4 Within the Linux bus framework, the FMC device is created and
5 registered by the carrier driver. For example, the PCI driver for the
6 SPEC card fills a data structure for each SPEC that it drives, and
7 registers an associated FMC device for each card. The SVEC driver can
8 do exactly the same for the VME carrier (actually, it should do it
9 twice, because the SVEC carries two FMC mezzanines). Similarly, an
10 Etherbone driver will be able to register its own FMC devices, offering
11 communication primitives through frame exchange.
13 The contents of the EEPROM within the FMC are used for identification
14 purposes, i.e. for matching the device with its own driver. For this
15 reason the device structure includes a complete copy of the EEPROM
16 (actually, the carrier driver may choose whether or not to return it -
17 for example we most likely won't have the whole EEPROM available for
20 The following listing shows the current structure defining a device.
21 Please note that all the machinery is in place but some details may
22 still change in the future. For this reason, there is a version field
23 at the beginning of the structure. As usual, the minor number will
24 change for compatible changes (like a new flag) and the major number
25 will increase when an incompatible change happens (for example, a
26 change in layout of some fmc data structures). Device writers should
27 just set it to the value FMC_VERSION, and be ready to get back -EINVAL
31 unsigned long version;
33 struct module *owner; /* char device must pin it */
34 struct fmc_fru_id id; /* for EEPROM-based match */
35 struct fmc_operations *op; /* carrier-provided */
36 int irq; /* according to host bus. 0 == none */
37 int eeprom_len; /* Usually 8kB, may be less */
38 int eeprom_addr; /* 0x50, 0x52 etc */
39 uint8_t *eeprom; /* Full contents or leading part */
40 char *carrier_name; /* "SPEC" or similar, for special use */
41 void *carrier_data; /* "struct spec *" or equivalent */
42 __iomem void *fpga_base; /* May be NULL (Etherbone) */
43 __iomem void *slot_base; /* Set by the driver */
44 struct fmc_device **devarray; /* Allocated by the bus */
45 int slot_id; /* Index in the slot array */
46 int nr_slots; /* Number of slots in this carrier */
47 unsigned long memlen; /* Used for the char device */
48 struct device dev; /* For Linux use */
49 struct device *hwdev; /* The underlying hardware device */
50 unsigned long sdbfs_entry;
51 struct sdb_array *sdb;
52 uint32_t device_id; /* Filled by the device */
53 char *mezzanine_name; /* Defaults to ``fmc'' */
57 The meaning of most fields is summarized in the code comment above.
59 The following fields must be filled by the carrier driver before
62 * version: must be set to FMC_VERSION.
64 * owner: set to MODULE_OWNER.
66 * op: the operations to act on the device.
68 * irq: number for the mezzanine; may be zero.
70 * eeprom_len: length of the following array.
72 * eeprom_addr: 0x50 for first mezzanine and so on.
74 * eeprom: the full content of the I2C EEPROM.
78 * carrier_data: a unique pointer for the carrier.
80 * fpga_base: the I/O memory address (may be NULL).
82 * slot_id: the index of this slot (starting from zero).
84 * memlen: if fpga_base is valid, the length of I/O memory.
86 * hwdev: to be used in some dev_err() calls.
88 * device_id: a slot-specific unique integer number.
91 Please note that the carrier should read its own EEPROM memory before
92 registering the device, as well as fill all other fields listed above.
94 The following fields should not be assigned, because they are filled
95 later by either the bus or the device driver:
99 * fru_id: filled by the bus, parsing the eeprom.
101 * slot_base: filled and used by the driver, if useful to it.
103 * devarray: an array og all mezzanines driven by a singe FPGA.
105 * nr_slots: set by the core at registration time.
107 * dev: used by Linux.
109 * sdb: FPGA contents, scanned according to driver's directions.
111 * sdbfs_entry: SDB entry point in EEPROM: autodetected.
113 * mezzanine_data: available for the driver.
115 * mezzanine_name: filled by fmc-bus during identification.
118 Note: mezzanine_data may be redundant, because Linux offers the drvdata
119 approach, so the field may be removed in later versions of this bus
122 As I write this, she SPEC carrier is already completely functional in
123 the fmc-bus environment, and is a good reference to look at.
