GitHub/moto-9609/android_kernel_motorola_exynos9610.git
7 years agoMerge branch 'kvm-ppc-next' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus...
Radim Krčmář [Thu, 7 Sep 2017 16:29:01 +0000 (18:29 +0200)]
Merge branch 'kvm-ppc-next' of git://git./linux/kernel/git/paulus/powerpc

KVM/PPC update for 4.14

There are various minor fixes and cleanups.  The only new feature is
that we now export information about storage key support to userspace,
so it can advertise it to the guest.

I have pulled in Michael Ellerman's topic/ppc-kvm branch from the
powerpc tree to get a couple of fixes that touch both KVM PPC code and
other PPC code.  That's why there is some arch/powerpc stuff in the
diffstat that isn't arch/powerpc/kvm.

7 years agoMerge tag 'kvm-arm-for-v4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm...
Radim Krčmář [Thu, 7 Sep 2017 16:22:04 +0000 (18:22 +0200)]
Merge tag 'kvm-arm-for-v4.14' of git://git./linux/kernel/git/kvmarm/kvmarm

KVM/ARM Changes for v4.14

Two minor cleanups and improvements, a fix for decoding external abort
types from guests, and added support for migrating the active priority
of interrupts when running a GICv2 guest on a GICv3 host.

7 years agoMerge tag 'kvm-s390-next-4.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Radim Krčmář [Thu, 7 Sep 2017 14:46:46 +0000 (16:46 +0200)]
Merge tag 'kvm-s390-next-4.14-2' of git://git./linux/kernel/git/kvms390/linux

KVM: s390: Fixes and features for 4.14

- merge of topic branch tlb-flushing from the s390 tree to get the
  no-dat base features
- merge of kvm/master to avoid conflicts with additional sthyi fixes
- wire up the no-dat enhancements in KVM
- multiple epoch facility (z14 feature)
- Configuration z/Architecture Mode
- more sthyi fixes
- gdb server range checking fix
- small code cleanups

7 years agoKVM: arm/arm64: Support uaccess of GICC_APRn
Christoffer Dall [Thu, 31 Aug 2017 20:24:25 +0000 (22:24 +0200)]
KVM: arm/arm64: Support uaccess of GICC_APRn

When migrating guests around we need to know the active priorities to
ensure functional virtual interrupt prioritization by the GIC.

This commit clarifies the API and how active priorities of interrupts in
different groups are represented, and implements the accessor functions
for the uaccess register range.

We live with a slight layering violation in accessing GICv3 data
structures from vgic-mmio-v2.c, because anything else just adds too much
complexity for us to deal with (it's not like there's a benefit
elsewhere in the code of an intermediate representation as is the case
with the VMCR).  We accept this, because while doing v3 processing from
a file named something-v2.c can look strange at first, this really is
specific to dealing with the user space interface for something that
looks like a GICv2.

Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
7 years agoKVM: arm/arm64: Extract GICv3 max APRn index calculation
Christoffer Dall [Fri, 1 Sep 2017 09:41:52 +0000 (11:41 +0200)]
KVM: arm/arm64: Extract GICv3 max APRn index calculation

As we are about to access the APRs from the GICv2 uaccess interface,
make this logic generally available.

Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
7 years agoKVM: arm/arm64: vITS: Drop its_ite->lpi field
Marc Zyngier [Fri, 1 Sep 2017 16:51:56 +0000 (17:51 +0100)]
KVM: arm/arm64: vITS: Drop its_ite->lpi field

For unknown reasons, the its_ite data structure carries an "lpi" field
which contains the intid of the LPI. This is an obvious duplication
of the vgic_irq->intid field, so let's fix the only user and remove
the now useless field.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Reviewed-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
7 years agoKVM: arm/arm64: vgic: constify seq_operations and file_operations
Arvind Yadav [Wed, 23 Aug 2017 06:55:36 +0000 (12:25 +0530)]
KVM: arm/arm64: vgic: constify seq_operations and file_operations

vgic_debug_seq_ops and file_operations are not supposed to change
at runtime and none of the structures is modified.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
7 years agoKVM: arm/arm64: Fix guest external abort matching
James Morse [Tue, 18 Jul 2017 12:37:41 +0000 (13:37 +0100)]
KVM: arm/arm64: Fix guest external abort matching

The ARM-ARM has two bits in the ESR/HSR relevant to external aborts.
A range of {I,D}FSC values (of which bit 5 is always set) and bit 9 'EA'
which provides:
> an IMPLEMENTATION DEFINED classification of External Aborts.

This bit is in addition to the {I,D}FSC range, and has an implementation
defined meaning. KVM should always ignore this bit when handling external
aborts from a guest.

Remove the ESR_ELx_EA definition and rewrite its helper
kvm_vcpu_dabt_isextabt() to check the {I,D}FSC range. This merges
kvm_vcpu_dabt_isextabt() and the recently added is_abort_sea() helper.

CC: Tyler Baicar <tbaicar@codeaurora.org>
Reported-by: gengdongjiu <gengdj.1984@gmail.com>
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
7 years agoKVM: PPC: Book3S HV: Fix memory leak in kvm_vm_ioctl_get_htab_fd
nixiaoming [Thu, 31 Aug 2017 10:51:09 +0000 (18:51 +0800)]
KVM: PPC: Book3S HV: Fix memory leak in kvm_vm_ioctl_get_htab_fd

We do ctx = kzalloc(sizeof(*ctx), GFP_KERNEL) and then later on call
anon_inode_getfd(), but if that fails we don't free ctx, so that
memory gets leaked.  To fix it, this adds kfree(ctx) in the failure
path.

Signed-off-by: nixiaoming <nixiaoming@huawei.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: s390: vsie: cleanup mcck reinjection
David Hildenbrand [Wed, 30 Aug 2017 16:06:02 +0000 (18:06 +0200)]
KVM: s390: vsie: cleanup mcck reinjection

The machine check information is part of the vsie_page.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170830160603.5452-4-david@redhat.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: s390: use WARN_ON_ONCE only for checking
David Hildenbrand [Wed, 30 Aug 2017 16:06:01 +0000 (18:06 +0200)]
KVM: s390: use WARN_ON_ONCE only for checking

Move the real logic that always has to be executed out of the
WARN_ON_ONCE.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170830160603.5452-3-david@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: s390: guestdbg: fix range check
David Hildenbrand [Wed, 30 Aug 2017 16:06:00 +0000 (18:06 +0200)]
KVM: s390: guestdbg: fix range check

Looks like the "overflowing" range check is wrong.

|=======b-------a=======|

addr >= a || addr <= b

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170830160603.5452-2-david@redhat.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoMerge remote-tracking branch 'remotes/powerpc/topic/ppc-kvm' into kvm-ppc-next
Paul Mackerras [Thu, 31 Aug 2017 02:37:03 +0000 (12:37 +1000)]
Merge remote-tracking branch 'remotes/powerpc/topic/ppc-kvm' into kvm-ppc-next

This merges in the 'ppc-kvm' topic branch from the powerpc tree in
order to bring in some fixes which touch both powerpc and KVM code.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: Book3S HV: Report storage key support to userspace
Paul Mackerras [Fri, 25 Aug 2017 09:53:39 +0000 (19:53 +1000)]
KVM: PPC: Book3S HV: Report storage key support to userspace

This adds information about storage keys to the struct returned by
the KVM_PPC_GET_SMMU_INFO ioctl.  The new fields replace a pad field,
which was zeroed by previous kernel versions.  Thus userspace that
knows about the new fields will see zeroes when running on an older
kernel, indicating that storage keys are not supported.  The size of
the structure has not changed.

The number of keys is hard-coded for the CPUs supported by HV KVM,
which is just POWER7, POWER8 and POWER9.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: Book3S HV: Fix case where HDEC is treated as 32-bit on POWER9
Paul Mackerras [Fri, 25 Aug 2017 09:52:12 +0000 (19:52 +1000)]
KVM: PPC: Book3S HV: Fix case where HDEC is treated as 32-bit on POWER9

Commit 2f2724630f7a ("KVM: PPC: Book3S HV: Cope with host using large
decrementer mode", 2017-05-22) added code to treat the hypervisor
decrementer (HDEC) as a 64-bit value on POWER9 rather than 32-bit.
Unfortunately, that commit missed one place where HDEC is treated
as a 32-bit value.  This fixes it.

This bug should not have any user-visible consequences that I can
think of, beyond an occasional unnecessary exit to the host kernel.
If the hypervisor decrementer has gone negative, then the bottom
32 bits will be negative for about 4 seconds after that, so as
long as we get out of the guest within those 4 seconds we won't
conclude that the HDEC interrupt is spurious.

Reported-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Fixes: 2f2724630f7a ("KVM: PPC: Book3S HV: Cope with host using large decrementer mode")
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: Book3S HV: Fix invalid use of register expression
Andreas Schwab [Tue, 15 Aug 2017 04:37:01 +0000 (14:37 +1000)]
KVM: PPC: Book3S HV: Fix invalid use of register expression

binutils >= 2.26 now warns about misuse of register expressions in
assembler operands that are actually literals. In this instance r0 is
being used where a literal 0 should be used.

Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
[mpe: Split into separate KVM patch, tweak change log]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: Book3S HV: Fix H_REGISTER_VPA VPA size validation
Nicholas Piggin [Sun, 13 Aug 2017 01:33:38 +0000 (11:33 +1000)]
KVM: PPC: Book3S HV: Fix H_REGISTER_VPA VPA size validation

KVM currently validates the size of the VPA registered by the client
against sizeof(struct lppaca), however we align (and therefore size)
that struct to 1kB to avoid crossing a 4kB boundary in the client.

PAPR calls for sizes >= 640 bytes to be accepted. Hard code this with
a comment.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: Book3S HV: Fix setting of storage key in H_ENTER
Ram Pai [Mon, 31 Jul 2017 21:39:59 +0000 (14:39 -0700)]
KVM: PPC: Book3S HV: Fix setting of storage key in H_ENTER

In handling a H_ENTER hypercall, the code in kvmppc_do_h_enter
clobbers the high-order two bits of the storage key, which is stored
in a split field in the second doubleword of the HPTE.  Any storage
key number above 7 hence fails to operate correctly.

This makes sure we preserve all the bits of the storage key.

Acked-by: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Ram Pai <linuxram@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: e500mc: Fix a NULL dereference
Dan Carpenter [Thu, 13 Jul 2017 07:38:49 +0000 (10:38 +0300)]
KVM: PPC: e500mc: Fix a NULL dereference

We should set "err = -ENOMEM;", otherwise it means we're returning
ERR_PTR(0) which is NULL.  It results in a NULL pointer dereference in
the caller.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: PPC: e500: Fix some NULL dereferences on error
Dan Carpenter [Thu, 13 Jul 2017 07:38:29 +0000 (10:38 +0300)]
KVM: PPC: e500: Fix some NULL dereferences on error

There are some error paths in kvmppc_core_vcpu_create_e500() where we
forget to set the error code.  It means that we return ERR_PTR(0) which
is NULL and it results in a NULL pointer dereference in the caller.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
7 years agoKVM: s390: we are always in czam mode
David Hildenbrand [Tue, 29 Aug 2017 14:31:08 +0000 (16:31 +0200)]
KVM: s390: we are always in czam mode

Independent of the underlying hardware, kvm will now always handle
SIGP SET ARCHITECTURE as if czam were enabled. Therefore, let's not
only forward that bit but always set it.

While at it, add a comment regarding STHYI.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170829143108.14703-1-david@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: s390: expose no-DAT to guest and migration support
Claudio Imbrenda [Mon, 29 Aug 2016 13:56:55 +0000 (15:56 +0200)]
KVM: s390: expose no-DAT to guest and migration support

The STFLE bit 147 indicates whether the ESSA no-DAT operation code is
valid, the bit is not normally provided to the host; the host is
instead provided with an SCLP bit that indicates whether guests can
support the feature.

This patch:
* enables the STFLE bit in the guest if the corresponding SCLP bit is
  present in the host.
* adds support for migrating the no-DAT bit in the PGSTEs
* fixes the software interpretation of the ESSA instruction that is
  used when migrating, both for the new operation code and for the old
  "set stable", as per specifications.

Signed-off-by: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: s390: sthyi: remove invalid guest write access
Heiko Carstens [Thu, 3 Aug 2017 11:05:40 +0000 (13:05 +0200)]
KVM: s390: sthyi: remove invalid guest write access

handle_sthyi() always writes to guest memory if the sthyi function
code is zero in order to fault in the page that later is written to.

However a function code of zero does not necessarily mean that a write
to guest memory happens: if the KVM host is running as a second level
guest under z/VM 6.2 the sthyi instruction is indicated to be
available to the KVM host, however if the instruction is executed it
will always return with a return code that indicates "unsupported
function code".

In such a case handle_sthyi() must not write to guest memory. This
means that the prior write access to fault in the guest page may
result in invalid guest exceptions, and/or invalid data modification.

In order to be architecture compliant simply remove the write_guest()
call.

Given that the guest assumed a write access anyway, this fix does not
qualify for -stable. This just makes sure the sthyi handler is
architecture compliant.

Fixes: 95ca2cb57985 ("KVM: s390: Add sthyi emulation")
Reviewed-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: s390: Multiple Epoch Facility support
Collin L. Walling [Tue, 26 Jul 2016 19:29:44 +0000 (15:29 -0400)]
KVM: s390: Multiple Epoch Facility support

Allow for the enablement of MEF and the support for the extended
epoch in SIE and VSIE for the extended guest TOD-Clock.

A new interface is used for getting/setting a guest's extended TOD-Clock
that uses a single ioctl invocation, KVM_S390_VM_TOD_EXT.  Since the
host time is a moving target that might see an epoch switch or STP sync
checks we need an atomic ioctl and cannot use the exisiting two
interfaces. The old method of getting and setting the guest TOD-Clock is
still retained and is used when the old ioctls are called.

Signed-off-by: Collin L. Walling <walling@linux.vnet.ibm.com>
Reviewed-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Reviewed-by: Jason J. Herne <jjherne@linux.vnet.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: PPC: Book3S HV: POWER9 does not require secondary thread management
Nicholas Piggin [Fri, 25 Aug 2017 04:30:33 +0000 (14:30 +1000)]
KVM: PPC: Book3S HV: POWER9 does not require secondary thread management

POWER9 CPUs have independent MMU contexts per thread, so KVM does not
need to quiesce secondary threads, so the hwthread_req/hwthread_state
protocol does not have to be used. So patch it away on POWER9, and patch
away the branch from the Linux idle wakeup to kvm_start_guest that is
never used.

Add a warning and error out of kvmppc_grab_hwthread in case it is ever
called on POWER9.

This avoids a hwsync in the idle wakeup path on POWER9.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
[mpe: Use WARN(...) instead of WARN_ON()/pr_err(...)]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agoKVM: s390: Support Configuration z/Architecture Mode
Jason J. Herne [Fri, 24 Feb 2017 15:01:30 +0000 (10:01 -0500)]
KVM: s390: Support Configuration z/Architecture Mode

kvm has always supported the concept of starting in z/Arch mode so let's
reflect the feature bit to the guest.

Also, we change sigp set architecture to reject any request to change
architecture modes.

Signed-off-by: Jason J. Herne <jjherne@linux.vnet.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoMerge tag 'kvm-s390-master-4.13-2' into kvms390/next
Christian Borntraeger [Mon, 28 Aug 2017 11:46:32 +0000 (13:46 +0200)]
Merge tag 'kvm-s390-master-4.13-2' into kvms390/next

Additional fixes on top of these two
 - missing inline assembly constraint
 - wrong exception handling

are necessary

7 years agokvm: nVMX: Validate the virtual-APIC address on nested VM-entry
Jim Mattson [Thu, 24 Aug 2017 20:24:47 +0000 (13:24 -0700)]
kvm: nVMX: Validate the virtual-APIC address on nested VM-entry

According to the SDM, if the "use TPR shadow" VM-execution control is
1, bits 11:0 of the virtual-APIC address must be 0 and the address
should set any bits beyond the processor's physical-address width.

Signed-off-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: nVMX: Fix trying to cancel vmlauch/vmresume
Wanpeng Li [Thu, 24 Aug 2017 10:35:11 +0000 (03:35 -0700)]
KVM: nVMX: Fix trying to cancel vmlauch/vmresume

------------[ cut here ]------------
WARNING: CPU: 7 PID: 3861 at /home/kernel/ssd/kvm/arch/x86/kvm//vmx.c:11299 nested_vmx_vmexit+0x176e/0x1980 [kvm_intel]
CPU: 7 PID: 3861 Comm: qemu-system-x86 Tainted: G        W  OE   4.13.0-rc4+ #11
RIP: 0010:nested_vmx_vmexit+0x176e/0x1980 [kvm_intel]
Call Trace:
 ? kvm_multiple_exception+0x149/0x170 [kvm]
 ? handle_emulation_failure+0x79/0x230 [kvm]
 ? load_vmcs12_host_state+0xa80/0xa80 [kvm_intel]
 ? check_chain_key+0x137/0x1e0
 ? reexecute_instruction.part.168+0x130/0x130 [kvm]
 nested_vmx_inject_exception_vmexit+0xb7/0x100 [kvm_intel]
 ? nested_vmx_inject_exception_vmexit+0xb7/0x100 [kvm_intel]
 vmx_queue_exception+0x197/0x300 [kvm_intel]
 kvm_arch_vcpu_ioctl_run+0x1b0c/0x2c90 [kvm]
 ? kvm_arch_vcpu_runnable+0x220/0x220 [kvm]
 ? preempt_count_sub+0x18/0xc0
 ? restart_apic_timer+0x17d/0x300 [kvm]
 ? kvm_lapic_restart_hv_timer+0x37/0x50 [kvm]
 ? kvm_arch_vcpu_load+0x1d8/0x350 [kvm]
 kvm_vcpu_ioctl+0x4e4/0x910 [kvm]
 ? kvm_vcpu_ioctl+0x4e4/0x910 [kvm]
 ? kvm_dev_ioctl+0xbe0/0xbe0 [kvm]

The flag "nested_run_pending", which can override the decision of which should run
next, L1 or L2. nested_run_pending=1 means that we *must* run L2 next, not L1. This
is necessary in particular when L1 did a VMLAUNCH of L2 and therefore expects L2 to
be run (and perhaps be injected with an event it specified, etc.). Nested_run_pending
is especially intended to avoid switching  to L1 in the injection decision-point.

