[PATCH 00/33] sfc version 2.3

[PATCH 00/33] sfc version 2.3

[Ben Hutchings]

The big changes here are to add support for Solarflare's new multi-speed
PHY and reference design.  There are also a number of small fixes, clean-
up and new features.

Ben.

Ben Hutchings (31):
  sfc: Board support fixes
  sfc: Change SPI lengths to type size_t
  sfc: Do not use pci_disable_device() to disable bus mastering
  sfc: Remove unneeded register write
  sfc: Correct interpretation of second param to ethtool phys_id()
  sfc: Clean up waits for flash/EEPROM operations
  sfc: Work around unreliable strap pins
  sfc: Restore phy_flash_cfg module parameter
  sfc: Provide hints to irqbalance daemon
  sfc: Abbreviate self-test names so they are not truncated
  sfc: Don't count RX checksum errors during loopback self-test
  sfc: Remove MII extension cruft
  sfc: Add support for MMDs numbered >15
  sfc: Add phy_type device attribute
  sfc: Clean up board identification
  sfc: Clean up MDIO flag setting
  sfc: Add support for sub-10G speeds
  sfc: Implement auto-negotiation
  sfc: Rework MAC, PHY and board event handling
  sfc: Add support for Solarflare 10Xpress SFT9001
  sfc: Add support for SFN4111T
  sfc: Fix format arguments for warning about MSI-X allocation
  sfc: Generate unique names for per-NIC workqueues
  sfc: Remove leading spaces
  sfc: Specify a meaningful component for loopback RX-side and PHY
    tests
  sfc: Use mutex_lock_interruptible() for ethtool EEPROM access
  sfc: Use model numbers for PHY type names
  sfc: Treat probe as unsuccessful if it scheduled a reset
  sfc: Use kzalloc() to ensure struct efx_spi_device is fully
    initialised
  sfc: Fix synchronisation of efx_mtd_{probe,rename,remove}
  sfc: Version 2.3

Neil Turton (1):
  sfc: Add option to use a separate channel for TX completions

Steve Hodgson (1):
  sfc: Make reset_workqueue driver-global rather than per-NIC

 drivers/net/sfc/Makefile        |    4 +-
 drivers/net/sfc/boards.c        |   70 ++---
 drivers/net/sfc/boards.h        |   14 +-
 drivers/net/sfc/efx.c           |  331 ++++++++++++-------
 drivers/net/sfc/enum.h          |   32 +-
 drivers/net/sfc/ethtool.c       |  137 ++++++---
 drivers/net/sfc/falcon.c        |  472 ++++++++++++++++-----------
 drivers/net/sfc/falcon.h        |    3 +
 drivers/net/sfc/falcon_gmac.c   |  229 +++++++++++++
 drivers/net/sfc/falcon_hwdefs.h |  160 +++++++++
 drivers/net/sfc/falcon_xmac.c   |  261 +++------------
 drivers/net/sfc/gmii.h          |  137 +--------
 drivers/net/sfc/mac.h           |   16 +-
 drivers/net/sfc/mdio_10g.c      |  453 ++++++++++++++++++--------
 drivers/net/sfc/mdio_10g.h      |   52 +++-
 drivers/net/sfc/mtd.c           |    2 +-
 drivers/net/sfc/net_driver.h    |  113 ++++++-
 drivers/net/sfc/phy.h           |    7 +-
 drivers/net/sfc/selftest.c      |   41 ++-
 drivers/net/sfc/selftest.h      |    2 -
 drivers/net/sfc/sfe4001.c       |  153 +++++++--
 drivers/net/sfc/spi.h           |    2 +-
 drivers/net/sfc/tenxpress.c     |  686 +++++++++++++++++++++++++++++----------
 drivers/net/sfc/workarounds.h   |   12 +-
 drivers/net/sfc/xfp_phy.c       |   38 +--
 25 files changed, 2232 insertions(+), 1195 deletions(-)
 create mode 100644 drivers/net/sfc/falcon_gmac.c

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

[PATCH 01/33] sfc: Board support fixes

[Ben Hutchings]

Set dummy monitor method for unrecognised boards.

Clean up board resources if efx_pci_probe_main() fails after board has
been initialised.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index e5024bb..bbc8402 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -1743,6 +1743,7 @@ static struct efx_board efx_dummy_board_info = {
 	.init		= efx_port_dummy_op_int,
 	.init_leds	= efx_port_dummy_op_int,
 	.set_fault_led	= efx_port_dummy_op_blink,
+	.monitor	= efx_port_dummy_op_int,
 	.blink		= efx_port_dummy_op_blink,
 	.fini		= efx_port_dummy_op_void,
 };
@@ -1983,6 +1984,7 @@ static int efx_pci_probe_main(struct efx_nic *efx)
 	efx_fini_port(efx);
  fail5:
  fail4:
+	efx->board_info.fini(efx);
  fail3:
 	efx_fini_napi(efx);
  fail2:
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 01/33] sfc: Board support fixes

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:48:26 +0000

> Set dummy monitor method for unrecognised boards.
> 
> Clean up board resources if efx_pci_probe_main() fails after board has
> been initialised.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 02/33] sfc: Change SPI lengths to type size_t

[Ben Hutchings]

Based on a patch by Andrew Morton.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |   19 ++++++++++---------
 1 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 71e0bed..271cbf8 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -1605,7 +1605,7 @@ void falcon_fini_interrupt(struct efx_nic *efx)
  **************************************************************************
  */
 
-#define FALCON_SPI_MAX_LEN ((unsigned) sizeof(efx_oword_t))
+#define FALCON_SPI_MAX_LEN sizeof(efx_oword_t)
 
 /* Wait for SPI command completion */
 static int falcon_spi_wait(struct efx_nic *efx)
@@ -1630,7 +1630,7 @@ static int falcon_spi_wait(struct efx_nic *efx)
 
 int falcon_spi_cmd(const struct efx_spi_device *spi,
 		   unsigned int command, int address,
-		   const void *in, void *out, unsigned int len)
+		   const void *in, void *out, size_t len)
 {
 	struct efx_nic *efx = spi->efx;
 	bool addressed = (address >= 0);
@@ -1686,8 +1686,8 @@ int falcon_spi_cmd(const struct efx_spi_device *spi,
 	return 0;
 }
 
-static unsigned int
-falcon_spi_write_limit(const struct efx_spi_device *spi, unsigned int start)
+static size_t
+falcon_spi_write_limit(const struct efx_spi_device *spi, size_t start)
 {
 	return min(FALCON_SPI_MAX_LEN,
 		   (spi->block_size - (start & (spi->block_size - 1))));
@@ -1725,12 +1725,12 @@ int falcon_spi_fast_wait(const struct efx_spi_device *spi)
 int falcon_spi_read(const struct efx_spi_device *spi, loff_t start,
 		    size_t len, size_t *retlen, u8 *buffer)
 {
-	unsigned int command, block_len, pos = 0;
+	size_t block_len, pos = 0;
+	unsigned int command;
 	int rc = 0;
 
 	while (pos < len) {
-		block_len = min((unsigned int)len - pos,
-				FALCON_SPI_MAX_LEN);
+		block_len = min(len - pos, FALCON_SPI_MAX_LEN);
 
 		command = efx_spi_munge_command(spi, SPI_READ, start + pos);
 		rc = falcon_spi_cmd(spi, command, start + pos, NULL,
@@ -1756,7 +1756,8 @@ int falcon_spi_write(const struct efx_spi_device *spi, loff_t start,
 		     size_t len, size_t *retlen, const u8 *buffer)
 {
 	u8 verify_buffer[FALCON_SPI_MAX_LEN];
-	unsigned int command, block_len, pos = 0;
+	size_t block_len, pos = 0;
+	unsigned int command;
 	int rc = 0;
 
 	while (pos < len) {
@@ -1764,7 +1765,7 @@ int falcon_spi_write(const struct efx_spi_device *spi, loff_t start,
 		if (rc)
 			break;
 
-		block_len = min((unsigned int)len - pos,
+		block_len = min(len - pos,
 				falcon_spi_write_limit(spi, start + pos));
 		command = efx_spi_munge_command(spi, SPI_WRITE, start + pos);
 		rc = falcon_spi_cmd(spi, command, start + pos,
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 02/33] sfc: Change SPI lengths to type size_t

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:48:51 +0000

> Based on a patch by Andrew Morton.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

Re: [PATCH 02/33] sfc: Change SPI lengths to type size_t

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:48:51 +0000

> Based on a patch by Andrew Morton.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Sigh, this isn't what you actually build tested, that's
for sure:

drivers/net/sfc/falcon.c:1664: error: conflicting types for 'falcon_spi_cmd'
drivers/net/sfc/spi.h:71: error: previous declaration of 'falcon_spi_cmd' was here

I'll fix it up this time, but never will I do this again.

Test what you send me.

Re: [PATCH 02/33] sfc: Change SPI lengths to type size_t

[Ben Hutchings]

David Miller wrote:
> From: Ben Hutchings <bhutchings@solarflare.com>
> Date: Fri, 12 Dec 2008 12:48:51 +0000
> 
> > Based on a patch by Andrew Morton.
> > 
> > Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
> 
> Sigh, this isn't what you actually build tested, that's
> for sure:

Of course it is.  But it looks like I forgot to test the final version
of this patch on a 64-bit architecture.  On i386 size_t = unsigned.

> drivers/net/sfc/falcon.c:1664: error: conflicting types for 'falcon_spi_cmd'
> drivers/net/sfc/spi.h:71: error: previous declaration of 'falcon_spi_cmd' was here
> 
> I'll fix it up this time, but never will I do this again.
 
Thanks.

Ben.

> Test what you send me.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

[PATCH 03/33] sfc: Do not use pci_disable_device() to disable bus mastering

[Ben Hutchings]

pci_disable_device() disables many features, like MSI-X, which we
never reenable in efx_reset().  Further, calls to pci_enable_device()
and pci_disable_device() must be matched since the nesting count was
introduced.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |   15 +++++++++++++--
 1 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 271cbf8..ac0b671 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -1344,6 +1344,17 @@ static inline void falcon_irq_ack_a1(struct efx_nic *efx)
 	falcon_readl(efx, &reg, WORK_AROUND_BROKEN_PCI_READS_REG_KER_A1);
 }
 
+static void efx_pci_clear_master(struct pci_dev *dev)
+{
+	u16 command;
+
+	if (dev->is_busmaster &&
+	    !pci_read_config_word(dev, PCI_COMMAND, &command) &&
+	    !pci_write_config_word(dev, PCI_COMMAND,
+				   command & ~PCI_COMMAND_MASTER))
+		dev->is_busmaster = false;
+}
+
 /* Process a fatal interrupt
  * Disable bus mastering ASAP and schedule a reset
  */
@@ -1375,9 +1386,9 @@ static irqreturn_t falcon_fatal_interrupt(struct efx_nic *efx)
 	}
 
 	/* Disable both devices */
-	pci_disable_device(efx->pci_dev);
+	efx_pci_clear_master(efx->pci_dev);
 	if (FALCON_IS_DUAL_FUNC(efx))
-		pci_disable_device(nic_data->pci_dev2);
+		efx_pci_clear_master(nic_data->pci_dev2);
 	falcon_disable_interrupts(efx);
 
 	if (++n_int_errors < FALCON_MAX_INT_ERRORS) {
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 03/33] sfc: Do not use pci_disable_device() to disable bus mastering

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:49:14 +0000

> pci_disable_device() disables many features, like MSI-X, which we
> never reenable in efx_reset().  Further, calls to pci_enable_device()
> and pci_disable_device() must be matched since the nesting count was
> introduced.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Since the PCI layer provides interfaces which do manage
this setting, there is no way you should be doing this
behind it's back.

You know why?

Because someone is going to change some of the internals
of this stuff in the PIC layer and it will break your
driver.

Please instead submit an interface to the PCI layer
maintainers that does what you want, then use it.

Patch not applied.

Re: [PATCH 03/33] sfc: Do not use pci_disable_device() to disable bus mastering

[Ben Hutchings]

David Miller wrote:
> From: Ben Hutchings <bhutchings@solarflare.com>
> Date: Fri, 12 Dec 2008 12:49:14 +0000
> 
> > pci_disable_device() disables many features, like MSI-X, which we
> > never reenable in efx_reset().  Further, calls to pci_enable_device()
> > and pci_disable_device() must be matched since the nesting count was
> > introduced.
> > 
> > Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
> 
> Since the PCI layer provides interfaces which do manage
> this setting, there is no way you should be doing this
> behind it's back.
> 
> You know why?
> 
> Because someone is going to change some of the internals
> of this stuff in the PIC layer and it will break your
> driver.
> 
> Please instead submit an interface to the PCI layer
> maintainers that does what you want, then use it.

I did ask on #linux-pci why this didn't exist; apparently it wasn't
considered to be worth providing.  So obviously I can submit a patch to
the PCI layer but I'm not sure it would be accepted.

Ben.

> Patch not applied.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

[PATCH 04/33] sfc: Remove unneeded register write

[Ben Hutchings]

This was only ever needed for an FPGA version of Falcon.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |    9 ---------
 1 files changed, 0 insertions(+), 9 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index ac0b671..3af361a 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -2881,15 +2881,6 @@ int falcon_init_nic(struct efx_nic *efx)
 	unsigned thresh;
 	int rc;
 
-	/* Set up the address region register. This is only needed
-	 * for the B0 FPGA, but since we are just pushing in the
-	 * reset defaults this may as well be unconditional. */
-	EFX_POPULATE_OWORD_4(temp, ADR_REGION0, 0,
-				   ADR_REGION1, (1 << 16),
-				   ADR_REGION2, (2 << 16),
-				   ADR_REGION3, (3 << 16));
-	falcon_write(efx, &temp, ADR_REGION_REG_KER);
-
 	/* Use on-chip SRAM */
 	falcon_read(efx, &temp, NIC_STAT_REG);
 	EFX_SET_OWORD_FIELD(temp, ONCHIP_SRAM, 1);
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 04/33] sfc: Remove unneeded register write

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:49:31 +0000

> This was only ever needed for an FPGA version of Falcon.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 05/33] sfc: Correct interpretation of second param to ethtool phys_id()

[Ben Hutchings]

A value of 0 means indefinite repetition (until interrupted).

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/ethtool.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index cd92c4d..9c533e0 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -182,12 +182,16 @@ static struct efx_ethtool_stat efx_ethtool_stats[] = {
  */
 
 /* Identify device by flashing LEDs */
-static int efx_ethtool_phys_id(struct net_device *net_dev, u32 seconds)
+static int efx_ethtool_phys_id(struct net_device *net_dev, u32 count)
 {
 	struct efx_nic *efx = netdev_priv(net_dev);
 
 	efx->board_info.blink(efx, 1);
-	schedule_timeout_interruptible(seconds * HZ);
+	set_current_state(TASK_INTERRUPTIBLE);
+	if (count)
+		schedule_timeout(count * HZ);
+	else
+		schedule();
 	efx->board_info.blink(efx, 0);
 	return 0;
 }
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 05/33] sfc: Correct interpretation of second param to ethtool phys_id()

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:49:50 +0000

> A value of 0 means indefinite repetition (until interrupted).
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 06/33] sfc: Make reset_workqueue driver-global rather than per-NIC

[Ben Hutchings]

From: Steve Hodgson <shodgson@solarflare.com>

Each reset is serialised by the rtnl_lock anyway, so there's no win
per-NIC.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c        |   45 ++++++++++++++++++-----------------------
 drivers/net/sfc/net_driver.h |    2 -
 2 files changed, 20 insertions(+), 27 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index bbc8402..127b4da 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -39,6 +39,12 @@
  */
 static struct workqueue_struct *refill_workqueue;
 
+/* Reset workqueue. If any NIC has a hardware failure then a reset will be
+ * queued onto this work queue. This is not a per-nic work queue, because
+ * efx_reset_work() acquires the rtnl lock, so resets are naturally serialised.
+ */
+static struct workqueue_struct *reset_workqueue;
+
 /**************************************************************************
  *
  * Configurable values
@@ -1697,7 +1703,7 @@ void efx_schedule_reset(struct efx_nic *efx, enum reset_type type)
 
 	efx->reset_pending = method;
 
-	queue_work(efx->reset_workqueue, &efx->reset_work);
+	queue_work(reset_workqueue, &efx->reset_work);
 }
 
 /**************************************************************************
@@ -1763,7 +1769,7 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
 	struct efx_channel *channel;
 	struct efx_tx_queue *tx_queue;
 	struct efx_rx_queue *rx_queue;
-	int i, rc;
+	int i;
 
 	/* Initialise common structures */
 	memset(efx, 0, sizeof(*efx));
@@ -1832,33 +1838,14 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
 				  interrupt_mode);
 
 	efx->workqueue = create_singlethread_workqueue("sfc_work");
-	if (!efx->workqueue) {
-		rc = -ENOMEM;
-		goto fail1;
-	}
-
-	efx->reset_workqueue = create_singlethread_workqueue("sfc_reset");
-	if (!efx->reset_workqueue) {
-		rc = -ENOMEM;
-		goto fail2;
-	}
+	if (!efx->workqueue)
+		return -ENOMEM;
 
 	return 0;
-
- fail2:
-	destroy_workqueue(efx->workqueue);
-	efx->workqueue = NULL;
-
- fail1:
-	return rc;
 }
 
 static void efx_fini_struct(struct efx_nic *efx)
 {
-	if (efx->reset_workqueue) {
-		destroy_workqueue(efx->reset_workqueue);
-		efx->reset_workqueue = NULL;
-	}
 	if (efx->workqueue) {
 		destroy_workqueue(efx->workqueue);
 		efx->workqueue = NULL;
@@ -1923,7 +1910,7 @@ static void efx_pci_remove(struct pci_dev *pci_dev)
 	 * scheduled from this point because efx_stop_all() has been
 	 * called, we are no longer registered with driverlink, and
 	 * the net_device's have been removed. */
-	flush_workqueue(efx->reset_workqueue);
+	cancel_work_sync(&efx->reset_work);
 
 	efx_pci_remove_main(efx);
 
@@ -2045,7 +2032,7 @@ static int __devinit efx_pci_probe(struct pci_dev *pci_dev,
 		 * scheduled since efx_stop_all() has been called, and we
 		 * have not and never have been registered with either
 		 * the rtnetlink or driverlink layers. */
-		flush_workqueue(efx->reset_workqueue);
+		cancel_work_sync(&efx->reset_work);
 
 		/* Retry if a recoverably reset event has been scheduled */
 		if ((efx->reset_pending != RESET_TYPE_INVISIBLE) &&
@@ -2121,6 +2108,11 @@ static int __init efx_init_module(void)
 		rc = -ENOMEM;
 		goto err_refill;
 	}
+	reset_workqueue = create_singlethread_workqueue("sfc_reset");
+	if (!reset_workqueue) {
+		rc = -ENOMEM;
+		goto err_reset;
+	}
 
 	rc = pci_register_driver(&efx_pci_driver);
 	if (rc < 0)
@@ -2129,6 +2121,8 @@ static int __init efx_init_module(void)
 	return 0;
 
  err_pci:
+	destroy_workqueue(reset_workqueue);
+ err_reset:
 	destroy_workqueue(refill_workqueue);
  err_refill:
 	unregister_netdevice_notifier(&efx_netdev_notifier);
@@ -2141,6 +2135,7 @@ static void __exit efx_exit_module(void)
 	printk(KERN_INFO "Solarflare NET driver unloading\n");
 
 	pci_unregister_driver(&efx_pci_driver);
+	destroy_workqueue(reset_workqueue);
 	destroy_workqueue(refill_workqueue);
 	unregister_netdevice_notifier(&efx_netdev_notifier);
 
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index e596c9a..f910609 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -635,7 +635,6 @@ union efx_multicast_hash {
  * @legacy_irq: IRQ number
  * @workqueue: Workqueue for port reconfigures and the HW monitor.
  *	Work items do not hold and must not acquire RTNL.
- * @reset_workqueue: Workqueue for resets.  Work item will acquire RTNL.
  * @reset_work: Scheduled reset workitem
  * @monitor_work: Hardware monitor workitem
  * @membase_phys: Memory BAR value as physical address
@@ -711,7 +710,6 @@ struct efx_nic {
 	const struct efx_nic_type *type;
 	int legacy_irq;
 	struct workqueue_struct *workqueue;
-	struct workqueue_struct *reset_workqueue;
 	struct work_struct reset_work;
 	struct delayed_work monitor_work;
 	resource_size_t membase_phys;
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 06/33] sfc: Make reset_workqueue driver-global rather than per-NIC

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:50:38 +0000

> From: Steve Hodgson <shodgson@solarflare.com>
> 
> Each reset is serialised by the rtnl_lock anyway, so there's no win
> per-NIC.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 07/33] sfc: Clean up waits for flash/EEPROM operations

[Ben Hutchings]

Make falcon_spi_wait() ignore the write timer - it is only relevant to
write commands, it only works for the device that contains VPD, and it
might not be initialised properly at all.

Rename falcon_spi_fast_wait() to falcon_spi_wait_write(), reflecting
its use, and make it wait up to 10 ms (not 1 ms) since buffered writes
to EEPROM may take this long to complete.

Make both wait functions sleep instead of busy-waiting.

Replace wait for command completion at top of falcon_spi_cmd() with a
single poll; no command should be running when the function starts.

Correct some comments.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |   58 +++++++++++++++++++++++++++++----------------
 drivers/net/sfc/mtd.c    |    2 +-
 drivers/net/sfc/spi.h    |    2 +-
 3 files changed, 39 insertions(+), 23 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 3af361a..4b2ec12 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -1618,24 +1618,37 @@ void falcon_fini_interrupt(struct efx_nic *efx)
 
 #define FALCON_SPI_MAX_LEN sizeof(efx_oword_t)
 
+static int falcon_spi_poll(struct efx_nic *efx)
+{
+	efx_oword_t reg;
+	falcon_read(efx, &reg, EE_SPI_HCMD_REG_KER);
+	return EFX_OWORD_FIELD(reg, EE_SPI_HCMD_CMD_EN) ? -EBUSY : 0;
+}
+
 /* Wait for SPI command completion */
 static int falcon_spi_wait(struct efx_nic *efx)
 {
-	unsigned long timeout = jiffies + DIV_ROUND_UP(HZ, 10);
-	efx_oword_t reg;
-	bool cmd_en, timer_active;
+	/* Most commands will finish quickly, so we start polling at
+	 * very short intervals.  Sometimes the command may have to
+	 * wait for VPD or expansion ROM access outside of our
+	 * control, so we allow up to 100 ms. */
+	unsigned long timeout = jiffies + 1 + DIV_ROUND_UP(HZ, 10);
+	int i;
+
+	for (i = 0; i < 10; i++) {
+		if (!falcon_spi_poll(efx))
+			return 0;
+		udelay(10);
+	}
 
 	for (;;) {
-		falcon_read(efx, &reg, EE_SPI_HCMD_REG_KER);
-		cmd_en = EFX_OWORD_FIELD(reg, EE_SPI_HCMD_CMD_EN);
-		timer_active = EFX_OWORD_FIELD(reg, EE_WR_TIMER_ACTIVE);
-		if (!cmd_en && !timer_active)
+		if (!falcon_spi_poll(efx))
 			return 0;
 		if (time_after_eq(jiffies, timeout)) {
 			EFX_ERR(efx, "timed out waiting for SPI\n");
 			return -ETIMEDOUT;
 		}
-		cpu_relax();
+		schedule_timeout_uninterruptible(1);
 	}
 }
 
@@ -1654,8 +1667,8 @@ int falcon_spi_cmd(const struct efx_spi_device *spi,
 		return -EINVAL;
 	BUG_ON(!mutex_is_locked(&efx->spi_lock));
 
-	/* Check SPI not currently being accessed */
-	rc = falcon_spi_wait(efx);
+	/* Check that previous command is not still running */
+	rc = falcon_spi_poll(efx);
 	if (rc)
 		return rc;
 
@@ -1711,26 +1724,29 @@ efx_spi_munge_command(const struct efx_spi_device *spi,
 	return command | (((address >> 8) & spi->munge_address) << 3);
 }
 
-int falcon_spi_fast_wait(const struct efx_spi_device *spi)
+/* Wait up to 10 ms for buffered write completion */
+int falcon_spi_wait_write(const struct efx_spi_device *spi)
 {
+	struct efx_nic *efx = spi->efx;
+	unsigned long timeout = jiffies + 1 + DIV_ROUND_UP(HZ, 100);
 	u8 status;
-	int i, rc;
-
-	/* Wait up to 1000us for flash/EEPROM to finish a fast operation. */
-	for (i = 0; i < 50; i++) {
-		udelay(20);
+	int rc;
 
+	for (;;) {
 		rc = falcon_spi_cmd(spi, SPI_RDSR, -1, NULL,
 				    &status, sizeof(status));
 		if (rc)
 			return rc;
 		if (!(status & SPI_STATUS_NRDY))
 			return 0;
+		if (time_after_eq(jiffies, timeout)) {
+			EFX_ERR(efx, "SPI write timeout on device %d"
+				" last status=0x%02x\n",
+				spi->device_id, status);
+			return -ETIMEDOUT;
+		}
+		schedule_timeout_uninterruptible(1);
 	}
-	EFX_ERR(spi->efx,
-		"timed out waiting for device %d last status=0x%02x\n",
-		spi->device_id, status);
-	return -ETIMEDOUT;
 }
 
 int falcon_spi_read(const struct efx_spi_device *spi, loff_t start,
@@ -1784,7 +1800,7 @@ int falcon_spi_write(const struct efx_spi_device *spi, loff_t start,
 		if (rc)
 			break;
 
-		rc = falcon_spi_fast_wait(spi);
+		rc = falcon_spi_wait_write(spi);
 		if (rc)
 			break;
 
diff --git a/drivers/net/sfc/mtd.c b/drivers/net/sfc/mtd.c
index a1e6c28..665cafb 100644
--- a/drivers/net/sfc/mtd.c
+++ b/drivers/net/sfc/mtd.c
@@ -76,7 +76,7 @@ static int efx_spi_unlock(const struct efx_spi_device *spi)
 	rc = falcon_spi_cmd(spi, SPI_WRSR, -1, &status, NULL, sizeof(status));
 	if (rc)
 		return rc;
-	rc = falcon_spi_fast_wait(spi);
+	rc = falcon_spi_wait_write(spi);
 	if (rc)
 		return rc;
 
diff --git a/drivers/net/sfc/spi.h b/drivers/net/sfc/spi.h
index c4aca13..4e45d54 100644
--- a/drivers/net/sfc/spi.h
+++ b/drivers/net/sfc/spi.h
@@ -69,7 +69,7 @@ struct efx_spi_device {
 
 int falcon_spi_cmd(const struct efx_spi_device *spi, unsigned int command,
 		   int address, const void* in, void *out, unsigned int len);
-int falcon_spi_fast_wait(const struct efx_spi_device *spi);
+int falcon_spi_wait_write(const struct efx_spi_device *spi);
 int falcon_spi_read(const struct efx_spi_device *spi, loff_t start,
 		    size_t len, size_t *retlen, u8 *buffer);
 int falcon_spi_write(const struct efx_spi_device *spi, loff_t start,
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 07/33] sfc: Clean up waits for flash/EEPROM operations

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:51:01 +0000

> Make falcon_spi_wait() ignore the write timer - it is only relevant to
> write commands, it only works for the device that contains VPD, and it
> might not be initialised properly at all.
> 
> Rename falcon_spi_fast_wait() to falcon_spi_wait_write(), reflecting
> its use, and make it wait up to 10 ms (not 1 ms) since buffered writes
> to EEPROM may take this long to complete.
> 
> Make both wait functions sleep instead of busy-waiting.
> 
> Replace wait for command completion at top of falcon_spi_cmd() with a
> single poll; no command should be running when the function starts.
> 
> Correct some comments.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 08/33] sfc: Work around unreliable strap pins

[Ben Hutchings]

The SFC4000 has strap pins indicating the presence of SPI flash and/or
EEPROM.  These pins are also used for GPIO, and in some cases they may
be read wrongly at reset.  However, on production boards it must boot
from one or the other device, so we can assume the boot device is
present and read the board config from there.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |  110 +++++++++++++++++----------------------------
 1 files changed, 42 insertions(+), 68 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 4b2ec12..9563c13 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -70,6 +70,20 @@ static int disable_dma_stats;
 #define RX_DC_ENTRIES_ORDER 2
 #define RX_DC_BASE 0x100000
 
+static const unsigned int
+/* "Large" EEPROM device: Atmel AT25640 or similar
+ * 8 KB, 16-bit address, 32 B write block */
+large_eeprom_type = ((13 << SPI_DEV_TYPE_SIZE_LBN)
+		     | (2 << SPI_DEV_TYPE_ADDR_LEN_LBN)
+		     | (5 << SPI_DEV_TYPE_BLOCK_SIZE_LBN)),
+/* Default flash device: Atmel AT25F1024
+ * 128 KB, 24-bit address, 32 KB erase block, 256 B write block */
+default_flash_type = ((17 << SPI_DEV_TYPE_SIZE_LBN)
+		      | (3 << SPI_DEV_TYPE_ADDR_LEN_LBN)
+		      | (0x52 << SPI_DEV_TYPE_ERASE_CMD_LBN)
+		      | (15 << SPI_DEV_TYPE_ERASE_SIZE_LBN)
+		      | (8 << SPI_DEV_TYPE_BLOCK_SIZE_LBN));
+
 /* RX FIFO XOFF watermark
  *
  * When the amount of the RX FIFO increases used increases past this
@@ -2281,12 +2295,15 @@ int falcon_read_nvram(struct efx_nic *efx, struct falcon_nvconfig *nvconfig_out)
 	__le16 *word, *limit;
 	u32 csum;
 
+	spi = efx->spi_flash ? efx->spi_flash : efx->spi_eeprom;
+	if (!spi)
+		return -EINVAL;
+
 	region = kmalloc(FALCON_NVCONFIG_END, GFP_KERNEL);
 	if (!region)
 		return -ENOMEM;
 	nvconfig = region + NVCONFIG_OFFSET;
 
-	spi = efx->spi_flash ? efx->spi_flash : efx->spi_eeprom;
 	mutex_lock(&efx->spi_lock);
 	rc = falcon_spi_read(spi, 0, FALCON_NVCONFIG_END, NULL, region);
 	mutex_unlock(&efx->spi_lock);
@@ -2724,80 +2741,37 @@ static int falcon_probe_nic_variant(struct efx_nic *efx)
 static void falcon_probe_spi_devices(struct efx_nic *efx)
 {
 	efx_oword_t nic_stat, gpio_ctl, ee_vpd_cfg;
-	bool has_flash, has_eeprom, boot_is_external;
+	int boot_dev;
 
 	falcon_read(efx, &gpio_ctl, GPIO_CTL_REG_KER);
 	falcon_read(efx, &nic_stat, NIC_STAT_REG);
 	falcon_read(efx, &ee_vpd_cfg, EE_VPD_CFG_REG_KER);
 
-	has_flash = EFX_OWORD_FIELD(nic_stat, SF_PRST);
-	has_eeprom = EFX_OWORD_FIELD(nic_stat, EE_PRST);
-	boot_is_external = EFX_OWORD_FIELD(gpio_ctl, BOOTED_USING_NVDEVICE);
-
-	if (has_flash) {
-		/* Default flash SPI device: Atmel AT25F1024
-		 * 128 KB, 24-bit address, 32 KB erase block,
-		 * 256 B write block
-		 */
-		u32 flash_device_type =
-			(17 << SPI_DEV_TYPE_SIZE_LBN)
-			| (3 << SPI_DEV_TYPE_ADDR_LEN_LBN)
-			| (0x52 << SPI_DEV_TYPE_ERASE_CMD_LBN)
-			| (15 << SPI_DEV_TYPE_ERASE_SIZE_LBN)
-			| (8 << SPI_DEV_TYPE_BLOCK_SIZE_LBN);
-
-		falcon_spi_device_init(efx, &efx->spi_flash,
-				       EE_SPI_FLASH, flash_device_type);
-
-		if (!boot_is_external) {
-			/* Disable VPD and set clock dividers to safe
-			 * values for initial programming.
-			 */
-			EFX_LOG(efx, "Booted from internal ASIC settings;"
-				" setting SPI config\n");
-			EFX_POPULATE_OWORD_3(ee_vpd_cfg, EE_VPD_EN, 0,
-					     /* 125 MHz / 7 ~= 20 MHz */
-					     EE_SF_CLOCK_DIV, 7,
-					     /* 125 MHz / 63 ~= 2 MHz */
-					     EE_EE_CLOCK_DIV, 63);
-			falcon_write(efx, &ee_vpd_cfg, EE_VPD_CFG_REG_KER);
-		}
-	}
-
-	if (has_eeprom) {
-		u32 eeprom_device_type;
-
-		/* If it has no flash, it must have a large EEPROM
-		 * for chip config; otherwise check whether 9-bit
-		 * addressing is used for VPD configuration
-		 */
-		if (has_flash &&
-		    (!boot_is_external ||
-		     EFX_OWORD_FIELD(ee_vpd_cfg, EE_VPD_EN_AD9_MODE))) {
-			/* Default SPI device: Atmel AT25040 or similar
-			 * 512 B, 9-bit address, 8 B write block
-			 */
-			eeprom_device_type =
-				(9 << SPI_DEV_TYPE_SIZE_LBN)
-				| (1 << SPI_DEV_TYPE_ADDR_LEN_LBN)
-				| (3 << SPI_DEV_TYPE_BLOCK_SIZE_LBN);
-		} else {
-			/* "Large" SPI device: Atmel AT25640 or similar
-			 * 8 KB, 16-bit address, 32 B write block
-			 */
-			eeprom_device_type =
-				(13 << SPI_DEV_TYPE_SIZE_LBN)
-				| (2 << SPI_DEV_TYPE_ADDR_LEN_LBN)
-				| (5 << SPI_DEV_TYPE_BLOCK_SIZE_LBN);
-		}
-
-		falcon_spi_device_init(efx, &efx->spi_eeprom,
-				       EE_SPI_EEPROM, eeprom_device_type);
-	}
-
-	EFX_LOG(efx, "flash is %s, EEPROM is %s\n",
-		(has_flash ? "present" : "absent"),
-		(has_eeprom ? "present" : "absent"));
+	if (EFX_OWORD_FIELD(gpio_ctl, BOOTED_USING_NVDEVICE)) {
+		boot_dev = (EFX_OWORD_FIELD(nic_stat, SF_PRST) ?
+			    EE_SPI_FLASH : EE_SPI_EEPROM);
+		EFX_LOG(efx, "Booted from %s\n",
+			boot_dev == EE_SPI_FLASH ? "flash" : "EEPROM");
+	} else {
+		/* Disable VPD and set clock dividers to safe
+		 * values for initial programming. */
+		boot_dev = -1;
+		EFX_LOG(efx, "Booted from internal ASIC settings;"
+			" setting SPI config\n");
+		EFX_POPULATE_OWORD_3(ee_vpd_cfg, EE_VPD_EN, 0,
+				     /* 125 MHz / 7 ~= 20 MHz */
+				     EE_SF_CLOCK_DIV, 7,
+				     /* 125 MHz / 63 ~= 2 MHz */
+				     EE_EE_CLOCK_DIV, 63);
+		falcon_write(efx, &ee_vpd_cfg, EE_VPD_CFG_REG_KER);
+	}
+
+	if (boot_dev == EE_SPI_FLASH)
+		falcon_spi_device_init(efx, &efx->spi_flash, EE_SPI_FLASH,
+				       default_flash_type);
+	if (boot_dev == EE_SPI_EEPROM)
+		falcon_spi_device_init(efx, &efx->spi_eeprom, EE_SPI_EEPROM,
+				       large_eeprom_type);
 }
 
 int falcon_probe_nic(struct efx_nic *efx)
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 08/33] sfc: Work around unreliable strap pins

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:51:28 +0000

> The SFC4000 has strap pins indicating the presence of SPI flash and/or
> EEPROM.  These pins are also used for GPIO, and in some cases they may
> be read wrongly at reset.  However, on production boards it must boot
> from one or the other device, so we can assume the boot device is
> present and read the board config from there.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 09/33] sfc: Restore phy_flash_cfg module parameter

[Ben Hutchings]

This is needed for recovery in case a PHY firmware upgrade is aborted.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 127b4da..15ae2fc 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -129,6 +129,10 @@ static unsigned int rss_cpus;
 module_param(rss_cpus, uint, 0444);
 MODULE_PARM_DESC(rss_cpus, "Number of CPUs to use for Receive-Side Scaling");
 
+static int phy_flash_cfg;
+module_param(phy_flash_cfg, int, 0644);
+MODULE_PARM_DESC(phy_flash_cfg, "Set PHYs into reflash mode initially");
+
 /**************************************************************************
  *
  * Utility functions and prototypes
@@ -609,6 +613,9 @@ static int efx_probe_port(struct efx_nic *efx)
 	if (rc)
 		goto err;
 
+	if (phy_flash_cfg)
+		efx->phy_mode = PHY_MODE_SPECIAL;
+
 	/* Sanity check MAC address */
 	if (is_valid_ether_addr(efx->mac_address)) {
 		memcpy(efx->net_dev->dev_addr, efx->mac_address, ETH_ALEN);
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 09/33] sfc: Restore phy_flash_cfg module parameter

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:51:48 +0000

> This is needed for recovery in case a PHY firmware upgrade is aborted.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 13/33] sfc: Don't count RX checksum errors during loopback self-test

[Ben Hutchings]

The loopback self-test checks that IP packets with incorrect checksums
are not altered when sent on a queue with checksum generation off.
These should not contribute to RX error statistics.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 7421945..dbab9db 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -784,15 +784,18 @@ static void falcon_handle_rx_not_ok(struct efx_rx_queue *rx_queue,
 			   rx_ev_buf_owner_id_err | rx_ev_eth_crc_err |
 			   rx_ev_frm_trunc | rx_ev_ip_hdr_chksum_err);
 
-	/* Count errors that are not in MAC stats. */
+	/* Count errors that are not in MAC stats.  Ignore expected
+	 * checksum errors during self-test. */
 	if (rx_ev_frm_trunc)
 		++rx_queue->channel->n_rx_frm_trunc;
 	else if (rx_ev_tobe_disc)
 		++rx_queue->channel->n_rx_tobe_disc;
-	else if (rx_ev_ip_hdr_chksum_err)
-		++rx_queue->channel->n_rx_ip_hdr_chksum_err;
-	else if (rx_ev_tcp_udp_chksum_err)
-		++rx_queue->channel->n_rx_tcp_udp_chksum_err;
+	else if (!efx->loopback_selftest) {
+		if (rx_ev_ip_hdr_chksum_err)
+			++rx_queue->channel->n_rx_ip_hdr_chksum_err;
+		else if (rx_ev_tcp_udp_chksum_err)
+			++rx_queue->channel->n_rx_tcp_udp_chksum_err;
+	}
 	if (rx_ev_ip_frag_err)
 		++rx_queue->channel->n_rx_ip_frag_err;
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 13/33] sfc: Don't count RX checksum errors during loopback self-test

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:53:07 +0000

> The loopback self-test checks that IP packets with incorrect checksums
> are not altered when sent on a queue with checksum generation off.
> These should not contribute to RX error statistics.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 14/33] sfc: Remove MII extension cruft

[Ben Hutchings]

Replace efx_nic::link_options bitfield with link_speed (speed in
Mbit/s) and link_fd (full duplex flag).

