Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
  firewire: use split transaction timeout only for split transactions
  firewire: ohci: consolidate context status flags
  firewire: ohci: cache the context run bit
  firewire: ohci: flush AT contexts after bus reset - addendum
  firewire: ohci: flush AT contexts after bus reset for OHCI 1.2
  firewire: net: set carrier state at ifup
  firewire: net: add carrier detection
  firewire: net: ratelimit error messages
  firewire: ohci: restart iso DMA contexts on resume from low power mode
  firewire: ohci: restore GUID on resume.
  firewire: ohci: use common buffer for self IDs and AR descriptors
  firewire: ohci: optimize iso context checks in the interrupt handler
  firewire: make PHY packet header format consistent
  firewire: ohci: properly clear posted write errors
  firewire: ohci: flush MMIO writes in the interrupt handler
  firewire: ohci: fix AT context initialization error handling
  firewire: ohci: Asynchronous Reception rewrite
  firewire: core: Update WARN uses
  firewire: nosy: char device is not seekable

7 files changed, 547 insertions, 246 deletions

diff --git a/drivers/firewire/Kconfig b/drivers/firewire/Kconfig
index 40a222e19b2d..68f942cb30f2 100644
--- a/drivers/firewire/Kconfig
+++ b/drivers/firewire/Kconfig
@@ -19,7 +19,7 @@ config FIREWIRE
 
 config FIREWIRE_OHCI
 	tristate "OHCI-1394 controllers"
-	depends on PCI && FIREWIRE
+	depends on PCI && FIREWIRE && MMU
 	help
 	  Enable this driver if you have a FireWire controller based
 	  on the OHCI specification.  For all practical purposes, this
diff --git a/drivers/firewire/core-cdev.c b/drivers/firewire/core-cdev.c
index 14bb7b7b5dd7..48ae712e2101 100644
--- a/drivers/firewire/core-cdev.c
+++ b/drivers/firewire/core-cdev.c
@@ -1501,9 +1501,10 @@ static int ioctl_send_phy_packet(struct client *client, union ioctl_arg *arg)
 	e->client		= client;
 	e->p.speed		= SCODE_100;
 	e->p.generation		= a->generation;
-	e->p.header[0]		= a->data[0];
-	e->p.header[1]		= a->data[1];
-	e->p.header_length	= 8;
+	e->p.header[0]		= TCODE_LINK_INTERNAL << 4;
+	e->p.header[1]		= a->data[0];
+	e->p.header[2]		= a->data[1];
+	e->p.header_length	= 12;
 	e->p.callback		= outbound_phy_packet_callback;
 	e->phy_packet.closure	= a->closure;
 	e->phy_packet.type	= FW_CDEV_EVENT_PHY_PACKET_SENT;
diff --git a/drivers/firewire/core-transaction.c b/drivers/firewire/core-transaction.c
index b42a0bde8494..d00f8ce902cc 100644
--- a/drivers/firewire/core-transaction.c
+++ b/drivers/firewire/core-transaction.c
@@ -72,6 +72,15 @@
 #define PHY_CONFIG_ROOT_ID(node_id)	((((node_id) & 0x3f) << 24) | (1 << 23))
 #define PHY_IDENTIFIER(id)		((id) << 30)
 
+/* returns 0 if the split timeout handler is already running */
+static int try_cancel_split_timeout(struct fw_transaction *t)
+{
+	if (t->is_split_transaction)
+		return del_timer(&t->split_timeout_timer);
+	else
+		return 1;
+}
+
 static int close_transaction(struct fw_transaction *transaction,
 			     struct fw_card *card, int rcode)
 {
@@ -81,7 +90,7 @@ static int close_transaction(struct fw_transaction *transaction,
 	spin_lock_irqsave(&card->lock, flags);
 	list_for_each_entry(t, &card->transaction_list, link) {
 		if (t == transaction) {
-			if (!del_timer(&t->split_timeout_timer)) {
+			if (!try_cancel_split_timeout(t)) {
 				spin_unlock_irqrestore(&card->lock, flags);
 				goto timed_out;
 			}
@@ -141,16 +150,28 @@ static void split_transaction_timeout_callback(unsigned long data)
 	card->tlabel_mask &= ~(1ULL << t->tlabel);
 	spin_unlock_irqrestore(&card->lock, flags);
 
-	card->driver->cancel_packet(card, &t->packet);
-
-	/*
-	 * At this point cancel_packet will never call the transaction
-	 * callback, since we just took the transaction out of the list.
-	 * So do it here.
-	 */
 	t->callback(card, RCODE_CANCELLED, NULL, 0, t->callback_data);
 }
 
