| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Single-stepping a userspace-emulated instruction that generates an
interrupt causes GDB to land on the instruction following it instead of
the respective interrupt handler.
The reason is that after arranging a KVM_EXIT_S390_SIEIC exit,
kvm_handle_sie_intercept() calls kvm_s390_handle_per_ifetch_icpt(),
which sets KVM_GUESTDBG_EXIT_PENDING. This bit, however, is not
processed immediately, but rather persists until the next ioctl(),
causing a spurious single-step exit.
Fix by clearing this bit in ioctl().
Reviewed-by: David Hildenbrand <[email protected]>
Reviewed-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Ilya Leoshkevich <[email protected]>
Message-ID: <[email protected]>
Signed-off-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Janosch Frank <[email protected]>
|
|
Single-stepping a kernel-emulated instruction that generates an
interrupt causes GDB to land on the instruction following it instead of
the respective interrupt handler.
The reason is that kvm_handle_sie_intercept(), after injecting the
interrupt, also processes the PER event and arranges a KVM_SINGLESTEP
exit. The interrupt is not yet delivered, however, so the userspace
sees the next instruction.
Fix by avoiding the KVM_SINGLESTEP exit when there is a pending
interrupt. The next __vcpu_run() loop iteration will arrange a
KVM_SINGLESTEP exit after delivering the interrupt.
Reviewed-by: David Hildenbrand <[email protected]>
Reviewed-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Ilya Leoshkevich <[email protected]>
Message-ID: <[email protected]>
Signed-off-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Janosch Frank <[email protected]>
|
|
Currently, after single-stepping an instruction that generates a
specification exception, GDB ends up on the instruction immediately
following it.
The reason is that vcpu_post_run() injects the interrupt and sets
KVM_GUESTDBG_EXIT_PENDING, causing a KVM_SINGLESTEP exit. The
interrupt is not delivered, however, therefore userspace sees the
address of the next instruction.
Fix by letting the __vcpu_run() loop go into the next iteration,
where vcpu_pre_run() delivers the interrupt and sets
KVM_GUESTDBG_EXIT_PENDING.
Reviewed-by: David Hildenbrand <[email protected]>
Signed-off-by: Ilya Leoshkevich <[email protected]>
Reviewed-by: Claudio Imbrenda <[email protected]>
Message-ID: <[email protected]>
Signed-off-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Janosch Frank <[email protected]>
|
|
After single-stepping an instruction that generates an interrupt, GDB
ends up on the second instruction of the respective interrupt handler.
The reason is that vcpu_pre_run() manually delivers the interrupt, and
then __vcpu_run() runs the first handler instruction using the
CPUSTAT_P flag. This causes a KVM_SINGLESTEP exit on the second handler
instruction.
Fix by delaying the KVM_SINGLESTEP exit until after the manual
interrupt delivery.
Acked-by: David Hildenbrand <[email protected]>
Reviewed-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Ilya Leoshkevich <[email protected]>
Message-ID: <[email protected]>
Signed-off-by: Claudio Imbrenda <[email protected]>
Signed-off-by: Janosch Frank <[email protected]>
|
|
Provide an immutable point in kvm-x86/selftests so that the guest printf()
support can be merged into other architectures' trees.
|
|
Add a test to ensure that setting both generic and fixed performance
event filters does not affect the consistency of the fixed event filter
behavior in KVM.
Signed-off-by: Jinrong Liang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add tests to cover that pmu event_filter works as expected when it's
applied to fixed performance counters, even if there is none fixed
counter exists (e.g. Intel guest pmu version=1 or AMD guest).
Signed-off-by: Jinrong Liang <[email protected]>
Reviewed-by: Isaku Yamahata <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add test cases to verify the handling of unsupported input values for the
PMU event filter. The tests cover unsupported "action" values, unsupported
"flags" values, and unsupported "nevents" values. All these cases should
return an error, as they are currently not supported by the filter.
Furthermore, the tests also cover the case where setting non-existent
fixed counters in the fixed bitmap does not fail.
Signed-off-by: Jinrong Liang <[email protected]>
Reviewed-by: Isaku Yamahata <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add custom "__kvm_pmu_event_filter" structure to improve pmu event
filter settings. Simplifies event filter setup by organizing event
filter parameters in a cleaner, more organized way.
Alternatively, selftests could use a struct overlay ala vcpu_set_msr()
to avoid dynamically allocating the array:
struct {
struct kvm_msrs header;
struct kvm_msr_entry entry;
} buffer = {};
memset(&buffer, 0, sizeof(buffer));
buffer.header.nmsrs = 1;
buffer.entry.index = msr_index;
buffer.entry.data = msr_value;
but the extra layer added by the nested structs is counterproductive
to writing efficient, clean code.
