1.. SPDX-License-Identifier: GPL-2.0 2 3================ 4EISA bus support 5================ 6 7:Author: Marc Zyngier <maz@wild-wind.fr.eu.org> 8 9This document groups random notes about porting EISA drivers to the 10new EISA/sysfs API. 11 12Starting from version 2.5.59, the EISA bus is almost given the same 13status as other much more mainstream buses such as PCI or USB. This 14has been possible through sysfs, which defines a nice enough set of 15abstractions to manage buses, devices and drivers. 16 17Although the new API is quite simple to use, converting existing 18drivers to the new infrastructure is not an easy task (mostly because 19detection code is generally also used to probe ISA cards). Moreover, 20most EISA drivers are among the oldest Linux drivers so, as you can 21imagine, some dust has settled here over the years. 22 23The EISA infrastructure is made up of three parts: 24 25 - The bus code implements most of the generic code. It is shared 26 among all the architectures that the EISA code runs on. It 27 implements bus probing (detecting EISA cards available on the bus), 28 allocates I/O resources, allows fancy naming through sysfs, and 29 offers interfaces for driver to register. 30 31 - The bus root driver implements the glue between the bus hardware 32 and the generic bus code. It is responsible for discovering the 33 device implementing the bus, and setting it up to be latter probed 34 by the bus code. This can go from something as simple as reserving 35 an I/O region on x86, to the rather more complex, like the hppa 36 EISA code. This is the part to implement in order to have EISA 37 running on an "new" platform. 38 39 - The driver offers the bus a list of devices that it manages, and 40 implements the necessary callbacks to probe and release devices 41 whenever told to. 42 43Every function/structure below lives in <linux/eisa.h>, which depends 44heavily on <linux/device.h>. 45 46Bus root driver 47=============== 48 49:: 50 51 int eisa_root_register (struct eisa_root_device *root); 52 53The eisa_root_register function is used to declare a device as the 54root of an EISA bus. The eisa_root_device structure holds a reference 55to this device, as well as some parameters for probing purposes:: 56 57 struct eisa_root_device { 58 struct device *dev; /* Pointer to bridge device */ 59 struct resource *res; 60 unsigned long bus_base_addr; 61 int slots; /* Max slot number */ 62 int force_probe; /* Probe even when no slot 0 */ 63 u64 dma_mask; /* from bridge device */ 64 int bus_nr; /* Set by eisa_root_register */ 65 struct resource eisa_root_res; /* ditto */ 66 }; 67 68============= ====================================================== 69node used for eisa_root_register internal purpose 70dev pointer to the root device 71res root device I/O resource 72bus_base_addr slot 0 address on this bus 73slots max slot number to probe 74force_probe Probe even when slot 0 is empty (no EISA mainboard) 75dma_mask Default DMA mask. Usually the bridge device dma_mask. 76bus_nr unique bus id, set by eisa_root_register 77============= ====================================================== 78 79Driver 80====== 81 82:: 83 84 int eisa_driver_register (struct eisa_driver *edrv); 85 void eisa_driver_unregister (struct eisa_driver *edrv); 86 87Clear enough ? 88 89:: 90 91 struct eisa_device_id { 92 char sig[EISA_SIG_LEN]; 93 unsigned long driver_data; 94 }; 95 96 struct eisa_driver { 97 const struct eisa_device_id *id_table; 98 struct device_driver driver; 99 }; 100 101=============== ==================================================== 102id_table an array of NULL terminated EISA id strings, 103 followed by an empty string. Each string can 104 optionally be paired with a driver-dependent value 105 (driver_data). 106 107driver a generic driver, such as described in 108 Documentation/driver-api/driver-model/driver.rst. Only .name, 109 .probe and .remove members are mandatory. 