This can be handled just like the other cases in vmx_check_nested_events, instead of
having a special case in vmx_queue_exception.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: X86: Fix loss of exception which has not yet been injected
Wanpeng Li [Thu, 24 Aug 2017 10:35:09 +0000 (03:35 -0700)]
KVM: X86: Fix loss of exception which has not yet been injected

vmx_complete_interrupts() assumes that the exception is always injected,
so it can be dropped by kvm_clear_exception_queue().  However,
an exception cannot be injected immediately if it is: 1) originally
destined to a nested guest; 2) trapped to cause a vmexit; 3) happening
right after VMLAUNCH/VMRESUME, i.e. when nested_run_pending is true.

This patch applies to exceptions the same algorithm that is used for
NMIs, replacing exception.reinject with "exception.injected" (equivalent
to nmi_injected).

exception.pending now represents an exception that is queued and whose
side effects (e.g., update RFLAGS.RF or DR7) have not been applied yet.
If exception.pending is true, the exception might result in a nested
vmexit instead, too (in which case the side effects must not be applied).

exception.injected instead represents an exception that is going to be
injected into the guest at the next vmentry.

Reported-by: Radim Krčmář <rkrcmar@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: VMX: use kvm_event_needs_reinjection
Wanpeng Li [Thu, 24 Aug 2017 10:35:08 +0000 (03:35 -0700)]
KVM: VMX: use kvm_event_needs_reinjection

Use kvm_event_needs_reinjection() encapsulation.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: speedup update_permission_bitmask
Paolo Bonzini [Thu, 24 Aug 2017 15:37:25 +0000 (17:37 +0200)]
KVM: MMU: speedup update_permission_bitmask

update_permission_bitmask currently does a 128-iteration loop to,
essentially, compute a constant array.  Computing the 8 bits in parallel
reduces it to 16 iterations, and is enough to speed it up substantially
because many boolean operations in the inner loop become constants or
simplify noticeably.

Because update_permission_bitmask is actually the top item in the profile
for nested vmexits, this speeds up an L2->L1 vmexit by about ten thousand
clock cycles, or up to 30%:

                                         before     after
   cpuid                                 35173      25954
   vmcall                                35122      27079
   inl_from_pmtimer                      52635      42675
   inl_from_qemu                         53604      44599
   inl_from_kernel                       38498      30798
   outl_to_kernel                        34508      28816
   wr_tsc_adjust_msr                     34185      26818
   rd_tsc_adjust_msr                     37409      27049
   mmio-no-eventfd:pci-mem               50563      45276
   mmio-wildcard-eventfd:pci-mem         34495      30823
   mmio-datamatch-eventfd:pci-mem        35612      31071
   portio-no-eventfd:pci-io              44925      40661
   portio-wildcard-eventfd:pci-io        29708      27269
   portio-datamatch-eventfd:pci-io       31135      27164

(I wrote a small C program to compare the tables for all values of CR0.WP,
CR4.SMAP and CR4.SMEP, and they match).

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: Expose the LA57 feature to VM.
Yu Zhang [Thu, 24 Aug 2017 12:27:56 +0000 (20:27 +0800)]
KVM: MMU: Expose the LA57 feature to VM.

This patch exposes 5 level page table feature to the VM.
At the same time, the canonical virtual address checking is
extended to support both 48-bits and 57-bits address width.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: Add 5 level EPT & Shadow page table support.
Yu Zhang [Thu, 24 Aug 2017 12:27:55 +0000 (20:27 +0800)]
KVM: MMU: Add 5 level EPT & Shadow page table support.

Extends the shadow paging code, so that 5 level shadow page
table can be constructed if VM is running in 5 level paging
mode.

Also extends the ept code, so that 5 level ept table can be
constructed if maxphysaddr of VM exceeds 48 bits. Unlike the
shadow logic, KVM should still use 4 level ept table for a VM
whose physical address width is less than 48 bits, even when
the VM is running in 5 level paging mode.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
[Unconditionally reset the MMU context in kvm_cpuid_update.
 Changing MAXPHYADDR invalidates the reserved bit bitmasks.
 - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: Rename PT64_ROOT_LEVEL to PT64_ROOT_4LEVEL.
Yu Zhang [Thu, 24 Aug 2017 12:27:54 +0000 (20:27 +0800)]
KVM: MMU: Rename PT64_ROOT_LEVEL to PT64_ROOT_4LEVEL.

Now we have 4 level page table and 5 level page table in 64 bits
long mode, let's rename the PT64_ROOT_LEVEL to PT64_ROOT_4LEVEL,
then we can use PT64_ROOT_5LEVEL for 5 level page table, it's
helpful to make the code more clear.

Also PT64_ROOT_MAX_LEVEL is defined as 4, so that we can just
redefine it to 5 whenever a replacement is needed for 5 level
paging.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: check guest CR3 reserved bits based on its physical address width.
Yu Zhang [Thu, 24 Aug 2017 12:27:53 +0000 (20:27 +0800)]
KVM: MMU: check guest CR3 reserved bits based on its physical address width.

Currently, KVM uses CR3_L_MODE_RESERVED_BITS to check the
reserved bits in CR3. Yet the length of reserved bits in
guest CR3 should be based on the physical address width
exposed to the VM. This patch changes CR3 check logic to
calculate the reserved bits at runtime.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: x86: Add return value to kvm_cpuid().
Yu Zhang [Thu, 24 Aug 2017 12:27:52 +0000 (20:27 +0800)]
KVM: x86: Add return value to kvm_cpuid().

Return false in kvm_cpuid() when it fails to find the cpuid
entry. Also, this routine(and its caller) is optimized with
a new argument - check_limit, so that the check_cpuid_limit()
fall back can be avoided.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agokvm: vmx: Raise #UD on unsupported XSAVES/XRSTORS
Paolo Bonzini [Thu, 24 Aug 2017 12:48:03 +0000 (14:48 +0200)]
kvm: vmx: Raise #UD on unsupported XSAVES/XRSTORS

A guest may not be configured to support XSAVES/XRSTORS, even when the host
does. If the guest does not support XSAVES/XRSTORS, clear the secondary
execution control so that the processor will raise #UD.

Also clear the "allowed-1" bit for XSAVES/XRSTORS exiting in the
IA32_VMX_PROCBASED_CTLS2 MSR, and pass through VMCS12's control in
the VMCS02.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agokvm: vmx: Raise #UD on unsupported RDSEED
Jim Mattson [Wed, 23 Aug 2017 23:32:03 +0000 (16:32 -0700)]
kvm: vmx: Raise #UD on unsupported RDSEED

A guest may not be configured to support RDSEED, even when the host
does. If the guest does not support RDSEED, intercept the instruction
and synthesize #UD. Also clear the "allowed-1" bit for RDSEED exiting
in the IA32_VMX_PROCBASED_CTLS2 MSR.

Signed-off-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agokvm: vmx: Raise #UD on unsupported RDRAND
Jim Mattson [Wed, 23 Aug 2017 23:32:04 +0000 (16:32 -0700)]
kvm: vmx: Raise #UD on unsupported RDRAND

A guest may not be configured to support RDRAND, even when the host
does. If the guest does not support RDRAND, intercept the instruction
and synthesize #UD. Also clear the "allowed-1" bit for RDRAND exiting
in the IA32_VMX_PROCBASED_CTLS2 MSR.

Signed-off-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: VMX: cache secondary exec controls
Paolo Bonzini [Thu, 24 Aug 2017 11:55:35 +0000 (13:55 +0200)]
KVM: VMX: cache secondary exec controls

Currently, secondary execution controls are divided in three groups:

- static, depending mostly on the module arguments or the processor
  (vmx_secondary_exec_control)

- static, depending on CPUID (vmx_cpuid_update)

- dynamic, depending on nested VMX or local APIC state

Because walking CPUID is expensive, prepare_vmcs02 is using only
the first group.  This however is unnecessarily complicated.  Just
cache the static secondary execution controls, and then prepare_vmcs02
does not need to compute them every time.  Computation of all static
secondary execution controls is now kept in a single function,
vmx_compute_secondary_exec_control.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: SVM: Enable Virtual GIF feature
Janakarajan Natarajan [Wed, 23 Aug 2017 14:57:19 +0000 (09:57 -0500)]
KVM: SVM: Enable Virtual GIF feature

Enable the Virtual GIF feature. This is done by setting bit 25 at position
60h in the vmcb.