Suggested-by: Sean Christopherson <[email protected]>
Signed-off-by: Jinrong Liang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
[sean: massage changelog to explain alternative]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
None of the callers consume remove_event(), and it incorrectly implies
that the incoming filter isn't modified. Drop the return.
Suggested-by: Sean Christopherson <[email protected]>
Signed-off-by: Jinrong Liang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add x86 properties for Intel PMU so that tests don't have to manually
retrieve the correct CPUID leaf+register, and so that the resulting code
is self-documenting.
Suggested-by: Sean Christopherson <[email protected]>
Signed-off-by: Jinrong Liang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Use GUEST_FAIL() in ARM's arch timer helpers now that printf-based
guest asserts are the default (and only) style of guest asserts, and
say goodbye to the GUEST_ASSERT_1() alias.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Use the newfanged printf-based guest assert framework to spit out the
guest RIP when an unhandled exception is detected, which makes debugging
such failures *much* easier.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Drop the param-based guest assert macros and enable the printf versions
for all selftests. Note! This change can affect tests even if they
don't use directly use guest asserts! E.g. via library code, or due to
the compiler making different optimization decisions.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's XCR0 vs. CPUID test to use printf-based guest asserts.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's VMX PMU capabilities test to use printf-based guest asserts.
Opportunstically add a helper to do the WRMSR+assert so as to reduce the
amount of copy+paste needed to spit out debug information.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's userspace I/O test to use printf-based guest asserts.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's TSC MSRs test, and it's liberal use of GUEST_ASSERT_EQ(), to
use printf-based guest assert reporting.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's nested SVM software interrupt injection test to use printf-
based guest asserts. Opportunistically use GUEST_ASSERT() and
GUEST_FAIL() in a few locations to spit out more debug information.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert the set_boot_cpu_id test to use printf-based guest asserts,
specifically the EQ and NE variants.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's nested exceptions test to printf-based guest asserts, and
use REPORT_GUEST_ASSERT() instead of TEST_FAIL() so that output is
formatted correctly.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's MONITOR/MWAIT test to use printf-based guest asserts. Add a
macro to handle reporting failures to reduce the amount of copy+paste
needed for MONITOR vs. MWAIT.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's KVM paravirtualization test to use the printf-based
GUEST_ASSERT_EQ().
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's Hyper-V feature test to use print-based guest asserts.
Opportunistically use the EQ and NE variants in a few places to capture
additional information.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's Hyper-V extended hypercalls test to use printf-based
GUEST_ASSERT_EQ().
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert x86's CPUID test to use printf-based GUEST_ASSERT_EQ() so that
the test prints out debug information. Note, the test previously used
REPORT_GUEST_ASSERT_2(), but that was pointless because none of the
guest-side code passed any parameters to the assert.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert the steal_time test to use printf-based GUEST_ASERT.
Opportunistically use GUEST_ASSERT_EQ() and GUEST_ASSERT_NE() so that the
test spits out debug information on failure.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert set_memory_region_test to print-based GUEST_ASSERT, using a combo
of newfangled macros to report (hopefully) useful information.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert s390's tprot test to printf-based GUEST_ASSERT.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert s390's memop test to printf-based GUEST_ASSERT, and
opportunistically use GUEST_FAIL() to report invalid sizes.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Use the printf-based GUEST_ASSERT_EQ() in the memslot perf test instead of
an half-baked open code version.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Use printf-based guest assert reporting in ARM's vGIC IRQ test. Note,
this is not as innocuous as it looks! The printf-based version of
GUEST_ASSERT_EQ() ensures the expressions are evaluated only once, whereas
the old version did not!
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Use GUEST_FAIL() in ARM's page fault test to report unexpected faults.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert ARM's hypercalls test to use printf-based GUEST_ASSERT().
Opportunistically use GUEST_FAIL() to complain about an unexpected stage.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert ARM's debug exceptions test to use printf-based GUEST_ASSERT().
Opportunistically Use GUEST_ASSERT_EQ() in guest_code_ss() so that the
expected vs. actual values get printed out.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Convert ARM's aarch_timer test to use printf-based GUEST_ASSERT().
To maintain existing functionality, manually print the host information,
e.g. stage and iteration, to stderr prior to reporting the guest assert.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add a test to exercise the various features in KVM selftest's local
snprintf() and compare them to LIBC's snprintf() to ensure they behave
the same.
This is not an exhaustive test. KVM's local snprintf() does not
implement all the features LIBC does, e.g. KVM's local snprintf() does
not support floats or doubles, so testing for those features were
excluded.
Testing was added for the features that are expected to work to
support a minimal version of printf() in the guest.
