Hello all,
We are happy to announce that libhugetlbfs 1.1-pre1 is available for
testing via either http://libhugetlbfs.ozlabs.org/ (grab the development
snapshot -- it's appropriately tagged by git as 1.1-pre1) or via the
master branch of git://ozlabs.org/~dgibson/git/libhugetlbfs.git.
The highlights:
* Segment sharing has been completely reworked by David Gibson to not
require a daemon binary, for security reasons. The new code is
definitely an improvement.
* HUGETLB_SHARE=2 is now deprecated, as it led to unsolvable issues with
address space randomization (noticed on x86_64, but the type of issue
seen could affect any architecture).
* Hugepage-backed malloc() now uses mlock() to guarantee pages to the
process. This should avoid the case where pages are available when a
process initially mmap()s a hugepage, but pages are not available when
they go to use them. NOTE: for NUMA machines, numactl should be used
to specify the hugepage allocation policy when using malloc.
* A bunch of new tests.
Here is the short log of what was changed:
Adam Litke (2):
tests: Fix zero_filesize_segment testcase
tests: Add direct-io testcase
Andy Whitcroft (1):
fix up duplicate share_path definition following miss merge
David Gibson (7):
Small tweaks to make debugging failing testcases easier
Testcase for misaligned offset maps
Make ar less verbose during build
Fix truncate_above_4GB test for wider range of pagesizes
Testcase for kernel bug attempting brk() into a hugepage area
Update test rationale comments
libhugetlbfs: Testcase for mappings overrunning TASK_SIZE
Nishanth Aravamudan (9):
Clean up hugepages before/after linkshare tests
Fix the linkshare testcase
Add and use a helper program for the hugetlbfs mountpoint
fix typo in testutils
make sure to use newlines in debugging/error paths
Reusing the linkhuge testcase to further test sharing
Remove newlines in some debugging to save space
remove sharing of writable segments
readd the mlock() call to morecore
Steve Fox (1):
the return of shrink get_extracopy()
aglitke (12):
tests: Testcase for MAP_PRIVATE OOM-liable race condition
tests: Always use two threads for the linkshare tests
tests: run_tests.sh uses bash extensions to sh
tests: use the syscall library function to access gettid()
build: Use literal tab instead of '\t'
elflink: Remove sharing daemon
tests: Add some truncation race tests
A misconversion of hugetlb_vmtruncate_list to a prio_tree has meant
elflink: Restore unmapped_abort debugging info
Merge deamon removal and new test additions
ldscripts: Fix .note.SuSE and .note.ABI-tag sections
revert: Revert commit 34c2ee9987c2945012bb4edd23ab924adcb57d46
Enjoy!
Thanks,
Nish
--
Nishanth Aravamudan <[EMAIL PROTECTED]>
IBM Linux Technology Center
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Libhugetlbfs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
