| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
BPF trampoline can be made to work with existing 5 bytes of BPF program
prologue, but let's add 5 bytes of NOPs to the beginning of every JITed BPF
program to make BPF trampoline job easier. They can be removed in the future.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20191114185720.1641606-14-ast@kernel.org
|
|
Introduce BPF trampoline concept to allow kernel code to call into BPF programs
with practically zero overhead. The trampoline generation logic is
architecture dependent. It's converting native calling convention into BPF
calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The
registers R1 to R5 are used to pass arguments into BPF functions. The main BPF
program accepts only single argument "ctx" in R1. Whereas CPU native calling
convention is different. x86-64 is passing first 6 arguments in registers
and the rest on the stack. x86-32 is passing first 3 arguments in registers.
sparc64 is passing first 6 in registers. And so on.
The trampolines between BPF and kernel already exist. BPF_CALL_x macros in
include/linux/filter.h statically compile trampolines from BPF into kernel
helpers. They convert up to five u64 arguments into kernel C pointers and
integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On
32-bit architecture they're meaningful.
The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and
__bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert
kernel function arguments into array of u64s that BPF program consumes via
R1=ctx pointer.
This patch set is doing the same job as __bpf_trace_##call() static
trampolines, but dynamically for any kernel function. There are ~22k global
kernel functions that are attachable via nop at function entry. The function
arguments and types are described in BTF. The job of btf_distill_func_proto()
function is to extract useful information from BTF into "function model" that
architecture dependent trampoline generators will use to generate assembly code
to cast kernel function arguments into array of u64s. For example the kernel
function eth_type_trans has two pointers. They will be casted to u64 and stored
into stack of generated trampoline. The pointer to that stack space will be
passed into BPF program in R1. On x86-64 such generated trampoline will consume
16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will
make sure that only two u64 are accessed read-only by BPF program. The verifier
will also recognize the precise type of the pointers being accessed and will
not allow typecasting of the pointer to a different type within BPF program.
The tracing use case in the datacenter demonstrated that certain key kernel
functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always
active. Other functions have both kprobe and kretprobe. So it is essential to
keep both kernel code and BPF programs executing at maximum speed. Hence
generated BPF trampoline is re-generated every time new program is attached or
detached to maintain maximum performance.
To avoid the high cost of retpoline the attached BPF programs are called
directly. __bpf_prog_enter/exit() are used to support per-program execution
stats. In the future this logic will be optimized further by adding support
for bpf_stats_enabled_key inside generated assembly code. Introduction of
preemptible and sleepable BPF programs will completely remove the need to call
to __bpf_prog_enter/exit().
Detach of a BPF program from the trampoline should not fail. To avoid memory
allocation in detach path the half of the page is used as a reserve and flipped
after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly
which is enough for BPF tracing use cases. This limit can be increased in the
future.
BPF_TRACE_FENTRY programs have access to raw kernel function arguments while
BPF_TRACE_FEXIT programs have access to kernel return value as well. Often
kprobe BPF program remembers function arguments in a map while kretprobe
fetches arguments from a map and analyzes them together with return value.
BPF_TRACE_FEXIT accelerates this typical use case.
Recursion prevention for kprobe BPF programs is done via per-cpu
bpf_prog_active counter. In practice that turned out to be a mistake. It
caused programs to randomly skip execution. The tracing tools missed results
they were looking for. Hence BPF trampoline doesn't provide builtin recursion
prevention. It's a job of BPF program itself and will be addressed in the
follow up patches.
BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases
in the future. For example to remove retpoline cost from XDP programs.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
|
|
Add bpf_arch_text_poke() helper that is used by BPF trampoline logic to patch
nops/calls in kernel text into calls into BPF trampoline and to patch
calls/nops inside BPF programs too.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20191114185720.1641606-4-ast@kernel.org
|
|
Refactor x86 JITing of LDX, STX, CALL instructions into separate helper
functions. No functional changes in LDX and STX helpers. There is a minor
change in CALL helper. It will populate target address correctly on the first
pass of JIT instead of second pass. That won't reduce total number of JIT
passes though.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20191114185720.1641606-3-ast@kernel.org
|
|
In preparation for static_call and variable size jump_label support,
teach text_poke_bp() to emulate instructions, namely:
JMP32, JMP8, CALL, NOP2, NOP_ATOMIC5, INT3
The current text_poke_bp() takes a @handler argument which is used as
a jump target when the temporary INT3 is hit by a different CPU.