Remove broken auto-negotiation functions.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c         |   23 +------
 drivers/net/sfc/ethtool.c     |    1 -
 drivers/net/sfc/falcon.c      |   16 ++---
 drivers/net/sfc/falcon_xmac.c |    1 -
 drivers/net/sfc/gmii.h        |  137 +----------------------------------------
 drivers/net/sfc/net_driver.h  |    6 +-
 drivers/net/sfc/tenxpress.c   |    4 +-
 drivers/net/sfc/xfp_phy.c     |    4 +-
 8 files changed, 18 insertions(+), 174 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 183a440..844580f 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -21,7 +21,6 @@
 #include <linux/ethtool.h>
 #include <linux/topology.h>
 #include "net_driver.h"
-#include "gmii.h"
 #include "ethtool.h"
 #include "tx.h"
 #include "rx.h"
@@ -551,26 +550,8 @@ static void efx_link_status_changed(struct efx_nic *efx)
 
 	/* Status message for kernel log */
 	if (efx->link_up) {
-		struct mii_if_info *gmii = &efx->mii;
-		unsigned adv, lpa;
-		/* NONE here means direct XAUI from the controller, with no
-		 * MDIO-attached device we can query. */
-		if (efx->phy_type != PHY_TYPE_NONE) {
-			adv = gmii_advertised(gmii);
-			lpa = gmii_lpa(gmii);
-		} else {
-			lpa = GM_LPA_10000 | LPA_DUPLEX;
-			adv = lpa;
-		}
-		EFX_INFO(efx, "link up at %dMbps %s-duplex "
-			 "(adv %04x lpa %04x) (MTU %d)%s\n",
-			 (efx->link_options & GM_LPA_10000 ? 10000 :
-			  (efx->link_options & GM_LPA_1000 ? 1000 :
-			   (efx->link_options & GM_LPA_100 ? 100 :
-			    10))),
-			 (efx->link_options & GM_LPA_DUPLEX ?
-			  "full" : "half"),
-			 adv, lpa,
+		EFX_INFO(efx, "link up at %uMbps %s-duplex (MTU %d)%s\n",
+			 efx->link_speed, efx->link_fd ? "full" : "half",
 			 efx->net_dev->mtu,
 			 (efx->promiscuous ? " [PROMISC]" : ""));
 	} else {
diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index 931ce14..43d6d8b 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -16,7 +16,6 @@
 #include "efx.h"
 #include "ethtool.h"
 #include "falcon.h"
-#include "gmii.h"
 #include "spi.h"
 #include "mac.h"
 
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index dbab9db..ac765cb 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -15,11 +15,11 @@
 #include <linux/seq_file.h>
 #include <linux/i2c.h>
 #include <linux/i2c-algo-bit.h>
+#include <linux/mii.h>
 #include "net_driver.h"
 #include "bitfield.h"
 #include "efx.h"
 #include "mac.h"
-#include "gmii.h"
 #include "spi.h"
 #include "falcon.h"
 #include "falcon_hwdefs.h"
@@ -1926,14 +1926,12 @@ void falcon_reconfigure_mac_wrapper(struct efx_nic *efx)
 	int link_speed;
 	bool tx_fc;
 
-	if (efx->link_options & GM_LPA_10000)
-		link_speed = 0x3;
-	else if (efx->link_options & GM_LPA_1000)
-		link_speed = 0x2;
-	else if (efx->link_options & GM_LPA_100)
-		link_speed = 0x1;
-	else
-		link_speed = 0x0;
+	switch (efx->link_speed) {
+	case 10000: link_speed = 3; break;
+	case 1000:  link_speed = 2; break;
+	case 100:   link_speed = 1; break;
+	default:    link_speed = 0; break;
+	}
 	/* MAC_LINK_STATUS controls MAC backpressure but doesn't work
 	 * as advertised.  Disable to ensure packets are not
 	 * indefinitely held and TX queue can be flushed at any point
diff --git a/drivers/net/sfc/falcon_xmac.c b/drivers/net/sfc/falcon_xmac.c
index d401231..4a54d09 100644
--- a/drivers/net/sfc/falcon_xmac.c
+++ b/drivers/net/sfc/falcon_xmac.c
@@ -15,7 +15,6 @@
 #include "falcon_hwdefs.h"
 #include "falcon_io.h"
 #include "mac.h"
-#include "gmii.h"
 #include "mdio_10g.h"
 #include "phy.h"
 #include "boards.h"
diff --git a/drivers/net/sfc/gmii.h b/drivers/net/sfc/gmii.h
index d25bbd1..dfccaa7 100644
--- a/drivers/net/sfc/gmii.h
+++ b/drivers/net/sfc/gmii.h
@@ -1,7 +1,7 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
  * Copyright 2005-2006 Fen Systems Ltd.
- * Copyright 2006 Solarflare Communications Inc.
+ * Copyright 2006-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -57,139 +57,4 @@
 #define ISR_POLARITY_CHG	0x0002	/* Bit 1 - polarity changed */
 #define ISR_JABBER		0x0001	/* Bit 0 - jabber */
 
-/* Logically extended advertisement register */
-#define GM_ADVERTISE_SLCT		ADVERTISE_SLCT
-#define GM_ADVERTISE_CSMA		ADVERTISE_CSMA
-#define GM_ADVERTISE_10HALF		ADVERTISE_10HALF
-#define GM_ADVERTISE_1000XFULL		ADVERTISE_1000XFULL
-#define GM_ADVERTISE_10FULL		ADVERTISE_10FULL
-#define GM_ADVERTISE_1000XHALF		ADVERTISE_1000XHALF
-#define GM_ADVERTISE_100HALF		ADVERTISE_100HALF
-#define GM_ADVERTISE_1000XPAUSE		ADVERTISE_1000XPAUSE
-#define GM_ADVERTISE_100FULL		ADVERTISE_100FULL
-#define GM_ADVERTISE_1000XPSE_ASYM	ADVERTISE_1000XPSE_ASYM
-#define GM_ADVERTISE_100BASE4		ADVERTISE_100BASE4
-#define GM_ADVERTISE_PAUSE_CAP		ADVERTISE_PAUSE_CAP
-#define GM_ADVERTISE_PAUSE_ASYM		ADVERTISE_PAUSE_ASYM
-#define GM_ADVERTISE_RESV		ADVERTISE_RESV
-#define GM_ADVERTISE_RFAULT		ADVERTISE_RFAULT
-#define GM_ADVERTISE_LPACK		ADVERTISE_LPACK
-#define GM_ADVERTISE_NPAGE		ADVERTISE_NPAGE
-#define GM_ADVERTISE_1000FULL		(ADVERTISE_1000FULL << 8)
-#define GM_ADVERTISE_1000HALF		(ADVERTISE_1000HALF << 8)
-#define GM_ADVERTISE_1000		(GM_ADVERTISE_1000FULL | \
-					 GM_ADVERTISE_1000HALF)
-#define GM_ADVERTISE_FULL		(GM_ADVERTISE_1000FULL | \
-					 ADVERTISE_FULL)
-#define GM_ADVERTISE_ALL		(GM_ADVERTISE_1000FULL | \
-					 GM_ADVERTISE_1000HALF | \
-					 ADVERTISE_ALL)
-
-/* Logically extended link partner ability register */
-#define GM_LPA_SLCT			LPA_SLCT
-#define GM_LPA_10HALF			LPA_10HALF
-#define GM_LPA_1000XFULL		LPA_1000XFULL
-#define GM_LPA_10FULL			LPA_10FULL
-#define GM_LPA_1000XHALF		LPA_1000XHALF
-#define GM_LPA_100HALF			LPA_100HALF
-#define GM_LPA_1000XPAUSE		LPA_1000XPAUSE
-#define GM_LPA_100FULL			LPA_100FULL
-#define GM_LPA_1000XPAUSE_ASYM		LPA_1000XPAUSE_ASYM
-#define GM_LPA_100BASE4			LPA_100BASE4
-#define GM_LPA_PAUSE_CAP		LPA_PAUSE_CAP
-#define GM_LPA_PAUSE_ASYM		LPA_PAUSE_ASYM
-#define GM_LPA_RESV			LPA_RESV
-#define GM_LPA_RFAULT			LPA_RFAULT
-#define GM_LPA_LPACK			LPA_LPACK
-#define GM_LPA_NPAGE			LPA_NPAGE
-#define GM_LPA_1000FULL			(LPA_1000FULL << 6)
-#define GM_LPA_1000HALF			(LPA_1000HALF << 6)
-#define GM_LPA_10000FULL		0x00040000
-#define GM_LPA_10000HALF		0x00080000
-#define GM_LPA_DUPLEX			(GM_LPA_1000FULL | GM_LPA_10000FULL \
-					 | LPA_DUPLEX)
-#define GM_LPA_10			(LPA_10FULL | LPA_10HALF)
-#define GM_LPA_100			LPA_100
-#define GM_LPA_1000			(GM_LPA_1000FULL | GM_LPA_1000HALF)
-#define GM_LPA_10000			(GM_LPA_10000FULL | GM_LPA_10000HALF)
-
-/* Retrieve GMII autonegotiation advertised abilities
- *
- * The MII advertisment register (MII_ADVERTISE) is logically extended
- * to include advertisement bits ADVERTISE_1000FULL and
- * ADVERTISE_1000HALF from MII_CTRL1000.  The result can be tested
- * against the GM_ADVERTISE_xxx constants.
- */
-static inline unsigned int gmii_advertised(struct mii_if_info *gmii)
-{
-	unsigned int advertise;
-	unsigned int ctrl1000;
-
-	advertise = gmii->mdio_read(gmii->dev, gmii->phy_id, MII_ADVERTISE);
-	ctrl1000 = gmii->mdio_read(gmii->dev, gmii->phy_id, MII_CTRL1000);
-	return (((ctrl1000 << 8) & GM_ADVERTISE_1000) | advertise);
-}
-
-/* Retrieve GMII autonegotiation link partner abilities
- *
- * The MII link partner ability register (MII_LPA) is logically
- * extended by adding bits LPA_1000HALF and LPA_1000FULL from
- * MII_STAT1000.  The result can be tested against the GM_LPA_xxx
- * constants.
- */
-static inline unsigned int gmii_lpa(struct mii_if_info *gmii)
-{
-	unsigned int lpa;
-	unsigned int stat1000;
-
-	lpa = gmii->mdio_read(gmii->dev, gmii->phy_id, MII_LPA);
-	stat1000 = gmii->mdio_read(gmii->dev, gmii->phy_id, MII_STAT1000);
-	return (((stat1000 << 6) & GM_LPA_1000) | lpa);
-}
-
-/* Calculate GMII autonegotiated link technology
- *
- * "negotiated" should be the result of gmii_advertised() logically
- * ANDed with the result of gmii_lpa().
- *
- * "tech" will be negotiated with the unused bits masked out.  For
- * example, if both ends of the link are capable of both
- * GM_LPA_1000FULL and GM_LPA_100FULL, GM_LPA_100FULL will be masked
- * out.
- */
-static inline unsigned int gmii_nway_result(unsigned int negotiated)
-{
-	unsigned int other_bits;
-
-	/* Mask out the speed and duplexity bits */
-	other_bits = negotiated & ~(GM_LPA_10 | GM_LPA_100 | GM_LPA_1000);
-
-	if (negotiated & GM_LPA_1000FULL)
-		return (other_bits | GM_LPA_1000FULL);
-	else if (negotiated & GM_LPA_1000HALF)
-		return (other_bits | GM_LPA_1000HALF);
-	else
-		return (other_bits | mii_nway_result(negotiated));
-}
-
-/* Calculate GMII non-autonegotiated link technology
- *
- * This provides an equivalent to gmii_nway_result for the case when
- * autonegotiation is disabled.
- */
-static inline unsigned int gmii_forced_result(unsigned int bmcr)
-{
-	unsigned int result;
-	int full_duplex;
-
-	full_duplex = bmcr & BMCR_FULLDPLX;
-	if (bmcr & BMCR_SPEED1000)
-		result = full_duplex ? GM_LPA_1000FULL : GM_LPA_1000HALF;
-	else if (bmcr & BMCR_SPEED100)
-		result = full_duplex ? GM_LPA_100FULL : GM_LPA_100HALF;
-	else
-		result = full_duplex ? GM_LPA_10FULL : GM_LPA_10HALF;
-	return result;
-}
-
 #endif /* EFX_GMII_H */
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index abff908..6cac5ed 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -694,7 +694,8 @@ union efx_multicast_hash {
  * @mii: PHY interface
  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
  * @link_up: Link status
- * @link_options: Link options (MII/GMII format)
+ * @link_fd: Link is full duplex
+ * @link_speed: Link speed (Mbps)
  * @n_link_state_changes: Number of times the link has changed state
  * @promiscuous: Promiscuous flag. Protected by netif_tx_lock.
  * @multicast_hash: Multicast hash table
@@ -772,7 +773,8 @@ struct efx_nic {
 	enum efx_phy_mode phy_mode;
 
 	bool link_up;
-	unsigned int link_options;
+	bool link_fd;
+	unsigned int link_speed;
 	unsigned int n_link_state_changes;
 
 	bool promiscuous;
diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c
index 8d41c29..3fa7ccb 100644
--- a/drivers/net/sfc/tenxpress.c
+++ b/drivers/net/sfc/tenxpress.c
@@ -10,7 +10,6 @@
 #include <linux/delay.h>
 #include <linux/seq_file.h>
 #include "efx.h"
-#include "gmii.h"
 #include "mdio_10g.h"
 #include "falcon.h"
 #include "phy.h"
@@ -362,7 +361,8 @@ static void tenxpress_phy_reconfigure(struct efx_nic *efx)
 	phy_data->loopback_mode = efx->loopback_mode;
 	phy_data->phy_mode = efx->phy_mode;
 	efx->link_up = tenxpress_link_ok(efx, false);
-	efx->link_options = GM_LPA_10000FULL;
+	efx->link_speed = 10000;
+	efx->link_fd = true;
 }
 
 static void tenxpress_phy_clear_interrupt(struct efx_nic *efx)
diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c
index 91f0246..971a24b 100644
--- a/drivers/net/sfc/xfp_phy.c
+++ b/drivers/net/sfc/xfp_phy.c
@@ -14,7 +14,6 @@
 #include <linux/timer.h>
 #include <linux/delay.h>
 #include "efx.h"
-#include "gmii.h"
 #include "mdio_10g.h"
 #include "xenpack.h"
 #include "phy.h"
@@ -154,7 +153,8 @@ static void xfp_phy_reconfigure(struct efx_nic *efx)
 
 	phy_data->phy_mode = efx->phy_mode;
 	efx->link_up = xfp_link_ok(efx);
-	efx->link_options = GM_LPA_10000FULL;
+	efx->link_speed = 10000;
+	efx->link_fd = true;
 }
 
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 14/33] sfc: Remove MII extension cruft

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:53:24 +0000

> Replace efx_nic::link_options bitfield with link_speed (speed in
> Mbit/s) and link_fd (full duplex flag).
> 
> Remove broken auto-negotiation functions.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 15/33] sfc: Add support for MMDs numbered >15

[Ben Hutchings]

Combine DEVS0 and DEVS1 registers into a 32-bit mask instead of
reading just DEVS0.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/mdio_10g.c  |   24 +++++++++++++-----------
 drivers/net/sfc/mdio_10g.h  |    8 ++++----
 drivers/net/sfc/tenxpress.c |    6 +++---
 drivers/net/sfc/xfp_phy.c   |    8 ++++----
 4 files changed, 24 insertions(+), 22 deletions(-)

diff --git a/drivers/net/sfc/mdio_10g.c b/drivers/net/sfc/mdio_10g.c
index 19e2521..1eed484 100644
--- a/drivers/net/sfc/mdio_10g.c
+++ b/drivers/net/sfc/mdio_10g.c
@@ -121,16 +121,18 @@ int mdio_clause45_wait_reset_mmds(struct efx_nic *efx,
 int mdio_clause45_check_mmds(struct efx_nic *efx,
 			     unsigned int mmd_mask, unsigned int fatal_mask)
 {
-	int devices, mmd = 0;
-	int probe_mmd;
+	u32 devices;
+	int mmd = 0, probe_mmd;
 
 	/* Historically we have probed the PHYXS to find out what devices are
 	 * present,but that doesn't work so well if the PHYXS isn't expected
 	 * to exist, if so just find the first item in the list supplied. */
-	probe_mmd = (mmd_mask & MDIO_MMDREG_DEVS0_PHYXS) ? MDIO_MMD_PHYXS :
+	probe_mmd = (mmd_mask & MDIO_MMDREG_DEVS_PHYXS) ? MDIO_MMD_PHYXS :
 	    __ffs(mmd_mask);
-	devices = mdio_clause45_read(efx, efx->mii.phy_id,
-				     probe_mmd, MDIO_MMDREG_DEVS0);
+	devices = (mdio_clause45_read(efx, efx->mii.phy_id,
+				      probe_mmd, MDIO_MMDREG_DEVS0) |
+		   mdio_clause45_read(efx, efx->mii.phy_id,
+				      probe_mmd, MDIO_MMDREG_DEVS1) << 16);
 
 	/* Check all the expected MMDs are present */
 	if (devices < 0) {
@@ -175,14 +177,14 @@ bool mdio_clause45_links_ok(struct efx_nic *efx, unsigned int mmd_mask)
 	else if (efx_phy_mode_disabled(efx->phy_mode))
 		return false;
 	else if (efx->loopback_mode == LOOPBACK_PHYXS)
-		mmd_mask &= ~(MDIO_MMDREG_DEVS0_PHYXS |
-			      MDIO_MMDREG_DEVS0_PCS |
-			      MDIO_MMDREG_DEVS0_PMAPMD);
+		mmd_mask &= ~(MDIO_MMDREG_DEVS_PHYXS |
+			      MDIO_MMDREG_DEVS_PCS |
+			      MDIO_MMDREG_DEVS_PMAPMD);
 	else if (efx->loopback_mode == LOOPBACK_PCS)
-		mmd_mask &= ~(MDIO_MMDREG_DEVS0_PCS |
-			      MDIO_MMDREG_DEVS0_PMAPMD);
+		mmd_mask &= ~(MDIO_MMDREG_DEVS_PCS |
+			      MDIO_MMDREG_DEVS_PMAPMD);
 	else if (efx->loopback_mode == LOOPBACK_PMAPMD)
-		mmd_mask &= ~MDIO_MMDREG_DEVS0_PMAPMD;
+		mmd_mask &= ~MDIO_MMDREG_DEVS_PMAPMD;
 
 	while (mmd_mask) {
 		if (mmd_mask & 1) {
diff --git a/drivers/net/sfc/mdio_10g.h b/drivers/net/sfc/mdio_10g.h
index db9f358..4516383 100644
--- a/drivers/net/sfc/mdio_10g.h
+++ b/drivers/net/sfc/mdio_10g.h
@@ -73,14 +73,14 @@
 #define MDIO_ID_MODEL(_id32)	((_id32 >> 4) & 0x3f)
 #define MDIO_ID_OUI(_id32)	(_id32 >> 10)
 
-/* Bits in MMDREG_DEVS0. Someone thoughtfully layed things out
+/* Bits in MMDREG_DEVS0/1. Someone thoughtfully layed things out
  * so the 'bit present' bit number of an MMD is the number of
  * that MMD */
 #define DEV_PRESENT_BIT(_b) (1 << _b)
 
-#define MDIO_MMDREG_DEVS0_PHYXS	 DEV_PRESENT_BIT(MDIO_MMD_PHYXS)
-#define MDIO_MMDREG_DEVS0_PCS	 DEV_PRESENT_BIT(MDIO_MMD_PCS)
-#define MDIO_MMDREG_DEVS0_PMAPMD DEV_PRESENT_BIT(MDIO_MMD_PMAPMD)
+#define MDIO_MMDREG_DEVS_PHYXS	DEV_PRESENT_BIT(MDIO_MMD_PHYXS)
+#define MDIO_MMDREG_DEVS_PCS	DEV_PRESENT_BIT(MDIO_MMD_PCS)
+#define MDIO_MMDREG_DEVS_PMAPMD	DEV_PRESENT_BIT(MDIO_MMD_PMAPMD)
 
 /* Bits in MMDREG_STAT2 */
 #define MDIO_MMDREG_STAT2_PRESENT_VAL	(2)
diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c
index 3fa7ccb..197b544 100644
--- a/drivers/net/sfc/tenxpress.c
+++ b/drivers/net/sfc/tenxpress.c
@@ -19,9 +19,9 @@
 
 /* We expect these MMDs to be in the package */
 /* AN not here as mdio_check_mmds() requires STAT2 support */
-#define TENXPRESS_REQUIRED_DEVS (MDIO_MMDREG_DEVS0_PMAPMD | \
-				 MDIO_MMDREG_DEVS0_PCS    | \
-				 MDIO_MMDREG_DEVS0_PHYXS)
+#define TENXPRESS_REQUIRED_DEVS (MDIO_MMDREG_DEVS_PMAPMD	| \
+				 MDIO_MMDREG_DEVS_PCS		| \
+				 MDIO_MMDREG_DEVS_PHYXS)
 
 #define TENXPRESS_LOOPBACKS ((1 << LOOPBACK_PHYXS) |	\
 			     (1 << LOOPBACK_PCS) |	\
diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c
index 971a24b..0413d13 100644
--- a/drivers/net/sfc/xfp_phy.c
+++ b/drivers/net/sfc/xfp_phy.c
@@ -19,9 +19,9 @@
 #include "phy.h"
 #include "mac.h"
 
-#define XFP_REQUIRED_DEVS (MDIO_MMDREG_DEVS0_PCS |	\
-			   MDIO_MMDREG_DEVS0_PMAPMD |	\
-			   MDIO_MMDREG_DEVS0_PHYXS)
+#define XFP_REQUIRED_DEVS (MDIO_MMDREG_DEVS_PCS |	\
+			   MDIO_MMDREG_DEVS_PMAPMD |	\
+			   MDIO_MMDREG_DEVS_PHYXS)
 
 #define XFP_LOOPBACKS ((1 << LOOPBACK_PCS) |		\
 		       (1 << LOOPBACK_PMAPMD) |		\
@@ -64,7 +64,7 @@ static int xfp_reset_phy(struct efx_nic *efx)
 	/* Check that all the MMDs we expect are present and responding. We
 	 * expect faults on some if the link is down, but not on the PHY XS */
 	rc = mdio_clause45_check_mmds(efx, XFP_REQUIRED_DEVS,
-				      MDIO_MMDREG_DEVS0_PHYXS);
+				      MDIO_MMDREG_DEVS_PHYXS);
 	if (rc < 0)
 		goto fail;
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 15/33] sfc: Add support for MMDs numbered >15

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:53:40 +0000

> Combine DEVS0 and DEVS1 registers into a 32-bit mask instead of
> reading just DEVS0.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 16/33] sfc: Add phy_type device attribute

[Ben Hutchings]

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 844580f..4fb5ba2 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -1496,6 +1496,14 @@ static struct notifier_block efx_netdev_notifier = {
 	.notifier_call = efx_netdev_event,
 };
 
+static ssize_t
+show_phy_type(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct efx_nic *efx = pci_get_drvdata(to_pci_dev(dev));
+	return sprintf(buf, "%d\n", efx->phy_type);
+}
+static DEVICE_ATTR(phy_type, 0644, show_phy_type, NULL);
+
 static int efx_register_netdev(struct efx_nic *efx)
 {
 	struct net_device *net_dev = efx->net_dev;
@@ -1522,7 +1530,17 @@ static int efx_register_netdev(struct efx_nic *efx)
 	strcpy(efx->name, net_dev->name);
 	efx_set_channel_names(efx);
 
+	rc = device_create_file(&efx->pci_dev->dev, &dev_attr_phy_type);
+	if (rc) {
+		EFX_ERR(efx, "failed to init net dev attributes\n");
+		goto fail_registered;
+	}
+
 	return 0;
+
+fail_registered:
+	unregister_netdev(net_dev);
+	return rc;
 }
 
 static void efx_unregister_netdev(struct efx_nic *efx)
@@ -1542,6 +1560,7 @@ static void efx_unregister_netdev(struct efx_nic *efx)
 
 	if (efx_dev_registered(efx)) {
 		strlcpy(efx->name, pci_name(efx->pci_dev), sizeof(efx->name));
+		device_remove_file(&efx->pci_dev->dev, &dev_attr_phy_type);
 		unregister_netdev(efx->net_dev);
 	}
 }
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 16/33] sfc: Add phy_type device attribute

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:54:02 +0000

> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 17/33] sfc: Clean up board identification

[Ben Hutchings]

Remove kluge for development boards with unspecified board type.

Remove assumption of contiguous board type code assignments.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/boards.c |   68 ++++++++++++---------------------------------
 drivers/net/sfc/boards.h |    7 +---
 2 files changed, 20 insertions(+), 55 deletions(-)

diff --git a/drivers/net/sfc/boards.c b/drivers/net/sfc/boards.c
index edf0262..08fa4e3 100644
--- a/drivers/net/sfc/boards.c
+++ b/drivers/net/sfc/boards.c
@@ -1,6 +1,6 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
- * Copyright 2007 Solarflare Communications Inc.
+ * Copyright 2007-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -231,70 +231,38 @@ static int sfe4002_init(struct efx_nic *efx)
 /* This will get expanded as board-specific details get moved out of the
  * PHY drivers. */
 struct efx_board_data {
+	enum efx_board_type type;
 	const char *ref_model;
 	const char *gen_type;
 	int (*init) (struct efx_nic *nic);
 };
 
-static int dummy_init(struct efx_nic *nic)
-{
-	return 0;
-}
 
 static struct efx_board_data board_data[] = {
-	[EFX_BOARD_INVALID] =
-	{NULL,	    NULL,                  dummy_init},
-	[EFX_BOARD_SFE4001] =
-	{"SFE4001", "10GBASE-T adapter",   sfe4001_init},
-	[EFX_BOARD_SFE4002] =
-	{"SFE4002", "XFP adapter",         sfe4002_init},
+	{ EFX_BOARD_SFE4001, "SFE4001", "10GBASE-T adapter", sfe4001_init },
+	{ EFX_BOARD_SFE4002, "SFE4002", "XFP adapter", sfe4002_init },
 };
 
-int efx_set_board_info(struct efx_nic *efx, u16 revision_info)
+void efx_set_board_info(struct efx_nic *efx, u16 revision_info)
 {
-	int rc = 0;
-	struct efx_board_data *data;
+	struct efx_board_data *data = NULL;
+	int i;
 
-	if (BOARD_TYPE(revision_info) >= EFX_BOARD_MAX) {
-		EFX_ERR(efx, "squashing unknown board type %d\n",
-			BOARD_TYPE(revision_info));
-		revision_info = 0;
-	}
-
-	if (BOARD_TYPE(revision_info) == 0) {
-		efx->board_info.major = 0;
-		efx->board_info.minor = 0;
-		/* For early boards that don't have revision info. there is
-		 * only 1 board for each PHY type, so we can work it out, with
-		 * the exception of the PHY-less boards. */
-		switch (efx->phy_type) {
-		case PHY_TYPE_10XPRESS:
-			efx->board_info.type = EFX_BOARD_SFE4001;
-			break;
-		case PHY_TYPE_XFP:
-			efx->board_info.type = EFX_BOARD_SFE4002;
-			break;
-		default:
-			efx->board_info.type = 0;
-			break;
-		}
-	} else {
-		efx->board_info.type = BOARD_TYPE(revision_info);
-		efx->board_info.major = BOARD_MAJOR(revision_info);
-		efx->board_info.minor = BOARD_MINOR(revision_info);
-	}
+	efx->board_info.type = BOARD_TYPE(revision_info);
+	efx->board_info.major = BOARD_MAJOR(revision_info);
+	efx->board_info.minor = BOARD_MINOR(revision_info);
 
-	data = &board_data[efx->board_info.type];
+	for (i = 0; i < ARRAY_SIZE(board_data); i++)
+		if (board_data[i].type == efx->board_info.type)
+			data = &board_data[i];
 
-	/* Report the board model number or generic type for recognisable
-	 * boards. */
-	if (efx->board_info.type != 0)
+	if (data) {
 		EFX_INFO(efx, "board is %s rev %c%d\n",
 			 (efx->pci_dev->subsystem_vendor == EFX_VENDID_SFC)
 			 ? data->ref_model : data->gen_type,
 			 'A' + efx->board_info.major, efx->board_info.minor);
-
-	efx->board_info.init = data->init;
-
-	return rc;
+		efx->board_info.init = data->init;
+	} else {
+		EFX_ERR(efx, "unknown board type %d\n", efx->board_info.type);
+	}
 }
diff --git a/drivers/net/sfc/boards.h b/drivers/net/sfc/boards.h
index c6e01b6..5e0dde5 100644
--- a/drivers/net/sfc/boards.h
+++ b/drivers/net/sfc/boards.h
@@ -1,6 +1,6 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
- * Copyright 2007 Solarflare Communications Inc.
+ * Copyright 2007-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -12,14 +12,11 @@
 
 /* Board IDs (must fit in 8 bits) */
 enum efx_board_type {
-	EFX_BOARD_INVALID = 0,
 	EFX_BOARD_SFE4001 = 1,   /* SFE4001 (10GBASE-T) */
 	EFX_BOARD_SFE4002 = 2,
-	/* Insert new types before here */
-	EFX_BOARD_MAX
 };
 
-extern int efx_set_board_info(struct efx_nic *efx, u16 revision_info);
+extern void efx_set_board_info(struct efx_nic *efx, u16 revision_info);
 extern int sfe4001_init(struct efx_nic *efx);
 
 #endif
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 17/33] sfc: Clean up board identification

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:54:19 +0000

> Remove kluge for development boards with unspecified board type.
> 
> Remove assumption of contiguous board type code assignments.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 18/33] sfc: Clean up MDIO flag setting

[Ben Hutchings]

We often want to set or clear a flag in an MDIO register, but avoid
writing if no change is required since this can have side-effects.
Encapsulate this in a function, mdio_clause45_set_flag().

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/mdio_10g.c |   87 ++++++++++++++------------------------------
 drivers/net/sfc/mdio_10g.h |    4 ++
 2 files changed, 32 insertions(+), 59 deletions(-)

diff --git a/drivers/net/sfc/mdio_10g.c b/drivers/net/sfc/mdio_10g.c
index 1eed484..8d91131 100644
--- a/drivers/net/sfc/mdio_10g.c
+++ b/drivers/net/sfc/mdio_10g.c
@@ -205,61 +205,24 @@ bool mdio_clause45_links_ok(struct efx_nic *efx, unsigned int mmd_mask)
 
 void mdio_clause45_transmit_disable(struct efx_nic *efx)
 {
-	int phy_id = efx->mii.phy_id;
-	int ctrl1, ctrl2;
-
-	ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
-					   MDIO_MMDREG_TXDIS);
-	if (efx->phy_mode & PHY_MODE_TX_DISABLED)
-		ctrl2 |= (1 << MDIO_MMDREG_TXDIS_GLOBAL_LBN);
-	else
-		ctrl1 &= ~(1 << MDIO_MMDREG_TXDIS_GLOBAL_LBN);
-	if (ctrl1 != ctrl2)
-		mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD,
-				    MDIO_MMDREG_TXDIS, ctrl2);
+	mdio_clause45_set_flag(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
+			       MDIO_MMDREG_TXDIS, MDIO_MMDREG_TXDIS_GLOBAL_LBN,
+			       efx->phy_mode & PHY_MODE_TX_DISABLED);
 }
 
 void mdio_clause45_phy_reconfigure(struct efx_nic *efx)
 {
 	int phy_id = efx->mii.phy_id;
-	int ctrl1, ctrl2;
-
-	/* Handle (with debouncing) PMA/PMD loopback */
-	ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
-					   MDIO_MMDREG_CTRL1);
-
-	if (efx->loopback_mode == LOOPBACK_PMAPMD)
-		ctrl2 |= (1 << MDIO_PMAPMD_CTRL1_LBACK_LBN);
-	else
-		ctrl2 &= ~(1 << MDIO_PMAPMD_CTRL1_LBACK_LBN);
-
-	if (ctrl1 != ctrl2)
-		mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD,
-				    MDIO_MMDREG_CTRL1, ctrl2);
-
-	/* Handle (with debouncing) PCS loopback */
-	ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PCS,
-					   MDIO_MMDREG_CTRL1);
-	if (efx->loopback_mode == LOOPBACK_PCS)
-		ctrl2 |= (1 << MDIO_MMDREG_CTRL1_LBACK_LBN);
-	else
-		ctrl2 &= ~(1 << MDIO_MMDREG_CTRL1_LBACK_LBN);
 
-	if (ctrl1 != ctrl2)
-		mdio_clause45_write(efx, phy_id, MDIO_MMD_PCS,
-				    MDIO_MMDREG_CTRL1, ctrl2);
-
-	/* Handle (with debouncing) PHYXS network loopback */
-	ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PHYXS,
-					   MDIO_MMDREG_CTRL1);
-	if (efx->loopback_mode == LOOPBACK_NETWORK)
-		ctrl2 |= (1 << MDIO_MMDREG_CTRL1_LBACK_LBN);
-	else
-		ctrl2 &= ~(1 << MDIO_MMDREG_CTRL1_LBACK_LBN);
-
-	if (ctrl1 != ctrl2)
-		mdio_clause45_write(efx, phy_id, MDIO_MMD_PHYXS,
-				    MDIO_MMDREG_CTRL1, ctrl2);
+	mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_PMAPMD,
+			       MDIO_MMDREG_CTRL1, MDIO_PMAPMD_CTRL1_LBACK_LBN,
+			       efx->loopback_mode == LOOPBACK_PMAPMD);
+	mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_PCS,
+			       MDIO_MMDREG_CTRL1, MDIO_MMDREG_CTRL1_LBACK_LBN,
+			       efx->loopback_mode == LOOPBACK_PCS);
+	mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_PHYXS,
+			       MDIO_MMDREG_CTRL1, MDIO_MMDREG_CTRL1_LBACK_LBN,
+			       efx->loopback_mode == LOOPBACK_NETWORK);
 }
 
 static void mdio_clause45_set_mmd_lpower(struct efx_nic *efx,
@@ -267,21 +230,13 @@ static void mdio_clause45_set_mmd_lpower(struct efx_nic *efx,
 {
 	int phy = efx->mii.phy_id;
 	int stat = mdio_clause45_read(efx, phy, mmd, MDIO_MMDREG_STAT1);
-	int ctrl1, ctrl2;
 
 	EFX_TRACE(efx, "Setting low power mode for MMD %d to %d\n",
 		  mmd, lpower);
 
 	if (stat & (1 << MDIO_MMDREG_STAT1_LPABLE_LBN)) {
-		ctrl1 = ctrl2 = mdio_clause45_read(efx, phy,
-						   mmd, MDIO_MMDREG_CTRL1);
-		if (lpower)
-			ctrl2 |= (1 << MDIO_MMDREG_CTRL1_LPOWER_LBN);
-		else
-			ctrl2 &= ~(1 << MDIO_MMDREG_CTRL1_LPOWER_LBN);
-		if (ctrl1 != ctrl2)
-			mdio_clause45_write(efx, phy, mmd,
-					    MDIO_MMDREG_CTRL1, ctrl2);
+		mdio_clause45_set_flag(efx, phy, mmd, MDIO_MMDREG_CTRL1,
+				       MDIO_MMDREG_CTRL1_LPOWER_LBN, lpower);
 	}
 }
 
@@ -395,3 +350,17 @@ int mdio_clause45_set_settings(struct efx_nic *efx,
 		return 0;
 	return -EOPNOTSUPP;
 }
+
+void mdio_clause45_set_flag(struct efx_nic *efx, u8 prt, u8 dev,
+			    u16 addr, int bit, bool sense)
+{
+	int old_val = mdio_clause45_read(efx, prt, dev, addr);
+	int new_val;
+
+	if (sense)
+		new_val = old_val | (1 << bit);
+	else
+		new_val = old_val & ~(1 << bit);
+	if (old_val != new_val)
+		mdio_clause45_write(efx, prt, dev, addr, new_val);
+}
diff --git a/drivers/net/sfc/mdio_10g.h b/drivers/net/sfc/mdio_10g.h
index 4516383..4830e0c 100644
--- a/drivers/net/sfc/mdio_10g.h
+++ b/drivers/net/sfc/mdio_10g.h
@@ -259,4 +259,8 @@ extern int mdio_clause45_set_settings(struct efx_nic *efx,
 extern int mdio_clause45_wait_reset_mmds(struct efx_nic *efx,
 					 unsigned int mmd_mask);
 
+/* Set or clear flag, debouncing */
+extern void mdio_clause45_set_flag(struct efx_nic *efx, u8 prt, u8 dev,
+				   u16 addr, int bit, bool sense);
+
 #endif /* EFX_MDIO_10G_H */
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 18/33] sfc: Clean up MDIO flag setting

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:54:34 +0000

> We often want to set or clear a flag in an MDIO register, but avoid
> writing if no change is required since this can have side-effects.
> Encapsulate this in a function, mdio_clause45_set_flag().
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 19/33] sfc: Add support for sub-10G speeds

[Ben Hutchings]

The SFC4000 has a separate MAC for use at sub-10G speeds.  Introduce
an efx_mac_operations structure with implementations for the two MACs.
Switch between the MACs as necessary.

