Snabb 2018.09 “Eggplant”
This release improves compatibility of the vhost-user driver with current QEMU versions, and adds rudimentary PMU support for AMD family 15h+ CPUs (including Ryzen). With “Eggplant”, our high-performance hash table implementation received a callback that can be invoked when it is resized, and a generic Token Bucket implementation (lib.token_bucket) and time stamp counter module (lib.tsc) are introduced. As usual, this release comes with numerous bug fixes and minor enhancements.
Changes
- #1369 apps.vhost: compatibility updates for current QEMU versions
- #1354 lib.pmu: rudimentary support for AMD family 15h+ CPUs
- #1359 Add optional callback to the CTable:resize() method
- #1355 Add support for RDTSC, apply to token bucket
- #1358 Add protocol header for ICMPv6 Packet-Too-Big type
- #1210 #1367 #1362 #1351 #1344 #1366 #1305 #1373 Various bug fixes and minor enhancements
Contributors
Alexander Gall (4):
- fb29cd0: Add support for RDTSC, apply to token bucket
- c7d772c: Add protocol header for ICMPv6 Packet-Too-Big type
- af98031: Add optional callback to the CTable:resize() method
- 8a1ebb0: ctable: fix power-of-2 issue
Andy Wingo (1):
- de790aa: Load potentially 64-bit values using mov64
Asumu Takikawa (5):
- 015dcea: Fix minor markdown typo
- 8d6e22d: Adjust ipfix README diagram
- 88079ad: vhost_user: Add VHOST_USER_F_PROTOCOL_FEATURES
- 27e8c13: vhost_user: fill out two additional messages
- 4700a96: vhost_user: asserts for wrong/missing link name
Ben Agricola (1):
- 270aea4: Fix length property testing
Luke Gorrie (3):
- 73223bd: Bump .version to 2018.08
- 8cfe828: fixup! Bump .version to 2018.08
- dc4cb79: Add engine.stop() to stop all apps.
Max Rottenkolber (4):
- 0077d21: core.memory: unmap tmpptr in allocate_huge_page
- 4a40691: lib.pmu: rudimentary support for AMD family 17h CPUs
- 3aa1cda: lib.pmu_x86: fix regression introduced in 4a40691
- a597b40: lib.numa: pretend non-NUMA systems are NUMA systems with one node
kullanici0606 (1):
- 73de63b: Fixes: #1210 hugetlb parameter causes problem on Centos and since memory is already mapped on hugetlbfs and with mlock, there is no need to explicitly use hugetlb parameter (See comment: #1210 (comment))