Merge branch 'dmaengine' into async-tx-next

Conflicts:
	crypto/async_tx/async_xor.c
	drivers/dma/ioat/dma_v2.h
	drivers/dma/ioat/pci.c
	drivers/md/raid5.c

13 files changed, 397 insertions, 211 deletions

diff --git a/drivers/net/wimax/i2400m/control.c b/drivers/net/wimax/i2400m/control.c
index b3cadb626fe..07308686dbc 100644
--- a/drivers/net/wimax/i2400m/control.c
+++ b/drivers/net/wimax/i2400m/control.c
@@ -292,8 +292,6 @@ void i2400m_report_tlv_system_state(struct i2400m *i2400m,
 
 	d_fnstart(3, dev, "(i2400m %p ss %p [%u])\n", i2400m, ss, i2400m_state);
 
-	if (unlikely(i2400m->ready == 0))	/* act if up */
-		goto out;
 	if (i2400m->state != i2400m_state) {
 		i2400m->state = i2400m_state;
 		wake_up_all(&i2400m->state_wq);
@@ -341,7 +339,6 @@ void i2400m_report_tlv_system_state(struct i2400m *i2400m,
 		i2400m->bus_reset(i2400m, I2400M_RT_WARM);
 		break;
 	};
-out:
 	d_fnend(3, dev, "(i2400m %p ss %p [%u]) = void\n",
 		i2400m, ss, i2400m_state);
 }
@@ -372,8 +369,6 @@ void i2400m_report_tlv_media_status(struct i2400m *i2400m,
 
 	d_fnstart(3, dev, "(i2400m %p ms %p [%u])\n", i2400m, ms, status);
 
-	if (unlikely(i2400m->ready == 0))	/* act if up */
-		goto out;
 	switch (status) {
 	case I2400M_MEDIA_STATUS_LINK_UP:
 		netif_carrier_on(net_dev);
@@ -393,14 +388,59 @@ void i2400m_report_tlv_media_status(struct i2400m *i2400m,
 		dev_err(dev, "HW BUG? unknown media status %u\n",
 			status);
 	};
-out:
 	d_fnend(3, dev, "(i2400m %p ms %p [%u]) = void\n",
 		i2400m, ms, status);
 }
 
 
 /*
- * Parse a 'state report' and extract carrier on/off information
+ * Process a TLV from a 'state report'
+ *
+ * @i2400m: device descriptor
+ * @tlv: pointer to the TLV header; it has been already validated for
+ *     consistent size.
+ * @tag: for error messages
+ *
+ * Act on the TLVs from a 'state report'.
+ */
+static
+void i2400m_report_state_parse_tlv(struct i2400m *i2400m,
+				   const struct i2400m_tlv_hdr *tlv,
+				   const char *tag)
+{
+	struct device *dev = i2400m_dev(i2400m);
+	const struct i2400m_tlv_media_status *ms;
+	const struct i2400m_tlv_system_state *ss;
+	const struct i2400m_tlv_rf_switches_status *rfss;
+
+	if (0 == i2400m_tlv_match(tlv, I2400M_TLV_SYSTEM_STATE, sizeof(*ss))) {
+		ss = container_of(tlv, typeof(*ss), hdr);
+		d_printf(2, dev, "%s: system state TLV "
+			 "found (0x%04x), state 0x%08x\n",
+			 tag, I2400M_TLV_SYSTEM_STATE,
+			 le32_to_cpu(ss->state));
+		i2400m_report_tlv_system_state(i2400m, ss);
+	}
+	if (0 == i2400m_tlv_match(tlv, I2400M_TLV_RF_STATUS, sizeof(*rfss))) {
+		rfss = container_of(tlv, typeof(*rfss), hdr);
+		d_printf(2, dev, "%s: RF status TLV "
+			 "found (0x%04x), sw 0x%02x hw 0x%02x\n",
+			 tag, I2400M_TLV_RF_STATUS,
+			 le32_to_cpu(rfss->sw_rf_switch),
+			 le32_to_cpu(rfss->hw_rf_switch));
+		i2400m_report_tlv_rf_switches_status(i2400m, rfss);
+	}
+	if (0 == i2400m_tlv_match(tlv, I2400M_TLV_MEDIA_STATUS, sizeof(*ms))) {
+		ms = container_of(tlv, typeof(*ms), hdr);
+		d_printf(2, dev, "%s: Media Status TLV: %u\n",
+			 tag, le32_to_cpu(ms->media_status));
+		i2400m_report_tlv_media_status(i2400m, ms);
+	}
+}
+
+
+/*
+ * Parse a 'state report' and extract information
  *
  * @i2400m: device descriptor
  * @l3l4_hdr: pointer to message; it has been already validated for
@@ -409,13 +449,7 @@ out:
  *        declaration is assumed to be congruent with @size (as in
  *        sizeof(*l3l4_hdr) + l3l4_hdr->length == size)
  *
- * Extract from the report state the system state TLV and infer from
- * there if we have a carrier or not. Update our local state and tell
- * netdev.
- *
- * When setting the carrier, it's fine to set OFF twice (for example),
- * as netif_carrier_off() will not generate two OFF events (just on
- * the transitions).
+ * Walk over the TLVs in a report state and act on them.
  */
 static
 void i2400m_report_state_hook(struct i2400m *i2400m,
@@ -424,9 +458,6 @@ void i2400m_report_state_hook(struct i2400m *i2400m,
 {
 	struct device *dev = i2400m_dev(i2400m);
 	const struct i2400m_tlv_hdr *tlv;
-	const struct i2400m_tlv_system_state *ss;
-	const struct i2400m_tlv_rf_switches_status *rfss;
-	const struct i2400m_tlv_media_status *ms;
 	size_t tlv_size = le16_to_cpu(l3l4_hdr->length);
 
 	d_fnstart(4, dev, "(i2400m %p, l3l4_hdr %p, size %zu, %s)\n",
@@ -434,34 +465,8 @@ void i2400m_report_state_hook(struct i2400m *i2400m,
 	tlv = NULL;
 
 	while ((tlv = i2400m_tlv_buffer_walk(i2400m, &l3l4_hdr->pl,
-					     tlv_size, tlv))) {
-		if (0 == i2400m_tlv_match(tlv, I2400M_TLV_SYSTEM_STATE,
-					  sizeof(*ss))) {
-			ss = container_of(tlv, typeof(*ss), hdr);
-			d_printf(2, dev, "%s: system state TLV "
-				 "found (0x%04x), state 0x%08x\n",
-				 tag, I2400M_TLV_SYSTEM_STATE,
-				 le32_to_cpu(ss->state));
-			i2400m_report_tlv_system_state(i2400m, ss);
-		}
-		if (0 == i2400m_tlv_match(tlv, I2400M_TLV_RF_STATUS,
-					  sizeof(*rfss))) {
-			rfss = container_of(tlv, typeof(*rfss), hdr);
-			d_printf(2, dev, "%s: RF status TLV "
-				 "found (0x%04x), sw 0x%02x hw 0x%02x\n",
-				 tag, I2400M_TLV_RF_STATUS,
-				 le32_to_cpu(rfss->sw_rf_switch),
-				 le32_to_cpu(rfss->hw_rf_switch));
-			i2400m_report_tlv_rf_switches_status(i2400m, rfss);
-		}
-		if (0 == i2400m_tlv_match(tlv, I2400M_TLV_MEDIA_STATUS,
-					  sizeof(*ms))) {
-			ms = container_of(tlv, typeof(*ms), hdr);
-			d_printf(2, dev, "%s: Media Status TLV: %u\n",
-				 tag, le32_to_cpu(ms->media_status));
-			i2400m_report_tlv_media_status(i2400m, ms);
-		}
-	}
+					     tlv_size, tlv)))
+		i2400m_report_state_parse_tlv(i2400m, tlv, tag);
 	d_fnend(4, dev, "(i2400m %p, l3l4_hdr %p, size %zu, %s) = void\n",
 		i2400m, l3l4_hdr, size, tag);
 }
@@ -500,8 +505,15 @@ void i2400m_report_hook(struct i2400m *i2400m,
 	 * it. */
 	case I2400M_MT_REPORT_POWERSAVE_READY:	/* zzzzz */
 		if (l3l4_hdr->status == cpu_to_le16(I2400M_MS_DONE_OK)) {
-			d_printf(1, dev, "ready for powersave, requesting\n");
-			i2400m_cmd_enter_powersave(i2400m);
+			if (i2400m_power_save_disabled)
+				d_printf(1, dev, "ready for powersave, "
+					 "not requesting (disabled by module "
+					 "parameter)\n");
+			else {
+				d_printf(1, dev, "ready for powersave, "
+					 "requesting\n");
+				i2400m_cmd_enter_powersave(i2400m);
+			}
 		}
 		break;
 	};
@@ -683,8 +695,9 @@ struct sk_buff *i2400m_msg_to_dev(struct i2400m *i2400m,
 	d_fnstart(3, dev, "(i2400m %p buf %p len %zu)\n",
 		  i2400m, buf, buf_len);
 