+static void start_split_transaction_timeout(struct fw_transaction *t,
+					    struct fw_card *card)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&card->lock, flags);
+
+	if (list_empty(&t->link) || WARN_ON(t->is_split_transaction)) {
+		spin_unlock_irqrestore(&card->lock, flags);
+		return;
+	}
+
+	t->is_split_transaction = true;
+	mod_timer(&t->split_timeout_timer,
+		  jiffies + card->split_timeout_jiffies);
+
+	spin_unlock_irqrestore(&card->lock, flags);
+}
+
 static void transmit_complete_callback(struct fw_packet *packet,
 				       struct fw_card *card, int status)
 {
@@ -162,7 +183,7 @@ static void transmit_complete_callback(struct fw_packet *packet,
 		close_transaction(t, card, RCODE_COMPLETE);
 		break;
 	case ACK_PENDING:
-		t->timestamp = packet->timestamp;
+		start_split_transaction_timeout(t, card);
 		break;
 	case ACK_BUSY_X:
 	case ACK_BUSY_A:
@@ -250,7 +271,7 @@ static void fw_fill_request(struct fw_packet *packet, int tcode, int tlabel,
 		break;
 
 	default:
-		WARN(1, "wrong tcode %d", tcode);
+		WARN(1, "wrong tcode %d\n", tcode);
 	}
  common:
 	packet->speed = speed;
@@ -349,11 +370,9 @@ void fw_send_request(struct fw_card *card, struct fw_transaction *t, int tcode,
 	t->node_id = destination_id;
 	t->tlabel = tlabel;
 	t->card = card;
+	t->is_split_transaction = false;
 	setup_timer(&t->split_timeout_timer,
 		    split_transaction_timeout_callback, (unsigned long)t);
-	/* FIXME: start this timer later, relative to t->timestamp */
-	mod_timer(&t->split_timeout_timer,
-		  jiffies + card->split_timeout_jiffies);
 	t->callback = callback;
 	t->callback_data = callback_data;
 
@@ -423,7 +442,8 @@ static void transmit_phy_packet_callback(struct fw_packet *packet,
 }
 
 static struct fw_packet phy_config_packet = {
-	.header_length	= 8,
+	.header_length	= 12,
+	.header[0]	= TCODE_LINK_INTERNAL << 4,
 	.payload_length	= 0,
 	.speed		= SCODE_100,
 	.callback	= transmit_phy_packet_callback,
@@ -451,8 +471,8 @@ void fw_send_phy_config(struct fw_card *card,
 
 	mutex_lock(&phy_config_mutex);
 
-	phy_config_packet.header[0] = data;
-	phy_config_packet.header[1] = ~data;
+	phy_config_packet.header[1] = data;
+	phy_config_packet.header[2] = ~data;
 	phy_config_packet.generation = generation;
 	INIT_COMPLETION(phy_config_done);
 
@@ -638,7 +658,7 @@ int fw_get_response_length(struct fw_request *r)
 		}
 
 	default:
-		WARN(1, "wrong tcode %d", tcode);
+		WARN(1, "wrong tcode %d\n", tcode);
 		return 0;
 	}
 }
@@ -694,7 +714,7 @@ void fw_fill_response(struct fw_packet *response, u32 *request_header,
 		break;
 
 	default:
-		WARN(1, "wrong tcode %d", tcode);
+		WARN(1, "wrong tcode %d\n", tcode);
 	}
 
 	response->payload_mapped = false;
@@ -925,7 +945,7 @@ void fw_core_handle_response(struct fw_card *card, struct fw_packet *p)
 	spin_lock_irqsave(&card->lock, flags);
 	list_for_each_entry(t, &card->transaction_list, link) {
 		if (t->node_id == source && t->tlabel == tlabel) {
-			if (!del_timer(&t->split_timeout_timer)) {
+			if (!try_cancel_split_timeout(t)) {
 				spin_unlock_irqrestore(&card->lock, flags);
 				goto timed_out;
 			}
diff --git a/drivers/firewire/core.h b/drivers/firewire/core.h
index e6239f971be6..f8dfcf1c6cbe 100644
--- a/drivers/firewire/core.h
+++ b/drivers/firewire/core.h
@@ -215,9 +215,11 @@ static inline bool is_next_generation(int new_generation, int old_generation)
 
 /* -transaction */
 
+#define TCODE_LINK_INTERNAL		0xe
+
 #define TCODE_IS_READ_REQUEST(tcode)	(((tcode) & ~1) == 4)
 #define TCODE_IS_BLOCK_PACKET(tcode)	(((tcode) &  1) != 0)
-#define TCODE_IS_LINK_INTERNAL(tcode)	((tcode) == 0xe)
+#define TCODE_IS_LINK_INTERNAL(tcode)	((tcode) == TCODE_LINK_INTERNAL)
 #define TCODE_IS_REQUEST(tcode)		(((tcode) &  2) == 0)
 #define TCODE_IS_RESPONSE(tcode)	(((tcode) &  2) != 0)
 #define TCODE_HAS_REQUEST_DATA(tcode)	(((tcode) & 12) != 4)
diff --git a/drivers/firewire/net.c b/drivers/firewire/net.c
index 1a467a91fb0b..c2e194c58667 100644
--- a/drivers/firewire/net.c
+++ b/drivers/firewire/net.c
@@ -9,6 +9,7 @@
 #include <linux/bug.h>
 #include <linux/delay.h>
 #include <linux/device.h>
+#include <linux/ethtool.h>
 #include <linux/firewire.h>
 #include <linux/firewire-constants.h>
 #include <linux/highmem.h>
@@ -179,6 +180,7 @@ struct fwnet_device {
 	/* Number of tx datagrams that have been queued but not yet acked */
 	int queued_datagrams;
 