Signed-off-by: Aaron Lewis <[email protected]>
[sean: use UCALL_EXIT_REASON, enable for all architectures]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Define the expected architecture specific exit reason for a successful
ucall so that common tests can assert that a ucall occurred without the
test needing to implement arch specific code.
Suggested-by: Andrew Jones <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add an architecture specific ucall.h and inline the simple arch hooks,
e.g. the init hook for everything except ARM, and the actual "do ucall"
hook for everything except x86 (which should be simple, but temporarily
isn't due to carrying a workaround).
Having a per-arch ucall header will allow adding a #define for the
expected KVM exit reason for a ucall that is colocated (for everything
except x86) with the ucall itself.
Reviewed-by: Andrew Jones <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add printf-based GUEST_ASSERT macros and accompanying host-side support to
provide an assert-specific versions of GUEST_PRINTF(). To make it easier
to parse assert messages, for humans and bots alike, preserve/use the same
layout as host asserts, e.g. in the example below, the reported expression,
file, line number, and message are from the guest assertion, not the host
reporting of the assertion.
The call stack still captures the host reporting, but capturing the guest
stack is a less pressing concern, i.e. can be done in the future, and an
optimal solution would capture *both* the host and guest stacks, i.e.
capturing the host stack isn't an outright bug.
Running soft int test
==== Test Assertion Failure ====
x86_64/svm_nested_soft_inject_test.c:39: regs->rip != (unsigned long)l2_guest_code_int
pid=214104 tid=214104 errno=4 - Interrupted system call
1 0x0000000000401b35: run_test at svm_nested_soft_inject_test.c:191
2 0x00000000004017d2: main at svm_nested_soft_inject_test.c:212
3 0x0000000000415b03: __libc_start_call_main at libc-start.o:?
4 0x000000000041714f: __libc_start_main_impl at ??:?
5 0x0000000000401660: _start at ??:?
Expected IRQ at RIP 0x401e50, received IRQ at 0x401e50
Don't bother sharing code between ucall_assert() and ucall_fmt(), as
forwarding the variable arguments would either require using macros or
building a va_list, i.e. would make the code less readable and/or require
just as much copy+paste code anyways.
Gate the new macros with a flag so that tests can more or less be switched
over one-by-one. The slow conversion won't be perfect, e.g. library code
won't pick up the flag, but the only asserts in library code are of the
vanilla GUEST_ASSERT() variety, i.e. don't print out variables.
Add a temporary alias to GUEST_ASSERT_1() to fudge around ARM's
arch_timer.h header using GUEST_ASSERT_1(), thus thwarting any attempt to
convert tests one-by-one.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add more flexibility to guest debugging and testing by adding
GUEST_PRINTF() and GUEST_ASSERT_FMT() to the ucall framework.
Add a sized buffer to the ucall structure to hold the formatted string,
i.e. to allow the guest to easily resolve the string, and thus avoid the
ugly pattern of the host side having to make assumptions about the desired
format, as well as having to pass around a large number of parameters.
The buffer size was chosen to accommodate most use cases, and based on
similar usage. E.g. printf() uses the same size buffer in
arch/x86/boot/printf.c. And 1KiB ought to be enough for anybody.
Signed-off-by: Aaron Lewis <[email protected]>
[sean: massage changelog, wrap macro param in ()]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add additional pages to the guest to account for the number of pages
the ucall headers need. The only reason things worked before is the
ucall headers are fairly small. If they were ever to increase in
size the guest could run out of memory.
This is done in preparation for adding string formatting options to
the guest through the ucall framework which increases the size of
the ucall headers.
Fixes: 426729b2cf2e ("KVM: selftests: Add ucall pool based implementation")
Signed-off-by: Aaron Lewis <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add a local version of guest_snprintf() for use in the guest.
Having a local copy allows the guest access to string formatting
options without dependencies on LIBC. LIBC is problematic because
it heavily relies on both AVX-512 instructions and a TLS, neither of
which are guaranteed to be set up in the guest.
The file guest_sprintf.c was lifted from arch/x86/boot/printf.c and
adapted to work in the guest, including the addition of buffer length.
I.e. s/sprintf/snprintf/
The functions where prefixed with "guest_" to allow guests to
explicitly call them.
A string formatted by this function is expected to succeed or die. If
something goes wrong during the formatting process a GUEST_ASSERT()
will be thrown.
Signed-off-by: Aaron Lewis <[email protected]>
Link: https://lore.kernel.org/all/mtdi6smhur5rqffvpu7qux7mptonw223y2653x2nwzvgm72nlo@zyc4w3kwl3rg
[sean: add a link to the discussion of other options]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Add strnlen() to the string overrides to allow it to be called in the
guest.
The implementation for strnlen() was taken from the kernel's generic
version, lib/string.c.
