deliverable/linux.git
11 years agoARM: KVM: Bugfix: vgic_bytemap_get_reg per cpu regs
Christoffer Dall [Thu, 29 Aug 2013 10:08:24 +0000 (11:08 +0100)] 
ARM: KVM: Bugfix: vgic_bytemap_get_reg per cpu regs

For bytemaps each IRQ field is 1 byte wide, so we pack 4 irq fields in
one word and since there are 32 private (per cpu) irqs, we have 8
private u32 fields on the vgic_bytemap struct.  We shift the offset from
the base of the register group right by 2, giving us the word index
instead of the field index.  But then there are 8 private words, not 4,
which is also why we subtract 8 words from the offset of the shared
words.

Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoARM: KVM: vgic: fix GICD_ICFGRn access
Marc Zyngier [Thu, 29 Aug 2013 10:08:23 +0000 (11:08 +0100)] 
ARM: KVM: vgic: fix GICD_ICFGRn access

All the code in handle_mmio_cfg_reg() assumes the offset has
been shifted right to accomodate for the 2:1 bit compression,
but this is only done when getting the register address.

Shift the offset early so the code works mostly unchanged.

Reported-by: Zhaobo (Bob, ERC) <zhaobo@huawei.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoARM: KVM: vgic: simplify vgic_get_target_reg
Marc Zyngier [Thu, 29 Aug 2013 10:08:22 +0000 (11:08 +0100)] 
ARM: KVM: vgic: simplify vgic_get_target_reg

vgic_get_target_reg is quite complicated, for no good reason.
Actually, it is fairly easy to write it in a much more efficient
way by using the target CPU array instead of the bitmap.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoMerge branch 'kvm-ppc-next' of git://github.com/agraf/linux-2.6 into queue
Gleb Natapov [Fri, 30 Aug 2013 12:33:11 +0000 (15:33 +0300)] 
Merge branch 'kvm-ppc-next' of git://github.com/agraf/linux-2.6 into queue

* 'kvm-ppc-next' of git://github.com/agraf/linux-2.6:
  KVM: PPC: Book3S PR: Rework kvmppc_mmu_book3s_64_xlate()
  KVM: PPC: Book3S PR: Make instruction fetch fallback work for system calls
  KVM: PPC: Book3S PR: Don't corrupt guest state when kernel uses VMX
  KVM: PPC: Book3S: Fix compile error in XICS emulation
  KVM: PPC: Book3S PR: return appropriate error when allocation fails
  arch: powerpc: kvm: add signed type cast for comparation
  powerpc/kvm: Copy the pvr value after memset
  KVM: PPC: Book3S PR: Load up SPRG3 register with guest value on guest entry
  kvm/ppc/booke: Don't call kvm_guest_enter twice
  kvm/ppc: Call trace_hardirqs_on before entry
  KVM: PPC: Book3S HV: Allow negative offsets to real-mode hcall handlers
  KVM: PPC: Book3S HV: Correct tlbie usage
  powerpc/kvm: Use 256K chunk to track both RMA and hash page table allocation.
  powerpc/kvm: Contiguous memory allocator based RMA allocation
  powerpc/kvm: Contiguous memory allocator based hash page table allocation
  KVM: PPC: Book3S: Ignore DABR register
  mm/cma: Move dma contiguous changes into a seperate config

11 years agoKVM: MMU: remove unused parameter
Xiao Guangrong [Tue, 30 Jul 2013 13:01:59 +0000 (21:01 +0800)] 
KVM: MMU: remove unused parameter

vcpu in page_fault_can_be_fast() is not used so remove it

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoMerge remote-tracking branch 'origin/next' into kvm-ppc-next
Alexander Graf [Wed, 28 Aug 2013 22:41:59 +0000 (00:41 +0200)] 
Merge remote-tracking branch 'origin/next' into kvm-ppc-next

Conflicts:
mm/Kconfig

CMA DMA split and ZSWAP introduction were conflicting, fix up manually.

11 years agoKVM: PPC: Book3S PR: Rework kvmppc_mmu_book3s_64_xlate()
Paul Mackerras [Tue, 6 Aug 2013 04:18:00 +0000 (14:18 +1000)] 
KVM: PPC: Book3S PR: Rework kvmppc_mmu_book3s_64_xlate()

This reworks kvmppc_mmu_book3s_64_xlate() to make it check the large
page bit in the hashed page table entries (HPTEs) it looks at, and
to simplify and streamline the code.  The checking of the first dword
of each HPTE is now done with a single mask and compare operation,
and all the code dealing with the matching HPTE, if we find one,
is consolidated in one place in the main line of the function flow.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: PPC: Book3S PR: Make instruction fetch fallback work for system calls
Paul Mackerras [Tue, 6 Aug 2013 04:15:19 +0000 (14:15 +1000)] 
KVM: PPC: Book3S PR: Make instruction fetch fallback work for system calls

It turns out that if we exit the guest due to a hcall instruction (sc 1),
and the loading of the instruction in the guest exit path fails for any
reason, the call to kvmppc_ld() in kvmppc_get_last_inst() fetches the
instruction after the hcall instruction rather than the hcall itself.
This in turn means that the instruction doesn't get recognized as an
hcall in kvmppc_handle_exit_pr() but gets passed to the guest kernel
as a sc instruction.  That usually results in the guest kernel getting
a return code of 38 (ENOSYS) from an hcall, which often triggers a
BUG_ON() or other failure.

This fixes the problem by adding a new variant of kvmppc_get_last_inst()
called kvmppc_get_last_sc(), which fetches the instruction if necessary
from pc - 4 rather than pc.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: PPC: Book3S PR: Don't corrupt guest state when kernel uses VMX
Paul Mackerras [Tue, 6 Aug 2013 04:14:33 +0000 (14:14 +1000)] 
KVM: PPC: Book3S PR: Don't corrupt guest state when kernel uses VMX

Currently the code assumes that once we load up guest FP/VSX or VMX
state into the CPU, it stays valid in the CPU registers until we
explicitly flush it to the thread_struct.  However, on POWER7,
copy_page() and memcpy() can use VMX.  These functions do flush the
VMX state to the thread_struct before using VMX instructions, but if
this happens while we have guest state in the VMX registers, and we
then re-enter the guest, we don't reload the VMX state from the
thread_struct, leading to guest corruption.  This has been observed
to cause guest processes to segfault.

To fix this, we check before re-entering the guest that all of the
bits corresponding to facilities owned by the guest, as expressed
in vcpu->arch.guest_owned_ext, are set in current->thread.regs->msr.
Any bits that have been cleared correspond to facilities that have
been used by kernel code and thus flushed to the thread_struct, so
for them we reload the state from the thread_struct.

We also need to check current->thread.regs->msr before calling
giveup_fpu() or giveup_altivec(), since if the relevant bit is
clear, the state has already been flushed to the thread_struct and
to flush it again would corrupt it.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: x86: update masterclock when kvmclock_offset is calculated (v2)
Marcelo Tosatti [Wed, 28 Aug 2013 02:55:29 +0000 (23:55 -0300)] 
KVM: x86: update masterclock when kvmclock_offset is calculated (v2)

The offset to add to the hosts monotonic time, kvmclock_offset, is
calculated against the monotonic time at KVM_SET_CLOCK ioctl time.