When patching CALL instructions, this doesn't work because we'd miss
the PUSH of the return address. Instead, teach poke_int3_handler() to
emulate an instruction, typically the instruction we're patching in.
This fits almost all text_poke_bp() users, except
arch_unoptimize_kprobe() which restores random text, and for that site
we have to build an explicit emulate instruction.
Tested-by: Alexei Starovoitov <ast@kernel.org>
Tested-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20191111132457.529086974@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 8c7eebc10687af45ac8e40ad1bac0cf7893dba9f)
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
The x86_capability array in cpuinfo_x86 is of type u32 and thus is
naturally aligned to 4 bytes. But, set_bit() and clear_bit() require the
array to be aligned to size of unsigned long (i.e. 8 bytes on 64-bit
systems).
The array pointer is handed into atomic bit operations. If the access is
not aligned to unsigned long then the atomic bit operations can end up
crossing a cache line boundary, which causes the CPU to do a full bus lock
as it can't lock both cache lines at once. The bus lock operation is heavy
weight and can cause severe performance degradation.
The upcoming #AC split lock detection mechanism will issue warnings for
this kind of access.
Force the alignment of the array to unsigned long. This avoids the massive
code changes which would be required when converting the array data type to
unsigned long.
[ tglx: Rewrote changelog so it contains information WHY this is required ]
Suggested-by: David Laight <David.Laight@aculab.com>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190916223958.27048-4-tony.luck@intel.com
|
|
cpu_caps_cleared[] and cpu_caps_set[] are arrays of type u32 and therefore
naturally aligned to 4 bytes, which is also unsigned long aligned on
32-bit, but not on 64-bit.
The array pointer is handed into atomic bit operations. If the access not
aligned to unsigned long then the atomic bit operations can end up crossing
a cache line boundary, which causes the CPU to do a full bus lock as it
can't lock both cache lines at once. The bus lock operation is heavy weight
and can cause severe performance degradation.
The upcoming #AC split lock detection mechanism will issue warnings for
this kind of access.
Force the alignment of these arrays to unsigned long. This avoids the
massive code changes which would be required when converting the array data
type to unsigned long.
[ tglx: Rewrote changelog ]
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20190916223958.27048-2-tony.luck@intel.com
|
|
There are two structures based on time_t that conflict between libc and
kernel: timeval and timespec. Both are now renamed to __kernel_old_timeval
and __kernel_old_timespec.
For time_t, the old typedef is still __kernel_time_t. There is nothing
wrong with that name, but it would be nice to not use that going forward
as this type is used almost only in deprecated interfaces because of
the y2038 overflow.
In the IPC headers (msgbuf.h, sembuf.h, shmbuf.h), __kernel_time_t is only
used for the 64-bit variants, which are not deprecated.
Change these to a plain 'long', which is the same type as __kernel_time_t
on all 64-bit architectures anyway, to reduce the number of users of the
old type.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
|
|
Only x86 uses the 'time' syscall in vdso, so change that to
__kernel_old_time_t as a preparation for removing 'time_t' and
'__kernel_time_t' later.
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
|
|
In order to remove 'timespec' completely from the kernel, all
internal uses should be converted to a y2038-safe type, while
those that are only for compatibity with existing user space
should be marked appropriately.
Change vdso to use __kernel_old_timespec in order to avoid
the deprecated type and mark these interfaces as outdated.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
|
|
The gettimeofday() function in vdso uses the traditional 'timeval'
structure layout, which will be incompatible with future versions of
glibc on 32-bit architectures that use a 64-bit time_t.
This interface is problematic for y2038, when time_t overflows on 32-bit
architectures, but the plan so far is that a libc with 64-bit time_t
will not call into the gettimeofday() vdso helper at all, and only
have a method for entering clock_gettime(). This means we don't have
to fix it here, though we probably want to add a new clock_gettime()
entry point using a 64-bit version of 'struct timespec' at some point.
Changing the vdso code to use __kernel_old_timeval helps isolate
this usage from the other ones that still need to be fixed properly,
and it gets us closer to removing the 'timeval' definition from the
kernel sources.