+	rmb();		/* Make sure we see what i2400m_dev_reset_handle() */
 	if (i2400m->boot_mode)
-		return ERR_PTR(-ENODEV);
+		return ERR_PTR(-EL3RST);
 
 	msg_l3l4_hdr = buf;
 	/* Check msg & payload consistency */
@@ -721,6 +734,8 @@ struct sk_buff *i2400m_msg_to_dev(struct i2400m *i2400m,
 		ack_timeout = HZ;
 	};
 
+	if (unlikely(i2400m->trace_msg_from_user))
+		wimax_msg(&i2400m->wimax_dev, "echo", buf, buf_len, GFP_KERNEL);
 	/* The RX path in rx.c will put any response for this message
 	 * in i2400m->ack_skb and wake us up. If we cancel the wait,
 	 * we need to change the value of i2400m->ack_skb to something
@@ -755,6 +770,9 @@ struct sk_buff *i2400m_msg_to_dev(struct i2400m *i2400m,
 	ack_l3l4_hdr = wimax_msg_data_len(ack_skb, &ack_len);
 
 	/* Check the ack and deliver it if it is ok */
+	if (unlikely(i2400m->trace_msg_from_user))
+		wimax_msg(&i2400m->wimax_dev, "echo",
+			  ack_l3l4_hdr, ack_len, GFP_KERNEL);
 	result = i2400m_msg_size_check(i2400m, ack_l3l4_hdr, ack_len);
 	if (result < 0) {
 		dev_err(dev, "HW BUG? reply to message 0x%04x: %d\n",
@@ -1379,16 +1397,16 @@ error:
  *
  * @i2400m: device descriptor
  *
- * Gracefully stops the device, moving it to the lowest power
- * consumption state possible.
+ * Release resources acquired during the running of the device; in
+ * theory, should also tell the device to go to sleep, switch off the
+ * radio, all that, but at this point, in most cases (driver
+ * disconnection, reset handling) we can't even talk to the device.
  */
 void i2400m_dev_shutdown(struct i2400m *i2400m)
 {
-	int result = -ENODEV;
 	struct device *dev = i2400m_dev(i2400m);
 
 	d_fnstart(3, dev, "(i2400m %p)\n", i2400m);
-	result = i2400m->bus_reset(i2400m, I2400M_RT_WARM);
-	d_fnend(3, dev, "(i2400m %p) = void [%d]\n", i2400m, result);
+	d_fnend(3, dev, "(i2400m %p) = void\n", i2400m);
 	return;
 }
diff --git a/drivers/net/wimax/i2400m/driver.c b/drivers/net/wimax/i2400m/driver.c
index 07a54bad237..304f0443ca4 100644
--- a/drivers/net/wimax/i2400m/driver.c
+++ b/drivers/net/wimax/i2400m/driver.c
@@ -62,6 +62,7 @@
  *   unregister_netdev()
  */
 #include "i2400m.h"
+#include <linux/etherdevice.h>
 #include <linux/wimax/i2400m.h>
 #include <linux/module.h>
 #include <linux/moduleparam.h>
@@ -81,6 +82,14 @@ module_param_named(rx_reorder_disabled, i2400m_rx_reorder_disabled, int, 0644);
 MODULE_PARM_DESC(rx_reorder_disabled,
 		 "If true, RX reordering will be disabled.");
 
+int i2400m_power_save_disabled;	/* 0 (power saving enabled) by default */
+module_param_named(power_save_disabled, i2400m_power_save_disabled, int, 0644);
+MODULE_PARM_DESC(power_save_disabled,
+		 "If true, the driver will not tell the device to enter "
+		 "power saving mode when it reports it is ready for it. "
+		 "False by default (so the device is told to do power "
+		 "saving).");
+
 /**
  * i2400m_queue_work - schedule work on a i2400m's queue
  *
@@ -171,7 +180,6 @@ int i2400m_schedule_work(struct i2400m *i2400m,
 	int result;
 	struct i2400m_work *iw;
 
-	BUG_ON(i2400m->work_queue == NULL);
 	result = -ENOMEM;
 	iw = kzalloc(sizeof(*iw), gfp_flags);
 	if (iw == NULL)
@@ -234,9 +242,6 @@ int i2400m_op_msg_from_user(struct wimax_dev *wimax_dev,
 	result = PTR_ERR(ack_skb);
 	if (IS_ERR(ack_skb))
 		goto error_msg_to_dev;
-	if (unlikely(i2400m->trace_msg_from_user))
-		wimax_msg(&i2400m->wimax_dev, "trace",
-			  msg_buf, msg_len, GFP_KERNEL);
 	result = wimax_msg_send(&i2400m->wimax_dev, ack_skb);
 error_msg_to_dev:
 	d_fnend(4, dev, "(wimax_dev %p [i2400m %p] msg_buf %p msg_len %zu "
@@ -379,6 +384,11 @@ error:
  * Uploads firmware and brings up all the resources needed to be able
  * to communicate with the device.
  *
+ * The workqueue has to be setup early, at least before RX handling
+ * (it's only real user for now) so it can process reports as they
+ * arrive. We also want to destroy it if we retry, to make sure it is
+ * flushed...easier like this.
+ *
  * TX needs to be setup before the bus-specific code (otherwise on
  * shutdown, the bus-tx code could try to access it).
  */
@@ -389,7 +399,7 @@ int __i2400m_dev_start(struct i2400m *i2400m, enum i2400m_bri flags)
 	struct wimax_dev *wimax_dev = &i2400m->wimax_dev;
 	struct net_device *net_dev = wimax_dev->net_dev;
 	struct device *dev = i2400m_dev(i2400m);
-	int times = 3;
+	int times = i2400m->bus_bm_retries;
 