+	int peer_count;
 	struct list_head peer_list;
 	struct fw_card *card;
 	struct net_device *netdev;
@@ -996,15 +998,23 @@ static void fwnet_transmit_packet_failed(struct fwnet_packet_task *ptask)
 static void fwnet_write_complete(struct fw_card *card, int rcode,
 				 void *payload, size_t length, void *data)
 {
-	struct fwnet_packet_task *ptask;
-
-	ptask = data;
+	struct fwnet_packet_task *ptask = data;
+	static unsigned long j;
+	static int last_rcode, errors_skipped;
 
 	if (rcode == RCODE_COMPLETE) {
 		fwnet_transmit_packet_done(ptask);
 	} else {
-		fw_error("fwnet_write_complete: failed: %x\n", rcode);
 		fwnet_transmit_packet_failed(ptask);
+
+		if (printk_timed_ratelimit(&j,  1000) || rcode != last_rcode) {
+			fw_error("fwnet_write_complete: "
+				"failed: %x (skipped %d)\n", rcode, errors_skipped);
+
+			errors_skipped = 0;
+			last_rcode = rcode;
+		} else
+			errors_skipped++;
 	}
 }
 
@@ -1213,6 +1223,14 @@ static int fwnet_broadcast_start(struct fwnet_device *dev)
 	return retval;
 }
 
+static void set_carrier_state(struct fwnet_device *dev)
+{
+	if (dev->peer_count > 1)
+		netif_carrier_on(dev->netdev);
+	else
+		netif_carrier_off(dev->netdev);
+}
+
 /* ifup */
 static int fwnet_open(struct net_device *net)
 {
@@ -1226,6 +1244,10 @@ static int fwnet_open(struct net_device *net)
 	}
 	netif_start_queue(net);
 
+	spin_lock_irq(&dev->lock);
+	set_carrier_state(dev);
+	spin_unlock_irq(&dev->lock);
+
 	return 0;
 }
 
@@ -1397,6 +1419,10 @@ static int fwnet_change_mtu(struct net_device *net, int new_mtu)
 	return 0;
 }
 
+static const struct ethtool_ops fwnet_ethtool_ops = {
+	.get_link	= ethtool_op_get_link,
+};
+
 static const struct net_device_ops fwnet_netdev_ops = {
 	.ndo_open       = fwnet_open,
 	.ndo_stop	= fwnet_stop,
@@ -1415,6 +1441,7 @@ static void fwnet_init_dev(struct net_device *net)
 	net->hard_header_len	= FWNET_HLEN;
 	net->type		= ARPHRD_IEEE1394;
 	net->tx_queue_len	= FWNET_TX_QUEUE_LEN;
+	net->ethtool_ops	= &fwnet_ethtool_ops;
 }
 
 /* caller must hold fwnet_device_mutex */
@@ -1455,6 +1482,8 @@ static int fwnet_add_peer(struct fwnet_device *dev,
 
 	spin_lock_irq(&dev->lock);
 	list_add_tail(&peer->peer_link, &dev->peer_list);
+	dev->peer_count++;
+	set_carrier_state(dev);
 	spin_unlock_irq(&dev->lock);
 
 	return 0;
@@ -1535,13 +1564,15 @@ static int fwnet_probe(struct device *_dev)
 	return ret;
 }
 
-static void fwnet_remove_peer(struct fwnet_peer *peer)
+static void fwnet_remove_peer(struct fwnet_peer *peer, struct fwnet_device *dev)
 {
 	struct fwnet_partial_datagram *pd, *pd_next;
 
-	spin_lock_irq(&peer->dev->lock);
+	spin_lock_irq(&dev->lock);
 	list_del(&peer->peer_link);
-	spin_unlock_irq(&peer->dev->lock);
+	dev->peer_count--;
+	set_carrier_state(dev);
+	spin_unlock_irq(&dev->lock);
 
 	list_for_each_entry_safe(pd, pd_next, &peer->pd_list, pd_link)
 		fwnet_pd_delete(pd);
@@ -1558,7 +1589,7 @@ static int fwnet_remove(struct device *_dev)
 
 	mutex_lock(&fwnet_device_mutex);
 
-	fwnet_remove_peer(peer);
+	fwnet_remove_peer(peer, dev);
 
 	if (list_empty(&dev->peer_list)) {
 		net = dev->netdev;
diff --git a/drivers/firewire/nosy.c b/drivers/firewire/nosy.c
index bf184fb59a5e..0618145376ad 100644
--- a/drivers/firewire/nosy.c
+++ b/drivers/firewire/nosy.c
@@ -302,7 +302,7 @@ nosy_open(struct inode *inode, struct file *file)
 
 	file->private_data = client;
 
-	return 0;
+	return nonseekable_open(inode, file);
 fail:
 	kfree(client);
 	lynx_put(lynx);
@@ -405,7 +405,6 @@ static const struct file_operations nosy_ops = {
 	.poll =			nosy_poll,
 	.open =			nosy_open,
 	.release =		nosy_release,
-	.llseek =		noop_llseek,
 };
 
 #define PHY_PACKET_SIZE 12 /* 1 payload, 1 inverse, 1 ack = 3 quadlets */
diff --git a/drivers/firewire/ohci.c b/drivers/firewire/ohci.c
index e3c8b60bd86b..d77d120ddc25 100644
--- a/drivers/firewire/ohci.c
+++ b/drivers/firewire/ohci.c
@@ -18,6 +18,7 @@
  * Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  */
 
+#include <linux/bitops.h>
 #include <linux/bug.h>
 #include <linux/compiler.h>
 #include <linux/delay.h>
@@ -40,6 +41,7 @@
 #include <linux/spinlock.h>
 #include <linux/string.h>
 #include <linux/time.h>
+#include <linux/vmalloc.h>
 