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
|
|
CFLAGS_REMOVE_*.o syntax is used to drop particular flags when
building objects from C files. It has no effect for assembly files.
vdso-note.o is compiled from the assembly file, vdso-note.S, hence
CFLAGS_REMOVE_vdso-note.o is meaningless.
Neither vvar.c nor vvar.S is found in the vdso directory. Since there
is no source file to create vvar.o, CFLAGS_REMOVE_vvar.o is also
meaningless.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: x86-ml <x86@kernel.org>
Link: https://lkml.kernel.org/r/20191114154922.30365-1-yamada.masahiro@socionext.com
|
|
In IOAPIC fixed delivery mode instead of flushing the scan
requests to all vCPUs, we should only send the requests to
vCPUs specified within the destination field.
This patch introduces kvm_get_dest_vcpus_mask() API which
retrieves an array of target vCPUs by using
kvm_apic_map_get_dest_lapic() and then based on the
vcpus_idx, it sets the bit in a bitmap. However, if the above
fails kvm_get_dest_vcpus_mask() finds the target vCPUs by
traversing all available vCPUs. Followed by setting the
bits in the bitmap.
If we had different vCPUs in the previous request for the
same redirection table entry then bits corresponding to
these vCPUs are also set. This to done to keep
ioapic_handled_vectors synchronized.
This bitmap is then eventually passed on to
kvm_make_vcpus_request_mask() to generate a masked request
only for the target vCPUs.
This would enable us to reduce the latency overhead on isolated
vCPUs caused by the IPI to process due to KVM_REQ_IOAPIC_SCAN.
Suggested-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Nitesh Narayan Lal <nitesh@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The L1 hypervisor may include the IA32_TIME_STAMP_COUNTER MSR in the
vmcs12 MSR VM-exit MSR-store area as a way of determining the highest
TSC value that might have been observed by L2 prior to VM-exit. The
current implementation does not capture a very tight bound on this
value. To tighten the bound, add the IA32_TIME_STAMP_COUNTER MSR to the
vmcs02 VM-exit MSR-store area whenever it appears in the vmcs12 VM-exit
MSR-store area. When L0 processes the vmcs12 VM-exit MSR-store area
during the emulation of an L2->L1 VM-exit, special-case the
IA32_TIME_STAMP_COUNTER MSR, using the value stored in the vmcs02
VM-exit MSR-store area to derive the value to be stored in the vmcs12
VM-exit MSR-store area.
Reviewed-by: Liran Alon <liran.alon@oracle.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Rename function find_msr() to vmx_find_msr_index() in preparation for an
upcoming patch where we export it and use it in nested.c.
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Rename NR_AUTOLOAD_MSRS to NR_LOADSTORE_MSRS. This needs to be done
due to the addition of the MSR-autostore area that will be added in a
future patch. After that the name AUTOLOAD will no longer make sense.
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Add the function read_and_check_msr_entry() which just pulls some code
out of nested_vmx_store_msr(). This will be useful as reusable code in
upcoming patches.
Reviewed-by: Liran Alon <liran.alon@oracle.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Correct a small inaccuracy in the shattering of vmx.c, which becomes
visible now that pmu_intel.c includes nested.h.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The "load IA32_PERF_GLOBAL_CTRL" bit for VM-entry and VM-exit should
only be exposed to the guest if IA32_PERF_GLOBAL_CTRL is a valid MSR.
Create a new helper to allow pmu_refresh() to update the VM-Entry and
VM-Exit controls to ensure PMU values are initialized when performing
the is_valid_msr() check.
Suggested-by: Jim Mattson <jmattson@google.com>
Co-developed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Add condition to prepare_vmcs02 which loads IA32_PERF_GLOBAL_CTRL on
VM-entry if the "load IA32_PERF_GLOBAL_CTRL" bit on the VM-entry control
is set. Use SET_MSR_OR_WARN() rather than directly writing to the field
to avoid overwrite by atomic_switch_perf_msrs().
Suggested-by: Jim Mattson <jmattson@google.com>
Co-developed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The existing implementation for loading the IA32_PERF_GLOBAL_CTRL MSR
on VM-exit was incorrect, as the next call to atomic_switch_perf_msrs()
could cause this value to be overwritten. Instead, call kvm_set_msr()
which will allow atomic_switch_perf_msrs() to correctly set the values.
