linux-2.6-microblaze.git
3 years agokvm: x86/mmu: Add TDP MMU SPTE changed trace point
Ben Gardon [Tue, 27 Oct 2020 17:59:44 +0000 (10:59 -0700)]
kvm: x86/mmu: Add TDP MMU SPTE changed trace point

Add an extremely verbose trace point to the TDP MMU to log all SPTE
changes, regardless of callstack / motivation. This is useful when a
complete picture of the paging structure is needed or a change cannot be
explained with the other, existing trace points.

Tested: ran the demand paging selftest on an Intel Skylake machine with
all the trace points used by the TDP MMU enabled and observed
them firing with expected values.

This patch can be viewed in Gerrit at:
https://linux-review.googlesource.com/c/virt/kvm/kvm/+/3813

Signed-off-by: Ben Gardon <bgardon@google.com>
Message-Id: <20201027175944.1183301-2-bgardon@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokvm: x86/mmu: Add existing trace points to TDP MMU
Ben Gardon [Tue, 27 Oct 2020 17:59:43 +0000 (10:59 -0700)]
kvm: x86/mmu: Add existing trace points to TDP MMU

The TDP MMU was initially implemented without some of the usual
tracepoints found in mmu.c. Correct this discrepancy by adding the
missing trace points to the TDP MMU.

Tested: ran the demand paging selftest on an Intel Skylake machine with
all the trace points used by the TDP MMU enabled and observed
them firing with expected values.

This patch can be viewed in Gerrit at:
https://linux-review.googlesource.com/c/virt/kvm/kvm/+/3812

Signed-off-by: Ben Gardon <bgardon@google.com>
Message-Id: <20201027175944.1183301-1-bgardon@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: SVM: check CR4 changes against vcpu->arch
Paolo Bonzini [Sun, 15 Nov 2020 14:44:18 +0000 (09:44 -0500)]
KVM: SVM: check CR4 changes against vcpu->arch

Similarly to what vmx/vmx.c does, use vcpu->arch.cr4 to check if CR4
bits PGE, PKE and OSXSAVE have changed.  When switching between VMCB01
and VMCB02, CPUID has to be adjusted every time if CR4.PKE or CR4.OSXSAVE
change; without this patch, instead, CR4 would be checked against the
previous value for L2 on vmentry, and against the previous value for
L1 on vmexit, and CPUID would not be updated.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: SVM: Move asid to vcpu_svm
Cathy Avery [Sun, 11 Oct 2020 18:48:17 +0000 (14:48 -0400)]
KVM: SVM: Move asid to vcpu_svm

KVM does not have separate ASIDs for L1 and L2; either the nested
hypervisor and nested guests share a single ASID, or on older processor
the ASID is used only to implement TLB flushing.

Either way, ASIDs are handled at the VM level.  In preparation
for having different VMCBs passed to VMLOAD/VMRUN/VMSAVE for L1 and
L2, store the current ASID to struct vcpu_svm and only move it to
the VMCB in svm_vcpu_run.  This way, TLB flushes can be applied
no matter which VMCB will be active during the next svm_vcpu_run.

Signed-off-by: Cathy Avery <cavery@redhat.com>
Message-Id: <20201011184818.3609-2-cavery@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agox86/kvm: remove unused macro HV_CLOCK_SIZE
Alex Shi [Fri, 6 Nov 2020 08:39:23 +0000 (16:39 +0800)]
x86/kvm: remove unused macro HV_CLOCK_SIZE

This macro is useless, and could cause gcc warning:
arch/x86/kernel/kvmclock.c:47:0: warning: macro "HV_CLOCK_SIZE" is not
used [-Wunused-macros]
Let's remove it.

Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
Cc: Wanpeng Li <wanpengli@tencent.com>
Cc: Jim Mattson <jmattson@google.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: x86@kernel.org
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: kvm@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Message-Id: <1604651963-10067-1-git-send-email-alex.shi@linux.alibaba.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: x86: Set supported CPUIDs on default VM
Andrew Jones [Wed, 11 Nov 2020 12:26:35 +0000 (13:26 +0100)]
KVM: selftests: x86: Set supported CPUIDs on default VM

Almost all tests do this anyway and the ones that don't don't
appear to care. Only vmx_set_nested_state_test assumes that
a feature (VMX) is disabled until later setting the supported
CPUIDs. It's better to disable that explicitly anyway.

Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <20201111122636.73346-11-drjones@redhat.com>
[Restore CPUID_VMX, or vmx_set_nested_state breaks. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Make test skipping consistent
Andrew Jones [Wed, 11 Nov 2020 12:26:36 +0000 (13:26 +0100)]
KVM: selftests: Make test skipping consistent

Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <20201111122636.73346-12-drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Also build dirty_log_perf_test on AArch64
Andrew Jones [Wed, 11 Nov 2020 12:26:34 +0000 (13:26 +0100)]
KVM: selftests: Also build dirty_log_perf_test on AArch64

Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <20201111122636.73346-10-drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Introduce vm_create_[default_]_with_vcpus
Andrew Jones [Wed, 11 Nov 2020 12:26:30 +0000 (13:26 +0100)]
KVM: selftests: Introduce vm_create_[default_]_with_vcpus

Introduce new vm_create variants that also takes a number of vcpus,
an amount of per-vcpu pages, and optionally a list of vcpuids. These
variants will create default VMs with enough additional pages to
cover the vcpu stacks, per-vcpu pages, and pagetable pages for all.
The new 'default' variant uses VM_MODE_DEFAULT, whereas the other
new variant accepts the mode as a parameter.

Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Ben Gardon <bgardon@google.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <20201111122636.73346-6-drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Make vm_create_default common
Andrew Jones [Wed, 11 Nov 2020 12:26:29 +0000 (13:26 +0100)]
KVM: selftests: Make vm_create_default common

The code is almost 100% the same anyway. Just move it to common
and add a few arch-specific macros.

Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Ben Gardon <bgardon@google.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <20201111122636.73346-5-drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: always use manual clear in dirty_log_perf_test
Paolo Bonzini [Fri, 13 Nov 2020 16:36:49 +0000 (11:36 -0500)]
KVM: selftests: always use manual clear in dirty_log_perf_test

Nothing sets USE_CLEAR_DIRTY_LOG anymore, so anything it surrounds
is dead code.

However, it is the recommended way to use the dirty page bitmap
for new enough kernel, so use it whenever KVM has the
KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 capability.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokvm: x86: Sink cpuid update into vendor-specific set_cr4 functions
Jim Mattson [Thu, 29 Oct 2020 17:06:48 +0000 (10:06 -0700)]
kvm: x86: Sink cpuid update into vendor-specific set_cr4 functions

On emulated VM-entry and VM-exit, update the CPUID bits that reflect
CR4.OSXSAVE and CR4.PKE.

This fixes a bug where the CPUID bits could continue to reflect L2 CR4
values after emulated VM-exit to L1. It also fixes a related bug where
the CPUID bits could continue to reflect L1 CR4 values after emulated
VM-entry to L2. The latter bug is mainly relevant to SVM, wherein
CPUID is not a required intercept. However, it could also be relevant
to VMX, because the code to conditionally update these CPUID bits
assumes that the guest CPUID and the guest CR4 are always in sync.

Fixes: 8eb3f87d903168 ("KVM: nVMX: fix guest CR4 loading when emulating L2 to L1 exit")
Fixes: 2acf923e38fb6a ("KVM: VMX: Enable XSAVE/XRSTOR for guest")
Fixes: b9baba86148904 ("KVM, pkeys: expose CPUID/CR4 to guest")
Reported-by: Abhiroop Dabral <adabral@paloaltonetworks.com>
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
Cc: Haozhong Zhang <haozhong.zhang@intel.com>
Cc: Dexuan Cui <dexuan.cui@intel.com>
Cc: Huaitong Han <huaitong.han@intel.com>
Message-Id: <20201029170648.483210-1-jmattson@google.com>

3 years agoselftests: kvm: keep .gitignore add to date
Paolo Bonzini [Fri, 6 Nov 2020 12:39:26 +0000 (07:39 -0500)]
selftests: kvm: keep .gitignore add to date

Add tsc_msrs_test, remove clear_dirty_log_test and alphabetize
everything.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Add "-c" parameter to dirty log test
Peter Xu [Thu, 1 Oct 2020 01:22:41 +0000 (21:22 -0400)]
KVM: selftests: Add "-c" parameter to dirty log test

It's only used to override the existing dirty ring size/count.  If
with a bigger ring count, we test async of dirty ring.  If with a
smaller ring count, we test ring full code path.  Async is default.

It has no use for non-dirty-ring tests.

Reviewed-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012241.6208-1-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Run dirty ring test asynchronously
Peter Xu [Thu, 1 Oct 2020 01:22:39 +0000 (21:22 -0400)]
KVM: selftests: Run dirty ring test asynchronously

Previously the dirty ring test was working in synchronous way, because
only with a vmexit (with that it was the ring full event) we'll know
the hardware dirty bits will be flushed to the dirty ring.

With this patch we first introduce a vcpu kick mechanism using SIGUSR1,
which guarantees a vmexit and also therefore the flushing of hardware
dirty bits.  Once this is in place, we can keep the vcpu dirty work
asynchronous of the whole collection procedure now.  Still, we need
to be very careful that when reaching the ring buffer soft limit
(KVM_EXIT_DIRTY_RING_FULL) we must collect the dirty bits before
continuing the vcpu.

Further increase the dirty ring size to current maximum to make sure
we torture more on the no-ring-full case, which should be the major
scenario when the hypervisors like QEMU would like to use this feature.

Reviewed-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012239.6159-1-peterx@redhat.com>
[Use KVM_SET_SIGNAL_MASK+sigwait instead of a signal handler. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Add dirty ring buffer test
Peter Xu [Thu, 1 Oct 2020 01:22:37 +0000 (21:22 -0400)]
KVM: selftests: Add dirty ring buffer test

Add the initial dirty ring buffer test.

