| Age | Commit message (Collapse) | Author | |
|---|---|---|---|
| 2002-12-06 | Don't use the RNG oscillator output directly, use the sha1'd version (the | Jason Wright | |
| the direct data does not pass 1/2 of the FIPS140-2 tests with any degree of regularity). | |||
| 2002-12-05 | Treat RNGSHA1 operations the same as RNGBYPASS for callback purposes (they | Jason Wright | |
| produce the same size/format data). | |||
| 2002-11-21 | From Angelos: | Jason Wright | |
| - simplistic load balancing across multiple cards - simplified registration process - a few style nits. | |||
| 2002-11-19 | Add a simplistic table driven lookup routine and use it where appropriate. | Jason Wright | |
| 2002-10-12 | Remove more '\n's from panic() statements. Both trailing and leading. | Kenneth R Westerback | |
| Diff generated by Chris Kuethe. | |||
| 2002-10-10 | Match the Sun Crypto Adapter 1000, it appears to be a broadcom 5821. | Jason Wright | |
| 2002-10-05 | match 5822 | Jason Wright | |
| 2002-09-24 | Don't use constants for the output parameter, use the iparam count as a ↵ | Jason Wright | |
| pointer to the first result location. | |||
| 2002-09-19 | remember: bits is bits and bytes is bytes... use -byte- count for bcopy not | Jason Wright | |
| bits. Also, the conversion between bits and bytes involves a division by 8 not 2. (The latter pointed out by Francis Cianfrocca <vze32r6m@verizon.net>) | |||
| 2002-09-12 | - Split out the hardware and software normalization versions of modexp... | Jason Wright | |
| I screwed something up when the function was trying to do both and it's much easier to read this way (and heck, even works). - Enable hardware normalization for chips that support it | |||
| 2002-09-11 | - On reset, disable hardware normalization for 582x and make sure the chip ↵ | Jason Wright | |
| is in little endian mode. - since sw normalization is now the only option, simplify normalization handling - remove some leftover #if 0 code | |||
| 2002-09-04 | Treat 5822 the same as 5820 | Jason Wright | |
| 2002-09-04 | 5801 has no pk or rng support | Jason Wright | |
| 2002-09-03 | add support for 5801 and 5802 which appear to be 5805's as far as sw is ↵ | Jason Wright | |
| concerned | |||
| 2002-07-08 | 5821 has two additional bits that must be ack'd (note they don't have | Jason Wright | |
| corresponding enable bits... they are always on... dain bramage). | |||
| 2002-07-05 | Never call crp_callback directly, use crypto_done() instead | Jason Wright | |
| 2002-07-05 | KNF (no space after casts) | Jason Wright | |
| 2002-07-03 | But when denormalizing we need the normalized length, too as well as the ↵ | Jason Wright | |
| destination length. | |||
| 2002-07-03 | result only needs to be as big as the number of real bits in the modulus | Jason Wright | |
| 2002-06-17 | remove noisy jason debug printf blatther. bad p | Theo de Raadt | |
| 2002-05-16 | Normalize the exponent too. This allows exponents with bit lengths different | Jason Wright | |
| from base/modulus to work on 5805. | |||
| 2002-05-16 | Re-enable RNG on the 5601 (it needs to be disabled on some 5805 variants, | Jason Wright | |
| but I'm not sure which revisions yet). | |||
| 2002-05-15 | Rework MODEXP: | Jason Wright | |
| 5805 (and 5601) require the modulus and base to be normalized to the right of one of several different register lengths. The result is also normalized to the same length. Provide functions for shifting the bits back and forth as appropriate. Note: for consistencies sake the exponent is NOT normalized. | |||
| 2002-05-13 | add and use three more flags: | Jason Wright | |
| RNG: chip has usable rng (5805/5820/5821) HWNORM: chip will automagically normalize bignums (5820/5821) BIGKEY: chip supports "large keys" (5820/5821) | |||
| 2002-05-08 | - Go ahead and register to handle CRK_MOD_EXP_CRT ops | Jason Wright | |
| - completely (almost) revamp kprocess_rsapriv to match what the chip expects - add and use a function to compute the significant bits of a given number (this will be necessary for normalization) [Still doesn't reproduce the broadcom or provos test data, but it's closer...] | |||