PHY settings are independent of the MAC, so add get_settings() and
set_settings() to efx_phy_operations.  Also add macs field to indicate
which MACs the PHY is connected to.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/Makefile        |    4 +-
 drivers/net/sfc/efx.c           |   62 +++++++---
 drivers/net/sfc/enum.h          |   29 +++--
 drivers/net/sfc/ethtool.c       |   38 +++----
 drivers/net/sfc/falcon.c        |  179 +++++++++++++++++++++++-------
 drivers/net/sfc/falcon.h        |    3 +
 drivers/net/sfc/falcon_gmac.c   |  233 +++++++++++++++++++++++++++++++++++++++
 drivers/net/sfc/falcon_hwdefs.h |  156 ++++++++++++++++++++++++++
 drivers/net/sfc/falcon_xmac.c   |  222 +++++--------------------------------
 drivers/net/sfc/mac.h           |   16 +--
 drivers/net/sfc/net_driver.h    |   31 +++++
 drivers/net/sfc/selftest.c      |   26 ++---
 drivers/net/sfc/selftest.h      |    2 -
 drivers/net/sfc/sfe4001.c       |    2 +-
 drivers/net/sfc/tenxpress.c     |   10 +-
 drivers/net/sfc/xfp_phy.c       |    7 +-
 16 files changed, 699 insertions(+), 321 deletions(-)
 create mode 100644 drivers/net/sfc/falcon_gmac.c

diff --git a/drivers/net/sfc/Makefile b/drivers/net/sfc/Makefile
index e507daa..b89f9be 100644
--- a/drivers/net/sfc/Makefile
+++ b/drivers/net/sfc/Makefile
@@ -1,5 +1,5 @@
-sfc-y			+= efx.o falcon.o tx.o rx.o falcon_xmac.o \
-			   selftest.o ethtool.o xfp_phy.o \
+sfc-y			+= efx.o falcon.o tx.o rx.o falcon_gmac.o \
+			   falcon_xmac.o selftest.o ethtool.o xfp_phy.o \
 			   mdio_10g.o tenxpress.o boards.o sfe4001.o
 sfc-$(CONFIG_SFC_MTD)	+= mtd.o
 
diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 4fb5ba2..0227dfe 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -27,7 +27,6 @@
 #include "efx.h"
 #include "mdio_10g.h"
 #include "falcon.h"
-#include "mac.h"
 
 #define EFX_MAX_MTU (9 * 1024)
 
@@ -575,10 +574,28 @@ void __efx_reconfigure_port(struct efx_nic *efx)
 		netif_addr_unlock_bh(efx->net_dev);
 	}
 
-	falcon_reconfigure_xmac(efx);
+	falcon_deconfigure_mac_wrapper(efx);
+
+	/* Reconfigure the PHY, disabling transmit in mac level loopback. */
+	if (LOOPBACK_INTERNAL(efx))
+		efx->phy_mode |= PHY_MODE_TX_DISABLED;
+	else
+		efx->phy_mode &= ~PHY_MODE_TX_DISABLED;
+	efx->phy_op->reconfigure(efx);
+
+	if (falcon_switch_mac(efx))
+		goto fail;
+
+	efx->mac_op->reconfigure(efx);
 
 	/* Inform kernel of loss/gain of carrier */
 	efx_link_status_changed(efx);
+	return;
+
+fail:
+	EFX_ERR(efx, "failed to reconfigure MAC\n");
+	efx->phy_op->fini(efx);
+	efx->port_initialized = false;
 }
 
 /* Reinitialise the MAC to pick up new PHY settings, even if the port is
@@ -648,18 +665,25 @@ static int efx_init_port(struct efx_nic *efx)
 
 	EFX_LOG(efx, "init port\n");
 
-	/* Initialise the MAC and PHY */
-	rc = falcon_init_xmac(efx);
+	rc = efx->phy_op->init(efx);
 	if (rc)
 		return rc;
+	efx->phy_op->reconfigure(efx);
+
+	mutex_lock(&efx->mac_lock);
+	rc = falcon_switch_mac(efx);
+	mutex_unlock(&efx->mac_lock);
+	if (rc)
+		goto fail;
+	efx->mac_op->reconfigure(efx);
 
 	efx->port_initialized = true;
 	efx->stats_enabled = true;
-
-	/* Reconfigure port to program MAC registers */
-	falcon_reconfigure_xmac(efx);
-
 	return 0;
+
+fail:
+	efx->phy_op->fini(efx);
+	return rc;
 }
 
 /* Allow efx_reconfigure_port() to be scheduled, and close the window
@@ -702,7 +726,7 @@ static void efx_fini_port(struct efx_nic *efx)
 	if (!efx->port_initialized)
 		return;
 
-	falcon_fini_xmac(efx);
+	efx->phy_op->fini(efx);
 	efx->port_initialized = false;
 
 	efx->link_up = false;
@@ -1179,7 +1203,6 @@ static void efx_monitor(struct work_struct *data)
 {
 	struct efx_nic *efx = container_of(data, struct efx_nic,
 					   monitor_work.work);
-	int rc = 0;
 
 	EFX_TRACE(efx, "hardware monitor executing on CPU %d\n",
 		  raw_smp_processor_id());
@@ -1195,7 +1218,7 @@ static void efx_monitor(struct work_struct *data)
 	}
 
 	if (efx->port_enabled)
-		rc = falcon_check_xmac(efx);
+		efx->mac_op->check_hw(efx);
 	mutex_unlock(&efx->mac_lock);
 
 	queue_delayed_work(efx->workqueue, &efx->monitor_work,
@@ -1331,7 +1354,7 @@ static struct net_device_stats *efx_net_stats(struct net_device *net_dev)
 	if (!spin_trylock(&efx->stats_lock))
 		return stats;
 	if (efx->stats_enabled) {
-		falcon_update_stats_xmac(efx);
+		efx->mac_op->update_stats(efx);
 		falcon_update_nic_stats(efx);
 	}
 	spin_unlock(&efx->stats_lock);
@@ -1519,7 +1542,7 @@ static int efx_register_netdev(struct efx_nic *efx)
 	netif_carrier_off(efx->net_dev);
 
 	/* Clear MAC statistics */
-	falcon_update_stats_xmac(efx);
+	efx->mac_op->update_stats(efx);
 	memset(&efx->mac_stats, 0, sizeof(efx->mac_stats));
 
 	rc = register_netdev(net_dev);
@@ -1575,8 +1598,6 @@ static void efx_unregister_netdev(struct efx_nic *efx)
  * before reset.  */
 void efx_reset_down(struct efx_nic *efx, struct ethtool_cmd *ecmd)
 {
-	int rc;
-
 	EFX_ASSERT_RESET_SERIALISED(efx);
 
 	/* The net_dev->get_stats handler is quite slow, and will fail
@@ -1589,9 +1610,7 @@ void efx_reset_down(struct efx_nic *efx, struct ethtool_cmd *ecmd)
 	mutex_lock(&efx->mac_lock);
 	mutex_lock(&efx->spi_lock);
 
-	rc = falcon_xmac_get_settings(efx, ecmd);
-	if (rc)
-		EFX_ERR(efx, "could not back up PHY settings\n");
+	efx->phy_op->get_settings(efx, ecmd);
 
 	efx_fini_channels(efx);
 }
@@ -1616,7 +1635,7 @@ int efx_reset_up(struct efx_nic *efx, struct ethtool_cmd *ecmd, bool ok)
 	if (ok) {
 		efx_init_channels(efx);
 
-		if (falcon_xmac_set_settings(efx, ecmd))
+		if (efx->phy_op->set_settings(efx, ecmd))
 			EFX_ERR(efx, "could not restore PHY settings\n");
 	}
 
@@ -1779,6 +1798,10 @@ int efx_port_dummy_op_int(struct efx_nic *efx)
 void efx_port_dummy_op_void(struct efx_nic *efx) {}
 void efx_port_dummy_op_blink(struct efx_nic *efx, bool blink) {}
 
+static struct efx_mac_operations efx_dummy_mac_operations = {
+	.reconfigure	= efx_port_dummy_op_void,
+};
+
 static struct efx_phy_operations efx_dummy_phy_operations = {
 	.init		 = efx_port_dummy_op_int,
 	.reconfigure	 = efx_port_dummy_op_void,
@@ -1831,6 +1854,7 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
 	spin_lock_init(&efx->netif_stop_lock);
 	spin_lock_init(&efx->stats_lock);
 	mutex_init(&efx->mac_lock);
+	efx->mac_op = &efx_dummy_mac_operations;
 	efx->phy_op = &efx_dummy_phy_operations;
 	efx->mii.dev = net_dev;
 	INIT_WORK(&efx->reconfigure_work, efx_reconfigure_work);
diff --git a/drivers/net/sfc/enum.h b/drivers/net/sfc/enum.h
index 41e758e..0c2e41c 100644
--- a/drivers/net/sfc/enum.h
+++ b/drivers/net/sfc/enum.h
@@ -1,6 +1,6 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
- * Copyright 2007 Solarflare Communications Inc.
+ * Copyright 2007-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -13,22 +13,24 @@
 /**
  * enum efx_loopback_mode - loopback modes
  * @LOOPBACK_NONE: no loopback
- * @LOOPBACK_XGMII: loopback within MAC at XGMII level
- * @LOOPBACK_XGXS: loopback within MAC at XGXS level
- * @LOOPBACK_XAUI: loopback within MAC at XAUI level
- * @LOOPBACK_PHYXS: loopback within PHY at PHYXS level
- * @LOOPBACK_PCS: loopback within PHY at PCS level
- * @LOOPBACK_PMAPMD: loopback within PHY at PMAPMD level
+ * @LOOPBACK_GMAC: loopback within GMAC at unspecified level
+ * @LOOPBACK_XGMII: loopback within XMAC at XGMII level
+ * @LOOPBACK_XGXS: loopback within XMAC at XGXS level
+ * @LOOPBACK_XAUI: loopback within XMAC at XAUI level
+ * @LOOPBACK_GPHY: loopback within 1G PHY at unspecified level
+ * @LOOPBACK_PHYXS: loopback within 10G PHY at PHYXS level
+ * @LOOPBACK_PCS: loopback within 10G PHY at PCS level
+ * @LOOPBACK_PMAPMD: loopback within 10G PHY at PMAPMD level
  * @LOOPBACK_NETWORK: reflecting loopback (even further than furthest!)
  */
 /* Please keep in order and up-to-date w.r.t the following two #defines */
 enum efx_loopback_mode {
 	LOOPBACK_NONE = 0,
-	LOOPBACK_MAC = 1,
+	LOOPBACK_GMAC = 1,
 	LOOPBACK_XGMII = 2,
 	LOOPBACK_XGXS = 3,
 	LOOPBACK_XAUI = 4,
-	LOOPBACK_PHY = 5,
+	LOOPBACK_GPHY = 5,
 	LOOPBACK_PHYXS = 6,
 	LOOPBACK_PCS = 7,
 	LOOPBACK_PMAPMD = 8,
@@ -45,15 +47,16 @@ extern const char *efx_loopback_mode_names[];
 	LOOPBACK_MODE_NAME(efx->loopback_mode)
 
 /* These loopbacks occur within the controller */
-#define LOOPBACKS_10G_INTERNAL ((1 << LOOPBACK_XGMII)| \
-				(1 << LOOPBACK_XGXS) | \
-				(1 << LOOPBACK_XAUI))
+#define LOOPBACKS_INTERNAL ((1 << LOOPBACK_GMAC) |     \
+			    (1 << LOOPBACK_XGMII)|     \
+			    (1 << LOOPBACK_XGXS) |     \
+			    (1 << LOOPBACK_XAUI))
 
 #define LOOPBACK_MASK(_efx)			\
 	(1 << (_efx)->loopback_mode)
 
 #define LOOPBACK_INTERNAL(_efx)				\
-	(!!(LOOPBACKS_10G_INTERNAL & LOOPBACK_MASK(_efx)))
+	(!!(LOOPBACKS_INTERNAL & LOOPBACK_MASK(_efx)))
 
 #define LOOPBACK_OUT_OF(_from, _to, _mask)				\
 	((LOOPBACK_MASK(_from) & (_mask)) && !(LOOPBACK_MASK(_to) & (_mask)))
diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index 43d6d8b..1b33f89 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -17,15 +17,14 @@
 #include "ethtool.h"
 #include "falcon.h"
 #include "spi.h"
-#include "mac.h"
 
 const char *efx_loopback_mode_names[] = {
 	[LOOPBACK_NONE]		= "NONE",
-	[LOOPBACK_MAC]		= "MAC",
+	[LOOPBACK_GMAC]		= "GMAC",
 	[LOOPBACK_XGMII]	= "XGMII",
 	[LOOPBACK_XGXS]		= "XGXS",
 	[LOOPBACK_XAUI] 	= "XAUI",
-	[LOOPBACK_PHY]		= "PHY",
+	[LOOPBACK_GPHY]		= "GPHY",
 	[LOOPBACK_PHYXS]	= "PHYXS",
 	[LOOPBACK_PCS]	 	= "PCS",
 	[LOOPBACK_PMAPMD]	= "PMA/PMD",
@@ -200,13 +199,15 @@ int efx_ethtool_get_settings(struct net_device *net_dev,
 			     struct ethtool_cmd *ecmd)
 {
 	struct efx_nic *efx = netdev_priv(net_dev);
-	int rc;
 
 	mutex_lock(&efx->mac_lock);
-	rc = falcon_xmac_get_settings(efx, ecmd);
+	efx->phy_op->get_settings(efx, ecmd);
 	mutex_unlock(&efx->mac_lock);
 
-	return rc;
+	/* Falcon GMAC does not support 1000Mbps HD */
+	ecmd->supported &= ~SUPPORTED_1000baseT_Half;
+
+	return 0;
 }
 
 /* This must be called with rtnl_lock held. */
@@ -216,8 +217,15 @@ int efx_ethtool_set_settings(struct net_device *net_dev,
 	struct efx_nic *efx = netdev_priv(net_dev);
 	int rc;
 
+	/* Falcon GMAC does not support 1000Mbps HD */
+	if (ecmd->speed == SPEED_1000 && ecmd->duplex != DUPLEX_FULL) {
+		EFX_LOG(efx, "rejecting unsupported 1000Mbps HD"
+			" setting\n");
+		return -EINVAL;
+	}
+
 	mutex_lock(&efx->mac_lock);
-	rc = falcon_xmac_set_settings(efx, ecmd);
+	rc = efx->phy_op->set_settings(efx, ecmd);
 	mutex_unlock(&efx->mac_lock);
 	if (!rc)
 		efx_reconfigure_port(efx);
@@ -362,10 +370,6 @@ static int efx_ethtool_fill_self_tests(struct efx_nic *efx,
 		      EFX_PORT_NAME, "phy", NULL);
 
 	/* Loopback tests */
-	efx_fill_test(n++, strings, data, &tests->loopback_speed,
-		      EFX_PORT_NAME, "loopback.speed", NULL);
-	efx_fill_test(n++, strings, data, &tests->loopback_full_duplex,
-		      EFX_PORT_NAME, "loopback.full_duplex", NULL);
 	for (mode = LOOPBACK_NONE; mode <= LOOPBACK_TEST_MAX; mode++) {
 		if (!(efx->loopback_modes & (1 << mode)))
 			continue;
@@ -671,22 +675,14 @@ static int efx_ethtool_set_pauseparam(struct net_device *net_dev,
 {
 	struct efx_nic *efx = netdev_priv(net_dev);
 	enum efx_fc_type flow_control = efx->flow_control;
-	int rc;
 
 	flow_control &= ~(EFX_FC_RX | EFX_FC_TX | EFX_FC_AUTO);
 	flow_control |= pause->rx_pause ? EFX_FC_RX : 0;
 	flow_control |= pause->tx_pause ? EFX_FC_TX : 0;
 	flow_control |= pause->autoneg ? EFX_FC_AUTO : 0;
 
-	/* Try to push the pause parameters */
-	mutex_lock(&efx->mac_lock);
-	rc = falcon_xmac_set_pause(efx, flow_control);
-	mutex_unlock(&efx->mac_lock);
-
-	if (!rc)
-		efx_reconfigure_port(efx);
-
-	return rc;
+	efx_reconfigure_port(efx);
+	return 0;
 }
 
 static void efx_ethtool_get_pauseparam(struct net_device *net_dev,
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index ac765cb..22419a0 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -1168,6 +1168,19 @@ void falcon_generate_test_event(struct efx_channel *channel, unsigned int magic)
 	falcon_generate_event(channel, &test_event);
 }
 
+void falcon_sim_phy_event(struct efx_nic *efx)
+{
+	efx_qword_t phy_event;
+
+	EFX_POPULATE_QWORD_1(phy_event, EV_CODE, GLOBAL_EV_DECODE);
+	if (EFX_IS10G(efx))
+		EFX_SET_OWORD_FIELD(phy_event, XG_PHY_INTR, 1);
+	else
+		EFX_SET_OWORD_FIELD(phy_event, G_PHY0_INTR, 1);
+
+	falcon_generate_event(&efx->channel[0], &phy_event);
+}
+
 /**************************************************************************
  *
  * Flush handling
@@ -1850,40 +1863,61 @@ int falcon_spi_write(const struct efx_spi_device *spi, loff_t start,
  *
  **************************************************************************
  */
-void falcon_drain_tx_fifo(struct efx_nic *efx)
+
+static int falcon_reset_macs(struct efx_nic *efx)
 {
-	efx_oword_t temp;
+	efx_oword_t reg;
 	int count;
 
-	if ((falcon_rev(efx) < FALCON_REV_B0) ||
-	    (efx->loopback_mode != LOOPBACK_NONE))
-		return;
+	if (falcon_rev(efx) < FALCON_REV_B0) {
+		/* It's not safe to use GLB_CTL_REG to reset the
+		 * macs, so instead use the internal MAC resets
+		 */
+		if (!EFX_IS10G(efx)) {
+			EFX_POPULATE_OWORD_1(reg, GM_SW_RST, 1);
+			falcon_write(efx, &reg, GM_CFG1_REG);
+			udelay(1000);
+
+			EFX_POPULATE_OWORD_1(reg, GM_SW_RST, 0);
+			falcon_write(efx, &reg, GM_CFG1_REG);
+			udelay(1000);
+			return 0;
+		} else {
+			EFX_POPULATE_OWORD_1(reg, XM_CORE_RST, 1);
+			falcon_write(efx, &reg, XM_GLB_CFG_REG);
+
+			for (count = 0; count < 10000; count++) {
+				falcon_read(efx, &reg, XM_GLB_CFG_REG);
+				if (EFX_OWORD_FIELD(reg, XM_CORE_RST) == 0)
+					return 0;
+				udelay(10);
+			}
 
-	falcon_read(efx, &temp, MAC0_CTRL_REG_KER);
-	/* There is no point in draining more than once */
-	if (EFX_OWORD_FIELD(temp, TXFIFO_DRAIN_EN_B0))
-		return;
+			EFX_ERR(efx, "timed out waiting for XMAC core reset\n");
+			return -ETIMEDOUT;
+		}
+	}
 
 	/* MAC stats will fail whilst the TX fifo is draining. Serialise
 	 * the drain sequence with the statistics fetch */
 	spin_lock(&efx->stats_lock);
 
-	EFX_SET_OWORD_FIELD(temp, TXFIFO_DRAIN_EN_B0, 1);
-	falcon_write(efx, &temp, MAC0_CTRL_REG_KER);
+	falcon_read(efx, &reg, MAC0_CTRL_REG_KER);
+	EFX_SET_OWORD_FIELD(reg, TXFIFO_DRAIN_EN_B0, 1);
+	falcon_write(efx, &reg, MAC0_CTRL_REG_KER);
 
-	/* Reset the MAC and EM block. */
-	falcon_read(efx, &temp, GLB_CTL_REG_KER);
-	EFX_SET_OWORD_FIELD(temp, RST_XGTX, 1);
-	EFX_SET_OWORD_FIELD(temp, RST_XGRX, 1);
-	EFX_SET_OWORD_FIELD(temp, RST_EM, 1);
-	falcon_write(efx, &temp, GLB_CTL_REG_KER);
+	falcon_read(efx, &reg, GLB_CTL_REG_KER);
+	EFX_SET_OWORD_FIELD(reg, RST_XGTX, 1);
+	EFX_SET_OWORD_FIELD(reg, RST_XGRX, 1);
+	EFX_SET_OWORD_FIELD(reg, RST_EM, 1);
+	falcon_write(efx, &reg, GLB_CTL_REG_KER);
 
 	count = 0;
 	while (1) {
-		falcon_read(efx, &temp, GLB_CTL_REG_KER);
-		if (!EFX_OWORD_FIELD(temp, RST_XGTX) &&
-		    !EFX_OWORD_FIELD(temp, RST_XGRX) &&
-		    !EFX_OWORD_FIELD(temp, RST_EM)) {
+		falcon_read(efx, &reg, GLB_CTL_REG_KER);
+		if (!EFX_OWORD_FIELD(reg, RST_XGTX) &&
+		    !EFX_OWORD_FIELD(reg, RST_XGRX) &&
+		    !EFX_OWORD_FIELD(reg, RST_EM)) {
 			EFX_LOG(efx, "Completed MAC reset after %d loops\n",
 				count);
 			break;
@@ -1900,21 +1934,39 @@ void falcon_drain_tx_fifo(struct efx_nic *efx)
 
 	/* If we've reset the EM block and the link is up, then
 	 * we'll have to kick the XAUI link so the PHY can recover */
-	if (efx->link_up && EFX_WORKAROUND_5147(efx))
+	if (efx->link_up && EFX_IS10G(efx) && EFX_WORKAROUND_5147(efx))
 		falcon_reset_xaui(efx);
+
+	return 0;
+}
+
+void falcon_drain_tx_fifo(struct efx_nic *efx)
+{
+	efx_oword_t reg;
+
+	if ((falcon_rev(efx) < FALCON_REV_B0) ||
+	    (efx->loopback_mode != LOOPBACK_NONE))
+		return;
+
+	falcon_read(efx, &reg, MAC0_CTRL_REG_KER);
+	/* There is no point in draining more than once */
+	if (EFX_OWORD_FIELD(reg, TXFIFO_DRAIN_EN_B0))
+		return;
+
+	falcon_reset_macs(efx);
 }
 
 void falcon_deconfigure_mac_wrapper(struct efx_nic *efx)
 {
-	efx_oword_t temp;
+	efx_oword_t reg;
 
 	if (falcon_rev(efx) < FALCON_REV_B0)
 		return;
 
 	/* Isolate the MAC -> RX */
-	falcon_read(efx, &temp, RX_CFG_REG_KER);
-	EFX_SET_OWORD_FIELD(temp, RX_INGR_EN_B0, 0);
-	falcon_write(efx, &temp, RX_CFG_REG_KER);
+	falcon_read(efx, &reg, RX_CFG_REG_KER);
+	EFX_SET_OWORD_FIELD(reg, RX_INGR_EN_B0, 0);
+	falcon_write(efx, &reg, RX_CFG_REG_KER);
 
 	if (!efx->link_up)
 		falcon_drain_tx_fifo(efx);
@@ -2041,7 +2093,8 @@ static int falcon_gmii_wait(struct efx_nic *efx)
 	efx_dword_t md_stat;
 	int count;
 
-	for (count = 0; count < 1000; count++) {	/* wait upto 10ms */
+	/* wait upto 50ms - taken max from datasheet */
+	for (count = 0; count < 5000; count++) {
 		falcon_readl(efx, &md_stat, MD_STAT_REG_KER);
 		if (EFX_DWORD_FIELD(md_stat, MD_BSY) == 0) {
 			if (EFX_DWORD_FIELD(md_stat, MD_LNFL) != 0 ||
@@ -2217,10 +2270,59 @@ static int falcon_probe_phy(struct efx_nic *efx)
 		return -1;
 	}
 
-	efx->loopback_modes = LOOPBACKS_10G_INTERNAL | efx->phy_op->loopbacks;
+	if (efx->phy_op->macs & EFX_XMAC)
+		efx->loopback_modes |= ((1 << LOOPBACK_XGMII) |
+					(1 << LOOPBACK_XGXS) |
+					(1 << LOOPBACK_XAUI));
+	if (efx->phy_op->macs & EFX_GMAC)
+		efx->loopback_modes |= (1 << LOOPBACK_GMAC);
+	efx->loopback_modes |= efx->phy_op->loopbacks;
+
 	return 0;
 }
 
+int falcon_switch_mac(struct efx_nic *efx)
+{
+	struct efx_mac_operations *old_mac_op = efx->mac_op;
+	efx_oword_t nic_stat;
+	unsigned strap_val;
+
+	/* Internal loopbacks override the phy speed setting */
+	if (efx->loopback_mode == LOOPBACK_GMAC) {
+		efx->link_speed = 1000;
+		efx->link_fd = true;
+	} else if (LOOPBACK_INTERNAL(efx)) {
+		efx->link_speed = 10000;
+		efx->link_fd = true;
+	}
+
+	efx->mac_op = (EFX_IS10G(efx) ?
+		       &falcon_xmac_operations : &falcon_gmac_operations);
+	if (old_mac_op == efx->mac_op)
+		return 0;
+
+	WARN_ON(!mutex_is_locked(&efx->mac_lock));
+
+	/* Not all macs support a mac-level link state */
+	efx->mac_up = true;
+
+	falcon_read(efx, &nic_stat, NIC_STAT_REG);
+	strap_val = EFX_IS10G(efx) ? 5 : 3;
+	if (falcon_rev(efx) >= FALCON_REV_B0) {
+		EFX_SET_OWORD_FIELD(nic_stat, EE_STRAP_EN, 1);
+		EFX_SET_OWORD_FIELD(nic_stat, EE_STRAP_OVR, strap_val);
+		falcon_write(efx, &nic_stat, NIC_STAT_REG);
+	} else {
+		/* Falcon A1 does not support 1G/10G speed switching
+		 * and must not be used with a PHY that does. */
+		BUG_ON(EFX_OWORD_FIELD(nic_stat, STRAP_PINS) != strap_val);
+	}
+
+
+	EFX_LOG(efx, "selected %cMAC\n", EFX_IS10G(efx) ? 'X' : 'G');
+	return falcon_reset_macs(efx);
+}
+
 /* This call is responsible for hooking in the MAC and PHY operations */
 int falcon_probe_port(struct efx_nic *efx)
 {
@@ -2373,6 +2475,10 @@ static struct {
 	  EFX_OWORD32(0x000003FF, 0x00000000, 0x00000000, 0x00000000) },
 	{ DP_CTRL_REG,
 	  EFX_OWORD32(0x00000FFF, 0x00000000, 0x00000000, 0x00000000) },
+	{ GM_CFG2_REG,
+	  EFX_OWORD32(0x00007337, 0x00000000, 0x00000000, 0x00000000) },
+	{ GMF_CFG0_REG,
+	  EFX_OWORD32(0x00001F1F, 0x00000000, 0x00000000, 0x00000000) },
 	{ XM_GLB_CFG_REG,
 	  EFX_OWORD32(0x00000C68, 0x00000000, 0x00000000, 0x00000000) },
 	{ XM_TX_CFG_REG,
@@ -2698,6 +2804,7 @@ static int falcon_probe_nvconfig(struct efx_nic *efx)
 static int falcon_probe_nic_variant(struct efx_nic *efx)
 {
 	efx_oword_t altera_build;
+	efx_oword_t nic_stat;
 
 	falcon_read(efx, &altera_build, ALTERA_BUILD_REG_KER);
 	if (EFX_OWORD_FIELD(altera_build, VER_ALL)) {
@@ -2705,27 +2812,20 @@ static int falcon_probe_nic_variant(struct efx_nic *efx)
 		return -ENODEV;
 	}
 
+	falcon_read(efx, &nic_stat, NIC_STAT_REG);
+
 	switch (falcon_rev(efx)) {
 	case FALCON_REV_A0:
 	case 0xff:
 		EFX_ERR(efx, "Falcon rev A0 not supported\n");
 		return -ENODEV;
 
-	case FALCON_REV_A1:{
-		efx_oword_t nic_stat;
-
-		falcon_read(efx, &nic_stat, NIC_STAT_REG);
-
+	case FALCON_REV_A1:
 		if (EFX_OWORD_FIELD(nic_stat, STRAP_PCIE) == 0) {
 			EFX_ERR(efx, "Falcon rev A1 PCI-X not supported\n");
 			return -ENODEV;
 		}
-		if (!EFX_OWORD_FIELD(nic_stat, STRAP_10G)) {
-			EFX_ERR(efx, "1G mode not supported\n");
-			return -ENODEV;
-		}
 		break;
-	}
 
 	case FALCON_REV_B0:
 		break;
@@ -2735,6 +2835,9 @@ static int falcon_probe_nic_variant(struct efx_nic *efx)
 		return -ENODEV;
 	}
 
+	/* Initial assumed speed */
+	efx->link_speed = EFX_OWORD_FIELD(nic_stat, STRAP_10G) ? 10000 : 1000;
+
 	return 0;
 }
 
diff --git a/drivers/net/sfc/falcon.h b/drivers/net/sfc/falcon.h
index be025ba..7869c3d 100644
--- a/drivers/net/sfc/falcon.h
+++ b/drivers/net/sfc/falcon.h
@@ -12,6 +12,7 @@
 #define EFX_FALCON_H
 
 #include "net_driver.h"
+#include "efx.h"
 
 /*
  * Falcon hardware control
@@ -65,6 +66,7 @@ extern int falcon_probe_port(struct efx_nic *efx);
 extern void falcon_remove_port(struct efx_nic *efx);
 
 /* MAC/PHY */
+extern int falcon_switch_mac(struct efx_nic *efx);
 extern bool falcon_xaui_link_ok(struct efx_nic *efx);
 extern int falcon_dma_stats(struct efx_nic *efx,
 			    unsigned int done_offset);
@@ -77,6 +79,7 @@ extern int falcon_init_interrupt(struct efx_nic *efx);
 extern void falcon_enable_interrupts(struct efx_nic *efx);
 extern void falcon_generate_test_event(struct efx_channel *channel,
 				       unsigned int magic);
+extern void falcon_sim_phy_event(struct efx_nic *efx);
 extern void falcon_generate_interrupt(struct efx_nic *efx);
 extern void falcon_set_int_moderation(struct efx_channel *channel);
 extern void falcon_disable_interrupts(struct efx_nic *efx);
diff --git a/drivers/net/sfc/falcon_gmac.c b/drivers/net/sfc/falcon_gmac.c
new file mode 100644
index 0000000..247f802
--- /dev/null
+++ b/drivers/net/sfc/falcon_gmac.c
@@ -0,0 +1,233 @@
+/****************************************************************************
+ * Driver for Solarflare Solarstorm network controllers and boards
+ * Copyright 2005-2006 Fen Systems Ltd.
+ * Copyright 2006-2008 Solarflare Communications Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation, incorporated herein by reference.
+ */
+
+#include <linux/delay.h>
+#include "net_driver.h"
+#include "efx.h"
+#include "falcon.h"
+#include "mac.h"
+#include "falcon_hwdefs.h"
+#include "falcon_io.h"
+#include "gmii.h"
+
+/**************************************************************************
+ *
+ * MAC operations
+ *
+ *************************************************************************/
+
+static void falcon_reconfigure_gmac(struct efx_nic *efx)
+{
+	bool loopback, tx_fc, rx_fc, bytemode;
+	int if_mode;
+	unsigned int max_frame_len;
+	efx_oword_t reg;
+
+	/* Configuration register 1 */
+	tx_fc = (efx->flow_control & EFX_FC_TX) || !efx->link_fd;
+	rx_fc = !!(efx->flow_control & EFX_FC_RX);
+	loopback = (efx->loopback_mode == LOOPBACK_GMAC);
+	bytemode = (efx->link_speed == 1000);
+
+	EFX_POPULATE_OWORD_5(reg,
+			     GM_LOOP, loopback,
+			     GM_TX_EN, 1,
+			     GM_TX_FC_EN, tx_fc,
+			     GM_RX_EN, 1,
+			     GM_RX_FC_EN, rx_fc);
+	falcon_write(efx, &reg, GM_CFG1_REG);
+	udelay(10);
+
+	/* Configuration register 2 */
+	if_mode = (bytemode) ? 2 : 1;
+	EFX_POPULATE_OWORD_5(reg,
+			     GM_IF_MODE, if_mode,
+			     GM_PAD_CRC_EN, 1,
+			     GM_LEN_CHK, 1,
+			     GM_FD, efx->link_fd,
+			     GM_PAMBL_LEN, 0x7/*datasheet recommended */);
+
+	falcon_write(efx, &reg, GM_CFG2_REG);
+	udelay(10);
+
+	/* Max frame len register */
+	max_frame_len = EFX_MAX_FRAME_LEN(efx->net_dev->mtu);
+	EFX_POPULATE_OWORD_1(reg, GM_MAX_FLEN, max_frame_len);
+	falcon_write(efx, &reg, GM_MAX_FLEN_REG);
+	udelay(10);
+
+	/* FIFO configuration register 0 */
+	EFX_POPULATE_OWORD_5(reg,
+			     GMF_FTFENREQ, 1,
+			     GMF_STFENREQ, 1,
+			     GMF_FRFENREQ, 1,
+			     GMF_SRFENREQ, 1,
+			     GMF_WTMENREQ, 1);
+	falcon_write(efx, &reg, GMF_CFG0_REG);
+	udelay(10);
+
+	/* FIFO configuration register 1 */
+	EFX_POPULATE_OWORD_2(reg,
+			     GMF_CFGFRTH, 0x12,
+			     GMF_CFGXOFFRTX, 0xffff);
+	falcon_write(efx, &reg, GMF_CFG1_REG);
+	udelay(10);
+
+	/* FIFO configuration register 2 */
+	EFX_POPULATE_OWORD_2(reg,
+			     GMF_CFGHWM, 0x3f,
+			     GMF_CFGLWM, 0xa);
+	falcon_write(efx, &reg, GMF_CFG2_REG);
+	udelay(10);
+
+	/* FIFO configuration register 3 */
+	EFX_POPULATE_OWORD_2(reg,
+			     GMF_CFGHWMFT, 0x1c,
+			     GMF_CFGFTTH, 0x08);
+	falcon_write(efx, &reg, GMF_CFG3_REG);
+	udelay(10);
+
+	/* FIFO configuration register 4 */
+	EFX_POPULATE_OWORD_1(reg, GMF_HSTFLTRFRM_PAUSE, 1);
+	falcon_write(efx, &reg, GMF_CFG4_REG);
+	udelay(10);
+
+	/* FIFO configuration register 5 */
+	falcon_read(efx, &reg, GMF_CFG5_REG);
+	EFX_SET_OWORD_FIELD(reg, GMF_CFGBYTMODE, bytemode);
+	EFX_SET_OWORD_FIELD(reg, GMF_CFGHDPLX, !efx->link_fd);
+	EFX_SET_OWORD_FIELD(reg, GMF_HSTDRPLT64, !efx->link_fd);
+	EFX_SET_OWORD_FIELD(reg, GMF_HSTFLTRFRMDC_PAUSE, 0);
+	falcon_write(efx, &reg, GMF_CFG5_REG);
+	udelay(10);
+
+	/* MAC address */
+	EFX_POPULATE_OWORD_4(reg,
+			     GM_HWADDR_5, efx->net_dev->dev_addr[5],
+			     GM_HWADDR_4, efx->net_dev->dev_addr[4],
+			     GM_HWADDR_3, efx->net_dev->dev_addr[3],
+			     GM_HWADDR_2, efx->net_dev->dev_addr[2]);
+	falcon_write(efx, &reg, GM_ADR1_REG);
+	udelay(10);
+	EFX_POPULATE_OWORD_2(reg,
+			     GM_HWADDR_1, efx->net_dev->dev_addr[1],
+			     GM_HWADDR_0, efx->net_dev->dev_addr[0]);
+	falcon_write(efx, &reg, GM_ADR2_REG);
+	udelay(10);
+
+	falcon_reconfigure_mac_wrapper(efx);
+}
+
+static void falcon_update_stats_gmac(struct efx_nic *efx)
+{
+	struct efx_mac_stats *mac_stats = &efx->mac_stats;
+	unsigned long old_rx_pause, old_tx_pause;
+	unsigned long new_rx_pause, new_tx_pause;
+	int rc;
+
+	rc = falcon_dma_stats(efx, GDmaDone_offset);
+	if (rc)
+		return;
+
+	/* Pause frames are erroneously counted as errors (SFC bug 3269) */
+	old_rx_pause = mac_stats->rx_pause;
+	old_tx_pause = mac_stats->tx_pause;
+
+	/* Update MAC stats from DMAed values */
+	FALCON_STAT(efx, GRxGoodOct, rx_good_bytes);
+	FALCON_STAT(efx, GRxBadOct, rx_bad_bytes);
+	FALCON_STAT(efx, GRxMissPkt, rx_missed);
+	FALCON_STAT(efx, GRxFalseCRS, rx_false_carrier);
+	FALCON_STAT(efx, GRxPausePkt, rx_pause);
+	FALCON_STAT(efx, GRxBadPkt, rx_bad);
+	FALCON_STAT(efx, GRxUcastPkt, rx_unicast);
+	FALCON_STAT(efx, GRxMcastPkt, rx_multicast);
+	FALCON_STAT(efx, GRxBcastPkt, rx_broadcast);
+	FALCON_STAT(efx, GRxGoodLt64Pkt, rx_good_lt64);
+	FALCON_STAT(efx, GRxBadLt64Pkt, rx_bad_lt64);
+	FALCON_STAT(efx, GRx64Pkt, rx_64);
+	FALCON_STAT(efx, GRx65to127Pkt, rx_65_to_127);
+	FALCON_STAT(efx, GRx128to255Pkt, rx_128_to_255);
+	FALCON_STAT(efx, GRx256to511Pkt, rx_256_to_511);
+	FALCON_STAT(efx, GRx512to1023Pkt, rx_512_to_1023);
+	FALCON_STAT(efx, GRx1024to15xxPkt, rx_1024_to_15xx);
+	FALCON_STAT(efx, GRx15xxtoJumboPkt, rx_15xx_to_jumbo);
+	FALCON_STAT(efx, GRxGtJumboPkt, rx_gtjumbo);
+	FALCON_STAT(efx, GRxFcsErr64to15xxPkt, rx_bad_64_to_15xx);
+	FALCON_STAT(efx, GRxFcsErr15xxtoJumboPkt, rx_bad_15xx_to_jumbo);
+	FALCON_STAT(efx, GRxFcsErrGtJumboPkt, rx_bad_gtjumbo);
+	FALCON_STAT(efx, GTxGoodBadOct, tx_bytes);
+	FALCON_STAT(efx, GTxGoodOct, tx_good_bytes);
+	FALCON_STAT(efx, GTxSglColPkt, tx_single_collision);
+	FALCON_STAT(efx, GTxMultColPkt, tx_multiple_collision);
+	FALCON_STAT(efx, GTxExColPkt, tx_excessive_collision);
+	FALCON_STAT(efx, GTxDefPkt, tx_deferred);
+	FALCON_STAT(efx, GTxLateCol, tx_late_collision);
+	FALCON_STAT(efx, GTxExDefPkt, tx_excessive_deferred);
+	FALCON_STAT(efx, GTxPausePkt, tx_pause);
+	FALCON_STAT(efx, GTxBadPkt, tx_bad);
+	FALCON_STAT(efx, GTxUcastPkt, tx_unicast);
+	FALCON_STAT(efx, GTxMcastPkt, tx_multicast);
+	FALCON_STAT(efx, GTxBcastPkt, tx_broadcast);
+	FALCON_STAT(efx, GTxLt64Pkt, tx_lt64);
+	FALCON_STAT(efx, GTx64Pkt, tx_64);
+	FALCON_STAT(efx, GTx65to127Pkt, tx_65_to_127);
+	FALCON_STAT(efx, GTx128to255Pkt, tx_128_to_255);
+	FALCON_STAT(efx, GTx256to511Pkt, tx_256_to_511);
+	FALCON_STAT(efx, GTx512to1023Pkt, tx_512_to_1023);
+	FALCON_STAT(efx, GTx1024to15xxPkt, tx_1024_to_15xx);
+	FALCON_STAT(efx, GTx15xxtoJumboPkt, tx_15xx_to_jumbo);
+	FALCON_STAT(efx, GTxGtJumboPkt, tx_gtjumbo);
+	FALCON_STAT(efx, GTxNonTcpUdpPkt, tx_non_tcpudp);
+	FALCON_STAT(efx, GTxMacSrcErrPkt, tx_mac_src_error);
+	FALCON_STAT(efx, GTxIpSrcErrPkt, tx_ip_src_error);
+
+	/* Pause frames are erroneously counted as errors (SFC bug 3269) */
+	new_rx_pause = mac_stats->rx_pause;
+	new_tx_pause = mac_stats->tx_pause;
+	mac_stats->rx_bad -= (new_rx_pause - old_rx_pause);
+	mac_stats->tx_bad -= (new_tx_pause - old_tx_pause);
+
+	/* Derive stats that the MAC doesn't provide directly */
+	mac_stats->tx_bad_bytes =
+		mac_stats->tx_bytes - mac_stats->tx_good_bytes;
+	mac_stats->tx_packets =
+		mac_stats->tx_lt64 + mac_stats->tx_64 +
+		mac_stats->tx_65_to_127 + mac_stats->tx_128_to_255 +
+		mac_stats->tx_256_to_511 + mac_stats->tx_512_to_1023 +
+		mac_stats->tx_1024_to_15xx + mac_stats->tx_15xx_to_jumbo +
+		mac_stats->tx_gtjumbo;
+	mac_stats->tx_collision =
+		mac_stats->tx_single_collision +
+		mac_stats->tx_multiple_collision +
+		mac_stats->tx_excessive_collision +
+		mac_stats->tx_late_collision;
+	mac_stats->rx_bytes =
+		mac_stats->rx_good_bytes + mac_stats->rx_bad_bytes;
+	mac_stats->rx_packets =
+		mac_stats->rx_good_lt64 + mac_stats->rx_bad_lt64 +
+		mac_stats->rx_64 + mac_stats->rx_65_to_127 +
+		mac_stats->rx_128_to_255 + mac_stats->rx_256_to_511 +
+		mac_stats->rx_512_to_1023 + mac_stats->rx_1024_to_15xx +
+		mac_stats->rx_15xx_to_jumbo + mac_stats->rx_gtjumbo;
+	mac_stats->rx_good = mac_stats->rx_packets - mac_stats->rx_bad;
+	mac_stats->rx_lt64 = mac_stats->rx_good_lt64 + mac_stats->rx_bad_lt64;
+}
+
+static int falcon_check_gmac(struct efx_nic *efx)
+{
+	return efx->phy_op->check_hw(efx);
+}
+
+struct efx_mac_operations falcon_gmac_operations = {
+	.reconfigure	= falcon_reconfigure_gmac,
+	.update_stats	= falcon_update_stats_gmac,
+	.check_hw	= falcon_check_gmac,
+};
diff --git a/drivers/net/sfc/falcon_hwdefs.h b/drivers/net/sfc/falcon_hwdefs.h
index 040e70e..5553df8 100644
--- a/drivers/net/sfc/falcon_hwdefs.h
+++ b/drivers/net/sfc/falcon_hwdefs.h
@@ -111,12 +111,18 @@
 