The current test implements the userspace dirty ring collection, by
only reaping the dirty ring when the ring is full.

So it's still running synchronously like this:

            vcpu                             main thread

  1. vcpu dirties pages
  2. vcpu gets dirty ring full
     (userspace exit)

                                       3. main thread waits until full
                                          (so hardware buffers flushed)
                                       4. main thread collects
                                       5. main thread continues vcpu

  6. vcpu continues, goes back to 1

We can't directly collects dirty bits during vcpu execution because
otherwise we can't guarantee the hardware dirty bits were flushed when
we collect and we're very strict on the dirty bits so otherwise we can
fail the future verify procedure.  A follow up patch will make this
test to support async just like the existing dirty log test, by adding
a vcpu kick mechanism.

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012237.6111-1-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Introduce after_vcpu_run hook for dirty log test
Peter Xu [Thu, 1 Oct 2020 01:22:35 +0000 (21:22 -0400)]
KVM: selftests: Introduce after_vcpu_run hook for dirty log test

Provide a hook for the checks after vcpu_run() completes.  Preparation
for the dirty ring test because we'll need to take care of another
exit reason.

Reviewed-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012235.6063-1-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: Don't allocate dirty bitmap if dirty ring is enabled
Peter Xu [Thu, 1 Oct 2020 01:22:26 +0000 (21:22 -0400)]
KVM: Don't allocate dirty bitmap if dirty ring is enabled

Because kvm dirty rings and kvm dirty log is used in an exclusive way,
Let's avoid creating the dirty_bitmap when kvm dirty ring is enabled.
At the meantime, since the dirty_bitmap will be conditionally created
now, we can't use it as a sign of "whether this memory slot enabled
dirty tracking".  Change users like that to check against the kvm
memory slot flags.

Note that there still can be chances where the kvm memory slot got its
dirty_bitmap allocated, _if_ the memory slots are created before
enabling of the dirty rings and at the same time with the dirty
tracking capability enabled, they'll still with the dirty_bitmap.
However it should not hurt much (e.g., the bitmaps will always be
freed if they are there), and the real users normally won't trigger
this because dirty bit tracking flag should in most cases only be
applied to kvm slots only before migration starts, that should be far
latter than kvm initializes (VM starts).

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012226.5868-1-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: Make dirty ring exclusive to dirty bitmap log
Peter Xu [Thu, 1 Oct 2020 01:22:24 +0000 (21:22 -0400)]
KVM: Make dirty ring exclusive to dirty bitmap log

There's no good reason to use both the dirty bitmap logging and the
new dirty ring buffer to track dirty bits.  We should be able to even
support both of them at the same time, but it could complicate things
which could actually help little.  Let's simply make it the rule
before we enable dirty ring on any arch, that we don't allow these two
interfaces to be used together.

The big world switch would be KVM_CAP_DIRTY_LOG_RING capability
enablement.  That's where we'll switch from the default dirty logging
way to the dirty ring way.  As long as kvm->dirty_ring_size is setup
correctly, we'll once and for all switch to the dirty ring buffer mode
for the current virtual machine.

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012224.5818-1-peterx@redhat.com>
[Change errno from EINVAL to ENXIO. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: X86: Implement ring-based dirty memory tracking
Peter Xu [Thu, 1 Oct 2020 01:22:22 +0000 (21:22 -0400)]
KVM: X86: Implement ring-based dirty memory tracking

This patch is heavily based on previous work from Lei Cao
<lei.cao@stratus.com> and Paolo Bonzini <pbonzini@redhat.com>. [1]

KVM currently uses large bitmaps to track dirty memory.  These bitmaps
are copied to userspace when userspace queries KVM for its dirty page
information.  The use of bitmaps is mostly sufficient for live
migration, as large parts of memory are be dirtied from one log-dirty
pass to another.  However, in a checkpointing system, the number of
dirty pages is small and in fact it is often bounded---the VM is
paused when it has dirtied a pre-defined number of pages. Traversing a
large, sparsely populated bitmap to find set bits is time-consuming,
as is copying the bitmap to user-space.

A similar issue will be there for live migration when the guest memory
is huge while the page dirty procedure is trivial.  In that case for
each dirty sync we need to pull the whole dirty bitmap to userspace
and analyse every bit even if it's mostly zeros.

The preferred data structure for above scenarios is a dense list of
guest frame numbers (GFN).  This patch series stores the dirty list in
kernel memory that can be memory mapped into userspace to allow speedy
harvesting.

This patch enables dirty ring for X86 only.  However it should be
easily extended to other archs as well.

[1] https://patchwork.kernel.org/patch/10471409/

Signed-off-by: Lei Cao <lei.cao@stratus.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012222.5767-1-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: Pass in kvm pointer into mark_page_dirty_in_slot()
Peter Xu [Thu, 1 Oct 2020 01:20:34 +0000 (21:20 -0400)]
KVM: Pass in kvm pointer into mark_page_dirty_in_slot()

The context will be needed to implement the kvm dirty ring.

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012044.5151-5-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: remove kvm_clear_guest_page
Paolo Bonzini [Fri, 6 Nov 2020 10:25:09 +0000 (05:25 -0500)]
KVM: remove kvm_clear_guest_page

kvm_clear_guest_page is not used anymore after "KVM: X86: Don't track dirty
for KVM_SET_[TSS_ADDR|IDENTITY_MAP_ADDR]", except from kvm_clear_guest.
We can just inline it in its sole user.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: X86: Don't track dirty for KVM_SET_[TSS_ADDR|IDENTITY_MAP_ADDR]
Peter Xu [Thu, 1 Oct 2020 01:20:33 +0000 (21:20 -0400)]
KVM: X86: Don't track dirty for KVM_SET_[TSS_ADDR|IDENTITY_MAP_ADDR]

Originally, we have three code paths that can dirty a page without
vcpu context for X86:

  - init_rmode_identity_map
  - init_rmode_tss
  - kvmgt_rw_gpa

init_rmode_identity_map and init_rmode_tss will be setup on
destination VM no matter what (and the guest cannot even see them), so
it does not make sense to track them at all.

To do this, allow __x86_set_memory_region() to return the userspace
address that just allocated to the caller.  Then in both of the
functions we directly write to the userspace address instead of
calling kvm_write_*() APIs.

Another trivial change is that we don't need to explicitly clear the
identity page table root in init_rmode_identity_map() because no
matter what we'll write to the whole page with 4M huge page entries.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20201001012044.5151-4-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: test KVM_GET_SUPPORTED_HV_CPUID as a system ioctl
Vitaly Kuznetsov [Tue, 29 Sep 2020 15:09:44 +0000 (17:09 +0200)]
KVM: selftests: test KVM_GET_SUPPORTED_HV_CPUID as a system ioctl

KVM_GET_SUPPORTED_HV_CPUID is now supported as both vCPU and VM ioctl,
test that.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20200929150944.1235688-3-vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: hyper-v: allow KVM_GET_SUPPORTED_HV_CPUID as a system ioctl
Vitaly Kuznetsov [Tue, 29 Sep 2020 15:09:43 +0000 (17:09 +0200)]
KVM: x86: hyper-v: allow KVM_GET_SUPPORTED_HV_CPUID as a system ioctl

KVM_GET_SUPPORTED_HV_CPUID is a vCPU ioctl but its output is now
independent from vCPU and in some cases VMMs may want to use it as a system
ioctl instead. In particular, QEMU doesn CPU feature expansion before any
vCPU gets created so KVM_GET_SUPPORTED_HV_CPUID can't be used.

Convert KVM_GET_SUPPORTED_HV_CPUID to 'dual' system/vCPU ioctl with the
same meaning.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20200929150944.1235688-2-vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokvm/eventfd: Drain events from eventfd in irqfd_wakeup()
David Woodhouse [Tue, 27 Oct 2020 13:55:23 +0000 (13:55 +0000)]
kvm/eventfd: Drain events from eventfd in irqfd_wakeup()

Don't allow the events to accumulate in the eventfd counter, drain them
as they are handled.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Message-Id: <20201027135523.646811-4-dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agovfio/virqfd: Drain events from eventfd in virqfd_wakeup()
David Woodhouse [Tue, 27 Oct 2020 13:55:22 +0000 (13:55 +0000)]
vfio/virqfd: Drain events from eventfd in virqfd_wakeup()

Don't allow the events to accumulate in the eventfd counter, drain them
as they are handled.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Message-Id: <20201027135523.646811-3-dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
3 years agoeventfd: Export eventfd_ctx_do_read()
David Woodhouse [Tue, 27 Oct 2020 13:55:21 +0000 (13:55 +0000)]
eventfd: Export eventfd_ctx_do_read()

Where events are consumed in the kernel, for example by KVM's
irqfd_wakeup() and VFIO's virqfd_wakeup(), they currently lack a
mechanism to drain the eventfd's counter.

Since the wait queue is already locked while the wakeup functions are
invoked, all they really need to do is call eventfd_ctx_do_read().

Add a check for the lock, and export it for them.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Message-Id: <20201027135523.646811-2-dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokvm/eventfd: Use priority waitqueue to catch events before userspace
David Woodhouse [Mon, 26 Oct 2020 17:53:25 +0000 (17:53 +0000)]
kvm/eventfd: Use priority waitqueue to catch events before userspace

As far as I can tell, when we use posted interrupts we silently cut off
the events from userspace, if it's listening on the same eventfd that
feeds the irqfd.

I like that behaviour. Let's do it all the time, even without posted
interrupts. It makes it much easier to handle IRQ remapping invalidation
without having to constantly add/remove the fd from the userspace poll
set. We can just leave userspace polling on it, and the bypass will...
well... bypass it.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Message-Id: <20201026175325.585623-2-dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agosched/wait: Add add_wait_queue_priority()
David Woodhouse [Tue, 27 Oct 2020 14:39:43 +0000 (14:39 +0000)]
sched/wait: Add add_wait_queue_priority()

This allows an exclusive wait_queue_entry to be added at the head of the
queue, instead of the tail as normal. Thus, it gets to consume events
first without allowing non-exclusive waiters to be woken at all.

