Root/Documentation/networking/phy.txt

1
2-------
3PHY Abstraction Layer
4(Updated 2008-04-08)
5
6Purpose
7
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.
14
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.
21
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:
25
26   1) Increase code-reuse
27   2) Increase overall code-maintainability
28   3) Speed development time for new network drivers, and for new systems
29 
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.
33
34The MDIO bus
35
36 Most network devices are connected to a PHY by means of a management bus.
37 Different devices use different busses (though some share common interfaces).
38 In order to take advantage of the PAL, each bus interface needs to be
39 registered as a distinct device.
40
41 1) read and write functions must be implemented. Their prototypes are:
42
43     int write(struct mii_bus *bus, int mii_id, int regnum, u16 value);
44     int read(struct mii_bus *bus, int mii_id, int regnum);
45
46   mii_id is the address on the bus for the PHY, and regnum is the register
47   number. These functions are guaranteed not to be called from interrupt
48   time, so it is safe for them to block, waiting for an interrupt to signal
49   the operation is complete
50 
51 2) A reset function is necessary. This is used to return the bus to an
52   initialized state.
53
54 3) A probe function is needed. This function should set up anything the bus
55   driver needs, setup the mii_bus structure, and register with the PAL using
56   mdiobus_register. Similarly, there's a remove function to undo all of
57   that (use mdiobus_unregister).
58 
59 4) Like any driver, the device_driver structure must be configured, and init
60   exit functions are used to register the driver.
61
62 5) The bus must also be declared somewhere as a device, and registered.
63
64 As an example for how one driver implemented an mdio bus driver, see
65 drivers/net/gianfar_mii.c and arch/ppc/syslib/mpc85xx_devices.c
66
67Connecting to a PHY
68
69 Sometime during startup, the network driver needs to establish a connection
70 between the PHY device, and the network device. At this time, the PHY's bus
71 and drivers need to all have been loaded, so it is ready for the connection.
72 At this point, there are several ways to connect to the PHY:
73
74 1) The PAL handles everything, and only calls the network driver when
75   the link state changes, so it can react.
76
77 2) The PAL handles everything except interrupts (usually because the
78   controller has the interrupt registers).
79
80 3) The PAL handles everything, but checks in with the driver every second,
81   allowing the network driver to react first to any changes before the PAL
82   does.
83 
84 4) The PAL serves only as a library of functions, with the network device
85   manually calling functions to update status, and configure the PHY
86
87
88Letting the PHY Abstraction Layer do Everything
89
90 If you choose option 1 (The hope is that every driver can, but to still be
91 useful to drivers that can't), connecting to the PHY is simple:
92
93 First, you need a function to react to changes in the link state. This
94 function follows this protocol:
95
96   static void adjust_link(struct net_device *dev);
97 
98 Next, you need to know the device name of the PHY connected to this device.
99 The name will look something like, "0:00", where the first number is the
100 bus id, and the second is the PHY's address on that bus. Typically,
101 the bus is responsible for making its ID unique.
102 
103 Now, to connect, just call this function:
104 
105   phydev = phy_connect(dev, phy_name, &adjust_link, flags, interface);
106
107 phydev is a pointer to the phy_device structure which represents the PHY. If
108 phy_connect is successful, it will return the pointer. dev, here, is the
109 pointer to your net_device. Once done, this function will have started the
110 PHY's software state machine, and registered for the PHY's interrupt, if it
111 has one. The phydev structure will be populated with information about the
112 current state, though the PHY will not yet be truly operational at this
113 point.
114
115 flags is a u32 which can optionally contain phy-specific flags.
116 This is useful if the system has put hardware restrictions on
117 the PHY/controller, of which the PHY needs to be aware.