 	d_fnstart(3, dev, "(i2400m %p)\n", i2400m);
 retry:
@@ -404,15 +414,15 @@ retry:
 	result = i2400m_rx_setup(i2400m);
 	if (result < 0)
 		goto error_rx_setup;
-	result = i2400m->bus_dev_start(i2400m);
-	if (result < 0)
-		goto error_bus_dev_start;
 	i2400m->work_queue = create_singlethread_workqueue(wimax_dev->name);
 	if (i2400m->work_queue == NULL) {
 		result = -ENOMEM;
 		dev_err(dev, "cannot create workqueue\n");
 		goto error_create_workqueue;
 	}
+	result = i2400m->bus_dev_start(i2400m);
+	if (result < 0)
+		goto error_bus_dev_start;
 	result = i2400m_firmware_check(i2400m);	/* fw versions ok? */
 	if (result < 0)
 		goto error_fw_check;
@@ -434,17 +444,17 @@ retry:
 error_dev_initialize:
 error_check_mac_addr:
 error_fw_check:
-	destroy_workqueue(i2400m->work_queue);
-error_create_workqueue:
 	i2400m->bus_dev_stop(i2400m);
 error_bus_dev_start:
+	destroy_workqueue(i2400m->work_queue);
+error_create_workqueue:
 	i2400m_rx_release(i2400m);
 error_rx_setup:
 	i2400m_tx_release(i2400m);
 error_tx_setup:
 error_bootstrap:
-	if (result == -ERESTARTSYS && times-- > 0) {
-		flags = I2400M_BRI_SOFT;
+	if (result == -EL3RST && times-- > 0) {
+		flags = I2400M_BRI_SOFT|I2400M_BRI_MAC_REINIT;
 		goto retry;
 	}
 	d_fnend(3, dev, "(net_dev %p [i2400m %p]) = %d\n",
@@ -473,7 +483,9 @@ int i2400m_dev_start(struct i2400m *i2400m, enum i2400m_bri bm_flags)
  *
  * Returns: 0 if ok, < 0 errno code on error.
  *
- * Releases all the resources allocated to communicate with the device.
+ * Releases all the resources allocated to communicate with the
+ * device. Note we cannot destroy the workqueue earlier as until RX is
+ * fully destroyed, it could still try to schedule jobs.
  */
 static
 void __i2400m_dev_stop(struct i2400m *i2400m)
@@ -485,8 +497,8 @@ void __i2400m_dev_stop(struct i2400m *i2400m)
 	wimax_state_change(wimax_dev, __WIMAX_ST_QUIESCING);
 	i2400m_dev_shutdown(i2400m);
 	i2400m->ready = 0;
-	destroy_workqueue(i2400m->work_queue);
 	i2400m->bus_dev_stop(i2400m);
+	destroy_workqueue(i2400m->work_queue);
 	i2400m_rx_release(i2400m);
 	i2400m_tx_release(i2400m);
 	wimax_state_change(wimax_dev, WIMAX_ST_DOWN);
@@ -548,7 +560,7 @@ void __i2400m_dev_reset_handle(struct work_struct *ws)
 		 * i2400m_dev_stop() [we are shutting down anyway, so
 		 * ignore it] or we are resetting somewhere else. */
 		dev_err(dev, "device rebooted\n");
-		i2400m_msg_to_dev_cancel_wait(i2400m, -ERESTARTSYS);
+		i2400m_msg_to_dev_cancel_wait(i2400m, -EL3RST);
 		complete(&i2400m->msg_completion);
 		goto out;
 	}
@@ -598,6 +610,8 @@ out:
  */
 int i2400m_dev_reset_handle(struct i2400m *i2400m)
 {
+	i2400m->boot_mode = 1;
+	wmb();		/* Make sure i2400m_msg_to_dev() sees boot_mode */
 	return i2400m_schedule_work(i2400m, __i2400m_dev_reset_handle,
 				    GFP_ATOMIC);
 }
@@ -650,6 +664,7 @@ int i2400m_setup(struct i2400m *i2400m, enum i2400m_bri bm_flags)
 	result = i2400m_read_mac_addr(i2400m);
 	if (result < 0)
 		goto error_read_mac_addr;
+	random_ether_addr(i2400m->src_mac_addr);
 
 	result = register_netdev(net_dev);	/* Okey dokey, bring it up */
 	if (result < 0) {
diff --git a/drivers/net/wimax/i2400m/fw.c b/drivers/net/wimax/i2400m/fw.c
index 675c6ce810c..e81750e5445 100644
--- a/drivers/net/wimax/i2400m/fw.c
+++ b/drivers/net/wimax/i2400m/fw.c
@@ -397,7 +397,7 @@ static int i2400m_download_chunk(struct i2400m *i2400m, const void *chunk,
 				 unsigned int direct, unsigned int do_csum)
 {
 	int ret;
-	size_t chunk_len = ALIGN(__chunk_len, I2400M_PL_PAD);
+	size_t chunk_len = ALIGN(__chunk_len, I2400M_PL_ALIGN);
 	struct device *dev = i2400m_dev(i2400m);
 	struct {
 		struct i2400m_bootrom_header cmd;
@@ -532,14 +532,14 @@ int i2400m_dnload_finalize(struct i2400m *i2400m,
 	cmd = (void *) bcf + offset;
 	if (i2400m->sboot == 0) {
 		struct i2400m_bootrom_header jump_ack;
-		d_printf(3, dev, "unsecure boot, jumping to 0x%08x\n",
+		d_printf(1, dev, "unsecure boot, jumping to 0x%08x\n",
 			le32_to_cpu(cmd->target_addr));
 		i2400m_brh_set_opcode(cmd, I2400M_BRH_JUMP);
 		cmd->data_size = 0;
 		ret = i2400m_bm_cmd(i2400m, cmd, sizeof(*cmd),
 				    &jump_ack, sizeof(jump_ack), 0);
 	} else {
-		d_printf(3, dev, "secure boot, jumping to 0x%08x\n",
+		d_printf(1, dev, "secure boot, jumping to 0x%08x\n",
 			 le32_to_cpu(cmd->target_addr));
 		cmd_buf = i2400m->bm_cmd_buf;
 		memcpy(&cmd_buf->cmd, cmd, sizeof(*cmd));
@@ -696,8 +696,7 @@ error_dev_gone:
 	return result;
 
 error_timeout:
-	dev_err(dev, "Timed out waiting for reboot ack, resetting\n");
-	i2400m->bus_reset(i2400m, I2400M_RT_BUS);
+	dev_err(dev, "Timed out waiting for reboot ack\n");
 	result = -ETIMEDOUT;
 	goto exit_timeout;
 }
@@ -770,40 +769,21 @@ error_read_mac:
 static
 int i2400m_dnload_init_nonsigned(struct i2400m *i2400m)
 {
-#define POKE(a, d) {					\
-	.address = cpu_to_le32(a),		\
-	.data = cpu_to_le32(d)		\
-}
-	static const struct {
-		__le32 address;
-		__le32 data;
-	} i2400m_pokes[] = {
-		POKE(0x081A58, 0xA7810230),
-		POKE(0x080040, 0x00000000),
-		POKE(0x080048, 0x00000082),
-		POKE(0x08004C, 0x0000081F),
-		POKE(0x080054, 0x00000085),
-		POKE(0x080058, 0x00000180),
-		POKE(0x08005C, 0x00000018),
-		POKE(0x080060, 0x00000010),
-		POKE(0x080574, 0x00000001),
-		POKE(0x080550, 0x00000005),
-		POKE(0xAE0000, 0x00000000),
-	};
-#undef POKE
-	unsigned i;
-	int ret;
+	unsigned i = 0;
+	int ret = 0;
 	struct device *dev = i2400m_dev(i2400m);
-
-	dev_warn(dev, "WARNING!!! non-signed boot UNTESTED PATH!\n");
-
 	d_fnstart(5, dev, "(i2400m %p)\n", i2400m);
-	for (i = 0; i < ARRAY_SIZE(i2400m_pokes); i++) {
-		ret = i2400m_download_chunk(i2400m, &i2400m_pokes[i].data,
-					    sizeof(i2400m_pokes[i].data),
-					    i2400m_pokes[i].address, 1, 1);
-		if (ret < 0)
-			break;
+	if (i2400m->bus_bm_pokes_table) {
+		while (i2400m->bus_bm_pokes_table[i].address) {
+			ret = i2400m_download_chunk(
+				i2400m,
+				&i2400m->bus_bm_pokes_table[i].data,
+				sizeof(i2400m->bus_bm_pokes_table[i].data),
+				i2400m->bus_bm_pokes_table[i].address, 1, 1);
+			if (ret < 0)
+				break;
+			i++;
+		}
 	}
 	d_fnend(5, dev, "(i2400m %p) = %d\n", i2400m, ret);
 	return ret;
@@ -980,11 +960,12 @@ int i2400m_fw_dnload(struct i2400m *i2400m, const struct i2400m_bcf_hdr *bcf,
 {
 	int ret = 0;
 	struct device *dev = i2400m_dev(i2400m);
-	int count = I2400M_BOOT_RETRIES;
+	int count = i2400m->bus_bm_retries;
 
 	d_fnstart(5, dev, "(i2400m %p bcf %p size %zu)\n",
 		  i2400m, bcf, bcf_size);
 	i2400m->boot_mode = 1;
+	wmb();		/* Make sure other readers see it */
 hw_reboot:
 	if (count-- == 0) {
 		ret = -ERESTARTSYS;
@@ -1033,6 +1014,7 @@ hw_reboot:
 	d_printf(2, dev, "fw %s successfully uploaded\n",
 		 i2400m->fw_name);
 	i2400m->boot_mode = 0;
+	wmb();		/* Make sure i2400m_msg_to_dev() sees boot_mode */
 error_dnload_finalize:
 error_dnload_bcf:
 error_dnload_init:
diff --git a/drivers/net/wimax/i2400m/i2400m-sdio.h b/drivers/net/wimax/i2400m/i2400m-sdio.h
index 08c2fb73923..9c4e3189f7b 100644
--- a/drivers/net/wimax/i2400m/i2400m-sdio.h
+++ b/drivers/net/wimax/i2400m/i2400m-sdio.h
@@ -78,6 +78,8 @@ enum {
 	/* The number of ticks to wait for the device to signal that
 	 * it is ready */
 	I2400MS_INIT_SLEEP_INTERVAL = 10,
+	/* How long to wait for the device to settle after reset */
+	I2400MS_SETTLE_TIME = 40,
 };
 