The (first) intended use is for KVM IRQFD, which currently has
inconsistent behaviour depending on whether posted interrupts are
available or not. If they are, KVM will bypass the eventfd completely
and deliver interrupts directly to the appropriate vCPU. If not, events
are delivered through the eventfd and userspace will receive them when
polling on the eventfd.

By using add_wait_queue_priority(), KVM will be able to consistently
consume events within the kernel without accidentally exposing them
to userspace when they're supposed to be bypassed. This, in turn, means
that userspace doesn't have to jump through hoops to avoid listening
on the erroneously noisy eventfd and injecting duplicate interrupts.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Message-Id: <20201027143944.648769-2-dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: emulate wait-for-SIPI and SIPI-VMExit
Yadong Qi [Fri, 6 Nov 2020 06:51:22 +0000 (14:51 +0800)]
KVM: x86: emulate wait-for-SIPI and SIPI-VMExit

Background: We have a lightweight HV, it needs INIT-VMExit and
SIPI-VMExit to wake-up APs for guests since it do not monitor
the Local APIC. But currently virtual wait-for-SIPI(WFS) state
is not supported in nVMX, so when running on top of KVM, the L1
HV cannot receive the INIT-VMExit and SIPI-VMExit which cause
the L2 guest cannot wake up the APs.

According to Intel SDM Chapter 25.2 Other Causes of VM Exits,
SIPIs cause VM exits when a logical processor is in
wait-for-SIPI state.

In this patch:
    1. introduce SIPI exit reason,
    2. introduce wait-for-SIPI state for nVMX,
    3. advertise wait-for-SIPI support to guest.

When L1 hypervisor is not monitoring Local APIC, L0 need to emulate
INIT-VMExit and SIPI-VMExit to L1 to emulate INIT-SIPI-SIPI for
L2. L2 LAPIC write would be traped by L0 Hypervisor(KVM), L0 should
emulate the INIT/SIPI vmexit to L1 hypervisor to set proper state
for L2's vcpu state.

Handle procdure:
Source vCPU:
    L2 write LAPIC.ICR(INIT).
    L0 trap LAPIC.ICR write(INIT): inject a latched INIT event to target
       vCPU.
Target vCPU:
    L0 emulate an INIT VMExit to L1 if is guest mode.
    L1 set guest VMCS, guest_activity_state=WAIT_SIPI, vmresume.
    L0 set vcpu.mp_state to INIT_RECEIVED if (vmcs12.guest_activity_state
       == WAIT_SIPI).

Source vCPU:
    L2 write LAPIC.ICR(SIPI).
    L0 trap LAPIC.ICR write(INIT): inject a latched SIPI event to traget
       vCPU.
Target vCPU:
    L0 emulate an SIPI VMExit to L1 if (vcpu.mp_state == INIT_RECEIVED).
    L1 set CS:IP, guest_activity_state=ACTIVE, vmresume.
    L0 resume to L2.
    L2 start-up.

Signed-off-by: Yadong Qi <yadong.qi@intel.com>
Message-Id: <20200922052343.84388-1-yadong.qi@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201106065122.403183-1-yadong.qi@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: fix apic_accept_events vs check_nested_events
Paolo Bonzini [Thu, 5 Nov 2020 16:20:49 +0000 (11:20 -0500)]
KVM: x86: fix apic_accept_events vs check_nested_events

vmx_apic_init_signal_blocked is buggy in that it returns true
even in VMX non-root mode.  In non-root mode, however, INITs
are not latched, they just cause a vmexit.  Previously,
KVM was waiting for them to be processed when kvm_apic_accept_events
and in the meanwhile it ate the SIPIs that the processor received.

However, in order to implement the wait-for-SIPI activity state,
KVM will have to process KVM_APIC_SIPI in vmx_check_nested_events,
and it will not be possible anymore to disregard SIPIs in non-root
mode as the code is currently doing.

By calling kvm_x86_ops.nested_ops->check_events, we can force a vmexit
(with the side-effect of latching INITs) before incorrectly injecting
an INIT or SIPI in a guest, and therefore vmx_apic_init_signal_blocked
can do the right thing.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: selftests: Verify supported CR4 bits can be set before KVM_SET_CPUID2
Sean Christopherson [Wed, 7 Oct 2020 01:44:17 +0000 (18:44 -0700)]
KVM: selftests: Verify supported CR4 bits can be set before KVM_SET_CPUID2

Extend the KVM_SET_SREGS test to verify that all supported CR4 bits, as
enumerated by KVM, can be set before KVM_SET_CPUID2, i.e. without first
defining the vCPU model.  KVM is supposed to skip guest CPUID checks
when host userspace is stuffing guest state.

Check the inverse as well, i.e. that KVM rejects KVM_SET_REGS if CR4
has one or more unsupported bits set.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20201007014417.29276-7-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: Return bool instead of int for CR4 and SREGS validity checks
Sean Christopherson [Wed, 7 Oct 2020 01:44:16 +0000 (18:44 -0700)]
KVM: x86: Return bool instead of int for CR4 and SREGS validity checks

Rework the common CR4 and SREGS checks to return a bool instead of an
int, i.e. true/false instead of 0/-EINVAL, and add "is" to the name to
clarify the polarity of the return value (which is effectively inverted
by this change).

No functional changed intended.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20201007014417.29276-6-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: Move vendor CR4 validity check to dedicated kvm_x86_ops hook
Sean Christopherson [Wed, 7 Oct 2020 01:44:15 +0000 (18:44 -0700)]
KVM: x86: Move vendor CR4 validity check to dedicated kvm_x86_ops hook

Split out VMX's checks on CR4.VMXE to a dedicated hook, .is_valid_cr4(),
and invoke the new hook from kvm_valid_cr4().  This fixes an issue where
KVM_SET_SREGS would return success while failing to actually set CR4.

Fixing the issue by explicitly checking kvm_x86_ops.set_cr4()'s return
in __set_sregs() is not a viable option as KVM has already stuffed a
variety of vCPU state.

Note, kvm_valid_cr4() and is_valid_cr4() have different return types and
inverted semantics.  This will be remedied in a future patch.

Fixes: 5e1746d6205d ("KVM: nVMX: Allow setting the VMXE bit in CR4")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20201007014417.29276-5-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: SVM: Drop VMXE check from svm_set_cr4()
Sean Christopherson [Wed, 7 Oct 2020 01:44:14 +0000 (18:44 -0700)]
KVM: SVM: Drop VMXE check from svm_set_cr4()

Drop svm_set_cr4()'s explicit check CR4.VMXE now that common x86 handles
the check by incorporating VMXE into the CR4 reserved bits, via
kvm_cpu_caps.  SVM obviously does not set X86_FEATURE_VMX.

No functional change intended.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20201007014417.29276-4-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: VMX: Drop explicit 'nested' check from vmx_set_cr4()
Sean Christopherson [Wed, 7 Oct 2020 01:44:13 +0000 (18:44 -0700)]
KVM: VMX: Drop explicit 'nested' check from vmx_set_cr4()

Drop vmx_set_cr4()'s explicit check on the 'nested' module param now
that common x86 handles the check by incorporating VMXE into the CR4
reserved bits, via kvm_cpu_caps.  X86_FEATURE_VMX is set in kvm_cpu_caps
(by vmx_set_cpu_caps()), if and only if 'nested' is true.

No functional change intended.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20201007014417.29276-3-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: VMX: Drop guest CPUID check for VMXE in vmx_set_cr4()
Sean Christopherson [Wed, 7 Oct 2020 01:44:12 +0000 (18:44 -0700)]
KVM: VMX: Drop guest CPUID check for VMXE in vmx_set_cr4()

Drop vmx_set_cr4()'s somewhat hidden guest_cpuid_has() check on VMXE now
that common x86 handles the check by incorporating VMXE into the CR4
reserved bits, i.e. in cr4_guest_rsvd_bits.  This fixes a bug where KVM
incorrectly rejects KVM_SET_SREGS with CR4.VMXE=1 if it's executed
before KVM_SET_CPUID{,2}.

Fixes: 5e1746d6205d ("KVM: nVMX: Allow setting the VMXE bit in CR4")
Reported-by: Stas Sergeev <stsp@users.sourceforge.net>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20201007014417.29276-2-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokvm: mmu: fix is_tdp_mmu_check when the TDP MMU is not in use
Paolo Bonzini [Sun, 15 Nov 2020 13:55:43 +0000 (08:55 -0500)]
kvm: mmu: fix is_tdp_mmu_check when the TDP MMU is not in use

In some cases where shadow paging is in use, the root page will
be either mmu->pae_root or vcpu->arch.mmu->lm_root.  Then it will
not have an associated struct kvm_mmu_page, because it is allocated
with alloc_page instead of kvm_mmu_alloc_page.

Just return false quickly from is_tdp_mmu_root if the TDP MMU is
not in use, which also includes the case where shadow paging is
enabled.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: SVM: Update cr3_lm_rsvd_bits for AMD SEV guests
Babu Moger [Thu, 12 Nov 2020 22:18:03 +0000 (16:18 -0600)]
KVM: SVM: Update cr3_lm_rsvd_bits for AMD SEV guests

For AMD SEV guests, update the cr3_lm_rsvd_bits to mask
the memory encryption bit in reserved bits.

Signed-off-by: Babu Moger <babu.moger@amd.com>
Message-Id: <160521948301.32054.5783800787423231162.stgit@bmoger-ubuntu>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: Introduce cr3_lm_rsvd_bits in kvm_vcpu_arch
Babu Moger [Thu, 12 Nov 2020 22:17:56 +0000 (16:17 -0600)]
KVM: x86: Introduce cr3_lm_rsvd_bits in kvm_vcpu_arch

SEV guests fail to boot on a system that supports the PCID feature.