 /* NIC status register */
 #define NIC_STAT_REG 0x0200
+#define EE_STRAP_EN_LBN 31
+#define EE_STRAP_EN_WIDTH 1
+#define EE_STRAP_OVR_LBN 24
+#define EE_STRAP_OVR_WIDTH 4
 #define ONCHIP_SRAM_LBN 16
 #define ONCHIP_SRAM_WIDTH 1
 #define SF_PRST_LBN 9
 #define SF_PRST_WIDTH 1
 #define EE_PRST_LBN 8
 #define EE_PRST_WIDTH 1
+#define STRAP_PINS_LBN 0
+#define STRAP_PINS_WIDTH 3
 /* These bit definitions are extrapolated from the list of numerical
  * values for STRAP_PINS.
  */
@@ -492,6 +498,107 @@
 #define MAC_MCAST_HASH_REG0_KER 0xca0
 #define MAC_MCAST_HASH_REG1_KER 0xcb0
 
+/* GMAC configuration register 1 */
+#define GM_CFG1_REG 0xe00
+#define GM_SW_RST_LBN 31
+#define GM_SW_RST_WIDTH 1
+#define GM_LOOP_LBN 8
+#define GM_LOOP_WIDTH 1
+#define GM_RX_FC_EN_LBN 5
+#define GM_RX_FC_EN_WIDTH 1
+#define GM_TX_FC_EN_LBN 4
+#define GM_TX_FC_EN_WIDTH 1
+#define GM_RX_EN_LBN 2
+#define GM_RX_EN_WIDTH 1
+#define GM_TX_EN_LBN 0
+#define GM_TX_EN_WIDTH 1
+
+/* GMAC configuration register 2 */
+#define GM_CFG2_REG 0xe10
+#define GM_PAMBL_LEN_LBN 12
+#define GM_PAMBL_LEN_WIDTH 4
+#define GM_IF_MODE_LBN 8
+#define GM_IF_MODE_WIDTH 2
+#define GM_LEN_CHK_LBN 4
+#define GM_LEN_CHK_WIDTH 1
+#define GM_PAD_CRC_EN_LBN 2
+#define GM_PAD_CRC_EN_WIDTH 1
+#define GM_FD_LBN 0
+#define GM_FD_WIDTH 1
+
+/* GMAC maximum frame length register */
+#define GM_MAX_FLEN_REG 0xe40
+#define GM_MAX_FLEN_LBN 0
+#define GM_MAX_FLEN_WIDTH 16
+
+/* GMAC station address register 1 */
+#define GM_ADR1_REG 0xf00
+#define GM_HWADDR_5_LBN 24
+#define GM_HWADDR_5_WIDTH 8
+#define GM_HWADDR_4_LBN 16
+#define GM_HWADDR_4_WIDTH 8
+#define GM_HWADDR_3_LBN 8
+#define GM_HWADDR_3_WIDTH 8
+#define GM_HWADDR_2_LBN 0
+#define GM_HWADDR_2_WIDTH 8
+
+/* GMAC station address register 2 */
+#define GM_ADR2_REG 0xf10
+#define GM_HWADDR_1_LBN 24
+#define GM_HWADDR_1_WIDTH 8
+#define GM_HWADDR_0_LBN 16
+#define GM_HWADDR_0_WIDTH 8
+
+/* GMAC FIFO configuration register 0 */
+#define GMF_CFG0_REG 0xf20
+#define GMF_FTFENREQ_LBN 12
+#define GMF_FTFENREQ_WIDTH 1
+#define GMF_STFENREQ_LBN 11
+#define GMF_STFENREQ_WIDTH 1
+#define GMF_FRFENREQ_LBN 10
+#define GMF_FRFENREQ_WIDTH 1
+#define GMF_SRFENREQ_LBN 9
+#define GMF_SRFENREQ_WIDTH 1
+#define GMF_WTMENREQ_LBN 8
+#define GMF_WTMENREQ_WIDTH 1
+
+/* GMAC FIFO configuration register 1 */
+#define GMF_CFG1_REG 0xf30
+#define GMF_CFGFRTH_LBN 16
+#define GMF_CFGFRTH_WIDTH 5
+#define GMF_CFGXOFFRTX_LBN 0
+#define GMF_CFGXOFFRTX_WIDTH 16
+
+/* GMAC FIFO configuration register 2 */
+#define GMF_CFG2_REG 0xf40
+#define GMF_CFGHWM_LBN 16
+#define GMF_CFGHWM_WIDTH 6
+#define GMF_CFGLWM_LBN 0
+#define GMF_CFGLWM_WIDTH 6
+
+/* GMAC FIFO configuration register 3 */
+#define GMF_CFG3_REG 0xf50
+#define GMF_CFGHWMFT_LBN 16
+#define GMF_CFGHWMFT_WIDTH 6
+#define GMF_CFGFTTH_LBN 0
+#define GMF_CFGFTTH_WIDTH 6
+
+/* GMAC FIFO configuration register 4 */
+#define GMF_CFG4_REG 0xf60
+#define GMF_HSTFLTRFRM_PAUSE_LBN 12
+#define GMF_HSTFLTRFRM_PAUSE_WIDTH 12
+
+/* GMAC FIFO configuration register 5 */
+#define GMF_CFG5_REG 0xf70
+#define GMF_CFGHDPLX_LBN 22
+#define GMF_CFGHDPLX_WIDTH 1
+#define GMF_CFGBYTMODE_LBN 19
+#define GMF_CFGBYTMODE_WIDTH 1
+#define GMF_HSTDRPLT64_LBN 18
+#define GMF_HSTDRPLT64_WIDTH 1
+#define GMF_HSTFLTRFRMDC_PAUSE_LBN 12
+#define GMF_HSTFLTRFRMDC_PAUSE_WIDTH 1
+
 /* XGMAC address register low */
 #define XM_ADR_LO_REG 0x1200
 #define XM_ADR_3_LBN 24
@@ -962,54 +1069,103 @@
  **************************************************************************
  *
  */
+
 #define GRxGoodOct_offset 0x0
+#define GRxGoodOct_WIDTH 48
 #define GRxBadOct_offset 0x8
+#define GRxBadOct_WIDTH 48
 #define GRxMissPkt_offset 0x10
+#define GRxMissPkt_WIDTH 32
 #define GRxFalseCRS_offset 0x14
+#define GRxFalseCRS_WIDTH 32
 #define GRxPausePkt_offset 0x18
+#define GRxPausePkt_WIDTH 32
 #define GRxBadPkt_offset 0x1C
+#define GRxBadPkt_WIDTH 32
 #define GRxUcastPkt_offset 0x20
+#define GRxUcastPkt_WIDTH 32
 #define GRxMcastPkt_offset 0x24
+#define GRxMcastPkt_WIDTH 32
 #define GRxBcastPkt_offset 0x28
+#define GRxBcastPkt_WIDTH 32
 #define GRxGoodLt64Pkt_offset 0x2C
+#define GRxGoodLt64Pkt_WIDTH 32
 #define GRxBadLt64Pkt_offset 0x30
+#define GRxBadLt64Pkt_WIDTH 32
 #define GRx64Pkt_offset 0x34
+#define GRx64Pkt_WIDTH 32
 #define GRx65to127Pkt_offset 0x38
+#define GRx65to127Pkt_WIDTH 32
 #define GRx128to255Pkt_offset 0x3C
+#define GRx128to255Pkt_WIDTH 32
 #define GRx256to511Pkt_offset 0x40
+#define GRx256to511Pkt_WIDTH 32
 #define GRx512to1023Pkt_offset 0x44
+#define GRx512to1023Pkt_WIDTH 32
 #define GRx1024to15xxPkt_offset 0x48
+#define GRx1024to15xxPkt_WIDTH 32
 #define GRx15xxtoJumboPkt_offset 0x4C
+#define GRx15xxtoJumboPkt_WIDTH 32
 #define GRxGtJumboPkt_offset 0x50
+#define GRxGtJumboPkt_WIDTH 32
 #define GRxFcsErr64to15xxPkt_offset 0x54
+#define GRxFcsErr64to15xxPkt_WIDTH 32
 #define GRxFcsErr15xxtoJumboPkt_offset 0x58
+#define GRxFcsErr15xxtoJumboPkt_WIDTH 32
 #define GRxFcsErrGtJumboPkt_offset 0x5C
+#define GRxFcsErrGtJumboPkt_WIDTH 32
 #define GTxGoodBadOct_offset 0x80
+#define GTxGoodBadOct_WIDTH 48
 #define GTxGoodOct_offset 0x88
+#define GTxGoodOct_WIDTH 48
 #define GTxSglColPkt_offset 0x90
+#define GTxSglColPkt_WIDTH 32
 #define GTxMultColPkt_offset 0x94
+#define GTxMultColPkt_WIDTH 32
 #define GTxExColPkt_offset 0x98
+#define GTxExColPkt_WIDTH 32
 #define GTxDefPkt_offset 0x9C
+#define GTxDefPkt_WIDTH 32
 #define GTxLateCol_offset 0xA0
+#define GTxLateCol_WIDTH 32
 #define GTxExDefPkt_offset 0xA4
+#define GTxExDefPkt_WIDTH 32
 #define GTxPausePkt_offset 0xA8
+#define GTxPausePkt_WIDTH 32
 #define GTxBadPkt_offset 0xAC
+#define GTxBadPkt_WIDTH 32
 #define GTxUcastPkt_offset 0xB0
+#define GTxUcastPkt_WIDTH 32
 #define GTxMcastPkt_offset 0xB4
+#define GTxMcastPkt_WIDTH 32
 #define GTxBcastPkt_offset 0xB8
+#define GTxBcastPkt_WIDTH 32
 #define GTxLt64Pkt_offset 0xBC
+#define GTxLt64Pkt_WIDTH 32
 #define GTx64Pkt_offset 0xC0
+#define GTx64Pkt_WIDTH 32
 #define GTx65to127Pkt_offset 0xC4
+#define GTx65to127Pkt_WIDTH 32
 #define GTx128to255Pkt_offset 0xC8
+#define GTx128to255Pkt_WIDTH 32
 #define GTx256to511Pkt_offset 0xCC
+#define GTx256to511Pkt_WIDTH 32
 #define GTx512to1023Pkt_offset 0xD0
+#define GTx512to1023Pkt_WIDTH 32
 #define GTx1024to15xxPkt_offset 0xD4
+#define GTx1024to15xxPkt_WIDTH 32
 #define GTx15xxtoJumboPkt_offset 0xD8
+#define GTx15xxtoJumboPkt_WIDTH 32
 #define GTxGtJumboPkt_offset 0xDC
+#define GTxGtJumboPkt_WIDTH 32
 #define GTxNonTcpUdpPkt_offset 0xE0
+#define GTxNonTcpUdpPkt_WIDTH 16
 #define GTxMacSrcErrPkt_offset 0xE4
+#define GTxMacSrcErrPkt_WIDTH 16
 #define GTxIpSrcErrPkt_offset 0xE8
+#define GTxIpSrcErrPkt_WIDTH 16
 #define GDmaDone_offset 0xEC
+#define GDmaDone_WIDTH 32
 
 #define XgRxOctets_offset 0x0
 #define XgRxOctets_WIDTH 48
diff --git a/drivers/net/sfc/falcon_xmac.c b/drivers/net/sfc/falcon_xmac.c
index 4a54d09..2206ead 100644
--- a/drivers/net/sfc/falcon_xmac.c
+++ b/drivers/net/sfc/falcon_xmac.c
@@ -25,24 +25,6 @@
  * MAC operations
  *
  *************************************************************************/
-static int falcon_reset_xmac(struct efx_nic *efx)
-{
-	efx_oword_t reg;
-	int count;
-
-	EFX_POPULATE_OWORD_1(reg, XM_CORE_RST, 1);
-	falcon_write(efx, &reg, XM_GLB_CFG_REG);
-
-	for (count = 0; count < 10000; count++) {	/* wait upto 100ms */
-		falcon_read(efx, &reg, XM_GLB_CFG_REG);
-		if (EFX_OWORD_FIELD(reg, XM_CORE_RST) == 0)
-			return 0;
-		udelay(10);
-	}
-
-	EFX_ERR(efx, "timed out waiting for XMAC core reset\n");
-	return -ETIMEDOUT;
-}
 
 /* Configure the XAUI driver that is an output from Falcon */
 static void falcon_setup_xaui(struct efx_nic *efx)
@@ -98,31 +80,20 @@ int falcon_reset_xaui(struct efx_nic *efx)
 	return -ETIMEDOUT;
 }
 
-static bool falcon_xgmii_status(struct efx_nic *efx)
+static void falcon_mask_status_intr(struct efx_nic *efx, bool enable)
 {
 	efx_oword_t reg;
 
-	if (falcon_rev(efx) < FALCON_REV_B0)
-		return true;
-
-	/* The ISR latches, so clear it and re-read */
-	falcon_read(efx, &reg, XM_MGT_INT_REG_B0);
-	falcon_read(efx, &reg, XM_MGT_INT_REG_B0);
-
-	if (EFX_OWORD_FIELD(reg, XM_LCLFLT) ||
-	    EFX_OWORD_FIELD(reg, XM_RMTFLT)) {
-		EFX_INFO(efx, "MGT_INT: "EFX_DWORD_FMT"\n", EFX_DWORD_VAL(reg));
-		return false;
-	}
-
-	return true;
-}
+	if ((falcon_rev(efx) != FALCON_REV_B0) || LOOPBACK_INTERNAL(efx))
+		return;
 
-static void falcon_mask_status_intr(struct efx_nic *efx, bool enable)
-{
-	efx_oword_t reg;
+	/* We expect xgmii faults if the wireside link is up */
+	if (!EFX_WORKAROUND_5147(efx) || !efx->link_up)
+		return;
 
-	if ((falcon_rev(efx) < FALCON_REV_B0) || LOOPBACK_INTERNAL(efx))
+	/* We can only use this interrupt to signal the negative edge of
+	 * xaui_align [we have to poll the positive edge]. */
+	if (!efx->mac_up)
 		return;
 
 	/* Flush the ISR */
@@ -135,35 +106,7 @@ static void falcon_mask_status_intr(struct efx_nic *efx, bool enable)
 	falcon_write(efx, &reg, XM_MGT_INT_MSK_REG_B0);
 }
 
-int falcon_init_xmac(struct efx_nic *efx)
-{
-	int rc;
-
-	/* Initialize the PHY first so the clock is around */
-	rc = efx->phy_op->init(efx);
-	if (rc)
-		goto fail1;
-
-	rc = falcon_reset_xaui(efx);
-	if (rc)
-		goto fail2;
-
-	/* Wait again. Give the PHY and MAC time to come back */
-	schedule_timeout_uninterruptible(HZ / 10);
-
-	rc = falcon_reset_xmac(efx);
-	if (rc)
-		goto fail2;
-
-	falcon_mask_status_intr(efx, true);
-	return 0;
-
- fail2:
-	efx->phy_op->fini(efx);
- fail1:
-	return rc;
-}
-
+/* Get status of XAUI link */
 bool falcon_xaui_link_ok(struct efx_nic *efx)
 {
 	efx_oword_t reg;
@@ -187,18 +130,10 @@ bool falcon_xaui_link_ok(struct efx_nic *efx)
 	EFX_SET_OWORD_FIELD(reg, XX_DISPERR, XX_DISPERR_RESET);
 	falcon_write(efx, &reg, XX_CORE_STAT_REG);
 
-	/* If the link is up, then check the phy side of the xaui link
-	 * (error conditions from the wire side propoagate back through
-	 * the phy to the xaui side). */
-	if (efx->link_up && link_ok) {
+	/* If the link is up, then check the phy side of the xaui link */
+	if (efx->link_up && link_ok)
 		if (efx->phy_op->mmds & (1 << MDIO_MMD_PHYXS))
 			link_ok = mdio_clause45_phyxgxs_lane_sync(efx);
-	}
-
-	/* If the PHY and XAUI links are up, then check the mac's xgmii
-	 * fault state */
-	if (efx->link_up && link_ok)
-		link_ok = falcon_xgmii_status(efx);
 
 	return link_ok;
 }
@@ -310,70 +245,39 @@ static void falcon_reconfigure_xgxs_core(struct efx_nic *efx)
 
 /* Try and bring the Falcon side of the Falcon-Phy XAUI link fails
  * to come back up. Bash it until it comes back up */
-static bool falcon_check_xaui_link_up(struct efx_nic *efx)
+static void falcon_check_xaui_link_up(struct efx_nic *efx, int tries)
 {
-	int max_tries, tries;
-	tries = EFX_WORKAROUND_5147(efx) ? 5 : 1;
-	max_tries = tries;
+	efx->mac_up = falcon_xaui_link_ok(efx);
 
 	if ((efx->loopback_mode == LOOPBACK_NETWORK) ||
-	    (efx->phy_type == PHY_TYPE_NONE) ||
 	    efx_phy_mode_disabled(efx->phy_mode))
-		return false;
-
-	while (tries) {
-		if (falcon_xaui_link_ok(efx))
-			return true;
+		/* XAUI link is expected to be down */
+		return;
 
-		EFX_LOG(efx, "%s Clobbering XAUI (%d tries left).\n",
-			__func__, tries);
+	while (!efx->mac_up && tries) {
+		EFX_LOG(efx, "bashing xaui\n");
 		falcon_reset_xaui(efx);
 		udelay(200);
-		tries--;
-	}
 
-	EFX_LOG(efx, "Failed to bring XAUI link back up in %d tries!\n",
-		max_tries);
-	return false;
+		efx->mac_up = falcon_xaui_link_ok(efx);
+		--tries;
+	}
 }
 
-void falcon_reconfigure_xmac(struct efx_nic *efx)
+static void falcon_reconfigure_xmac(struct efx_nic *efx)
 {
-	bool xaui_link_ok;
-
 	falcon_mask_status_intr(efx, false);
 
-	falcon_deconfigure_mac_wrapper(efx);
-
-	/* Reconfigure the PHY, disabling transmit in mac level loopback. */
-	if (LOOPBACK_INTERNAL(efx))
-		efx->phy_mode |= PHY_MODE_TX_DISABLED;
-	else
-		efx->phy_mode &= ~PHY_MODE_TX_DISABLED;
-	efx->phy_op->reconfigure(efx);
-
 	falcon_reconfigure_xgxs_core(efx);
 	falcon_reconfigure_xmac_core(efx);
 
 	falcon_reconfigure_mac_wrapper(efx);
 
-	/* Ensure XAUI link is up */
-	xaui_link_ok = falcon_check_xaui_link_up(efx);
-
-	if (xaui_link_ok && efx->link_up)
-		falcon_mask_status_intr(efx, true);
-}
-
-void falcon_fini_xmac(struct efx_nic *efx)
-{
-	/* Isolate the MAC - PHY */
-	falcon_deconfigure_mac_wrapper(efx);
-
-	/* Potentially power down the PHY */
-	efx->phy_op->fini(efx);
+	falcon_check_xaui_link_up(efx, 5);
+	falcon_mask_status_intr(efx, true);
 }
 
-void falcon_update_stats_xmac(struct efx_nic *efx)
+static void falcon_update_stats_xmac(struct efx_nic *efx)
 {
 	struct efx_mac_stats *mac_stats = &efx->mac_stats;
 	int rc;
@@ -438,7 +342,7 @@ void falcon_update_stats_xmac(struct efx_nic *efx)
 		 mac_stats->rx_control * 64);
 }
 
-int falcon_check_xmac(struct efx_nic *efx)
+static int falcon_check_xmac(struct efx_nic *efx)
 {
 	bool xaui_link_ok;
 	int rc;
@@ -463,72 +367,8 @@ int falcon_check_xmac(struct efx_nic *efx)
 	return rc;
 }
 
-/* Simulate a PHY event */
-void falcon_xmac_sim_phy_event(struct efx_nic *efx)
-{
-	efx_qword_t phy_event;
-
-	EFX_POPULATE_QWORD_2(phy_event,
-			     EV_CODE, GLOBAL_EV_DECODE,
-			     XG_PHY_INTR, 1);
-	falcon_generate_event(&efx->channel[0], &phy_event);
-}
-
-int falcon_xmac_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
-{
-	mdio_clause45_get_settings(efx, ecmd);
-	ecmd->transceiver = XCVR_INTERNAL;
-	ecmd->phy_address = efx->mii.phy_id;
-	ecmd->autoneg = AUTONEG_DISABLE;
-	ecmd->duplex = DUPLEX_FULL;
-	return 0;
-}
-
-int falcon_xmac_set_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
-{
-	if (ecmd->transceiver != XCVR_INTERNAL)
-		return -EINVAL;
-	if (ecmd->autoneg != AUTONEG_DISABLE)
-		return -EINVAL;
-	if (ecmd->duplex != DUPLEX_FULL)
-		return -EINVAL;
-
-	return mdio_clause45_set_settings(efx, ecmd);
-}
-
-
-int falcon_xmac_set_pause(struct efx_nic *efx, enum efx_fc_type flow_control)
-{
-	bool reset;
-
-	if (flow_control & EFX_FC_AUTO) {
-		EFX_LOG(efx, "10G does not support flow control "
-			"autonegotiation\n");
-		return -EINVAL;
-	}
-
-	if ((flow_control & EFX_FC_TX) && !(flow_control & EFX_FC_RX))
-		return -EINVAL;
-
-	/* TX flow control may automatically turn itself off if the
-	 * link partner (intermittently) stops responding to pause
-	 * frames. There isn't any indication that this has happened,
-	 * so the best we do is leave it up to the user to spot this
-	 * and fix it be cycling transmit flow control on this end. */
-	reset = ((flow_control & EFX_FC_TX) &&
-		 !(efx->flow_control & EFX_FC_TX));
-	if (EFX_WORKAROUND_11482(efx) && reset) {
-		if (falcon_rev(efx) >= FALCON_REV_B0) {
-			/* Recover by resetting the EM block */
-			if (efx->link_up)
-				falcon_drain_tx_fifo(efx);
-		} else {
-			/* Schedule a reset to recover */
-			efx_schedule_reset(efx, RESET_TYPE_INVISIBLE);
-		}
-	}
-
-	efx->flow_control = flow_control;
-
-	return 0;
-}
+struct efx_mac_operations falcon_xmac_operations = {
+	.reconfigure	= falcon_reconfigure_xmac,
+	.update_stats	= falcon_update_stats_xmac,
+	.check_hw	= falcon_check_xmac,
+};
diff --git a/drivers/net/sfc/mac.h b/drivers/net/sfc/mac.h
index a31571c..4e70742 100644
--- a/drivers/net/sfc/mac.h
+++ b/drivers/net/sfc/mac.h
@@ -1,7 +1,7 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
  * Copyright 2005-2006 Fen Systems Ltd.
- * Copyright 2006-2007 Solarflare Communications Inc.
+ * Copyright 2006-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -13,17 +13,7 @@
 
 #include "net_driver.h"
 
-extern int falcon_init_xmac(struct efx_nic *efx);
-extern void falcon_reconfigure_xmac(struct efx_nic *efx);
-extern void falcon_update_stats_xmac(struct efx_nic *efx);
-extern void falcon_fini_xmac(struct efx_nic *efx);
-extern int falcon_check_xmac(struct efx_nic *efx);
-extern void falcon_xmac_sim_phy_event(struct efx_nic *efx);
-extern int falcon_xmac_get_settings(struct efx_nic *efx,
-				    struct ethtool_cmd *ecmd);
-extern int falcon_xmac_set_settings(struct efx_nic *efx,
-				    struct ethtool_cmd *ecmd);
-extern int falcon_xmac_set_pause(struct efx_nic *efx,
-				 enum efx_fc_type pause_params);
+extern struct efx_mac_operations falcon_gmac_operations;
+extern struct efx_mac_operations falcon_xmac_operations;
 
 #endif
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index 6cac5ed..883086e 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -463,6 +463,8 @@ enum phy_type {
 
 #define PHY_ADDR_INVALID 0xff
 
+#define EFX_IS10G(efx) ((efx)->link_speed == 10000)
+
 enum nic_state {
 	STATE_INIT = 0,
 	STATE_RUNNING = 1,
@@ -503,6 +505,24 @@ enum efx_fc_type {
 	EFX_FC_AUTO = 4,
 };
 
+/* Supported MAC bit-mask */
+enum efx_mac_type {
+	EFX_GMAC = 1,
+	EFX_XMAC = 2,
+};
+
+/**
+ * struct efx_mac_operations - Efx MAC operations table
+ * @reconfigure: Reconfigure MAC. Serialised by the mac_lock
+ * @update_stats: Update statistics
+ * @check_hw: Check hardware. Serialised by the mac_lock
+ */
+struct efx_mac_operations {
+	void (*reconfigure) (struct efx_nic *efx);
+	void (*update_stats) (struct efx_nic *efx);
+	int (*check_hw) (struct efx_nic *efx);
+};
+
 /**
  * struct efx_phy_operations - Efx PHY operations table
  * @init: Initialise PHY
@@ -511,16 +531,23 @@ enum efx_fc_type {
  * @clear_interrupt: Clear down interrupt
  * @blink: Blink LEDs
  * @check_hw: Check hardware
+ * @get_settings: Get ethtool settings. Serialised by the mac_lock.
+ * @set_settings: Set ethtool settings. Serialised by the mac_lock.
  * @mmds: MMD presence mask
  * @loopbacks: Supported loopback modes mask
  */
 struct efx_phy_operations {
+	enum efx_mac_type macs;
 	int (*init) (struct efx_nic *efx);
 	void (*fini) (struct efx_nic *efx);
 	void (*reconfigure) (struct efx_nic *efx);
 	void (*clear_interrupt) (struct efx_nic *efx);
 	int (*check_hw) (struct efx_nic *efx);
 	int (*test) (struct efx_nic *efx);
+	void (*get_settings) (struct efx_nic *efx,
+			      struct ethtool_cmd *ecmd);
+	int (*set_settings) (struct efx_nic *efx,
+			     struct ethtool_cmd *ecmd);
 	int mmds;
 	unsigned loopbacks;
 };
@@ -686,6 +713,7 @@ union efx_multicast_hash {
  * @stats_lock: Statistics update lock. Serialises statistics fetches
  * @stats_enabled: Temporarily disable statistics fetches.
  *	Serialised by @stats_lock
+ * @mac_op: MAC interface
  * @mac_address: Permanent MAC address
  * @phy_type: PHY type
  * @phy_lock: PHY access lock
@@ -693,6 +721,7 @@ union efx_multicast_hash {
  * @phy_data: PHY private data (including PHY-specific stats)
  * @mii: PHY interface
  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
+ * @mac_up: MAC link state
  * @link_up: Link status
  * @link_fd: Link is full duplex
  * @link_speed: Link speed (Mbps)
@@ -763,6 +792,7 @@ struct efx_nic {
 	spinlock_t stats_lock;
 	bool stats_enabled;
 
+	struct efx_mac_operations *mac_op;
 	unsigned char mac_address[ETH_ALEN];
 
 	enum phy_type phy_type;
@@ -772,6 +802,7 @@ struct efx_nic {
 	struct mii_if_info mii;
 	enum efx_phy_mode phy_mode;
 
+	bool mac_up;
 	bool link_up;
 	bool link_fd;
 	unsigned int link_speed;
diff --git a/drivers/net/sfc/selftest.c b/drivers/net/sfc/selftest.c
index 362956e..8142e37 100644
--- a/drivers/net/sfc/selftest.c
+++ b/drivers/net/sfc/selftest.c
@@ -26,7 +26,6 @@
 #include "selftest.h"
 #include "boards.h"
 #include "workarounds.h"
-#include "mac.h"
 #include "spi.h"
 #include "falcon_io.h"
 #include "mdio_10g.h"
@@ -105,9 +104,11 @@ static int efx_test_mii(struct efx_nic *efx, struct efx_self_tests *tests)
 		goto out;
 	}
 
-	rc = mdio_clause45_check_mmds(efx, efx->phy_op->mmds, 0);
-	if (rc)
-		goto out;
+	if (EFX_IS10G(efx)) {
+		rc = mdio_clause45_check_mmds(efx, efx->phy_op->mmds, 0);
+		if (rc)
+			goto out;
+	}
 
 out:
 	mutex_unlock(&efx->mac_lock);
@@ -598,7 +599,7 @@ static int efx_test_loopbacks(struct efx_nic *efx, struct ethtool_cmd ecmd,
 		do {
 			struct efx_channel *channel = &efx->channel[0];
 
-			falcon_check_xmac(efx);
+			efx->mac_op->check_hw(efx);
 			schedule_timeout_uninterruptible(HZ / 10);
 			if (channel->work_pending)
 				efx_process_channel_now(channel);
@@ -606,13 +607,12 @@ static int efx_test_loopbacks(struct efx_nic *efx, struct ethtool_cmd ecmd,
 			flush_workqueue(efx->workqueue);
 			rmb();
 
-			/* efx->link_up can be 1 even if the XAUI link is down,
-			 * (bug5762). Usually, it's not worth bothering with the
-			 * difference, but for selftests, we need that extra
-			 * guarantee that the link is really, really, up.
-			 */
+			/* We need both the phy and xaui links to be ok.
+			 * rather than relying on the falcon_xmac irq/poll
+			 * regime, just poll xaui directly */
 			link_up = efx->link_up;
-			if (!falcon_xaui_link_ok(efx))
+			if (link_up && EFX_IS10G(efx) &&
+			    !falcon_xaui_link_ok(efx))
 				link_up = false;
 
 		} while ((++count < 20) && !link_up);
@@ -721,7 +721,6 @@ int efx_offline_test(struct efx_nic *efx,
 	if (ecmd_test.autoneg == AUTONEG_ENABLE) {
 		ecmd_test.autoneg = AUTONEG_DISABLE;
 		ecmd_test.duplex = DUPLEX_FULL;
-		ecmd_test.speed = SPEED_10000;
 	}
 	efx->loopback_mode = LOOPBACK_NONE;
 
@@ -732,9 +731,6 @@ int efx_offline_test(struct efx_nic *efx,
 		return rc;
 	}
 
-	tests->loopback_speed = ecmd_test.speed;
-	tests->loopback_full_duplex = ecmd_test.duplex;
-
 	rc = efx_test_phy(efx, tests);
 	if (rc && !rc2)
 		rc2 = rc;
diff --git a/drivers/net/sfc/selftest.h b/drivers/net/sfc/selftest.h
index fc15df1..252f7d7 100644
--- a/drivers/net/sfc/selftest.h
+++ b/drivers/net/sfc/selftest.h
@@ -39,8 +39,6 @@ struct efx_self_tests {
 	/* offline tests */
 	int registers;
 	int phy;
-	int loopback_speed;
-	int loopback_full_duplex;
 	struct efx_loopback_self_tests loopback[LOOPBACK_TEST_MAX + 1];
 };
 
diff --git a/drivers/net/sfc/sfe4001.c b/drivers/net/sfc/sfe4001.c
index aa576c5..af65284 100644
--- a/drivers/net/sfc/sfe4001.c
+++ b/drivers/net/sfc/sfe4001.c
@@ -176,7 +176,7 @@ static int sfe4001_check_hw(struct efx_nic *efx)
 	s32 status;
 
 	/* If XAUI link is up then do not monitor */
-	if (EFX_WORKAROUND_7884(efx) && falcon_xaui_link_ok(efx))
+	if (EFX_WORKAROUND_7884(efx) && efx->mac_up)
 		return 0;
 
 	/* Check the powered status of the PHY. Lack of power implies that
diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c
index 197b544..d60353b 100644
--- a/drivers/net/sfc/tenxpress.c
+++ b/drivers/net/sfc/tenxpress.c
@@ -1,6 +1,6 @@
 /****************************************************************************
- * Driver for Solarflare 802.3an compliant PHY
- * Copyright 2007 Solarflare Communications Inc.
+ * Driver for Solarflare Solarstorm network controllers and boards
+ * Copyright 2007-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -15,7 +15,6 @@
 #include "phy.h"
 #include "falcon_hwdefs.h"
 #include "boards.h"
-#include "mac.h"
 
 /* We expect these MMDs to be in the package */
 /* AN not here as mdio_check_mmds() requires STAT2 support */
@@ -381,7 +380,7 @@ static int tenxpress_phy_check_hw(struct efx_nic *efx)
 	link_ok = tenxpress_link_ok(efx, true);
 
 	if (link_ok != efx->link_up)
-		falcon_xmac_sim_phy_event(efx);
+		falcon_sim_phy_event(efx);
 
 	if (phy_data->phy_mode != PHY_MODE_NORMAL)
 		return 0;
@@ -453,12 +452,15 @@ static int tenxpress_phy_test(struct efx_nic *efx)
 }
 
 struct efx_phy_operations falcon_tenxpress_phy_ops = {
+	.macs		  = EFX_XMAC,
 	.init             = tenxpress_phy_init,
 	.reconfigure      = tenxpress_phy_reconfigure,
 	.check_hw         = tenxpress_phy_check_hw,
 	.fini             = tenxpress_phy_fini,
 	.clear_interrupt  = tenxpress_phy_clear_interrupt,
 	.test             = tenxpress_phy_test,
+	.get_settings	  = mdio_clause45_get_settings,
+	.set_settings	  = mdio_clause45_set_settings,
 	.mmds             = TENXPRESS_REQUIRED_DEVS,
 	.loopbacks        = TENXPRESS_LOOPBACKS,
 };
diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c
index 0413d13..d4e203d 100644
--- a/drivers/net/sfc/xfp_phy.c
+++ b/drivers/net/sfc/xfp_phy.c
@@ -17,7 +17,7 @@
 #include "mdio_10g.h"
 #include "xenpack.h"
 #include "phy.h"
-#include "mac.h"
+#include "falcon.h"
 
 #define XFP_REQUIRED_DEVS (MDIO_MMDREG_DEVS_PCS |	\
 			   MDIO_MMDREG_DEVS_PMAPMD |	\
@@ -125,7 +125,7 @@ static int xfp_phy_check_hw(struct efx_nic *efx)
 	int link_up = xfp_link_ok(efx);
 	/* Simulate a PHY event if link state has changed */
 	if (link_up != efx->link_up)
-		falcon_xmac_sim_phy_event(efx);
+		falcon_sim_phy_event(efx);
 
 	rc = efx->board_info.monitor(efx);
 	if (rc) {
@@ -169,11 +169,14 @@ static void xfp_phy_fini(struct efx_nic *efx)
 }
 
 struct efx_phy_operations falcon_xfp_phy_ops = {
+	.macs		 = EFX_XMAC,
 	.init            = xfp_phy_init,
 	.reconfigure     = xfp_phy_reconfigure,
 	.check_hw        = xfp_phy_check_hw,
 	.fini            = xfp_phy_fini,
 	.clear_interrupt = xfp_phy_clear_interrupt,
+	.get_settings    = mdio_clause45_get_settings,
+	.set_settings	 = mdio_clause45_set_settings,
 	.mmds            = XFP_REQUIRED_DEVS,
 	.loopbacks       = XFP_LOOPBACKS,
 };
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 19/33] sfc: Add support for sub-10G speeds

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:54:51 +0000

> The SFC4000 has a separate MAC for use at sub-10G speeds.  Introduce
> an efx_mac_operations structure with implementations for the two MACs.
> Switch between the MACs as necessary.
> 
> PHY settings are independent of the MAC, so add get_settings() and
> set_settings() to efx_phy_operations.  Also add macs field to indicate
> which MACs the PHY is connected to.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 20/33] sfc: Implement auto-negotiation

[Ben Hutchings]

Add infrastructure for auto-negotiation of speed, duplex and flow
control.