 
@@ -105,6 +107,10 @@ struct i2400ms {
 	char tx_wq_name[32];
 
 	struct dentry *debugfs_dentry;
+
+	wait_queue_head_t bm_wfa_wq;
+	int bm_wait_result;
+	size_t bm_ack_size;
 };
 
 
@@ -129,4 +135,7 @@ extern ssize_t i2400ms_bus_bm_cmd_send(struct i2400m *,
 extern ssize_t i2400ms_bus_bm_wait_for_ack(struct i2400m *,
 					   struct i2400m_bootrom_header *,
 					   size_t);
+extern void i2400ms_bus_bm_release(struct i2400m *);
+extern int i2400ms_bus_bm_setup(struct i2400m *);
+
 #endif /* #ifndef __I2400M_SDIO_H__ */
diff --git a/drivers/net/wimax/i2400m/i2400m.h b/drivers/net/wimax/i2400m/i2400m.h
index 3ae2df38b59..60330f313f2 100644
--- a/drivers/net/wimax/i2400m/i2400m.h
+++ b/drivers/net/wimax/i2400m/i2400m.h
@@ -150,11 +150,33 @@
 enum {
 	/* Firmware uploading */
 	I2400M_BOOT_RETRIES = 3,
+	I3200_BOOT_RETRIES = 3,
 	/* Size of the Boot Mode Command buffer */
 	I2400M_BM_CMD_BUF_SIZE = 16 * 1024,
 	I2400M_BM_ACK_BUF_SIZE = 256,
 };
 
+/**
+ * struct i2400m_poke_table - Hardware poke table for the Intel 2400m
+ *
+ * This structure will be used to create a device specific poke table
+ * to put the device in a consistant state at boot time.
+ *
+ * @address: The device address to poke
+ *
+ * @data: The data value to poke to the device address
+ *
+ */
+struct i2400m_poke_table{
+	__le32 address;
+	__le32 data;
+};
+
+#define I2400M_FW_POKE(a, d) {		\
+	.address = cpu_to_le32(a),	\
+	.data = cpu_to_le32(d)		\
+}
+
 
 /**
  * i2400m_reset_type - methods to reset a device
@@ -224,6 +246,17 @@ struct i2400m_roq;
  *     process, so it cannot rely on common infrastructure being laid
  *     out.
  *
+ * @bus_bm_retries: [fill] How many times shall a firmware upload /
+ *     device initialization be retried? Different models of the same
+ *     device might need different values, hence it is set by the
+ *     bus-specific driver. Note this value is used in two places,
+ *     i2400m_fw_dnload() and __i2400m_dev_start(); they won't become
+ *     multiplicative (__i2400m_dev_start() calling N times
+ *     i2400m_fw_dnload() and this trying N times to download the
+ *     firmware), as if __i2400m_dev_start() only retries if the
+ *     firmware crashed while initializing the device (not in a
+ *     general case).
+ *
  * @bus_bm_cmd_send: [fill] Function called to send a boot-mode
  *     command. Flags are defined in 'enum i2400m_bm_cmd_flags'. This
  *     is synchronous and has to return 0 if ok or < 0 errno code in
@@ -252,6 +285,12 @@ struct i2400m_roq;
  *     address provided in boot mode is kind of broken and needs to
  *     be re-read later on.
  *
+ * @bus_bm_pokes_table: [fill/optional] A table of device addresses
+ *     and values that will be poked at device init time to move the
+ *     device to the correct state for the type of boot/firmware being
+ *     used.  This table MUST be terminated with (0x000000,
+ *     0x00000000) or bad things will happen.
+ *
  *
  * @wimax_dev: WiMAX generic device for linkage into the kernel WiMAX
  *     stack. Due to the way a net_device is allocated, we need to
@@ -323,6 +362,10 @@ struct i2400m_roq;
  *     delivered. Then the driver can release them to the host. See
  *     drivers/net/i2400m/rx.c for details.
  *
+ * @src_mac_addr: MAC address used to make ethernet packets be coming
+ *     from. This is generated at i2400m_setup() time and used during
+ *     the life cycle of the instance. See i2400m_fake_eth_header().
+ *
  * @init_mutex: Mutex used for serializing the device bringup
  *     sequence; this way if the device reboots in the middle, we
  *     don't try to do a bringup again while we are tearing down the
@@ -389,12 +432,14 @@ struct i2400m {
 	unsigned ready:1;		/* all probing steps done */
 	unsigned rx_reorder:1;		/* RX reorder is enabled */
 	u8 trace_msg_from_user;		/* echo rx msgs to 'trace' pipe */
-					/* typed u8 so debugfs/u8 can tweak */
+					/* typed u8 so /sys/kernel/debug/u8 can tweak */
 	enum i2400m_system_state state;
 	wait_queue_head_t state_wq;	/* Woken up when on state updates */
 
 	size_t bus_tx_block_size;
 	size_t bus_pl_size_max;
+	unsigned bus_bm_retries;
+
 	int (*bus_dev_start)(struct i2400m *);
 	void (*bus_dev_stop)(struct i2400m *);
 	void (*bus_tx_kick)(struct i2400m *);
@@ -406,6 +451,7 @@ struct i2400m {
 				       struct i2400m_bootrom_header *, size_t);
 	const char **bus_fw_names;
 	unsigned bus_bm_mac_addr_impaired:1;
+	const struct i2400m_poke_table *bus_bm_pokes_table;
 
 	spinlock_t tx_lock;		/* protect TX state */
 	void *tx_buf;
@@ -421,6 +467,7 @@ struct i2400m {
 	unsigned rx_pl_num, rx_pl_max, rx_pl_min,
 		rx_num, rx_size_acc, rx_size_min, rx_size_max;
 	struct i2400m_roq *rx_roq;	/* not under rx_lock! */
+	u8 src_mac_addr[ETH_HLEN];
 
 	struct mutex msg_mutex;		/* serialize command execution */
 	struct completion msg_completion;
@@ -704,6 +751,7 @@ static const __le32 i2400m_SBOOT_BARKER[4] = {
 	cpu_to_le32(I2400M_SBOOT_BARKER)
 };
 
+extern int i2400m_power_save_disabled;
 