 #include <asm/byteorder.h>
 #include <asm/page.h>
@@ -80,17 +82,23 @@ struct descriptor {
 #define COMMAND_PTR(regs)	((regs) + 12)
 #define CONTEXT_MATCH(regs)	((regs) + 16)
 
-struct ar_buffer {
-	struct descriptor descriptor;
-	struct ar_buffer *next;
-	__le32 data[0];
-};
+#define AR_BUFFER_SIZE	(32*1024)
+#define AR_BUFFERS_MIN	DIV_ROUND_UP(AR_BUFFER_SIZE, PAGE_SIZE)
+/* we need at least two pages for proper list management */
+#define AR_BUFFERS	(AR_BUFFERS_MIN >= 2 ? AR_BUFFERS_MIN : 2)
+
+#define MAX_ASYNC_PAYLOAD	4096
+#define MAX_AR_PACKET_SIZE	(16 + MAX_ASYNC_PAYLOAD + 4)
+#define AR_WRAPAROUND_PAGES	DIV_ROUND_UP(MAX_AR_PACKET_SIZE, PAGE_SIZE)
 
 struct ar_context {
 	struct fw_ohci *ohci;
-	struct ar_buffer *current_buffer;
-	struct ar_buffer *last_buffer;
+	struct page *pages[AR_BUFFERS];
+	void *buffer;
+	struct descriptor *descriptors;
+	dma_addr_t descriptors_bus;
 	void *pointer;
+	unsigned int last_buffer_index;
 	u32 regs;
 	struct tasklet_struct tasklet;
 };
@@ -117,6 +125,8 @@ struct context {
 	struct fw_ohci *ohci;
 	u32 regs;
 	int total_allocation;
+	bool running;
+	bool flushing;
 
 	/*
 	 * List of page-sized buffers for storing DMA descriptors.
@@ -161,6 +171,9 @@ struct iso_context {
 	int excess_bytes;
 	void *header;
 	size_t header_length;
+
+	u8 sync;
+	u8 tags;
 };
 
 #define CONFIG_ROM_SIZE 1024
@@ -177,7 +190,8 @@ struct fw_ohci {
 	u32 bus_time;
 	bool is_root;
 	bool csr_state_setclear_abdicate;
-
+	int n_ir;
+	int n_it;
 	/*
 	 * Spinlock for accessing fw_ohci data.  Never call out of
 	 * this driver with this lock held.
@@ -186,6 +200,9 @@ struct fw_ohci {
 
 	struct mutex phy_reg_mutex;
 
+	void *misc_buffer;
+	dma_addr_t misc_buffer_bus;
+
 	struct ar_context ar_request_ctx;
 	struct ar_context ar_response_ctx;
 	struct context at_request_ctx;
@@ -411,10 +428,6 @@ static const char *tcodes[] = {
 	[0xc] = "-reserved-",		[0xd] = "-reserved-",
 	[0xe] = "link internal",	[0xf] = "-reserved-",
 };
-static const char *phys[] = {
-	[0x0] = "phy config packet",	[0x1] = "link-on packet",
-	[0x2] = "self-id packet",	[0x3] = "-reserved-",
-};
 
 static void log_ar_at_event(char dir, int speed, u32 *header, int evt)
 {
@@ -433,12 +446,6 @@ static void log_ar_at_event(char dir, int speed, u32 *header, int evt)
 		return;
 	}
 
-	if (header[0] == ~header[1]) {
-		fw_notify("A%c %s, %s, %08x\n",
-		    dir, evts[evt], phys[header[0] >> 30 & 0x3], header[0]);
-		return;
-	}
-
 	switch (tcode) {
 	case 0x0: case 0x6: case 0x8:
 		snprintf(specific, sizeof(specific), " = %08x",
@@ -453,9 +460,13 @@ static void log_ar_at_event(char dir, int speed, u32 *header, int evt)
 	}
 
 	switch (tcode) {
-	case 0xe: case 0xa:
+	case 0xa:
 		fw_notify("A%c %s, %s\n", dir, evts[evt], tcodes[tcode]);
 		break;
+	case 0xe:
+		fw_notify("A%c %s, PHY %08x %08x\n",
+			  dir, evts[evt], header[1], header[2]);
+		break;
 	case 0x0: case 0x1: case 0x4: case 0x5: case 0x9:
 		fw_notify("A%c spd %x tl %02x, "
 		    "%04x -> %04x, %s, "
@@ -594,59 +605,150 @@ static int ohci_update_phy_reg(struct fw_card *card, int addr,
 	return ret;
 }
 
-static void ar_context_link_page(struct ar_context *ctx,
-				 struct ar_buffer *ab, dma_addr_t ab_bus)
+static inline dma_addr_t ar_buffer_bus(struct ar_context *ctx, unsigned int i)
 {
-	size_t offset;
+	return page_private(ctx->pages[i]);
+}
 