While emulating the RSM instruction, KVM reads the guest CR3
and calls kvm_set_cr3(). If the vCPU is in the long mode,
kvm_set_cr3() does a sanity check for the CR3 value. In this case,
it validates whether the value has any reserved bits set. The
reserved bit range is 63:cpuid_maxphysaddr(). When AMD memory
encryption is enabled, the memory encryption bit is set in the CR3
value. The memory encryption bit may fall within the KVM reserved
bit range, causing the KVM emulation failure.

Introduce a new field cr3_lm_rsvd_bits in kvm_vcpu_arch which will
cache the reserved bits in the CR3 value. This will be initialized
to rsvd_bits(cpuid_maxphyaddr(vcpu), 63).

If the architecture has any special bits(like AMD SEV encryption bit)
that needs to be masked from the reserved bits, should be cleared
in vendor specific kvm_x86_ops.vcpu_after_set_cpuid handler.

Fixes: a780a3ea628268b2 ("KVM: X86: Fix reserved bits check for MOV to CR3")
Signed-off-by: Babu Moger <babu.moger@amd.com>
Message-Id: <160521947657.32054.3264016688005356563.stgit@bmoger-ubuntu>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoKVM: x86: clflushopt should be treated as a no-op by emulation
David Edmondson [Tue, 3 Nov 2020 12:04:00 +0000 (12:04 +0000)]
KVM: x86: clflushopt should be treated as a no-op by emulation

The instruction emulator ignores clflush instructions, yet fails to
support clflushopt. Treat both similarly.

Fixes: 13e457e0eebf ("KVM: x86: Emulator does not decode clflush well")
Signed-off-by: David Edmondson <david.edmondson@oracle.com>
Message-Id: <20201103120400.240882-1-david.edmondson@oracle.com>
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoMerge tag 'kvmarm-fixes-5.10-3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Paolo Bonzini [Fri, 13 Nov 2020 11:28:23 +0000 (06:28 -0500)]
Merge tag 'kvmarm-fixes-5.10-3' of git://git./linux/kernel/git/kvmarm/kvmarm into HEAD

KVM/arm64 fixes for v5.10, take #3

- Allow userspace to downgrade ID_AA64PFR0_EL1.CSV2
- Inject UNDEF on SCXTNUM_ELx access

3 years agoMerge tag 'fscrypt-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/fscrypt
Linus Torvalds [Fri, 13 Nov 2020 00:39:58 +0000 (16:39 -0800)]
Merge tag 'fscrypt-for-linus' of git://git./fs/fscrypt/fscrypt

Pull fscrypt fix from Eric Biggers:
 "Fix a regression where new files weren't using inline encryption when
  they should be"

* tag 'fscrypt-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/fscrypt:
  fscrypt: fix inline encryption not used on new files

3 years agoMerge tag 'gfs2-v5.10-rc3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 13 Nov 2020 00:37:14 +0000 (16:37 -0800)]
Merge tag 'gfs2-v5.10-rc3-fixes' of git://git./linux/kernel/git/gfs2/linux-gfs2

Pull gfs2 fixes from Andreas Gruenbacher:
 "Fix jdata data corruption and glock reference leak"

* tag 'gfs2-v5.10-rc3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
  gfs2: Fix case in which ail writes are done to jdata holes
  Revert "gfs2: Ignore journal log writes for jdata holes"
  gfs2: fix possible reference leak in gfs2_check_blk_type

3 years agoMerge tag 'net-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 12 Nov 2020 22:02:04 +0000 (14:02 -0800)]
Merge tag 'net-5.10-rc4' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Current release - regressions:

   - arm64: dts: fsl-ls1028a-kontron-sl28: specify in-band mode for
     ENETC

  Current release - bugs in new features:

   - mptcp: provide rmem[0] limit offset to fix oops

  Previous release - regressions:

   - IPv6: Set SIT tunnel hard_header_len to zero to fix path MTU
     calculations

   - lan743x: correctly handle chips with internal PHY

   - bpf: Don't rely on GCC __attribute__((optimize)) to disable GCSE

   - mlx5e: Fix VXLAN port table synchronization after function reload

  Previous release - always broken:

   - bpf: Zero-fill re-used per-cpu map element

   - fix out-of-order UDP packets when forwarding with UDP GSO fraglists
     turned on:
       - fix UDP header access on Fast/frag0 UDP GRO
       - fix IP header access and skb lookup on Fast/frag0 UDP GRO

   - ethtool: netlink: add missing netdev_features_change() call

   - net: Update window_clamp if SOCK_RCVBUF is set

   - igc: Fix returning wrong statistics

   - ch_ktls: fix multiple leaks and corner cases in Chelsio TLS offload

   - tunnels: Fix off-by-one in lower MTU bounds for ICMP/ICMPv6 replies

   - r8169: disable hw csum for short packets on all chip versions

   - vrf: Fix fast path output packet handling with async Netfilter
     rules"

* tag 'net-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (65 commits)
  lan743x: fix use of uninitialized variable
  net: udp: fix IP header access and skb lookup on Fast/frag0 UDP GRO
  net: udp: fix UDP header access on Fast/frag0 UDP GRO
  devlink: Avoid overwriting port attributes of registered port
  vrf: Fix fast path output packet handling with async Netfilter rules
  cosa: Add missing kfree in error path of cosa_write
  net: switch to the kernel.org patchwork instance
  ch_ktls: stop the txq if reaches threshold
  ch_ktls: tcb update fails sometimes
  ch_ktls/cxgb4: handle partial tag alone SKBs
  ch_ktls: don't free skb before sending FIN
  ch_ktls: packet handling prior to start marker
  ch_ktls: Correction in middle record handling
  ch_ktls: missing handling of header alone
  ch_ktls: Correction in trimmed_len calculation
  cxgb4/ch_ktls: creating skbs causes panic
  ch_ktls: Update cheksum information
  ch_ktls: Correction in finding correct length
  cxgb4/ch_ktls: decrypted bit is not enough
  net/x25: Fix null-ptr-deref in x25_connect
  ...

3 years agoMerge tag 'nfs-for-5.10-2' of git://git.linux-nfs.org/projects/anna/linux-nfs
Linus Torvalds [Thu, 12 Nov 2020 21:49:12 +0000 (13:49 -0800)]
Merge tag 'nfs-for-5.10-2' of git://git.linux-nfs.org/projects/anna/linux-nfs

Pull NFS client bugfixes from Anna Schumaker:
 "Stable fixes:
  - Fix failure to unregister shrinker

  Other fixes:
  - Fix unnecessary locking to clear up some contention
  - Fix listxattr receive buffer size
  - Fix default mount options for nfsroot"

* tag 'nfs-for-5.10-2' of git://git.linux-nfs.org/projects/anna/linux-nfs:
  NFS: Remove unnecessary inode lock in nfs_fsync_dir()
  NFS: Remove unnecessary inode locking in nfs_llseek_dir()
  NFS: Fix listxattr receive buffer size
  NFSv4.2: fix failure to unregister shrinker
  nfsroot: Default mount option should ask for built-in NFS version

3 years agoKVM: arm64: Handle SCXTNUM_ELx traps
Marc Zyngier [Tue, 10 Nov 2020 14:13:08 +0000 (14:13 +0000)]
KVM: arm64: Handle SCXTNUM_ELx traps

As the kernel never sets HCR_EL2.EnSCXT, accesses to SCXTNUM_ELx
will trap to EL2. Let's handle that as gracefully as possible
by injecting an UNDEF exception into the guest. This is consistent
with the guest's view of ID_AA64PFR0_EL1.CSV2 being at most 1.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20201110141308.451654-4-maz@kernel.org
3 years agoKVM: arm64: Unify trap handlers injecting an UNDEF
Marc Zyngier [Tue, 10 Nov 2020 14:13:07 +0000 (14:13 +0000)]
KVM: arm64: Unify trap handlers injecting an UNDEF

A large number of system register trap handlers only inject an
UNDEF exeption, and yet each class of sysreg seems to provide its
own, identical function.

Let's unify them all, saving us introducing yet another one later.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20201110141308.451654-3-maz@kernel.org
3 years agoKVM: arm64: Allow setting of ID_AA64PFR0_EL1.CSV2 from userspace
Marc Zyngier [Tue, 10 Nov 2020 14:13:06 +0000 (14:13 +0000)]
KVM: arm64: Allow setting of ID_AA64PFR0_EL1.CSV2 from userspace

We now expose ID_AA64PFR0_EL1.CSV2=1 to guests running on hosts
that are immune to Spectre-v2, but that don't have this field set,
most likely because they predate the specification.

However, this prevents the migration of guests that have started on
a host the doesn't fake this CSV2 setting to one that does, as KVM
rejects the write to ID_AA64PFR0_EL2 on the grounds that it isn't
what is already there.

In order to fix this, allow userspace to set this field as long as
this doesn't result in a promising more than what is already there
(setting CSV2 to 0 is acceptable, but setting it to 1 when it is
already set to 0 isn't).

Fixes: e1026237f9067 ("KVM: arm64: Set CSV2 for guests on hardware unaffected by Spectre-v2")
Reported-by: Peng Liang <liangpeng10@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20201110141308.451654-2-maz@kernel.org
3 years agoMerge tag 'v5.10-rc1' into kvmarm-master/next
Marc Zyngier [Thu, 12 Nov 2020 21:20:43 +0000 (21:20 +0000)]
Merge tag 'v5.10-rc1' into kvmarm-master/next

Linux 5.10-rc1

Signed-off-by: Marc Zyngier <maz@kernel.org>
3 years agoMerge tag 'acpi-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Thu, 12 Nov 2020 19:06:53 +0000 (11:06 -0800)]
Merge tag 'acpi-5.10-rc4' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fixes from Rafael Wysocki:
 "These are mostly docmentation fixes and janitorial changes plus some
  new device IDs and a new quirk.