118
119 interface is a u32 which specifies the connection type used
120 between the controller and the PHY. Examples are GMII, MII,
121 RGMII, and SGMII. For a full list, see include/linux/phy.h
122
123 Now just make sure that phydev->supported and phydev->advertising have any
124 values pruned from them which don't make sense for your controller (a 10/100
125 controller may be connected to a gigabit capable PHY, so you would need to
126 mask off SUPPORTED_1000baseT*). See include/linux/ethtool.h for definitions
127 for these bitfields. Note that you should not SET any bits, or the PHY may
128 get put into an unsupported state.
129
130 Lastly, once the controller is ready to handle network traffic, you call
131 phy_start(phydev). This tells the PAL that you are ready, and configures the
132 PHY to connect to the network. If you want to handle your own interrupts,
133 just set phydev->irq to PHY_IGNORE_INTERRUPT before you call phy_start.
134 Similarly, if you don't want to use interrupts, set phydev->irq to PHY_POLL.
135
136 When you want to disconnect from the network (even if just briefly), you call
137 phy_stop(phydev).
138
139Keeping Close Tabs on the PAL
140
141 It is possible that the PAL's built-in state machine needs a little help to
142 keep your network device and the PHY properly in sync. If so, you can
143 register a helper function when connecting to the PHY, which will be called
144 every second before the state machine reacts to any changes. To do this, you
145 need to manually call phy_attach() and phy_prepare_link(), and then call
146 phy_start_machine() with the second argument set to point to your special
147 handler.
148
149 Currently there are no examples of how to use this functionality, and testing
150 on it has been limited because the author does not have any drivers which use
151 it (they all use option 1). So Caveat Emptor.
152
153Doing it all yourself
154
155 There's a remote chance that the PAL's built-in state machine cannot track
156 the complex interactions between the PHY and your network device. If this is
157 so, you can simply call phy_attach(), and not call phy_start_machine or
158 phy_prepare_link(). This will mean that phydev->state is entirely yours to
159 handle (phy_start and phy_stop toggle between some of the states, so you
160 might need to avoid them).
161
162 An effort has been made to make sure that useful functionality can be
163 accessed without the state-machine running, and most of these functions are
164 descended from functions which did not interact with a complex state-machine.
165 However, again, no effort has been made so far to test running without the
166 state machine, so tryer beware.
167
168 Here is a brief rundown of the functions:
169
170 int phy_read(struct phy_device *phydev, u16 regnum);
171 int phy_write(struct phy_device *phydev, u16 regnum, u16 val);
172
173   Simple read/write primitives. They invoke the bus's read/write function
174   pointers.
175
176 void phy_print_status(struct phy_device *phydev);
177 
178   A convenience function to print out the PHY status neatly.
179
180 int phy_start_interrupts(struct phy_device *phydev);
181 int phy_stop_interrupts(struct phy_device *phydev);
182
183   Requests the IRQ for the PHY interrupts, then enables them for
184   start, or disables then frees them for stop.
185
186 struct phy_device * phy_attach(struct net_device *dev, const char *phy_id,
187         u32 flags, phy_interface_t interface);
188
189   Attaches a network device to a particular PHY, binding the PHY to a generic
190   driver if none was found during bus initialization. Passes in
191   any phy-specific flags as needed.
192
193 int phy_start_aneg(struct phy_device *phydev);
194   
195   Using variables inside the phydev structure, either configures advertising
196   and resets autonegotiation, or disables autonegotiation, and configures
197   forced settings.
198
199 static inline int phy_read_status(struct phy_device *phydev);
200
201   Fills the phydev structure with up-to-date information about the current
202   settings in the PHY.
203
204 int phy_ethtool_sset(struct phy_device *phydev, struct ethtool_cmd *cmd);
205 int phy_ethtool_gset(struct phy_device *phydev, struct ethtool_cmd *cmd);
206
207   Ethtool convenience functions.
208
209 int phy_mii_ioctl(struct phy_device *phydev,
210                 struct mii_ioctl_data *mii_data, int cmd);
211
212   The MII ioctl. Note that this function will completely screw up the state
213   machine if you write registers like BMCR, BMSR, ADVERTISE, etc. Best to
214   use this only to write registers which are not standard, and don't set off
215   a renegotiation.