Define a macro, SET_MSR_OR_WARN(), to set the MSR with kvm_set_msr()
and WARN on failure.
Suggested-by: Jim Mattson <jmattson@google.com>
Co-developed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Add a consistency check on nested vm-entry for host's
IA32_PERF_GLOBAL_CTRL from vmcs12. Per Intel's SDM Vol 3 26.2.2:
If the "load IA32_PERF_GLOBAL_CTRL"
VM-exit control is 1, bits reserved in the IA32_PERF_GLOBAL_CTRL
MSR must be 0 in the field for that register"
Suggested-by: Jim Mattson <jmattson@google.com>
Co-developed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Add condition to nested_vmx_check_guest_state() to check the validity of
GUEST_IA32_PERF_GLOBAL_CTRL. Per Intel's SDM Vol 3 26.3.1.1:
If the "load IA32_PERF_GLOBAL_CTRL" VM-entry control is 1, bits
reserved in the IA32_PERF_GLOBAL_CTRL MSR must be 0 in the field for that
register.
Suggested-by: Jim Mattson <jmattson@google.com>
Co-developed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Create a helper function to check the validity of a proposed value for
IA32_PERF_GLOBAL_CTRL from the existing check in intel_pmu_set_msr().
Per Intel's SDM, the reserved bits in IA32_PERF_GLOBAL_CTRL must be
cleared for the corresponding host/guest state fields.
Suggested-by: Jim Mattson <jmattson@google.com>
Co-developed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
When KVM emulates a nested VMEntry (L1->L2 VMEntry), it switches mmu root
page. If nEPT is used, this will happen from
kvm_init_shadow_ept_mmu()->__kvm_mmu_new_cr3() and otherwise it will
happpen from nested_vmx_load_cr3()->kvm_mmu_new_cr3(). Either case,
__kvm_mmu_new_cr3() will use fast_cr3_switch() in attempt to switch to a
previously cached root page.
In case fast_cr3_switch() finds a matching cached root page, it will
set it in mmu->root_hpa and request KVM_REQ_LOAD_CR3 such that on
next entry to guest, KVM will set root HPA in appropriate hardware
fields (e.g. vmcs->eptp). In addition, fast_cr3_switch() calls
kvm_x86_ops->tlb_flush() in order to flush TLB as MMU root page
was replaced.
This works as mmu->root_hpa, which vmx_flush_tlb() use, was
already replaced in cached_root_available(). However, this may
result in unnecessary INVEPT execution because a KVM_REQ_TLB_FLUSH
may have already been requested. For example, by prepare_vmcs02()
in case L1 don't use VPID.
Therefore, change fast_cr3_switch() to just request TLB flush on
next entry to guest.
Reviewed-by: Bhavesh Davda <bhavesh.davda@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Currently, a host perf_event is created for a vPMC functionality emulation.
It’s unpredictable to determine if a disabled perf_event will be reused.
If they are disabled and are not reused for a considerable period of time,
those obsolete perf_events would increase host context switch overhead that
could have been avoided.
If the guest doesn't WRMSR any of the vPMC's MSRs during an entire vcpu
sched time slice, and its independent enable bit of the vPMC isn't set,
we can predict that the guest has finished the use of this vPMC, and then
do request KVM_REQ_PMU in kvm_arch_sched_in and release those perf_events
in the first call of kvm_pmu_handle_event() after the vcpu is scheduled in.
This lazy mechanism delays the event release time to the beginning of the
next scheduled time slice if vPMC's MSRs aren't changed during this time
slice. If guest comes back to use this vPMC in next time slice, a new perf
event would be re-created via perf_event_create_kernel_counter() as usual.
Suggested-by: Wei Wang <wei.w.wang@intel.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Like Xu <like.xu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The perf_event_create_kernel_counter() in the pmc_reprogram_counter() is
a heavyweight and high-frequency operation, especially when host disables
the watchdog (maximum 21000000 ns) which leads to an unacceptable latency
of the guest NMI handler. It limits the use of vPMUs in the guest.
When a vPMC is fully enabled, the legacy reprogram_*_counter() would stop
and release its existing perf_event (if any) every time EVEN in most cases
almost the same requested perf_event will be created and configured again.
