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