126 The API Offered by Carriers
127 ===========================
129 The carrier provides a number of methods by means of the
130 `fmc_operations' structure, which currently is defined like this
131 (again, it is a moving target, please refer to the header rather than
134 struct fmc_operations {
135 uint32_t (*readl)(struct fmc_device *fmc, int offset);
136 void (*writel)(struct fmc_device *fmc, uint32_t value, int offset);
137 int (*reprogram)(struct fmc_device *f, struct fmc_driver *d, char *gw);
138 int (*validate)(struct fmc_device *fmc, struct fmc_driver *drv);
139 int (*irq_request)(struct fmc_device *fmc, irq_handler_t h,
140 char *name, int flags);
141 void (*irq_ack)(struct fmc_device *fmc);
142 int (*irq_free)(struct fmc_device *fmc);
143 int (*gpio_config)(struct fmc_device *fmc, struct fmc_gpio *gpio,
145 int (*read_ee)(struct fmc_device *fmc, int pos, void *d, int l);
146 int (*write_ee)(struct fmc_device *fmc, int pos, const void *d, int l);
149 The individual methods perform the following tasks:
153 These functions access FPGA registers by whatever means the
154 carrier offers. They are not expected to fail, and most of the time
155 they will just make a memory access to the host bus. If the
156 carrier provides a fpga_base pointer, the driver may use direct
157 access through that pointer. For this reason the header offers the
158 inline functions fmc_readl and fmc_writel that access fpga_base if
159 the respective method is NULL. A driver that wants to be portable
160 and efficient should use fmc_readl and fmc_writel. For Etherbone,
161 or other non-local carriers, error-management is still to be
165 Module parameters are used to manage different applications for
166 two or more boards of the same kind. Validation is based on the
167 busid module parameter, if provided, and returns the matching
168 index in the associated array. See *note Module Parameters:: in in
169 doubt. If no match is found, `-ENOENT' is returned; if the user
170 didn't pass `busid=', all devices will pass validation. The value
171 returned by the validate method can be used as index into other
172 parameters (for example, some drivers use the `lm32=' parameter in
173 this way). Such "generic parameters" are documented in *note
174 Module Parameters::, below. The validate method is used by
175 `fmc-trivial.ko', described in *note fmc-trivial::.
178 The carrier enumerates FMC devices by loading a standard (or
179 golden) FPGA binary that allows EEPROM access. Each driver, then,
180 will need to reprogram the FPGA by calling this function. If the
181 name argument is NULL, the carrier should reprogram the golden
182 binary. If the gateware name has been overridden through module
183 parameters (in a carrier-specific way) the file loaded will match
184 the parameters. Per-device gateware names can be specified using
185 the `gateware=' parameter, see *note Module Parameters::. Note:
186 Clients should call rhe new helper, fmc_reprogram, which both
187 calls this method and parse the SDB tree of the FPGA.
192 Interrupt management is carrier-specific, so it is abstracted as
193 operations. The interrupt number is listed in the device
194 structure, and for the mezzanine driver the number is only
195 informative. The handler will receive the fmc pointer as dev_id;
196 the flags argument is passed to the Linux request_irq function,
197 but fmc-specific flags may be added in the future. You'll most
198 likely want to pass the `IRQF_SHARED' flag.
201 The method allows to configure a GPIO pin in the carrier, and read
202 its current value if it is configured as input. See *note The GPIO
203 Abstraction:: for details.
207 Read or write the EEPROM. The functions are expected to be only
208 called before reprogramming and the carrier should refuse them
209 with `ENODEV' after reprogramming. The offset is expected to be
210 within 8kB (the current size), but addresses up to 1MB are
211 reserved to fit bigger I2C devices in the future. Carriers may
212 offer access to other internal flash memories using these same
213 methods: for example the SPEC driver may define that its carrier
214 I2C memory is seen at offset 1M and the internal SPI flash is seen
215 at offset 16M. This multiplexing of several flash memories in the
216 same address space is carrier-specific and should only be used
217 by a driver that has verified the `carrier_name' field.
224 Support for GPIO pins in the fmc-bus environment is not very
225 straightforward and deserves special discussion.