For each vPMC, if the reuqested config ('u64 eventsel' for gp and 'u8 ctrl'
for fixed) is the same as its current config AND a new sample period based
on pmc->counter is accepted by host perf interface, the current event could
be reused safely as a new created one does. Otherwise, do release the
undesirable perf_event and reprogram a new one as usual.
It's light-weight to call pmc_pause_counter (disable, read and reset event)
and pmc_resume_counter (recalibrate period and re-enable event) as guest
expects instead of release-and-create again on any condition. Compared to
use the filterable event->attr or hw.config, a new 'u64 current_config'
field is added to save the last original programed config for each vPMC.
Based on this implementation, the number of calls to pmc_reprogram_counter
is reduced by ~82.5% for a gp sampling event and ~99.9% for a fixed event.
In the usage of multiplexing perf sampling mode, the average latency of the
guest NMI handler is reduced from 104923 ns to 48393 ns (~2.16x speed up).
If host disables watchdog, the minimum latecy of guest NMI handler could be
speed up at ~3413x (from 20407603 to 5979 ns) and at ~786x in the average.
Suggested-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Like Xu <like.xu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Introduce a new callback msr_idx_to_pmc that returns a struct kvm_pmc*,
and change kvm_pmu_is_valid_msr to return ".msr_idx_to_pmc(vcpu, msr) ||
.is_valid_msr(vcpu, msr)" and AMD just returns false from .is_valid_msr.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Like Xu <like.xu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
The leagcy pmu_ops->msr_idx_to_pmc is only called in kvm_pmu_rdpmc, so
this function actually receives the contents of ECX before RDPMC, and
translates it to a kvm_pmc. Let's clarify its semantic by renaming the
existing msr_idx_to_pmc to rdpmc_ecx_to_pmc, and is_valid_msr_idx to
is_valid_rdpmc_ecx; likewise for the wrapper kvm_pmu_is_valid_msr_idx.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Like Xu <like.xu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
When L1 don't use TPR-Shadow to run L2, L0 configures vmcs02 without
TPR-Shadow and install intercepts on CR8 access (load and store).
If L1 do not intercept L2 CR8 access, L0 intercepts on those accesses
will emulate load/store on L1's LAPIC TPR. If in this case L2 lowers
TPR such that there is now an injectable interrupt to L1,
apic_update_ppr() will request a KVM_REQ_EVENT which will trigger a call
to update_cr8_intercept() to update TPR-Threshold to highest pending IRR
priority.
However, this update to TPR-Threshold is done while active vmcs is
vmcs02 instead of vmcs01. Thus, when later at some point L0 will
emulate an exit from L2 to L1, L1 will still run with high
TPR-Threshold. This will result in every VMEntry to L1 to immediately
exit on TPR_BELOW_THRESHOLD and continue to do so infinitely until
some condition will cause KVM_REQ_EVENT to be set.
(Note that TPR_BELOW_THRESHOLD exit handler do not set KVM_REQ_EVENT
until apic_update_ppr() will notice a new injectable interrupt for PPR)
To fix this issue, change update_cr8_intercept() such that if L2 lowers
L1's TPR in a way that requires to lower L1's TPR-Threshold, save update
to TPR-Threshold and apply it to vmcs01 when L0 emulates an exit from
L2 to L1.
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
No functional changes.
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
This check is unnecessary as x86 update_cr8_intercept() which calls
this VMX/SVM specific callback already performs this check.
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
There are some duplicate code in kvm_pv_send_ipi when deal with ipi
bitmap. Add helper func to remove it, and eliminate odd out label,
get rid of unnecessary kvm_lapic_irq field init and so on.
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
When SYSCALL/SYSENTER ability check failed, cs and ss is inited but
remain not used. Delay initializing cs and ss until SYSCALL/SYSENTER
ability check passed.
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Intel SDM section 25.2 OTHER CAUSES OF VM EXITS specifies the following
on INIT signals: "Such exits do not modify register state or clear pending
events as they would outside of VMX operation."
When commit 4b9852f4f389 ("KVM: x86: Fix INIT signal handling in various CPU states")
was applied, I interepted above Intel SDM statement such that
INIT_SIGNAL exit don’t consume the LAPIC INIT pending event.