With this feature enabled, the processor uses bit 9 at position 60h as the
virtual GIF when executing STGI/CLGI instructions.

Since the execution of STGI by the L1 hypervisor does not cause a return to
the outermost (L0) hypervisor, the enable_irq_window and enable_nmi_window
are modified.

The IRQ window will be opened even if GIF is not set, under the assumption
that on resuming the L1 hypervisor the IRQ will be held pending until the
processor executes the STGI instruction.

For the NMI window, the STGI intercept is set. This will assist in opening
the window only when GIF=1.

Signed-off-by: Janakarajan Natarajan <Janakarajan.Natarajan@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: SVM: Add Virtual GIF feature definition
Janakarajan Natarajan [Wed, 23 Aug 2017 14:57:18 +0000 (09:57 -0500)]
KVM: SVM: Add Virtual GIF feature definition

Add a new cpufeature definition for Virtual GIF.

Signed-off-by: Janakarajan Natarajan <Janakarajan.Natarajan@amd.com>
Reviewed-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: s390: sthyi: fix specification exception detection
Heiko Carstens [Thu, 3 Aug 2017 12:27:30 +0000 (14:27 +0200)]
KVM: s390: sthyi: fix specification exception detection

sthyi should only generate a specification exception if the function
code is zero and the response buffer is not on a 4k boundary.

The current code would also test for unknown function codes if the
response buffer, that is currently only defined for function code 0,
is not on a 4k boundary and incorrectly inject a specification
exception instead of returning with condition code 3 and return code 4
(unsupported function code).

Fix this by moving the boundary check.

Fixes: 95ca2cb57985 ("KVM: s390: Add sthyi emulation")
Cc: <stable@vger.kernel.org> # 4.8+
Reviewed-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: s390: sthyi: fix sthyi inline assembly
Heiko Carstens [Thu, 3 Aug 2017 11:05:11 +0000 (13:05 +0200)]
KVM: s390: sthyi: fix sthyi inline assembly

The sthyi inline assembly misses register r3 within the clobber
list. The sthyi instruction will always write a return code to
register "R2+1", which in this case would be r3. Due to that we may
have register corruption and see host crashes or data corruption
depending on how gcc decided to allocate and use registers during
compile time.

Fixes: 95ca2cb57985 ("KVM: s390: Add sthyi emulation")
Cc: <stable@vger.kernel.org> # 4.8+
Reviewed-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
7 years agoKVM: VMX: always require WB memory type for EPT
David Hildenbrand [Thu, 10 Aug 2017 21:15:29 +0000 (23:15 +0200)]
KVM: VMX: always require WB memory type for EPT

We already always set that type but don't check if it is supported. Also
for nVMX, we only support WB for now. Let's just require it.

Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: VMX: cleanup EPTP definitions
David Hildenbrand [Thu, 10 Aug 2017 21:15:28 +0000 (23:15 +0200)]
KVM: VMX: cleanup EPTP definitions

Don't use shifts, tag them correctly as EPTP and use better matching
names (PWL vs. GAW).

Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: SVM: delete avic_vm_id_bitmap (2 megabyte static array)
Denys Vlasenko [Fri, 11 Aug 2017 20:11:58 +0000 (22:11 +0200)]
KVM: SVM: delete avic_vm_id_bitmap (2 megabyte static array)

With lightly tweaked defconfig:

    text    data     bss      dec     hex filename
11259661 5109408 2981888 19350957 12745ad vmlinux.before
11259661 5109408  884736 17253805 10745ad vmlinux.after

Only compile-tested.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: pbonzini@redhat.com
Cc: rkrcmar@redhat.com
Cc: tglx@linutronix.de
Cc: mingo@redhat.com
Cc: hpa@zytor.com
Cc: x86@kernel.org
Cc: kvm@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: x86: fix use of L1 MMIO areas in nested guests
Paolo Bonzini [Thu, 17 Aug 2017 16:36:58 +0000 (18:36 +0200)]
KVM: x86: fix use of L1 MMIO areas in nested guests

There is currently some confusion between nested and L1 GPAs.  The
assignment to "direct" in kvm_mmu_page_fault tries to fix that, but
it is not enough.  What this patch does is fence off the MMIO cache
completely when using shadow nested page tables, since we have neither
a GVA nor an L1 GPA to put in the cache.  This also allows some
simplifications in kvm_mmu_page_fault and FNAME(page_fault).

The EPT misconfig likewise does not have an L1 GPA to pass to
kvm_io_bus_write, so that must be skipped for guest mode.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
[Changed comment to say "GPAs" instead of "L1's physical addresses", as
 per David's review. - Radim]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: x86: Avoid guest page table walk when gpa_available is set
Brijesh Singh [Thu, 17 Aug 2017 16:36:57 +0000 (18:36 +0200)]
KVM: x86: Avoid guest page table walk when gpa_available is set

When a guest causes a page fault which requires emulation, the
vcpu->arch.gpa_available flag is set to indicate that cr2 contains a
valid GPA.

Currently, emulator_read_write_onepage() makes use of gpa_available flag
to avoid a guest page walk for a known MMIO regions. Lets not limit
the gpa_available optimization to just MMIO region. The patch extends
the check to avoid page walk whenever gpa_available flag is set.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
[Fix EPT=0 according to Wanpeng Li's fix, plus ensure VMX also uses the
 new code. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
[Moved "ret < 0" to the else brach, as per David's review. - Radim]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: x86: simplify ept_misconfig
Paolo Bonzini [Thu, 17 Aug 2017 16:36:56 +0000 (18:36 +0200)]
KVM: x86: simplify ept_misconfig

Calling handle_mmio_page_fault() has been unnecessary since commit
e9ee956e311d ("KVM: x86: MMU: Move handle_mmio_page_fault() call to
kvm_mmu_page_fault()", 2016-02-22).

handle_mmio_page_fault() can now be made static.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agopowerpc/mm: Rename find_linux_pte_or_hugepte()
Aneesh Kumar K.V [Thu, 27 Jul 2017 06:24:53 +0000 (11:54 +0530)]
powerpc/mm: Rename find_linux_pte_or_hugepte()

Add newer helpers to make the function usage simpler. It is always
recommended to use find_current_mm_pte() for walking the page table.
If we cannot use find_current_mm_pte(), it should be documented why
the said usage of __find_linux_pte() is safe against a parallel THP
split.

For now we have KVM code using __find_linux_pte(). This is because kvm
code ends up calling __find_linux_pte() in real mode with MSR_EE=0 but
with PACA soft_enabled = 1. We may want to fix that later and make
sure we keep the MSR_EE and PACA soft_enabled in sync. When we do that
we can switch kvm to use find_linux_pte().

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agokvm: avoid uninitialized-variable warnings
Arnd Bergmann [Thu, 10 Aug 2017 12:14:39 +0000 (14:14 +0200)]
kvm: avoid uninitialized-variable warnings

When PAGE_OFFSET is not a compile-time constant, we run into
warnings from the use of kvm_is_error_hva() that the compiler
cannot optimize out:

arch/arm/kvm/../../../virt/kvm/kvm_main.c: In function '__kvm_gfn_to_hva_cache_init':
arch/arm/kvm/../../../virt/kvm/kvm_main.c:1978:14: error: 'nr_pages_avail' may be used uninitialized in this function [-Werror=maybe-uninitialized]
arch/arm/kvm/../../../virt/kvm/kvm_main.c: In function 'gfn_to_page_many_atomic':
arch/arm/kvm/../../../virt/kvm/kvm_main.c:1660:5: error: 'entry' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This adds fake initializations to the two instances I ran into.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agokvm: x86: Disallow illegal IA32_APIC_BASE MSR values
Jim Mattson [Thu, 10 Aug 2017 17:14:13 +0000 (10:14 -0700)]
kvm: x86: Disallow illegal IA32_APIC_BASE MSR values

Host-initiated writes to the IA32_APIC_BASE MSR do not have to follow
local APIC state transition constraints, but the value written must be
valid.

Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: Bail out immediately if there is no available mmu page
Wanpeng Li [Thu, 10 Aug 2017 23:28:02 +0000 (16:28 -0700)]
KVM: MMU: Bail out immediately if there is no available mmu page

Bailing out immediately if there is no available mmu page to alloc.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MMU: Fix softlockup due to mmu_lock is held too long
Wanpeng Li [Thu, 10 Aug 2017 13:55:51 +0000 (06:55 -0700)]
KVM: MMU: Fix softlockup due to mmu_lock is held too long

watchdog: BUG: soft lockup - CPU#5 stuck for 22s! [warn_test:3089]
 irq event stamp: 20532
 hardirqs last  enabled at (20531): [<ffffffff8e9b6908>] restore_regs_and_iret+0x0/0x1d
 hardirqs last disabled at (20532): [<ffffffff8e9b7ae8>] apic_timer_interrupt+0x98/0xb0
 softirqs last  enabled at (8266): [<ffffffff8e9badc6>] __do_softirq+0x206/0x4c1
 softirqs last disabled at (8253): [<ffffffff8e083918>] irq_exit+0xf8/0x100
 CPU: 5 PID: 3089 Comm: warn_test Tainted: G           OE   4.13.0-rc3+ #8
 RIP: 0010:kvm_mmu_prepare_zap_page+0x72/0x4b0 [kvm]
 Call Trace:
  make_mmu_pages_available.isra.120+0x71/0xc0 [kvm]
  kvm_mmu_load+0x1cf/0x410 [kvm]
  kvm_arch_vcpu_ioctl_run+0x1316/0x1bf0 [kvm]
  kvm_vcpu_ioctl+0x340/0x700 [kvm]
  ? kvm_vcpu_ioctl+0x340/0x700 [kvm]
  ? __fget+0xfc/0x210
  do_vfs_ioctl+0xa4/0x6a0
  ? __fget+0x11d/0x210
  SyS_ioctl+0x79/0x90
  entry_SYSCALL_64_fastpath+0x23/0xc2
  ? __this_cpu_preempt_check+0x13/0x20

This can be reproduced readily by ept=N and running syzkaller tests since
many syzkaller testcases don't setup any memory regions. However, if ept=Y
rmode identity map will be created, then kvm_mmu_calculate_mmu_pages() will
extend the number of VM's mmu pages to at least KVM_MIN_ALLOC_MMU_PAGES
which just hide the issue.

I saw the scenario kvm->arch.n_max_mmu_pages == 0 && kvm->arch.n_used_mmu_pages == 1,
so there is one active mmu page on the list, kvm_mmu_prepare_zap_page() fails
to zap any pages, however prepare_zap_oldest_mmu_page() always returns true.
It incurs infinite loop in make_mmu_pages_available() which causes mmu->lock
softlockup.

This patch fixes it by setting the return value of prepare_zap_oldest_mmu_page()
according to whether or not there is mmu page zapped.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: nVMX: validate eptp pointer
David Hildenbrand [Thu, 10 Aug 2017 21:36:54 +0000 (23:36 +0200)]
KVM: nVMX: validate eptp pointer

Let's reuse the function introduced with eptp switching.

We don't explicitly have to check against enable_ept_ad_bits, as this
is implicitly done when checking against nested_vmx_ept_caps in
valid_ept_address().

Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: MAINTAINERS improvements
Andrew Jones [Fri, 11 Aug 2017 12:29:20 +0000 (14:29 +0200)]
KVM: MAINTAINERS improvements

Remove nonexistent files, allow less awkward expressions when
extracting arch-specific information, and only return relevant
information when using arch-specific expressions. Additionally
add include/trace/events/kvm.h, arch/*/include/uapi/asm/kvm*,
and arch/powerpc/kernel/kvm* to appropriate sections. The arch-
specific expressions are now:

 /KVM/                                        -- All KVM
 /\(KVM\)|\(KVM\/x86\)/                       -- X86
 /\(KVM\)|\(KVM\/x86\)|\(KVM\/amd\)/          -- X86 plus AMD
 /\(KVM\)|\(KVM\/arm\)/                       -- ARM
 /\(KVM\)|\(KVM\/arm\)|\(KVM\/arm64\)/        -- ARM plus ARM64
 /\(KVM\)|\(KVM\/powerpc\)/                   -- POWERPC
 /\(KVM\)|\(KVM\/s390\)/                      -- S390
 /\(KVM\)|\(KVM\/mips\)/                      -- MIPS

Signed-off-by: Andrew Jones <drjones@redhat.com>
Acked-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agokvm: nVMX: Add support for fast unprotection of nested guest page tables
Paolo Bonzini [Mon, 28 Nov 2016 13:39:58 +0000 (14:39 +0100)]
kvm: nVMX: Add support for fast unprotection of nested guest page tables

This is the same as commit 147277540bbc ("kvm: svm: Add support for
additional SVM NPF error codes", 2016-11-23), but for Intel processors.
In this case, the exit qualification field's bit 8 says whether the
EPT violation occurred while translating the guest's final physical
address or rather while translating the guest page tables.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: SVM: Limit PFERR_NESTED_GUEST_PAGE error_code check to L1 guest
Brijesh Singh [Mon, 7 Aug 2017 19:11:30 +0000 (14:11 -0500)]
KVM: SVM: Limit PFERR_NESTED_GUEST_PAGE error_code check to L1 guest

Commit 147277540bbc ("kvm: svm: Add support for additional SVM NPF error
codes", 2016-11-23) added a new error code to aid nested page fault
handling.  The commit unprotects (kvm_mmu_unprotect_page) the page when
we get a NPF due to guest page table walk where the page was marked RO.

However, if an L0->L2 shadow nested page table can also be marked read-only
when a page is read only in L1's nested page table.  If such a page
is accessed by L2 while walking page tables it can cause a nested
page fault (page table walks are write accesses).  However, after
kvm_mmu_unprotect_page we may get another page fault, and again in an
endless stream.

To cover this use case, we qualify the new error_code check with
vcpu->arch.mmu_direct_map so that the error_code check would run on L1
guest, and not the L2 guest.  This avoids hitting the above scenario.

Fixes: 147277540bbc54119172481c8ef6d930cc9fbfc2
Cc: stable@vger.kernel.org
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: Thomas Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: X86: Fix residual mmio emulation request to userspace
Wanpeng Li [Thu, 10 Aug 2017 05:33:12 +0000 (22:33 -0700)]
KVM: X86: Fix residual mmio emulation request to userspace

Reported by syzkaller:

The kvm-intel.unrestricted_guest=0

   WARNING: CPU: 5 PID: 1014 at /home/kernel/data/kvm/arch/x86/kvm//x86.c:7227 kvm_arch_vcpu_ioctl_run+0x38b/0x1be0 [kvm]
   CPU: 5 PID: 1014 Comm: warn_test Tainted: G        W  OE   4.13.0-rc3+ #8
   RIP: 0010:kvm_arch_vcpu_ioctl_run+0x38b/0x1be0 [kvm]
   Call Trace:
    ? put_pid+0x3a/0x50
    ? rcu_read_lock_sched_held+0x79/0x80
    ? kmem_cache_free+0x2f2/0x350
    kvm_vcpu_ioctl+0x340/0x700 [kvm]
    ? kvm_vcpu_ioctl+0x340/0x700 [kvm]
    ? __fget+0xfc/0x210
    do_vfs_ioctl+0xa4/0x6a0
    ? __fget+0x11d/0x210
    SyS_ioctl+0x79/0x90
    entry_SYSCALL_64_fastpath+0x23/0xc2
    ? __this_cpu_preempt_check+0x13/0x20

The syszkaller folks reported a residual mmio emulation request to userspace
due to vm86 fails to emulate inject real mode interrupt(fails to read CS) and
incurs a triple fault. The vCPU returns to userspace with vcpu->mmio_needed == true
and KVM_EXIT_SHUTDOWN exit reason. However, the syszkaller testcase constructs
several threads to launch the same vCPU, the thread which lauch this vCPU after
the thread whichs get the vcpu->mmio_needed == true and KVM_EXIT_SHUTDOWN will
trigger the warning.

   #define _GNU_SOURCE
   #include <pthread.h>
   #include <stdio.h>
   #include <stdlib.h>
   #include <string.h>
   #include <sys/wait.h>
   #include <sys/types.h>
   #include <sys/stat.h>
   #include <sys/mman.h>
   #include <fcntl.h>
   #include <unistd.h>
   #include <linux/kvm.h>
   #include <stdio.h>

   int kvmcpu;
   struct kvm_run *run;

   void* thr(void* arg)
   {
     int res;
     res = ioctl(kvmcpu, KVM_RUN, 0);
     printf("ret1=%d exit_reason=%d suberror=%d\n",
         res, run->exit_reason, run->internal.suberror);
     return 0;
   }

   void test()
   {
     int i, kvm, kvmvm;
     pthread_t th[4];

     kvm = open("/dev/kvm", O_RDWR);
     kvmvm = ioctl(kvm, KVM_CREATE_VM, 0);
     kvmcpu = ioctl(kvmvm, KVM_CREATE_VCPU, 0);
     run = (struct kvm_run*)mmap(0, 4096, PROT_READ|PROT_WRITE, MAP_SHARED, kvmcpu, 0);
     srand(getpid());
     for (i = 0; i < 4; i++) {
       pthread_create(&th[i], 0, thr, 0);
       usleep(rand() % 10000);
     }
     for (i = 0; i < 4; i++)
       pthread_join(th[i], 0);
   }

   int main()
   {
     for (;;) {
       int pid = fork();
       if (pid < 0)
         exit(1);
       if (pid == 0) {
         test();
         exit(0);
       }
       int status;
       while (waitpid(pid, &status, __WALL) != pid) {}
     }
     return 0;
   }

This patch fixes it by resetting the vcpu->mmio_needed once we receive
the triple fault to avoid the residue.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: arm: implements the kvm_arch_vcpu_in_kernel()
Longpeng(Mike) [Tue, 8 Aug 2017 04:05:35 +0000 (12:05 +0800)]
KVM: arm: implements the kvm_arch_vcpu_in_kernel()

This implements the kvm_arch_vcpu_in_kernel() for ARM, and adjusts
the calls to kvm_vcpu_on_spin().

Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: s390: implements the kvm_arch_vcpu_in_kernel()
Longpeng(Mike) [Tue, 8 Aug 2017 04:05:34 +0000 (12:05 +0800)]
KVM: s390: implements the kvm_arch_vcpu_in_kernel()

This implements kvm_arch_vcpu_in_kernel() for s390.  DIAG is a privileged
operation, so it cannot be called from problem state (user mode).

Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: X86: implement the logic for spinlock optimization
Longpeng(Mike) [Tue, 8 Aug 2017 04:05:33 +0000 (12:05 +0800)]
KVM: X86: implement the logic for spinlock optimization

get_cpl requires vcpu_load, so we must cache the result (whether the
vcpu was preempted when its cpl=0) in kvm_vcpu_arch.

Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: add spinlock optimization framework
Longpeng(Mike) [Tue, 8 Aug 2017 04:05:32 +0000 (12:05 +0800)]
KVM: add spinlock optimization framework

If a vcpu exits due to request a user mode spinlock, then
the spinlock-holder may be preempted in user mode or kernel mode.
(Note that not all architectures trap spin loops in user mode,
only AMD x86 and ARM/ARM64 currently do).

But if a vcpu exits in kernel mode, then the holder must be
preempted in kernel mode, so we should choose a vcpu in kernel mode
as a more likely candidate for the lock holder.

This introduces kvm_arch_vcpu_in_kernel() to decide whether the
vcpu is in kernel-mode when it's preempted.  kvm_vcpu_on_spin's
new argument says the same of the spinning VCPU.

Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: x86: use general helpers for some cpuid manipulation
Radim Krčmář [Fri, 4 Aug 2017 22:12:50 +0000 (00:12 +0200)]
KVM: x86: use general helpers for some cpuid manipulation

Add guest_cpuid_clear() and use it instead of kvm_find_cpuid_entry().
Also replace some uses of kvm_find_cpuid_entry() with guest_cpuid_has().

Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: x86: generalize guest_cpuid_has_ helpers
Radim Krčmář [Fri, 4 Aug 2017 22:12:49 +0000 (00:12 +0200)]
KVM: x86: generalize guest_cpuid_has_ helpers

This patch turns guest_cpuid_has_XYZ(cpuid) into guest_cpuid_has(cpuid,
X86_FEATURE_XYZ), which gets rid of many very similar helpers.

When seeing a X86_FEATURE_*, we can know which cpuid it belongs to, but
this information isn't in common code, so we recreate it for KVM.

Add some BUILD_BUG_ONs to make sure that it runs nicely.

Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: x86: X86_FEATURE_NRIPS is not scattered anymore
Radim Krčmář [Fri, 4 Aug 2017 22:12:48 +0000 (00:12 +0200)]
KVM: x86: X86_FEATURE_NRIPS is not scattered anymore

bit(X86_FEATURE_NRIPS) is 3 since 2ccd71f1b278 ("x86/cpufeature: Move
some of the scattered feature bits to x86_capability"), so we can
simplify the code.

Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: nVMX: Emulate EPTP switching for the L1 hypervisor
Bandan Das [Thu, 3 Aug 2017 19:54:43 +0000 (15:54 -0400)]
KVM: nVMX: Emulate EPTP switching for the L1 hypervisor

When L2 uses vmfunc, L0 utilizes the associated vmexit to
emulate a switching of the ept pointer by reloading the
guest MMU.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Bandan Das <bsd@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: nVMX: Enable VMFUNC for the L1 hypervisor
Bandan Das [Thu, 3 Aug 2017 19:54:42 +0000 (15:54 -0400)]
KVM: nVMX: Enable VMFUNC for the L1 hypervisor

Expose VMFUNC in MSRs and VMCS fields. No actual VMFUNCs are enabled.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Bandan Das <bsd@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: vmx: Enable VMFUNCs
Bandan Das [Thu, 3 Aug 2017 19:54:41 +0000 (15:54 -0400)]
KVM: vmx: Enable VMFUNCs

Enable VMFUNC in the secondary execution controls.  This simplifies the
changes necessary to expose it to nested hypervisors.  VMFUNCs still
cause #UD when invoked.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Bandan Das <bsd@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: nVMX: get rid of nested_release_page*
David Hildenbrand [Thu, 3 Aug 2017 16:11:05 +0000 (18:11 +0200)]
KVM: nVMX: get rid of nested_release_page*

Let's also just use the underlying functions directly here.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
[Rebased on top of 9f744c597460 ("KVM: nVMX: do not pin the VMCS12")]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
7 years agoKVM: nVMX: get rid of nested_get_page()
David Hildenbrand [Thu, 3 Aug 2017 16:11:04 +0000 (18:11 +0200)]
KVM: nVMX: get rid of nested_get_page()

nested_get_page() just sounds confusing. All we want is a page from G1.
This is even unrelated to nested.

Let's introduce kvm_vcpu_gpa_to_page() so we don't get too lengthy
lines.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
[Squash pasto fix from Wanpeng Li. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: nVMX: INVPCID support
Paolo Bonzini [Thu, 27 Jul 2017 11:22:13 +0000 (13:22 +0200)]
KVM: nVMX: INVPCID support

Expose the "Enable INVPCID" secondary execution control to the guest
and properly reflect the exit reason.

In addition, before this patch the guest was always running with
INVPCID enabled, causing pcid.flat's "Test on INVPCID when disabled"
test to fail.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: hyperv: support HV_X64_MSR_TSC_FREQUENCY and HV_X64_MSR_APIC_FREQUENCY
Ladi Prosek [Wed, 26 Jul 2017 11:32:59 +0000 (13:32 +0200)]
KVM: hyperv: support HV_X64_MSR_TSC_FREQUENCY and HV_X64_MSR_APIC_FREQUENCY

It has been experimentally confirmed that supporting these two MSRs is one
of the necessary conditions for nested Hyper-V to use the TSC page. Modern
Windows guests are noticeably slower when they fall back to reading
timestamps from the HV_X64_MSR_TIME_REF_COUNT MSR instead of using the TSC
page.

The newly supported MSRs are advertised with the AccessFrequencyRegs
partition privilege flag and CPUID.40000003H:EDX[8] "Support for
determining timer frequencies is available" (both outside of the scope of
this KVM patch).

Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoLinux 4.13-rc4
Linus Torvalds [Mon, 7 Aug 2017 01:44:49 +0000 (18:44 -0700)]
Linux 4.13-rc4

7 years agoMerge tag 'platform-drivers-x86-v4.13-4' of git://git.infradead.org/linux-platform...
Linus Torvalds [Sun, 6 Aug 2017 23:11:34 +0000 (16:11 -0700)]
Merge tag 'platform-drivers-x86-v4.13-4' of git://git.infradead.org/linux-platform-drivers-x86

Pull x86 platform driver fix from Darren Hart:
 "Fix loop preventing some platforms from waking up via the power button
  in s2idle:

   - intel-vbtn: match power button on press rather than release"

* tag 'platform-drivers-x86-v4.13-4' of git://git.infradead.org/linux-platform-drivers-x86:
  platform/x86: intel-vbtn: match power button on press rather than release

7 years agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 6 Aug 2017 19:31:17 +0000 (12:31 -0700)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "A large number of ext4 bug fixes and cleanups for v4.13"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix copy paste error in ext4_swap_extents()
  ext4: fix overflow caused by missing cast in ext4_resize_fs()
  ext4, project: expand inode extra size if possible
  ext4: cleanup ext4_expand_extra_isize_ea()
  ext4: restructure ext4_expand_extra_isize
  ext4: fix forgetten xattr lock protection in ext4_expand_extra_isize
  ext4: make xattr inode reads faster
  ext4: inplace xattr block update fails to deduplicate blocks
  ext4: remove unused mode parameter
  ext4: fix warning about stack corruption
  ext4: fix dir_nlink behaviour
  ext4: silence array overflow warning
  ext4: fix SEEK_HOLE/SEEK_DATA for blocksize < pagesize
  ext4: release discard bio after sending discard commands
  ext4: convert swap_inode_data() over to use swap() on most of the fields
  ext4: error should be cleared if ea_inode isn't added to the cache
  ext4: Don't clear SGID when inheriting ACLs
  ext4: preserve i_mode if __ext4_set_acl() fails
  ext4: remove unused metadata accounting variables
  ext4: correct comment references to ext4_ext_direct_IO()

7 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Sun, 6 Aug 2017 18:52:01 +0000 (11:52 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:
 "This fixes two build issues for ralink platforms, both due to missing
  #includes which used to be included indirectly via other headers"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: ralink: mt7620: Add missing header
  MIPS: ralink: Fix build error due to missing header

7 years agoFix compat_sys_sigpending breakage
Dmitry V. Levin [Sat, 5 Aug 2017 20:00:50 +0000 (23:00 +0300)]
Fix compat_sys_sigpending breakage

The latest change of compat_sys_sigpending in commit 8f13621abced
("sigpending(): move compat to native") has broken it in two ways.