Request a master clock update at this time, to reduce a potentially
unbounded difference between the values of the masterclock and
the clock value used to calculate kvmclock_offset.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: PPC: Book3S: Fix compile error in XICS emulation
Paul Mackerras [Tue, 6 Aug 2013 04:13:44 +0000 (14:13 +1000)] 
KVM: PPC: Book3S: Fix compile error in XICS emulation

Commit 8e44ddc3f3 ("powerpc/kvm/book3s: Add support for H_IPOLL and
H_XIRR_X in XICS emulation") added a call to get_tb() but didn't
include the header that defines it, and on some configs this means
book3s_xics.c fails to compile:

arch/powerpc/kvm/book3s_xics.c: In function ‘kvmppc_xics_hcall’:
arch/powerpc/kvm/book3s_xics.c:812:3: error: implicit declaration of function ‘get_tb’ [-Werror=implicit-function-declaration]

Cc: stable@vger.kernel.org [v3.10, v3.11]
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: PPC: Book3S PR: return appropriate error when allocation fails
Thadeu Lima de Souza Cascardo [Wed, 17 Jul 2013 15:10:29 +0000 (12:10 -0300)] 
KVM: PPC: Book3S PR: return appropriate error when allocation fails

err was overwritten by a previous function call, and checked to be 0. If
the following page allocation fails, 0 is going to be returned instead
of -ENOMEM.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoarch: powerpc: kvm: add signed type cast for comparation
Chen Gang [Mon, 22 Jul 2013 06:32:35 +0000 (14:32 +0800)] 
arch: powerpc: kvm: add signed type cast for comparation

'rmls' is 'unsigned long', lpcr_rmls() will return negative number when
failure occurs, so it need a type cast for comparing.

'lpid' is 'unsigned long', kvmppc_alloc_lpid() return negative number
when failure occurs, so it need a type cast for comparing.

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: x86: add comments where MMIO does not return to the emulator
Paolo Bonzini [Tue, 27 Aug 2013 13:41:43 +0000 (15:41 +0200)] 
KVM: x86: add comments where MMIO does not return to the emulator

Support for single-step in the emulator (new in 3.12) does not work for
MMIO or PIO writes, because they are completed without returning to
the emulator.  This is not worse than what we had in 3.11; still, add
comments so that the issue is not forgotten.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: vmx: count exits to userspace during invalid guest emulation
Paolo Bonzini [Tue, 27 Aug 2013 13:41:42 +0000 (15:41 +0200)] 
KVM: vmx: count exits to userspace during invalid guest emulation

These will happen due to MMIO.

Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: rename __kvm_io_bus_sort_cmp to kvm_io_bus_cmp
Paolo Bonzini [Tue, 27 Aug 2013 13:41:41 +0000 (15:41 +0200)] 
KVM: rename __kvm_io_bus_sort_cmp to kvm_io_bus_cmp

This is the type-safe comparison function, so the double-underscore is
not related.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agokvm: optimize away THP checks in kvm_is_mmio_pfn()
Andrea Arcangeli [Thu, 25 Jul 2013 01:04:38 +0000 (03:04 +0200)] 
kvm: optimize away THP checks in kvm_is_mmio_pfn()

The checks on PG_reserved in the page structure on head and tail pages
aren't necessary because split_huge_page wouldn't transfer the
PG_reserved bit from head to tail anyway.

This was a forward-thinking check done in the case PageReserved was
set by a driver-owned page mapped in userland with something like
remap_pfn_range in a VM_PFNMAP region, but using hugepmds (not
possible right now). It was meant to be very safe, but it's overkill
as it's unlikely split_huge_page could ever run without the driver
noticing and tearing down the hugepage itself.

And if a driver in the future will really want to map a reserved
hugepage in userland using an huge pmd it should simply take care of
marking all subpages reserved too to keep KVM safe. This of course
would require such a hypothetical driver to tear down the huge pmd
itself and splitting the hugepage itself, instead of relaying on
split_huge_page, but that sounds very reasonable, especially
considering split_huge_page wouldn't currently transfer the reserved
bit anyway.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: PPC: reserve a capability number for multitce support
Alexey Kardashevskiy [Thu, 1 Aug 2013 04:44:24 +0000 (14:44 +1000)] 
KVM: PPC: reserve a capability number for multitce support

This is to reserve a capablity number for upcoming support
of H_PUT_TCE_INDIRECT and H_STUFF_TCE pseries hypercalls
which support mulptiple DMA map/unmap operations per one call.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoppc: kvm: use anon_inode_getfd() with O_CLOEXEC flag
Yann Droneaud [Sat, 24 Aug 2013 20:14:08 +0000 (22:14 +0200)] 
ppc: kvm: use anon_inode_getfd() with O_CLOEXEC flag

KVM uses anon_inode_get() to allocate file descriptors as part
of some of its ioctls. But those ioctls are lacking a flag argument
allowing userspace to choose options for the newly opened file descriptor.

In such case it's advised to use O_CLOEXEC by default so that
userspace is allowed to choose, without race, if the file descriptor
is going to be inherited across exec().

This patch set O_CLOEXEC flag on all file descriptors created
with anon_inode_getfd() to not leak file descriptors across exec().

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1377372576.git.ydroneaud@opteya.com
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agokvm: use anon_inode_getfd() with O_CLOEXEC flag
Yann Droneaud [Sat, 24 Aug 2013 20:14:07 +0000 (22:14 +0200)] 
kvm: use anon_inode_getfd() with O_CLOEXEC flag

KVM uses anon_inode_get() to allocate file descriptors as part
of some of its ioctls. But those ioctls are lacking a flag argument
allowing userspace to choose options for the newly opened file descriptor.

In such case it's advised to use O_CLOEXEC by default so that
userspace is allowed to choose, without race, if the file descriptor
is going to be inherited across exec().

This patch set O_CLOEXEC flag on all file descriptors created
with anon_inode_getfd() to not leak file descriptors across exec().

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1377372576.git.ydroneaud@opteya.com
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoDocumentation/kvm : Add documentation on Hypercalls and features used for PV spinlock
Raghavendra K T [Mon, 26 Aug 2013 08:48:36 +0000 (14:18 +0530)] 
Documentation/kvm : Add documentation on Hypercalls and features used for PV spinlock

KVM_HC_KICK_CPU  hypercall added to wakeup halted vcpu in paravirtual spinlock
enabled guest.

KVM_FEATURE_PV_UNHALT enables guest to check whether pv spinlock can be enabled
in guest.

Thanks Vatsa for rewriting KVM_HC_KICK_CPU
Cc: Rob Landley <rob@landley.net>
Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agokvm hypervisor: Simplify kvm_for_each_vcpu with kvm_irq_delivery_to_apic
Raghavendra K T [Mon, 26 Aug 2013 08:48:35 +0000 (14:18 +0530)] 
kvm hypervisor: Simplify kvm_for_each_vcpu with kvm_irq_delivery_to_apic

Note that we are using APIC_DM_REMRD which has reserved usage.
In future if APIC_DM_REMRD usage is standardized, then we should
find some other way or go back to old method.

Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agokvm hypervisor : Add a hypercall to KVM hypervisor to support pv-ticketlocks
Srivatsa Vaddagiri [Mon, 26 Aug 2013 08:48:34 +0000 (14:18 +0530)] 
kvm hypervisor : Add a hypercall to KVM hypervisor to support pv-ticketlocks

kvm_hc_kick_cpu allows the calling vcpu to kick another vcpu out of halt state.
the presence of these hypercalls is indicated to guest via
kvm_feature_pv_unhalt.

Fold pv_unhalt flag into GET_MP_STATE ioctl to aid migration
During migration, any vcpu that got kicked but did not become runnable
(still in halted state) should be runnable after migration.

Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Suzuki Poulose <suzuki@in.ibm.com>
[Raghu: Apic related changes, folding pvunhalted into vcpu_runnable
 Added flags for future use (suggested by Gleb)]
[ Raghu: fold pv_unhalt flag as suggested by Eric Northup]
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agokvm uapi: Add KICK_CPU and PV_UNHALT definition to uapi
Raghavendra K T [Mon, 26 Aug 2013 08:48:33 +0000 (14:18 +0530)] 
kvm uapi: Add KICK_CPU and PV_UNHALT definition to uapi

this is needed by both guest and host.

Originally-from: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agomips/kvm: Make kvm_locore.S 64-bit buildable/safe.
David Daney [Thu, 1 Aug 2013 20:22:35 +0000 (13:22 -0700)] 
mips/kvm: Make kvm_locore.S 64-bit buildable/safe.

We need to use more of the Macros in asm.h to allow kvm_locore.S to
build in a 64-bit kernel.

For 32-bit there is no change in the generated object code.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agomips/kvm: Cleanup .push/.pop directives in kvm_locore.S
David Daney [Thu, 1 Aug 2013 20:22:34 +0000 (13:22 -0700)] 
mips/kvm: Cleanup .push/.pop directives in kvm_locore.S

There are:
.set push
.set noreorder
.set noat
 .
 .
 .
.set pop

Sequences all over the place in this file, but in some places the
final ".set pop" is erroneously converted to ".set push", so none of
these really do what they appear to.

Clean up the whole mess by moving ".set noreorder", ".set noat" to the
top, and get rid of everything else.

Generated object code is unchanged.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agomips/kvm: Improve code formatting in arch/mips/kvm/kvm_locore.S
David Daney [Thu, 1 Aug 2013 20:22:33 +0000 (13:22 -0700)] 
mips/kvm: Improve code formatting in arch/mips/kvm/kvm_locore.S

No code changes, just reflowing some comments and consistently using
tabs and spaces.  Object code is verified to be unchanged.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agopowerpc/kvm: Copy the pvr value after memset
Aneesh Kumar K.V [Thu, 22 Aug 2013 11:38:39 +0000 (17:08 +0530)] 
powerpc/kvm: Copy the pvr value after memset

Otherwise we would clear the pvr value

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: x86: Update symbolic exit codes
Jan Kiszka [Mon, 12 Aug 2013 06:50:41 +0000 (08:50 +0200)] 
KVM: x86: Update symbolic exit codes

Add decoding for INVEPT and reorder the list according to the reason
numbers.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: nVMX: Advertise IA32_PAT in VM exit control
Arthur Chunqi Li [Tue, 6 Aug 2013 10:41:45 +0000 (18:41 +0800)] 
KVM: nVMX: Advertise IA32_PAT in VM exit control

Advertise VM_EXIT_SAVE_IA32_PAT and VM_EXIT_LOAD_IA32_PAT.

Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: nVMX: Fix up VM_ENTRY_IA32E_MODE control feature reporting
Jan Kiszka [Tue, 6 Aug 2013 08:39:56 +0000 (10:39 +0200)] 
KVM: nVMX: Fix up VM_ENTRY_IA32E_MODE control feature reporting

Do not report that we can enter the guest in 64-bit mode if the host is
32-bit only. This is not supported by KVM.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: nEPT: Advertise WB type EPTP
Jan Kiszka [Tue, 6 Aug 2013 08:39:55 +0000 (10:39 +0200)] 
KVM: nEPT: Advertise WB type EPTP

At least WB must be possible.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonVMX: Keep arch.pat in sync on L1-L2 switches
Jan Kiszka [Sun, 4 Aug 2013 15:17:27 +0000 (17:17 +0200)] 
nVMX: Keep arch.pat in sync on L1-L2 switches

When asking vmx to load the PAT MSR for us while switching from L1 to L2
or vice versa, we have to update arch.pat as well as it may later be
used again to load or read out the MSR content.

Reviewed-by: Gleb Natapov <gleb@redhat.com>
Tested-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Miscelleneous cleanups
Nadav Har'El [Mon, 5 Aug 2013 08:07:20 +0000 (11:07 +0300)] 
nEPT: Miscelleneous cleanups

Some trivial code cleanups not really related to nested EPT.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Some additional comments
Nadav Har'El [Mon, 5 Aug 2013 08:07:19 +0000 (11:07 +0300)] 
nEPT: Some additional comments

Some additional comments to preexisting code:
Explain who (L0 or L1) handles EPT violation and misconfiguration exits.
Don't mention "shadow on either EPT or shadow" as the only two options.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoAdvertise the support of EPT to the L1 guest, through the appropriate MSR.
Nadav Har'El [Wed, 7 Aug 2013 12:59:22 +0000 (14:59 +0200)] 
Advertise the support of EPT to the L1 guest, through the appropriate MSR.

This is the last patch of the basic Nested EPT feature, so as to allow
bisection through this patch series: The guest will not see EPT support until
this last patch, and will not attempt to use the half-applied feature.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Nested INVEPT
Nadav Har'El [Mon, 5 Aug 2013 08:07:17 +0000 (11:07 +0300)] 
nEPT: Nested INVEPT

If we let L1 use EPT, we should probably also support the INVEPT instruction.

In our current nested EPT implementation, when L1 changes its EPT table
for L2 (i.e., EPT12), L0 modifies the shadow EPT table (EPT02), and in
the course of this modification already calls INVEPT. But if last level
of shadow page is unsync not all L1's changes to EPT12 are intercepted,
which means roots need to be synced when L1 calls INVEPT. Global INVEPT
should not be different since roots are synced by kvm_mmu_load() each
time EPTP02 changes.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: MMU context for nested EPT
Nadav Har'El [Mon, 5 Aug 2013 08:07:16 +0000 (11:07 +0300)] 
nEPT: MMU context for nested EPT

KVM's existing shadow MMU code already supports nested TDP. To use it, we
need to set up a new "MMU context" for nested EPT, and create a few callbacks
for it (nested_ept_*()). This context should also use the EPT versions of
the page table access functions (defined in the previous patch).
Then, we need to switch back and forth between this nested context and the
regular MMU context when switching between L1 and L2 (when L1 runs this L2
with EPT).