However, when Nadav Amit run matching kvm-unit-test on a bare-metal
machine, it turned out my interpetation was wrong. i.e. INIT_SIGNAL
exit does consume the LAPIC INIT pending event.
(See: https://www.spinics.net/lists/kvm/msg196757.html)
Therefore, fix KVM code to behave as observed on bare-metal.
Fixes: 4b9852f4f389 ("KVM: x86: Fix INIT signal handling in various CPU states")
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Mihai Carabas <mihai.carabas@oracle.com>
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Commit 4b9852f4f389 ("KVM: x86: Fix INIT signal handling in various CPU states")
fixed KVM to also latch pending LAPIC INIT event when vCPU is in VMX
operation.
However, current API of KVM_SET_MP_STATE allows userspace to put vCPU
into KVM_MP_STATE_SIPI_RECEIVED or KVM_MP_STATE_INIT_RECEIVED even when
vCPU is in VMX operation.
Fix this by introducing a util method to check if vCPU state latch INIT
signals and use it in KVM_SET_MP_STATE handler.
Fixes: 4b9852f4f389 ("KVM: x86: Fix INIT signal handling in various CPU states")
Reported-by: Sean Christopherson <sean.j.christopherson@intel.com>
Reviewed-by: Mihai Carabas <mihai.carabas@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Commit 4b9852f4f389 ("KVM: x86: Fix INIT signal handling in various CPU states")
fixed KVM to also latch pending LAPIC INIT event when vCPU is in VMX
operation.
However, current API of KVM_SET_VCPU_EVENTS defines this field as
part of SMM state and only set pending LAPIC INIT event if vCPU is
specified to be in SMM mode (events->smi.smm is set).
Change KVM_SET_VCPU_EVENTS handler to set pending LAPIC INIT event
by latched_init field regardless of if vCPU is in SMM mode or not.
Fixes: 4b9852f4f389 ("KVM: x86: Fix INIT signal handling in various CPU states")
Reviewed-by: Mihai Carabas <mihai.carabas@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
It's enough to check the value and issue the direct call.
After this commit is applied, here the most common retpolines executed
under a high resolution timer workload in the guest on a VMX host:
[..]
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 267
@[]: 2256
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
__kvm_wait_lapic_expire+284
vmx_vcpu_run.part.97+1091
vcpu_enter_guest+377
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 2390
@[]: 33410
@total: 315707
Note the highest hit above is __delay so probably not worth optimizing
even if it would be more frequent than 2k hits per sec.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
It's enough to check the exit value and issue a direct call to avoid
the retpoline for all the common vmexit reasons.
After this commit is applied, here the most common retpolines executed
under a high resolution timer workload in the guest on a SVM host:
[..]
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
ktime_get_update_offsets_now+70
hrtimer_interrupt+131
smp_apic_timer_interrupt+106
apic_timer_interrupt+15
start_sw_timer+359
restart_apic_timer+85
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 1940
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_r12+33
force_qs_rnp+217
rcu_gp_kthread+1270
kthread+268
ret_from_fork+34
]: 4644
@[]: 25095
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
lapic_next_event+28
clockevents_program_event+148
hrtimer_start_range_ns+528
start_sw_timer+356
restart_apic_timer+85
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 41474
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
clockevents_program_event+148
hrtimer_start_range_ns+528
start_sw_timer+356
restart_apic_timer+85
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 41474
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
ktime_get+58
clockevents_program_event+84
hrtimer_start_range_ns+528
start_sw_timer+356
restart_apic_timer+85
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 41887
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
lapic_next_event+28
clockevents_program_event+148
hrtimer_try_to_cancel+168
hrtimer_cancel+21
kvm_set_lapic_tscdeadline_msr+43
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 42723
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
clockevents_program_event+148
hrtimer_try_to_cancel+168
hrtimer_cancel+21
kvm_set_lapic_tscdeadline_msr+43
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 42766
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
ktime_get+58
clockevents_program_event+84
hrtimer_try_to_cancel+168
hrtimer_cancel+21
kvm_set_lapic_tscdeadline_msr+43
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 42848
@[
trace_retpoline+1
__trace_retpoline+30
__x86_indirect_thunk_rax+33
ktime_get+58
start_sw_timer+279
restart_apic_timer+85
kvm_set_msr_common+1497
msr_interception+142
vcpu_enter_guest+684
kvm_arch_vcpu_ioctl_run+261
kvm_vcpu_ioctl+559
do_vfs_ioctl+164
ksys_ioctl+96
__x64_sys_ioctl+22
do_syscall_64+89
entry_SYSCALL_64_after_hwframe+68
]: 499845
@total: 1780243
SVM has no TSC based programmable preemption timer so it is invoking
ktime_get() frequently.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
It's enough to check the exit value and issue a direct call to avoid
the retpoline for all the common vmexit reasons.