When using 10Xpress, auto-negotiate flow control.  While we're
at it, clean up the code to warn when partner is not 10GBASE-T
capable.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/ethtool.c     |   57 ++++++-
 drivers/net/sfc/falcon.c      |    6 +-
 drivers/net/sfc/falcon_gmac.c |    4 +-
 drivers/net/sfc/falcon_xmac.c |    2 +-
 drivers/net/sfc/mdio_10g.c    |  348 ++++++++++++++++++++++++++++++++---------
 drivers/net/sfc/mdio_10g.h    |   32 ++++
 drivers/net/sfc/net_driver.h  |   38 +++++-
 drivers/net/sfc/tenxpress.c   |  138 +++++++++--------
 drivers/net/sfc/xfp_phy.c     |    1 +
 9 files changed, 472 insertions(+), 154 deletions(-)

diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index 1b33f89..0e81af6 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -12,11 +12,13 @@
 #include <linux/ethtool.h>
 #include <linux/rtnetlink.h>
 #include "net_driver.h"
+#include "workarounds.h"
 #include "selftest.h"
 #include "efx.h"
 #include "ethtool.h"
 #include "falcon.h"
 #include "spi.h"
+#include "mdio_10g.h"
 
 const char *efx_loopback_mode_names[] = {
 	[LOOPBACK_NONE]		= "NONE",
@@ -674,14 +676,51 @@ static int efx_ethtool_set_pauseparam(struct net_device *net_dev,
 				      struct ethtool_pauseparam *pause)
 {
 	struct efx_nic *efx = netdev_priv(net_dev);
-	enum efx_fc_type flow_control = efx->flow_control;
+	enum efx_fc_type wanted_fc;
+	bool reset;
 
-	flow_control &= ~(EFX_FC_RX | EFX_FC_TX | EFX_FC_AUTO);
-	flow_control |= pause->rx_pause ? EFX_FC_RX : 0;
-	flow_control |= pause->tx_pause ? EFX_FC_TX : 0;
-	flow_control |= pause->autoneg ? EFX_FC_AUTO : 0;
+	wanted_fc = ((pause->rx_pause ? EFX_FC_RX : 0) |
+		     (pause->tx_pause ? EFX_FC_TX : 0) |
+		     (pause->autoneg ? EFX_FC_AUTO : 0));
+
+	if ((wanted_fc & EFX_FC_TX) && !(wanted_fc & EFX_FC_RX)) {
+		EFX_LOG(efx, "Flow control unsupported: tx ON rx OFF\n");
+		return -EINVAL;
+	}
+
+	if (!(efx->phy_op->mmds & DEV_PRESENT_BIT(MDIO_MMD_AN)) &&
+	    (wanted_fc & EFX_FC_AUTO)) {
+		EFX_LOG(efx, "PHY does not support flow control "
+			"autonegotiation\n");
+		return -EINVAL;
+	}
+
+	/* TX flow control may automatically turn itself off if the
+	 * link partner (intermittently) stops responding to pause
+	 * frames. There isn't any indication that this has happened,
+	 * so the best we do is leave it up to the user to spot this
+	 * and fix it be cycling transmit flow control on this end. */
+	reset = (wanted_fc & EFX_FC_TX) && !(efx->wanted_fc & EFX_FC_TX);
+	if (EFX_WORKAROUND_11482(efx) && reset) {
+		if (falcon_rev(efx) >= FALCON_REV_B0) {
+			/* Recover by resetting the EM block */
+			if (efx->link_up)
+				falcon_drain_tx_fifo(efx);
+		} else {
+			/* Schedule a reset to recover */
+			efx_schedule_reset(efx, RESET_TYPE_INVISIBLE);
+		}
+	}
+
+	/* Try to push the pause parameters */
+	mutex_lock(&efx->mac_lock);
+
+	efx->wanted_fc = wanted_fc;
+	mdio_clause45_set_pause(efx);
+	__efx_reconfigure_port(efx);
+
+	mutex_unlock(&efx->mac_lock);
 
-	efx_reconfigure_port(efx);
 	return 0;
 }
 
@@ -690,9 +729,9 @@ static void efx_ethtool_get_pauseparam(struct net_device *net_dev,
 {
 	struct efx_nic *efx = netdev_priv(net_dev);
 
-	pause->rx_pause = !!(efx->flow_control & EFX_FC_RX);
-	pause->tx_pause = !!(efx->flow_control & EFX_FC_TX);
-	pause->autoneg = !!(efx->flow_control & EFX_FC_AUTO);
+	pause->rx_pause = !!(efx->wanted_fc & EFX_FC_RX);
+	pause->tx_pause = !!(efx->wanted_fc & EFX_FC_TX);
+	pause->autoneg = !!(efx->wanted_fc & EFX_FC_AUTO);
 }
 
 
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 22419a0..60b3b95 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -2009,7 +2009,7 @@ void falcon_reconfigure_mac_wrapper(struct efx_nic *efx)
 	/* Transmission of pause frames when RX crosses the threshold is
 	 * covered by RX_XOFF_MAC_EN and XM_TX_CFG_REG:XM_FCNTL.
 	 * Action on receipt of pause frames is controller by XM_DIS_FCNTL */
-	tx_fc = !!(efx->flow_control & EFX_FC_TX);
+	tx_fc = !!(efx->link_fc & EFX_FC_TX);
 	falcon_read(efx, &reg, RX_CFG_REG_KER);
 	EFX_SET_OWORD_FIELD_VER(efx, reg, RX_XOFF_MAC_EN, tx_fc);
 
@@ -2339,9 +2339,9 @@ int falcon_probe_port(struct efx_nic *efx)
 
 	/* Hardware flow ctrl. FalconA RX FIFO too small for pause generation */
 	if (falcon_rev(efx) >= FALCON_REV_B0)
-		efx->flow_control = EFX_FC_RX | EFX_FC_TX;
+		efx->wanted_fc = EFX_FC_RX | EFX_FC_TX;
 	else
-		efx->flow_control = EFX_FC_RX;
+		efx->wanted_fc = EFX_FC_RX;
 
 	/* Allocate buffer for stats */
 	rc = falcon_alloc_buffer(efx, &efx->stats_buffer,
diff --git a/drivers/net/sfc/falcon_gmac.c b/drivers/net/sfc/falcon_gmac.c
index 247f802..b6e6eb9 100644
--- a/drivers/net/sfc/falcon_gmac.c
+++ b/drivers/net/sfc/falcon_gmac.c
@@ -31,8 +31,8 @@ static void falcon_reconfigure_gmac(struct efx_nic *efx)
 	efx_oword_t reg;
 
 	/* Configuration register 1 */
-	tx_fc = (efx->flow_control & EFX_FC_TX) || !efx->link_fd;
-	rx_fc = !!(efx->flow_control & EFX_FC_RX);
+	tx_fc = (efx->link_fc & EFX_FC_TX) || !efx->link_fd;
+	rx_fc = !!(efx->link_fc & EFX_FC_RX);
 	loopback = (efx->loopback_mode == LOOPBACK_GMAC);
 	bytemode = (efx->link_speed == 1000);
 
diff --git a/drivers/net/sfc/falcon_xmac.c b/drivers/net/sfc/falcon_xmac.c
index 2206ead..0ce8f01 100644
--- a/drivers/net/sfc/falcon_xmac.c
+++ b/drivers/net/sfc/falcon_xmac.c
@@ -142,7 +142,7 @@ static void falcon_reconfigure_xmac_core(struct efx_nic *efx)
 {
 	unsigned int max_frame_len;
 	efx_oword_t reg;
-	bool rx_fc = !!(efx->flow_control & EFX_FC_RX);
+	bool rx_fc = !!(efx->link_fc & EFX_FC_RX);
 
 	/* Configure MAC  - cut-thru mode is hard wired on */
 	EFX_POPULATE_DWORD_3(reg,
diff --git a/drivers/net/sfc/mdio_10g.c b/drivers/net/sfc/mdio_10g.c
index 8d91131..037601e 100644
--- a/drivers/net/sfc/mdio_10g.c
+++ b/drivers/net/sfc/mdio_10g.c
@@ -47,13 +47,16 @@ static int mdio_clause45_check_mmd(struct efx_nic *efx, int mmd,
 	if (LOOPBACK_INTERNAL(efx))
 		return 0;
 
-	/* Read MMD STATUS2 to check it is responding. */
-	status = mdio_clause45_read(efx, phy_id, mmd, MDIO_MMDREG_STAT2);
-	if (((status >> MDIO_MMDREG_STAT2_PRESENT_LBN) &
-	     ((1 << MDIO_MMDREG_STAT2_PRESENT_WIDTH) - 1)) !=
-	    MDIO_MMDREG_STAT2_PRESENT_VAL) {
-		EFX_ERR(efx, "PHY MMD %d not responding.\n", mmd);
-		return -EIO;
+	if (mmd != MDIO_MMD_AN) {
+		/* Read MMD STATUS2 to check it is responding. */
+		status = mdio_clause45_read(efx, phy_id, mmd,
+					    MDIO_MMDREG_STAT2);
+		if (((status >> MDIO_MMDREG_STAT2_PRESENT_LBN) &
+		     ((1 << MDIO_MMDREG_STAT2_PRESENT_WIDTH) - 1)) !=
+		    MDIO_MMDREG_STAT2_PRESENT_VAL) {
+			EFX_ERR(efx, "PHY MMD %d not responding.\n", mmd);
+			return -EIO;
+		}
 	}
 
 	/* Read MMD STATUS 1 to check for fault. */
@@ -179,12 +182,15 @@ bool mdio_clause45_links_ok(struct efx_nic *efx, unsigned int mmd_mask)
 	else if (efx->loopback_mode == LOOPBACK_PHYXS)
 		mmd_mask &= ~(MDIO_MMDREG_DEVS_PHYXS |
 			      MDIO_MMDREG_DEVS_PCS |
-			      MDIO_MMDREG_DEVS_PMAPMD);
+			      MDIO_MMDREG_DEVS_PMAPMD |
+			      MDIO_MMDREG_DEVS_AN);
 	else if (efx->loopback_mode == LOOPBACK_PCS)
 		mmd_mask &= ~(MDIO_MMDREG_DEVS_PCS |
-			      MDIO_MMDREG_DEVS_PMAPMD);
+			      MDIO_MMDREG_DEVS_PMAPMD |
+			      MDIO_MMDREG_DEVS_AN);
 	else if (efx->loopback_mode == LOOPBACK_PMAPMD)
-		mmd_mask &= ~MDIO_MMDREG_DEVS_PMAPMD;
+		mmd_mask &= ~(MDIO_MMDREG_DEVS_PMAPMD |
+			      MDIO_MMDREG_DEVS_AN);
 
 	while (mmd_mask) {
 		if (mmd_mask & 1) {
@@ -244,6 +250,7 @@ void mdio_clause45_set_mmds_lpower(struct efx_nic *efx,
 				   int low_power, unsigned int mmd_mask)
 {
 	int mmd = 0;
+	mmd_mask &= ~MDIO_MMDREG_DEVS_AN;
 	while (mmd_mask) {
 		if (mmd_mask & 1)
 			mdio_clause45_set_mmd_lpower(efx, low_power, mmd);
@@ -252,103 +259,302 @@ void mdio_clause45_set_mmds_lpower(struct efx_nic *efx,
 	}
 }
 
+static u32 mdio_clause45_get_an(struct efx_nic *efx, u16 addr, u32 xnp)
+{
+	int phy_id = efx->mii.phy_id;
+	u32 result = 0;
+	int reg;
+
+	reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN, addr);
+	if (reg & ADVERTISE_10HALF)
+		result |= ADVERTISED_10baseT_Half;
+	if (reg & ADVERTISE_10FULL)
+		result |= ADVERTISED_10baseT_Full;
+	if (reg & ADVERTISE_100HALF)
+		result |= ADVERTISED_100baseT_Half;
+	if (reg & ADVERTISE_100FULL)
+		result |= ADVERTISED_100baseT_Full;
+	if (reg & LPA_RESV)
+		result |= xnp;
+
+	return result;
+}
+
 /**
  * mdio_clause45_get_settings - Read (some of) the PHY settings over MDIO.
  * @efx:		Efx NIC
  * @ecmd: 		Buffer for settings
  *
  * On return the 'port', 'speed', 'supported' and 'advertising' fields of
- * ecmd have been filled out based on the PMA type.
+ * ecmd have been filled out.
  */
 void mdio_clause45_get_settings(struct efx_nic *efx,
 				struct ethtool_cmd *ecmd)
 {
-	int pma_type;
+	mdio_clause45_get_settings_ext(efx, ecmd, 0, 0);
+}
 
-	/* If no PMA is present we are presumably talking something XAUI-ish
-	 * like CX4. Which we report as FIBRE (see below) */
-	if ((efx->phy_op->mmds & DEV_PRESENT_BIT(MDIO_MMD_PMAPMD)) == 0) {
-		ecmd->speed = SPEED_10000;
-		ecmd->port = PORT_FIBRE;
-		ecmd->supported = SUPPORTED_FIBRE;
-		ecmd->advertising = ADVERTISED_FIBRE;
-		return;
-	}
+/**
+ * mdio_clause45_get_settings_ext - Read (some of) the PHY settings over MDIO.
+ * @efx:		Efx NIC
+ * @ecmd: 		Buffer for settings
+ * @xnp:		Advertised Extended Next Page state
+ * @xnp_lpa:		Link Partner's advertised XNP state
+ *
+ * On return the 'port', 'speed', 'supported' and 'advertising' fields of
+ * ecmd have been filled out.
+ */
+void mdio_clause45_get_settings_ext(struct efx_nic *efx,
+				    struct ethtool_cmd *ecmd,
+				    u32 xnp, u32 xnp_lpa)
+{
+	int phy_id = efx->mii.phy_id;
+	int reg;
 
-	pma_type = mdio_clause45_read(efx, efx->mii.phy_id,
-				      MDIO_MMD_PMAPMD, MDIO_MMDREG_CTRL2);
-	pma_type &= MDIO_PMAPMD_CTRL2_TYPE_MASK;
+	ecmd->transceiver = XCVR_INTERNAL;
+	ecmd->phy_address = phy_id;
 
-	switch (pma_type) {
-		/* We represent CX4 as fibre in the absence of anything
-		   better. */
-	case MDIO_PMAPMD_CTRL2_10G_CX4:
-		ecmd->speed = SPEED_10000;
-		ecmd->port = PORT_FIBRE;
-		ecmd->supported = SUPPORTED_FIBRE;
-		ecmd->advertising = ADVERTISED_FIBRE;
-		break;
-		/* 10G Base-T */
+	reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+				 MDIO_MMDREG_CTRL2);
+	switch (reg & MDIO_PMAPMD_CTRL2_TYPE_MASK) {
 	case MDIO_PMAPMD_CTRL2_10G_BT:
-		ecmd->speed = SPEED_10000;
-		ecmd->port = PORT_TP;
-		ecmd->supported = SUPPORTED_TP | SUPPORTED_10000baseT_Full;
-		ecmd->advertising = (ADVERTISED_FIBRE
-				     | ADVERTISED_10000baseT_Full);
-		break;
 	case MDIO_PMAPMD_CTRL2_1G_BT:
-		ecmd->speed = SPEED_1000;
-		ecmd->port = PORT_TP;
-		ecmd->supported = SUPPORTED_TP | SUPPORTED_1000baseT_Full;
-		ecmd->advertising = (ADVERTISED_FIBRE
-				     | ADVERTISED_1000baseT_Full);
-		break;
 	case MDIO_PMAPMD_CTRL2_100_BT:
-		ecmd->speed = SPEED_100;
-		ecmd->port = PORT_TP;
-		ecmd->supported = SUPPORTED_TP | SUPPORTED_100baseT_Full;
-		ecmd->advertising = (ADVERTISED_FIBRE
-				     | ADVERTISED_100baseT_Full);
-		break;
 	case MDIO_PMAPMD_CTRL2_10_BT:
-		ecmd->speed = SPEED_10;
 		ecmd->port = PORT_TP;
-		ecmd->supported = SUPPORTED_TP | SUPPORTED_10baseT_Full;
-		ecmd->advertising = ADVERTISED_FIBRE | ADVERTISED_10baseT_Full;
+		ecmd->supported = SUPPORTED_TP;
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+					 MDIO_MMDREG_SPEED);
+		if (reg & (1 << MDIO_MMDREG_SPEED_10G_LBN))
+			ecmd->supported |= SUPPORTED_10000baseT_Full;
+		if (reg & (1 << MDIO_MMDREG_SPEED_1000M_LBN))
+			ecmd->supported |= (SUPPORTED_1000baseT_Full |
+					    SUPPORTED_1000baseT_Half);
+		if (reg & (1 << MDIO_MMDREG_SPEED_100M_LBN))
+			ecmd->supported |= (SUPPORTED_100baseT_Full |
+					    SUPPORTED_100baseT_Half);
+		if (reg & (1 << MDIO_MMDREG_SPEED_10M_LBN))
+			ecmd->supported |= (SUPPORTED_10baseT_Full |
+					    SUPPORTED_10baseT_Half);
+		ecmd->advertising = ADVERTISED_TP;
 		break;
-	/* All the other defined modes are flavours of
-	 * 10G optical */
+
+	/* We represent CX4 as fibre in the absence of anything better */
+	case MDIO_PMAPMD_CTRL2_10G_CX4:
+	/* All the other defined modes are flavours of optical */
 	default:
-		ecmd->speed = SPEED_10000;
 		ecmd->port = PORT_FIBRE;
 		ecmd->supported = SUPPORTED_FIBRE;
 		ecmd->advertising = ADVERTISED_FIBRE;
 		break;
 	}
+
+	if (efx->phy_op->mmds & DEV_PRESENT_BIT(MDIO_MMD_AN)) {
+		ecmd->supported |= SUPPORTED_Autoneg;
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN,
+					 MDIO_MMDREG_CTRL1);
+		if (reg & BMCR_ANENABLE) {
+			ecmd->autoneg = AUTONEG_ENABLE;
+			ecmd->advertising |=
+				ADVERTISED_Autoneg |
+				mdio_clause45_get_an(efx,
+						     MDIO_AN_ADVERTISE, xnp);
+		} else
+			ecmd->autoneg = AUTONEG_DISABLE;
+	} else
+		ecmd->autoneg = AUTONEG_DISABLE;
+
+	/* If AN is enabled and complete, report best common mode */
+	if (ecmd->autoneg &&
+	    (mdio_clause45_read(efx, phy_id, MDIO_MMD_AN, MDIO_MMDREG_STAT1) &
+	     (1 << MDIO_AN_STATUS_AN_DONE_LBN))) {
+		u32 common, lpa;
+		lpa = mdio_clause45_get_an(efx, MDIO_AN_LPA, xnp_lpa);
+		common = ecmd->advertising & lpa;
+		if (common & ADVERTISED_10000baseT_Full) {
+			ecmd->speed = SPEED_10000;
+			ecmd->duplex = DUPLEX_FULL;
+		} else if (common & (ADVERTISED_1000baseT_Full |
+				     ADVERTISED_1000baseT_Half)) {
+			ecmd->speed = SPEED_1000;
+			ecmd->duplex = !!(common & ADVERTISED_1000baseT_Full);
+		} else if (common & (ADVERTISED_100baseT_Full |
+				     ADVERTISED_100baseT_Half)) {
+			ecmd->speed = SPEED_100;
+			ecmd->duplex = !!(common & ADVERTISED_100baseT_Full);
+		} else {
+			ecmd->speed = SPEED_10;
+			ecmd->duplex = !!(common & ADVERTISED_10baseT_Full);
+		}
+	} else {
+		/* Report forced settings */
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+					 MDIO_MMDREG_CTRL1);
+		ecmd->speed = (((reg & BMCR_SPEED1000) ? 100 : 1) *
+			       ((reg & BMCR_SPEED100) ? 100 : 10));
+		ecmd->duplex = (reg & BMCR_FULLDPLX ||
+				ecmd->speed == SPEED_10000);
+	}
 }
 
 /**
  * mdio_clause45_set_settings - Set (some of) the PHY settings over MDIO.
  * @efx:		Efx NIC
  * @ecmd: 		New settings
- *
- * Currently this just enforces that we are _not_ changing the
- * 'port', 'speed', 'supported' or 'advertising' settings as these
- * cannot be changed on any currently supported PHY.
  */
 int mdio_clause45_set_settings(struct efx_nic *efx,
 			       struct ethtool_cmd *ecmd)
 {
-	struct ethtool_cmd tmpcmd;
-	mdio_clause45_get_settings(efx, &tmpcmd);
-	/* None of the current PHYs support more than one mode
-	 * of operation (and only 10GBT ever will), so keep things
-	 * simple for now */
-	if ((ecmd->speed == tmpcmd.speed) && (ecmd->port == tmpcmd.port) &&
-	    (ecmd->supported == tmpcmd.supported) &&
-	    (ecmd->advertising == tmpcmd.advertising))
+	int phy_id = efx->mii.phy_id;
+	struct ethtool_cmd prev;
+	u32 required;
+	int ctrl1_bits, reg;
+
+	efx->phy_op->get_settings(efx, &prev);
+
+	if (ecmd->advertising == prev.advertising &&
+	    ecmd->speed == prev.speed &&
+	    ecmd->duplex == prev.duplex &&
+	    ecmd->port == prev.port &&
+	    ecmd->autoneg == prev.autoneg)
 		return 0;
-	return -EOPNOTSUPP;
+
+	/* We can only change these settings for -T PHYs */
+	if (prev.port != PORT_TP || ecmd->port != PORT_TP)
+		return -EINVAL;
+
+	/* Check that PHY supports these settings and work out the
+	 * basic control bits */
+	if (ecmd->duplex) {
+		switch (ecmd->speed) {
+		case SPEED_10:
+			ctrl1_bits = BMCR_FULLDPLX;
+			required = SUPPORTED_10baseT_Full;
+			break;
+		case SPEED_100:
+			ctrl1_bits = BMCR_SPEED100 | BMCR_FULLDPLX;
+			required = SUPPORTED_100baseT_Full;
+			break;
+		case SPEED_1000:
+			ctrl1_bits = BMCR_SPEED1000 | BMCR_FULLDPLX;
+			required = SUPPORTED_1000baseT_Full;
+			break;
+		case SPEED_10000:
+			ctrl1_bits = (BMCR_SPEED1000 | BMCR_SPEED100 |
+				      BMCR_FULLDPLX);
+			required = SUPPORTED_10000baseT_Full;
+			break;
+		default:
+			return -EINVAL;
+		}
+	} else {
+		switch (ecmd->speed) {
+		case SPEED_10:
+			ctrl1_bits = 0;
+			required = SUPPORTED_10baseT_Half;
+			break;
+		case SPEED_100:
+			ctrl1_bits = BMCR_SPEED100;
+			required = SUPPORTED_100baseT_Half;
+			break;
+		case SPEED_1000:
+			ctrl1_bits = BMCR_SPEED1000;
+			required = SUPPORTED_1000baseT_Half;
+			break;
+		default:
+			return -EINVAL;
+		}
+	}
+	if (ecmd->autoneg)
+		required |= SUPPORTED_Autoneg;
+	required |= ecmd->advertising;
+	if (required & ~prev.supported)
+		return -EINVAL;
+
+	/* Set the basic control bits */
+	reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+				 MDIO_MMDREG_CTRL1);
+	reg &= ~(BMCR_SPEED1000 | BMCR_SPEED100 | BMCR_FULLDPLX | 0x003c);
+	reg |= ctrl1_bits;
+	mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD, MDIO_MMDREG_CTRL1,
+			    reg);
+
+	/* Set the AN registers */
+	if (ecmd->autoneg != prev.autoneg ||
+	    ecmd->advertising != prev.advertising) {
+		bool xnp = false;
+
+		if (efx->phy_op->set_xnp_advertise)
+			xnp = efx->phy_op->set_xnp_advertise(efx,
+							     ecmd->advertising);
+
+		if (ecmd->autoneg) {
+			reg = 0;
+			if (ecmd->advertising & ADVERTISED_10baseT_Half)
+				reg |= ADVERTISE_10HALF;
+			if (ecmd->advertising & ADVERTISED_10baseT_Full)
+				reg |= ADVERTISE_10FULL;
+			if (ecmd->advertising & ADVERTISED_100baseT_Half)
+				reg |= ADVERTISE_100HALF;
+			if (ecmd->advertising & ADVERTISED_100baseT_Full)
+				reg |= ADVERTISE_100FULL;
+			if (xnp)
+				reg |= ADVERTISE_RESV;
+			mdio_clause45_write(efx, phy_id, MDIO_MMD_AN,
+					    MDIO_AN_ADVERTISE, reg);
+		}
+
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN,
+					 MDIO_MMDREG_CTRL1);
+		if (ecmd->autoneg)
+			reg |= BMCR_ANENABLE | BMCR_ANRESTART;
+		else
+			reg &= ~BMCR_ANENABLE;
+		if (xnp)
+			reg |= 1 << MDIO_AN_CTRL_XNP_LBN;
+		else
+			reg &= ~(1 << MDIO_AN_CTRL_XNP_LBN);
+		mdio_clause45_write(efx, phy_id, MDIO_MMD_AN,
+				    MDIO_MMDREG_CTRL1, reg);
+	}
+
+	return 0;
+}
+
+void mdio_clause45_set_pause(struct efx_nic *efx)
+{
+	int phy_id = efx->mii.phy_id;
+	int reg;
+
+	if (efx->phy_op->mmds & DEV_PRESENT_BIT(MDIO_MMD_AN)) {
+		/* Set pause capability advertising */
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN,
+					 MDIO_AN_ADVERTISE);
+		reg &= ~(ADVERTISE_PAUSE_CAP | ADVERTISE_PAUSE_ASYM);
+		reg |= efx_fc_advertise(efx->wanted_fc);
+		mdio_clause45_write(efx, phy_id, MDIO_MMD_AN,
+				    MDIO_AN_ADVERTISE, reg);
+
+		/* Restart auto-negotiation */
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN,
+					 MDIO_MMDREG_CTRL1);
+		if (reg & BMCR_ANENABLE) {
+			reg |= BMCR_ANRESTART;
+			mdio_clause45_write(efx, phy_id, MDIO_MMD_AN,
+					    MDIO_MMDREG_CTRL1, reg);
+		}
+	}
+}
+
+enum efx_fc_type mdio_clause45_get_pause(struct efx_nic *efx)
+{
+	int phy_id = efx->mii.phy_id;
+	int lpa;
+
+	if (!(efx->phy_op->mmds & DEV_PRESENT_BIT(MDIO_MMD_AN)))
+		return efx->wanted_fc;
+	lpa = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN, MDIO_AN_LPA);
+	return efx_fc_resolve(efx->wanted_fc, lpa);
 }
 
 void mdio_clause45_set_flag(struct efx_nic *efx, u8 prt, u8 dev,
diff --git a/drivers/net/sfc/mdio_10g.h b/drivers/net/sfc/mdio_10g.h
index 4830e0c..80c63dd 100644
--- a/drivers/net/sfc/mdio_10g.h
+++ b/drivers/net/sfc/mdio_10g.h
@@ -81,6 +81,17 @@
 #define MDIO_MMDREG_DEVS_PHYXS	DEV_PRESENT_BIT(MDIO_MMD_PHYXS)
 #define MDIO_MMDREG_DEVS_PCS	DEV_PRESENT_BIT(MDIO_MMD_PCS)
 #define MDIO_MMDREG_DEVS_PMAPMD	DEV_PRESENT_BIT(MDIO_MMD_PMAPMD)
+#define MDIO_MMDREG_DEVS_AN	DEV_PRESENT_BIT(MDIO_MMD_AN)
+
+/* Bits in MMDREG_SPEED */
+#define MDIO_MMDREG_SPEED_10G_LBN	0
+#define MDIO_MMDREG_SPEED_10G_WIDTH	1
+#define MDIO_MMDREG_SPEED_1000M_LBN	4
+#define MDIO_MMDREG_SPEED_1000M_WIDTH	1
+#define MDIO_MMDREG_SPEED_100M_LBN	5
+#define MDIO_MMDREG_SPEED_100M_WIDTH	1
+#define MDIO_MMDREG_SPEED_10M_LBN	6
+#define MDIO_MMDREG_SPEED_10M_WIDTH	1
 
 /* Bits in MMDREG_STAT2 */
 #define MDIO_MMDREG_STAT2_PRESENT_VAL	(2)
@@ -119,12 +130,20 @@
 #define MDIO_PHYXS_LANE_ALIGNED_LBN	(12)
 
 /* AN registers */
+#define MDIO_AN_CTRL_XNP_LBN		13
 #define MDIO_AN_STATUS			(1)
 #define MDIO_AN_STATUS_XNP_LBN		(7)
 #define MDIO_AN_STATUS_PAGE_LBN		(6)
 #define MDIO_AN_STATUS_AN_DONE_LBN	(5)
 #define MDIO_AN_STATUS_LP_AN_CAP_LBN	(0)
 
+#define MDIO_AN_ADVERTISE		16
+#define MDIO_AN_ADVERTISE_XNP_LBN	12
+#define MDIO_AN_LPA			19
+#define MDIO_AN_XNP			22
+#define MDIO_AN_LPA_XNP			25
+
+#define MDIO_AN_10GBT_ADVERTISE		32
 #define MDIO_AN_10GBT_STATUS		(33)
 #define MDIO_AN_10GBT_STATUS_MS_FLT_LBN (15) /* MASTER/SLAVE config fault */
 #define MDIO_AN_10GBT_STATUS_MS_LBN     (14) /* MASTER/SLAVE config */
@@ -251,10 +270,23 @@ extern void mdio_clause45_set_mmds_lpower(struct efx_nic *efx,
 extern void mdio_clause45_get_settings(struct efx_nic *efx,
 				       struct ethtool_cmd *ecmd);
 
+/* Read (some of) the PHY settings over MDIO */
+extern void
+mdio_clause45_get_settings_ext(struct efx_nic *efx, struct ethtool_cmd *ecmd,
+			       u32 xnp, u32 xnp_lpa);
+
 /* Set (some of) the PHY settings over MDIO */
 extern int mdio_clause45_set_settings(struct efx_nic *efx,
 				      struct ethtool_cmd *ecmd);
 
+/* Set pause parameters to be advertised through AN (if available) */
+extern void mdio_clause45_set_pause(struct efx_nic *efx);
+
+/* Get pause parameters from AN if available (otherwise return
+ * requested pause parameters)
+ */
+enum efx_fc_type mdio_clause45_get_pause(struct efx_nic *efx);
+
 /* Wait for specified MMDs to exit reset within a timeout */
 extern int mdio_clause45_wait_reset_mmds(struct efx_nic *efx,
 					 unsigned int mmd_mask);
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index 883086e..fb8d725 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -511,6 +511,35 @@ enum efx_mac_type {
 	EFX_XMAC = 2,
 };
 
+static inline unsigned int efx_fc_advertise(enum efx_fc_type wanted_fc)
+{
+	unsigned int adv = 0;
+	if (wanted_fc & EFX_FC_RX)
+		adv = ADVERTISE_PAUSE_CAP | ADVERTISE_PAUSE_ASYM;
+	if (wanted_fc & EFX_FC_TX)
+		adv ^= ADVERTISE_PAUSE_ASYM;
+	return adv;
+}
+
+static inline enum efx_fc_type efx_fc_resolve(enum efx_fc_type wanted_fc,
+					      unsigned int lpa)
+{
+	unsigned int adv = efx_fc_advertise(wanted_fc);
+
+	if (!(wanted_fc & EFX_FC_AUTO))
+		return wanted_fc;
+
+	if (adv & lpa & ADVERTISE_PAUSE_CAP)
+		return EFX_FC_RX | EFX_FC_TX;
+	if (adv & lpa & ADVERTISE_PAUSE_ASYM) {
+		if (adv & ADVERTISE_PAUSE_CAP)
+			return EFX_FC_RX;
+		if (lpa & ADVERTISE_PAUSE_CAP)
+			return EFX_FC_TX;
+	}
+	return 0;
+}
+
 /**
  * struct efx_mac_operations - Efx MAC operations table
  * @reconfigure: Reconfigure MAC. Serialised by the mac_lock
@@ -533,6 +562,8 @@ struct efx_mac_operations {
  * @check_hw: Check hardware
  * @get_settings: Get ethtool settings. Serialised by the mac_lock.
  * @set_settings: Set ethtool settings. Serialised by the mac_lock.
+ * @set_xnp_advertise: Set abilities advertised in Extended Next Page
+ *	(only needed where AN bit is set in mmds)
  * @mmds: MMD presence mask
  * @loopbacks: Supported loopback modes mask
  */
@@ -548,6 +579,7 @@ struct efx_phy_operations {
 			      struct ethtool_cmd *ecmd);
 	int (*set_settings) (struct efx_nic *efx,
 			     struct ethtool_cmd *ecmd);
+	bool (*set_xnp_advertise) (struct efx_nic *efx, u32);
 	int mmds;
 	unsigned loopbacks;
 };
@@ -724,11 +756,12 @@ union efx_multicast_hash {
  * @mac_up: MAC link state
  * @link_up: Link status
  * @link_fd: Link is full duplex
+ * @link_fc: Actualy flow control flags
  * @link_speed: Link speed (Mbps)
  * @n_link_state_changes: Number of times the link has changed state
  * @promiscuous: Promiscuous flag. Protected by netif_tx_lock.
  * @multicast_hash: Multicast hash table
- * @flow_control: Flow control flags - separate RX/TX so can't use link_options
+ * @wanted_fc: Wanted flow control flags
  * @reconfigure_work: work item for dealing with PHY events
  * @loopback_mode: Loopback status
  * @loopback_modes: Supported loopback mode bitmask
@@ -805,12 +838,13 @@ struct efx_nic {
 	bool mac_up;
 	bool link_up;
 	bool link_fd;
+	enum efx_fc_type link_fc;
 	unsigned int link_speed;
 	unsigned int n_link_state_changes;
 
 	bool promiscuous;
 	union efx_multicast_hash multicast_hash;
-	enum efx_fc_type flow_control;
+	enum efx_fc_type wanted_fc;
 	struct work_struct reconfigure_work;
 
 	atomic_t rx_reset;
diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c
index d60353b..634ff91 100644
--- a/drivers/net/sfc/tenxpress.c
+++ b/drivers/net/sfc/tenxpress.c
@@ -17,10 +17,10 @@
 #include "boards.h"
 
 /* We expect these MMDs to be in the package */
-/* AN not here as mdio_check_mmds() requires STAT2 support */
 #define TENXPRESS_REQUIRED_DEVS (MDIO_MMDREG_DEVS_PMAPMD	| \
 				 MDIO_MMDREG_DEVS_PCS		| \
-				 MDIO_MMDREG_DEVS_PHYXS)
+				 MDIO_MMDREG_DEVS_PHYXS		| \
+				 MDIO_MMDREG_DEVS_AN)
 
 #define TENXPRESS_LOOPBACKS ((1 << LOOPBACK_PHYXS) |	\
 			     (1 << LOOPBACK_PCS) |	\
@@ -57,6 +57,7 @@
 #define	PMA_PMD_LED_ON		(1)
 #define	PMA_PMD_LED_OFF		(2)
 #define PMA_PMD_LED_FLASH	(3)
+#define PMA_PMD_LED_MASK	3
 /* All LEDs under hardware control */
 #define PMA_PMD_LED_FULL_AUTO	(0)
 /* Green and Amber under hardware control, Red off */
@@ -242,78 +243,60 @@ unlock:
 	return rc;
 }
 
-static void tenxpress_set_bad_lp(struct efx_nic *efx, bool bad_lp)
+static void tenxpress_check_bad_lp(struct efx_nic *efx, bool link_ok)
 {
 	struct tenxpress_phy_data *pd = efx->phy_data;
+	int phy_id = efx->mii.phy_id;
+	bool bad_lp;
 	int reg;
 
+	if (link_ok) {
+		bad_lp = false;
+	} else {
+		/* Check that AN has started but not completed. */
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN,
+					 MDIO_AN_STATUS);
+		if (!(reg & (1 << MDIO_AN_STATUS_LP_AN_CAP_LBN)))
+			return; /* LP status is unknown */
+		bad_lp = !(reg & (1 << MDIO_AN_STATUS_AN_DONE_LBN));
+		if (bad_lp)
+			pd->bad_lp_tries++;
+	}
+
 	/* Nothing to do if all is well and was previously so. */
-	if (!(bad_lp || pd->bad_lp_tries))
+	if (!pd->bad_lp_tries)
 		return;
 
-	reg = mdio_clause45_read(efx, efx->mii.phy_id,
-				 MDIO_MMD_PMAPMD, PMA_PMD_LED_OVERR_REG);
-
-	if (bad_lp)
-		pd->bad_lp_tries++;
-	else
-		pd->bad_lp_tries = 0;
-
-	if (pd->bad_lp_tries == MAX_BAD_LP_TRIES) {
-		pd->bad_lp_tries = 0;	/* Restart count */
-		reg &= ~(PMA_PMD_LED_FLASH << PMA_PMD_LED_RX_LBN);
-		reg |= (PMA_PMD_LED_FLASH << PMA_PMD_LED_RX_LBN);
-		EFX_ERR(efx, "This NIC appears to be plugged into"
-			" a port that is not 10GBASE-T capable.\n"
-			" This PHY is 10GBASE-T ONLY, so no link can"
-			" be established.\n");
-	} else {
-		reg |= (PMA_PMD_LED_OFF << PMA_PMD_LED_RX_LBN);
+	/* Use the RX (red) LED as an error indicator once we've seen AN
+	 * failure several times in a row, and also log a message. */
+	if (!bad_lp || pd->bad_lp_tries == MAX_BAD_LP_TRIES) {
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+					 PMA_PMD_LED_OVERR_REG);
+		reg &= ~(PMA_PMD_LED_MASK << PMA_PMD_LED_RX_LBN);
+		if (!bad_lp) {
+			reg |= PMA_PMD_LED_OFF << PMA_PMD_LED_RX_LBN;
+		} else {
+			reg |= PMA_PMD_LED_FLASH << PMA_PMD_LED_RX_LBN;
+			EFX_ERR(efx, "appears to be plugged into a port"
+				" that is not 10GBASE-T capable. The PHY"
+				" supports 10GBASE-T ONLY, so no link can"
+				" be established\n");
+		}
+		mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD,
+				    PMA_PMD_LED_OVERR_REG, reg);
+		pd->bad_lp_tries = bad_lp;
 	}
-	mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
-			    PMA_PMD_LED_OVERR_REG, reg);
 }
 
-/* Check link status and return a boolean OK value. If the link is NOT
- * OK we have a quick rummage round to see if we appear to be plugged
- * into a non-10GBT port and if so warn the user that they won't get
- * link any time soon as we are 10GBT only, unless caller specified
- * not to do this check (it isn't useful in loopback) */
-static bool tenxpress_link_ok(struct efx_nic *efx, bool check_lp)
+static bool tenxpress_link_ok(struct efx_nic *efx)
 {
-	bool ok = mdio_clause45_links_ok(efx, TENXPRESS_REQUIRED_DEVS);
-
-	if (ok) {
-		tenxpress_set_bad_lp(efx, false);
-	} else if (check_lp) {
-		/* Are we plugged into the wrong sort of link? */
-		bool bad_lp = false;
-		int phy_id = efx->mii.phy_id;
-		int an_stat = mdio_clause45_read(efx, phy_id, MDIO_MMD_AN,
-						 MDIO_AN_STATUS);
-		int xphy_stat = mdio_clause45_read(efx, phy_id,
-						   MDIO_MMD_PMAPMD,
-						   PMA_PMD_XSTATUS_REG);
-		/* Are we plugged into anything that sends FLPs? If
-		 * not we can't distinguish between not being plugged
-		 * in and being plugged into a non-AN antique. The FLP
-		 * bit has the advantage of not clearing when autoneg
-		 * restarts. */
-		if (!(xphy_stat & (1 << PMA_PMD_XSTAT_FLP_LBN))) {
-			tenxpress_set_bad_lp(efx, false);
-			return ok;
-		}
-
-		/* If it can do 10GBT it must be XNP capable */
-		bad_lp = !(an_stat & (1 << MDIO_AN_STATUS_XNP_LBN));
-		if (!bad_lp && (an_stat & (1 << MDIO_AN_STATUS_PAGE_LBN))) {
-			bad_lp = !(mdio_clause45_read(efx, phy_id,
-					MDIO_MMD_AN, MDIO_AN_10GBT_STATUS) &
-					(1 << MDIO_AN_10GBT_STATUS_LP_10G_LBN));
-		}
-		tenxpress_set_bad_lp(efx, bad_lp);
-	}
-	return ok;
+	if (efx->loopback_mode == LOOPBACK_NONE)
+		return mdio_clause45_links_ok(efx, MDIO_MMDREG_DEVS_AN);
+	else
+		return mdio_clause45_links_ok(efx,
+					      MDIO_MMDREG_DEVS_PMAPMD |
+					      MDIO_MMDREG_DEVS_PCS |
+					      MDIO_MMDREG_DEVS_PHYXS);
 }
 
 static void tenxpress_phyxs_loopback(struct efx_nic *efx)
@@ -359,9 +342,10 @@ static void tenxpress_phy_reconfigure(struct efx_nic *efx)
 
 	phy_data->loopback_mode = efx->loopback_mode;
 	phy_data->phy_mode = efx->phy_mode;
-	efx->link_up = tenxpress_link_ok(efx, false);
+	efx->link_up = tenxpress_link_ok(efx);
 	efx->link_speed = 10000;
 	efx->link_fd = true;
+	efx->link_fc = mdio_clause45_get_pause(efx);
 }
 
 static void tenxpress_phy_clear_interrupt(struct efx_nic *efx)
@@ -377,7 +361,8 @@ static int tenxpress_phy_check_hw(struct efx_nic *efx)
 	bool link_ok;
 	int rc = 0;
 
-	link_ok = tenxpress_link_ok(efx, true);
+	link_ok = tenxpress_link_ok(efx);
+	tenxpress_check_bad_lp(efx, link_ok);
 
 	if (link_ok != efx->link_up)
 		falcon_sim_phy_event(efx);
@@ -451,6 +436,27 @@ static int tenxpress_phy_test(struct efx_nic *efx)
 	return tenxpress_special_reset(efx);
 }
 