  Specifics:

   - Fix documentation regarding GPIO properties (Andy Shevchenko)

   - Fix spelling mistakes in ACPI documentation (Flavio Suligoi)

   - Fix white space inconsistencies in ACPI code (Maximilian Luz)

   - Fix string formatting in the ACPI Generic Event Device (GED) driver
     (Nick Desaulniers)

   - Add Intel Alder Lake device IDs to the ACPI drivers used by the
     Dynamic Platform and Thermal Framework (Srinivas Pandruvada)

   - Add lid-related DMI quirk for Medion Akoya E2228T to the ACPI
     button driver (Hans de Goede)"

* tag 'acpi-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI: DPTF: Support Alder Lake
  Documentation: ACPI: fix spelling mistakes
  ACPI: button: Add DMI quirk for Medion Akoya E2228T
  ACPI: GED: fix -Wformat
  ACPI: Fix whitespace inconsistencies
  ACPI: scan: Fix acpi_dma_configure_id() kerneldoc name
  Documentation: firmware-guide: gpio-properties: Clarify initial output state
  Documentation: firmware-guide: gpio-properties: active_low only for GpioIo()
  Documentation: firmware-guide: gpio-properties: Fix factual mistakes

3 years agoMerge tag 'pm-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Thu, 12 Nov 2020 19:03:38 +0000 (11:03 -0800)]
Merge tag 'pm-5.10-rc4' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "Make the intel_pstate driver behave as expected when it operates in
  the passive mode with HWP enabled and the 'powersave' governor on top
  of it"

* tag 'pm-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq: intel_pstate: Take CPUFREQ_GOV_STRICT_TARGET into account
  cpufreq: Add strict_target to struct cpufreq_policy
  cpufreq: Introduce CPUFREQ_GOV_STRICT_TARGET
  cpufreq: Introduce governor flags

3 years agolan743x: fix use of uninitialized variable
Sven Van Asbroeck [Thu, 12 Nov 2020 15:25:13 +0000 (10:25 -0500)]
lan743x: fix use of uninitialized variable

When no devicetree is present, the driver will use an
uninitialized variable.

Fix by initializing this variable.

Fixes: 902a66e08cea ("lan743x: correctly handle chips with internal PHY")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
Link: https://lore.kernel.org/r/20201112152513.1941-1-TheSven73@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge branch 'net-udp-fix-fast-frag0-udp-gro'
Jakub Kicinski [Thu, 12 Nov 2020 17:55:58 +0000 (09:55 -0800)]
Merge branch 'net-udp-fix-fast-frag0-udp-gro'

Alexander Lobakin says:

====================
net: udp: fix Fast/frag0 UDP GRO

While testing UDP GSO fraglists forwarding through driver that uses
Fast GRO (via napi_gro_frags()), I was observing lots of out-of-order
iperf packets:

[ ID] Interval           Transfer     Bitrate         Jitter
[SUM]  0.0-40.0 sec  12106 datagrams received out-of-order

Simple switch to napi_gro_receive() or any other method without frag0
shortcut completely resolved them.

I've found two incorrect header accesses in GRO receive callback(s):
 - udp_hdr() (instead of udp_gro_udphdr()) that always points to junk
   in "fast" mode and could probably do this in "regular".
   This was the actual bug that caused all out-of-order delivers;
 - udp{4,6}_lib_lookup_skb() -> ip{,v6}_hdr() (instead of
   skb_gro_network_header()) that potentionally might return odd
   pointers in both modes.

Each patch addresses one of these two issues.

This doesn't cover a support for nested tunnels as it's out of the
subject and requires more invasive changes. It will be handled
separately in net-next series.

Credits:
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Willem de Bruijn <willemb@google.com>
Since v4 [0]:
 - split the fix into two logical ones (Willem);
 - replace ternaries with plain ifs to beautify the code (Jakub);
 - drop p->data part to reintroduce it later in abovementioned set.

Since v3 [1]:
 - restore the original {,__}udp{4,6}_lib_lookup_skb() and use
   private versions of them inside GRO code (Willem).

Since v2 [2]:
 - dropped redundant check introduced in v2 as it's performed right
   before (thanks to Eric);
 - udp_hdr() switched to data + off for skbs from list (also Eric);
 - fixed possible malfunction of {,__}udp{4,6}_lib_lookup_skb() with
   Fast/frag0 due to ip{,v6}_hdr() usage (Willem).

Since v1 [3]:
 - added a NULL pointer check for "uh" as suggested by Willem.

[0] https://lore.kernel.org/netdev/Ha2hou5eJPcblo4abjAqxZRzIl1RaLs2Hy0oOAgFs@cp4-web-036.plabs.ch
[1] https://lore.kernel.org/netdev/MgZce9htmEtCtHg7pmWxXXfdhmQ6AHrnltXC41zOoo@cp7-web-042.plabs.ch
[2] https://lore.kernel.org/netdev/0eaG8xtbtKY1dEKCTKUBubGiC9QawGgB3tVZtNqVdY@cp4-web-030.plabs.ch
[3] https://lore.kernel.org/netdev/YazU6GEzBdpyZMDMwJirxDX7B4sualpDG68ADZYvJI@cp4-web-034.plabs.ch
====================

Link: https://lore.kernel.org/r/hjGOh0iCOYyo1FPiZh6TMXcx3YCgNs1T1eGKLrDz8@cp4-web-037.plabs.ch
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: udp: fix IP header access and skb lookup on Fast/frag0 UDP GRO
Alexander Lobakin [Wed, 11 Nov 2020 20:45:38 +0000 (20:45 +0000)]
net: udp: fix IP header access and skb lookup on Fast/frag0 UDP GRO

udp{4,6}_lib_lookup_skb() use ip{,v6}_hdr() to get IP header of the
packet. While it's probably OK for non-frag0 paths, this helpers
will also point to junk on Fast/frag0 GRO when all headers are
located in frags. As a result, sk/skb lookup may fail or give wrong
results. To support both GRO modes, skb_gro_network_header() might
be used. To not modify original functions, add private versions of
udp{4,6}_lib_lookup_skb() only to perform correct sk lookups on GRO.

Present since the introduction of "application-level" UDP GRO
in 4.7-rc1.

Misc: replace totally unneeded ternaries with plain ifs.

Fixes: a6024562ffd7 ("udp: Add GRO functions to UDP socket")
Suggested-by: Willem de Bruijn <willemb@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: udp: fix UDP header access on Fast/frag0 UDP GRO
Alexander Lobakin [Wed, 11 Nov 2020 20:45:25 +0000 (20:45 +0000)]
net: udp: fix UDP header access on Fast/frag0 UDP GRO

UDP GRO uses udp_hdr(skb) in its .gro_receive() callback. While it's
probably OK for non-frag0 paths (when all headers or even the entire
frame are already in skb head), this inline points to junk when
using Fast GRO (napi_gro_frags() or napi_gro_receive() with only
Ethernet header in skb head and all the rest in the frags) and breaks
GRO packet compilation and the packet flow itself.
To support both modes, skb_gro_header_fast() + skb_gro_header_slow()
are typically used. UDP even has an inline helper that makes use of
them, udp_gro_udphdr(). Use that instead of troublemaking udp_hdr()
to get rid of the out-of-order delivers.

Present since the introduction of plain UDP GRO in 5.0-rc1.

Fixes: e20cf8d3f1f7 ("udp: implement GRO for plain UDP sockets.")
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agogfs2: Fix case in which ail writes are done to jdata holes
Bob Peterson [Thu, 12 Nov 2020 16:02:48 +0000 (10:02 -0600)]
gfs2: Fix case in which ail writes are done to jdata holes

Patch b2a846dbef4e ("gfs2: Ignore journal log writes for jdata holes")
tried (unsuccessfully) to fix a case in which writes were done to jdata
blocks, the blocks are sent to the ail list, then a punch_hole or truncate
operation caused the blocks to be freed. In other words, the ail items
are for jdata holes. Before b2a846dbef4e, the jdata hole caused function
gfs2_block_map to return -EIO, which was eventually interpreted as an
IO error to the journal, and then withdraw.

This patch changes function gfs2_get_block_noalloc, which is only used
for jdata writes, so it returns -ENODATA rather than -EIO, and when
-ENODATA is returned to gfs2_ail1_start_one, the error is ignored.
We can safely ignore it because gfs2_ail1_start_one is only called
when the jdata pages have already been written and truncated, so the
ail1 content no longer applies.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
3 years agoRevert "gfs2: Ignore journal log writes for jdata holes"
Bob Peterson [Wed, 11 Nov 2020 17:09:55 +0000 (11:09 -0600)]
Revert "gfs2: Ignore journal log writes for jdata holes"

This reverts commit b2a846dbef4ef54ef032f0f5ee188c609a0278a7.

That commit changed the behavior of function gfs2_block_map to return
-ENODATA in cases where a hole (IOMAP_HOLE) is encountered and create is
false.  While that fixed the intended problem for jdata, it also broke
other callers of gfs2_block_map such as some jdata block reads.  Before
the patch, an encountered hole would be skipped and the buffer seen as
unmapped by the caller.  The patch changed the behavior to return
-ENODATA, which is interpreted as an error by the caller.

The -ENODATA return code should be restricted to the specific case where
jdata holes are encountered during ail1 writes.  That will be done in a
later patch.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
3 years agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue
Jakub Kicinski [Thu, 12 Nov 2020 16:47:22 +0000 (08:47 -0800)]
Merge branch '40GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2020-11-10

This series contains updates to i40e and igc drivers and the MAINTAINERS
file.

Slawomir fixes updating VF MAC addresses to fix various issues related
to reporting and setting of these addresses for i40e.

Dan Carpenter fixes a possible used before being initialized issue for
i40e.

Vinicius fixes reporting of netdev stats for igc.

