1 Hardware driver for Intel i810 Random Number Generator (RNG) 2 Copyright 2000,2001 Jeff Garzik <jgarzik@pobox.com> 3 Copyright 2000,2001 Philipp Rumpf <prumpf@mandrakesoft.com> 4 5Introduction: 6 7 The i810_rng device driver is software that makes use of a 8 special hardware feature on the Intel i8xx-based chipsets, 9 a Random Number Generator (RNG). 10 11 In order to make effective use of this device driver, you 12 should download the support software as well. Download the 13 latest version of the "intel-rng-tools" package from the 14 i810_rng driver's official Web site: 15 16 http://sourceforge.net/projects/gkernel/ 17 18About the Intel RNG hardware, from the firmware hub datasheet: 19 20 The Firmware Hub integrates a Random Number Generator (RNG) 21 using thermal noise generated from inherently random quantum 22 mechanical properties of silicon. When not generating new random 23 bits the RNG circuitry will enter a low power state. Intel will 24 provide a binary software driver to give third party software 25 access to our RNG for use as a security feature. At this time, 26 the RNG is only to be used with a system in an OS-present state. 27 28Theory of operation: 29 30 Character driver. Using the standard open() 31 and read() system calls, you can read random data from 32 the i810 RNG device. This data is NOT CHECKED by any 33 fitness tests, and could potentially be bogus (if the 34 hardware is faulty or has been tampered with). Data is only 35 output if the hardware "has-data" flag is set, but nevertheless 36 a security-conscious person would run fitness tests on the 37 data before assuming it is truly random. 38 39 /dev/intel_rng is char device major 10, minor 183. 40 41Driver notes: 42 43 * FIXME: support poll(2) 44 45 NOTE: request_mem_region was removed, for two reasons: 46 1) Only one RNG is supported by this driver, 2) The location 47 used by the RNG is a fixed location in MMIO-addressable memory, 48 3) users with properly working BIOS e820 handling will always 49 have the region in which the RNG is located reserved, so 50 request_mem_region calls always fail for proper setups. 51 However, for people who use mem=XX, BIOS e820 information is 52 -not- in /proc/iomem, and request_mem_region(RNG_ADDR) can 53 succeed. 54 55Driver details: 56 57 Based on: 58 Intel 82802AB/82802AC Firmware Hub (FWH) Datasheet 59 May 1999 Order Number: 290658-002 R 60 61 Intel 82802 Firmware Hub: Random Number Generator 62 Programmer's Reference Manual 63 December 1999 Order Number: 298029-001 R 64 65 Intel 82802 Firmware HUB Random Number Generator Driver 66 Copyright (c) 2000 Matt Sottek <msottek@quiknet.com> 67 68 Special thanks to Matt Sottek. I did the "guts", he 69 did the "brains" and all the testing. 70 71Change history: 72 73 Version 0.9.8: 74 * Support other i8xx chipsets by adding 82801E detection 75 * 82801DB detection is the same as for 82801CA. 76 77 Version 0.9.7: 78 * Support other i8xx chipsets too (by adding 82801BA(M) and 79 82801CA(M) detection) 80 81 Version 0.9.6: 82 * Internal driver cleanups, prep for 1.0.0 release. 83 84 Version 0.9.5: 85 * Rip out entropy injection via timer. It never ever worked, 86 and a better solution (rngd) is now available. 87 88 Version 0.9.4: 89 * Fix: Remove request_mem_region 90 * Fix: Horrible bugs in FIPS calculation and test execution 91 92 Version 0.9.3: 93 * Clean up rng_read a bit. 94 * Update i810_rng driver Web site URL. 95 * Increase default timer interval to 4 samples per second. 96 * Abort if mem region is not available. 97 * BSS zero-initialization cleanup. 98 * Call misc_register() from rng_init_one. 99 * Fix O_NONBLOCK to occur before we schedule. 100 101 Version 0.9.2: 102 * Simplify open blocking logic 103 104 Version 0.9.1: 105 * Support i815 chipsets too (Matt Sottek) 106 * Fix reference counting when statically compiled (prumpf) 107 * Rewrite rng_dev_read (prumpf) 108 * Make module races less likely (prumpf) 109 * Small miscellaneous bug fixes (prumpf) 110 * Use pci table for PCI id list 111 112 Version 0.9.0: 113 * Don't register a pci_driver, because we are really 114 using PCI bridge vendor/device ids, and someone 115 may want to register a driver for the bridge. (bug fix) 116 * Don't let the usage count go negative (bug fix) 117 * Clean up spinlocks (bug fix) 118 * Enable PCI device, if necessary (bug fix) 119 * iounmap on module unload (bug fix) 120 * If RNG chrdev is already in use when open(2) is called, 121 sleep until it is available. 122 * Remove redundant globals rng_allocated, rng_use_count 123 * Convert numeric globals to unsigned 124 * Module unload cleanup 125 126 Version 0.6.2: 127 * Clean up spinlocks. Since we don't have any interrupts 128 to worry about, but we do have a timer to worry about, 129 we use spin_lock_bh everywhere except the timer function 130 itself. 131 * Fix module load/unload. 132 * Fix timer function and h/w enable/disable logic 133 * New timer interval sysctl 134 * Clean up sysctl names 135