- 19 May, 2017 5 commits
-
-
Alex Bennée authored
This is slightly clumsy way to ensure the directory structure for the out-of-tree build is made first. The PHONY target is depended on by make all first which runs mkdir for all OBJDIRS. If an architecture target needs deeper nesting it needs to add directories to OBJDIRS so it can be picked up by the directories rule. Signed-off-by:
Alex Bennée <alex.bennee@linaro.org> Signed-off-by:
Radim Krčmář <rkrcmar@redhat.com>
-
Alex Bennée authored
I would of thought VPATH took care of this but apparently not. Signed-off-by:
-
Alex Bennée authored
Setting the VPATH prompts make to search VPATH for source files. There are still some fix-ups needed for linking and including other Makefile fragments for each architecture. Signed-off-by:
Andrew Jones <drjones@redhat.com> Signed-off-by:
-
Alex Bennée authored
Pass -C $(SRCDIR) to the git command that generates the build-head stamp. Signed-off-by:
-
Alex Bennée authored
This is a first step to enabling out-of-tree builds for kvm-unit-tests. When you invoke configure like this: ../tree.git/configure [args] It will detect we the case and: - link ../tree.git/Makefile to the build-dir - link ../tree.git/$(arch) and test scripts - link ../tree.git/run_tests.sh - ensure lib is created with a correct lib/asm - set SRCDIR in the config.mk Signed-off-by:
-
- 17 May, 2017 1 commit
-
-
David Hildenbrand authored
Both, -lpthread and -lrt seem to be relevant for api/ only. x86 already adds these flags manually, when linking api/%. So we should be safe to just drop it. This makes it possible to easily cross-compile when e.g. the librt is not available (or not linked to libc). Suggested-by:
Thomas Huth <thuth@redhat.com> Signed-off-by:
David Hildenbrand <david@redhat.com> Signed-off-by:
-
- 16 May, 2017 3 commits
-
-
David Hildenbrand authored
Signed-off-by:
-
David Hildenbrand authored
Since the ppc64 implementation is currently only a defunc dummy implementation, let's replace it by the generic one. Acked-by:
-
David Hildenbrand authored
Let's provide a basic lock implementation that should work on most architectures. Signed-off-by:
-
- 12 May, 2017 31 commits
-
-
Radim Krčmář authored
Fixes this GCC error: In file included from lib/report.c:13:0: lib/libcflat.h:134:10: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 6 has type ‘unsigned int’ [-Werror=format=] printf("%s:%d: assert failed: %s: " fmt "\n", \ ^ lib/report.c:53:2: note: in expansion of macro ‘assert_msg’ assert_msg(len < sizeof(prefixes), "%d >= %lu", len, sizeof(prefixes)); ^~~~~~~~~~ lib/report.c:53:46: note: format string is defined here assert_msg(len < sizeof(prefixes), "%d >= %lu", len, sizeof(prefixes)); ~~^ %u Reviewed-by:Paolo Bonzini <pbonzini@redhat.com> Reviewed-by:
Peter Feiner <pfeiner@google.com> Signed-off-by:
-
Paolo Bonzini authored
These are useful to avoid warnings from the compiler. Signed-off-by:
-
David Hildenbrand authored
GCC doesn't like comparison of signed and unsigned values. Signed-off-by:
-
Thomas Petazzoni authored
With gcc 4.7.2, the build fails with: x86/hyperv_clock.c: Assembler messages: x86/hyperv_clock.c:21: Error: no instruction mnemonic suffix given and no register operands; can't size instruction In order to avoid this, make the mul instruction data size explicit by adding the appropriate suffix. It operates on 64-bit data, so use "mulq". Signed-off-by:
Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by:
-
Bandan Das authored
Verify that a gpa is logged in the pml buffer when it's written to. Also verify that when the PML buffer overflows, a PML FULL event occurs. Signed-off-by:
Bandan Das <bsd@redhat.com> Signed-off-by:
-
David Matlack authored
The assertion macros in the vmx tests only call report() if the test fails. This avoids a lot of logging (e.g. vmx_ept_access_test_reserved_bits checks 3460 assertions). However the test runner interprets passing runs as "skipped" because it looks like no testcases were run. Add a function called report_pass(), which lets tests report that a test passed without printing anything to the console. Signed-off-by:
David Matlack <dmatlack@google.com> Message-Id: <20170421005004.137260-33-dmatlack@google.com> Signed-off-by:
-
Peter Feiner authored
EPT tests don't exercise 2m support unless it's advertised properly. This test asserts that we actually advertise it and that some basic 2m stuff works. Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Convenience functions for manipulating and querying page tables. Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Upcoming test allocates a 1GiB huge page. The VM needs more memory and the default container limit of 2 GiB isn't enough. Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
The upper bounds on level in set_ept_pte and install_ept_entry were just wrong. There's nothing wrong with setting an EPT PTE at the highest (512GiB) level. Got rid of unused error return values and replaced them with assertions. Nobody was checking the return values, so these functions were silently failing. Change-Id: I4db1e1c2a0c050f5f69cce28df460b4c8ce10d1c Signed-off-by:
-
Peter Feiner authored
Need this for testing huge pages. Change-Id: Iecf4f916a3d3d24844957dcd4e81c8aa10b1cdda Signed-off-by:
-
Peter Feiner authored
Makes allocation of contiguous physical pages simpler. Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Was returning the number of characters printed plus, incorrectly, one for the null terminator. Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
Peter Feiner authored
Signed-off-by:
-
David Matlack authored
Signed-off-by:
-
David Matlack authored
The Intel SDM states that following a VMCLEAR instruction, the VMCS given as an operand to VMCLEAR should be resident in memory. If the CPU caches VMCS data outside of memory, VMCLEAR must flush this cache to memory. Signed-off-by:
-
David Matlack authored
This test VMCLEARs and VMPTRLDs two VMCS's to put each VMCS into different states of active/not-active and current/not-current and then verifies that VMREAD/VMWRITE to the current VMCS work correctly. Signed-off-by:
-
David Matlack authored
Issues VMWRITE to every VMCS field and the checks that VMREAD returns the expected result. Some tricky cases: read-only fields (skipped), not-yet-implemented fields (skipped, VMREAD fails with VMfailValid), guest segment access rights fields (reserved bits are zeroed by the CPU, so only check non-reserved bits). Signed-off-by:
-
GanShun authored
Added a check in test_vmptrld to verfiy that VMXERR_VMPTRLD_VMXON_POINTER is returned instead of VMXERR_VMCLEAR_VMXON_POINTER. Signed-off-by:
GanShun <ganshun@google.com> Signed-off-by:
-
Peter Feiner authored
Before applying "KVM: nVMX: fix lifetime issues for vmcs0", exiting while vmcs02 was loaded would trigger a VM_BUG_ON. Filter out exit_monitor_from_l2_test when running the full vmx test suite (i.e. the "[vmx]" test suite in unittest.cfg). Since exit_monitor_from_l2_test causes the VM to shutdown, many of the tests in vmx_tests.c end up getting skipped. Signed-off-by:
-
Peter Feiner authored
Limited glob matching with leading or trailing '*'. Tests (and globs of tests) can be excluded by prepending with a '-'. Signed-off-by:
-