Tony updates repositories for Intel Ethernet Drivers.

* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  MAINTAINERS: Update repositories for Intel Ethernet Drivers
  igc: Fix returning wrong statistics
  i40e, xsk: uninitialized variable in i40e_clean_rx_irq_zc()
  i40e: Fix MAC address setting for a VF via Host/VM
====================

Link: https://lore.kernel.org/r/20201111001955.533210-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agodevlink: Avoid overwriting port attributes of registered port
Parav Pandit [Wed, 11 Nov 2020 03:47:44 +0000 (05:47 +0200)]
devlink: Avoid overwriting port attributes of registered port

Cited commit in fixes tag overwrites the port attributes for the
registered port.

Avoid such error by checking registered flag before setting attributes.

Fixes: 71ad8d55f8e5 ("devlink: Replace devlink_port_attrs_set parameters with a struct")
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20201111034744.35554-1-parav@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agovrf: Fix fast path output packet handling with async Netfilter rules
Martin Willi [Fri, 6 Nov 2020 07:30:30 +0000 (08:30 +0100)]
vrf: Fix fast path output packet handling with async Netfilter rules

VRF devices use an optimized direct path on output if a default qdisc
is involved, calling Netfilter hooks directly. This path, however, does
not consider Netfilter rules completing asynchronously, such as with
NFQUEUE. The Netfilter okfn() is called for asynchronously accepted
packets, but the VRF never passes that packet down the stack to send
it out over the slave device. Using the slower redirect path for this
seems not feasible, as we do not know beforehand if a Netfilter hook
has asynchronously completing rules.

Fix the use of asynchronously completing Netfilter rules in OUTPUT and
POSTROUTING by using a special completion function that additionally
calls dst_output() to pass the packet down the stack. Also, slightly
adjust the use of nf_reset_ct() so that is called in the asynchronous
case, too.

Fixes: dcdd43c41e60 ("net: vrf: performance improvements for IPv4")
Fixes: a9ec54d1b0cd ("net: vrf: performance improvements for IPv6")
Signed-off-by: Martin Willi <martin@strongswan.org>
Link: https://lore.kernel.org/r/20201106073030.3974927-1-martin@strongswan.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoNFS: Remove unnecessary inode lock in nfs_fsync_dir()
Trond Myklebust [Fri, 30 Oct 2020 21:57:30 +0000 (17:57 -0400)]
NFS: Remove unnecessary inode lock in nfs_fsync_dir()

nfs_inc_stats() is already thread-safe, and there are no other reasons
to hold the inode lock here.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
3 years agoNFS: Remove unnecessary inode locking in nfs_llseek_dir()
Trond Myklebust [Fri, 30 Oct 2020 21:57:29 +0000 (17:57 -0400)]
NFS: Remove unnecessary inode locking in nfs_llseek_dir()

Remove the contentious inode lock, and instead provide thread safety
using the file->f_lock spinlock.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
3 years agoNFS: Fix listxattr receive buffer size
Chuck Lever [Sat, 31 Oct 2020 16:44:25 +0000 (12:44 -0400)]
NFS: Fix listxattr receive buffer size

Certain NFSv4.2/RDMA tests fail with v5.9-rc1.

rpcrdma_convert_kvec() runs off the end of the rl_segments array
because rq_rcv_buf.tail[0].iov_len holds a very large positive
value. The resultant kernel memory corruption is enough to crash
the client system.

Callers of rpc_prepare_reply_pages() must reserve an extra XDR_UNIT
in the maximum decode size for a possible XDR pad of the contents
of the xdr_buf's pages. That guarantees the allocated receive buffer
will be large enough to accommodate the usual contents plus that XDR
pad word.

encode_op_hdr() cannot add that extra word. If it does,
xdr_inline_pages() underruns the length of the tail iovec.

Fixes: 3e1f02123fba ("NFSv4.2: add client side XDR handling for extended attributes")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
3 years agoNFSv4.2: fix failure to unregister shrinker
J. Bruce Fields [Wed, 21 Oct 2020 14:34:15 +0000 (10:34 -0400)]
NFSv4.2: fix failure to unregister shrinker

We forgot to unregister the nfs4_xattr_large_entry_shrinker.

That leaves the global list of shrinkers corrupted after unload of the
nfs module, after which possibly unrelated code that calls
register_shrinker() or unregister_shrinker() gets a BUG() with
"supervisor write access in kernel mode".

And similarly for the nfs4_xattr_large_entry_lru.

Reported-by: Kris Karas <bugs-a17@moonlit-rail.com>
Tested-By: Kris Karas <bugs-a17@moonlit-rail.com>
Fixes: 95ad37f90c33 "NFSv4.2: add client side xattr caching."
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
CC: stable@vger.kernel.org
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
3 years agoMerge branches 'acpi-scan', 'acpi-misc', 'acpi-button' and 'acpi-dptf'
Rafael J. Wysocki [Thu, 12 Nov 2020 15:11:48 +0000 (16:11 +0100)]
Merge branches 'acpi-scan', 'acpi-misc', 'acpi-button' and 'acpi-dptf'

* acpi-scan:
  ACPI: scan: Fix acpi_dma_configure_id() kerneldoc name

* acpi-misc:
  ACPI: GED: fix -Wformat
  ACPI: Fix whitespace inconsistencies

* acpi-button:
  ACPI: button: Add DMI quirk for Medion Akoya E2228T

* acpi-dptf:
  ACPI: DPTF: Support Alder Lake

3 years agogfs2: fix possible reference leak in gfs2_check_blk_type
Zhang Qilong [Sun, 8 Nov 2020 09:27:41 +0000 (17:27 +0800)]
gfs2: fix possible reference leak in gfs2_check_blk_type

In the fail path of gfs2_check_blk_type, forgetting to call
gfs2_glock_dq_uninit will result in rgd_gh reference leak.

Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
3 years agofscrypt: fix inline encryption not used on new files
Eric Biggers [Wed, 11 Nov 2020 01:52:24 +0000 (17:52 -0800)]
fscrypt: fix inline encryption not used on new files

The new helper function fscrypt_prepare_new_inode() runs before
S_ENCRYPTED has been set on the new inode.  This accidentally made
fscrypt_select_encryption_impl() never enable inline encryption on newly
created files, due to its use of fscrypt_needs_contents_encryption()
which only returns true when S_ENCRYPTED is set.

Fix this by using S_ISREG() directly instead of
fscrypt_needs_contents_encryption(), analogous to what
select_encryption_mode() does.

I didn't notice this earlier because by design, the user-visible
behavior is the same (other than performance, potentially) regardless of
whether inline encryption is used or not.

Fixes: a992b20cd4ee ("fscrypt: add fscrypt_prepare_new_inode() and fscrypt_set_context()")
Reviewed-by: Satya Tangirala <satyat@google.com>
Link: https://lore.kernel.org/r/20201111015224.303073-1-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
3 years agocosa: Add missing kfree in error path of cosa_write
Wang Hai [Tue, 10 Nov 2020 14:46:14 +0000 (22:46 +0800)]
cosa: Add missing kfree in error path of cosa_write

If memory allocation for 'kbuf' succeed, cosa_write() doesn't have a
corresponding kfree() in exception handling. Thus add kfree() for this
function implementation.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Acked-by: Jan "Yenya" Kasprzak <kas@fi.muni.cz>
Link: https://lore.kernel.org/r/20201110144614.43194-1-wanghai38@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: switch to the kernel.org patchwork instance
Jakub Kicinski [Tue, 10 Nov 2020 03:51:20 +0000 (19:51 -0800)]
net: switch to the kernel.org patchwork instance

Move to the kernel.org patchwork instance, it has significantly
lower latency for accessing from Europe and the US. Other quirks
include the reply bot.

Link: https://lore.kernel.org/r/20201110035120.642746-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge branch 'cxgb4-ch_ktls-fixes-in-nic-tls-code'
Jakub Kicinski [Thu, 12 Nov 2020 00:30:41 +0000 (16:30 -0800)]
Merge branch 'cxgb4-ch_ktls-fixes-in-nic-tls-code'

Rohit Maheshwari says:

====================
cxgb4/ch_ktls: Fixes in nic tls code

This series helps in fixing multiple nic ktls issues. Series is broken
into 12 patches.

Patch 1 avoids deciding tls packet based on decrypted bit. If its a
retransmit packet which has tls handshake and finish (for encryption),
decrypted bit won't be set there, and so we can't rely on decrypted
bit.

Patch 2 helps supporting linear skb. SKBs were assumed non-linear.
Corrected the length extraction.

Patch 3 fixes the checksum offload update in WR.

Patch 4 fixes kernel panic happening due to creating new skb for each
record. As part of fix driver will use same skb to send out one tls
record (partial data) of the same SKB.

Patch 5 fixes the problem of skb data length smaller than remaining data
of the record.

Patch 6 fixes the handling of SKBs which has tls header alone pkt, but
not starting from beginning.

Patch 7 avoids sending extra data which is used to make a record 16 byte
aligned. We don't need to retransmit those extra few bytes.

Patch 8 handles the cases where retransmit packet has tls starting
exchanges which are prior to tls start marker.

Patch 9 fixes the problem os skb free before HW knows about tcp FIN.

Patch 10 handles the small packet case which has partial TAG bytes only.
HW can't handle those, hence using sw crypto for such pkts.

Patch 11 corrects the potential tcb update problem.

Patch 12 stops the queue if queue reaches threshold value.

v1->v2:
- Corrected fixes tag issue.
- Marked chcr_ktls_sw_fallback() static.

v2->v3:
- Replaced GFP_KERNEL with GFP_ATOMIC.
- Removed mixed fixes.

v3->v4:
- Corrected fixes tag issue.

v4->v5:
- Separated mixed fixes from patch 4.

v5-v6:
- Fixes tag should be at the end.
====================

Link: https://lore.kernel.org/r/20201109105142.15398-1-rohitm@chelsio.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: stop the txq if reaches threshold
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:42 +0000 (16:21 +0530)]
ch_ktls: stop the txq if reaches threshold

Stop the queue and ask for the credits if queue reaches to
threashold.