227 While the general idea of a carrier-independent driver seems to fly,
228 configuration of specific signals within the carrier needs at least
229 some knowledge of the carrier itself. For this reason, the specific
230 driver can request to configure carrier-specific GPIO pins, numbered
231 from 0 to at most 4095. Configuration is performed by passing a
232 pointer to an array of struct fmc_gpio items, as well as the length of
233 the array. This is the data structure:
238 int _gpio; /* internal use by the carrier */
239 int mode; /* GPIOF_DIR_OUT etc, from <linux/gpio.h> */
240 int irqmode; /* IRQF_TRIGGER_LOW and so on */
243 By specifying a carrier_name for each pin, the driver may access
244 different pins in different carriers. The gpio_config method is
245 expected to return the number of pins successfully configured, ignoring
246 requests for other carriers. However, if no pin is configured (because
247 no structure at all refers to the current carrier_name), the operation
248 returns an error so the caller will know that it is running under a
249 yet-unsupported carrier.
251 So, for example, a driver that has been developed and tested on both
252 the SPEC and the SVEC may request configuration of two different GPIO
253 pins, and expect one such configuration to succeed - if none succeeds
254 it most likely means that the current carrier is a still-unknown one.
256 If, however, your GPIO pin has a specific known role, you can pass a
257 special number in the gpio field, using one of the following macros:
259 #define FMC_GPIO_RAW(x) (x) /* 4096 of them */
260 #define FMC_GPIO_IRQ(x) ((x) + 0x1000) /* 256 of them */
261 #define FMC_GPIO_LED(x) ((x) + 0x1100) /* 256 of them */
262 #define FMC_GPIO_KEY(x) ((x) + 0x1200) /* 256 of them */
263 #define FMC_GPIO_TP(x) ((x) + 0x1300) /* 256 of them */
264 #define FMC_GPIO_USER(x) ((x) + 0x1400) /* 256 of them */
266 Use of virtual GPIO numbers (anything but FMC_GPIO_RAW) is allowed
267 provided the carrier_name field in the data structure is left
268 unspecified (NULL). Each carrier is responsible for providing a mapping
269 between virtual and physical GPIO numbers. The carrier may then use the
270 _gpio field to cache the result of this mapping.
272 All carriers must map their I/O lines to the sets above starting from
273 zero. The SPEC, for example, maps interrupt pins 0 and 1, and test
274 points 0 through 3 (even if the test points on the PCB are called
277 If, for example, a driver requires a free LED and a test point (for a
278 scope probe to be plugged at some point during development) it may ask
279 for FMC_GPIO_LED(0) and FMC_GPIO_TP(0). Each carrier will provide
280 suitable GPIO pins. Clearly, the person running the drivers will know
281 the order used by the specific carrier driver in assigning leds and
282 testpoints, so to make a carrier-dependent use of the diagnostic tools.
284 In theory, some form of autodetection should be possible: a driver like
285 the wr-nic (which uses IRQ(1) on the SPEC card) should configure
286 IRQ(0), make a test with software-generated interrupts and configure
287 IRQ(1) if the test fails. This probing step should be used because even
288 if the wr-nic gateware is known to use IRQ1 on the SPEC, the driver
289 should be carrier-independent and thus use IRQ(0) as a first bet -
290 actually, the knowledge that IRQ0 may fail is carrier-dependent
291 information, but using it doesn't make the driver unsuitable for other
294 The return value of gpio_config is defined as follows:
296 * If no pin in the array can be used by the carrier, `-ENODEV'.
298 * If at least one virtual GPIO number cannot be mapped, `-ENOENT'.
300 * On success, 0 or positive. The value returned is the number of
301 high input bits (if no input is configured, the value for success
304 While I admit the procedure is not completely straightforward, it
305 allows configuration, input and output with a single carrier operation.
306 Given the typical use case of FMC devices, GPIO operations are not
307 expected to ever by in hot paths, and GPIO access so fare has only been
308 used to configure the interrupt pin, mode and polarity. Especially
309 reading inputs is not expected to be common. If your device has GPIO
310 capabilities in the hot path, you should consider using the kernel's