| 2002-05-06 | basic infrastructure for handling RSA with CRT parameters. Just need to | Jason Wright | |
| figure out how p, q, dp, dq, and pinv fit into the context. | |||
| 2002-05-06 | - Only copy the significant bits of the result out (and make sure the buffer | Jason Wright | |
| is long enough to handle it) and bzero the rest. - Increase key buffer sizes to 2048 bits. | |||
| 2002-05-02 | the exponent parameter is: | Jason Wright | |
| 1. byte counted (not word count * 4 counted) 2. not normalized (we don't normalize anything yet, so no biggy) Increase maximum size of exponent to 2048 bits. (Better length checks coming soon) | |||
| 2002-05-02 | More of previous (don't refer to parameters by index directly) | Jason Wright | |
| Also, add a sanity check that the result bits must be >= modulus bits. | |||
| 2002-05-02 | Don't refer to the parameters directly by number, use macro's so it can be | Jason Wright | |
| changed later if necessary. Also, don't bother bzero'n the mcr, all of the fields are initialized anyway | |||
| 2002-05-01 | - make sure 'me' is initialized | Jason Wright | |
| - compute modulus bits early (if its too big, return E2BIG) - modulus bits must be rounded to 512/768/1024 (and/or 1536/2048 for 5820) - allocate the result based on modulus bits and bzero it - add two diagnostic checks that will hang the chip: unaligned result/length [score so far: 655 out of 1000 test cases work for modexp on 5820] | |||
| 2002-04-30 | - Output chain length must be equal to the byte count of the modulus length | Jason Wright | |
| - document parameter order (it may yet change) - don't bother setting mcr_pktlen, it isn't used | |||
| 2002-04-28 | make UBSEC_NO_RNG work correctly (ie. it means no RNG, not no MCR2 operations) | Jason Wright | |
| 2002-04-26 | debug stuff inside #ifdef | Theo de Raadt | |
| 2002-04-26 | missed one; deraadt | Jason Wright | |
| 2002-04-26 | 1024 bit max | Jason Wright | |
| 2002-04-26 | register as supporting modular exponentiation | Jason Wright | |
| still contains much debugging code and isn't quite done, but its a start | |||
| 2002-04-22 | clean up attach messages | Jason Wright | |
| 2002-04-08 | Credit DARPA/USAF appropriately. | Jason Wright | |
| 2002-03-14 | First round of __P removal in sys | Todd C. Miller | |
| 2002-01-28 | Try to share a common src/dst map where possible, and cope with that | Jason Wright | |
| eventuality later in the code. | |||
| 2002-01-28 | First round of post-bus_dma cleanups: | Jason Wright | |
| - remove the packl/packp arrays and rework handling appropriately - destination map may or may not exist, cope. - remove a redundant bus_dmamap_sync() in _process (real sync is in _feed) - remove long deprecated q_{src,dst}pkt stuff from queue structure | |||
| 2002-01-28 | Quick (and a little dirty) conversion to bus_dma(9). | Jason Wright | |
| 2002-01-24 | More cleaning | Jason Wright | |
| 2002-01-19 | From Patrik Lindergren (patrik@ipunplugged.com): | Jason Wright | |
| * make the driver big-endian aware * handling for DMA errors * move some allocations to attach From me: whitespace clean up and vtophys removal (almost works on sparc64) | |||
| 2002-01-02 | at least ; required after label or case; openbsd@davidkrause.com | Theo de Raadt | |
| 2001-12-07 | Don't disable TRDY/RETRY, this doesn't have the expected behavior. | Jason Wright | |
| 2001-11-20 | Match bcom 5821 (this is untested and is based on the datasheet's claim that | Jason Wright | |
| the 5821 is "register and software compatible with Broadcom 5820"). | |||
| 2001-11-14 | - Move rng buffer allocation (bus_dmamem_map/unmap) to attach instead of during | Jason Wright | |
| the first timeout which is the wrong time to be calling bus_dmamem_map/unmap - Make sure that mastering is really enabled. - remove some debugging stuff that would be a pain to get working on sparc64. | |||
| 2001-11-09 | be way more sure that software cannot be used | Theo de Raadt | |
 |
index : src | |
| OpenBSD base system |
| summaryrefslogtreecommitdiff |