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Add nEPT violation/misconfigration support
Yang Zhang [Tue, 6 Aug 2013 09:00:32 +0000 (12:00 +0300)] 
nEPT: Add nEPT violation/misconfigration support

Inject nEPT fault to L1 guest. This patch is original from Xinhao.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: correctly check if remote tlb flush is needed for shadowed EPT tables
Gleb Natapov [Mon, 5 Aug 2013 08:07:14 +0000 (11:07 +0300)] 
nEPT: correctly check if remote tlb flush is needed for shadowed EPT tables

need_remote_flush() assumes that shadow page is in PT64 format, but
with addition of nested EPT this is no longer always true. Fix it by
bits definitions that depend on host shadow page type.

Reported-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Redefine EPT-specific link_shadow_page()
Yang Zhang [Mon, 5 Aug 2013 08:07:13 +0000 (11:07 +0300)] 
nEPT: Redefine EPT-specific link_shadow_page()

Since nEPT doesn't support A/D bit, so we should not set those bit
when build shadow page table.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Add EPT tables support to paging_tmpl.h
Nadav Har'El [Mon, 5 Aug 2013 08:07:12 +0000 (11:07 +0300)] 
nEPT: Add EPT tables support to paging_tmpl.h

This is the first patch in a series which adds nested EPT support to KVM's
nested VMX. Nested EPT means emulating EPT for an L1 guest so that L1 can use
EPT when running a nested guest L2. When L1 uses EPT, it allows the L2 guest
to set its own cr3 and take its own page faults without either of L0 or L1
getting involved. This often significanlty improves L2's performance over the
previous two alternatives (shadow page tables over EPT, and shadow page
tables over shadow page tables).

This patch adds EPT support to paging_tmpl.h.

paging_tmpl.h contains the code for reading and writing page tables. The code
for 32-bit and 64-bit tables is very similar, but not identical, so
paging_tmpl.h is #include'd twice in mmu.c, once with PTTTYPE=32 and once
with PTTYPE=64, and this generates the two sets of similar functions.

There are subtle but important differences between the format of EPT tables
and that of ordinary x86 64-bit page tables, so for nested EPT we need a
third set of functions to read the guest EPT table and to write the shadow
EPT table.

So this patch adds third PTTYPE, PTTYPE_EPT, which creates functions (prefixed
with "EPT") which correctly read and write EPT tables.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Support shadow paging for guest paging without A/D bits
Gleb Natapov [Mon, 5 Aug 2013 08:07:11 +0000 (11:07 +0300)] 
nEPT: Support shadow paging for guest paging without A/D bits

Some guest paging modes do not support A/D bits. Add support for such
modes in shadow page code. For such modes PT_GUEST_DIRTY_MASK,
PT_GUEST_ACCESSED_MASK, PT_GUEST_DIRTY_SHIFT and PT_GUEST_ACCESSED_SHIFT
should be set to zero.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: make guest's A/D bits depends on guest's paging mode
Gleb Natapov [Mon, 5 Aug 2013 08:07:10 +0000 (11:07 +0300)] 
nEPT: make guest's A/D bits depends on guest's paging mode

This patch makes guest A/D bits definition to be dependable on paging
mode, so when EPT support will be added it will be able to define them
differently.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Move common code to paging_tmpl.h
Nadav Har'El [Mon, 5 Aug 2013 08:07:09 +0000 (11:07 +0300)] 
nEPT: Move common code to paging_tmpl.h

For preparation, we just move gpte_access(), prefetch_invalid_gpte(),
s_rsvd_bits_set(), protect_clean_gpte() and is_dirty_gpte() from mmu.c
to paging_tmpl.h.

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Fix wrong test in kvm_set_cr3
Nadav Har'El [Mon, 5 Aug 2013 08:07:08 +0000 (11:07 +0300)] 
nEPT: Fix wrong test in kvm_set_cr3

kvm_set_cr3() attempts to check if the new cr3 is a valid guest physical
address. The problem is that with nested EPT, cr3 is an *L2* physical
address, not an L1 physical address as this test expects.

As the comment above this test explains, it isn't necessary, and doesn't
correspond to anything a real processor would do. So this patch removes it.

Note that this wrong test could have also theoretically caused problems
in nested NPT, not just in nested EPT. However, in practice, the problem
was avoided: nested_svm_vmexit()/vmrun() do not call kvm_set_cr3 in the
nested NPT case, and instead set the vmcb (and arch.cr3) directly, thus
circumventing the problem. Additional potential calls to the buggy function
are avoided in that we don't trap cr3 modifications when nested NPT is
enabled. However, because in nested VMX we did want to use kvm_set_cr3()
(as requested in Avi Kivity's review of the original nested VMX patches),
we can't avoid this problem and need to fix it.

Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Fix cr3 handling in nested exit and entry
Nadav Har'El [Mon, 5 Aug 2013 08:07:07 +0000 (11:07 +0300)] 
nEPT: Fix cr3 handling in nested exit and entry

The existing code for handling cr3 and related VMCS fields during nested
exit and entry wasn't correct in all cases:

If L2 is allowed to control cr3 (and this is indeed the case in nested EPT),
during nested exit we must copy the modified cr3 from vmcs02 to vmcs12, and
we forgot to do so. This patch adds this copy.

If L0 isn't controlling cr3 when running L2 (i.e., L0 is using EPT), and
whoever does control cr3 (L1 or L2) is using PAE, the processor might have
saved PDPTEs and we should also save them in vmcs12 (and restore later).

Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonEPT: Support LOAD_IA32_EFER entry/exit controls for L1
Nadav Har'El [Mon, 5 Aug 2013 08:07:06 +0000 (11:07 +0300)] 
nEPT: Support LOAD_IA32_EFER entry/exit controls for L1

Recent KVM, since http://kerneltrap.org/mailarchive/linux-kvm/2010/5/2/6261577
switch the EFER MSR when EPT is used and the host and guest have different
NX bits. So if we add support for nested EPT (L1 guest using EPT to run L2)
and want to be able to run recent KVM as L1, we need to allow L1 to use this
EFER switching feature.

To do this EFER switching, KVM uses VM_ENTRY/EXIT_LOAD_IA32_EFER if available,
and if it isn't, it uses the generic VM_ENTRY/EXIT_MSR_LOAD. This patch adds
support for the former (the latter is still unsupported).

Nested entry and exit emulation (prepare_vmcs_02 and load_vmcs12_host_state,
respectively) already handled VM_ENTRY/EXIT_LOAD_IA32_EFER correctly. So all
that's left to do in this patch is to properly advertise this feature to L1.

Note that vmcs12's VM_ENTRY/EXIT_LOAD_IA32_EFER are emulated by L0, by using
vmx_set_efer (which itself sets one of several vmcs02 fields), so we always
support this feature, regardless of whether the host supports it.

Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Nadav Har'El <nyh@il.ibm.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xinhao Xu <xinhao.xu@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: MMU: fix check the reserved bits on the gpte of L2
Xiao Guangrong [Mon, 5 Aug 2013 04:59:19 +0000 (12:59 +0800)] 
KVM: MMU: fix check the reserved bits on the gpte of L2

Current code always uses arch.mmu to check the reserved bits on guest gpte
which is valid only for L1 guest, we should use arch.nested_mmu instead when
we translate gva to gpa for the L2 guest

Fix it by using @mmu instead since it is adapted to the current mmu mode
automatically

The bug can be triggered when nested npt is used and L1 guest and L2 guest
use different mmu mode

Reported-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: nVMX: correctly set tr base on nested vmexit emulation
Gleb Natapov [Sun, 4 Aug 2013 12:08:06 +0000 (15:08 +0300)] 
KVM: nVMX: correctly set tr base on nested vmexit emulation

After commit 21feb4eb64e21f8dc91136b91ee886b978ce6421 tr base is zeroed
during vmexit. Set it to L1's HOST_TR_BASE. This should fix
https://bugzilla.kernel.org/show_bug.cgi?id=60679

Reported-by: Yongjie Ren <yongjie.ren@intel.com>
Reviewed-by: Arthur Chunqi Li <yzt356@gmail.com>
Tested-by: Yongjie Ren <yongjie.ren@intel.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agonVMX: reset rflags register cache during nested vmentry.
Gleb Natapov [Sun, 28 Jul 2013 15:31:06 +0000 (18:31 +0300)] 
nVMX: reset rflags register cache during nested vmentry.

During nested vmentry into vm86 mode a vcpu state is found to be incorrect
because rflags does not have VM flag set since it is read from the cache
and has L1's value instead of L2's. If emulate_invalid_guest_state=1 L0
KVM tries to emulate it, but emulation does not work for nVMX and it
never should happen anyway. Fix that by using vmx_set_rflags() to set
rflags during nested vmentry which takes care of updating register cache.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: Make KVM_HVA_ERR_BAD usable on s390
Dominik Dingel [Fri, 26 Jul 2013 13:04:07 +0000 (15:04 +0200)] 
KVM: s390: Make KVM_HVA_ERR_BAD usable on s390

Current common code uses PAGE_OFFSET to indicate a bad host virtual address.
As this check won't work on architectures that don't map kernel and user memory
into the same address space (e.g. s390), such architectures can now provide
their own KVM_HVA_ERR_BAD defines.

Signed-off-by: Dominik Dingel <dingel@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: Add helper function for setting condition code
Thomas Huth [Fri, 26 Jul 2013 13:04:06 +0000 (15:04 +0200)] 
KVM: s390: Add helper function for setting condition code

Introduced a helper function for setting the CC in the
guest PSW to improve the readability of the code.

Signed-off-by: Thomas Huth <thuth@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: Fix sparse warnings in priv.c
Thomas Huth [Fri, 26 Jul 2013 13:04:05 +0000 (15:04 +0200)] 
KVM: s390: Fix sparse warnings in priv.c

sparse complained about the missing UL postfix for long constants.

Signed-off-by: Thomas Huth <thuth@linux.vnet.ibm.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: declare virtual HW facilities
Michael Mueller [Fri, 26 Jul 2013 13:04:04 +0000 (15:04 +0200)] 
KVM: s390: declare virtual HW facilities

The patch renames the array holding the HW facility bitmaps.
This allows to interprete the variable as set of virtual
machine specific "virtual" facilities. The basic idea is
to make virtual facilities externally managable in future.
An availability test for virtual facilites has been added
as well.

Signed-off-by: Michael Mueller <mimu@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: fix task size check
Martin Schwidefsky [Fri, 26 Jul 2013 13:04:03 +0000 (15:04 +0200)] 
KVM: s390: fix task size check

The gmap_map_segment function uses PGDIR_SIZE in the check for the
maximum address in the tasks address space. This incorrectly limits
the amount of memory usable for a kvm guest to 4TB. The correct limit
is (1UL << 53). As the TASK_SIZE has different values (4TB vs 8PB)
dependent on the existance of the fourth page table level, create
a new define 'TASK_MAX_SIZE' for (1UL << 53).

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: allow sie enablement for multi-threaded programs
Martin Schwidefsky [Fri, 26 Jul 2013 13:04:02 +0000 (15:04 +0200)] 
KVM: s390: allow sie enablement for multi-threaded programs

Improve the code to upgrade the standard 2K page tables to 4K page tables
with PGSTEs to allow the operation to happen when the program is already
multi-threaded.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: x86: handle singlestep during emulation
Paolo Bonzini [Tue, 25 Jun 2013 16:32:07 +0000 (18:32 +0200)] 
KVM: x86: handle singlestep during emulation

This lets debugging work better during emulation of invalid
guest state.

This time the check is done after emulation, but before writeback
of the flags; we need to check the flags *before* execution of the
instruction, we cannot check singlestep_rip because the CS base may
have already been modified.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Conflicts:
arch/x86/kvm/x86.c

11 years agoKVM: x86: handle hardware breakpoints during emulation
Paolo Bonzini [Thu, 30 May 2013 09:48:30 +0000 (11:48 +0200)] 
KVM: x86: handle hardware breakpoints during emulation

This lets debugging work better during emulation of invalid
guest state.

The check is done before emulating the instruction, and (in the case
of guest debugging) reuses EMULATE_DO_MMIO to exit with KVM_EXIT_DEBUG.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: x86: rename EMULATE_DO_MMIO
Paolo Bonzini [Tue, 25 Jun 2013 16:24:41 +0000 (18:24 +0200)] 
KVM: x86: rename EMULATE_DO_MMIO

The next patch will reuse it for other userspace exits than MMIO,
namely debug events.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: introduce __kvm_io_bus_sort_cmp
Paolo Bonzini [Tue, 16 Jul 2013 11:03:29 +0000 (13:03 +0200)] 
KVM: introduce __kvm_io_bus_sort_cmp

kvm_io_bus_sort_cmp is used also directly, not just as a callback for
sort and bsearch.  In these cases, it is handy to have a type-safe
variant.  This patch introduces such a variant, __kvm_io_bus_sort_cmp,
and uses it throughout kvm_main.c.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: PPC: Book3S PR: Load up SPRG3 register with guest value on guest entry
Paul Mackerras [Thu, 11 Jul 2013 11:49:43 +0000 (21:49 +1000)] 
KVM: PPC: Book3S PR: Load up SPRG3 register with guest value on guest entry

Unlike the other general-purpose SPRs, SPRG3 can be read by usermode
code, and is used in recent kernels to store the CPU and NUMA node
numbers so that they can be read by VDSO functions.  Thus we need to
load the guest's SPRG3 value into the real SPRG3 register when entering
the guest, and restore the host's value when exiting the guest.  We don't
need to save the guest SPRG3 value when exiting the guest as usermode
code can't modify SPRG3.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
11 years agoKVM: x86: Drop some unused functions from lapic
Jan Kiszka [Thu, 25 Jul 2013 07:59:22 +0000 (09:59 +0200)] 
KVM: x86: Drop some unused functions from lapic

Both have no users anymore.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: x86: Simplify __apic_accept_irq
Jan Kiszka [Thu, 25 Jul 2013 07:58:45 +0000 (09:58 +0200)] 
KVM: x86: Simplify __apic_accept_irq

If posted interrupts are enabled, we can no longer track if an IRQ was
coalesced based on IRR. So drop this logic also from the classic
software path and simplify apic_test_and_set_irr to apic_set_irr.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoperf, kvm: Support the in_tx/in_tx_cp modifiers in KVM arch perfmon emulation v5
Andi Kleen [Thu, 18 Jul 2013 22:57:02 +0000 (15:57 -0700)] 
perf, kvm: Support the in_tx/in_tx_cp modifiers in KVM arch perfmon emulation v5

[KVM maintainers:
The underlying support for this is in perf/core now. So please merge
this patch into the KVM tree.]

