- 08 Mar, 2005 40 commits
-
-
Matt Mackall authored
Drop the cryptolib SHA implementation and use the faster and much smaller SHA implementation from lib/. Saves about 5K. This also saves time by doing one memset per update call rather than one per SHA block. Signed-off-by:
Matt Mackall <mpm@selenic.com> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Matt Mackall authored
A replacement SHA routine that's slightly larger, but over twice as fast. It's also faster and much smaller than the cryptolib version. size speed buffer size original: 350B 2.3us 320B cryptolib: 5776B 1.2us 80B this code: 466B 1.0us 320B alternate: 2112B 1.0us 80B Signed-off-by: -
Matt Mackall authored
Move random SHA code to lib/. Signed-off-by:
-
Matt Mackall authored
Clean up SHA hash function for moving to lib/ Do proper endian conversion Provide sha_init function Add kerneldoc Signed-off-by:
-
Matt Mackall authored
Kill the unrolled SHA variants, they're unused and duplicate code in cryptoapi. Signed-off-by:
-
Matt Mackall authored
Move users of private rotl/rotr functions to rol32/ror32. Crypto bits verified with tcrypt. Signed-off-by:
-
Matt Mackall authored
Add rol32 and ror32 bitops to bitops.h Signed-off-by:
-
Matt Mackall authored
Rather than batching up entropy samples, resulting in longer lock hold times when we actually process the samples, mix in samples immediately. The trickle code should eliminate almost all the additional interrupt-time overhead this would otherwise incur, with or without locking. Signed-off-by:
-
Matt Mackall authored
Clean up buffer usage for SHA and reseed. This makes the code more readable and reduces worst-case stack usage. Signed-off-by:
-
Matt Mackall authored
Simplify output hash folding Signed-off-by:
-
Matt Mackall authored
Remove long-dead md5 code. Signed-off-by:
-
Matt Mackall authored
Break apart extract_entropy into kernel and user versions, remove last extract flag and some unnecessary variables. This makes the code more readable and amenable to sparse. Signed-off-by:
-
Matt Mackall authored
Put pointer to reseed pool in pool struct and automatically pull entropy from it if it is set. This lets us remove the EXTRACT_SECONDARY flag. Signed-off-by:
-
Matt Mackall authored
Move the limit flag to the pool struct, begin process of eliminating extract flags. Signed-off-by:
-
Matt Mackall authored
Additional parameter to allow keeping an entropy reserve in the input pool. Groundwork for proper /dev/urandom vs /dev/random starvation prevention. Signed-off-by:
-
Matt Mackall authored
When reseeding, we must always do a "catastrophic reseed" where we pull enough new bits to make the new state unguessable from outputs even if we knew the old state. So we must do the checks against the minimum reseed amount under the pool lock in extract_entropy. Signed-off-by:
-
Matt Mackall authored
Static initialization for sysctl support Signed-off-by:
-
Matt Mackall authored
As we no longer allow resizing of pools, it makes sense to allocate and initialize them statically. Remove create_entropy_store and simplify rand_initialize. Signed-off-by:
-
Matt Mackall authored
Give pools more meaningful names. Signed-off-by:
-
Matt Mackall authored
Remove incorrectly named ln (it's log2!) and x86 asm function and replace with fls bitop. Signed-off-by:
-
Matt Mackall authored
This thinko.. makes things a bit more arbitrary than we'd like. I've re-audited the other rotate conversions. Signed-off-by:
-
Matt Mackall authored
We've got three definitions of rotate_left. Remove x86 and duplicate rotate definitions. Remaining definition is fixed up such that recent gcc will generate rol instructions on x86 at least. A later patch will move this to bitops and clean up the other tree users. Signed-off-by:
-
Matt Mackall authored
Remove Linux 2.2 compatibility cruft. Signed-off-by:
-
Matt Mackall authored
Remove unused extract_timer_state struct. It was formerly used to feedback zero-entropy timing samples while extracting entropy, but that had a tendency to overwhelm the batch processing queue and prevent storing real samples. Signed-off-by:
-
Matt Mackall authored
Remove redundant memsets of BSS data Signed-off-by:
-
Matt Mackall authored
Simplify the init code Signed-off-by:
-
Matt Mackall authored
- Re-init all three pools in ioctls - Clear entropy count in init_std_data under a lock - Add kerneldoc comment Signed-off-by:
-
Matt Mackall authored
ZAPENTCNT is now effectively identical to RNDCLEARPOOL, fall through Signed-off-by:
-
Matt Mackall authored
Remove pool clearing. We've only ever cleared one of three pools and there's no good reason to do it. Instead just reset the entropy count. Signed-off-by:
-
Matt Mackall authored