First, it tries to write 4 bytes more than userspace expects:
sizeof(old_sigset_t) == sizeof(long) == 8 instead of
sizeof(compat_old_sigset_t) == sizeof(u32) == 4.

Second, on big endian architectures these bytes are being written in the
wrong order.

This bug was found by strace test suite.

Reported-by: Anatoly Pugachev <matorola@gmail.com>
Inspired-by: Eugene Syromyatnikov <evgsyr@gmail.com>
Fixes: 8f13621abced ("sigpending(): move compat to native")
Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoext4: fix copy paste error in ext4_swap_extents()
Maninder Singh [Sun, 6 Aug 2017 05:33:07 +0000 (01:33 -0400)]
ext4: fix copy paste error in ext4_swap_extents()

This bug was found by a static code checker tool for copy paste
problems.

Signed-off-by: Maninder Singh <maninder1.s@samsung.com>
Signed-off-by: Vaneet Narang <v.narang@samsung.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: fix overflow caused by missing cast in ext4_resize_fs()
Jerry Lee [Sun, 6 Aug 2017 05:18:31 +0000 (01:18 -0400)]
ext4: fix overflow caused by missing cast in ext4_resize_fs()

On a 32-bit platform, the value of n_blcoks_count may be wrong during
the file system is resized to size larger than 2^32 blocks.  This may
caused the superblock being corrupted with zero blocks count.

Fixes: 1c6bd7173d66
Signed-off-by: Jerry Lee <jerrylee@qnap.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org # 3.7+
7 years agoext4, project: expand inode extra size if possible
Miao Xie [Sun, 6 Aug 2017 05:00:49 +0000 (01:00 -0400)]
ext4, project: expand inode extra size if possible

When upgrading from old format, try to set project id
to old file first time, it will return EOVERFLOW, but if
that file is dirtied(touch etc), changing project id will
be allowed, this might be confusing for users, we could
try to expand @i_extra_isize here too.

Reported-by: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Miao Xie <miaoxie@huawei.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: cleanup ext4_expand_extra_isize_ea()
Miao Xie [Sun, 6 Aug 2017 04:55:48 +0000 (00:55 -0400)]
ext4: cleanup ext4_expand_extra_isize_ea()

Clean up some goto statement, make ext4_expand_extra_isize_ea() clearer.

Signed-off-by: Miao Xie <miaoxie@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Wang Shilong <wshilong@ddn.com>
7 years agoext4: restructure ext4_expand_extra_isize
Miao Xie [Sun, 6 Aug 2017 04:40:01 +0000 (00:40 -0400)]
ext4: restructure ext4_expand_extra_isize

Current ext4_expand_extra_isize just tries to expand extra isize, if
someone is holding xattr lock or some check fails, it will give up.
So rename its name to ext4_try_to_expand_extra_isize.

Besides that, we clean up unnecessary check and move some relative checks
into it.

Signed-off-by: Miao Xie <miaoxie@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Wang Shilong <wshilong@ddn.com>
7 years agoext4: fix forgetten xattr lock protection in ext4_expand_extra_isize
Miao Xie [Sun, 6 Aug 2017 04:27:38 +0000 (00:27 -0400)]
ext4: fix forgetten xattr lock protection in ext4_expand_extra_isize

We should avoid the contention between the i_extra_isize update and
the inline data insertion, so move the xattr trylock in front of
i_extra_isize update.

Signed-off-by: Miao Xie <miaoxie@huawei.com>
Reviewed-by: Wang Shilong <wshilong@ddn.com>
7 years agoext4: make xattr inode reads faster
Tahsin Erdogan [Sun, 6 Aug 2017 04:07:01 +0000 (00:07 -0400)]
ext4: make xattr inode reads faster

ext4_xattr_inode_read() currently reads each block sequentially while
waiting for io operation to complete before moving on to the next
block. This prevents request merging in block layer.

Add a ext4_bread_batch() function that starts reads for all blocks
then optionally waits for them to complete. A similar logic is used
in ext4_find_entry(), so update that code to use the new function.

Signed-off-by: Tahsin Erdogan <tahsin@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: inplace xattr block update fails to deduplicate blocks
Tahsin Erdogan [Sun, 6 Aug 2017 02:41:42 +0000 (22:41 -0400)]
ext4: inplace xattr block update fails to deduplicate blocks

When an xattr block has a single reference, block is updated inplace
and it is reinserted to the cache. Later, a cache lookup is performed
to see whether an existing block has the same contents. This cache
lookup will most of the time return the just inserted entry so
deduplication is not achieved.

Running the following test script will produce two xattr blocks which
can be observed in "File ACL: " line of debugfs output:

  mke2fs -b 1024 -I 128 -F -O extent /dev/sdb 1G
  mount /dev/sdb /mnt/sdb

  touch /mnt/sdb/{x,y}

  setfattr -n user.1 -v aaa /mnt/sdb/x
  setfattr -n user.2 -v bbb /mnt/sdb/x

  setfattr -n user.1 -v aaa /mnt/sdb/y
  setfattr -n user.2 -v bbb /mnt/sdb/y

  debugfs -R 'stat x' /dev/sdb | cat
  debugfs -R 'stat y' /dev/sdb | cat

This patch defers the reinsertion to the cache so that we can locate
other blocks with the same contents.

Signed-off-by: Tahsin Erdogan <tahsin@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
7 years agoext4: remove unused mode parameter
Tahsin Erdogan [Sun, 6 Aug 2017 02:15:45 +0000 (22:15 -0400)]
ext4: remove unused mode parameter

ext4_alloc_file_blocks() does not use its mode parameter. Remove it.

Signed-off-by: Tahsin Erdogan <tahsin@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: fix warning about stack corruption
Arnd Bergmann [Sun, 6 Aug 2017 01:57:46 +0000 (21:57 -0400)]
ext4: fix warning about stack corruption

After commit 62d1034f53e3 ("fortify: use WARN instead of BUG for now"),
we get a warning about possible stack overflow from a memcpy that
was not strictly bounded to the size of the local variable:

    inlined from 'ext4_mb_seq_groups_show' at fs/ext4/mballoc.c:2322:2:
include/linux/string.h:309:9: error: '__builtin_memcpy': writing between 161 and 1116 bytes into a region of size 160 overflows the destination [-Werror=stringop-overflow=]

We actually had a bug here that would have been found by the warning,
but it was already fixed last year in commit 30a9d7afe70e ("ext4: fix
stack memory corruption with 64k block size").

This replaces the fixed-length structure on the stack with a variable-length
structure, using the correct upper bound that tells the compiler that
everything is really fine here. I also change the loop count to check
for the same upper bound for consistency, but the existing code is
already correct here.

Note that while clang won't allow certain kinds of variable-length arrays
in structures, this particular instance is fine, as the array is at the
end of the structure, and the size is strictly bounded.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: fix dir_nlink behaviour
Andreas Dilger [Sat, 5 Aug 2017 23:47:34 +0000 (19:47 -0400)]
ext4: fix dir_nlink behaviour

The dir_nlink feature has been enabled by default for new ext4
filesystems since e2fsprogs-1.41 in 2008, and was automatically
enabled by the kernel for older ext4 filesystems since the
dir_nlink feature was added with ext4 in kernel 2.6.28+ when
the subdirectory count exceeded EXT4_LINK_MAX-1.

Automatically adding the file system features such as dir_nlink is
generally frowned upon, since it could cause the file system to not be
mountable on older kernel, thus preventing the administrator from
rolling back to an older kernel if necessary.

In this case, the administrator might also want to disable the feature
because glibc's fts_read() function does not correctly optimize
directory traversal for directories that use st_nlinks field of 1 to
indicate that the number of links in the directory are not tracked by
the file system, and could fail to traverse the full directory
hierarchy.  Fortunately, in the past ten years very few users have
complained about incomplete file system traversal by glibc's
fts_read().

This commit also changes ext4_inc_count() to allow i_nlinks to reach
the full EXT4_LINK_MAX links on the parent directory (including "."
and "..") before changing i_links_count to be 1.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=196405
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: silence array overflow warning
Dan Carpenter [Sat, 5 Aug 2017 23:00:31 +0000 (19:00 -0400)]
ext4: silence array overflow warning

I get a static checker warning:

    fs/ext4/ext4.h:3091 ext4_set_de_type()
    error: buffer overflow 'ext4_type_by_mode' 15 <= 15

It seems unlikely that we would hit this read overflow in real life, but
it's also simple enough to make the array 16 bytes instead of 15.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 years agoext4: fix SEEK_HOLE/SEEK_DATA for blocksize < pagesize
Jan Kara [Sat, 5 Aug 2017 21:43:24 +0000 (17:43 -0400)]
ext4: fix SEEK_HOLE/SEEK_DATA for blocksize < pagesize

ext4_find_unwritten_pgoff() does not properly handle a situation when
starting index is in the middle of a page and blocksize < pagesize. The
following command shows the bug on filesystem with 1k blocksize:

  xfs_io -f -c "falloc 0 4k" \
            -c "pwrite 1k 1k" \
            -c "pwrite 3k 1k" \
            -c "seek -a -r 0" foo

In this example, neither lseek(fd, 1024, SEEK_HOLE) nor lseek(fd, 2048,
SEEK_DATA) will return the correct result.