This is not arch perfmon, but older CPUs will just ignore it. This makes
it possible to do at least some TSX measurements from a KVM guest

v2: Various fixes to address review feedback
v3: Ignore the bits when no CPUID. No #GP. Force raw events with TSX bits.
v4: Use reserved bits for #GP
v5: Remove obsolete argument
Acked-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: nVMX: Set segment infomation of L1 when L2 exits
Arthur Chunqi Li [Mon, 15 Jul 2013 08:04:08 +0000 (16:04 +0800)] 
KVM: nVMX: Set segment infomation of L1 when L2 exits

When L2 exits to L1, segment infomations of L1 are not set correctly.
According to Intel SDM 27.5.2(Loading Host Segment and Descriptor
Table Registers), segment base/limit/access right of L1 should be
set to some designed value when L2 exits to L1. This patch fixes
this.

Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Reviewed-by: Gleb Natapov <gnatapov@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoremove sched notifier for cross-cpu migrations
Marcelo Tosatti [Thu, 11 Jul 2013 01:21:57 +0000 (22:21 -0300)] 
remove sched notifier for cross-cpu migrations

Linux as a guest on KVM hypervisor, the only user of the pvclock
vsyscall interface, does not require notification on task migration
because:

1. cpu ID number maps 1:1 to per-CPU pvclock time info.
2. per-CPU pvclock time info is updated if the
   underlying CPU changes.
3. that version is increased whenever underlying CPU
   changes.

Which is sufficient to guarantee nanoseconds counter
is calculated properly.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: nVMX: Fix read/write to MSR_IA32_FEATURE_CONTROL
Nadav Har'El [Mon, 8 Jul 2013 11:12:35 +0000 (19:12 +0800)] 
KVM: nVMX: Fix read/write to MSR_IA32_FEATURE_CONTROL

Fix read/write to IA32_FEATURE_CONTROL MSR in nested environment.

This patch simulate this MSR in nested_vmx and the default value is
0x0. BIOS should set it to 0x5 before VMXON. After setting the lock
bit, write to it will cause #GP(0).

Another QEMU patch is also needed to handle emulation of reset
and migration. Reset to vCPU should clear this MSR and migration
should reserve value of it.

This patch is based on Nadav's previous commit.
http://permalink.gmane.org/gmane.comp.emulators.kvm.devel/88478

Signed-off-by: Nadav Har'El <nyh@math.technion.ac.il>
Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: x86: Drop useless cast
Mathias Krause [Wed, 26 Jun 2013 18:36:23 +0000 (20:36 +0200)] 
KVM: x86: Drop useless cast

Void pointers don't need no casting, drop it.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: VMX: Use proper types to access const arrays
Mathias Krause [Wed, 26 Jun 2013 18:36:21 +0000 (20:36 +0200)] 
KVM: VMX: Use proper types to access const arrays

Use a const pointer type instead of casting away the const qualifier
from const arrays. Keep the pointer array on the stack, nonetheless.
Making it static just increases the object size.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: nVMX: Set success rflags when emulate VMXON/VMXOFF in nested virt
Arthur Chunqi Li [Thu, 4 Jul 2013 07:03:33 +0000 (15:03 +0800)] 
KVM: nVMX: Set success rflags when emulate VMXON/VMXOFF in nested virt

Set rflags after successfully emulateing VMXON/VMXOFF in VMX.

Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: nVMX: Change location of 3 functions in vmx.c
Arthur Chunqi Li [Thu, 4 Jul 2013 07:03:32 +0000 (15:03 +0800)] 
KVM: nVMX: Change location of 3 functions in vmx.c

Move nested_vmx_succeed/nested_vmx_failInvalid/nested_vmx_failValid
ahead of handle_vmon to eliminate double declaration in the same
file

Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: x86: Avoid zapping mmio sptes twice for generation wraparound
Takuya Yoshikawa [Thu, 4 Jul 2013 04:41:26 +0000 (13:41 +0900)] 
KVM: x86: Avoid zapping mmio sptes twice for generation wraparound

Now that kvm_arch_memslots_updated() catches every increment of the
memslots->generation, checking if the mmio generation has reached its
maximum value is enough.

Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: Introduce kvm_arch_memslots_updated()
Takuya Yoshikawa [Thu, 4 Jul 2013 04:40:29 +0000 (13:40 +0900)] 
KVM: Introduce kvm_arch_memslots_updated()

This is called right after the memslots is updated, i.e. when the result
of update_memslots() gets installed in install_new_memslots().  Since
the memslots needs to be updated twice when we delete or move a memslot,
kvm_arch_commit_memory_region() does not correspond to this exactly.

In the following patch, x86 will use this new API to check if the mmio
generation has reached its maximum value, in which case mmio sptes need
to be flushed out.

Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Acked-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: s390: use cookies for ioeventfd
Cornelia Huck [Wed, 3 Jul 2013 14:30:54 +0000 (16:30 +0200)] 
KVM: s390: use cookies for ioeventfd

Make use of cookies for the virtio ccw notification hypercall to speed up
lookup of devices on the io bus.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
[Small fix to a comment. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoKVM: kvm-io: support cookies
Cornelia Huck [Wed, 3 Jul 2013 14:30:53 +0000 (16:30 +0200)] 
KVM: kvm-io: support cookies

Add new functions kvm_io_bus_{read,write}_cookie() that allows users of
the kvm io infrastructure to use a cookie value to speed up lookup of a
device on an io bus.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
11 years agoKVM: MMU: avoid fast page fault fixing mmio page fault
Xiao Guangrong [Thu, 18 Jul 2013 04:52:37 +0000 (12:52 +0800)] 
KVM: MMU: avoid fast page fault fixing mmio page fault

Currently, fast page fault incorrectly tries to fix mmio page fault when
the generation number is invalid (spte.gen != kvm.gen).  It then returns
to guest to retry the fault since it sees the last spte is nonpresent.
This causes an infinite loop.

