8 Most network devices consist of set of registers which provide an interface
9 to a MAC layer, which communicates with the physical connection through a
10 PHY. The PHY concerns itself with negotiating link parameters with the link
11 partner on the other side of the network connection (typically, an ethernet
12 cable), and provides a register interface to allow drivers to determine what
13 settings were chosen, and to configure what settings are allowed.
15 While these devices are distinct from the network devices, and conform to a
16 standard layout for the registers, it has been common practice to integrate
17 the PHY management code with the network driver. This has resulted in large
18 amounts of redundant code. Also, on embedded systems with multiple (and
19 sometimes quite different) ethernet controllers connected to the same
20 management bus, it is difficult to ensure safe use of the bus.
22 Since the PHYs are devices, and the management busses through which they are
23 accessed are, in fact, busses, the PHY Abstraction Layer treats them as such.
24 In doing so, it has these goals:
26 #. Increase code-reuse
27 #. Increase overall code-maintainability
28 #. Speed development time for new network drivers, and for new systems
30 Basically, this layer is meant to provide an interface to PHY devices which
31 allows network driver writers to write as little code as possible, while
32 still providing a full feature set.
37 Most network devices are connected to a PHY by means of a management bus.
38 Different devices use different busses (though some share common interfaces).
39 In order to take advantage of the PAL, each bus interface needs to be
40 registered as a distinct device.
42 #. read and write functions must be implemented. Their prototypes are::
44 int write(struct mii_bus *bus, int mii_id, int regnum, u16 value);
45 int read(struct mii_bus *bus, int mii_id, int regnum);
47 mii_id is the address on the bus for the PHY, and regnum is the register
48 number. These functions are guaranteed not to be called from interrupt
49 time, so it is safe for them to block, waiting for an interrupt to signal
50 the operation is complete
52 #. A reset function is optional. This is used to return the bus to an
55 #. A probe function is needed. This function should set up anything the bus
56 driver needs, setup the mii_bus structure, and register with the PAL using
57 mdiobus_register. Similarly, there's a remove function to undo all of
58 that (use mdiobus_unregister).
60 #. Like any driver, the device_driver structure must be configured, and init
61 exit functions are used to register the driver.
63 #. The bus must also be declared somewhere as a device, and registered.
65 As an example for how one driver implemented an mdio bus driver, see
66 drivers/net/ethernet/freescale/fsl_pq_mdio.c and an associated DTS file
67 for one of the users. (e.g. "git grep fsl,.*-mdio arch/powerpc/boot/dts/")
69 (RG)MII/electrical interface considerations
70 ===========================================
72 The Reduced Gigabit Medium Independent Interface (RGMII) is a 12-pin
73 electrical signal interface using a synchronous 125Mhz clock signal and several
74 data lines. Due to this design decision, a 1.5ns to 2ns delay must be added
75 between the clock line (RXC or TXC) and the data lines to let the PHY (clock
76 sink) have a large enough setup and hold time to sample the data lines correctly. The
77 PHY library offers different types of PHY_INTERFACE_MODE_RGMII* values to let
78 the PHY driver and optionally the MAC driver, implement the required delay. The
79 values of phy_interface_t must be understood from the perspective of the PHY
80 device itself, leading to the following:
82 * PHY_INTERFACE_MODE_RGMII: the PHY is not responsible for inserting any
83 internal delay by itself, it assumes that either the Ethernet MAC (if capable
84 or the PCB traces) insert the correct 1.5-2ns delay
86 * PHY_INTERFACE_MODE_RGMII_TXID: the PHY should insert an internal delay
87 for the transmit data lines (TXD[3:0]) processed by the PHY device
89 * PHY_INTERFACE_MODE_RGMII_RXID: the PHY should insert an internal delay
90 for the receive data lines (RXD[3:0]) processed by the PHY device
92 * PHY_INTERFACE_MODE_RGMII_ID: the PHY should insert internal delays for
93 both transmit AND receive data lines from/to the PHY device
95 Whenever possible, use the PHY side RGMII delay for these reasons:
97 * PHY devices may offer sub-nanosecond granularity in how they allow a
98 receiver/transmitter side delay (e.g: 0.5, 1.0, 1.5ns) to be specified. Such
99 precision may be required to account for differences in PCB trace lengths