-	ab->next = NULL;
-	memset(&ab->descriptor, 0, sizeof(ab->descriptor));
-	ab->descriptor.control        = cpu_to_le16(DESCRIPTOR_INPUT_MORE |
-						    DESCRIPTOR_STATUS |
-						    DESCRIPTOR_BRANCH_ALWAYS);
-	offset = offsetof(struct ar_buffer, data);
-	ab->descriptor.req_count      = cpu_to_le16(PAGE_SIZE - offset);
-	ab->descriptor.data_address   = cpu_to_le32(ab_bus + offset);
-	ab->descriptor.res_count      = cpu_to_le16(PAGE_SIZE - offset);
-	ab->descriptor.branch_address = 0;
+static void ar_context_link_page(struct ar_context *ctx, unsigned int index)
+{
+	struct descriptor *d;
+
+	d = &ctx->descriptors[index];
+	d->branch_address  &= cpu_to_le32(~0xf);
+	d->res_count       =  cpu_to_le16(PAGE_SIZE);
+	d->transfer_status =  0;
 
 	wmb(); /* finish init of new descriptors before branch_address update */
-	ctx->last_buffer->descriptor.branch_address = cpu_to_le32(ab_bus | 1);
-	ctx->last_buffer->next = ab;
-	ctx->last_buffer = ab;
+	d = &ctx->descriptors[ctx->last_buffer_index];
+	d->branch_address  |= cpu_to_le32(1);
+
+	ctx->last_buffer_index = index;
 
 	reg_write(ctx->ohci, CONTROL_SET(ctx->regs), CONTEXT_WAKE);
 	flush_writes(ctx->ohci);
 }
 
-static int ar_context_add_page(struct ar_context *ctx)
+static void ar_context_release(struct ar_context *ctx)
 {
-	struct device *dev = ctx->ohci->card.device;
-	struct ar_buffer *ab;
-	dma_addr_t uninitialized_var(ab_bus);
+	unsigned int i;
 
-	ab = dma_alloc_coherent(dev, PAGE_SIZE, &ab_bus, GFP_ATOMIC);
-	if (ab == NULL)
-		return -ENOMEM;
+	if (ctx->buffer)
+		vm_unmap_ram(ctx->buffer, AR_BUFFERS + AR_WRAPAROUND_PAGES);
 
-	ar_context_link_page(ctx, ab, ab_bus);
+	for (i = 0; i < AR_BUFFERS; i++)
+		if (ctx->pages[i]) {
+			dma_unmap_page(ctx->ohci->card.device,
+				       ar_buffer_bus(ctx, i),
+				       PAGE_SIZE, DMA_FROM_DEVICE);
+			__free_page(ctx->pages[i]);
+		}
+}
 
-	return 0;
+static void ar_context_abort(struct ar_context *ctx, const char *error_msg)
+{
+	if (reg_read(ctx->ohci, CONTROL_CLEAR(ctx->regs)) & CONTEXT_RUN) {
+		reg_write(ctx->ohci, CONTROL_CLEAR(ctx->regs), CONTEXT_RUN);
+		flush_writes(ctx->ohci);
+
+		fw_error("AR error: %s; DMA stopped\n", error_msg);
+	}
+	/* FIXME: restart? */
 }
 
-static void ar_context_release(struct ar_context *ctx)
+static inline unsigned int ar_next_buffer_index(unsigned int index)
+{
+	return (index + 1) % AR_BUFFERS;
+}
+
+static inline unsigned int ar_prev_buffer_index(unsigned int index)
+{
+	return (index - 1 + AR_BUFFERS) % AR_BUFFERS;
+}
+
+static inline unsigned int ar_first_buffer_index(struct ar_context *ctx)
+{
+	return ar_next_buffer_index(ctx->last_buffer_index);
+}
+
+/*
+ * We search for the buffer that contains the last AR packet DMA data written
+ * by the controller.
+ */
+static unsigned int ar_search_last_active_buffer(struct ar_context *ctx,
+						 unsigned int *buffer_offset)
 {
-	struct ar_buffer *ab, *ab_next;
-	size_t offset;
-	dma_addr_t ab_bus;
+	unsigned int i, next_i, last = ctx->last_buffer_index;
+	__le16 res_count, next_res_count;
+
+	i = ar_first_buffer_index(ctx);
+	res_count = ACCESS_ONCE(ctx->descriptors[i].res_count);
+
+	/* A buffer that is not yet completely filled must be the last one. */
+	while (i != last && res_count == 0) {
+
+		/* Peek at the next descriptor. */
+		next_i = ar_next_buffer_index(i);
+		rmb(); /* read descriptors in order */
+		next_res_count = ACCESS_ONCE(
+				ctx->descriptors[next_i].res_count);
+		/*
+		 * If the next descriptor is still empty, we must stop at this
+		 * descriptor.
+		 */
+		if (next_res_count == cpu_to_le16(PAGE_SIZE)) {
+			/*
+			 * The exception is when the DMA data for one packet is
+			 * split over three buffers; in this case, the middle
+			 * buffer's descriptor might be never updated by the
+			 * controller and look still empty, and we have to peek
+			 * at the third one.
+			 */
+			if (MAX_AR_PACKET_SIZE > PAGE_SIZE && i != last) {
+				next_i = ar_next_buffer_index(next_i);
+				rmb();
+				next_res_count = ACCESS_ONCE(
+					ctx->descriptors[next_i].res_count);
+				if (next_res_count != cpu_to_le16(PAGE_SIZE))
+					goto next_buffer_is_active;
+			}
 
