| Age | Commit message (Collapse) | Author | |
|---|---|---|---|
| 2000-09-21 | style | Jason Wright | |
| 2000-08-19 | - introduce new function ubsec_mcopy() which copies the header and trailer | Jason Wright | |
| of one mbuf into another since ubsec doesn't copy data through - get all of the offsets right for the !nicealign case - style fixes Authentication/Encryption on the same packet now appears to work reliably | |||
| 2000-08-17 | use destination lengths when trimming the output packet (cut and pasto) | Jason Wright | |
| 2000-08-17 | - auth+enc sorta works: fixup all of the offsets for encryption/authentication | Jason Wright | |
| - also check for cases that ubsec cannot handle (and should not happen anyway) - actually use the destination length to trim the packet in the output descriptor - don't copy the IV into the packet unless it doesn't already have one (old debugging code...) - style fixes | |||
| 2000-08-15 | use pci_mapreg_map, deallocate resources on failure in attachment; jason@ ok | Michael Shalayeff | |
| 2000-08-15 | Put the bcopy back in place for the decryption CRD_F_EXPLICIT_IV case (missed | Jason Wright | |
| this yesterday when fixing the encryption side of this case). Only used for old IPSec xforms anyway. | |||
| 2000-08-15 | don't forget the initial swizzling with IPAD for inner state (fixes | Jason Wright | |
| auth-only mode which was broken with new session code). | |||
| 2000-08-15 | - Can't avoid one of those bcopy's so easily (only happens with old ESP | Jason Wright | |
| transforms... those with CRD_F_IV_EXPLICIT, so we can take the function call overhead there). - >> instead of << for coffset (auth+enc still isn't working, but this was an obvious bug). - #ifdef UBSEC_DEBUG out the auth-only debugging code | |||
| 2000-08-13 | do iv copying by hand, to avoid bcopy overhead | Theo de Raadt | |
| 2000-08-13 | fix session code | Theo de Raadt | |
| 2000-08-13 | not completely working session code from jason | Theo de Raadt | |
| 2000-08-13 | indent | Theo de Raadt | |
| 2000-08-12 | mcr aggregation now works, best i have seen yet is 5 (huh?) | Theo de Raadt | |
| 2000-08-11 | move mcr out of q; and write prelim mcr aggregation code, which does not yet | Theo de Raadt | |
| work for some reason or another, so it is currently disabled. | |||
| 2000-07-31 | since byte order bugs are gone, interrupts work ok | Theo de Raadt | |
| 2000-07-29 | remove the (unused) sc_intrmask, and be more careful about initializing | Jason Wright | |
| the BS_CTRL register... The BE32 & BE64 bits do NOT do what they imply, so leave the bits set, and add the ones we want. This allows ubsec to interop with our software implementation (at least for encryption). More work to be done in this driver though... Many thanks to Jimmy Ruane at Broadcom for the pointer about BE32 & BE64! | |||
| 2000-07-20 | work around broken A0 5805 silicon that fails to set the status result word ↵ | Theo de Raadt | |
| in the mcr | |||
| 2000-06-20 | call crypto_done() | Jason Wright | |
| 2000-06-19 | oops, nuke unused variable | Jason Wright | |
| 2000-06-19 | add authentication-only handling (easy setup). | Jason Wright | |
| remove most of the parameters from ubsec_callback() since they can be found later. ok, so ubsec can now authenticate to itself (doesn't like software, tho) | |||
| 2000-06-18 | Use the same field data types as the reference code and adjust offsets | Jason Wright | |
| appropriately. Byte swap key/iv fields because they are given to us as "network order", but the chip operates as little endian. coffset is in WORDS not bytes | |||
| 2000-06-14 | readd queue limiting code that was backed out yesterday. (ip_esp.c change | Jason Wright | |
| seems to have fixed this). | |||
| 2000-06-13 | backout previous 2 changes... causes panics down the line. | Jason Wright | |
| 2000-06-13 | Oops: initialize q to avoid using the unitialized value when the queue fills | Jason Wright | |
| 2000-06-13 | put an upperbound on queue length | Jason Wright | |
| 2000-06-13 | correct 5805 test | Theo de Raadt | |
| 2000-06-12 | ESP 3des now works, after squishing 4 bugs | Theo de Raadt | |
| 2000-06-10 | correct SIMPLEQ bugs, ack interrupt. 8 packets have crypted, but kernel | Theo de Raadt | |
| memory got corrupted. | |||
| 2000-06-10 | another vaddr used as paddr | Jason Wright | |
| 2000-06-10 | paddr not vaddr for mcr1 | Jason Wright | |
| 2000-06-03 | Move everything to where is supposed to be (reg definitions, etc). | Jason Wright | |
| Add some of the skip logic. | |||
| 2000-06-02 | squeeze basic framework into place. process generates SIMPLEQ of requests, | Theo de Raadt | |
| they get fed in, irq recovers old one, feeds new one in, callback schreds and calls back to crypto(9)... mac result buffers and packet offsets need work. | |||
| 2000-05-18 | work in progress: driver for BlueSteel (Broadcom) 5[56]01 crypto accelerator | Jason Wright | |
 |
index : src | |
| OpenBSD base system |
| summaryrefslogtreecommitdiff |