Instead of implementing downloading of external resources in every
Makefile again, use one script. This script adds the following
features:
- Always check the download against a known SHA-256 hash.
- Support caching of downloaded resources (set DL_CACHE=/some/dir).
- Allow offline builds if all files are cached. If DL_OFFLINE=true the
build will never attempt to download anything.
For detection of SGX/non-SGX (for example in regression tests) always
use the SGX environment variable. To generate launch/EINIT tokens use
the new 'sgx-tokens' Make target.
- Deduplicate much of setup code.
- Allow running non-sandbox and sandbox code in single run.
- Use pytest.
- Generate JUnit-XML file for Jenkins.
- Document running a subset of regression tests.
This fixes two problems with subprocess handling:
1. Use subprocess.communicate to avoid that the pipe can fill up.
2. Use a process group to ensure that we also kill the subprocess'
children in case of a timeout.
Note: This does not cover the case that the child returns successfully
but leaves other child processes running. AFAICS this is not easy since
subprocess.communicate() wait(2)s for the process so the pid is no
longer guaranteed to be valid to kill. So we leave this to the outside
environment (Jenkins or whatever).
The regression python scripts are not marked executable. Also they
expect that the Makefile has setup the environment correctly so just
marking them executable would not be useful. So drop the unused shebang
to avoid encoding the interpreter in multiple places.
Script/regression.py is a library so it also doesn't need a shebang.
.packed directories are originally created for testing on non-Linux platforms. The usage is mostly obsolete right now, and the binaries are completely out-dated. A better strategy would be to create a hook in the github to release binary packages.
- Fix a severe bug caused by merging #87. The patch does not check the return values from get_config_entries_size(), while the return values can easily be negative (-PAL_ERROR_INVAL). Any usage of this function should carefully check the return values and also pass the size into get_config_entries() to prevent buffer overflow.
- Fix a minor bug in handling IPC disconnection. The callback ipc_child_exit() never receives a "term_signal" parameter. The callback is called when thepipe to a child process unexpectedly terminates, which can only mean the child process has crashed. I believe the proper signal to send is SIGKILL.
- Remove a compilation warning caused by type mismatch of __malloc().
- Allow binding TCP socket to ANY address
- enable LD_PRELOAD in glibc
- disable output buffering in LTP tests to reveal omitted passes
- Remove a double-unlock in shim_async.c
- Replace all int with size_t or ssize_t in Graphene configuration API
- Add new passed LTP tests
* add memusg to pal_loader script
* Stop bleeding PAL handles.
Deprecating DkOjectReference and reference counting in PAL handle.
Deprecating DkSemaphoreDestroy and DkEventDestroy (replaced by DkObjectClose).
Cleaning unused PAL handles in the library OS.
Adding a heap tracing feature to profile usage of PAL handles.
* fix a bug in SGX mode that mapping untrusted files into memory never got free by DkVirtualMemoryFree()
* adding lighttpd SSL option
* fixing the freeing convention of PAL handles; On SGX, event and mutex handles need to be freed seperately.
* changing how mutexes and events are allocated on SGX
* fixing GCC regression tests (for both Linux and SGX)
* fix a double-free problem of the first thread handle
* Fix a PAL unit test
* Make the PAL unit tests fail properly if one that should work doesn't.
* Get consistent indexes with and without graphene ipc module
* A little debugging output for ltp flakiness. Seems to make things a little more stable.
* A bugfix for issue #80. The migration is not complete when a file-backed VMA is larger than
the file size, but the file size is not page-aligned. In Linux, if the file size is smaller
than the mapped memory, accessing the remaining memory that is not backed by the file will
trigger a SIGBUS. However, it is fine to access the remaining memory within the last page that
still overlaps with file, and won't trigger a SIGBUS. This area is commonly used in ELF binary
to store uninitialized, static variables. To improve fork latency, Graphene chooses to
truncate the migration size as the file size, but missed the memory which belongs to the last
file-backed page. The migration size should be aligned up to the page size.
* Fix assertion lines in PAL and LibOS
* Make sure the return code is correct for the LTP script, bump up some timeouts
* Only run gipc tests when the module is loaded, for the purposes of getting CI to work.
* Build fix
- Bugfixes in PAL and SHIM
- Improvement of migration scheme in SHIM, which largely speed up
forking
- Upgrade glibc to 2.19
- FreeBSD port
- Support OpenJDK, python and R
Isaku Yamahata