| 2009-08-20 19:05 toby |
| |
| * cryptocteon/Makefile, ep80579/Makefile: |
| Add some makefile |
| changes to let a linux-2.4 build from head work when doing 'make |
| dep'. |
| |
| 2009-08-12 21:42 davidm |
| |
| * cryptodev.c: |
| |
| if crypto_dispatch fails, make sure we bail immediately rather |
| than go on to do other error checking that we should only do if |
| crypto_dispatch worked. |
| |
| 2009-08-12 21:30 davidm |
| |
| * crypto.c: |
| |
| To avoid a race while we have the Q's unlcked we must stage the |
| blocking of the driver, so, we assume the driver to be blocked, |
| but if any unblocks come in while we are unlocked, we do not block |
| the driver. |
| |
| At worst this means we will try again and get another block. |
| |
| Without this change we can lose an unblock request and give up on |
| the driver. |
| |
| 2009-08-12 21:04 davidm |
| |
| * crypto.c: |
| |
| kill_proc needs to be defined for kernels at or after 2.6.27 |
| |
| 2009-08-12 21:03 davidm |
| |
| * ep80579/environment.mk: |
| |
| Allow OCF to be with the kernel, or within the uClinux-dist |
| modules dir. |
| |
| 2009-08-12 20:52 davidm |
| |
| * ocf-compat.h: |
| |
| if IS_ERR_VALUE is not defined, try and get it, if it doesn't |
| exist, make our own. |
| |
| 2009-05-27 01:52 davidm |
| |
| * cryptodev.c: |
| |
| Order a crd combination of cipher/hmac just as ipsec would. That |
| way it's easier to test results against cryptosoft and the HW |
| drivers that automatically fix this. |
| |
| cryptosoft processes algs in crd order, thus they get different |
| results to when compared to cryptosoft. |
| |
| encrypt = cipher then hmac decrypt = hmac then cipher |
| |
| 2009-05-27 01:48 davidm |
| |
| * ocf-compat.h: |
| |
| Make sure we have an late_initcall macro even on old systems. |
| |
| 2009-05-27 01:47 davidm |
| |
| * cryptosoft.c: |
| |
| make sure that cryptosoft it started after the kernel crypto, |
| otherwise it will not detect the kernels algs correctly. |
| |
| 2009-05-27 01:41 davidm |
| |
| * cryptocteon/: cavium_crypto.c, cryptocteon.c: |
| |
| aes/des/3des + md5/sha1 working with ipsec. |
| |
| 2009-05-20 19:11 davidm |
| |
| * cryptocteon/: README.txt, cavium_crypto.c, cryptocteon.c: |
| |
| Complete rework of the driver to setup a framework that can handle |
| all the combinations we need properly. |
| |
| Currently only DES is working. Need to knock out a DES-MD5 version |
| so we can bench ipsec, then cleanup the AES/SHA bits. |
| |
| 2009-05-14 20:53 asallawa |
| |
| * ixp4xx/ixp4xx.c: |
| |
| Remove call to flush_scheduled_work(). Doesn't appear to be |
| necessary and can cause 'scheduling while atomic' kernel errors. |
| |
| 2009-05-14 01:17 asallawa |
| |
| * crypto.c: |
| |
| In particular situations, the crypto kernel thread can run forever |
| without ever calling wait_event_interruptible() which meant that it |
| could starve the rest of the system of CPU time. To fix this, set |
| a maximum number of crypto operations to do consecutively before |
| calling schedule(). Default value is 1000, which should be less |
| than a second on most of our platforms. |
| |
| 2009-05-08 20:10 davidm |
| |
| * crypto.c: |
| |
| Currently we are queuing requests that fail for callback. The |
| callback will never happen because the request failed. This can |
| cause "busy loops" in the crypto support threads. |
| |
| Make sure we track the error status when processing a request to |
| the driver and only return 0 if everything worked. |
| |
| The original code had this right, just bad locking, the locking |
| fixes broke the error reporting. |
| |
| 2009-03-31 20:41 davidm |
| |
| * crypto.c: |
| |
| We cannot use sys_kill (not experted), so switch to using send_sig |
| and the appropriate task finding functions. |
| |
| 2009-03-27 03:50 gerg |
| |
| * crypto.c: |
| |
| From linux-2.6.29 onwards there is no kill_proc() routine. On |
| these newer systems call direct to the sys_kill() system call. |
| |
| 2009-03-14 07:00 davidm |
| |
| * ep80579/Makefile: |
| |
| merge makefile fix from head so that make clean works properly. |
| |
| 2009-03-10 22:57 davidm |
| |
| * hifn/hifn7751.c: |
| |
| Fix up cache line size for hifns, not all get it right. Patch |
| suggested and tested by Donald Lai <dlai@hifn.com>. |
| |
| 2009-03-10 21:36 davidm |
| |
| * ep80579/Makefile: |
| |
| Make sure we do not break non-tolapai builds (when running clean |
| etc) |
| |
| 2009-02-07 00:32 davidm |
| |
| * Kconfig, Makefile, cryptocteon/Makefile, cryptocteon/README.txt, |
| cryptocteon/cavium_crypto.c, cryptocteon/cryptocteon.c: |
| |
| First pass slightly working (3des/aes) crypto driver for the cavium |
| 5010. |
| |
| Still a fair bit needed here (hmac+cipher and hash processing). |
| Need to revisit the faster cavium example ipsec code. |
| |
| 2009-01-20 01:06 gerg |
| |
| * cryptosoft.c: |
| |
| Include linux/scatterlist.h instead of asm/scatterlist.h. Otherwise |
| we don't get the generic sg macros (only the asm specific |
| difinitions). |
| |
| 2009-01-16 20:05 davidm |
| |
| * cryptosoft.c: |
| |
| clean up some debug output (sign extension fo chars) |
| |
| 2008-11-20 22:50 davidm |
| |
| * ocf-compat.h: |
| |
| We need to include mm.h to get page_address on some kernels. |
| reported by Paul Wouters <paul@xelerance.com> |
| |
| 2008-11-14 06:14 davidm |
| |
| * patches/linux-2.6.26-ocf.patch: |
| |
| make sure we patch into the kernel properly. Paul Wouters |
| <paul@xelerance.com> |
| |
| 2008-09-20 02:20 davidm |
| |
| * crypto.c, cryptodev.h: |
| |
| Fix up session migration so that a driver can be removed while |
| ipsec is up and running and the tunnel will auto-migrate to a |
| backup driver (ie., cryptosoft). |
| |
| Brad Vrabete <brad.vrabete@intel.com> |