110=============== ==================================================== 111 112An example is the 3c59x driver:: 113 114 static struct eisa_device_id vortex_eisa_ids[] = { 115 { "TCM5920", EISA_3C592_OFFSET }, 116 { "TCM5970", EISA_3C597_OFFSET }, 117 { "" } 118 }; 119 120 static struct eisa_driver vortex_eisa_driver = { 121 .id_table = vortex_eisa_ids, 122 .driver = { 123 .name = "3c59x", 124 .probe = vortex_eisa_probe, 125 .remove = vortex_eisa_remove 126 } 127 }; 128 129Device 130====== 131 132The sysfs framework calls .probe and .remove functions upon device 133discovery and removal (note that the .remove function is only called 134when driver is built as a module). 135 136Both functions are passed a pointer to a 'struct device', which is 137encapsulated in a 'struct eisa_device' described as follows:: 138 139 struct eisa_device { 140 struct eisa_device_id id; 141 int slot; 142 int state; 143 unsigned long base_addr; 144 struct resource res[EISA_MAX_RESOURCES]; 145 u64 dma_mask; 146 struct device dev; /* generic device */ 147 }; 148 149======== ============================================================ 150id EISA id, as read from device. id.driver_data is set from the 151 matching driver EISA id. 152slot slot number which the device was detected on 153state set of flags indicating the state of the device. Current 154 flags are EISA_CONFIG_ENABLED and EISA_CONFIG_FORCED. 155res set of four 256 bytes I/O regions allocated to this device 156dma_mask DMA mask set from the parent device. 157dev generic device (see Documentation/driver-api/driver-model/device.rst) 158======== ============================================================ 159 160You can get the 'struct eisa_device' from 'struct device' using the 161'to_eisa_device' macro. 162 163Misc stuff 164========== 165 166:: 167 168 void eisa_set_drvdata (struct eisa_device *edev, void *data); 169 170Stores data into the device's driver_data area. 171 172:: 173 174 void *eisa_get_drvdata (struct eisa_device *edev): 175 176Gets the pointer previously stored into the device's driver_data area. 177 178:: 179 180 int eisa_get_region_index (void *addr); 181 182Returns the region number (0 <= x < EISA_MAX_RESOURCES) of a given 183address. 184 185Kernel parameters 186================= 187 188eisa_bus.enable_dev 189 A comma-separated list of slots to be enabled, even if the firmware 190 set the card as disabled. The driver must be able to properly 191 initialize the device in such conditions. 192 193eisa_bus.disable_dev 194 A comma-separated list of slots to be disabled, even if the firmware 195 set the card as enabled. The driver won't be called to handle this 196 device. 197 198virtual_root.force_probe 199 Force the probing code to probe EISA slots even when it cannot find an 200 EISA compliant mainboard (nothing appears on slot 0). Defaults to 0 201 (don't force), and set to 1 (force probing) when 202 CONFIG_EISA_VLB_PRIMING is set. 203 204Random notes 205============ 206 207Converting an EISA driver to the new API mostly involves *deleting* 208code (since probing is now in the core EISA code). Unfortunately, most 209drivers share their probing routine between ISA, and EISA. Special 210care must be taken when ripping out the EISA code, so other buses 211won't suffer from these surgical strikes... 212 213You *must not* expect any EISA device to be detected when returning 214from eisa_driver_register, since the chances are that the bus has not 215yet been probed. In fact, that's what happens most of the time (the 216bus root driver usually kicks in rather late in the boot process). 217Unfortunately, most drivers are doing the probing by themselves, and 218expect to have explored the whole machine when they exit their probe 219routine. 220 221For example, switching your favorite EISA SCSI card to the "hotplug" 222model is "the right thing"(tm). 223 224Thanks 225====== 226 227I'd like to thank the following people for their help: 228 229- Xavier Benigni for lending me a wonderful Alpha Jensen, 230- James Bottomley, Jeff Garzik for getting this stuff into the kernel, 231- Andries Brouwer for contributing numerous EISA ids, 232- Catrin Jones for coping with far too many machines at home. 233