101 * PHY devices are typically qualified for a large range of applications
102 (industrial, medical, automotive...), and they provide a constant and
103 reliable delay across temperature/pressure/voltage ranges
105 * PHY device drivers in PHYLIB being reusable by nature, being able to
106 configure correctly a specified delay enables more designs with similar delay
107 requirements to be operate correctly
109 For cases where the PHY is not capable of providing this delay, but the
110 Ethernet MAC driver is capable of doing so, the correct phy_interface_t value
111 should be PHY_INTERFACE_MODE_RGMII, and the Ethernet MAC driver should be
112 configured correctly in order to provide the required transmit and/or receive
113 side delay from the perspective of the PHY device. Conversely, if the Ethernet
114 MAC driver looks at the phy_interface_t value, for any other mode but
115 PHY_INTERFACE_MODE_RGMII, it should make sure that the MAC-level delays are
118 In case neither the Ethernet MAC, nor the PHY are capable of providing the
119 required delays, as defined per the RGMII standard, several options may be
122 * Some SoCs may offer a pin pad/mux/controller capable of configuring a given
123 set of pins'strength, delays, and voltage; and it may be a suitable
124 option to insert the expected 2ns RGMII delay.
126 * Modifying the PCB design to include a fixed delay (e.g: using a specifically
127 designed serpentine), which may not require software configuration at all.
129 Common problems with RGMII delay mismatch
130 -----------------------------------------
132 When there is a RGMII delay mismatch between the Ethernet MAC and the PHY, this
133 will most likely result in the clock and data line signals to be unstable when
134 the PHY or MAC take a snapshot of these signals to translate them into logical
135 1 or 0 states and reconstruct the data being transmitted/received. Typical
138 * Transmission/reception partially works, and there is frequent or occasional
141 * Ethernet MAC may report some or all packets ingressing with a FCS/CRC error,
142 or just discard them all
144 * Switching to lower speeds such as 10/100Mbits/sec makes the problem go away
145 (since there is enough setup/hold time in that case)
150 Sometime during startup, the network driver needs to establish a connection
151 between the PHY device, and the network device. At this time, the PHY's bus
152 and drivers need to all have been loaded, so it is ready for the connection.
153 At this point, there are several ways to connect to the PHY:
155 #. The PAL handles everything, and only calls the network driver when
156 the link state changes, so it can react.
158 #. The PAL handles everything except interrupts (usually because the
159 controller has the interrupt registers).
161 #. The PAL handles everything, but checks in with the driver every second,
162 allowing the network driver to react first to any changes before the PAL
165 #. The PAL serves only as a library of functions, with the network device
166 manually calling functions to update status, and configure the PHY
169 Letting the PHY Abstraction Layer do Everything
170 ===============================================
172 If you choose option 1 (The hope is that every driver can, but to still be
173 useful to drivers that can't), connecting to the PHY is simple:
175 First, you need a function to react to changes in the link state. This
176 function follows this protocol::
178 static void adjust_link(struct net_device *dev);
180 Next, you need to know the device name of the PHY connected to this device.
181 The name will look something like, "0:00", where the first number is the
182 bus id, and the second is the PHY's address on that bus. Typically,
183 the bus is responsible for making its ID unique.
185 Now, to connect, just call this function::
187 phydev = phy_connect(dev, phy_name, &adjust_link, interface);
189 *phydev* is a pointer to the phy_device structure which represents the PHY.
190 If phy_connect is successful, it will return the pointer. dev, here, is the
191 pointer to your net_device. Once done, this function will have started the
192 PHY's software state machine, and registered for the PHY's interrupt, if it
193 has one. The phydev structure will be populated with information about the
194 current state, though the PHY will not yet be truly operational at this
197 PHY-specific flags should be set in phydev->dev_flags prior to the call
198 to phy_connect() such that the underlying PHY driver can check for flags
199 and perform specific operations based on them.
200 This is useful if the system has put hardware restrictions on
201 the PHY/controller, of which the PHY needs to be aware.