-	for (ab = ctx->current_buffer; ab; ab = ab_next) {
-		ab_next = ab->next;
-		offset = offsetof(struct ar_buffer, data);
-		ab_bus = le32_to_cpu(ab->descriptor.data_address) - offset;
-		dma_free_coherent(ctx->ohci->card.device, PAGE_SIZE,
-				  ab, ab_bus);
+			break;
+		}
+
+next_buffer_is_active:
+		i = next_i;
+		res_count = next_res_count;
+	}
+
+	rmb(); /* read res_count before the DMA data */
+
+	*buffer_offset = PAGE_SIZE - le16_to_cpu(res_count);
+	if (*buffer_offset > PAGE_SIZE) {
+		*buffer_offset = 0;
+		ar_context_abort(ctx, "corrupted descriptor");
+	}
+
+	return i;
+}
+
+static void ar_sync_buffers_for_cpu(struct ar_context *ctx,
+				    unsigned int end_buffer_index,
+				    unsigned int end_buffer_offset)
+{
+	unsigned int i;
+
+	i = ar_first_buffer_index(ctx);
+	while (i != end_buffer_index) {
+		dma_sync_single_for_cpu(ctx->ohci->card.device,
+					ar_buffer_bus(ctx, i),
+					PAGE_SIZE, DMA_FROM_DEVICE);
+		i = ar_next_buffer_index(i);
 	}
+	if (end_buffer_offset > 0)
+		dma_sync_single_for_cpu(ctx->ohci->card.device,
+					ar_buffer_bus(ctx, i),
+					end_buffer_offset, DMA_FROM_DEVICE);
 }
 
 #if defined(CONFIG_PPC_PMAC) && defined(CONFIG_PPC32)
@@ -689,6 +791,10 @@ static __le32 *handle_ar_packet(struct ar_context *ctx, __le32 *buffer)
 		p.header[3] = cond_le32_to_cpu(buffer[3]);
 		p.header_length = 16;
 		p.payload_length = p.header[3] >> 16;
+		if (p.payload_length > MAX_ASYNC_PAYLOAD) {
+			ar_context_abort(ctx, "invalid packet length");
+			return NULL;
+		}
 		break;
 
 	case TCODE_WRITE_RESPONSE:
@@ -699,9 +805,8 @@ static __le32 *handle_ar_packet(struct ar_context *ctx, __le32 *buffer)
 		break;
 
 	default:
-		/* FIXME: Stop context, discard everything, and restart? */
-		p.header_length = 0;
-		p.payload_length = 0;
+		ar_context_abort(ctx, "invalid tcode");
+		return NULL;
 	}
 
 	p.payload = (void *) buffer + p.header_length;
@@ -751,121 +856,147 @@ static __le32 *handle_ar_packet(struct ar_context *ctx, __le32 *buffer)
 	return buffer + length + 1;
 }
 
+static void *handle_ar_packets(struct ar_context *ctx, void *p, void *end)
+{
+	void *next;
+
+	while (p < end) {
+		next = handle_ar_packet(ctx, p);
+		if (!next)
+			return p;
+		p = next;
+	}
+
+	return p;
+}
+
+static void ar_recycle_buffers(struct ar_context *ctx, unsigned int end_buffer)
+{
+	unsigned int i;
+
+	i = ar_first_buffer_index(ctx);
+	while (i != end_buffer) {
+		dma_sync_single_for_device(ctx->ohci->card.device,
+					   ar_buffer_bus(ctx, i),
+					   PAGE_SIZE, DMA_FROM_DEVICE);
+		ar_context_link_page(ctx, i);
+		i = ar_next_buffer_index(i);
+	}
+}
+
 static void ar_context_tasklet(unsigned long data)
 {
 	struct ar_context *ctx = (struct ar_context *)data;
-	struct ar_buffer *ab;
-	struct descriptor *d;
-	void *buffer, *end;
-	__le16 res_count;
+	unsigned int end_buffer_index, end_buffer_offset;
+	void *p, *end;
 
-	ab = ctx->current_buffer;
-	d = &ab->descriptor;
+	p = ctx->pointer;
+	if (!p)
+		return;
 
-	res_count = ACCESS_ONCE(d->res_count);
-	if (res_count == 0) {
-		size_t size, size2, rest, pktsize, size3, offset;
-		dma_addr_t start_bus;
-		void *start;
+	end_buffer_index = ar_search_last_active_buffer(ctx,
+							&end_buffer_offset);
+	ar_sync_buffers_for_cpu(ctx, end_buffer_index, end_buffer_offset);
+	end = ctx->buffer + end_buffer_index * PAGE_SIZE + end_buffer_offset;
 
+	if (end_buffer_index < ar_first_buffer_index(ctx)) {
 		/*
-		 * This descriptor is finished and we may have a
-		 * packet split across this and the next buffer. We
-		 * reuse the page for reassembling the split packet.
+		 * The filled part of the overall buffer wraps around; handle
+		 * all packets up to the buffer end here.  If the last packet
+		 * wraps around, its tail will be visible after the buffer end
+		 * because the buffer start pages are mapped there again.
 		 */
+		void *buffer_end = ctx->buffer + AR_BUFFERS * PAGE_SIZE;
+		p = handle_ar_packets(ctx, p, buffer_end);
+		if (p < buffer_end)
+			goto error;
+		/* adjust p to point back into the actual buffer */
+		p -= AR_BUFFERS * PAGE_SIZE;
+	}
 