Fixes: 5a4b9fe7fece ("cxgb4/chcr: complete record tx handling")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: tcb update fails sometimes
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:41 +0000 (16:21 +0530)]
ch_ktls: tcb update fails sometimes

context id and port id should be filled while sending tcb update.

Fixes: 5a4b9fe7fece ("cxgb4/chcr: complete record tx handling")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls/cxgb4: handle partial tag alone SKBs
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:40 +0000 (16:21 +0530)]
ch_ktls/cxgb4: handle partial tag alone SKBs

If TCP congestion caused a very small packets which only has some
part fo the TAG, and that too is not till the end. HW can't handle
such case, so falling back to sw crypto in such cases.

v1->v2:
- Marked chcr_ktls_sw_fallback() static.

Fixes: dc05f3df8fac ("chcr: Handle first or middle part of record")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: don't free skb before sending FIN
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:39 +0000 (16:21 +0530)]
ch_ktls: don't free skb before sending FIN

If its a last packet and fin is set. Make sure FIN is informed
to HW before skb gets freed.

Fixes: 429765a149f1 ("chcr: handle partial end part of a record")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: packet handling prior to start marker
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:38 +0000 (16:21 +0530)]
ch_ktls: packet handling prior to start marker

There could be a case where ACK for tls exchanges prior to start
marker is missed out, and by the time tls is offloaded. This pkt
should not be discarded and handled carefully. It could be
plaintext alone or plaintext + finish as well.

Fixes: 5a4b9fe7fece ("cxgb4/chcr: complete record tx handling")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: Correction in middle record handling
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:37 +0000 (16:21 +0530)]
ch_ktls: Correction in middle record handling

If a record starts in middle, reset TCB UNA so that we could
avoid sending out extra packet which is needed to make it 16
byte aligned to start AES CTR.
Check also considers prev_seq, which should be what is
actually sent, not the skb data length.
Avoid updating partial TAG to HW at any point of time, that's
why we need to check if remaining part is smaller than TAG
size, then reset TX_MAX to be TAG starting sequence number.

Fixes: 5a4b9fe7fece ("cxgb4/chcr: complete record tx handling")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: missing handling of header alone
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:36 +0000 (16:21 +0530)]
ch_ktls: missing handling of header alone

If an skb has only header part which doesn't start from
beginning, is not being handled properly.

Fixes: dc05f3df8fac ("chcr: Handle first or middle part of record")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: Correction in trimmed_len calculation
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:35 +0000 (16:21 +0530)]
ch_ktls: Correction in trimmed_len calculation

trimmed length calculation goes wrong if skb has only tag part
to send. It should be zero if there is no data bytes apart from
TAG.

Fixes: dc05f3df8fac ("chcr: Handle first or middle part of record")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agocxgb4/ch_ktls: creating skbs causes panic
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:34 +0000 (16:21 +0530)]
cxgb4/ch_ktls: creating skbs causes panic

Creating SKB per tls record and freeing the original one causes
panic. There will be race if connection reset is requested. By
freeing original skb, refcnt will be decremented and that means,
there is no pending record to send, and so tls_dev_del will be
requested in control path while SKB of related connection is in
queue.
 Better approach is to use same SKB to send one record (partial
data) at a time. We still have to create a new SKB when partial
last part of a record is requested.
 This fix introduces new API cxgb4_write_partial_sgl() to send
partial part of skb. Present cxgb4_write_sgl can only provide
feasibility to start from an offset which limits to header only
and it can write sgls for the whole skb len. But this new API
will help in both. It can start from any offset and can end
writing in middle of the skb.

v4->v5:
- Removed extra changes.

Fixes: 429765a149f1 ("chcr: handle partial end part of a record")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: Update cheksum information
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:33 +0000 (16:21 +0530)]
ch_ktls: Update cheksum information

Checksum update was missing in the WR.

Fixes: 429765a149f1 ("chcr: handle partial end part of a record")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoch_ktls: Correction in finding correct length
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:32 +0000 (16:21 +0530)]
ch_ktls: Correction in finding correct length

There is a possibility of linear skbs coming in. Correcting
the length extraction logic.

v2->v3:
- Separated un-related changes from this patch.

Fixes: 5a4b9fe7fece ("cxgb4/chcr: complete record tx handling")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agocxgb4/ch_ktls: decrypted bit is not enough
Rohit Maheshwari [Mon, 9 Nov 2020 10:51:31 +0000 (16:21 +0530)]
cxgb4/ch_ktls: decrypted bit is not enough

If skb has retransmit data starting before start marker, e.g. ccs,
decrypted bit won't be set for that, and if it has some data to
encrypt, then it must be given to crypto ULD. So in place of
decrypted, check if socket is tls offloaded. Also, unless skb has
some data to encrypt, no need to give it for tls offload handling.

v2->v3:
- Removed ifdef.

Fixes: 5a4b9fe7fece ("cxgb4/chcr: complete record tx handling")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet/x25: Fix null-ptr-deref in x25_connect
Martin Schiller [Mon, 9 Nov 2020 06:54:49 +0000 (07:54 +0100)]
net/x25: Fix null-ptr-deref in x25_connect

This fixes a regression for blocking connects introduced by commit
4becb7ee5b3d ("net/x25: Fix x25_neigh refcnt leak when x25 disconnect").

The x25->neighbour is already set to "NULL" by x25_disconnect() now,
while a blocking connect is waiting in
x25_wait_for_connection_establishment(). Therefore x25->neighbour must
not be accessed here again and x25->state is also already set to
X25_STATE_0 by x25_disconnect().

Fixes: 4becb7ee5b3d ("net/x25: Fix x25_neigh refcnt leak when x25 disconnect")
Signed-off-by: Martin Schiller <ms@dev.tdt.de>
Reviewed-by: Xie He <xie.he.0141@gmail.com>
Link: https://lore.kernel.org/r/20201109065449.9014-1-ms@dev.tdt.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoarm64: dts: fsl-ls1028a-kontron-sl28: specify in-band mode for ENETC
Michael Walle [Mon, 9 Nov 2020 11:04:36 +0000 (12:04 +0100)]
arm64: dts: fsl-ls1028a-kontron-sl28: specify in-band mode for ENETC

Since commit 71b77a7a27a3 ("enetc: Migrate to PHYLINK and PCS_LYNX") the
network port of the Kontron sl28 board is broken. After the migration to
phylink the device tree has to specify the in-band-mode property. Add
it.

Fixes: 71b77a7a27a3 ("enetc: Migrate to PHYLINK and PCS_LYNX")
Suggested-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20201109110436.5906-1-michael@walle.cc
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agotipc: fix memory leak in tipc_topsrv_start()
Wang Hai [Mon, 9 Nov 2020 14:09:13 +0000 (22:09 +0800)]
tipc: fix memory leak in tipc_topsrv_start()

kmemleak report a memory leak as follows:

unreferenced object 0xffff88810a596800 (size 512):
  comm "ip", pid 21558, jiffies 4297568990 (age 112.120s)
  hex dump (first 32 bytes):
    00 00 00 00 ad 4e ad de ff ff ff ff 00 00 00 00  .....N..........
    ff ff ff ff ff ff ff ff 00 83 60 b0 ff ff ff ff  ..........`.....
  backtrace:
    [<0000000022bbe21f>] tipc_topsrv_init_net+0x1f3/0xa70
    [<00000000fe15ddf7>] ops_init+0xa8/0x3c0
    [<00000000138af6f2>] setup_net+0x2de/0x7e0
    [<000000008c6807a3>] copy_net_ns+0x27d/0x530
    [<000000006b21adbd>] create_new_namespaces+0x382/0xa30
    [<00000000bb169746>] unshare_nsproxy_namespaces+0xa1/0x1d0
    [<00000000fe2e42bc>] ksys_unshare+0x39c/0x780
    [<0000000009ba3b19>] __x64_sys_unshare+0x2d/0x40
    [<00000000614ad866>] do_syscall_64+0x56/0xa0
    [<00000000a1b5ca3c>] entry_SYSCALL_64_after_hwframe+0x44/0xa9

'srv' is malloced in tipc_topsrv_start() but not free before
leaving from the error handling cases. We need to free it.

Fixes: 5c45ab24ac77 ("tipc: make struct tipc_server private for server.c")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Link: https://lore.kernel.org/r/20201109140913.47370-1-wanghai38@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge branch 'stable/for-linus-5.10-rc2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 11 Nov 2020 22:15:06 +0000 (14:15 -0800)]
Merge branch 'stable/for-linus-5.10-rc2' of git://git./linux/kernel/git/konrad/swiotlb

Pull swiotlb fixes from Konrad Rzeszutek Wilk:
 "Two tiny fixes for issues that make drivers under Xen unhappy under
  certain conditions"

* 'stable/for-linus-5.10-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb:
  swiotlb: remove the tbl_dma_addr argument to swiotlb_tbl_map_single
  swiotlb: fix "x86: Don't panic if can not alloc buffer for swiotlb"

3 years agoMerge branch 'net-iucv-fixes-2020-11-09'
Jakub Kicinski [Wed, 11 Nov 2020 02:08:21 +0000 (18:08 -0800)]
Merge branch 'net-iucv-fixes-2020-11-09'

Julian Wiedmann says:

====================
net/iucv: fixes 2020-11-09

One fix in the shutdown path for af_iucv sockets. This is relevant for
stable as well.
Also sending along an update for the Maintainers file.

v1 -> v2: use the correct Fixes tag in patch 1 (Jakub)
====================

Link: https://lore.kernel.org/r/20201109075706.56573-1-jwi@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMAINTAINERS: remove Ursula Braun as s390 network maintainer
Ursula Braun [Mon, 9 Nov 2020 07:57:06 +0000 (08:57 +0100)]
MAINTAINERS: remove Ursula Braun as s390 network maintainer

I am retiring soon. Thus this patch removes myself from the
MAINTAINERS file (s390 network).