+static u32 tenxpress_get_xnp_lpa(struct efx_nic *efx)
+{
+	int phy = efx->mii.phy_id;
+	u32 lpa = 0;
+	int reg;
+
+	reg = mdio_clause45_read(efx, phy, MDIO_MMD_AN, MDIO_AN_10GBT_STATUS);
+	if (reg & (1 << MDIO_AN_10GBT_STATUS_LP_10G_LBN))
+		lpa |= ADVERTISED_10000baseT_Full;
+	return lpa;
+}
+
+static void
+tenxpress_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
+{
+	mdio_clause45_get_settings_ext(efx, ecmd, ADVERTISED_10000baseT_Full,
+				       tenxpress_get_xnp_lpa(efx));
+	ecmd->supported |= SUPPORTED_10000baseT_Full;
+	ecmd->advertising |= ADVERTISED_10000baseT_Full;
+}
+
 struct efx_phy_operations falcon_tenxpress_phy_ops = {
 	.macs		  = EFX_XMAC,
 	.init             = tenxpress_phy_init,
@@ -459,7 +465,7 @@ struct efx_phy_operations falcon_tenxpress_phy_ops = {
 	.fini             = tenxpress_phy_fini,
 	.clear_interrupt  = tenxpress_phy_clear_interrupt,
 	.test             = tenxpress_phy_test,
-	.get_settings	  = mdio_clause45_get_settings,
+	.get_settings	  = tenxpress_get_settings,
 	.set_settings	  = mdio_clause45_set_settings,
 	.mmds             = TENXPRESS_REQUIRED_DEVS,
 	.loopbacks        = TENXPRESS_LOOPBACKS,
diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c
index d4e203d..fbe8e25 100644
--- a/drivers/net/sfc/xfp_phy.c
+++ b/drivers/net/sfc/xfp_phy.c
@@ -155,6 +155,7 @@ static void xfp_phy_reconfigure(struct efx_nic *efx)
 	efx->link_up = xfp_link_ok(efx);
 	efx->link_speed = 10000;
 	efx->link_fd = true;
+	efx->link_fc = efx->wanted_fc;
 }
 
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 20/33] sfc: Implement auto-negotiation

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:55:10 +0000

> Add infrastructure for auto-negotiation of speed, duplex and flow
> control.
> 
> When using 10Xpress, auto-negotiate flow control.  While we're
> at it, clean up the code to warn when partner is not 10GBASE-T
> capable.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 21/33] sfc: Rework MAC, PHY and board event handling

[Ben Hutchings]

From: Steve Hodgson <shodgson@solarflare.com>

MAC, PHY and board events may be separately enabled and signalled.
Our current arrangement of chaining the polling functions can result
in events being missed.  Change them to be more independent.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c           |   63 ++++++++++++++++++++++++++-------------
 drivers/net/sfc/falcon.c        |   20 +++++-------
 drivers/net/sfc/falcon_gmac.c   |    8 +----
 drivers/net/sfc/falcon_hwdefs.h |    2 +
 drivers/net/sfc/falcon_xmac.c   |   42 +++++++++++++------------
 drivers/net/sfc/net_driver.h    |   24 ++++++++------
 drivers/net/sfc/selftest.c      |    6 ++-
 drivers/net/sfc/tenxpress.c     |   44 ++++++++------------------
 drivers/net/sfc/xfp_phy.c       |   16 +--------
 9 files changed, 111 insertions(+), 114 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 0227dfe..b99ccfb 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -612,10 +612,9 @@ void efx_reconfigure_port(struct efx_nic *efx)
 /* Asynchronous efx_reconfigure_port work item. To speed up efx_flush_all()
  * we don't efx_reconfigure_port() if the port is disabled. Care is taken
  * in efx_stop_all() and efx_start_port() to prevent PHY events being lost */
-static void efx_reconfigure_work(struct work_struct *data)
+static void efx_phy_work(struct work_struct *data)
 {
-	struct efx_nic *efx = container_of(data, struct efx_nic,
-					   reconfigure_work);
+	struct efx_nic *efx = container_of(data, struct efx_nic, phy_work);
 
 	mutex_lock(&efx->mac_lock);
 	if (efx->port_enabled)
@@ -623,6 +622,16 @@ static void efx_reconfigure_work(struct work_struct *data)
 	mutex_unlock(&efx->mac_lock);
 }
 
+static void efx_mac_work(struct work_struct *data)
+{
+	struct efx_nic *efx = container_of(data, struct efx_nic, mac_work);
+
+	mutex_lock(&efx->mac_lock);
+	if (efx->port_enabled)
+		efx->mac_op->irq(efx);
+	mutex_unlock(&efx->mac_lock);
+}
+
 static int efx_probe_port(struct efx_nic *efx)
 {
 	int rc;
@@ -688,7 +697,7 @@ fail:
 
 /* Allow efx_reconfigure_port() to be scheduled, and close the window
  * between efx_stop_port and efx_flush_all whereby a previously scheduled
- * efx_reconfigure_port() may have been cancelled */
+ * efx_phy_work()/efx_mac_work() may have been cancelled */
 static void efx_start_port(struct efx_nic *efx)
 {
 	EFX_LOG(efx, "start port\n");
@@ -697,13 +706,14 @@ static void efx_start_port(struct efx_nic *efx)
 	mutex_lock(&efx->mac_lock);
 	efx->port_enabled = true;
 	__efx_reconfigure_port(efx);
+	efx->mac_op->irq(efx);
 	mutex_unlock(&efx->mac_lock);
 }
 
-/* Prevent efx_reconfigure_work and efx_monitor() from executing, and
- * efx_set_multicast_list() from scheduling efx_reconfigure_work.
- * efx_reconfigure_work can still be scheduled via NAPI processing
- * until efx_flush_all() is called */
+/* Prevent efx_phy_work, efx_mac_work, and efx_monitor() from executing,
+ * and efx_set_multicast_list() from scheduling efx_phy_work. efx_phy_work
+ * and efx_mac_work may still be scheduled via NAPI processing until
+ * efx_flush_all() is called */
 static void efx_stop_port(struct efx_nic *efx)
 {
 	EFX_LOG(efx, "stop port\n");
@@ -1094,7 +1104,8 @@ static void efx_flush_all(struct efx_nic *efx)
 		cancel_delayed_work_sync(&rx_queue->work);
 
 	/* Stop scheduled port reconfigurations */
-	cancel_work_sync(&efx->reconfigure_work);
+	cancel_work_sync(&efx->mac_work);
+	cancel_work_sync(&efx->phy_work);
 
 }
 
@@ -1131,7 +1142,7 @@ static void efx_stop_all(struct efx_nic *efx)
 	 * window to loose phy events */
 	efx_stop_port(efx);
 
-	/* Flush reconfigure_work, refill_workqueue, monitor_work */
+	/* Flush efx_phy_work, efx_mac_work, refill_workqueue, monitor_work */
 	efx_flush_all(efx);
 
 	/* Isolate the MAC from the TX and RX engines, so that queue
@@ -1203,24 +1214,31 @@ static void efx_monitor(struct work_struct *data)
 {
 	struct efx_nic *efx = container_of(data, struct efx_nic,
 					   monitor_work.work);
+	int rc;
 
 	EFX_TRACE(efx, "hardware monitor executing on CPU %d\n",
 		  raw_smp_processor_id());
 
-
 	/* If the mac_lock is already held then it is likely a port
 	 * reconfiguration is already in place, which will likely do
 	 * most of the work of check_hw() anyway. */
-	if (!mutex_trylock(&efx->mac_lock)) {
-		queue_delayed_work(efx->workqueue, &efx->monitor_work,
-				   efx_monitor_interval);
-		return;
+	if (!mutex_trylock(&efx->mac_lock))
+		goto out_requeue;
+	if (!efx->port_enabled)
+		goto out_unlock;
+	rc = efx->board_info.monitor(efx);
+	if (rc) {
+		EFX_ERR(efx, "Board sensor %s; shutting down PHY\n",
+			(rc == -ERANGE) ? "reported fault" : "failed");
+		efx->phy_mode |= PHY_MODE_LOW_POWER;
+		falcon_sim_phy_event(efx);
 	}
+	efx->phy_op->poll(efx);
+	efx->mac_op->poll(efx);
 
-	if (efx->port_enabled)
-		efx->mac_op->check_hw(efx);
+out_unlock:
 	mutex_unlock(&efx->mac_lock);
-
+out_requeue:
 	queue_delayed_work(efx->workqueue, &efx->monitor_work,
 			   efx_monitor_interval);
 }
@@ -1477,7 +1495,7 @@ static void efx_set_multicast_list(struct net_device *net_dev)
 		return;
 
 	if (changed)
-		queue_work(efx->workqueue, &efx->reconfigure_work);
+		queue_work(efx->workqueue, &efx->phy_work);
 
 	/* Create and activate new global multicast hash table */
 	falcon_set_multicast_hash(efx);
@@ -1800,12 +1818,14 @@ void efx_port_dummy_op_blink(struct efx_nic *efx, bool blink) {}
 
 static struct efx_mac_operations efx_dummy_mac_operations = {
 	.reconfigure	= efx_port_dummy_op_void,
+	.poll		= efx_port_dummy_op_void,
+	.irq		= efx_port_dummy_op_void,
 };
 
 static struct efx_phy_operations efx_dummy_phy_operations = {
 	.init		 = efx_port_dummy_op_int,
 	.reconfigure	 = efx_port_dummy_op_void,
-	.check_hw        = efx_port_dummy_op_int,
+	.poll		 = efx_port_dummy_op_void,
 	.fini		 = efx_port_dummy_op_void,
 	.clear_interrupt = efx_port_dummy_op_void,
 };
@@ -1857,7 +1877,8 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
 	efx->mac_op = &efx_dummy_mac_operations;
 	efx->phy_op = &efx_dummy_phy_operations;
 	efx->mii.dev = net_dev;
-	INIT_WORK(&efx->reconfigure_work, efx_reconfigure_work);
+	INIT_WORK(&efx->phy_work, efx_phy_work);
+	INIT_WORK(&efx->mac_work, efx_mac_work);
 	atomic_set(&efx->netif_stop_count, 1);
 
 	for (i = 0; i < EFX_MAX_CHANNELS; i++) {
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 60b3b95..734add1 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -910,22 +910,20 @@ static void falcon_handle_global_event(struct efx_channel *channel,
 				       efx_qword_t *event)
 {
 	struct efx_nic *efx = channel->efx;
-	bool is_phy_event = false, handled = false;
+	bool handled = false;
 
-	/* Check for interrupt on either port.  Some boards have a
-	 * single PHY wired to the interrupt line for port 1. */
 	if (EFX_QWORD_FIELD(*event, G_PHY0_INTR) ||
 	    EFX_QWORD_FIELD(*event, G_PHY1_INTR) ||
-	    EFX_QWORD_FIELD(*event, XG_PHY_INTR))
-		is_phy_event = true;
+	    EFX_QWORD_FIELD(*event, XG_PHY_INTR) ||
+	    EFX_QWORD_FIELD(*event, XFP_PHY_INTR)) {
+		efx->phy_op->clear_interrupt(efx);
+		queue_work(efx->workqueue, &efx->phy_work);
+		handled = true;
+	}
 
 	if ((falcon_rev(efx) >= FALCON_REV_B0) &&
-	    EFX_QWORD_FIELD(*event, XG_MNT_INTR_B0))
-		is_phy_event = true;
-
-	if (is_phy_event) {
-		efx->phy_op->clear_interrupt(efx);
-		queue_work(efx->workqueue, &efx->reconfigure_work);
+	    EFX_QWORD_FIELD(*event, XG_MNT_INTR_B0)) {
+		queue_work(efx->workqueue, &efx->mac_work);
 		handled = true;
 	}
 
diff --git a/drivers/net/sfc/falcon_gmac.c b/drivers/net/sfc/falcon_gmac.c
index b6e6eb9..8865eae 100644
--- a/drivers/net/sfc/falcon_gmac.c
+++ b/drivers/net/sfc/falcon_gmac.c
@@ -221,13 +221,9 @@ static void falcon_update_stats_gmac(struct efx_nic *efx)
 	mac_stats->rx_lt64 = mac_stats->rx_good_lt64 + mac_stats->rx_bad_lt64;
 }
 
-static int falcon_check_gmac(struct efx_nic *efx)
-{
-	return efx->phy_op->check_hw(efx);
-}
-
 struct efx_mac_operations falcon_gmac_operations = {
 	.reconfigure	= falcon_reconfigure_gmac,
 	.update_stats	= falcon_update_stats_gmac,
-	.check_hw	= falcon_check_gmac,
+	.irq		= efx_port_dummy_op_void,
+	.poll		= efx_port_dummy_op_void,
 };
diff --git a/drivers/net/sfc/falcon_hwdefs.h b/drivers/net/sfc/falcon_hwdefs.h
index 5553df8..a58c627 100644
--- a/drivers/net/sfc/falcon_hwdefs.h
+++ b/drivers/net/sfc/falcon_hwdefs.h
@@ -1051,6 +1051,8 @@
 #define XG_MNT_INTR_B0_WIDTH 1
 #define RX_RECOVERY_A1_LBN 11
 #define RX_RECOVERY_A1_WIDTH 1
+#define XFP_PHY_INTR_LBN 10
+#define XFP_PHY_INTR_WIDTH 1
 #define XG_PHY_INTR_LBN 9
 #define XG_PHY_INTR_WIDTH 1
 #define G_PHY1_INTR_LBN 8
diff --git a/drivers/net/sfc/falcon_xmac.c b/drivers/net/sfc/falcon_xmac.c
index 0ce8f01..5a03713 100644
--- a/drivers/net/sfc/falcon_xmac.c
+++ b/drivers/net/sfc/falcon_xmac.c
@@ -342,33 +342,35 @@ static void falcon_update_stats_xmac(struct efx_nic *efx)
 		 mac_stats->rx_control * 64);
 }
 
-static int falcon_check_xmac(struct efx_nic *efx)
+static void falcon_xmac_irq(struct efx_nic *efx)
 {
-	bool xaui_link_ok;
-	int rc;
+	/* The XGMII link has a transient fault, which indicates either:
+	 *   - there's a transient xgmii fault
+	 *   - falcon's end of the xaui link may need a kick
+	 *   - the wire-side link may have gone down, but the lasi/poll()
+	 *     hasn't noticed yet.
+	 *
+	 * We only want to even bother polling XAUI if we're confident it's
+	 * not (1) or (3). In both cases, the only reliable way to spot this
+	 * is to wait a bit. We do this here by forcing the mac link state
+	 * to down, and waiting for the mac poll to come round and check
+	 */
+	efx->mac_up = false;
+}
 
-	if ((efx->loopback_mode == LOOPBACK_NETWORK) ||
-	    efx_phy_mode_disabled(efx->phy_mode))
-		return 0;
+static void falcon_poll_xmac(struct efx_nic *efx)
+{
+	if (!EFX_WORKAROUND_5147(efx) || !efx->link_up || efx->mac_up)
+		return;
 
 	falcon_mask_status_intr(efx, false);
-	xaui_link_ok = falcon_xaui_link_ok(efx);
-
-	if (EFX_WORKAROUND_5147(efx) && !xaui_link_ok)
-		falcon_reset_xaui(efx);
-
-	/* Call the PHY check_hw routine */
-	rc = efx->phy_op->check_hw(efx);
-
-	/* Unmask interrupt if everything was (and still is) ok */
-	if (xaui_link_ok && efx->link_up)
-		falcon_mask_status_intr(efx, true);
-
-	return rc;
+	falcon_check_xaui_link_up(efx, 1);
+	falcon_mask_status_intr(efx, true);
 }
 
 struct efx_mac_operations falcon_xmac_operations = {
 	.reconfigure	= falcon_reconfigure_xmac,
 	.update_stats	= falcon_update_stats_xmac,
-	.check_hw	= falcon_check_xmac,
+	.irq		= falcon_xmac_irq,
+	.poll		= falcon_poll_xmac,
 };
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index fb8d725..61d09a2 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -544,12 +544,14 @@ static inline enum efx_fc_type efx_fc_resolve(enum efx_fc_type wanted_fc,
  * struct efx_mac_operations - Efx MAC operations table
  * @reconfigure: Reconfigure MAC. Serialised by the mac_lock
  * @update_stats: Update statistics
- * @check_hw: Check hardware. Serialised by the mac_lock
+ * @irq: Hardware MAC event callback. Serialised by the mac_lock
+ * @poll: Poll for hardware state. Serialised by the mac_lock
  */
 struct efx_mac_operations {
 	void (*reconfigure) (struct efx_nic *efx);
 	void (*update_stats) (struct efx_nic *efx);
-	int (*check_hw) (struct efx_nic *efx);
+	void (*irq) (struct efx_nic *efx);
+	void (*poll) (struct efx_nic *efx);
 };
 
 /**
@@ -559,7 +561,7 @@ struct efx_mac_operations {
  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
  * @clear_interrupt: Clear down interrupt
  * @blink: Blink LEDs
- * @check_hw: Check hardware
+ * @poll: Poll for hardware state. Serialised by the mac_lock.
  * @get_settings: Get ethtool settings. Serialised by the mac_lock.
  * @set_settings: Set ethtool settings. Serialised by the mac_lock.
  * @set_xnp_advertise: Set abilities advertised in Extended Next Page
@@ -573,7 +575,7 @@ struct efx_phy_operations {
 	void (*fini) (struct efx_nic *efx);
 	void (*reconfigure) (struct efx_nic *efx);
 	void (*clear_interrupt) (struct efx_nic *efx);
-	int (*check_hw) (struct efx_nic *efx);
+	void (*poll) (struct efx_nic *efx);
 	int (*test) (struct efx_nic *efx);
 	void (*get_settings) (struct efx_nic *efx,
 			      struct ethtool_cmd *ecmd);
@@ -728,10 +730,10 @@ union efx_multicast_hash {
  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
  *	@port_inhibited, efx_monitor() and efx_reconfigure_port()
  * @port_enabled: Port enabled indicator.
- *	Serialises efx_stop_all(), efx_start_all() and efx_monitor() and
- *	efx_reconfigure_work with kernel interfaces. Safe to read under any
- *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
- *	be held to modify it.
+ *	Serialises efx_stop_all(), efx_start_all(), efx_monitor(),
+ *	efx_phy_work(), and efx_mac_work() with kernel interfaces. Safe to read
+ *	under any one of the rtnl_lock, mac_lock, or netif_tx_lock, but all
+ *	three must be held to modify it.
  * @port_inhibited: If set, the netif_carrier is always off. Hold the mac_lock
  * @port_initialized: Port initialized?
  * @net_dev: Operating system network device. Consider holding the rtnl lock
@@ -762,7 +764,8 @@ union efx_multicast_hash {
  * @promiscuous: Promiscuous flag. Protected by netif_tx_lock.
  * @multicast_hash: Multicast hash table
  * @wanted_fc: Wanted flow control flags
- * @reconfigure_work: work item for dealing with PHY events
+ * @phy_work: work item for dealing with PHY events
+ * @mac_work: work item for dealing with MAC events
  * @loopback_mode: Loopback status
  * @loopback_modes: Supported loopback mode bitmask
  * @loopback_selftest: Offline self-test private state
@@ -810,6 +813,7 @@ struct efx_nic {
 	struct falcon_nic_data *nic_data;
 
 	struct mutex mac_lock;
+	struct work_struct mac_work;
 	bool port_enabled;
 	bool port_inhibited;
 
@@ -830,6 +834,7 @@ struct efx_nic {
 
 	enum phy_type phy_type;
 	spinlock_t phy_lock;
+	struct work_struct phy_work;
 	struct efx_phy_operations *phy_op;
 	void *phy_data;
 	struct mii_if_info mii;
@@ -845,7 +850,6 @@ struct efx_nic {
 	bool promiscuous;
 	union efx_multicast_hash multicast_hash;
 	enum efx_fc_type wanted_fc;
-	struct work_struct reconfigure_work;
 
 	atomic_t rx_reset;
 	enum efx_loopback_mode loopback_mode;
diff --git a/drivers/net/sfc/selftest.c b/drivers/net/sfc/selftest.c
index 8142e37..578b7f4 100644
--- a/drivers/net/sfc/selftest.c
+++ b/drivers/net/sfc/selftest.c
@@ -594,12 +594,14 @@ static int efx_test_loopbacks(struct efx_nic *efx, struct ethtool_cmd ecmd,
 		efx->loopback_mode = mode;
 		efx_reconfigure_port(efx);
 
-		/* Wait for the PHY to signal the link is up */
+		/* Wait for the PHY to signal the link is up. Interrupts
+		 * are enabled for PHY's using LASI, otherwise we poll()
+		 * quickly */
 		count = 0;
 		do {
 			struct efx_channel *channel = &efx->channel[0];
 
-			efx->mac_op->check_hw(efx);
+			efx->phy_op->poll(efx);
 			schedule_timeout_uninterruptible(HZ / 10);
 			if (channel->work_pending)
 				efx_process_channel_now(channel);
diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c
index 634ff91..7256ea4 100644
--- a/drivers/net/sfc/tenxpress.c
+++ b/drivers/net/sfc/tenxpress.c
@@ -348,50 +348,34 @@ static void tenxpress_phy_reconfigure(struct efx_nic *efx)
 	efx->link_fc = mdio_clause45_get_pause(efx);
 }
 
-static void tenxpress_phy_clear_interrupt(struct efx_nic *efx)
-{
-	/* Nothing done here - LASI interrupts aren't reliable so poll  */
-}
-
-
 /* Poll PHY for interrupt */
-static int tenxpress_phy_check_hw(struct efx_nic *efx)
+static void tenxpress_phy_poll(struct efx_nic *efx)
 {
 	struct tenxpress_phy_data *phy_data = efx->phy_data;
-	bool link_ok;
-	int rc = 0;
+	bool change = false, link_ok;
+	unsigned link_fc;
 
 	link_ok = tenxpress_link_ok(efx);
+	if (link_ok != efx->link_up) {
+		change = true;
+	} else {
+		link_fc = mdio_clause45_get_pause(efx);
+		if (link_fc != efx->link_fc)
+			change = true;
+	}
 	tenxpress_check_bad_lp(efx, link_ok);
 
-	if (link_ok != efx->link_up)
+	if (change)
 		falcon_sim_phy_event(efx);
 
 	if (phy_data->phy_mode != PHY_MODE_NORMAL)
-		return 0;
+		return;
 
 	if (atomic_read(&phy_data->bad_crc_count) > crc_error_reset_threshold) {
 		EFX_ERR(efx, "Resetting XAUI due to too many CRC errors\n");
 		falcon_reset_xaui(efx);
 		atomic_set(&phy_data->bad_crc_count, 0);
 	}
-
-	rc = efx->board_info.monitor(efx);
-	if (rc) {
-		EFX_ERR(efx, "Board sensor %s; shutting down PHY\n",
-			(rc == -ERANGE) ? "reported fault" : "failed");
-		if (efx->phy_mode & PHY_MODE_OFF) {
-			/* Assume that board has shut PHY off */
-			phy_data->phy_mode = PHY_MODE_OFF;
-		} else {
-			efx->phy_mode |= PHY_MODE_LOW_POWER;
-			mdio_clause45_set_mmds_lpower(efx, true,
-						      efx->phy_op->mmds);
-			phy_data->phy_mode |= PHY_MODE_LOW_POWER;
-		}
-	}
-
-	return rc;
 }
 
 static void tenxpress_phy_fini(struct efx_nic *efx)
@@ -461,9 +445,9 @@ struct efx_phy_operations falcon_tenxpress_phy_ops = {
 	.macs		  = EFX_XMAC,
 	.init             = tenxpress_phy_init,
 	.reconfigure      = tenxpress_phy_reconfigure,
-	.check_hw         = tenxpress_phy_check_hw,
+	.poll             = tenxpress_phy_poll,
 	.fini             = tenxpress_phy_fini,
-	.clear_interrupt  = tenxpress_phy_clear_interrupt,
+	.clear_interrupt  = efx_port_dummy_op_void,
 	.test             = tenxpress_phy_test,
 	.get_settings	  = tenxpress_get_settings,
 	.set_settings	  = mdio_clause45_set_settings,
diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c
index fbe8e25..345ffc3 100644
--- a/drivers/net/sfc/xfp_phy.c
+++ b/drivers/net/sfc/xfp_phy.c
@@ -119,24 +119,12 @@ static int xfp_link_ok(struct efx_nic *efx)
 	return mdio_clause45_links_ok(efx, XFP_REQUIRED_DEVS);
 }
 
-static int xfp_phy_check_hw(struct efx_nic *efx)
+static void xfp_phy_poll(struct efx_nic *efx)
 {
-	int rc = 0;
 	int link_up = xfp_link_ok(efx);
 	/* Simulate a PHY event if link state has changed */
 	if (link_up != efx->link_up)
 		falcon_sim_phy_event(efx);
-
-	rc = efx->board_info.monitor(efx);
-	if (rc) {
-		struct xfp_phy_data *phy_data = efx->phy_data;
-		EFX_ERR(efx, "XFP sensor alert; putting PHY into low power\n");
-		efx->phy_mode |= PHY_MODE_LOW_POWER;
-		mdio_clause45_set_mmds_lpower(efx, 1, XFP_REQUIRED_DEVS);
-		phy_data->phy_mode |= PHY_MODE_LOW_POWER;
-	}
-
-	return rc;
 }
 
 static void xfp_phy_reconfigure(struct efx_nic *efx)
@@ -173,7 +161,7 @@ struct efx_phy_operations falcon_xfp_phy_ops = {
 	.macs		 = EFX_XMAC,
 	.init            = xfp_phy_init,
 	.reconfigure     = xfp_phy_reconfigure,
-	.check_hw        = xfp_phy_check_hw,
+	.poll            = xfp_phy_poll,
 	.fini            = xfp_phy_fini,
 	.clear_interrupt = xfp_phy_clear_interrupt,
 	.get_settings    = mdio_clause45_get_settings,
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 21/33] sfc: Rework MAC, PHY and board event handling

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:55:47 +0000

> From: Steve Hodgson <shodgson@solarflare.com>
> 
> MAC, PHY and board events may be separately enabled and signalled.
> Our current arrangement of chaining the polling functions can result
> in events being missed.  Change them to be more independent.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 22/33] sfc: Add support for Solarflare 10Xpress SFT9001

[Ben Hutchings]

Add type codes for the new PHY and rename the SFX7101 type code.

Add definition of clause 22 extension MMD.

Adapt the 10Xpress SFX7101 code to support the SFT9001 as well.
Clean up register definitions.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/enum.h        |    3 +
 drivers/net/sfc/ethtool.c     |    3 +
 drivers/net/sfc/falcon.c      |   10 +-
 drivers/net/sfc/mdio_10g.h    |    8 +
 drivers/net/sfc/net_driver.h  |    4 +-
 drivers/net/sfc/phy.h         |    7 +-
 drivers/net/sfc/tenxpress.c   |  528 +++++++++++++++++++++++++++++++++--------
 drivers/net/sfc/workarounds.h |    9 +-
 8 files changed, 467 insertions(+), 105 deletions(-)

diff --git a/drivers/net/sfc/enum.h b/drivers/net/sfc/enum.h
index 0c2e41c..60cbc6e 100644
--- a/drivers/net/sfc/enum.h
+++ b/drivers/net/sfc/enum.h
@@ -58,6 +58,9 @@ extern const char *efx_loopback_mode_names[];
 #define LOOPBACK_INTERNAL(_efx)				\
 	(!!(LOOPBACKS_INTERNAL & LOOPBACK_MASK(_efx)))
 
+#define LOOPBACK_CHANGED(_from, _to, _mask)				\
+	(!!((LOOPBACK_MASK(_from) ^ LOOPBACK_MASK(_to)) & (_mask)))
+
 #define LOOPBACK_OUT_OF(_from, _to, _mask)				\
 	((LOOPBACK_MASK(_from) & (_mask)) && !(LOOPBACK_MASK(_to) & (_mask)))
 
diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index 0e81af6..aad0d1b 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -219,6 +219,9 @@ int efx_ethtool_set_settings(struct net_device *net_dev,
 	struct efx_nic *efx = netdev_priv(net_dev);
 	int rc;
 
+	if (EFX_WORKAROUND_13963(efx) && !ecmd->autoneg)
+		return -EINVAL;
+
 	/* Falcon GMAC does not support 1000Mbps HD */
 	if (ecmd->speed == SPEED_1000 && ecmd->duplex != DUPLEX_FULL) {
 		EFX_LOG(efx, "rejecting unsupported 1000Mbps HD"
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 734add1..68747c8 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -826,7 +826,7 @@ static void falcon_handle_rx_not_ok(struct efx_rx_queue *rx_queue,
 #endif
 
 	if (unlikely(rx_ev_eth_crc_err && EFX_WORKAROUND_10750(efx) &&
-		     efx->phy_type == PHY_TYPE_10XPRESS))
+		     efx->phy_type == PHY_TYPE_SFX7101))
 		tenxpress_crc_err(efx);
 }
 
@@ -2256,8 +2256,12 @@ static void falcon_init_mdio(struct mii_if_info *gmii)
 static int falcon_probe_phy(struct efx_nic *efx)
 {
 	switch (efx->phy_type) {
-	case PHY_TYPE_10XPRESS:
-		efx->phy_op = &falcon_tenxpress_phy_ops;
+	case PHY_TYPE_SFX7101:
+		efx->phy_op = &falcon_sfx7101_phy_ops;
+		break;
+	case PHY_TYPE_SFT9001A:
+	case PHY_TYPE_SFT9001B:
+		efx->phy_op = &falcon_sft9001_phy_ops;
 		break;
 	case PHY_TYPE_XFP:
 		efx->phy_op = &falcon_xfp_phy_ops;
diff --git a/drivers/net/sfc/mdio_10g.h b/drivers/net/sfc/mdio_10g.h
index 80c63dd..4091182 100644
--- a/drivers/net/sfc/mdio_10g.h
+++ b/drivers/net/sfc/mdio_10g.h
@@ -33,6 +33,8 @@
 #define MDIO_MMD_TC	(6)
 /* Auto negotiation */
 #define MDIO_MMD_AN	(7)
+/* Clause 22 extension */
+#define MDIO_MMD_C22EXT	29
 
 /* Generic register locations */
 #define MDIO_MMDREG_CTRL1	(0)
@@ -82,6 +84,7 @@
 #define MDIO_MMDREG_DEVS_PCS	DEV_PRESENT_BIT(MDIO_MMD_PCS)
 #define MDIO_MMDREG_DEVS_PMAPMD	DEV_PRESENT_BIT(MDIO_MMD_PMAPMD)
 #define MDIO_MMDREG_DEVS_AN	DEV_PRESENT_BIT(MDIO_MMD_AN)
+#define MDIO_MMDREG_DEVS_C22EXT	DEV_PRESENT_BIT(MDIO_MMD_C22EXT)
 
 /* Bits in MMDREG_SPEED */
 #define MDIO_MMDREG_SPEED_10G_LBN	0
@@ -125,6 +128,11 @@
 #define MDIO_PMAPMD_CTRL2_10_BT		(0xf)
 #define MDIO_PMAPMD_CTRL2_TYPE_MASK	(0xf)
 
+/* PMA 10GBT registers */
+#define MDIO_PMAPMD_10GBT_TXPWR		(131)
+#define MDIO_PMAPMD_10GBT_TXPWR_SHORT_LBN (0)
+#define MDIO_PMAPMD_10GBT_TXPWR_SHORT_WIDTH (1)
+
 /* PHY XGXS lane state */
 #define MDIO_PHYXS_LANE_STATE		(0x18)
 #define MDIO_PHYXS_LANE_ALIGNED_LBN	(12)
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index 61d09a2..1728329 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -455,9 +455,11 @@ enum phy_type {
 	PHY_TYPE_NONE = 0,
 	PHY_TYPE_CX4_RTMR = 1,
 	PHY_TYPE_1G_ALASKA = 2,
-	PHY_TYPE_10XPRESS = 3,
+	PHY_TYPE_SFX7101 = 3,
 	PHY_TYPE_XFP = 4,
 	PHY_TYPE_PM8358 = 6,
+	PHY_TYPE_SFT9001A = 8,
+	PHY_TYPE_SFT9001B = 10,
 	PHY_TYPE_MAX	/* Insert any new items before this */
 };
 
diff --git a/drivers/net/sfc/phy.h b/drivers/net/sfc/phy.h
index f746536..58c493e 100644
--- a/drivers/net/sfc/phy.h
+++ b/drivers/net/sfc/phy.h
@@ -1,6 +1,6 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
- * Copyright 2007 Solarflare Communications Inc.
+ * Copyright 2007-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -11,9 +11,10 @@
 #define EFX_PHY_H
 
 /****************************************************************************
- * 10Xpress (SFX7101) PHY
+ * 10Xpress (SFX7101 and SFT9001) PHYs
  */
-extern struct efx_phy_operations falcon_tenxpress_phy_ops;
+extern struct efx_phy_operations falcon_sfx7101_phy_ops;
+extern struct efx_phy_operations falcon_sft9001_phy_ops;
 
 extern void tenxpress_phy_blink(struct efx_nic *efx, bool blink);
 extern void tenxpress_crc_err(struct efx_nic *efx);
diff --git a/drivers/net/sfc/tenxpress.c b/drivers/net/sfc/tenxpress.c
index 7256ea4..b3ca2dc 100644
--- a/drivers/net/sfc/tenxpress.c
+++ b/drivers/net/sfc/tenxpress.c
@@ -15,38 +15,71 @@
 #include "phy.h"
 #include "falcon_hwdefs.h"
 #include "boards.h"
+#include "workarounds.h"
+#include "selftest.h"
 
-/* We expect these MMDs to be in the package */
+/* We expect these MMDs to be in the package.  SFT9001 also has a
+ * clause 22 extension MMD, but since it doesn't have all the generic
+ * MMD registers it is pointless to include it here.
+ */
 #define TENXPRESS_REQUIRED_DEVS (MDIO_MMDREG_DEVS_PMAPMD	| \
 				 MDIO_MMDREG_DEVS_PCS		| \
 				 MDIO_MMDREG_DEVS_PHYXS		| \
 				 MDIO_MMDREG_DEVS_AN)
 
-#define TENXPRESS_LOOPBACKS ((1 << LOOPBACK_PHYXS) |	\
-			     (1 << LOOPBACK_PCS) |	\
-			     (1 << LOOPBACK_PMAPMD) |	\
-			     (1 << LOOPBACK_NETWORK))
+#define SFX7101_LOOPBACKS ((1 << LOOPBACK_PHYXS) |	\
+			   (1 << LOOPBACK_PCS) |	\
+			   (1 << LOOPBACK_PMAPMD) |	\
+			   (1 << LOOPBACK_NETWORK))
+
+#define SFT9001_LOOPBACKS ((1 << LOOPBACK_GPHY) |	\
+			   (1 << LOOPBACK_PHYXS) |	\
+			   (1 << LOOPBACK_PCS) |	\
+			   (1 << LOOPBACK_PMAPMD) |	\
+			   (1 << LOOPBACK_NETWORK))
 
 /* We complain if we fail to see the link partner as 10G capable this many
  * times in a row (must be > 1 as sampling the autoneg. registers is racy)
  */
 #define MAX_BAD_LP_TRIES	(5)
 
+/* LASI Control */
+#define PMA_PMD_LASI_CTRL	36866
+#define PMA_PMD_LASI_STATUS	36869
+#define PMA_PMD_LS_ALARM_LBN	0
+#define PMA_PMD_LS_ALARM_WIDTH	1
+#define PMA_PMD_TX_ALARM_LBN	1
+#define PMA_PMD_TX_ALARM_WIDTH	1
+#define PMA_PMD_RX_ALARM_LBN	2
+#define PMA_PMD_RX_ALARM_WIDTH	1
+#define PMA_PMD_AN_ALARM_LBN	3
+#define PMA_PMD_AN_ALARM_WIDTH	1
+
 /* Extended control register */
-#define	PMA_PMD_XCONTROL_REG 0xc000
-#define	PMA_PMD_LNPGA_POWERDOWN_LBN 8
-#define	PMA_PMD_LNPGA_POWERDOWN_WIDTH 1
+#define PMA_PMD_XCONTROL_REG	49152
+#define PMA_PMD_EXT_GMII_EN_LBN	1
+#define PMA_PMD_EXT_GMII_EN_WIDTH 1
+#define PMA_PMD_EXT_CLK_OUT_LBN	2
+#define PMA_PMD_EXT_CLK_OUT_WIDTH 1
+#define PMA_PMD_LNPGA_POWERDOWN_LBN 8	/* SFX7101 only */
+#define PMA_PMD_LNPGA_POWERDOWN_WIDTH 1
+#define PMA_PMD_EXT_CLK312_LBN	8	/* SFT9001 only */
+#define PMA_PMD_EXT_CLK312_WIDTH 1
+#define PMA_PMD_EXT_LPOWER_LBN  12
+#define PMA_PMD_EXT_LPOWER_WIDTH 1
+#define PMA_PMD_EXT_SSR_LBN	15
+#define PMA_PMD_EXT_SSR_WIDTH	1
 
 /* extended status register */
-#define PMA_PMD_XSTATUS_REG 0xc001
+#define PMA_PMD_XSTATUS_REG	49153
 #define PMA_PMD_XSTAT_FLP_LBN   (12)
 
 /* LED control register */
-#define PMA_PMD_LED_CTRL_REG	(0xc007)
+#define PMA_PMD_LED_CTRL_REG	49159
 #define PMA_PMA_LED_ACTIVITY_LBN	(3)
 
 /* LED function override register */
-#define PMA_PMD_LED_OVERR_REG	(0xc009)
+#define PMA_PMD_LED_OVERR_REG	49161
 /* Bit positions for different LEDs (there are more but not wired on SFE4001)*/
 #define PMA_PMD_LED_LINK_LBN	(0)
 #define PMA_PMD_LED_SPEED_LBN	(2)
@@ -63,36 +96,74 @@
 /* Green and Amber under hardware control, Red off */
 #define PMA_PMD_LED_DEFAULT	(PMA_PMD_LED_OFF << PMA_PMD_LED_RX_LBN)
 
-
-/* Special Software reset register */
-#define PMA_PMD_EXT_CTRL_REG 49152
-#define PMA_PMD_EXT_SSR_LBN 15
-
-/* Misc register defines */
-#define PCS_CLOCK_CTRL_REG 0xd801
+#define PMA_PMD_SPEED_ENABLE_REG 49192
+#define PMA_PMD_100TX_ADV_LBN    1
+#define PMA_PMD_100TX_ADV_WIDTH  1
+#define PMA_PMD_1000T_ADV_LBN    2
+#define PMA_PMD_1000T_ADV_WIDTH  1
+#define PMA_PMD_10000T_ADV_LBN   3
+#define PMA_PMD_10000T_ADV_WIDTH 1
+#define PMA_PMD_SPEED_LBN        4
+#define PMA_PMD_SPEED_WIDTH      4
+
+/* Serdes control registers - SFT9001 only */
+#define PMA_PMD_CSERDES_CTRL_REG 64258
+/* Set the 156.25 MHz output to 312.5 MHz to drive Falcon's XMAC */
+#define PMA_PMD_CSERDES_DEFAULT	0x000f
+
+/* Misc register defines - SFX7101 only */
+#define PCS_CLOCK_CTRL_REG	55297
 #define PLL312_RST_N_LBN 2
 
-#define PCS_SOFT_RST2_REG 0xd806
+#define PCS_SOFT_RST2_REG	55302
 #define SERDES_RST_N_LBN 13
 #define XGXS_RST_N_LBN 12
 
-#define	PCS_TEST_SELECT_REG 0xd807	/* PRM 10.5.8 */
+#define	PCS_TEST_SELECT_REG	55303	/* PRM 10.5.8 */
 #define	CLK312_EN_LBN 3
 
 /* PHYXS registers */
+#define PHYXS_XCONTROL_REG	49152
+#define PHYXS_RESET_LBN		15
+#define PHYXS_RESET_WIDTH	1
+
 #define PHYXS_TEST1         (49162)
 #define LOOPBACK_NEAR_LBN   (8)
 #define LOOPBACK_NEAR_WIDTH (1)
 