This will be needed when printf() is introduced.
Signed-off-by: Aaron Lewis <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Preserve or clobber all GPRs (except RIP and RSP, as they're saved and
restored via the VMCS) when performing a ucall on x86 to fudge around a
horrific long-standing bug in selftests' nested VMX support where L2's
GPRs are not preserved across a nested VM-Exit. I.e. if a test triggers a
nested VM-Exit to L1 in response to a ucall, e.g. GUEST_SYNC(), then L2's
GPR state can be corrupted.
The issues manifests as an unexpected #GP in clear_bit() when running the
hyperv_evmcs test due to RBX being used to track the ucall object, and RBX
being clobbered by the nested VM-Exit. The problematic hyperv_evmcs
testcase is where L0 (test's host userspace) injects an NMI in response to
GUEST_SYNC(8) from L2, but the bug could "randomly" manifest in any test
that induces a nested VM-Exit from L0. The bug hasn't caused failures in
the past due to sheer dumb luck.
The obvious fix is to rework the nVMX helpers to save/restore L2 GPRs
across VM-Exit and VM-Enter, but that is a much bigger task and carries
its own risks, e.g. nSVM does save/restore GPRs, but not in a thread-safe
manner, and there is a _lot_ of cleanup that can be done to unify code
for doing VM-Enter on nVMX, nSVM, and eVMCS.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Clean up TEST_ASSERT_EQ() so that the (mostly) raw code is captured in the
main assert message, not the helper macro's code. E.g. make this:
x86_64/tsc_msrs_test.c:106: __a == __b
pid=40470 tid=40470 errno=0 - Success
1 0x000000000040170e: main at tsc_msrs_test.c:106
2 0x0000000000416f23: __libc_start_call_main at libc-start.o:?
3 0x000000000041856f: __libc_start_main_impl at ??:?
4 0x0000000000401ef0: _start at ??:?
TEST_ASSERT_EQ(rounded_host_rdmsr(MSR_IA32_TSC), val + 1) failed.
rounded_host_rdmsr(MSR_IA32_TSC) is 0
val + 1 is 0x1
look like this:
x86_64/tsc_msrs_test.c:106: rounded_host_rdmsr(MSR_IA32_TSC) == val + 1
pid=5737 tid=5737 errno=0 - Success
1 0x0000000000401714: main at tsc_msrs_test.c:106
2 0x0000000000415c23: __libc_start_call_main at libc-start.o:?
3 0x000000000041726f: __libc_start_main_impl at ??:?
4 0x0000000000401e60: _start at ??:?
0 != 0x1 (rounded_host_rdmsr(MSR_IA32_TSC) != val + 1)
Opportunstically clean up the formatting of the entire macro.
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
There is already an ASSERT_EQ macro in the file
tools/testing/selftests/kselftest_harness.h, so currently KVM selftests
can't include test_util.h from the KVM selftests together with that file.
Rename the macro in the KVM selftests to TEST_ASSERT_EQ to avoid the
problem - it is also more similar to the other macros in test_util.h that
way.
Suggested-by: Sean Christopherson <[email protected]>
Signed-off-by: Thomas Huth <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Don't nullify "nodep" to NULL one line before it's set to "tmp".
Signed-off-by: Minjie Du <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
[sean: massage shortlog+changelog]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
With test case kvm_page_table_test, start time is acquired with
time type CLOCK_MONOTONIC_RAW, however end time in timespec_elapsed()
is acquired with time type CLOCK_MONOTONIC. This can cause inaccurate
elapsed time calculation due to mixing timebases, e.g. LoongArch in
particular will see weirdness.
Modify kvm_page_table_test to use unified time type CLOCK_MONOTONIC for
start time.
Signed-off-by: Bibo Mao <[email protected]>
Reviewed-by: Oliver Upton <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
[sean: massage changelog]
Signed-off-by: Sean Christopherson <[email protected]>
|
|
Attempt to set the to-be-queued exception to be both pending and injected
_after_ KVM_CAP_SYNC_REGS's kvm_vcpu_ioctl_x86_set_vcpu_events() squashes
the pending exception (if there's also an injected exception). Buggy KVM
versions will eventually yell loudly about having impossible state when
processing queued excpetions, e.g.
WARNING: CPU: 0 PID: 1115 at arch/x86/kvm/x86.c:10095 kvm_check_and_inject_events+0x220/0x500 [kvm]
arch/x86/kvm/x86.c:kvm_check_and_inject_events():
WARN_ON_ONCE(vcpu->arch.exception.injected &&
vcpu->arch.exception.pending);
Signed-off-by: Michal Luczaj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
[sean: split to separate patch, massage changelog and comment]
Signed-off-by: Sean Christopherson <[email protected]>
|