 /*
  * Utility functions
diff --git a/drivers/net/wimax/i2400m/netdev.c b/drivers/net/wimax/i2400m/netdev.c
index 6b1fe7a81f2..9653f478b38 100644
--- a/drivers/net/wimax/i2400m/netdev.c
+++ b/drivers/net/wimax/i2400m/netdev.c
@@ -404,10 +404,12 @@ static
 void i2400m_rx_fake_eth_header(struct net_device *net_dev,
 			       void *_eth_hdr, __be16 protocol)
 {
+	struct i2400m *i2400m = net_dev_to_i2400m(net_dev);
 	struct ethhdr *eth_hdr = _eth_hdr;
 
 	memcpy(eth_hdr->h_dest, net_dev->dev_addr, sizeof(eth_hdr->h_dest));
-	memset(eth_hdr->h_source, 0, sizeof(eth_hdr->h_dest));
+	memcpy(eth_hdr->h_source, i2400m->src_mac_addr,
+	       sizeof(eth_hdr->h_source));
 	eth_hdr->h_proto = protocol;
 }
 
diff --git a/drivers/net/wimax/i2400m/op-rfkill.c b/drivers/net/wimax/i2400m/op-rfkill.c
index 487ec58cea4..43927b5d7ad 100644
--- a/drivers/net/wimax/i2400m/op-rfkill.c
+++ b/drivers/net/wimax/i2400m/op-rfkill.c
@@ -54,8 +54,10 @@ int i2400m_radio_is(struct i2400m *i2400m, enum wimax_rf_state state)
 		/* state == WIMAX_RF_ON */
 		return i2400m->state != I2400M_SS_RF_OFF
 			&& i2400m->state != I2400M_SS_RF_SHUTDOWN;
-	else
+	else {
 		BUG();
+		return -EINVAL;	/* shut gcc warnings on certain arches */
+	}
 }
 
 
diff --git a/drivers/net/wimax/i2400m/rx.c b/drivers/net/wimax/i2400m/rx.c
index f9fc3890232..07c32e68909 100644
--- a/drivers/net/wimax/i2400m/rx.c
+++ b/drivers/net/wimax/i2400m/rx.c
@@ -177,7 +177,8 @@ void i2400m_report_hook_work(struct work_struct *ws)
 	struct i2400m_work *iw =
 		container_of(ws, struct i2400m_work, ws);
 	struct i2400m_report_hook_args *args = (void *) iw->pl;
-	i2400m_report_hook(iw->i2400m, args->l3l4_hdr, args->size);
+	if (iw->i2400m->ready)
+		i2400m_report_hook(iw->i2400m, args->l3l4_hdr, args->size);
 	kfree_skb(args->skb_rx);
 	i2400m_put(iw->i2400m);
 	kfree(iw);
@@ -309,6 +310,9 @@ void i2400m_rx_ctl(struct i2400m *i2400m, struct sk_buff *skb_rx,
 		skb_get(skb_rx);
 		i2400m_queue_work(i2400m, i2400m_report_hook_work,
 				  GFP_KERNEL, &args, sizeof(args));
+		if (unlikely(i2400m->trace_msg_from_user))
+			wimax_msg(&i2400m->wimax_dev, "echo",
+				  l3l4_hdr, size, GFP_KERNEL);
 		result = wimax_msg(&i2400m->wimax_dev, NULL, l3l4_hdr, size,
 				   GFP_KERNEL);
 		if (result < 0)
@@ -1144,7 +1148,7 @@ int i2400m_rx(struct i2400m *i2400m, struct sk_buff *skb)
 	num_pls = le16_to_cpu(msg_hdr->num_pls);
 	pl_itr = sizeof(*msg_hdr) +	/* Check payload descriptor(s) */
 		num_pls * sizeof(msg_hdr->pld[0]);
-	pl_itr = ALIGN(pl_itr, I2400M_PL_PAD);
+	pl_itr = ALIGN(pl_itr, I2400M_PL_ALIGN);
 	if (pl_itr > skb->len) {	/* got all the payload descriptors? */
 		dev_err(dev, "RX: HW BUG? message too short (%u bytes) for "
 			"%u payload descriptors (%zu each, total %zu)\n",
@@ -1162,7 +1166,7 @@ int i2400m_rx(struct i2400m *i2400m, struct sk_buff *skb)
 		single_last = num_pls == 1 || i == num_pls - 1;
 		i2400m_rx_payload(i2400m, skb, single_last, &msg_hdr->pld[i],
 				  skb->data + pl_itr);
-		pl_itr += ALIGN(pl_size, I2400M_PL_PAD);
+		pl_itr += ALIGN(pl_size, I2400M_PL_ALIGN);
 		cond_resched();		/* Don't monopolize */
 	}
 	kfree_skb(skb);
diff --git a/drivers/net/wimax/i2400m/sdio-fw.c b/drivers/net/wimax/i2400m/sdio-fw.c
index 3487205d8f5..7d6ec0f475f 100644
--- a/drivers/net/wimax/i2400m/sdio-fw.c
+++ b/drivers/net/wimax/i2400m/sdio-fw.c
@@ -46,17 +46,24 @@
  * Inaky Perez-Gonzalez <inaky.perez-gonzalez@intel.com>
  *  - SDIO rehash for changes in the bus-driver model
  *
+ * Dirk Brandewie <dirk.j.brandewie@intel.com>
+ *  - Make it IRQ based, not polling
+ *
  * THE PROCEDURE
  *
  * See fw.c for the generic description of this procedure.
  *
  * This file implements only the SDIO specifics. It boils down to how
  * to send a command and waiting for an acknowledgement from the
- * device. We do polled reads.
+ * device.
+ *
+ * All this code is sequential -- all i2400ms_bus_bm_*() functions are
+ * executed in the same thread, except i2400ms_bm_irq() [on its own by
+ * the SDIO driver]. This makes it possible to avoid locking.
  *
  * COMMAND EXECUTION
  *
- * THe generic firmware upload code will call i2400m_bus_bm_cmd_send()
+ * The generic firmware upload code will call i2400m_bus_bm_cmd_send()
  * to send commands.
  *
  * The SDIO devices expects things in 256 byte blocks, so it will pad
@@ -64,12 +71,15 @@
  *
  * ACK RECEPTION
  *
- * This works in polling mode -- the fw loader says when to wait for
- * data and for that it calls i2400ms_bus_bm_wait_for_ack().
+ * This works in IRQ mode -- the fw loader says when to wait for data
+ * and for that it calls i2400ms_bus_bm_wait_for_ack().
  *
- * This will poll the device for data until it is received. We need to
- * receive at least as much bytes as where asked for (although it'll
- * always be a multiple of 256 bytes).
+ * This checks if there is any data available (RX size > 0); if not,
+ * waits for the IRQ handler to notify about it. Once there is data,
+ * it is read and passed to the caller. Doing it this way we don't
+ * need much coordination/locking, and it makes it much more difficult
+ * for an interrupt to be lost and the wait_for_ack() function getting
+ * stuck even when data is pending.
  */
 #include <linux/mmc/sdio_func.h>
 #include "i2400m-sdio.h"
@@ -78,6 +88,7 @@
 #define D_SUBMODULE fw
 #include "sdio-debug-levels.h"
 
+
 /*
  * Send a boot-mode command to the SDIO function
  *
@@ -139,7 +150,7 @@ error_too_big:
 
 
 /*
- * Read an ack from the device's boot-mode (polling)
+ * Read an ack from the device's boot-mode
  *
  * @i2400m:
  * @_ack: pointer to where to store the read data
@@ -150,75 +161,49 @@ error_too_big:
  * The ACK for a BM command is always at least sizeof(*ack) bytes, so
  * check for that. We don't need to check for device reboots
  *
- * NOTE: We do an artificial timeout of 1 sec over the SDIO timeout;
- *     this way we have control over it...there is no way that I know
- *     of setting an SDIO transaction timeout.
  */
 ssize_t i2400ms_bus_bm_wait_for_ack(struct i2400m *i2400m,
 				    struct i2400m_bootrom_header *ack,
 				    size_t ack_size)
 {
-	int result;
-	ssize_t rx_size;
-	u64 timeout;
+	ssize_t result;
 	struct i2400ms *i2400ms = container_of(i2400m, struct i2400ms, i2400m);
 	struct sdio_func *func = i2400ms->func;
 	struct device *dev = &func->dev;
+	int size;
 
 	BUG_ON(sizeof(*ack) > ack_size);
 
 	d_fnstart(5, dev, "(i2400m %p ack %p size %zu)\n",
 		  i2400m, ack, ack_size);
 