203 *interface* is a u32 which specifies the connection type used
204 between the controller and the PHY. Examples are GMII, MII,
205 RGMII, and SGMII. See "PHY interface mode" below. For a full
206 list, see include/linux/phy.h
208 Now just make sure that phydev->supported and phydev->advertising have any
209 values pruned from them which don't make sense for your controller (a 10/100
210 controller may be connected to a gigabit capable PHY, so you would need to
211 mask off SUPPORTED_1000baseT*). See include/linux/ethtool.h for definitions
212 for these bitfields. Note that you should not SET any bits, except the
213 SUPPORTED_Pause and SUPPORTED_AsymPause bits (see below), or the PHY may get
214 put into an unsupported state.
216 Lastly, once the controller is ready to handle network traffic, you call
217 phy_start(phydev). This tells the PAL that you are ready, and configures the
218 PHY to connect to the network. If the MAC interrupt of your network driver
219 also handles PHY status changes, just set phydev->irq to PHY_IGNORE_INTERRUPT
220 before you call phy_start and use phy_mac_interrupt() from the network
221 driver. If you don't want to use interrupts, set phydev->irq to PHY_POLL.
222 phy_start() enables the PHY interrupts (if applicable) and starts the
223 phylib state machine.
225 When you want to disconnect from the network (even if just briefly), you call
226 phy_stop(phydev). This function also stops the phylib state machine and
227 disables PHY interrupts.
232 The PHY interface mode supplied in the phy_connect() family of functions
233 defines the initial operating mode of the PHY interface. This is not
234 guaranteed to remain constant; there are PHYs which dynamically change
235 their interface mode without software interaction depending on the
238 Some of the interface modes are described below:
240 ``PHY_INTERFACE_MODE_1000BASEX``
241 This defines the 1000BASE-X single-lane serdes link as defined by the
242 802.3 standard section 36. The link operates at a fixed bit rate of
243 1.25Gbaud using a 10B/8B encoding scheme, resulting in an underlying
244 data rate of 1Gbps. Embedded in the data stream is a 16-bit control
245 word which is used to negotiate the duplex and pause modes with the
246 remote end. This does not include "up-clocked" variants such as 2.5Gbps
249 ``PHY_INTERFACE_MODE_2500BASEX``
250 This defines a variant of 1000BASE-X which is clocked 2.5 times faster,
251 than the 802.3 standard giving a fixed bit rate of 3.125Gbaud.
253 ``PHY_INTERFACE_MODE_SGMII``
254 This is used for Cisco SGMII, which is a modification of 1000BASE-X
255 as defined by the 802.3 standard. The SGMII link consists of a single
256 serdes lane running at a fixed bit rate of 1.25Gbaud with 10B/8B
257 encoding. The underlying data rate is 1Gbps, with the slower speeds of
258 100Mbps and 10Mbps being achieved through replication of each data symbol.
259 The 802.3 control word is re-purposed to send the negotiated speed and
260 duplex information from to the MAC, and for the MAC to acknowledge
261 receipt. This does not include "up-clocked" variants such as 2.5Gbps
264 Note: mismatched SGMII vs 1000BASE-X configuration on a link can
265 successfully pass data in some circumstances, but the 16-bit control
266 word will not be correctly interpreted, which may cause mismatches in
267 duplex, pause or other settings. This is dependent on the MAC and/or
270 ``PHY_INTERFACE_MODE_10GBASER``
271 This is the IEEE 802.3 Clause 49 defined 10GBASE-R protocol used with
272 various different mediums. Please refer to the IEEE standard for a
275 Note: 10GBASE-R is just one protocol that can be used with XFI and SFI.
276 XFI and SFI permit multiple protocols over a single SERDES lane, and
277 also defines the electrical characteristics of the signals with a host
278 compliance board plugged into the host XFP/SFP connector. Therefore,
279 XFI and SFI are not PHY interface types in their own right.
281 ``PHY_INTERFACE_MODE_10GKR``
282 This is the IEEE 802.3 Clause 49 defined 10GBASE-R with Clause 73
283 autonegotiation. Please refer to the IEEE standard for further
286 Note: due to legacy usage, some 10GBASE-R usage incorrectly makes
287 use of this definition.