Original code checked in output pool for missed wakeup avoidance, while waker (batch_entropy_process) checked input pool which could result in a missed wakeup. - Move to wait_event_interruptible style - Delete superfluous waitqueue Signed-off-by:
-
Matt Mackall authored
x86_64 wasn't doing anything special in its sort_extable. Use the generic lib/extable sort. Signed-off-by:
-
Matt Mackall authored
Switch IA64 exception tables to lib/sort. Signed-off-by:
-
Matt Mackall authored
Replace exception table insertion sort with lib/sort Signed-off-by:
-
Matt Mackall authored
Point XFS qsort at lib/sort in a way that makes it happy. Signed-off-by:
-
Andrew Morton authored
We cannot tell at build time whether some module may want it. Signed-off-by:
-
Matt Mackall authored
This patch adds a generic array sorting library routine. This is meant to replace qsort, which has two problem areas for kernel use. The first issue is quadratic worst-case performance. While quicksort worst-case datasets are rarely encountered in normal scenarios, it is in fact quite easy to construct worst cases for almost all quicksort algorithms given source or access to an element comparison callback. This could allow attackers to cause sorts that would otherwise take less than a millisecond to take seconds and sorts that should take less than a second to take weeks or months. Fixing this problem requires randomizing pivot selection with a secure random number generator, which is rather expensive. The second is that quicksort's recursion tracking requires either nontrivial amounts of stack space or dynamic memory allocation and out of memory error handling. By comparison, heapsort has both O(n log n) average and worst-case performance and practically no extra storage requirements. This version runs within 70-90% of the average performance of optimized quicksort so it should be an acceptable replacement wherever quicksort would be used in the kernel. Note that this function has an extra parameter for passing in an optimized swapping function. This is worth 10% or more over the typical byte-by-byte exchange functions. Benchmarks: qsort: glibc variant 1189 bytes (+ 256/1024 stack) qsort_3f: my simplified variant 459 bytes (+ 256/1024 stack) heapsort: the version below 346 bytes shellsort: an optimized shellsort 196 bytes P4 1.8GHz Opteron 1.4GHz (32-bit) size algorithm cycles relative cycles relative 100: qsort: 38682 100.00% 27631 100.00% qsort_3f: 36277 106.63% 22406 123.32% heapsort: 43574 88.77% 30301 91.19% shellsort: 39087 98.97% 25139 109.91% 200: qsort: 86468 100.00% 61148 100.00% qsort_3f: 78918 109.57% 48959 124.90% heapsort: 98040 88.20% 68235 89.61% shellsort: 95688 90.36% 62279 98.18% 400: qsort: 187720 100.00% 131313 100.00% qsort_3f: 174905 107.33% 107954 121.64% heapsort: 223896 83.84% 154241 85.13% shellsort: 223037 84.17% 148990 88.14% 800: qsort: 407060 100.00% 287460 100.00% qsort_3f: 385106 105.70% 239131 120.21% heapsort: 484662 83.99% 340099 84.52% shellsort: 537110 75.79% 354755 81.03% 1600: qsort: 879596 100.00% 621331 100.00% qsort_3f: 861568 102.09% 522013 119.03% heapsort: 1079750 81.46% 746677 83.21% shellsort: 1234243 71.27% 820782 75.70% 3200: qsort: 1903902 100.00% 1342126 100.00% qsort_3f: 1908816 99.74% 1131496 118.62% heapsort: 2515493 75.69% 1630333 82.32% shellsort: 2985339 63.78% 1964794 68.31% 6400: qsort: 4046370 100.00% 2909215 100.00% qsort_3f: 4164468 97.16% 2468393 117.86% heapsort: 5150659 78.56% 3533585 82.33% shellsort: 6650225 60.85% 4429849 65.67% 12800: qsort: 8729730 100.00% 6185097 100.00% qsort_3f: 8776885 99.46% 5288826 116.95% heapsort: 11064224 78.90% 7603061 81.35% shellsort: 15487905 56.36% 10305163 60.02% 25600: qsort: 18357770 100.00% 13172205 100.00% qsort_3f: 18687842 98.23% 11337115 116.19% heapsort: 24121241 76.11% 16612122 79.29% shellsort: 35552814 51.64% 24106987 54.64% 51200: qsort: 38658883 100.00% 28008505 100.00% qsort_3f: 39498463 97.87% 24339675 115.07% heapsort: 50553552 76.47% 37013828 75.67% shellsort: 82602416 46.80% 56201889 49.84% 102400: qsort: 81197794 100.00% 58918933 100.00% qsort_3f: 84257930 96.37% 51986219 113.34% heapsort: 110540577 73.46% 81419675 72.36% shellsort: 191303132 42.44% 129786472 45.40% From: Zou Nan hai <nanhai.zou@intel.com> The new sort routine only works if there are an even number of entries in the ia64 exception fix-up tables. If the number of entries is odd the sort fails, and then random get_user/put_user calls can fail. Signed-off-by: -
Matt Mackall authored
CONFIG_BASE_SMALL reduce console transfer buffer Signed-off-by:
-
Matt Mackall authored
CONFIG_BASE_SMALL reduce timer list hashes Signed-off-by:
-
Matt Mackall authored
CONFIG_BASE_SMALL reduce futex hash table Signed-off-by:
-
Matt Mackall authored
CONFIG_BASE_SMALL reduce UID lookup hash Signed-off-by:
-