216
217
218PHY Device Drivers
219
220 With the PHY Abstraction Layer, adding support for new PHYs is
221 quite easy. In some cases, no work is required at all! However,
222 many PHYs require a little hand-holding to get up-and-running.
223
224Generic PHY driver
225
226 If the desired PHY doesn't have any errata, quirks, or special
227 features you want to support, then it may be best to not add
228 support, and let the PHY Abstraction Layer's Generic PHY Driver
229 do all of the work.
230
231Writing a PHY driver
232
233 If you do need to write a PHY driver, the first thing to do is
234 make sure it can be matched with an appropriate PHY device.
235 This is done during bus initialization by reading the device's
236 UID (stored in registers 2 and 3), then comparing it to each
237 driver's phy_id field by ANDing it with each driver's
238 phy_id_mask field. Also, it needs a name. Here's an example:
239
240   static struct phy_driver dm9161_driver = {
241         .phy_id = 0x0181b880,
242     .name = "Davicom DM9161E",
243     .phy_id_mask = 0x0ffffff0,
244     ...
245   }
246
247 Next, you need to specify what features (speed, duplex, autoneg,
248 etc) your PHY device and driver support. Most PHYs support
249 PHY_BASIC_FEATURES, but you can look in include/mii.h for other
250 features.
251
252 Each driver consists of a number of function pointers:
253
254   config_init: configures PHY into a sane state after a reset.
255     For instance, a Davicom PHY requires descrambling disabled.
256   probe: Does any setup needed by the driver
257   suspend/resume: power management
258   config_aneg: Changes the speed/duplex/negotiation settings
259   read_status: Reads the current speed/duplex/negotiation settings
260   ack_interrupt: Clear a pending interrupt
261   config_intr: Enable or disable interrupts
262   remove: Does any driver take-down
263
264 Of these, only config_aneg and read_status are required to be
265 assigned by the driver code. The rest are optional. Also, it is
266 preferred to use the generic phy driver's versions of these two
267 functions if at all possible: genphy_read_status and
268 genphy_config_aneg. If this is not possible, it is likely that
269 you only need to perform some actions before and after invoking
270 these functions, and so your functions will wrap the generic
271 ones.
272
273 Feel free to look at the Marvell, Cicada, and Davicom drivers in
274 drivers/net/phy/ for examples (the lxt and qsemi drivers have
275 not been tested as of this writing)
276
277Board Fixups
278
279 Sometimes the specific interaction between the platform and the PHY requires
280 special handling. For instance, to change where the PHY's clock input is,
281 or to add a delay to account for latency issues in the data path. In order
282 to support such contingencies, the PHY Layer allows platform code to register
283 fixups to be run when the PHY is brought up (or subsequently reset).
284
285 When the PHY Layer brings up a PHY it checks to see if there are any fixups
286 registered for it, matching based on UID (contained in the PHY device's phy_id
287 field) and the bus identifier (contained in phydev->dev.bus_id). Both must
288 match, however two constants, PHY_ANY_ID and PHY_ANY_UID, are provided as
289 wildcards for the bus ID and UID, respectively.
290
291 When a match is found, the PHY layer will invoke the run function associated
292 with the fixup. This function is passed a pointer to the phy_device of
293 interest. It should therefore only operate on that PHY.
294
295 The platform code can either register the fixup using phy_register_fixup():
296
297    int phy_register_fixup(const char *phy_id,
298        u32 phy_uid, u32 phy_uid_mask,
299        int (*run)(struct phy_device *));
300
301 Or using one of the two stubs, phy_register_fixup_for_uid() and
302 phy_register_fixup_for_id():
303
304 int phy_register_fixup_for_uid(u32 phy_uid, u32 phy_uid_mask,
305        int (*run)(struct phy_device *));
306 int phy_register_fixup_for_id(const char *phy_id,
307        int (*run)(struct phy_device *));
308
309 The stubs set one of the two matching criteria, and set the other one to
310 match anything.
311
312

Archive Download this file



interactive