Fix the problem by neglecting buffers in a page before starting offset.

Reported-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Jan Kara <jack@suse.cz>
CC: stable@vger.kernel.org # 3.8+
7 years agoplatform/x86: intel-vbtn: match power button on press rather than release
Mario Limonciello [Fri, 4 Aug 2017 17:00:06 +0000 (12:00 -0500)]
platform/x86: intel-vbtn: match power button on press rather than release

This fixes a problem where the system gets stuck in a loop
unable to wakeup via power button in s2idle.

The problem happens because:
 - press power button:
   - system emits 0xc0 (power press), event ignored
   - system emits 0xc1 (power release), event processed,
     emited as KEY_POWER
   - set wakeup_mode to true
   - system goes to s2idle
 - press power button
   - system emits 0xc0 (power press), wakeup_mode is true,
     system wakes
   - system emits 0xc1 (power release), event processed,
     emited as KEY_POWER
   - system goes to s2idle again

To avoid this situation, process the presses (which matches what
intel-hid does too).

Verified on an Dell XPS 9365

Signed-off-by: Mario Limonciello <mario.limonciello@dell.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Darren Hart (VMware) <dvhart@infradead.org>
7 years agoMerge tag 'media/v4.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Sat, 5 Aug 2017 21:09:26 +0000 (14:09 -0700)]
Merge tag 'media/v4.13-2' of git://git./linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab:
 "This series is larger than I would like to submit for -rc4. My
  original intent were to sent it to either -rc2 or -rc3. Unfortunately,
  due to my vacations, I got a lot of pending stuff after my return, and
  had to do some biz trips, with prevented me to send this earlier.

  Several fixes:

   - some fixes at atomisp staging driver

   - several gcc 7 warning fixes

   - cleanup media SVG files, in order to fix PDF build on some distros

   - fix random Kconfig build of venus driver

   - some fixes for the venus driver

   - some changes from semaphone to mutex in ngene's driver

   - some locking fixes at dib0700 driver

   - several fixes on ngene's driver and frontends to make it properly
     support some new boards added on Kernel 4.13

   - some fixes to CEC drivers

   - omap_vout: vrfb: convert to dmaengine

   - docs-rst: document EBUSY for VIDIOC_S_FMT

  Please notice that the big diffstat changes here are at the SVG files.

  Visually, the images look the same, but the file size is now a lot
  smaller than before, and they don't use some XML tags that would cause
  them to be badly parsed by some ImageMagick versions, or to require a
  lot of memory by TeTex, with would break PDF output on some
  distributions"

* tag 'media/v4.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media: (68 commits)
  media: atomisp2: array underflow in imx_enum_frame_size()
  media: atomisp2: array underflow in ap1302_enum_frame_size()
  media: atomisp2: Array underflow in atomisp_enum_input()
  media: platform: davinci: drop VPFE_CMD_S_CCDC_RAW_PARAMS
  media: platform: davinci: return -EINVAL for VPFE_CMD_S_CCDC_RAW_PARAMS ioctl
  media: venus: don't abuse dma_alloc for non-DMA allocations
  media: venus: hfi: fix error handling in hfi_sys_init_done()
  media: venus: fix compile-test build on non-qcom ARM platform
  media: venus: mark PM functions as __maybe_unused
  media: cec-notifier: small improvements
  media: pulse8-cec: persistent_config should be off by default
  media: cec: cec_transmit_attempt_done: ignore CEC_TX_STATUS_MAX_RETRIES
  media: staging: atomisp: array underflow in ioctl
  media: lirc: LIRC_GET_REC_RESOLUTION should return microseconds
  media: svg: avoid too long lines
  media: svg files: simplify files
  media: selection.svg: simplify the SVG file
  media: vimc: set id_table for platform drivers
  media: staging: atomisp: disable warnings with cc-disable-warning
  media: davinci: variable 'common' set but not used
  ...

7 years agoext4: release discard bio after sending discard commands
Daeho Jeong [Sat, 5 Aug 2017 17:11:57 +0000 (13:11 -0400)]
ext4: release discard bio after sending discard commands

We've changed the discard command handling into parallel manner.
But, in this change, I forgot decreasing the usage count of the bio
which was used to send discard request. I'm sorry about that.

Fixes: a015434480dc ("ext4: send parallel discards on commit completions")
Signed-off-by: Daeho Jeong <daeho.jeong@samsung.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
7 years agoMerge tag 'gpio-v4.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Sat, 5 Aug 2017 13:55:13 +0000 (06:55 -0700)]
Merge tag 'gpio-v4.13-2' of git://git./linux/kernel/git/linusw/linux-gpio

Pull GPIO fixes from Linus Walleij:

 - LP87565: set the proper output level for direction_output.

 - stm32: fix the kernel build by selecting the hierarchical irqdomain
   symbol properly - this happens to be done in the pin control
   framework but whatever, it had dependencies to GPIO so we need to
   apply it here.

 - Select the hierarchical IRQ domain also for Xgene.

 - Fix wakeups to work on MXC.

 - Fix up the device tree binding on Exar that went astray, also add the
   right bindings.

 - Fix the unwanted events for edges from the library.

 - Fix the unbalanced chanined IRQ on the Tegra.

* tag 'gpio-v4.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
  gpio: tegra: fix unbalanced chained_irq_enter/exit
  gpiolib: skip unwanted events, don't convert them to opposite edge
  gpio: exar: Use correct property prefix and document bindings
  gpio: gpio-mxc: Fix: higher 16 GPIOs usable as wake source
  gpio: xgene-sb: select IRQ_DOMAIN_HIERARCHY
  pinctrl: stm32: select IRQ_DOMAIN_HIERARCHY instead of depends on
  gpio: lp87565: Set proper output level and direction for direction_output
  MAINTAINERS: Add entry for Whiskey Cove PMIC GPIO driver

7 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 4 Aug 2017 23:45:29 +0000 (16:45 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "A handful of critical fixes for changes introduce this merge window.

   - The TI sci_clk_get() API was pretty broken and nobody noticed.

   - There were some CPUfreq crashes on C.H.I.P devices because we
     failed to propagate rates up the clk tree.

   - Also, the Intel Atom PMC clk driver needs to mark a clk critical if
     the firmware has it enabled already so that audio doesn't get
     killed on Baytrail.

   - Gemini devices have a dead serial console because the reset control
     usage in the serial driver assume one method of reset that gemini
     doesn't support (this will be fixed in the next version in the
     reset framework so this is the small fix for -rc series).

   - Finally we have two rate calculation fixes, one for Exynos and one
     for Meson SoCs, that fix rate inconsistencies"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: keystone: sci-clk: Fix sci_clk_get
  clk: meson: mpll: fix mpll0 fractional part ignored
  clk: samsung: exynos5420: The EPLL rate table corrections
  clk: sunxi-ng: sun5i: Add clk_set_rate_parent to the CPU clock
  clk: x86: Do not gate clocks enabled by the firmware
  clk: gemini: Fix reset regression

7 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 4 Aug 2017 22:18:27 +0000 (15:18 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Radim Krčmář:
 "ARM:

   - Yet another race with VM destruction plugged

   - A set of small vgic fixes

  x86:

   - Preserve pending INIT

   - RCU fixes in paravirtual async pf, VM teardown, and VMXOFF
     emulation

   - nVMX interrupt injection and dirty tracking fixes

   - initialize to make UBSAN happy"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: arm/arm64: vgic: Use READ_ONCE fo cmpxchg
  KVM: nVMX: Fix interrupt window request with "Acknowledge interrupt on exit"
  KVM: nVMX: mark vmcs12 pages dirty on L2 exit
  kvm: nVMX: don't flush VMCS12 during VMXOFF or VCPU teardown
  KVM: nVMX: do not pin the VMCS12
  KVM: avoid using rcu_dereference_protected
  KVM: X86: init irq->level in kvm_pv_kick_cpu_op
  KVM: X86: Fix loss of pending INIT due to race
  KVM: async_pf: make rcu irq exit if not triggered from idle task
  KVM: nVMX: fixes to nested virt interrupt injection
  KVM: nVMX: do not fill vm_exit_intr_error_code in prepare_vmcs12
  KVM: arm/arm64: Handle hva aging while destroying the vm
  KVM: arm/arm64: PMU: Fix overflow interrupt injection
  KVM: arm/arm64: Fix bug in advertising KVM_CAP_MSI_DEVID capability

7 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 4 Aug 2017 22:16:09 +0000 (15:16 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fix from Thomas Gleixner:
 "The recent irq core changes unearthed API abuse in the HPET code,
  which manifested itself in a suspend/resume regression.

  The fix replaces the cruft with the proper function calls and cures
  the regression"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/hpet: Cure interface abuse in the resume path