-	timeout = get_jiffies_64() + 2 * HZ;
-	sdio_claim_host(func);
-	while (1) {
-		if (time_after64(get_jiffies_64(), timeout)) {
-			rx_size = -ETIMEDOUT;
-			dev_err(dev, "timeout waiting for ack data\n");
-			goto error_timedout;
-		}
+	spin_lock(&i2400m->rx_lock);
+	i2400ms->bm_ack_size = -EINPROGRESS;
+	spin_unlock(&i2400m->rx_lock);
 
-		/* Find the RX size, check if it fits or not -- it if
-		 * doesn't fit, fail, as we have no way to dispose of
-		 * the extra data. */
-		rx_size = __i2400ms_rx_get_size(i2400ms);
-		if (rx_size < 0)
-			goto error_rx_get_size;
-		result = -ENOSPC;		/* Check it fits */
-		if (rx_size < sizeof(*ack)) {
-			rx_size = -EIO;
-			dev_err(dev, "HW BUG? received is too small (%zu vs "
-				"%zu needed)\n", sizeof(*ack), rx_size);
-			goto error_too_small;
-		}
-		if (rx_size > I2400M_BM_ACK_BUF_SIZE) {
-			dev_err(dev, "SW BUG? BM_ACK_BUF is too small (%u vs "
-				"%zu needed)\n", I2400M_BM_ACK_BUF_SIZE,
-				rx_size);
-			goto error_too_small;
-		}
+	result = wait_event_timeout(i2400ms->bm_wfa_wq,
+				    i2400ms->bm_ack_size != -EINPROGRESS,
+				    2 * HZ);
+	if (result == 0) {
+		result = -ETIMEDOUT;
+		dev_err(dev, "BM: error waiting for an ack\n");
+		goto error_timeout;
+	}
 
-		/* Read it */
-		result = sdio_memcpy_fromio(func, i2400m->bm_ack_buf,
-					    I2400MS_DATA_ADDR, rx_size);
-		if (result == -ETIMEDOUT || result == -ETIME)
-			continue;
-		if (result < 0) {
-			dev_err(dev, "BM SDIO receive (%zu B) failed: %d\n",
-				rx_size, result);
-			goto error_read;
-		} else
-			break;
+	spin_lock(&i2400m->rx_lock);
+	result = i2400ms->bm_ack_size;
+	BUG_ON(result == -EINPROGRESS);
+	if (result < 0)        /* so we exit when rx_release() is called */
+		dev_err(dev, "BM: %s failed: %zd\n", __func__, result);
+	else {
+		size = min(ack_size, i2400ms->bm_ack_size);
+		memcpy(ack, i2400m->bm_ack_buf, size);
 	}
-	rx_size = min((ssize_t)ack_size, rx_size);
-	memcpy(ack, i2400m->bm_ack_buf, rx_size);
-error_read:
-error_too_small:
-error_rx_get_size:
-error_timedout:
-	sdio_release_host(func);
-	d_fnend(5, dev, "(i2400m %p ack %p size %zu) = %ld\n",
-		i2400m, ack, ack_size, (long) rx_size);
-	return rx_size;
+	i2400ms->bm_ack_size = -EINPROGRESS;
+	spin_unlock(&i2400m->rx_lock);
+
+error_timeout:
+	d_fnend(5, dev, "(i2400m %p ack %p size %zu) = %zd\n",
+		i2400m, ack, ack_size, result);
+	return result;
 }
diff --git a/drivers/net/wimax/i2400m/sdio-rx.c b/drivers/net/wimax/i2400m/sdio-rx.c
index a3008b904f7..321beadf6e4 100644
--- a/drivers/net/wimax/i2400m/sdio-rx.c
+++ b/drivers/net/wimax/i2400m/sdio-rx.c
@@ -69,6 +69,13 @@
 #define D_SUBMODULE rx
 #include "sdio-debug-levels.h"
 
+static const __le32 i2400m_ACK_BARKER[4] = {
+	__constant_cpu_to_le32(I2400M_ACK_BARKER),
+	__constant_cpu_to_le32(I2400M_ACK_BARKER),
+	__constant_cpu_to_le32(I2400M_ACK_BARKER),
+	__constant_cpu_to_le32(I2400M_ACK_BARKER)
+};
+
 
 /*
  * Read and return the amount of bytes available for RX
@@ -131,25 +138,35 @@ void i2400ms_rx(struct i2400ms *i2400ms)
 		ret = rx_size;
 		goto error_get_size;
 	}
+
 	ret = -ENOMEM;
 	skb = alloc_skb(rx_size, GFP_ATOMIC);
 	if (NULL == skb) {
 		dev_err(dev, "RX: unable to alloc skb\n");
 		goto error_alloc_skb;
 	}
-
 	ret = sdio_memcpy_fromio(func, skb->data,
 				 I2400MS_DATA_ADDR, rx_size);
 	if (ret < 0) {
 		dev_err(dev, "RX: SDIO data read failed: %d\n", ret);
 		goto error_memcpy_fromio;
 	}
-	/* Check if device has reset */
-	if (!memcmp(skb->data, i2400m_NBOOT_BARKER,
-		    sizeof(i2400m_NBOOT_BARKER))
-	    || !memcmp(skb->data, i2400m_SBOOT_BARKER,
-		       sizeof(i2400m_SBOOT_BARKER))) {
+
+	rmb();	/* make sure we get boot_mode from dev_reset_handle */
+	if (i2400m->boot_mode == 1) {
+		spin_lock(&i2400m->rx_lock);
+		i2400ms->bm_ack_size = rx_size;
+		spin_unlock(&i2400m->rx_lock);
+		memcpy(i2400m->bm_ack_buf, skb->data, rx_size);
+		wake_up(&i2400ms->bm_wfa_wq);
+		dev_err(dev, "RX: SDIO boot mode message\n");
+		kfree_skb(skb);
+	} else if (unlikely(!memcmp(skb->data, i2400m_NBOOT_BARKER,
+				    sizeof(i2400m_NBOOT_BARKER))
+			    || !memcmp(skb->data, i2400m_SBOOT_BARKER,
+				       sizeof(i2400m_SBOOT_BARKER)))) {
 		ret = i2400m_dev_reset_handle(i2400m);
+		dev_err(dev, "RX: SDIO reboot barker\n");
 		kfree_skb(skb);
 	} else {
 		skb_put(skb, rx_size);
@@ -179,7 +196,6 @@ void i2400ms_irq(struct sdio_func *func)
 {
 	int ret;
 	struct i2400ms *i2400ms = sdio_get_drvdata(func);
-	struct i2400m *i2400m = &i2400ms->i2400m;
 	struct device *dev = &func->dev;
 	int val;
 
@@ -194,10 +210,7 @@ void i2400ms_irq(struct sdio_func *func)
 		goto error_no_irq;
 	}
 	sdio_writeb(func, 1, I2400MS_INTR_CLEAR_ADDR, &ret);
-	if (WARN_ON(i2400m->boot_mode != 0))
-		dev_err(dev, "RX: SW BUG? boot mode and IRQ is up?\n");
-	else
-		i2400ms_rx(i2400ms);
+	i2400ms_rx(i2400ms);
 error_no_irq:
 	d_fnend(6, dev, "(i2400ms %p) = void\n", i2400ms);
 	return;
@@ -214,8 +227,15 @@ int i2400ms_rx_setup(struct i2400ms *i2400ms)
 	int result;
 	struct sdio_func *func = i2400ms->func;
 	struct device *dev = &func->dev;
+	struct i2400m *i2400m = &i2400ms->i2400m;
 
 	d_fnstart(5, dev, "(i2400ms %p)\n", i2400ms);
+
+	init_waitqueue_head(&i2400ms->bm_wfa_wq);
+	spin_lock(&i2400m->rx_lock);
+	i2400ms->bm_wait_result = -EINPROGRESS;
+	spin_unlock(&i2400m->rx_lock);
+
 	sdio_claim_host(func);
 	result = sdio_claim_irq(func, i2400ms_irq);
 	if (result < 0) {
@@ -245,8 +265,13 @@ void i2400ms_rx_release(struct i2400ms *i2400ms)
 	int result;
 	struct sdio_func *func = i2400ms->func;
 	struct device *dev = &func->dev;
+	struct i2400m *i2400m = &i2400ms->i2400m;
 
 	d_fnstart(5, dev, "(i2400ms %p)\n", i2400ms);
+	spin_lock(&i2400m->rx_lock);
+	i2400ms->bm_ack_size = -EINTR;
+	spin_unlock(&i2400m->rx_lock);
+	wake_up_all(&i2400ms->bm_wfa_wq);
 	sdio_claim_host(func);
 	sdio_writeb(func, 0, I2400MS_INTR_ENABLE_ADDR, &result);
 	sdio_release_irq(func);
diff --git a/drivers/net/wimax/i2400m/sdio.c b/drivers/net/wimax/i2400m/sdio.c
index 5ac5e76701c..2538825d1c6 100644
--- a/drivers/net/wimax/i2400m/sdio.c
+++ b/drivers/net/wimax/i2400m/sdio.c
@@ -78,6 +78,14 @@ static const char *i2400ms_bus_fw_names[] = {
 };
 