-		offset = offsetof(struct ar_buffer, data);
-		start = ab;
-		start_bus = le32_to_cpu(ab->descriptor.data_address) - offset;
-		buffer = ab->data;
-
-		ab = ab->next;
-		d = &ab->descriptor;
-		size = start + PAGE_SIZE - ctx->pointer;
-		/* valid buffer data in the next page */
-		rest = le16_to_cpu(d->req_count) - le16_to_cpu(d->res_count);
-		/* what actually fits in this page */
-		size2 = min(rest, (size_t)PAGE_SIZE - offset - size);
-		memmove(buffer, ctx->pointer, size);
-		memcpy(buffer + size, ab->data, size2);
-
-		while (size > 0) {
-			void *next = handle_ar_packet(ctx, buffer);
-			pktsize = next - buffer;
-			if (pktsize >= size) {
-				/*
-				 * We have handled all the data that was
-				 * originally in this page, so we can now
-				 * continue in the next page.
-				 */
-				buffer = next;
-				break;
-			}
-			/* move the next packet to the start of the buffer */
-			memmove(buffer, next, size + size2 - pktsize);
-			size -= pktsize;
-			/* fill up this page again */
-			size3 = min(rest - size2,
-				    (size_t)PAGE_SIZE - offset - size - size2);
-			memcpy(buffer + size + size2,
-			       (void *) ab->data + size2, size3);
-			size2 += size3;
-		}
-
-		if (rest > 0) {
-			/* handle the packets that are fully in the next page */
-			buffer = (void *) ab->data +
-					(buffer - (start + offset + size));
-			end = (void *) ab->data + rest;
-
-			while (buffer < end)
-				buffer = handle_ar_packet(ctx, buffer);
+	p = handle_ar_packets(ctx, p, end);
+	if (p != end) {
+		if (p > end)
+			ar_context_abort(ctx, "inconsistent descriptor");
+		goto error;
+	}
 
-			ctx->current_buffer = ab;
-			ctx->pointer = end;
+	ctx->pointer = p;
+	ar_recycle_buffers(ctx, end_buffer_index);
 
-			ar_context_link_page(ctx, start, start_bus);
-		} else {
-			ctx->pointer = start + PAGE_SIZE;
-		}
-	} else {
-		buffer = ctx->pointer;
-		ctx->pointer = end =
-			(void *) ab + PAGE_SIZE - le16_to_cpu(res_count);
+	return;
 
-		while (buffer < end)
-			buffer = handle_ar_packet(ctx, buffer);
-	}
+error:
+	ctx->pointer = NULL;
 }
 
-static int ar_context_init(struct ar_context *ctx,
-			   struct fw_ohci *ohci, u32 regs)
+static int ar_context_init(struct ar_context *ctx, struct fw_ohci *ohci,
+			   unsigned int descriptors_offset, u32 regs)
 {
-	struct ar_buffer ab;
+	unsigned int i;
+	dma_addr_t dma_addr;
+	struct page *pages[AR_BUFFERS + AR_WRAPAROUND_PAGES];
+	struct descriptor *d;
 
 	ctx->regs        = regs;
 	ctx->ohci        = ohci;
-	ctx->last_buffer = &ab;
 	tasklet_init(&ctx->tasklet, ar_context_tasklet, (unsigned long)ctx);
 
-	ar_context_add_page(ctx);
-	ar_context_add_page(ctx);
-	ctx->current_buffer = ab.next;
-	ctx->pointer = ctx->current_buffer->data;
+	for (i = 0; i < AR_BUFFERS; i++) {
+		ctx->pages[i] = alloc_page(GFP_KERNEL | GFP_DMA32);
+		if (!ctx->pages[i])
+			goto out_of_memory;
+		dma_addr = dma_map_page(ohci->card.device, ctx->pages[i],
+					0, PAGE_SIZE, DMA_FROM_DEVICE);
+		if (dma_mapping_error(ohci->card.device, dma_addr)) {
+			__free_page(ctx->pages[i]);
+			ctx->pages[i] = NULL;
+			goto out_of_memory;
+		}
+		set_page_private(ctx->pages[i], dma_addr);
+	}
+
+	for (i = 0; i < AR_BUFFERS; i++)
+		pages[i]              = ctx->pages[i];
+	for (i = 0; i < AR_WRAPAROUND_PAGES; i++)
+		pages[AR_BUFFERS + i] = ctx->pages[i];
+	ctx->buffer = vm_map_ram(pages, AR_BUFFERS + AR_WRAPAROUND_PAGES,
+				 -1, PAGE_KERNEL_RO);
+	if (!ctx->buffer)
+		goto out_of_memory;
+
+	ctx->descriptors     = ohci->misc_buffer     + descriptors_offset;
+	ctx->descriptors_bus = ohci->misc_buffer_bus + descriptors_offset;
+
+	for (i = 0; i < AR_BUFFERS; i++) {
+		d = &ctx->descriptors[i];
+		d->req_count      = cpu_to_le16(PAGE_SIZE);
+		d->control        = cpu_to_le16(DESCRIPTOR_INPUT_MORE |
+						DESCRIPTOR_STATUS |
+						DESCRIPTOR_BRANCH_ALWAYS);
+		d->data_address   = cpu_to_le32(ar_buffer_bus(ctx, i));
+		d->branch_address = cpu_to_le32(ctx->descriptors_bus +
+			ar_next_buffer_index(i) * sizeof(struct descriptor));
+	}
 
 	return 0;
+
+out_of_memory:
+	ar_context_release(ctx);
+
+	return -ENOMEM;
 }
 
 static void ar_context_run(struct ar_context *ctx)
 {
-	struct ar_buffer *ab = ctx->current_buffer;
-	dma_addr_t ab_bus;
-	size_t offset;
+	unsigned int i;
+
+	for (i = 0; i < AR_BUFFERS; i++)
+		ar_context_link_page(ctx, i);
 