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
[jwi: fix up the subject]
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet/af_iucv: fix null pointer dereference on shutdown
Ursula Braun [Mon, 9 Nov 2020 07:57:05 +0000 (08:57 +0100)]
net/af_iucv: fix null pointer dereference on shutdown

syzbot reported the following KASAN finding:

BUG: KASAN: nullptr-dereference in iucv_send_ctrl+0x390/0x3f0 net/iucv/af_iucv.c:385
Read of size 2 at addr 000000000000021e by task syz-executor907/519

CPU: 0 PID: 519 Comm: syz-executor907 Not tainted 5.9.0-syzkaller-07043-gbcf9877ad213 #0
Hardware name: IBM 3906 M04 701 (KVM/Linux)
Call Trace:
 [<00000000c576af60>] unwind_start arch/s390/include/asm/unwind.h:65 [inline]
 [<00000000c576af60>] show_stack+0x180/0x228 arch/s390/kernel/dumpstack.c:135
 [<00000000c9dcd1f8>] __dump_stack lib/dump_stack.c:77 [inline]
 [<00000000c9dcd1f8>] dump_stack+0x268/0x2f0 lib/dump_stack.c:118
 [<00000000c5fed016>] print_address_description.constprop.0+0x5e/0x218 mm/kasan/report.c:383
 [<00000000c5fec82a>] __kasan_report mm/kasan/report.c:517 [inline]
 [<00000000c5fec82a>] kasan_report+0x11a/0x168 mm/kasan/report.c:534
 [<00000000c98b5b60>] iucv_send_ctrl+0x390/0x3f0 net/iucv/af_iucv.c:385
 [<00000000c98b6262>] iucv_sock_shutdown+0x44a/0x4c0 net/iucv/af_iucv.c:1457
 [<00000000c89d3a54>] __sys_shutdown+0x12c/0x1c8 net/socket.c:2204
 [<00000000c89d3b70>] __do_sys_shutdown net/socket.c:2212 [inline]
 [<00000000c89d3b70>] __s390x_sys_shutdown+0x38/0x48 net/socket.c:2210
 [<00000000c9e36eac>] system_call+0xe0/0x28c arch/s390/kernel/entry.S:415

There is nothing to shutdown if a connection has never been established.
Besides that iucv->hs_dev is not yet initialized if a socket is in
IUCV_OPEN state and iucv->path is not yet initialized if socket is in
IUCV_BOUND state.
So, just skip the shutdown calls for a socket in these states.

Fixes: eac3731bd04c ("[S390]: Add AF_IUCV socket support")
Fixes: 82492a355fac ("af_iucv: add shutdown for HS transport")
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
[jwi: correct one Fixes tag]
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agolan743x: fix "BUG: invalid wait context" when setting rx mode
Sven Van Asbroeck [Mon, 9 Nov 2020 20:38:28 +0000 (15:38 -0500)]
lan743x: fix "BUG: invalid wait context" when setting rx mode

In the net core, the struct net_device_ops -> ndo_set_rx_mode()
callback is called with the dev->addr_list_lock spinlock held.

However, this driver's ndo_set_rx_mode callback eventually calls
lan743x_dp_write(), which acquires a mutex. Mutex acquisition
may sleep, and this is not allowed when holding a spinlock.

Fix by removing the dp_lock mutex entirely. Its purpose is to
prevent concurrent accesses to the data port. No concurrent
accesses are possible, because the dev->addr_list_lock
spinlock in the core only lets through one thread at a time.

Fixes: 23f0703c125b ("lan743x: Add main source files for new lan743x driver")
Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
Link: https://lore.kernel.org/r/20201109203828.5115-1-TheSven73@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: dsa: mv88e6xxx: Fix memleak in mv88e6xxx_region_atu_snapshot
zhangxiaoxu [Mon, 9 Nov 2020 14:44:16 +0000 (09:44 -0500)]
net: dsa: mv88e6xxx: Fix memleak in mv88e6xxx_region_atu_snapshot

When mv88e6xxx_fid_map return error, we lost free the table.

Fix it.

Fixes: bfb255428966 ("net: dsa: mv88e6xxx: Add devlink regions")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhangxiaoxu <zhangxiaoxu5@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20201109144416.1540867-1-zhangxiaoxu5@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: Update window_clamp if SOCK_RCVBUF is set
Mao Wenan [Tue, 10 Nov 2020 00:16:31 +0000 (08:16 +0800)]
net: Update window_clamp if SOCK_RCVBUF is set

When net.ipv4.tcp_syncookies=1 and syn flood is happened,
cookie_v4_check or cookie_v6_check tries to redo what
tcp_v4_send_synack or tcp_v6_send_synack did,
rsk_window_clamp will be changed if SOCK_RCVBUF is set,
which will make rcv_wscale is different, the client
still operates with initial window scale and can overshot
granted window, the client use the initial scale but local
server use new scale to advertise window value, and session
work abnormally.

Fixes: e88c64f0a425 ("tcp: allow effective reduction of TCP's rcv-buffer via setsockopt")
Signed-off-by: Mao Wenan <wenan.mao@linux.alibaba.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/1604967391-123737-1-git-send-email-wenan.mao@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: phy: realtek: support paged operations on RTL8201CP
Heiner Kallweit [Sun, 8 Nov 2020 21:44:02 +0000 (22:44 +0100)]
net: phy: realtek: support paged operations on RTL8201CP

The RTL8401-internal PHY identifies as RTL8201CP, and the init
sequence in r8169, copied from vendor driver r8168, uses paged
operations. Therefore set the same paged operation callbacks as
for the other Realtek PHY's.

Fixes: cdafdc29ef75 ("r8169: sync support for RTL8401 with vendor driver")
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/69882f7a-ca2f-e0c7-ae83-c9b6937282cd@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agolan743x: correctly handle chips with internal PHY
Sven Van Asbroeck [Sun, 8 Nov 2020 17:12:24 +0000 (12:12 -0500)]
lan743x: correctly handle chips with internal PHY

Commit 6f197fb63850 ("lan743x: Added fixed link and RGMII support")
assumes that chips with an internal PHY will never have a devicetree
entry. This is incorrect: even for these chips, a devicetree entry
can be useful e.g. to pass the mac address from bootloader to chip:

    &pcie {
            status = "okay";

            host@0 {
                    reg = <0 0 0 0 0>;

                    #address-cells = <3>;
                    #size-cells = <2>;

                    lan7430: ethernet@0 {
                            /* LAN7430 with internal PHY */
                            compatible = "microchip,lan743x";
                            status = "okay";
                            reg = <0 0 0 0 0>;
                            /* filled in by bootloader */
                            local-mac-address = [00 00 00 00 00 00];
                    };
            };
    };

If a devicetree entry is present, the driver will not attach the chip
to its internal phy, and the chip will be non-operational.

Fix by tweaking the phy connection algorithm:
- first try to connect to a phy specified in the devicetree
  (could be 'real' phy, or just a 'fixed-link')
- if that doesn't succeed, try to connect to an internal phy, even
  if the chip has a devnode

Tested on a LAN7430 with internal PHY. I cannot test a device using
fixed-link, as I do not have access to one.

Fixes: 6f197fb63850 ("lan743x: Added fixed link and RGMII support")
Tested-by: Sven Van Asbroeck <thesven73@gmail.com> # lan7430
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Sven Van Asbroeck <thesven73@gmail.com>
Link: https://lore.kernel.org/r/20201108171224.23829-1-TheSven73@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonetlabel: fix our progress tracking in netlbl_unlabel_staticlist()
Paul Moore [Sun, 8 Nov 2020 14:08:26 +0000 (09:08 -0500)]
netlabel: fix our progress tracking in netlbl_unlabel_staticlist()

The current NetLabel code doesn't correctly keep track of the netlink
dump state in some cases, in particular when multiple interfaces with
large configurations are loaded.  The problem manifests itself by not
reporting the full configuration to userspace, even though it is
loaded and active in the kernel.  This patch fixes this by ensuring
that the dump state is properly reset when necessary inside the
netlbl_unlabel_staticlist() function.

Fixes: 8cc44579d1bd ("NetLabel: Introduce static network labels for unlabeled connections")
Signed-off-by: Paul Moore <paul@paul-moore.com>
Link: https://lore.kernel.org/r/160484450633.3752.16512718263560813473.stgit@sifl
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMAINTAINERS: Update repositories for Intel Ethernet Drivers
Tony Nguyen [Tue, 10 Nov 2020 00:07:35 +0000 (16:07 -0800)]
MAINTAINERS: Update repositories for Intel Ethernet Drivers

Update Intel Ethernet Drivers repositories to new locations.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
3 years agoigc: Fix returning wrong statistics
Vinicius Costa Gomes [Fri, 25 Sep 2020 18:35:37 +0000 (11:35 -0700)]
igc: Fix returning wrong statistics

'igc_update_stats()' was not updating 'netdev->stats', so the returned
statistics, for example, requested by:

$ ip -s link show dev enp3s0

were not being updated and were always zero.

Fix by returning a set of statistics that are actually being
updated (adapter->stats64).

Fixes: c9a11c23ceb6 ("igc: Add netdev")
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
3 years agoi40e, xsk: uninitialized variable in i40e_clean_rx_irq_zc()
Dan Carpenter [Wed, 16 Sep 2020 14:32:28 +0000 (17:32 +0300)]
i40e, xsk: uninitialized variable in i40e_clean_rx_irq_zc()

The "failure" variable is used without being initialized.  It should be
set to false.

Fixes: 8cbf74149903 ("i40e, xsk: move buffer allocation out of the Rx processing loop")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: George Kuruvinakunnel <george.kuruvinakunnel@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>