 
+static const struct i2400m_poke_table i2400ms_pokes[] = {
+	I2400M_FW_POKE(0x6BE260, 0x00000088),
+	I2400M_FW_POKE(0x080550, 0x00000005),
+	I2400M_FW_POKE(0xAE0000, 0x00000000),
+	I2400M_FW_POKE(0x000000, 0x00000000), /* MUST be 0 terminated or bad
+					       * things will happen */
+};
+
 /*
  * Enable the SDIO function
  *
@@ -148,19 +156,14 @@ int i2400ms_bus_dev_start(struct i2400m *i2400m)
 
 	d_fnstart(3, dev, "(i2400m %p)\n", i2400m);
 	msleep(200);
-	result = i2400ms_rx_setup(i2400ms);
-	if (result < 0)
-		goto error_rx_setup;
 	result = i2400ms_tx_setup(i2400ms);
 	if (result < 0)
 		goto error_tx_setup;
 	d_fnend(3, dev, "(i2400m %p) = %d\n", i2400m, result);
 	return result;
 
-	i2400ms_tx_release(i2400ms);
 error_tx_setup:
-	i2400ms_rx_release(i2400ms);
-error_rx_setup:
+	i2400ms_tx_release(i2400ms);
 	d_fnend(3, dev, "(i2400m %p) = void\n", i2400m);
 	return result;
 }
@@ -174,7 +177,6 @@ void i2400ms_bus_dev_stop(struct i2400m *i2400m)
 	struct device *dev = &func->dev;
 
 	d_fnstart(3, dev, "(i2400m %p)\n", i2400m);
-	i2400ms_rx_release(i2400ms);
 	i2400ms_tx_release(i2400ms);
 	d_fnend(3, dev, "(i2400m %p) = void\n", i2400m);
 }
@@ -255,7 +257,7 @@ error_kzalloc:
 static
 int i2400ms_bus_reset(struct i2400m *i2400m, enum i2400m_reset_type rt)
 {
-	int result;
+	int result = 0;
 	struct i2400ms *i2400ms =
 		container_of(i2400m, struct i2400ms, i2400m);
 	struct device *dev = i2400m_dev(i2400m);
@@ -280,8 +282,25 @@ int i2400ms_bus_reset(struct i2400m *i2400m, enum i2400m_reset_type rt)
 					       sizeof(i2400m_COLD_BOOT_BARKER));
 	else if (rt == I2400M_RT_BUS) {
 do_bus_reset:
-		dev_err(dev, "FIXME: SDIO bus reset not implemented\n");
-		result = rt == I2400M_RT_WARM ? -ENODEV : -ENOSYS;
+		/* call netif_tx_disable() before sending IOE disable,
+		 * so that all the tx from network layer are stopped
+		 * while IOE is being reset. Make sure it is called
+		 * only after register_netdev() was issued.
+		 */
+		if (i2400m->wimax_dev.net_dev->reg_state == NETREG_REGISTERED)
+			netif_tx_disable(i2400m->wimax_dev.net_dev);
+
+		i2400ms_rx_release(i2400ms);
+		sdio_claim_host(i2400ms->func);
+		sdio_disable_func(i2400ms->func);
+		sdio_release_host(i2400ms->func);
+
+		/* Wait for the device to settle */
+		msleep(40);
+
+		result = i2400ms_enable_function(i2400ms->func);
+		if (result >= 0)
+			i2400ms_rx_setup(i2400ms);
 	} else
 		BUG();
 	if (result < 0 && rt != I2400M_RT_BUS) {
@@ -404,24 +423,32 @@ int i2400ms_probe(struct sdio_func *func,
 	i2400m->bus_dev_stop = i2400ms_bus_dev_stop;
 	i2400m->bus_tx_kick = i2400ms_bus_tx_kick;
 	i2400m->bus_reset = i2400ms_bus_reset;
+	/* The iwmc3200-wimax sometimes requires the driver to try
+	 * hard when we paint it into a corner. */
+	i2400m->bus_bm_retries = I3200_BOOT_RETRIES;
 	i2400m->bus_bm_cmd_send = i2400ms_bus_bm_cmd_send;
 	i2400m->bus_bm_wait_for_ack = i2400ms_bus_bm_wait_for_ack;
 	i2400m->bus_fw_names = i2400ms_bus_fw_names;
 	i2400m->bus_bm_mac_addr_impaired = 1;
-
-	result = i2400ms_enable_function(i2400ms->func);
-	if (result < 0) {
-		dev_err(dev, "Cannot enable SDIO function: %d\n", result);
-		goto error_func_enable;
-	}
+	i2400m->bus_bm_pokes_table = &i2400ms_pokes[0];
 
 	sdio_claim_host(func);
 	result = sdio_set_block_size(func, I2400MS_BLK_SIZE);
+	sdio_release_host(func);
 	if (result < 0) {
 		dev_err(dev, "Failed to set block size: %d\n", result);
 		goto error_set_blk_size;
 	}
-	sdio_release_host(func);
+
+	result = i2400ms_enable_function(i2400ms->func);
+	if (result < 0) {
+		dev_err(dev, "Cannot enable SDIO function: %d\n", result);
+		goto error_func_enable;
+	}
+
+	result = i2400ms_rx_setup(i2400ms);
+	if (result < 0)
+		goto error_rx_setup;
 
 	result = i2400m_setup(i2400m, I2400M_BRI_NO_REBOOT);
 	if (result < 0) {
@@ -440,12 +467,14 @@ int i2400ms_probe(struct sdio_func *func,
 error_debugfs_add:
 	i2400m_release(i2400m);
 error_setup:
-	sdio_set_drvdata(func, NULL);
+	i2400ms_rx_release(i2400ms);
+error_rx_setup:
 	sdio_claim_host(func);
-error_set_blk_size:
 	sdio_disable_func(func);
 	sdio_release_host(func);
 error_func_enable:
+error_set_blk_size:
+	sdio_set_drvdata(func, NULL);
 	free_netdev(net_dev);
 error_alloc_netdev:
 	return result;
@@ -462,6 +491,7 @@ void i2400ms_remove(struct sdio_func *func)
 
 	d_fnstart(3, dev, "SDIO func %p\n", func);
 	debugfs_remove_recursive(i2400ms->debugfs_dentry);
+	i2400ms_rx_release(i2400ms);
 	i2400m_release(i2400m);
 	sdio_set_drvdata(func, NULL);
 	sdio_claim_host(func);
diff --git a/drivers/net/wimax/i2400m/tx.c b/drivers/net/wimax/i2400m/tx.c
index 613a88ffd65..fa16ccf8e26 100644
--- a/drivers/net/wimax/i2400m/tx.c
+++ b/drivers/net/wimax/i2400m/tx.c
@@ -278,6 +278,48 @@ enum {
 #define TAIL_FULL ((void *)~(unsigned long)NULL)
 
