d->last_buffer = v.buffer;
 
@@ -1217,14 +1166,14 @@ static int video1394_ioctl(struct inode *inode, struct file *file,
 			d->it_prg[d->last_buffer]
 				[ d->last_used_cmd[d->last_buffer]
 					].end.branchAddress = 
-				(virt_to_bus(&(d->it_prg[v.buffer][0].begin.control)) 
-				 & 0xfffffff0) | 0x3;
+				cpu_to_le32((virt_to_bus(&(d->it_prg[v.buffer][0].begin.control)) 
+				 & 0xfffffff0) | 0x3);
 
 			d->it_prg[d->last_buffer]
 				[d->last_used_cmd[d->last_buffer]
 					].begin.branchAddress = 
-				(virt_to_bus(&(d->it_prg[v.buffer][0].begin.control)) 
-				 & 0xfffffff0) | 0x3;
+				cpu_to_le32((virt_to_bus(&(d->it_prg[v.buffer][0].begin.control)) 
+				 & 0xfffffff0) | 0x3);
 			d->next_buffer[d->last_buffer] = v.buffer;
 		}
 		d->last_buffer = v.buffer;