-	offset = offsetof(struct ar_buffer, data);
-	ab_bus = le32_to_cpu(ab->descriptor.data_address) - offset;
+	ctx->pointer = ctx->buffer;
 
-	reg_write(ctx->ohci, COMMAND_PTR(ctx->regs), ab_bus | 1);
+	reg_write(ctx->ohci, COMMAND_PTR(ctx->regs), ctx->descriptors_bus | 1);
 	reg_write(ctx->ohci, CONTROL_SET(ctx->regs), CONTEXT_RUN);
 	flush_writes(ctx->ohci);
 }
@@ -1042,6 +1173,7 @@ static void context_run(struct context *ctx, u32 extra)
 		  le32_to_cpu(ctx->last->branch_address));
 	reg_write(ohci, CONTROL_CLEAR(ctx->regs), ~0);
 	reg_write(ohci, CONTROL_SET(ctx->regs), CONTEXT_RUN | extra);
+	ctx->running = true;
 	flush_writes(ohci);
 }
 
@@ -1069,6 +1201,7 @@ static void context_stop(struct context *ctx)
 	int i;
 
 	reg_write(ctx->ohci, CONTROL_CLEAR(ctx->regs), CONTEXT_RUN);
+	ctx->running = false;
 	flush_writes(ctx->ohci);
 
 	for (i = 0; i < 10; i++) {
@@ -1099,7 +1232,6 @@ static int at_context_queue_packet(struct context *ctx,
 	struct descriptor *d, *last;
 	__le32 *header;
 	int z, tcode;
-	u32 reg;
 
 	d = context_get_descriptors(ctx, 4, &d_bus);
 	if (d == NULL) {
@@ -1113,21 +1245,27 @@ static int at_context_queue_packet(struct context *ctx,
 	/*
 	 * The DMA format for asyncronous link packets is different
 	 * from the IEEE1394 layout, so shift the fields around
-	 * accordingly.  If header_length is 8, it's a PHY packet, to
-	 * which we need to prepend an extra quadlet.
+	 * accordingly.
 	 */
 
+	tcode = (packet->header[0] >> 4) & 0x0f;
 	header = (__le32 *) &d[1];
-	switch (packet->header_length) {
-	case 16:
-	case 12:
+	switch (tcode) {
+	case TCODE_WRITE_QUADLET_REQUEST:
+	case TCODE_WRITE_BLOCK_REQUEST:
+	case TCODE_WRITE_RESPONSE:
+	case TCODE_READ_QUADLET_REQUEST:
+	case TCODE_READ_BLOCK_REQUEST:
+	case TCODE_READ_QUADLET_RESPONSE:
+	case TCODE_READ_BLOCK_RESPONSE:
+	case TCODE_LOCK_REQUEST:
+	case TCODE_LOCK_RESPONSE:
 		header[0] = cpu_to_le32((packet->header[0] & 0xffff) |
 					(packet->speed << 16));
 		header[1] = cpu_to_le32((packet->header[1] & 0xffff) |
 					(packet->header[0] & 0xffff0000));
 		header[2] = cpu_to_le32(packet->header[2]);
 
-		tcode = (packet->header[0] >> 4) & 0x0f;
 		if (TCODE_IS_BLOCK_PACKET(tcode))
 			header[3] = cpu_to_le32(packet->header[3]);
 		else
@@ -1136,18 +1274,18 @@ static int at_context_queue_packet(struct context *ctx,
 		d[0].req_count = cpu_to_le16(packet->header_length);
 		break;
 
-	case 8:
+	case TCODE_LINK_INTERNAL:
 		header[0] = cpu_to_le32((OHCI1394_phy_tcode << 4) |
 					(packet->speed << 16));
-		header[1] = cpu_to_le32(packet->header[0]);
-		header[2] = cpu_to_le32(packet->header[1]);
+		header[1] = cpu_to_le32(packet->header[1]);
+		header[2] = cpu_to_le32(packet->header[2]);
 		d[0].req_count = cpu_to_le16(12);
 
-		if (is_ping_packet(packet->header))
+		if (is_ping_packet(&packet->header[1]))
 			d[0].control |= cpu_to_le16(DESCRIPTOR_PING);
 		break;
 
-	case 4:
+	case TCODE_STREAM_DATA:
 		header[0] = cpu_to_le32((packet->header[0] & 0xffff) |
 					(packet->speed << 16));
 		header[1] = cpu_to_le32(packet->header[0] & 0xffff0000);
@@ -1197,6 +1335,8 @@ static int at_context_queue_packet(struct context *ctx,
 	 * some controllers (like a JMicron JMB381 PCI-e) misbehave and wind
 	 * up stalling out.  So we just bail out in software and try again
 	 * later, and everyone is happy.
+	 * FIXME: Test of IntEvent.busReset may no longer be necessary since we
+	 *        flush AT queues in bus_reset_tasklet.
 	 * FIXME: Document how the locking works.
 	 */
 	if (ohci->generation != packet->generation ||
@@ -1210,14 +1350,23 @@ static int at_context_queue_packet(struct context *ctx,
 
 	context_append(ctx, d, z, 4 - z);
 
-	/* If the con