Of course CONFIG_RETPOLINE already forbids gcc to use indirect jumps
while compiling all switch() statements, however switch() would still
allow the compiler to bisect the case value. It's more efficient to
prioritize the most frequent vmexits instead.
The halt may be slow paths from the point of the guest, but not
necessarily so from the point of the host if the host runs at full CPU
capacity and no host CPU is ever left idle.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
Eliminate wasteful call/ret non RETPOLINE case and unnecessary fentry
dynamic tracing hooking points.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
pci.h is not a UAPI header, so the __KERNEL__ ifdef is rather pointless.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20191113071836.21041-4-hch@lst.de
|
|
This file only contains external declarations for two non-existing
function pointers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20191113071836.21041-3-hch@lst.de
|
|
The calgary IOMMU was only used on high-end IBM systems in the early
x86_64 age and has no known users left. Remove it to avoid having to
touch it for pending changes to the DMA API.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20191113071836.21041-2-hch@lst.de
|
|
Hyper-V has historically initialized stimer-based clockevents late in the
process of onlining a CPU because clockevents depend on stimer
interrupts. In the original Hyper-V design, stimer interrupts generate a
VMbus message, so the VMbus machinery must be running first, and VMbus
can't be initialized until relatively late. On x86/64, LAPIC timer based
clockevents are used during early initialization before VMbus and
stimer-based clockevents are ready, and again during CPU offlining after
the stimer clockevents have been shut down.
Unfortunately, this design creates problems when offlining CPUs for
hibernation or other purposes. stimer-based clockevents are shut down
relatively early in the offlining process, so clockevents_unbind_device()
must be used to fallback to the LAPIC-based clockevents for the remainder
of the offlining process. Furthermore, the late initialization and early
shutdown of stimer-based clockevents doesn't work well on ARM64 since there
is no other timer like the LAPIC to fallback to. So CPU onlining and
offlining doesn't work properly.
Fix this by recognizing that stimer Direct Mode is the normal path for
newer versions of Hyper-V on x86/64, and the only path on other
architectures. With stimer Direct Mode, stimer interrupts don't require any
VMbus machinery. stimer clockevents can be initialized and shut down
consistent with how it is done for other clockevent devices. While the old
VMbus-based stimer interrupts must still be supported for backward
compatibility on x86, that mode of operation can be treated as legacy.
So add a new Hyper-V stimer entry in the CPU hotplug state list, and use
that new state when in Direct Mode. Update the Hyper-V clocksource driver
to allocate and initialize stimer clockevents earlier during boot. Update
Hyper-V initialization and the VMbus driver to use this new design. As a
result, the LAPIC timer is no longer used during boot or CPU
onlining/offlining and clockevents_unbind_device() is not called. But
retain the old design as a legacy implementation for older versions of
Hyper-V that don't support Direct Mode.
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Dexuan Cui <decui@microsoft.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Link: https://lkml.kernel.org/r/1573607467-9456-1-git-send-email-mikelley@microsoft.com
|
|
Pick up upstream fixes to avoid conflicts.
|
|
... or let function calls stick out and thus remain on a single line,
even if the 80 cols rule is violated by a couple of chars, for better
readability.
No functional changes.
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: x86@kernel.org
Link: https://lkml.kernel.org/r/20191114172200.19563-1-bp@alien8.de
|
|
When the crashkernel kernel command line option is specified, the low
1M memory will always be reserved now. Therefore, it's not necessary to
create a backup region anymore and also no need to copy the contents of
the first 640k to it.
Remove all the code related to handling that backup region.