 /*
+ * Calculate how much tail room is available
+ *
+ * Note the trick here. This path is ONLY caleed for Case A (see
+ * i2400m_tx_fifo_push() below), where we have:
+ *
+ *       Case A
+ * N  ___________
+ *   | tail room |
+ *   |           |
+ *   |<-  IN   ->|
+ *   |           |
+ *   |   data    |
+ *   |           |
+ *   |<-  OUT  ->|
+ *   |           |
+ *   | head room |
+ * 0  -----------
+ *
+ * When calculating the tail_room, tx_in might get to be zero if
+ * i2400m->tx_in is right at the end of the buffer (really full
+ * buffer) if there is no head room. In this case, tail_room would be
+ * I2400M_TX_BUF_SIZE, although it is actually zero. Hence the final
+ * mod (%) operation. However, when doing this kind of optimization,
+ * i2400m->tx_in being zero would fail, so we treat is an a special
+ * case.
+ */
+static inline
+size_t __i2400m_tx_tail_room(struct i2400m *i2400m)
+{
+	size_t tail_room;
+	size_t tx_in;
+
+	if (unlikely(i2400m->tx_in) == 0)
+		return I2400M_TX_BUF_SIZE;
+	tx_in = i2400m->tx_in % I2400M_TX_BUF_SIZE;
+	tail_room = I2400M_TX_BUF_SIZE - tx_in;
+	tail_room %= I2400M_TX_BUF_SIZE;
+	return tail_room;
+}
+
+
+/*
  * Allocate @size bytes in the TX fifo, return a pointer to it
  *
  * @i2400m: device descriptor
@@ -338,7 +380,7 @@ void *i2400m_tx_fifo_push(struct i2400m *i2400m, size_t size, size_t padding)
 		return NULL;
 	}
 	/* Is there space at the tail? */
-	tail_room = I2400M_TX_BUF_SIZE - i2400m->tx_in % I2400M_TX_BUF_SIZE;
+	tail_room = __i2400m_tx_tail_room(i2400m);
 	if (tail_room < needed_size) {
 		if (i2400m->tx_out % I2400M_TX_BUF_SIZE
 		    < i2400m->tx_in % I2400M_TX_BUF_SIZE) {
@@ -367,17 +409,29 @@ void *i2400m_tx_fifo_push(struct i2400m *i2400m, size_t size, size_t padding)
  * (I2400M_PL_PAD for the payloads, I2400M_TX_PLD_SIZE for the
  * header).
  *
+ * Tail room can get to be zero if a message was opened when there was
+ * space only for a header. _tx_close() will mark it as to-skip (as it
+ * will have no payloads) and there will be no more space to flush, so
+ * nothing has to be done here. This is probably cheaper than ensuring
+ * in _tx_new() that there is some space for payloads...as we could
+ * always possibly hit the same problem if the payload wouldn't fit.
+ *
  * Note:
  *
  *     Assumes i2400m->tx_lock is taken, and we use that as a barrier
+ *
+ *     This path is only taken for Case A FIFO situations [see
+ *     i2400m_tx_fifo_push()]
  */
 static
 void i2400m_tx_skip_tail(struct i2400m *i2400m)
 {
 	struct device *dev = i2400m_dev(i2400m);
 	size_t tx_in = i2400m->tx_in % I2400M_TX_BUF_SIZE;
-	size_t tail_room = I2400M_TX_BUF_SIZE - tx_in;
+	size_t tail_room = __i2400m_tx_tail_room(i2400m);
 	struct i2400m_msg_hdr *msg = i2400m->tx_buf + tx_in;
+	if (unlikely(tail_room == 0))
+		return;
 	BUG_ON(tail_room < sizeof(*msg));
 	msg->size = tail_room | I2400M_TX_SKIP;
 	d_printf(2, dev, "skip tail: skipping %zu bytes @%zu\n",
@@ -474,10 +528,18 @@ void i2400m_tx_close(struct i2400m *i2400m)
 	struct i2400m_msg_hdr *tx_msg_moved;
 	size_t aligned_size, padding, hdr_size;
 	void *pad_buf;
+	unsigned num_pls;
 
 	if (tx_msg->size & I2400M_TX_SKIP)	/* a skipper? nothing to do */
 		goto out;
-
+	num_pls = le16_to_cpu(tx_msg->num_pls);
+	/* We can get this situation when a new message was started
+	 * and there was no space to add payloads before hitting the
+	 tail (and taking padding into consideration). */
+	if (num_pls == 0) {
+		tx_msg->size |= I2400M_TX_SKIP;
+		goto out;
+	}
 	/* Relocate the message header
 	 *
 	 * Find the current header size, align it to 16 and if we need
@@ -491,7 +553,7 @@ void i2400m_tx_close(struct i2400m *i2400m)
 	 */
 	hdr_size = sizeof(*tx_msg)
 		+ le16_to_cpu(tx_msg->num_pls) * sizeof(tx_msg->pld[0]);
-	hdr_size = ALIGN(hdr_size, I2400M_PL_PAD);
+	hdr_size = ALIGN(hdr_size, I2400M_PL_ALIGN);
 	tx_msg->offset = I2400M_TX_PLD_SIZE - hdr_size;
 	tx_msg_moved = (void *) tx_msg + tx_msg->offset;
 	memmove(tx_msg_moved, tx_msg, hdr_size);
@@ -574,7 +636,7 @@ int i2400m_tx(struct i2400m *i2400m, const void *buf, size_t buf_len,
 
 	d_fnstart(3, dev, "(i2400m %p skb %p [%zu bytes] pt %u)\n",
 		  i2400m, buf, buf_len, pl_type);
-	padded_len = ALIGN(buf_len, I2400M_PL_PAD);
+	padded_len = ALIGN(buf_len, I2400M_PL_ALIGN);
 	d_printf(5, dev, "padded_len %zd buf_len %zd\n", padded_len, buf_len);
 	/* If there is no current TX message, create one; if the
 	 * current one is out of payload slots or we have a singleton,
@@ -591,6 +653,8 @@ try_new:
 		i2400m_tx_close(i2400m);
 		i2400m_tx_new(i2400m);
 	}
+	if (i2400m->tx_msg == NULL)
+		goto error_tx_new;
 	if (i2400m->tx_msg->size + padded_len > I2400M_TX_BUF_SIZE / 2) {
 		d_printf(2, dev, "TX: message too big, going new\n");
 		i2400m_tx_close(i2400m);
@@ -773,7 +837,6 @@ void i2400m_tx_msg_sent(struct i2400m *i2400m)
 	n = i2400m->tx_out / I2400M_TX_BUF_SIZE;
 	i2400m->tx_out %= I2400M_TX_BUF_SIZE;
 	i2400m->tx_in -= n * I2400M_TX_BUF_SIZE;
-	netif_start_queue(i2400m->wimax_dev.net_dev);
 	spin_unlock_irqrestore(&i2400m->tx_lock, flags);
 	d_fnend(3, dev, "(i2400m %p) = void\n", i2400m);
 }
diff --git a/drivers/net/wimax/i2400m/usb.c b/drivers/net/wimax/i2400m/usb.c
index 17851321b7f..cfdaf69da9d 100644
--- a/drivers/net/wimax/i2400m/usb.c
+++ b/drivers/net/wimax/i2400m/usb.c
@@ -254,8 +254,10 @@ do_bus_reset:
 			dev_err(dev, "USB reset failed (%d), giving up!\n",
 				result);
 		}
-	} else
+	} else {
+		result = -EINVAL;	/* shut gcc up in certain arches */
 		BUG();
+	}
 	if (result < 0
 	    && result != -EINVAL	/* device is gone */
 	    && rt != I2400M_RT_BUS) {
@@ -399,6 +401,7 @@ int i2400mu_probe(struct usb_interface *iface,
 	i2400m->bus_dev_stop = i2400mu_bus_dev_stop;
 	i2400m->bus_tx_kick = i2400mu_bus_tx_kick;
 	i2400m->bus_reset = i2400mu_bus_reset;
+	i2400m->bus_bm_retries = I2400M_BOOT_RETRIES;
 	i2400m->bus_bm_cmd_send = i2400mu_bus_bm_cmd_send;
 	i2400m->bus_bm_wait_for_ack = i2400mu_bus_bm_wait_for_ack;
 	i2400m->bus_fw_names = i2400mu_bus_fw_names;