+#define PCS_10GBASET_STAT1       32
+#define PCS_10GBASET_BLKLK_LBN   0
+#define PCS_10GBASET_BLKLK_WIDTH 1
+
 /* Boot status register */
-#define PCS_BOOT_STATUS_REG	(0xd000)
+#define PCS_BOOT_STATUS_REG	53248
 #define PCS_BOOT_FATAL_ERR_LBN	(0)
 #define PCS_BOOT_PROGRESS_LBN	(1)
 #define PCS_BOOT_PROGRESS_WIDTH	(2)
 #define PCS_BOOT_COMPLETE_LBN	(3)
+
 #define PCS_BOOT_MAX_DELAY	(100)
 #define PCS_BOOT_POLL_DELAY	(10)
 
+/* 100M/1G PHY registers */
+#define GPHY_XCONTROL_REG	49152
+#define GPHY_ISOLATE_LBN	10
+#define GPHY_ISOLATE_WIDTH	1
+#define GPHY_DUPLEX_LBN	  	8
+#define GPHY_DUPLEX_WIDTH	1
+#define GPHY_LOOPBACK_NEAR_LBN	14
+#define GPHY_LOOPBACK_NEAR_WIDTH 1
+
+#define C22EXT_STATUS_REG       49153
+#define C22EXT_STATUS_LINK_LBN  2
+#define C22EXT_STATUS_LINK_WIDTH 1
+
+#define C22EXT_MSTSLV_REG       49162
+#define C22EXT_MSTSLV_1000_HD_LBN 10
+#define C22EXT_MSTSLV_1000_HD_WIDTH 1
+#define C22EXT_MSTSLV_1000_FD_LBN 11
+#define C22EXT_MSTSLV_1000_FD_WIDTH 1
+
 /* Time to wait between powering down the LNPGA and turning off the power
  * rails */
 #define LNPGA_PDOWN_WAIT	(HZ / 5)
@@ -116,6 +187,38 @@ void tenxpress_crc_err(struct efx_nic *efx)
 		atomic_inc(&phy_data->bad_crc_count);
 }
 
+static ssize_t show_phy_short_reach(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	struct efx_nic *efx = pci_get_drvdata(to_pci_dev(dev));
+	int reg;
+
+	reg = mdio_clause45_read(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
+				 MDIO_PMAPMD_10GBT_TXPWR);
+	return sprintf(buf, "%d\n",
+		       !!(reg & (1 << MDIO_PMAPMD_10GBT_TXPWR_SHORT_LBN)));
+}
+
+static ssize_t set_phy_short_reach(struct device *dev,
+				   struct device_attribute *attr,
+				   const char *buf, size_t count)
+{
+	struct efx_nic *efx = pci_get_drvdata(to_pci_dev(dev));
+
+	rtnl_lock();
+	mdio_clause45_set_flag(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
+			       MDIO_PMAPMD_10GBT_TXPWR,
+			       MDIO_PMAPMD_10GBT_TXPWR_SHORT_LBN,
+			       count != 0 && *buf != '0');
+	efx_reconfigure_port(efx);
+	rtnl_unlock();
+
+	return count;
+}
+
+static DEVICE_ATTR(phy_short_reach, 0644, show_phy_short_reach,
+		   set_phy_short_reach);
+
 /* Check that the C166 has booted successfully */
 static int tenxpress_phy_check(struct efx_nic *efx)
 {
@@ -147,27 +250,42 @@ static int tenxpress_phy_check(struct efx_nic *efx)
 
 static int tenxpress_init(struct efx_nic *efx)
 {
-	int rc, reg;
+	int phy_id = efx->mii.phy_id;
+	int reg;
+	int rc;
 
-	/* Turn on the clock  */
-	reg = (1 << CLK312_EN_LBN);
-	mdio_clause45_write(efx, efx->mii.phy_id,
-			    MDIO_MMD_PCS, PCS_TEST_SELECT_REG, reg);
+	if (efx->phy_type == PHY_TYPE_SFX7101) {
+		/* Enable 312.5 MHz clock */
+		mdio_clause45_write(efx, phy_id,
+				    MDIO_MMD_PCS, PCS_TEST_SELECT_REG,
+				    1 << CLK312_EN_LBN);
+	} else {
+		/* Enable 312.5 MHz clock and GMII */
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+					 PMA_PMD_XCONTROL_REG);
+		reg |= ((1 << PMA_PMD_EXT_GMII_EN_LBN) |
+			(1 << PMA_PMD_EXT_CLK_OUT_LBN) |
+			(1 << PMA_PMD_EXT_CLK312_LBN));
+		mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD,
+				    PMA_PMD_XCONTROL_REG, reg);
+		mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_C22EXT,
+				       GPHY_XCONTROL_REG, GPHY_ISOLATE_LBN,
+				       false);
+	}
 
 	rc = tenxpress_phy_check(efx);
 	if (rc < 0)
 		return rc;
 
 	/* Set the LEDs up as: Green = Link, Amber = Link/Act, Red = Off */
-	reg = mdio_clause45_read(efx, efx->mii.phy_id,
-				 MDIO_MMD_PMAPMD, PMA_PMD_LED_CTRL_REG);
-	reg |= (1 << PMA_PMA_LED_ACTIVITY_LBN);
-	mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
-			    PMA_PMD_LED_CTRL_REG, reg);
-
-	reg = PMA_PMD_LED_DEFAULT;
-	mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
-			    PMA_PMD_LED_OVERR_REG, reg);
+	if (efx->phy_type == PHY_TYPE_SFX7101) {
+		mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_PMAPMD,
+				       PMA_PMD_LED_CTRL_REG,
+				       PMA_PMA_LED_ACTIVITY_LBN,
+				       true);
+		mdio_clause45_write(efx, phy_id, MDIO_MMD_PMAPMD,
+				    PMA_PMD_LED_OVERR_REG, PMA_PMD_LED_DEFAULT);
+	}
 
 	return rc;
 }
@@ -183,22 +301,43 @@ static int tenxpress_phy_init(struct efx_nic *efx)
 	efx->phy_data = phy_data;
 	phy_data->phy_mode = efx->phy_mode;
 
-	rc = mdio_clause45_wait_reset_mmds(efx,
-					   TENXPRESS_REQUIRED_DEVS);
-	if (rc < 0)
-		goto fail;
+	if (!(efx->phy_mode & PHY_MODE_SPECIAL)) {
+		if (efx->phy_type == PHY_TYPE_SFT9001A) {
+			int reg;
+			reg = mdio_clause45_read(efx, efx->mii.phy_id,
+						 MDIO_MMD_PMAPMD,
+						 PMA_PMD_XCONTROL_REG);
+			reg |= (1 << PMA_PMD_EXT_SSR_LBN);
+			mdio_clause45_write(efx, efx->mii.phy_id,
+					    MDIO_MMD_PMAPMD,
+					    PMA_PMD_XCONTROL_REG, reg);
+			mdelay(200);
+		}
 
-	rc = mdio_clause45_check_mmds(efx, TENXPRESS_REQUIRED_DEVS, 0);
-	if (rc < 0)
-		goto fail;
+		rc = mdio_clause45_wait_reset_mmds(efx,
+						   TENXPRESS_REQUIRED_DEVS);
+		if (rc < 0)
+			goto fail;
+
+		rc = mdio_clause45_check_mmds(efx, TENXPRESS_REQUIRED_DEVS, 0);
+		if (rc < 0)
+			goto fail;
+	}
 
 	rc = tenxpress_init(efx);
 	if (rc < 0)
 		goto fail;
 
+	if (efx->phy_type == PHY_TYPE_SFT9001B) {
+		rc = device_create_file(&efx->pci_dev->dev,
+					&dev_attr_phy_short_reach);
+		if (rc)
+			goto fail;
+	}
+
 	schedule_timeout_uninterruptible(HZ / 5); /* 200ms */
 
-	/* Let XGXS and SerDes out of reset and resets 10XPress */
+	/* Let XGXS and SerDes out of reset */
 	falcon_reset_xaui(efx);
 
 	return 0;
@@ -209,21 +348,24 @@ static int tenxpress_phy_init(struct efx_nic *efx)
 	return rc;
 }
 
+/* Perform a "special software reset" on the PHY. The caller is
+ * responsible for saving and restoring the PHY hardware registers
+ * properly, and masking/unmasking LASI */
 static int tenxpress_special_reset(struct efx_nic *efx)
 {
 	int rc, reg;
 
 	/* The XGMAC clock is driven from the SFC7101/SFT9001 312MHz clock, so
 	 * a special software reset can glitch the XGMAC sufficiently for stats
-	 * requests to fail. Since we don't ofen special_reset, just lock. */
+	 * requests to fail. Since we don't often special_reset, just lock. */
 	spin_lock(&efx->stats_lock);
 
 	/* Initiate reset */
 	reg = mdio_clause45_read(efx, efx->mii.phy_id,
-				 MDIO_MMD_PMAPMD, PMA_PMD_EXT_CTRL_REG);
+				 MDIO_MMD_PMAPMD, PMA_PMD_XCONTROL_REG);
 	reg |= (1 << PMA_PMD_EXT_SSR_LBN);
 	mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
-			    PMA_PMD_EXT_CTRL_REG, reg);
+			    PMA_PMD_XCONTROL_REG, reg);
 
 	mdelay(200);
 
@@ -238,12 +380,14 @@ static int tenxpress_special_reset(struct efx_nic *efx)
 	if (rc < 0)
 		goto unlock;
 
+	/* Wait for the XGXS state machine to churn */
+	mdelay(10);
 unlock:
 	spin_unlock(&efx->stats_lock);
 	return rc;
 }
 
-static void tenxpress_check_bad_lp(struct efx_nic *efx, bool link_ok)
+static void sfx7101_check_bad_lp(struct efx_nic *efx, bool link_ok)
 {
 	struct tenxpress_phy_data *pd = efx->phy_data;
 	int phy_id = efx->mii.phy_id;
@@ -288,63 +432,142 @@ static void tenxpress_check_bad_lp(struct efx_nic *efx, bool link_ok)
 	}
 }
 
-static bool tenxpress_link_ok(struct efx_nic *efx)
+static bool sfx7101_link_ok(struct efx_nic *efx)
 {
-	if (efx->loopback_mode == LOOPBACK_NONE)
-		return mdio_clause45_links_ok(efx, MDIO_MMDREG_DEVS_AN);
-	else
+	return mdio_clause45_links_ok(efx,
+				      MDIO_MMDREG_DEVS_PMAPMD |
+				      MDIO_MMDREG_DEVS_PCS |
+				      MDIO_MMDREG_DEVS_PHYXS);
+}
+
+static bool sft9001_link_ok(struct efx_nic *efx, struct ethtool_cmd *ecmd)
+{
+	int phy_id = efx->mii.phy_id;
+	u32 reg;
+
+	if (efx->loopback_mode == LOOPBACK_GPHY)
+		return true;
+	else if (efx_phy_mode_disabled(efx->phy_mode))
+		return false;
+	else if (efx->loopback_mode)
 		return mdio_clause45_links_ok(efx,
 					      MDIO_MMDREG_DEVS_PMAPMD |
 					      MDIO_MMDREG_DEVS_PCS |
 					      MDIO_MMDREG_DEVS_PHYXS);
+
+	/* We must use the same definition of link state as LASI,
+	 * otherwise we can miss a link state transition
+	 */
+	if (ecmd->speed == 10000) {
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PCS,
+					 PCS_10GBASET_STAT1);
+		return reg & (1 << PCS_10GBASET_BLKLK_LBN);
+	} else {
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_C22EXT,
+					 C22EXT_STATUS_REG);
+		return reg & (1 << C22EXT_STATUS_LINK_LBN);
+	}
 }
 
-static void tenxpress_phyxs_loopback(struct efx_nic *efx)
+static void tenxpress_ext_loopback(struct efx_nic *efx)
 {
 	int phy_id = efx->mii.phy_id;
-	int ctrl1, ctrl2;
 
-	ctrl1 = ctrl2 = mdio_clause45_read(efx, phy_id, MDIO_MMD_PHYXS,
-					   PHYXS_TEST1);
-	if (efx->loopback_mode == LOOPBACK_PHYXS)
-		ctrl2 |= (1 << LOOPBACK_NEAR_LBN);
+	mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_PHYXS,
+			       PHYXS_TEST1, LOOPBACK_NEAR_LBN,
+			       efx->loopback_mode == LOOPBACK_PHYXS);
+	if (efx->phy_type != PHY_TYPE_SFX7101)
+		mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_C22EXT,
+				       GPHY_XCONTROL_REG,
+				       GPHY_LOOPBACK_NEAR_LBN,
+				       efx->loopback_mode == LOOPBACK_GPHY);
+}
+
+static void tenxpress_low_power(struct efx_nic *efx)
+{
+	int phy_id = efx->mii.phy_id;
+
+	if (efx->phy_type == PHY_TYPE_SFX7101)
+		mdio_clause45_set_mmds_lpower(
+			efx, !!(efx->phy_mode & PHY_MODE_LOW_POWER),
+			TENXPRESS_REQUIRED_DEVS);
 	else
-		ctrl2 &= ~(1 << LOOPBACK_NEAR_LBN);
-	if (ctrl1 != ctrl2)
-		mdio_clause45_write(efx, phy_id, MDIO_MMD_PHYXS,
-				    PHYXS_TEST1, ctrl2);
+		mdio_clause45_set_flag(
+			efx, phy_id, MDIO_MMD_PMAPMD,
+			PMA_PMD_XCONTROL_REG, PMA_PMD_EXT_LPOWER_LBN,
+			!!(efx->phy_mode & PHY_MODE_LOW_POWER));
 }
 
 static void tenxpress_phy_reconfigure(struct efx_nic *efx)
 {
 	struct tenxpress_phy_data *phy_data = efx->phy_data;
-	bool loop_change = LOOPBACK_OUT_OF(phy_data, efx,
-					   TENXPRESS_LOOPBACKS);
+	struct ethtool_cmd ecmd;
+	bool phy_mode_change, loop_reset, loop_toggle, loopback;
 
-	if (efx->phy_mode & PHY_MODE_SPECIAL) {
+	if (efx->phy_mode & (PHY_MODE_OFF | PHY_MODE_SPECIAL)) {
 		phy_data->phy_mode = efx->phy_mode;
 		return;
 	}
 
-	/* When coming out of transmit disable, coming out of low power
-	 * mode, or moving out of any PHY internal loopback mode,
-	 * perform a special software reset */
-	if ((efx->phy_mode == PHY_MODE_NORMAL &&
-	     phy_data->phy_mode != PHY_MODE_NORMAL) ||
-	    loop_change) {
-		tenxpress_special_reset(efx);
-		falcon_reset_xaui(efx);
+	tenxpress_low_power(efx);
+
+	phy_mode_change = (efx->phy_mode == PHY_MODE_NORMAL &&
+			   phy_data->phy_mode != PHY_MODE_NORMAL);
+	loopback = LOOPBACK_MASK(efx) & efx->phy_op->loopbacks;
+	loop_toggle = LOOPBACK_CHANGED(phy_data, efx, efx->phy_op->loopbacks);
+	loop_reset = (LOOPBACK_OUT_OF(phy_data, efx, efx->phy_op->loopbacks) ||
+		      LOOPBACK_CHANGED(phy_data, efx, 1 << LOOPBACK_GPHY));
+
+	if (loop_reset || loop_toggle || loopback || phy_mode_change) {
+		int rc;
+
+		efx->phy_op->get_settings(efx, &ecmd);
+
+		if (loop_reset || phy_mode_change) {
+			tenxpress_special_reset(efx);
+
+			/* Reset XAUI if we were in 10G, and are staying
+			 * in 10G. If we're moving into and out of 10G
+			 * then xaui will be reset anyway */
+			if (EFX_IS10G(efx))
+				falcon_reset_xaui(efx);
+		}
+
+		if (efx->phy_type != PHY_TYPE_SFX7101) {
+			/* Only change autoneg once, on coming out or
+			 * going into loopback */
+			if (loop_toggle)
+				ecmd.autoneg = !loopback;
+			if (loopback) {
+				ecmd.duplex = DUPLEX_FULL;
+				if (efx->loopback_mode == LOOPBACK_GPHY)
+					ecmd.speed = SPEED_1000;
+				else
+					ecmd.speed = SPEED_10000;
+			}
+		}
+
+		rc = efx->phy_op->set_settings(efx, &ecmd);
+		WARN_ON(rc);
 	}
 
 	mdio_clause45_transmit_disable(efx);
 	mdio_clause45_phy_reconfigure(efx);
-	tenxpress_phyxs_loopback(efx);
+	tenxpress_ext_loopback(efx);
 
 	phy_data->loopback_mode = efx->loopback_mode;
 	phy_data->phy_mode = efx->phy_mode;
-	efx->link_up = tenxpress_link_ok(efx);
-	efx->link_speed = 10000;
-	efx->link_fd = true;
+
+	if (efx->phy_type == PHY_TYPE_SFX7101) {
+		efx->link_speed = 10000;
+		efx->link_fd = true;
+		efx->link_up = sfx7101_link_ok(efx);
+	} else {
+		efx->phy_op->get_settings(efx, &ecmd);
+		efx->link_speed = ecmd.speed;
+		efx->link_fd = ecmd.duplex == DUPLEX_FULL;
+		efx->link_up = sft9001_link_ok(efx, &ecmd);
+	}
 	efx->link_fc = mdio_clause45_get_pause(efx);
 }
 
@@ -355,15 +578,23 @@ static void tenxpress_phy_poll(struct efx_nic *efx)
 	bool change = false, link_ok;
 	unsigned link_fc;
 
-	link_ok = tenxpress_link_ok(efx);
-	if (link_ok != efx->link_up) {
-		change = true;
+	if (efx->phy_type == PHY_TYPE_SFX7101) {
+		link_ok = sfx7101_link_ok(efx);
+		if (link_ok != efx->link_up) {
+			change = true;
+		} else {
+			link_fc = mdio_clause45_get_pause(efx);
+			if (link_fc != efx->link_fc)
+				change = true;
+		}
+		sfx7101_check_bad_lp(efx, link_ok);
 	} else {
-		link_fc = mdio_clause45_get_pause(efx);
-		if (link_fc != efx->link_fc)
+		u32 status = mdio_clause45_read(efx, efx->mii.phy_id,
+						MDIO_MMD_PMAPMD,
+						PMA_PMD_LASI_STATUS);
+		if (status & (1 << PMA_PMD_LS_ALARM_LBN))
 			change = true;
 	}
-	tenxpress_check_bad_lp(efx, link_ok);
 
 	if (change)
 		falcon_sim_phy_event(efx);
@@ -371,7 +602,8 @@ static void tenxpress_phy_poll(struct efx_nic *efx)
 	if (phy_data->phy_mode != PHY_MODE_NORMAL)
 		return;
 
-	if (atomic_read(&phy_data->bad_crc_count) > crc_error_reset_threshold) {
+	if (EFX_WORKAROUND_10750(efx) &&
+	    atomic_read(&phy_data->bad_crc_count) > crc_error_reset_threshold) {
 		EFX_ERR(efx, "Resetting XAUI due to too many CRC errors\n");
 		falcon_reset_xaui(efx);
 		atomic_set(&phy_data->bad_crc_count, 0);
@@ -382,15 +614,20 @@ static void tenxpress_phy_fini(struct efx_nic *efx)
 {
 	int reg;
 
-	/* Power down the LNPGA */
-	reg = (1 << PMA_PMD_LNPGA_POWERDOWN_LBN);
-	mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
-			    PMA_PMD_XCONTROL_REG, reg);
-
-	/* Waiting here ensures that the board fini, which can turn off the
-	 * power to the PHY, won't get run until the LNPGA powerdown has been
-	 * given long enough to complete. */
-	schedule_timeout_uninterruptible(LNPGA_PDOWN_WAIT); /* 200 ms */
+	if (efx->phy_type == PHY_TYPE_SFT9001B) {
+		device_remove_file(&efx->pci_dev->dev,
+				   &dev_attr_phy_short_reach);
+	} else {
+		/* Power down the LNPGA */
+		reg = (1 << PMA_PMD_LNPGA_POWERDOWN_LBN);
+		mdio_clause45_write(efx, efx->mii.phy_id, MDIO_MMD_PMAPMD,
+				    PMA_PMD_XCONTROL_REG, reg);
+
+		/* Waiting here ensures that the board fini, which can turn
+		 * off the power to the PHY, won't get run until the LNPGA
+		 * powerdown has been given long enough to complete. */
+		schedule_timeout_uninterruptible(LNPGA_PDOWN_WAIT); /* 200 ms */
+	}
 
 	kfree(efx->phy_data);
 	efx->phy_data = NULL;
@@ -426,14 +663,21 @@ static u32 tenxpress_get_xnp_lpa(struct efx_nic *efx)
 	u32 lpa = 0;
 	int reg;
 
+	if (efx->phy_type != PHY_TYPE_SFX7101) {
+		reg = mdio_clause45_read(efx, phy, MDIO_MMD_C22EXT,
+					 C22EXT_MSTSLV_REG);
+		if (reg & (1 << C22EXT_MSTSLV_1000_HD_LBN))
+			lpa |= ADVERTISED_1000baseT_Half;
+		if (reg & (1 << C22EXT_MSTSLV_1000_FD_LBN))
+			lpa |= ADVERTISED_1000baseT_Full;
+	}
 	reg = mdio_clause45_read(efx, phy, MDIO_MMD_AN, MDIO_AN_10GBT_STATUS);
 	if (reg & (1 << MDIO_AN_10GBT_STATUS_LP_10G_LBN))
 		lpa |= ADVERTISED_10000baseT_Full;
 	return lpa;
 }
 
-static void
-tenxpress_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
+static void sfx7101_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
 {
 	mdio_clause45_get_settings_ext(efx, ecmd, ADVERTISED_10000baseT_Full,
 				       tenxpress_get_xnp_lpa(efx));
@@ -441,7 +685,82 @@ tenxpress_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
 	ecmd->advertising |= ADVERTISED_10000baseT_Full;
 }
 
-struct efx_phy_operations falcon_tenxpress_phy_ops = {
+static void sft9001_get_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
+{
+	int phy_id = efx->mii.phy_id;
+	u32 xnp_adv = 0;
+	int reg;
+
+	reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_PMAPMD,
+				 PMA_PMD_SPEED_ENABLE_REG);
+	if (EFX_WORKAROUND_13204(efx) && (reg & (1 << PMA_PMD_100TX_ADV_LBN)))
+		xnp_adv |= ADVERTISED_100baseT_Full;
+	if (reg & (1 << PMA_PMD_1000T_ADV_LBN))
+		xnp_adv |= ADVERTISED_1000baseT_Full;
+	if (reg & (1 << PMA_PMD_10000T_ADV_LBN))
+		xnp_adv |= ADVERTISED_10000baseT_Full;
+
+	mdio_clause45_get_settings_ext(efx, ecmd, xnp_adv,
+				       tenxpress_get_xnp_lpa(efx));
+
+	ecmd->supported |= (SUPPORTED_100baseT_Half |
+			    SUPPORTED_100baseT_Full |
+			    SUPPORTED_1000baseT_Full);
+
+	/* Use the vendor defined C22ext register for duplex settings */
+	if (ecmd->speed != SPEED_10000 && !ecmd->autoneg) {
+		reg = mdio_clause45_read(efx, phy_id, MDIO_MMD_C22EXT,
+					 GPHY_XCONTROL_REG);
+		ecmd->duplex = (reg & (1 << GPHY_DUPLEX_LBN) ?
+				DUPLEX_FULL : DUPLEX_HALF);
+	}
+}
+
+static int sft9001_set_settings(struct efx_nic *efx, struct ethtool_cmd *ecmd)
+{
+	int phy_id = efx->mii.phy_id;
+	int rc;
+
+	rc = mdio_clause45_set_settings(efx, ecmd);
+	if (rc)
+		return rc;
+
+	if (ecmd->speed != SPEED_10000 && !ecmd->autoneg)
+		mdio_clause45_set_flag(efx, phy_id, MDIO_MMD_C22EXT,
+				       GPHY_XCONTROL_REG, GPHY_DUPLEX_LBN,
+				       ecmd->duplex == DUPLEX_FULL);
+
+	return rc;
+}
+
+static bool sft9001_set_xnp_advertise(struct efx_nic *efx, u32 advertising)
+{
+	int phy = efx->mii.phy_id;
+	int reg = mdio_clause45_read(efx, phy, MDIO_MMD_PMAPMD,
+				     PMA_PMD_SPEED_ENABLE_REG);
+	bool enabled;
+
+	reg &= ~((1 << 2) | (1 << 3));
+	if (EFX_WORKAROUND_13204(efx) &&
+	    (advertising & ADVERTISED_100baseT_Full))
+		reg |= 1 << PMA_PMD_100TX_ADV_LBN;
+	if (advertising & ADVERTISED_1000baseT_Full)
+		reg |= 1 << PMA_PMD_1000T_ADV_LBN;
+	if (advertising & ADVERTISED_10000baseT_Full)
+		reg |= 1 << PMA_PMD_10000T_ADV_LBN;
+	mdio_clause45_write(efx, phy, MDIO_MMD_PMAPMD,
+			    PMA_PMD_SPEED_ENABLE_REG, reg);
+
+	enabled = (advertising &
+		   (ADVERTISED_1000baseT_Half |
+		    ADVERTISED_1000baseT_Full |
+		    ADVERTISED_10000baseT_Full));
+	if (EFX_WORKAROUND_13204(efx))
+		enabled |= (advertising & ADVERTISED_100baseT_Full);
+	return enabled;
+}
+
+struct efx_phy_operations falcon_sfx7101_phy_ops = {
 	.macs		  = EFX_XMAC,
 	.init             = tenxpress_phy_init,
 	.reconfigure      = tenxpress_phy_reconfigure,
@@ -449,8 +768,23 @@ struct efx_phy_operations falcon_tenxpress_phy_ops = {
 	.fini             = tenxpress_phy_fini,
 	.clear_interrupt  = efx_port_dummy_op_void,
 	.test             = tenxpress_phy_test,
-	.get_settings	  = tenxpress_get_settings,
+	.get_settings	  = sfx7101_get_settings,
 	.set_settings	  = mdio_clause45_set_settings,
 	.mmds             = TENXPRESS_REQUIRED_DEVS,
-	.loopbacks        = TENXPRESS_LOOPBACKS,
+	.loopbacks        = SFX7101_LOOPBACKS,
+};
+
+struct efx_phy_operations falcon_sft9001_phy_ops = {
+	.macs		  = EFX_GMAC | EFX_XMAC,
+	.init             = tenxpress_phy_init,
+	.reconfigure      = tenxpress_phy_reconfigure,
+	.poll             = tenxpress_phy_poll,
+	.fini             = tenxpress_phy_fini,
+	.clear_interrupt  = efx_port_dummy_op_void,
+	.test             = tenxpress_phy_test,
+	.get_settings	  = sft9001_get_settings,
+	.set_settings	  = sft9001_set_settings,
+	.set_xnp_advertise = sft9001_set_xnp_advertise,
+	.mmds             = TENXPRESS_REQUIRED_DEVS,
+	.loopbacks        = SFT9001_LOOPBACKS,
 };
diff --git a/drivers/net/sfc/workarounds.h b/drivers/net/sfc/workarounds.h
index ec50b90..ecebff2 100644
--- a/drivers/net/sfc/workarounds.h
+++ b/drivers/net/sfc/workarounds.h
@@ -17,6 +17,8 @@
 
 #define EFX_WORKAROUND_ALWAYS(efx) 1
 #define EFX_WORKAROUND_FALCON_A(efx) (falcon_rev(efx) <= FALCON_REV_A1)
+#define EFX_WORKAROUND_SFX7101(efx) ((efx)->phy_type == PHY_TYPE_SFX7101)
+#define EFX_WORKAROUND_SFT9001A(efx) ((efx)->phy_type == PHY_TYPE_SFT9001A)
 
 /* XAUI resets if link not detected */
 #define EFX_WORKAROUND_5147 EFX_WORKAROUND_ALWAYS
@@ -27,7 +29,7 @@
 /* TX pkt parser problem with <= 16 byte TXes */
 #define EFX_WORKAROUND_9141 EFX_WORKAROUND_ALWAYS
 /* Low rate CRC errors require XAUI reset */
-#define EFX_WORKAROUND_10750 EFX_WORKAROUND_ALWAYS
+#define EFX_WORKAROUND_10750 EFX_WORKAROUND_SFX7101
 /* TX_EV_PKT_ERR can be caused by a dangling TX descriptor
  * or a PCIe error (bug 11028) */
 #define EFX_WORKAROUND_10727 EFX_WORKAROUND_ALWAYS
@@ -51,4 +53,9 @@
 /* Leak overlength packets rather than free */
 #define EFX_WORKAROUND_8071 EFX_WORKAROUND_FALCON_A
 
+/* Need to send XNP pages for 100BaseT */
+#define EFX_WORKAROUND_13204 EFX_WORKAROUND_SFT9001A
+/* Need to keep AN enabled */
+#define EFX_WORKAROUND_13963 EFX_WORKAROUND_SFT9001A
+
 #endif /* EFX_WORKAROUNDS_H */
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 22/33] sfc: Add support for Solarflare 10Xpress SFT9001

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:56:06 +0000

> Add type codes for the new PHY and rename the SFX7101 type code.
> 
> Add definition of clause 22 extension MMD.
> 
> Adapt the 10Xpress SFX7101 code to support the SFT9001 as well.
> Clean up register definitions.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 23/33] sfc: Add support for SFN4111T

[Ben Hutchings]

Add support code for the SFN4111T 100/1000/10GBASE-T reference design,
based in part on the existing code for the SFE4001.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/boards.c        |    2 +
 drivers/net/sfc/boards.h        |    7 ++-
 drivers/net/sfc/falcon.c        |    7 ++
 drivers/net/sfc/falcon_hwdefs.h |    2 +
 drivers/net/sfc/selftest.c      |   11 +++-
 drivers/net/sfc/sfe4001.c       |  153 +++++++++++++++++++++++++++++++--------
 drivers/net/sfc/workarounds.h   |    3 +-
 7 files changed, 150 insertions(+), 35 deletions(-)

diff --git a/drivers/net/sfc/boards.c b/drivers/net/sfc/boards.c
index 08fa4e3..6490349 100644
--- a/drivers/net/sfc/boards.c
+++ b/drivers/net/sfc/boards.c
@@ -241,6 +241,8 @@ struct efx_board_data {
 static struct efx_board_data board_data[] = {
 	{ EFX_BOARD_SFE4001, "SFE4001", "10GBASE-T adapter", sfe4001_init },
 	{ EFX_BOARD_SFE4002, "SFE4002", "XFP adapter", sfe4002_init },
+	{ EFX_BOARD_SFN4111T, "SFN4111T", "100/1000/10GBASE-T adapter",
+	  sfn4111t_init },
 };
 
 void efx_set_board_info(struct efx_nic *efx, u16 revision_info)
diff --git a/drivers/net/sfc/boards.h b/drivers/net/sfc/boards.h
index 5e0dde5..d93c6c6 100644
--- a/drivers/net/sfc/boards.h
+++ b/drivers/net/sfc/boards.h
@@ -12,11 +12,16 @@
 
 /* Board IDs (must fit in 8 bits) */
 enum efx_board_type {
-	EFX_BOARD_SFE4001 = 1,   /* SFE4001 (10GBASE-T) */
+	EFX_BOARD_SFE4001 = 1,
 	EFX_BOARD_SFE4002 = 2,
+	EFX_BOARD_SFN4111T = 0x51,
 };
 
 extern void efx_set_board_info(struct efx_nic *efx, u16 revision_info);
+
+/* SFE4001 (10GBASE-T) */
 extern int sfe4001_init(struct efx_nic *efx);
+/* SFN4111T (100/1000/10GBASE-T) */
+extern int sfn4111t_init(struct efx_nic *efx);
 
 #endif
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 68747c8..2c53ef4 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -2982,6 +2982,13 @@ int falcon_init_nic(struct efx_nic *efx)
 	EFX_SET_OWORD_FIELD(temp, ONCHIP_SRAM, 1);
 	falcon_write(efx, &temp, NIC_STAT_REG);
 
+	/* Set the source of the GMAC clock */
+	if (falcon_rev(efx) == FALCON_REV_B0) {
+		falcon_read(efx, &temp, GPIO_CTL_REG_KER);
+		EFX_SET_OWORD_FIELD(temp, GPIO_USE_NIC_CLK, true);
+		falcon_write(efx, &temp, GPIO_CTL_REG_KER);
+	}
+
 	/* Set buffer table mode */
 	EFX_POPULATE_OWORD_1(temp, BUF_TBL_MODE, BUF_TBL_MODE_FULL);
 	falcon_write(efx, &temp, BUF_TBL_CFG_REG_KER);
diff --git a/drivers/net/sfc/falcon_hwdefs.h b/drivers/net/sfc/falcon_hwdefs.h
index a58c627..bda8d5b 100644
--- a/drivers/net/sfc/falcon_hwdefs.h
+++ b/drivers/net/sfc/falcon_hwdefs.h
@@ -136,6 +136,8 @@
 
 /* GPIO control register */
 #define GPIO_CTL_REG_KER 0x0210
+#define GPIO_USE_NIC_CLK_LBN (30)
+#define GPIO_USE_NIC_CLK_WIDTH (1)
 #define GPIO_OUTPUTS_LBN   (16)
 #define GPIO_OUTPUTS_WIDTH (4)
 #define GPIO_INPUTS_LBN (8)
diff --git a/drivers/net/sfc/selftest.c b/drivers/net/sfc/selftest.c
index 578b7f4..6bb09f2 100644
--- a/drivers/net/sfc/selftest.c
+++ b/drivers/net/sfc/selftest.c
@@ -702,8 +702,15 @@ int efx_offline_test(struct efx_nic *efx,
 	 */
 	mutex_lock(&efx->mac_lock);
 	efx->port_inhibited = true;
-	if (efx->loopback_modes)
-		efx->loopback_mode = __ffs(efx->loopback_modes);
+	if (efx->loopback_modes) {
+		/* We need the 312 clock from the PHY to test the XMAC
+		 * registers, so move into XGMII loopback if available */
+		if (efx->loopback_modes & (1 << LOOPBACK_XGMII))
+			efx->loopback_mode = LOOPBACK_XGMII;
+		else
+			efx->loopback_mode = __ffs(efx->loopback_modes);
+	}
+
 	__efx_reconfigure_port(efx);
 	mutex_unlock(&efx->mac_lock);
 
diff --git a/drivers/net/sfc/sfe4001.c b/drivers/net/sfc/sfe4001.c
index af65284..16b80ac 100644
--- a/drivers/net/sfc/sfe4001.c
+++ b/drivers/net/sfc/sfe4001.c
@@ -1,6 +1,6 @@
 /****************************************************************************
  * Driver for Solarflare Solarstorm network controllers and boards
- * Copyright 2007 Solarflare Communications Inc.
+ * Copyright 2007-2008 Solarflare Communications Inc.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published
@@ -8,10 +8,21 @@
  */
 
 /*****************************************************************************
- * Support for the SFE4001 NIC: driver code for the PCA9539 I/O expander that
- * controls the PHY power rails, and for the MAX6647 temp. sensor used to check
- * the PHY
+ * Support for the SFE4001 and SFN4111T NICs.
+ *
+ * The SFE4001 does not power-up fully at reset due to its high power
+ * consumption.  We control its power via a PCA9539 I/O expander.
+ * Both boards have a MAX6647 temperature monitor which we expose to
+ * the lm90 driver.
+ *
+ * This also provides minimal support for reflashing the PHY, which is
+ * initiated by resetting it with the FLASH_CFG_1 pin pulled down.
+ * On SFE4001 rev A2 and later this is connected to the 3V3X output of
+ * the IO-expander; on the SFN4111T it is connected to Falcon's GPIO3.
+ * We represent reflash mode as PHY_MODE_SPECIAL and make it mutually
+ * exclusive with the network device being open.
  */
+
 #include <linux/delay.h>
 #include "net_driver.h"
 #include "efx.h"
@@ -171,39 +182,30 @@ fail_on:
 	return rc;
 }
 
-static int sfe4001_check_hw(struct efx_nic *efx)
+static int sfn4111t_reset(struct efx_nic *efx)
 {
-	s32 status;
+	efx_oword_t reg;
 
-	/* If XAUI link is up then do not monitor */
-	if (EFX_WORKAROUND_7884(efx) && efx->mac_up)
-		return 0;
+	/* GPIO pins are also used for I2C, so block that temporarily */
+	mutex_lock(&efx->i2c_adap.bus_lock);
 
-	/* Check the powered status of the PHY. Lack of power implies that
-	 * the MAX6647 has shut down power to it, probably due to a temp.
-	 * alarm. Reading the power status rather than the MAX6647 status
-	 * directly because the later is read-to-clear and would thus
-	 * start to power up the PHY again when polled, causing us to blip
-	 * the power undesirably.
-	 * We know we can read from the IO expander because we did
-	 * it during power-on. Assume failure now is bad news. */
-	status = i2c_smbus_read_byte_data(efx->board_info.ioexp_client, P1_IN);
-	if (status >= 0 &&
-	    (status & ((1 << P1_AFE_PWD_LBN) | (1 << P1_DSP_PWD25_LBN))) != 0)
-		return 0;
+	falcon_read(efx, &reg, GPIO_CTL_REG_KER);
+	EFX_SET_OWORD_FIELD(reg, GPIO2_OEN, true);
+	EFX_SET_OWORD_FIELD(reg, GPIO2_OUT, false);
+	falcon_write(efx, &reg, GPIO_CTL_REG_KER);
+	msleep(1000);
+	EFX_SET_OWORD_FIELD(reg, GPIO2_OUT, true);
+	EFX_SET_OWORD_FIELD(reg, GPIO3_OEN, true);
+	EFX_SET_OWORD_FIELD(reg, GPIO3_OUT,
+			    !(efx->phy_mode & PHY_MODE_SPECIAL));
+	falcon_write(efx, &reg, GPIO_CTL_REG_KER);
 
-	/* Use board power control, not PHY power control */
-	sfe4001_poweroff(efx);
-	efx->phy_mode = PHY_MODE_OFF;
+	mutex_unlock(&efx->i2c_adap.bus_lock);
 
-	return (status < 0) ? -EIO : -ERANGE;
+	ssleep(1);
+	return 0;
 }
 
-/* On SFE4001 rev A2 and later, we can control the FLASH_CFG_1 pin
- * using the 3V3X output of the IO-expander.  Allow the user to set
- * this when the device is stopped, and keep it stopped then.
- */
-
 static ssize_t show_phy_flash_cfg(struct device *dev,
 				  struct device_attribute *attr, char *buf)
 {
@@ -231,7 +233,10 @@ static ssize_t set_phy_flash_cfg(struct device *dev,
 		err = -EBUSY;
 	} else {
 		efx->phy_mode = new_mode;
-		err = sfe4001_poweron(efx);
+		if (efx->board_info.type == EFX_BOARD_SFE4001)
+			err = sfe4001_poweron(efx);
+		else
+			err = sfn4111t_reset(efx);
 		efx_reconfigure_port(efx);
 	}
 	rtnl_unlock();
@@ -251,6 +256,34 @@ static void sfe4001_fini(struct efx_nic *efx)
 	i2c_unregister_device(efx->board_info.hwmon_client);
 }
 