289 Pause frames / flow control
290 ===========================
292 The PHY does not participate directly in flow control/pause frames except by
293 making sure that the SUPPORTED_Pause and SUPPORTED_AsymPause bits are set in
294 MII_ADVERTISE to indicate towards the link partner that the Ethernet MAC
295 controller supports such a thing. Since flow control/pause frames generation
296 involves the Ethernet MAC driver, it is recommended that this driver takes care
297 of properly indicating advertisement and support for such features by setting
298 the SUPPORTED_Pause and SUPPORTED_AsymPause bits accordingly. This can be done
299 either before or after phy_connect() and/or as a result of implementing the
300 ethtool::set_pauseparam feature.
303 Keeping Close Tabs on the PAL
304 =============================
306 It is possible that the PAL's built-in state machine needs a little help to
307 keep your network device and the PHY properly in sync. If so, you can
308 register a helper function when connecting to the PHY, which will be called
309 every second before the state machine reacts to any changes. To do this, you
310 need to manually call phy_attach() and phy_prepare_link(), and then call
311 phy_start_machine() with the second argument set to point to your special
314 Currently there are no examples of how to use this functionality, and testing
315 on it has been limited because the author does not have any drivers which use
316 it (they all use option 1). So Caveat Emptor.
318 Doing it all yourself
319 =====================
321 There's a remote chance that the PAL's built-in state machine cannot track
322 the complex interactions between the PHY and your network device. If this is
323 so, you can simply call phy_attach(), and not call phy_start_machine or
324 phy_prepare_link(). This will mean that phydev->state is entirely yours to
325 handle (phy_start and phy_stop toggle between some of the states, so you
326 might need to avoid them).
328 An effort has been made to make sure that useful functionality can be
329 accessed without the state-machine running, and most of these functions are
330 descended from functions which did not interact with a complex state-machine.
331 However, again, no effort has been made so far to test running without the
332 state machine, so tryer beware.
334 Here is a brief rundown of the functions::
336 int phy_read(struct phy_device *phydev, u16 regnum);
337 int phy_write(struct phy_device *phydev, u16 regnum, u16 val);
339 Simple read/write primitives. They invoke the bus's read/write function
343 void phy_print_status(struct phy_device *phydev);
345 A convenience function to print out the PHY status neatly.
348 void phy_request_interrupt(struct phy_device *phydev);
350 Requests the IRQ for the PHY interrupts.
353 struct phy_device * phy_attach(struct net_device *dev, const char *phy_id,
354 phy_interface_t interface);
356 Attaches a network device to a particular PHY, binding the PHY to a generic
357 driver if none was found during bus initialization.
360 int phy_start_aneg(struct phy_device *phydev);
362 Using variables inside the phydev structure, either configures advertising
363 and resets autonegotiation, or disables autonegotiation, and configures
367 static inline int phy_read_status(struct phy_device *phydev);
369 Fills the phydev structure with up-to-date information about the current
373 int phy_ethtool_ksettings_set(struct phy_device *phydev,
374 const struct ethtool_link_ksettings *cmd);
376 Ethtool convenience functions.
379 int phy_mii_ioctl(struct phy_device *phydev,
380 struct mii_ioctl_data *mii_data, int cmd);
382 The MII ioctl. Note that this function will completely screw up the state
383 machine if you write registers like BMCR, BMSR, ADVERTISE, etc. Best to
384 use this only to write registers which are not standard, and don't set off
390 With the PHY Abstraction Layer, adding support for new PHYs is
391 quite easy. In some cases, no work is required at all! However,
392 many PHYs require a little hand-holding to get up-and-running.
397 If the desired PHY doesn't have any errata, quirks, or special
398 features you want to support, then it may be best to not add
399 support, and let the PHY Abstraction Layer's Generic PHY Driver
405 If you do need to write a PHY driver, the first thing to do is
406 make sure it can be matched with an appropriate PHY device.