[ bp: Massage commit message. ]
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: bhe@redhat.com
Cc: Dave Young <dyoung@redhat.com>
Cc: d.hatayama@fujitsu.com
Cc: dhowells@redhat.com
Cc: ebiederm@xmission.com
Cc: horms@verge.net.au
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jürgen Gross <jgross@suse.com>
Cc: kexec@lists.infradead.org
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: vgoyal@redhat.com
Cc: x86-ml <x86@kernel.org>
Link: https://lkml.kernel.org/r/20191108090027.11082-3-lijiang@redhat.com
|
|
Acquire the per-VM slots_lock when zapping all shadow pages as part of
toggling nx_huge_pages. The fast zap algorithm relies on exclusivity
(via slots_lock) to identify obsolete vs. valid shadow pages, because it
uses a single bit for its generation number. Holding slots_lock also
obviates the need to acquire a read lock on the VM's srcu.
Failing to take slots_lock when toggling nx_huge_pages allows multiple
instances of kvm_mmu_zap_all_fast() to run concurrently, as the other
user, KVM_SET_USER_MEMORY_REGION, does not take the global kvm_lock.
(kvm_mmu_zap_all_fast() does take kvm->mmu_lock, but it can be
temporarily dropped by kvm_zap_obsolete_pages(), so it is not enough
to enforce exclusivity).
Concurrent fast zap instances causes obsolete shadow pages to be
incorrectly identified as valid due to the single bit generation number
wrapping, which results in stale shadow pages being left in KVM's MMU
and leads to all sorts of undesirable behavior.
The bug is easily confirmed by running with CONFIG_PROVE_LOCKING and
toggling nx_huge_pages via its module param.
Note, until commit 4ae5acbc4936 ("KVM: x86/mmu: Take slots_lock when
using kvm_mmu_zap_all_fast()", 2019-11-13) the fast zap algorithm used
an ulong-sized generation instead of relying on exclusivity for
correctness, but all callers except the recently added set_nx_huge_pages()
needed to hold slots_lock anyways. Therefore, this patch does not have
to be backported to stable kernels.
Given that toggling nx_huge_pages is by no means a fast path, force it
to conform to the current approach instead of reintroducing the previous
generation count.
Fixes: b8e8c8303ff28 ("kvm: mmu: ITLB_MULTIHIT mitigation", but NOT FOR STABLE)
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
|
|
On x86, purgatory() copies the first 640K of memory to a backup region
because the kernel needs those first 640K for the real mode trampoline
during boot, among others.
However, when SME is enabled, the kernel cannot properly copy the old
memory to the backup area but reads only its encrypted contents. The
result is that the crash tool gets invalid pointers when parsing vmcore:
crash> kmem -s|grep -i invalid
kmem: dma-kmalloc-512: slab:ffffd77680001c00 invalid freepointer:a6086ac099f0c5a4
kmem: dma-kmalloc-512: slab:ffffd77680001c00 invalid freepointer:a6086ac099f0c5a4
crash>
So reserve the remaining low 1M memory when the crashkernel option is
specified (after reserving real mode memory) so that allocated memory
does not fall into the low 1M area and thus the copying of the contents
of the first 640k to a backup region in purgatory() can be avoided
altogether.
This way, it does not need to be included in crash dumps or used for
anything except the trampolines that must live in the low 1M.
[ bp: Heavily rewrite commit message, flip check logic in
crash_reserve_low_1M().]
Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: bhe@redhat.com
Cc: Dave Young <dyoung@redhat.com>
Cc: d.hatayama@fujitsu.com
Cc: dhowells@redhat.com
Cc: ebiederm@xmission.com
Cc: horms@verge.net.au
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jürgen Gross <jgross@suse.com>
Cc: kexec@lists.infradead.org
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: vgoyal@redhat.com
Cc: x86-ml <x86@kernel.org>
Link: https://lkml.kernel.org/r/20191108090027.11082-2-lijiang@redhat.com
Link: https://bugzilla.kernel.org/show_bug.cgi?id=204793
|
Alexei Starovoitov
Peter Zijlstra
Fenghua Yu
Arnd Bergmann
Masahiro Yamada
Nitesh Narayan Lal
Aaron Lewis
Paolo Bonzini
Oliver Upton
Liran Alon
Like Xu
Miaohe Lin
Andrea Arcangeli
Christoph Hellwig
Michael Kelley
Thomas Gleixner
Borislav Petkov
Lianbo Jiang
Sean Christopherson