+static int sfe4001_check_hw(struct efx_nic *efx)
+{
+	s32 status;
+
+	/* If XAUI link is up then do not monitor */
+	if (EFX_WORKAROUND_7884(efx) && efx->mac_up)
+		return 0;
+
+	/* Check the powered status of the PHY. Lack of power implies that
+	 * the MAX6647 has shut down power to it, probably due to a temp.
+	 * alarm. Reading the power status rather than the MAX6647 status
+	 * directly because the later is read-to-clear and would thus
+	 * start to power up the PHY again when polled, causing us to blip
+	 * the power undesirably.
+	 * We know we can read from the IO expander because we did
+	 * it during power-on. Assume failure now is bad news. */
+	status = i2c_smbus_read_byte_data(efx->board_info.ioexp_client, P1_IN);
+	if (status >= 0 &&
+	    (status & ((1 << P1_AFE_PWD_LBN) | (1 << P1_DSP_PWD25_LBN))) != 0)
+		return 0;
+
+	/* Use board power control, not PHY power control */
+	sfe4001_poweroff(efx);
+	efx->phy_mode = PHY_MODE_OFF;
+
+	return (status < 0) ? -EIO : -ERANGE;
+}
+
 static struct i2c_board_info sfe4001_hwmon_info = {
 	I2C_BOARD_INFO("max6647", 0x4e),
 	.irq		= -1,
@@ -312,3 +345,61 @@ fail_hwmon:
 	i2c_unregister_device(efx->board_info.hwmon_client);
 	return rc;
 }
+
+static int sfn4111t_check_hw(struct efx_nic *efx)
+{
+	s32 status;
+
+	/* If XAUI link is up then do not monitor */
+	if (EFX_WORKAROUND_7884(efx) && efx->mac_up)
+		return 0;
+
+	/* Test LHIGH, RHIGH, FAULT, EOT and IOT alarms */
+	status = i2c_smbus_read_byte_data(efx->board_info.hwmon_client,
+					  MAX664X_REG_RSL);
+	if (status < 0)
+		return -EIO;
+	if (status & 0x57)
+		return -ERANGE;
+	return 0;
+}
+
+static void sfn4111t_fini(struct efx_nic *efx)
+{
+	EFX_INFO(efx, "%s\n", __func__);
+
+	device_remove_file(&efx->pci_dev->dev, &dev_attr_phy_flash_cfg);
+	i2c_unregister_device(efx->board_info.hwmon_client);
+}
+
+static struct i2c_board_info sfn4111t_hwmon_info = {
+	I2C_BOARD_INFO("max6647", 0x4e),
+	.irq		= -1,
+};
+
+int sfn4111t_init(struct efx_nic *efx)
+{
+	int rc;
+
+	efx->board_info.hwmon_client =
+		i2c_new_device(&efx->i2c_adap, &sfn4111t_hwmon_info);
+	if (!efx->board_info.hwmon_client)
+		return -EIO;
+
+	efx->board_info.blink = tenxpress_phy_blink;
+	efx->board_info.monitor = sfn4111t_check_hw;
+	efx->board_info.fini = sfn4111t_fini;
+
+	rc = device_create_file(&efx->pci_dev->dev, &dev_attr_phy_flash_cfg);
+	if (rc)
+		goto fail_hwmon;
+
+	if (efx->phy_mode & PHY_MODE_SPECIAL)
+		sfn4111t_reset(efx);
+
+	return 0;
+
+fail_hwmon:
+	i2c_unregister_device(efx->board_info.hwmon_client);
+	return rc;
+}
diff --git a/drivers/net/sfc/workarounds.h b/drivers/net/sfc/workarounds.h
index ecebff2..82e03e1 100644
--- a/drivers/net/sfc/workarounds.h
+++ b/drivers/net/sfc/workarounds.h
@@ -17,6 +17,7 @@
 
 #define EFX_WORKAROUND_ALWAYS(efx) 1
 #define EFX_WORKAROUND_FALCON_A(efx) (falcon_rev(efx) <= FALCON_REV_A1)
+#define EFX_WORKAROUND_10G(efx) EFX_IS10G(efx)
 #define EFX_WORKAROUND_SFX7101(efx) ((efx)->phy_type == PHY_TYPE_SFX7101)
 #define EFX_WORKAROUND_SFT9001A(efx) ((efx)->phy_type == PHY_TYPE_SFT9001A)
 
@@ -25,7 +26,7 @@
 /* RX PCIe double split performance issue */
 #define EFX_WORKAROUND_7575 EFX_WORKAROUND_ALWAYS
 /* Bit-bashed I2C reads cause performance drop */
-#define EFX_WORKAROUND_7884 EFX_WORKAROUND_ALWAYS
+#define EFX_WORKAROUND_7884 EFX_WORKAROUND_10G
 /* TX pkt parser problem with <= 16 byte TXes */
 #define EFX_WORKAROUND_9141 EFX_WORKAROUND_ALWAYS
 /* Low rate CRC errors require XAUI reset */
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 23/33] sfc: Add support for SFN4111T

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:56:24 +0000

> Add support code for the SFN4111T 100/1000/10GBASE-T reference design,
> based in part on the existing code for the SFE4001.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 25/33] sfc: Generate unique names for per-NIC workqueues

[Ben Hutchings]

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index ce1c7d3..c094845 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -1854,6 +1854,7 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
 	struct efx_channel *channel;
 	struct efx_tx_queue *tx_queue;
 	struct efx_rx_queue *rx_queue;
+	char name[16];
 	int i;
 
 	/* Initialise common structures */
@@ -1924,7 +1925,9 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
 	efx->interrupt_mode = max(efx->type->max_interrupt_mode,
 				  interrupt_mode);
 
-	efx->workqueue = create_singlethread_workqueue("sfc_work");
+	/* Would be good to use the net_dev name, but we're too early */
+	snprintf(name, sizeof(name), "sfc%s", pci_name(pci_dev));
+	efx->workqueue = create_singlethread_workqueue(name);
 	if (!efx->workqueue)
 		return -ENOMEM;
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 25/33] sfc: Generate unique names for per-NIC workqueues

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:56:57 +0000

> @@ -1854,6 +1854,7 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
>  	struct efx_channel *channel;
>  	struct efx_tx_queue *tx_queue;
>  	struct efx_rx_queue *rx_queue;
> +	char name[16];
...
> @@ -1924,7 +1925,9 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
>  	efx->interrupt_mode = max(efx->type->max_interrupt_mode,
>  				  interrupt_mode);
>  
> -	efx->workqueue = create_singlethread_workqueue("sfc_work");
> +	/* Would be good to use the net_dev name, but we're too early */
> +	snprintf(name, sizeof(name), "sfc%s", pci_name(pci_dev));
> +	efx->workqueue = create_singlethread_workqueue(name);
>  	if (!efx->workqueue)

This change is buggy.

create_singlethread_workqueue() is going to reference this
name buffer on the stack for the life of the workqueue, but
once this function returns that reference will no longer
be valid.

You'll need to use kmalloc()'d memory and free it later,
or something like that.

Patch not applied.

Re: [PATCH 25/33] sfc: Generate unique names for per-NIC workqueues

[Ben Hutchings]

David Miller wrote:
> From: Ben Hutchings <bhutchings@solarflare.com>
> Date: Fri, 12 Dec 2008 12:56:57 +0000
> 
> > @@ -1854,6 +1854,7 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
> >  	struct efx_channel *channel;
> >  	struct efx_tx_queue *tx_queue;
> >  	struct efx_rx_queue *rx_queue;
> > +	char name[16];
> ...
> > @@ -1924,7 +1925,9 @@ static int efx_init_struct(struct efx_nic *efx, struct efx_nic_type *type,
> >  	efx->interrupt_mode = max(efx->type->max_interrupt_mode,
> >  				  interrupt_mode);
> >  
> > -	efx->workqueue = create_singlethread_workqueue("sfc_work");
> > +	/* Would be good to use the net_dev name, but we're too early */
> > +	snprintf(name, sizeof(name), "sfc%s", pci_name(pci_dev));
> > +	efx->workqueue = create_singlethread_workqueue(name);
> >  	if (!efx->workqueue)
> 
> This change is buggy.
> 
> create_singlethread_workqueue() is going to reference this
> name buffer on the stack for the life of the workqueue, but
> once this function returns that reference will no longer
> be valid.
 
No, it's copied into the "comm" field of task_struct.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 25/33] sfc: Generate unique names for per-NIC workqueues

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Sat, 13 Dec 2008 06:23:03 +0000

> David Miller wrote:
> > create_singlethread_workqueue() is going to reference this
> > name buffer on the stack for the life of the workqueue, but
> > once this function returns that reference will no longer
> > be valid.
>  
> No, it's copied into the "comm" field of task_struct.

Hmmm...

create_singlethread_workqueue(name...
	--> __create_workqueue(name...
	--> __create_workqueue_key(name...

which goes:
	wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
 ...
	wq->name = name;

Which looks like a dynamic memory assumption to me.

Re: [PATCH 25/33] sfc: Generate unique names for per-NIC workqueues

[Ben Hutchings]

David Miller wrote:
> From: Ben Hutchings <bhutchings@solarflare.com>
> Date: Sat, 13 Dec 2008 06:23:03 +0000
> 
> > David Miller wrote:
> > > create_singlethread_workqueue() is going to reference this
> > > name buffer on the stack for the life of the workqueue, but
> > > once this function returns that reference will no longer
> > > be valid.
> >  
> > No, it's copied into the "comm" field of task_struct.
> 
> Hmmm...
> 
> create_singlethread_workqueue(name...
> 	--> __create_workqueue(name...
> 	--> __create_workqueue_key(name...
> 
> which goes:
> 	wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
>  ...
> 	wq->name = name;
> 
> Which looks like a dynamic memory assumption to me.

If you look carefully you'll see that wq->name is never used after
initialisation of single-threaded workqueues.  But I'd agree it's not
nice to have that dangling pointer around, so I'll respin this.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

[PATCH 26/33] sfc: Remove leading spaces

[Ben Hutchings]

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 2c53ef4..8107460 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -2940,10 +2940,10 @@ int falcon_probe_nic(struct efx_nic *efx)
 		goto fail5;
 
 	/* Initialise I2C adapter */
- 	efx->i2c_adap.owner = THIS_MODULE;
+	efx->i2c_adap.owner = THIS_MODULE;
 	nic_data->i2c_data = falcon_i2c_bit_operations;
 	nic_data->i2c_data.data = efx;
- 	efx->i2c_adap.algo_data = &nic_data->i2c_data;
+	efx->i2c_adap.algo_data = &nic_data->i2c_data;
 	efx->i2c_adap.dev.parent = &efx->pci_dev->dev;
 	strlcpy(efx->i2c_adap.name, "SFC4000 GPIO", sizeof(efx->i2c_adap.name));
 	rc = i2c_bit_add_bus(&efx->i2c_adap);
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 26/33] sfc: Remove leading spaces

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:57:18 +0000

> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 27/33] sfc: Specify a meaningful component for loopback RX-side and PHY tests

[Ben Hutchings]

Our ethtool self-test result names each begin with a component name.  For
some results this is "port0", which is not very meaningful.  Change that
to "rx" or "phy" as appropriate.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/ethtool.c |   14 ++++++++------
 1 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index aad0d1b..eab3e5c 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -274,8 +274,11 @@ static void efx_fill_test(unsigned int test_index,
 
 	/* Fill string, if applicable */
 	if (strings) {
-		snprintf(unit_str.name, sizeof(unit_str.name),
-			 unit_format, unit_id);
+		if (strchr(unit_format, '%'))
+			snprintf(unit_str.name, sizeof(unit_str.name),
+				 unit_format, unit_id);
+		else
+			strcpy(unit_str.name, unit_format);
 		snprintf(test_str.name, sizeof(test_str.name),
 			 test_format, test_id);
 		snprintf(strings[test_index].name,
@@ -284,7 +287,6 @@ static void efx_fill_test(unsigned int test_index,
 	}
 }
 
-#define EFX_PORT_NAME "port%d", 0
 #define EFX_CHANNEL_NAME(_channel) "chan%d", _channel->channel
 #define EFX_TX_QUEUE_NAME(_tx_queue) "txq%d", _tx_queue->queue
 #define EFX_RX_QUEUE_NAME(_rx_queue) "rxq%d", _rx_queue->queue
@@ -320,11 +322,11 @@ static int efx_fill_loopback_test(struct efx_nic *efx,
 	}
 	efx_fill_test(test_index++, strings, data,
 		      &lb_tests->rx_good,
-		      EFX_PORT_NAME,
+		      "rx", 0,
 		      EFX_LOOPBACK_NAME(mode, "rx_good"));
 	efx_fill_test(test_index++, strings, data,
 		      &lb_tests->rx_bad,
-		      EFX_PORT_NAME,
+		      "rx", 0,
 		      EFX_LOOPBACK_NAME(mode, "rx_bad"));
 
 	return test_index;
@@ -372,7 +374,7 @@ static int efx_ethtool_fill_self_tests(struct efx_nic *efx,
 	efx_fill_test(n++, strings, data, &tests->registers,
 		      "core", 0, "registers", NULL);
 	efx_fill_test(n++, strings, data, &tests->phy,
-		      EFX_PORT_NAME, "phy", NULL);
+		      "phy", 0, "bist", NULL);
 
 	/* Loopback tests */
 	for (mode = LOOPBACK_NONE; mode <= LOOPBACK_TEST_MAX; mode++) {
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 27/33] sfc: Specify a meaningful component for loopback RX-side and PHY tests

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:57:34 +0000

> Our ethtool self-test result names each begin with a component name.  For
> some results this is "port0", which is not very meaningful.  Change that
> to "rx" or "phy" as appropriate.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 28/33] sfc: Use mutex_lock_interruptible() for ethtool EEPROM access

[Ben Hutchings]

ethtool must contend with the MTD driver for the SPI bus lock, which
may carry out long operations such as flash erase.  Allow it to be
interrupted while waiting.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/ethtool.c |   10 ++++++++--
 1 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index eab3e5c..3aaece6 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -568,10 +568,13 @@ static int efx_ethtool_get_eeprom(struct net_device *net_dev,
 	size_t len;
 	int rc;
 
-	mutex_lock(&efx->spi_lock);
+	rc = mutex_lock_interruptible(&efx->spi_lock);
+	if (rc)
+		return rc;
 	rc = falcon_spi_read(spi, eeprom->offset + EFX_EEPROM_BOOTCONFIG_START,
 			     eeprom->len, &len, buf);
 	mutex_unlock(&efx->spi_lock);
+
 	eeprom->magic = EFX_ETHTOOL_EEPROM_MAGIC;
 	eeprom->len = len;
 	return rc;
@@ -588,10 +591,13 @@ static int efx_ethtool_set_eeprom(struct net_device *net_dev,
 	if (eeprom->magic != EFX_ETHTOOL_EEPROM_MAGIC)
 		return -EINVAL;
 
-	mutex_lock(&efx->spi_lock);
+	rc = mutex_lock_interruptible(&efx->spi_lock);
+	if (rc)
+		return rc;
 	rc = falcon_spi_write(spi, eeprom->offset + EFX_EEPROM_BOOTCONFIG_START,
 			      eeprom->len, &len, buf);
 	mutex_unlock(&efx->spi_lock);
+
 	eeprom->len = len;
 	return rc;
 }
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 28/33] sfc: Use mutex_lock_interruptible() for ethtool EEPROM access

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:59:23 +0000

> ethtool must contend with the MTD driver for the SPI bus lock, which
> may carry out long operations such as flash erase.  Allow it to be
> interrupted while waiting.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 29/33] sfc: Use model numbers for PHY type names

[Ben Hutchings]

Some of the PHY type names are overly generic.  Change them to include
the model numbers of the PHYs they represent.

Correct the model number reference at the top of xfp_phy.c.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c     |    2 +-
 drivers/net/sfc/net_driver.h |    6 +++---
 drivers/net/sfc/xfp_phy.c    |    2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 8107460..63414ce 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -2263,7 +2263,7 @@ static int falcon_probe_phy(struct efx_nic *efx)
 	case PHY_TYPE_SFT9001B:
 		efx->phy_op = &falcon_sft9001_phy_ops;
 		break;
-	case PHY_TYPE_XFP:
+	case PHY_TYPE_QT2022C2:
 		efx->phy_op = &falcon_xfp_phy_ops;
 		break;
 	default:
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index 1728329..7033dba 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -453,10 +453,10 @@ enum efx_int_mode {
 
 enum phy_type {
 	PHY_TYPE_NONE = 0,
-	PHY_TYPE_CX4_RTMR = 1,
-	PHY_TYPE_1G_ALASKA = 2,
+	PHY_TYPE_TXC43128 = 1,
+	PHY_TYPE_88E1111 = 2,
 	PHY_TYPE_SFX7101 = 3,
-	PHY_TYPE_XFP = 4,
+	PHY_TYPE_QT2022C2 = 4,
 	PHY_TYPE_PM8358 = 6,
 	PHY_TYPE_SFT9001A = 8,
 	PHY_TYPE_SFT9001B = 10,
diff --git a/drivers/net/sfc/xfp_phy.c b/drivers/net/sfc/xfp_phy.c
index 345ffc3..2d50b6e 100644
--- a/drivers/net/sfc/xfp_phy.c
+++ b/drivers/net/sfc/xfp_phy.c
@@ -7,7 +7,7 @@
  * by the Free Software Foundation, incorporated herein by reference.
  */
 /*
- * Driver for XFP optical PHYs (plus some support specific to the Quake 2032)
+ * Driver for XFP optical PHYs (plus some support specific to the Quake 2022/32)
  * See www.amcc.com for details (search for qt2032)
  */
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 29/33] sfc: Use model numbers for PHY type names

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:59:44 +0000

> Some of the PHY type names are overly generic.  Change them to include
> the model numbers of the PHYs they represent.
> 
> Correct the model number reference at the top of xfp_phy.c.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 30/33] sfc: Treat probe as unsuccessful if it scheduled a reset

[Ben Hutchings]

From: Steve Hodgson <shodgson@solarflare.com>

efx_pci_probe_main() can return success despite a reset being scheduled.
Catch this and retry or abort probe depending on the reset type.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |   13 +++++++++++--
 1 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index c094845..0d57c6d 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -2115,8 +2115,6 @@ static int __devinit efx_pci_probe(struct pci_dev *pci_dev,
 	 * we're in STATE_INIT. */
 	for (i = 0; i < 5; i++) {
 		rc = efx_pci_probe_main(efx);
-		if (rc == 0)
-			break;
 
 		/* Serialise against efx_reset(). No more resets will be
 		 * scheduled since efx_stop_all() has been called, and we
@@ -2124,6 +2122,17 @@ static int __devinit efx_pci_probe(struct pci_dev *pci_dev,
 		 * the rtnetlink or driverlink layers. */
 		cancel_work_sync(&efx->reset_work);
 
+		if (rc == 0) {
+			if (efx->reset_pending != RESET_TYPE_NONE) {
+				/* If there was a scheduled reset during
+				 * probe, the NIC is probably hosed anyway */
+				efx_pci_remove_main(efx);
+				rc = -EIO;
+			} else {
+				break;
+			}
+		}
+
 		/* Retry if a recoverably reset event has been scheduled */
 		if ((efx->reset_pending != RESET_TYPE_INVISIBLE) &&
 		    (efx->reset_pending != RESET_TYPE_ALL))
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 30/33] sfc: Treat probe as unsuccessful if it scheduled a reset

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 13:00:12 +0000

> From: Steve Hodgson <shodgson@solarflare.com>
> 
> efx_pci_probe_main() can return success despite a reset being scheduled.
> Catch this and retry or abort probe depending on the reset type.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 31/33] sfc: Use kzalloc() to ensure struct efx_spi_device is fully initialised

[Ben Hutchings]

Currently the mtd field is not initialised early enough.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/falcon.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 63414ce..e09a01d 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -2701,7 +2701,7 @@ static int falcon_spi_device_init(struct efx_nic *efx,
 	struct efx_spi_device *spi_device;
 
 	if (device_type != 0) {
-		spi_device = kmalloc(sizeof(*spi_device), GFP_KERNEL);
+		spi_device = kzalloc(sizeof(*spi_device), GFP_KERNEL);
 		if (!spi_device)
 			return -ENOMEM;
 		spi_device->device_id = device_id;
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 31/33] sfc: Use kzalloc() to ensure struct efx_spi_device is fully initialised

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 13:00:30 +0000

> Currently the mtd field is not initialised early enough.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 10/33] sfc: Add option to use a separate channel for TX completions

[Ben Hutchings]

From: Neil Turton <nturton@solarflare.com>

In a bidirectional forwarding test, we find that the best performance
is achieved by sending the TX completion interrupts from one NIC to a
CPU which shares an L2 cache with RX completion interrupts from the
other NIC.  To facilitate this, add an option (through a module
parameter) to create separate channels for RX and TX completion with
separate IRQs when MSI-X is available.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c        |   41 ++++++++++++++++++++++++++---------------
 drivers/net/sfc/net_driver.h |    2 ++
 2 files changed, 28 insertions(+), 15 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 15ae2fc..58cbbd5 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -64,13 +64,15 @@ MODULE_PARM_DESC(lro, "Large receive offload acceleration");
 /*
  * Use separate channels for TX and RX events
  *
- * Set this to 1 to use separate channels for TX and RX. It allows us to
- * apply a higher level of interrupt moderation to TX events.
+ * Set this to 1 to use separate channels for TX and RX. It allows us
+ * to control interrupt affinity separately for TX and RX.
  *
- * This is forced to 0 for MSI interrupt mode as the interrupt vector
- * is not written
+ * This is only used in MSI-X interrupt mode
  */
-static unsigned int separate_tx_and_rx_channels = true;
+static unsigned int separate_tx_channels;
+module_param(separate_tx_channels, uint, 0644);
+MODULE_PARM_DESC(separate_tx_channels,
+		 "Use separate channels for TX and RX");
 
 /* This is the weight assigned to each of the (per-channel) virtual
  * NAPI devices.
@@ -846,26 +848,33 @@ static void efx_probe_interrupts(struct efx_nic *efx)
 	if (efx->interrupt_mode == EFX_INT_MODE_MSIX) {
 		struct msix_entry xentries[EFX_MAX_CHANNELS];
 		int wanted_ints;
+		int rx_queues;
 
 		/* We want one RX queue and interrupt per CPU package
 		 * (or as specified by the rss_cpus module parameter).
 		 * We will need one channel per interrupt.
 		 */
-		wanted_ints = rss_cpus ? rss_cpus : efx_wanted_rx_queues();
-		efx->n_rx_queues = min(wanted_ints, max_channels);
+		rx_queues = rss_cpus ? rss_cpus : efx_wanted_rx_queues();
+		wanted_ints = rx_queues + (separate_tx_channels ? 1 : 0);
+		wanted_ints = min(wanted_ints, max_channels);
 
-		for (i = 0; i < efx->n_rx_queues; i++)
+		for (i = 0; i < wanted_ints; i++)
 			xentries[i].entry = i;
-		rc = pci_enable_msix(efx->pci_dev, xentries, efx->n_rx_queues);
+		rc = pci_enable_msix(efx->pci_dev, xentries, wanted_ints);
 		if (rc > 0) {
-			EFX_BUG_ON_PARANOID(rc >= efx->n_rx_queues);
-			efx->n_rx_queues = rc;
+			EFX_ERR(efx, "WARNING: Insufficient MSI-X vectors"
+				" available (%d < %d).\n", wanted_ints, rc);
+			EFX_ERR(efx, "WARNING: Performance may be reduced.\n");
+			EFX_BUG_ON_PARANOID(rc >= wanted_ints);
+			wanted_ints = rc;
 			rc = pci_enable_msix(efx->pci_dev, xentries,
-					     efx->n_rx_queues);
+					     wanted_ints);
 		}
 
 		if (rc == 0) {
-			for (i = 0; i < efx->n_rx_queues; i++)
+			efx->n_rx_queues = min(rx_queues, wanted_ints);
+			efx->n_channels = wanted_ints;
+			for (i = 0; i < wanted_ints; i++)
 				efx->channel[i].irq = xentries[i].vector;
 		} else {
 			/* Fall back to single channel MSI */
@@ -877,6 +886,7 @@ static void efx_probe_interrupts(struct efx_nic *efx)
 	/* Try single interrupt MSI */
 	if (efx->interrupt_mode == EFX_INT_MODE_MSI) {
 		efx->n_rx_queues = 1;
+		efx->n_channels = 1;
 		rc = pci_enable_msi(efx->pci_dev);
 		if (rc == 0) {
 			efx->channel[0].irq = efx->pci_dev->irq;
@@ -889,6 +899,7 @@ static void efx_probe_interrupts(struct efx_nic *efx)
 	/* Assume legacy interrupts */
 	if (efx->interrupt_mode == EFX_INT_MODE_LEGACY) {
 		efx->n_rx_queues = 1;
+		efx->n_channels = 1 + (separate_tx_channels ? 1 : 0);
 		efx->legacy_irq = efx->pci_dev->irq;
 	}
 }
@@ -913,8 +924,8 @@ static void efx_set_channels(struct efx_nic *efx)
 	struct efx_rx_queue *rx_queue;
 
 	efx_for_each_tx_queue(tx_queue, efx) {
-		if (!EFX_INT_MODE_USE_MSI(efx) && separate_tx_and_rx_channels)
-			tx_queue->channel = &efx->channel[1];
+		if (separate_tx_channels)
+			tx_queue->channel = &efx->channel[efx->n_channels-1];
 		else
 			tx_queue->channel = &efx->channel[0];
 		tx_queue->channel->used_flags |= EFX_USED_BY_TX;
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index f910609..2c5b5fa 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -649,6 +649,7 @@ union efx_multicast_hash {
  * @rx_queue: RX DMA queues
  * @channel: Channels
  * @n_rx_queues: Number of RX queues
+ * @n_channels: Number of channels in use
  * @rx_buffer_len: RX buffer length
  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
  * @irq_status: Interrupt status buffer
@@ -728,6 +729,7 @@ struct efx_nic {
 	struct efx_channel channel[EFX_MAX_CHANNELS];
 
 	int n_rx_queues;
+	int n_channels;
 	unsigned int rx_buffer_len;
 	unsigned int rx_buffer_order;
 
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 10/33] sfc: Add option to use a separate channel for TX completions

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:52:16 +0000

> From: Neil Turton <nturton@solarflare.com>
> 
> In a bidirectional forwarding test, we find that the best performance
> is achieved by sending the TX completion interrupts from one NIC to a
> CPU which shares an L2 cache with RX completion interrupts from the
> other NIC.  To facilitate this, add an option (through a module
> parameter) to create separate channels for RX and TX completion with
> separate IRQs when MSI-X is available.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 11/33] sfc: Provide hints to irqbalance daemon

[Ben Hutchings]

Allocate IRQs with the name format <device>[-<type>]-<number> so that
future versions of irqbalanced understand what we're doing.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c        |   24 ++++++++++++++++++++++++
 drivers/net/sfc/falcon.c     |    2 +-
 drivers/net/sfc/net_driver.h |    2 ++
 3 files changed, 27 insertions(+), 1 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 58cbbd5..183a440 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -356,6 +356,27 @@ static int efx_probe_channel(struct efx_channel *channel)
 }
 
 
+static void efx_set_channel_names(struct efx_nic *efx)
+{
+	struct efx_channel *channel;
+	const char *type = "";
+	int number;
+
+	efx_for_each_channel(channel, efx) {
+		number = channel->channel;
+		if (efx->n_channels > efx->n_rx_queues) {
+			if (channel->channel < efx->n_rx_queues) {
+				type = "-rx";
+			} else {
+				type = "-tx";
+				number -= efx->n_rx_queues;
+			}
+		}
+		snprintf(channel->name, sizeof(channel->name),
+			 "%s%s-%d", efx->name, type, number);
+	}
+}
+
 /* Channels are shutdown and reinitialised whilst the NIC is running
  * to propagate configuration changes (mtu, checksum offload), or
  * to clear hardware error conditions
@@ -1002,6 +1023,7 @@ static int efx_probe_all(struct efx_nic *efx)
 			goto fail3;
 		}
 	}
+	efx_set_channel_names(efx);
 
 	return 0;
 
@@ -1483,6 +1505,7 @@ static int efx_netdev_event(struct notifier_block *this,
 
 		strcpy(efx->name, net_dev->name);
 		efx_mtd_rename(efx);
+		efx_set_channel_names(efx);
 	}
 
 	return NOTIFY_DONE;
@@ -1516,6 +1539,7 @@ static int efx_register_netdev(struct efx_nic *efx)
 		return rc;
 	}
 	strcpy(efx->name, net_dev->name);
+	efx_set_channel_names(efx);
 
 	return 0;
 }
diff --git a/drivers/net/sfc/falcon.c b/drivers/net/sfc/falcon.c
index 9563c13..7421945 100644
--- a/drivers/net/sfc/falcon.c
+++ b/drivers/net/sfc/falcon.c
@@ -1585,7 +1585,7 @@ int falcon_init_interrupt(struct efx_nic *efx)
 	efx_for_each_channel(channel, efx) {
 		rc = request_irq(channel->irq, falcon_msi_interrupt,
 				 IRQF_PROBE_SHARED, /* Not shared */
-				 efx->name, channel);
+				 channel->name, channel);
 		if (rc) {
 			EFX_ERR(efx, "failed to hook IRQ %d\n", channel->irq);
 			goto fail2;
diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index 2c5b5fa..abff908 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -327,6 +327,7 @@ enum efx_rx_alloc_method {
  *
  * @efx: Associated Efx NIC
  * @channel: Channel instance number
+ * @name: Name for channel and IRQ
  * @used_flags: Channel is used by net driver
  * @enabled: Channel enabled indicator
  * @irq: IRQ number (MSI and MSI-X only)
@@ -357,6 +358,7 @@ enum efx_rx_alloc_method {
 struct efx_channel {
 	struct efx_nic *efx;
 	int channel;
+	char name[IFNAMSIZ + 6];
 	int used_flags;
 	bool enabled;
 	int irq;
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 11/33] sfc: Provide hints to irqbalance daemon

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:52:30 +0000

> Allocate IRQs with the name format <device>[-<type>]-<number> so that
> future versions of irqbalanced understand what we're doing.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied, thanks for doing this Ben.

[PATCH 12/33] sfc: Abbreviate self-test names so they are not truncated

[Ben Hutchings]

Change "channel" to "chan".

Shorten PHY loopback names.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/ethtool.c |   16 ++++++++--------
 1 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/drivers/net/sfc/ethtool.c b/drivers/net/sfc/ethtool.c
index 9c533e0..931ce14 100644
--- a/drivers/net/sfc/ethtool.c
+++ b/drivers/net/sfc/ethtool.c
@@ -27,9 +27,9 @@ const char *efx_loopback_mode_names[] = {
 	[LOOPBACK_XGXS]		= "XGXS",
 	[LOOPBACK_XAUI] 	= "XAUI",
 	[LOOPBACK_PHY]		= "PHY",
-	[LOOPBACK_PHYXS]	= "PHY(XS)",
-	[LOOPBACK_PCS]	 	= "PHY(PCS)",
-	[LOOPBACK_PMAPMD]	= "PHY(PMAPMD)",
+	[LOOPBACK_PHYXS]	= "PHYXS",
+	[LOOPBACK_PCS]	 	= "PCS",
+	[LOOPBACK_PMAPMD]	= "PMA/PMD",
 	[LOOPBACK_NETWORK]	= "NETWORK",
 };
 
@@ -242,10 +242,10 @@ static void efx_ethtool_get_drvinfo(struct net_device *net_dev,
  * @strings:		Ethtool strings, or %NULL
  * @data:		Ethtool test results, or %NULL
  * @test:		Pointer to test result (used only if data != %NULL)
- * @unit_format:	Unit name format (e.g. "channel\%d")
- * @unit_id:		Unit id (e.g. 0 for "channel0")
+ * @unit_format:	Unit name format (e.g. "chan\%d")
+ * @unit_id:		Unit id (e.g. 0 for "chan0")
  * @test_format:	Test name format (e.g. "loopback.\%s.tx.sent")
- * @test_id:		Test id (e.g. "PHY" for "loopback.PHY.tx_sent")
+ * @test_id:		Test id (e.g. "PHYXS" for "loopback.PHYXS.tx_sent")
  *
  * Fill in an individual self-test entry.
  */
@@ -268,12 +268,12 @@ static void efx_fill_test(unsigned int test_index,
 			 test_format, test_id);
 		snprintf(strings[test_index].name,
 			 sizeof(strings[test_index].name),
-			 "%-9s%-17s", unit_str.name, test_str.name);
+			 "%-6s %-24s", unit_str.name, test_str.name);
 	}
 }
 
 #define EFX_PORT_NAME "port%d", 0
-#define EFX_CHANNEL_NAME(_channel) "channel%d", _channel->channel
+#define EFX_CHANNEL_NAME(_channel) "chan%d", _channel->channel
 #define EFX_TX_QUEUE_NAME(_tx_queue) "txq%d", _tx_queue->queue
 #define EFX_RX_QUEUE_NAME(_rx_queue) "rxq%d", _rx_queue->queue
 #define EFX_LOOPBACK_NAME(_mode, _counter)			\
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 12/33] sfc: Abbreviate self-test names so they are not truncated

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 12:52:45 +0000

> Change "channel" to "chan".
> 
> Shorten PHY loopback names.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 24/33] sfc: Fix format arguments for warning about MSI-X allocation

[Ben Hutchings]

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index b99ccfb..ce1c7d3 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -899,7 +899,7 @@ static void efx_probe_interrupts(struct efx_nic *efx)
 		rc = pci_enable_msix(efx->pci_dev, xentries, wanted_ints);
 		if (rc > 0) {
 			EFX_ERR(efx, "WARNING: Insufficient MSI-X vectors"
-				" available (%d < %d).\n", wanted_ints, rc);
+				" available (%d < %d).\n", rc, wanted_ints);
 			EFX_ERR(efx, "WARNING: Performance may be reduced.\n");
 			EFX_BUG_ON_PARANOID(rc >= wanted_ints);
 			wanted_ints = rc;
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 24/33] sfc: Fix format arguments for warning about MSI-X allocation

[Ben Hutchings]

On Fri, 2008-12-12 at 12:56 +0000, Ben Hutchings wrote:
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
> ---
>  drivers/net/sfc/efx.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
> index b99ccfb..ce1c7d3 100644
> --- a/drivers/net/sfc/efx.c
> +++ b/drivers/net/sfc/efx.c
> @@ -899,7 +899,7 @@ static void efx_probe_interrupts(struct efx_nic *efx)
>  		rc = pci_enable_msix(efx->pci_dev, xentries, wanted_ints);
>  		if (rc > 0) {
>  			EFX_ERR(efx, "WARNING: Insufficient MSI-X vectors"
> -				" available (%d < %d).\n", wanted_ints, rc);
> +				" available (%d < %d).\n", rc, wanted_ints);
>  			EFX_ERR(efx, "WARNING: Performance may be reduced.\n");
>  			EFX_BUG_ON_PARANOID(rc >= wanted_ints);
>  			wanted_ints = rc;

I thought this warning was already present, but it is actually
introduced by patch 10/33.  There's no point in introducing it wrongly,
so please combine the two.  Or I can post the combined patch myself.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 24/33] sfc: Fix format arguments for warning about MSI-X allocation

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 20:34:47 +0000

> I thought this warning was already present, but it is actually
> introduced by patch 10/33.  There's no point in introducing it wrongly,
> so please combine the two.  Or I can post the combined patch myself.

I'll take care of this, thanks Ben.

[PATCH 32/33] sfc: Fix synchronisation of efx_mtd_{probe,rename,remove}

[Ben Hutchings]

Currently efx_mtd_rename() can race with the probe() and remove()
functions.

Move probe() before device registration and remove() after
unregistration.  Move initialisation/update of all names based on the
netdev name into a new function and call it under the RTNL immediately
after registration.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/efx.c |   33 ++++++++++++++++++---------------
 1 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c
index 0d57c6d..8d9d0b4 100644
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -1517,18 +1517,21 @@ static const struct net_device_ops efx_netdev_ops = {
 #endif
 };
 
+static void efx_update_name(struct efx_nic *efx)
+{
+	strcpy(efx->name, efx->net_dev->name);
+	efx_mtd_rename(efx);
+	efx_set_channel_names(efx);
+}
+
 static int efx_netdev_event(struct notifier_block *this,
 			    unsigned long event, void *ptr)
 {
 	struct net_device *net_dev = ptr;
 
-	if (net_dev->netdev_ops == &efx_netdev_ops && event == NETDEV_CHANGENAME) {
-		struct efx_nic *efx = netdev_priv(net_dev);
-
-		strcpy(efx->name, net_dev->name);
-		efx_mtd_rename(efx);
-		efx_set_channel_names(efx);
-	}
+	if (net_dev->netdev_ops == &efx_netdev_ops &&
+	    event == NETDEV_CHANGENAME)
+		efx_update_name(netdev_priv(net_dev));
 
 	return NOTIFY_DONE;
 }
@@ -1568,8 +1571,10 @@ static int efx_register_netdev(struct efx_nic *efx)
 		EFX_ERR(efx, "could not register net dev\n");
 		return rc;
 	}
-	strcpy(efx->name, net_dev->name);
-	efx_set_channel_names(efx);
+
+	rtnl_lock();
+	efx_update_name(efx);
+	rtnl_unlock();
 
 	rc = device_create_file(&efx->pci_dev->dev, &dev_attr_phy_type);
 	if (rc) {
@@ -1981,8 +1986,6 @@ static void efx_pci_remove(struct pci_dev *pci_dev)
 	if (!efx)
 		return;
 
-	efx_mtd_remove(efx);
-
 	/* Mark the NIC as fini, then stop the interface */
 	rtnl_lock();
 	efx->state = STATE_FINI;
@@ -1996,6 +1999,8 @@ static void efx_pci_remove(struct pci_dev *pci_dev)
 
 	efx_unregister_netdev(efx);
 
+	efx_mtd_remove(efx);
+
 	/* Wait for any scheduled resets to complete. No more will be
 	 * scheduled from this point because efx_stop_all() has been
 	 * called, we are no longer registered with driverlink, and
@@ -2149,17 +2154,15 @@ static int __devinit efx_pci_probe(struct pci_dev *pci_dev,
 	/* Switch to the running state before we expose the device to
 	 * the OS.  This is to ensure that the initial gathering of
 	 * MAC stats succeeds. */
-	rtnl_lock();
 	efx->state = STATE_RUNNING;
-	rtnl_unlock();
+
+	efx_mtd_probe(efx); /* allowed to fail */
 
 	rc = efx_register_netdev(efx);
 	if (rc)
 		goto fail5;
 
 	EFX_LOG(efx, "initialisation successful\n");
-
-	efx_mtd_probe(efx); /* allowed to fail */
 	return 0;
 
  fail5:
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 32/33] sfc: Fix synchronisation of efx_mtd_{probe,rename,remove}

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 13:00:51 +0000

> Currently efx_mtd_rename() can race with the probe() and remove()
> functions.
> 
> Move probe() before device registration and remove() after
> unregistration.  Move initialisation/update of all names based on the
> netdev name into a new function and call it under the RTNL immediately
> after registration.
> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Applied.

[PATCH 33/33] sfc: Version 2.3

[Ben Hutchings]

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
 drivers/net/sfc/net_driver.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/net/sfc/net_driver.h b/drivers/net/sfc/net_driver.h
index 7033dba..03feaee 100644
--- a/drivers/net/sfc/net_driver.h
+++ b/drivers/net/sfc/net_driver.h
@@ -42,7 +42,7 @@
 #ifndef EFX_DRIVER_NAME
 #define EFX_DRIVER_NAME	"sfc"
 #endif
-#define EFX_DRIVER_VERSION	"2.2"
+#define EFX_DRIVER_VERSION	"2.3"
 
 #ifdef EFX_ENABLE_DEBUG
 #define EFX_BUG_ON_PARANOID(x) BUG_ON(x)
-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: [PATCH 33/33] sfc: Version 2.3

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Fri, 12 Dec 2008 13:01:06 +0000

> 
> Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

Also applied.