407 This is done during bus initialization by reading the device's
408 UID (stored in registers 2 and 3), then comparing it to each
409 driver's phy_id field by ANDing it with each driver's
410 phy_id_mask field. Also, it needs a name. Here's an example::
412 static struct phy_driver dm9161_driver = {
413 .phy_id = 0x0181b880,
414 .name = "Davicom DM9161E",
415 .phy_id_mask = 0x0ffffff0,
419 Next, you need to specify what features (speed, duplex, autoneg,
420 etc) your PHY device and driver support. Most PHYs support
421 PHY_BASIC_FEATURES, but you can look in include/mii.h for other
424 Each driver consists of a number of function pointers, documented
425 in include/linux/phy.h under the phy_driver structure.
427 Of these, only config_aneg and read_status are required to be
428 assigned by the driver code. The rest are optional. Also, it is
429 preferred to use the generic phy driver's versions of these two
430 functions if at all possible: genphy_read_status and
431 genphy_config_aneg. If this is not possible, it is likely that
432 you only need to perform some actions before and after invoking
433 these functions, and so your functions will wrap the generic
436 Feel free to look at the Marvell, Cicada, and Davicom drivers in
437 drivers/net/phy/ for examples (the lxt and qsemi drivers have
438 not been tested as of this writing).
440 The PHY's MMD register accesses are handled by the PAL framework
441 by default, but can be overridden by a specific PHY driver if
442 required. This could be the case if a PHY was released for
443 manufacturing before the MMD PHY register definitions were
444 standardized by the IEEE. Most modern PHYs will be able to use
445 the generic PAL framework for accessing the PHY's MMD registers.
446 An example of such usage is for Energy Efficient Ethernet support,
447 implemented in the PAL. This support uses the PAL to access MMD
448 registers for EEE query and configuration if the PHY supports
449 the IEEE standard access mechanisms, or can use the PHY's specific
450 access interfaces if overridden by the specific PHY driver. See
451 the Micrel driver in drivers/net/phy/ for an example of how this
457 Sometimes the specific interaction between the platform and the PHY requires
458 special handling. For instance, to change where the PHY's clock input is,
459 or to add a delay to account for latency issues in the data path. In order
460 to support such contingencies, the PHY Layer allows platform code to register
461 fixups to be run when the PHY is brought up (or subsequently reset).
463 When the PHY Layer brings up a PHY it checks to see if there are any fixups
464 registered for it, matching based on UID (contained in the PHY device's phy_id
465 field) and the bus identifier (contained in phydev->dev.bus_id). Both must
466 match, however two constants, PHY_ANY_ID and PHY_ANY_UID, are provided as
467 wildcards for the bus ID and UID, respectively.
469 When a match is found, the PHY layer will invoke the run function associated
470 with the fixup. This function is passed a pointer to the phy_device of
471 interest. It should therefore only operate on that PHY.
473 The platform code can either register the fixup using phy_register_fixup()::
475 int phy_register_fixup(const char *phy_id,
476 u32 phy_uid, u32 phy_uid_mask,
477 int (*run)(struct phy_device *));
479 Or using one of the two stubs, phy_register_fixup_for_uid() and
480 phy_register_fixup_for_id()::
482 int phy_register_fixup_for_uid(u32 phy_uid, u32 phy_uid_mask,
483 int (*run)(struct phy_device *));
484 int phy_register_fixup_for_id(const char *phy_id,
485 int (*run)(struct phy_device *));
487 The stubs set one of the two matching criteria, and set the other one to
490 When phy_register_fixup() or \*_for_uid()/\*_for_id() is called at module load
491 time, the module needs to unregister the fixup and free allocated memory when
494 Call one of following function before unloading module::
496 int phy_unregister_fixup(const char *phy_id, u32 phy_uid, u32 phy_uid_mask);
497 int phy_unregister_fixup_for_uid(u32 phy_uid, u32 phy_uid_mask);
498 int phy_register_fixup_for_id(const char *phy_id);
503 IEEE Standard 802.3: CSMA/CD Access Method and Physical Layer Specifications, Section Two:
504 http://standards.ieee.org/getieee802/download/802.3-2008_section2.pdf
507 http://web.archive.org/web/20160303212629/http://www.hp.com/rnd/pdfs/RGMIIv1_3.pdf
510 http://web.archive.org/web/20160303171328/http://www.hp.com/rnd/pdfs/RGMIIv2_0_final_hp.pdf