Since fast page fault only works for direct mmu, the issue exists when
1) tdp is enabled. It is only triggered only on AMD host since on Intel host
   the mmio page fault is recognized as ept-misconfig whose handler call
   fault-page path with error_code = 0

2) guest paging is disabled. Under this case, the issue is hardly discovered
   since paging disable is short-lived and the sptes will be invalid after
   memslot changed for 150 times

Fix it by filtering out MMIO page faults in page_fault_can_be_fast.

Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Tested-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
11 years agoLinux 3.11-rc1
Linus Torvalds [Sun, 14 Jul 2013 22:18:27 +0000 (15:18 -0700)] 
Linux 3.11-rc1

11 years agoMerge branch 'slab/for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penber...
Linus Torvalds [Sun, 14 Jul 2013 22:14:29 +0000 (15:14 -0700)] 
Merge branch 'slab/for-linus' of git://git./linux/kernel/git/penberg/linux

Pull slab update from Pekka Enberg:
 "Highlights:

  - Fix for boot-time problems on some architectures due to
    init_lock_keys() not respecting kmalloc_caches boundaries
    (Christoph Lameter)

  - CONFIG_SLUB_CPU_PARTIAL requested by RT folks (Joonsoo Kim)

  - Fix for excessive slab freelist draining (Wanpeng Li)

  - SLUB and SLOB cleanups and fixes (various people)"

I ended up editing the branch, and this avoids two commits at the end
that were immediately reverted, and I instead just applied the oneliner
fix in between myself.

* 'slab/for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/linux
  slub: Check for page NULL before doing the node_match check
  mm/slab: Give s_next and s_stop slab-specific names
  slob: Check for NULL pointer before calling ctor()
  slub: Make cpu partial slab support configurable
  slab: add kmalloc() to kernel API documentation
  slab: fix init_lock_keys
  slob: use DIV_ROUND_UP where possible
  slub: do not put a slab to cpu partial list when cpu_partial is 0
  mm/slub: Use node_nr_slabs and node_nr_objs in get_slabinfo
  mm/slub: Drop unnecessary nr_partials
  mm/slab: Fix /proc/slabinfo unwriteable for slab
  mm/slab: Sharing s_next and s_stop between slab and slub
  mm/slab: Fix drain freelist excessively
  slob: Rework #ifdeffery in slab.h
  mm, slab: moved kmem_cache_alloc_node comment to correct place

11 years agoslub: Check for page NULL before doing the node_match check
Steven Rostedt [Thu, 17 Jan 2013 18:10:58 +0000 (13:10 -0500)] 
slub: Check for page NULL before doing the node_match check

In the -rt kernel (mrg), we hit the following dump:

BUG: unable to handle kernel NULL pointer dereference at           (null)
IP: [<ffffffff811573f1>] kmem_cache_alloc_node+0x51/0x180
PGD a2d39067 PUD b1641067 PMD 0
Oops: 0000 [#1] PREEMPT SMP
Modules linked in: sunrpc cpufreq_ondemand ipv6 tg3 joydev sg serio_raw pcspkr k8temp amd64_edac_mod edac_core i2c_piix4 e100 mii shpchp ext4 mbcache jbd2 sd_mod crc_t10dif sr_mod cdrom sata_svw ata_generic pata_acpi pata_serverworks radeon ttm drm_kms_helper drm hwmon i2c_algo_bit i2c_core dm_mirror dm_region_hash dm_log dm_mod
CPU 3
Pid: 20878, comm: hackbench Not tainted 3.6.11-rt25.14.el6rt.x86_64 #1 empty empty/Tyan Transport GT24-B3992
RIP: 0010:[<ffffffff811573f1>]  [<ffffffff811573f1>] kmem_cache_alloc_node+0x51/0x180
RSP: 0018:ffff8800a9b17d70  EFLAGS: 00010213
RAX: 0000000000000000 RBX: 0000000001200011 RCX: ffff8800a06d8000
RDX: 0000000004d92a03 RSI: 00000000000000d0 RDI: ffff88013b805500
RBP: ffff8800a9b17dc0 R08: ffff88023fd14d10 R09: ffffffff81041cbd
R10: 00007f4e3f06e9d0 R11: 0000000000000246 R12: ffff88013b805500
R13: ffff8801ff46af40 R14: 0000000000000001 R15: 0000000000000000
FS:  00007f4e3f06e700(0000) GS:ffff88023fd00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 0000000000000000 CR3: 00000000a2d3a000 CR4: 00000000000007e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process hackbench (pid: 20878, threadinfo ffff8800a9b16000, task ffff8800a06d8000)
Stack:
 ffff8800a9b17da0 ffffffff81202e08 ffff8800a9b17de0 000000d001200011
 0000000001200011 0000000001200011 0000000000000000 0000000000000000
 00007f4e3f06e9d0 0000000000000000 ffff8800a9b17e60 ffffffff81041cbd
Call Trace:
 [<ffffffff81202e08>] ? current_has_perm+0x68/0x80
 [<ffffffff81041cbd>] copy_process+0xdd/0x15b0
 [<ffffffff810a2125>] ? rt_up_read+0x25/0x30
 [<ffffffff8104369a>] do_fork+0x5a/0x360
 [<ffffffff8107c66b>] ? migrate_enable+0xeb/0x220
 [<ffffffff8100b068>] sys_clone+0x28/0x30
 [<ffffffff81527423>] stub_clone+0x13/0x20
 [<ffffffff81527152>] ? system_call_fastpath+0x16/0x1b
Code: 89 fc 89 75 cc 41 89 d6 4d 8b 04 24 65 4c 03 04 25 48 ae 00 00 49 8b 50 08 4d 8b 28 49 8b 40 10 4d 85 ed 74 12 41 83 fe ff 74 27 <48> 8b 00 48 c1 e8 3a 41 39 c6 74 1b 8b 75 cc 4c 89 c9 44 89 f2
RIP  [<ffffffff811573f1>] kmem_cache_alloc_node+0x51/0x180
 RSP <ffff8800a9b17d70>
CR2: 0000000000000000
---[ end trace 0000000000000002 ]---

Now, this uses SLUB pretty much unmodified, but as it is the -rt kernel
with CONFIG_PREEMPT_RT set, spinlocks are mutexes, although they do
disable migration. But the SLUB code is relatively lockless, and the
spin_locks there are raw_spin_locks (not converted to mutexes), thus I
believe this bug can happen in mainline without -rt features. The -rt
patch is just good at triggering mainline bugs ;-)

Anyway, looking at where this crashed, it seems that the page variable
can be NULL when passed to the node_match() function (which does not
check if it is NULL). When this happens we get the above panic.

As page is only used in slab_alloc() to check if the node matches, if
it's NULL I'm assuming that we can say it doesn't and call the
__slab_alloc() code. Is this a correct assumption?

Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Sun, 14 Jul 2013 18:42:26 +0000 (11:42 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull more vfs stuff from Al Viro:
 "O_TMPFILE ABI changes, Oleg's fput() series, misc cleanups, including
  making simple_lookup() usable for filesystems with non-NULL s_d_op,
  which allows us to get rid of quite a bit of ugliness"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  sunrpc: now we can just set ->s_d_op
  cgroup: we can use simple_lookup() now
  efivarfs: we can use simple_lookup() now
  make simple_lookup() usable for filesystems that set ->s_d_op
  configfs: don't open-code d_alloc_name()
  __rpc_lookup_create_exclusive: pass string instead of qstr
  rpc_create_*_dir: don't bother with qstr
  llist: llist_add() can use llist_add_batch()
  llist: fix/simplify llist_add() and llist_add_batch()
  fput: turn "list_head delayed_fput_list" into llist_head
  fs/file_table.c:fput(): add comment
  Safer ABI for O_TMPFILE

11 years agosunrpc: now we can just set ->s_d_op
Al Viro [Sun, 14 Jul 2013 13:55:39 +0000 (17:55 +0400)] 
sunrpc: now we can just set ->s_d_op

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agocgroup: we can use simple_lookup() now
Al Viro [Sun, 14 Jul 2013 13:50:23 +0000 (17:50 +0400)] 
cgroup: we can use simple_lookup() now

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agoefivarfs: we can use simple_lookup() now
Al Viro [Sun, 14 Jul 2013 13:48:35 +0000 (17:48 +0400)] 
efivarfs: we can use simple_lookup() now

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agomake simple_lookup() usable for filesystems that set ->s_d_op
Al Viro [Sun, 14 Jul 2013 13:43:25 +0000 (17:43 +0400)] 
make simple_lookup() usable for filesystems that set ->s_d_op

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agoconfigfs: don't open-code d_alloc_name()
Al Viro [Sun, 14 Jul 2013 13:16:52 +0000 (17:16 +0400)] 
configfs: don't open-code d_alloc_name()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years ago__rpc_lookup_create_exclusive: pass string instead of qstr
Al Viro [Sun, 14 Jul 2013 13:09:57 +0000 (17:09 +0400)] 
__rpc_lookup_create_exclusive: pass string instead of qstr

... and use d_hash_and_lookup() instead of open-coding it, for fsck sake...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agorpc_create_*_dir: don't bother with qstr
Al Viro [Sun, 14 Jul 2013 12:43:54 +0000 (16:43 +0400)] 
rpc_create_*_dir: don't bother with qstr

just pass the name

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agoMerge branch 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86
Linus Torvalds [Sun, 14 Jul 2013 01:08:23 +0000 (18:08 -0700)] 
Merge branch 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86

Pull x86 platform driver updates from Matthew Garrett:
 "Nothing overly exciting here - a couple of new drivers that don't do a
  great deal, along with some miscellaneous fixes and a couple of small
  feature enablement patches"

* 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86:
  x86 platform drivers: fix gpio leak
  toshiba_acpi: Add dependency on SERIO_I8042
  asus-nb-wmi: set wapf=4 for ASUSTeK COMPUTER INC. 1015E/U
  Add trivial driver to disable Intel Smart Connect
  Add support driver for Intel Rapid Start Technology
  hp-wmi: add supports for POST code error
  asus-wmi: control wlan-led only if wapf == 4
  drivers/platform/x86/intel_ips: Convert to module_pci_driver
  asus-nb-wmi: ignore ALS notification key code
  asus-wmi: append newline to messages
  x86: asus-laptop: fix invalid point access
  x86: msi-laptop: fix memleak
  amilo-rfkill: Add dependency on SERIO_I8042
  dell-laptop: fix error return code in dell_init()
  hp-wmi: Enable hotkeys on some systems

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sun, 14 Jul 2013 01:05:13 +0000 (18:05 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull second round of input updates from Dmitry Torokhov:
 "An update to Elantech driver to support hardware v7, fix to the new
  cyttsp4 driver to use proper addressing, ads7846 device tree support
  and nspire-keypad got a small cleanup."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: nspire-keypad - replace magic offset with define
  Input: elantech - fix for newer hardware versions (v7)
  Input: cyttsp4 - use 16bit address for I2C/SPI communication
  Input: ads7846 - add device tree bindings
  Input: ads7846 - make sure we do not change platform data

11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sun, 14 Jul 2013 00:42:22 +0000 (17:42 -0700)] 
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "Just a bunch of small fixes and tidy ups:

   1) Finish the "busy_poll" renames, from Eliezer Tamir.

   2) Fix RCU stalls in IFB driver, from Ding Tianhong.

   3) Linearize buffers properly in tun/macvtap zerocopy code.

   4) Don't crash on rmmod in vxlan, from Pravin B Shelar.

   5) Spinlock used before init in alx driver, from Maarten Lankhorst.

   6) A sparse warning fix in bnx2x broke TSO checksums, fix from Dmitry
      Kravkov.

   7) Dummy and ifb driver load failure paths can oops, fixes from Tan
      Xiaojun and Ding Tianhong.

   8) Correct MTU calculations in IP tunnels, from Alexander Duyck.

   9) Account all TCP retransmits in SNMP stats properly, from Yuchung
      Cheng.

  10) atl1e and via-rhine do not handle DMA mapping failures properly,
      from Neil Horman.

  11) Various equal-cost multipath route fixes in ipv6 from Hannes
      Frederic Sowa"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (36 commits)
  ipv6: only static routes qualify for equal cost multipathing
  via-rhine: fix dma mapping errors
  atl1e: fix dma mapping warnings
  tcp: account all retransmit failures
  usb/net/r815x: fix cast to restricted __le32
  usb/net/r8152: fix integer overflow in expression
  net: access page->private by using page_private
  net: strict_strtoul is obsolete, use kstrtoul instead
  drivers/net/ieee802154: don't use devm_pinctrl_get_select_default() in probe
  drivers/net/ethernet/cadence: don't use devm_pinctrl_get_select_default() in probe
  drivers/net/can/c_can: don't use devm_pinctrl_get_select_default() in probe
  net/usb: add relative mii functions for r815x
  net/tipc: use %*phC to dump small buffers in hex form
  qlcnic: Adding Maintainers.
  gre: Fix MTU sizing check for gretap tunnels
  pkt_sched: sch_qfq: remove forward declaration of qfq_update_agg_ts
  pkt_sched: sch_qfq: improve efficiency of make_eligible
  gso: Update tunnel segmentation to support Tx checksum offload
  inet: fix spacing in assignment
  ifb: fix oops when loading the ifb failed
  ...

11 years agoMerge tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb...
Linus Torvalds [Sun, 14 Jul 2013 00:41:21 +0000 (17:41 -0700)] 
Merge tag 'scsi-for-linus' of git://git./linux/kernel/git/jejb/scsi

Pull final round of SCSI updates from James Bottomley:
 "This is the remaining set of SCSI patches for the merge window.  It's
  mostly driver updates (scsi_debug, qla2xxx, storvsc, mp3sas).  There
  are also several bug fixes in fcoe, libfc, and megaraid_sas.  We also
  have a couple of core changes to try to make device destruction more
  deterministic"

* tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (46 commits)
  [SCSI] scsi constants: command, sense key + additional sense strings
  fcoe: Reduce number of sparse warnings
  fcoe: Stop fc_rport_priv structure leak
  libfcoe: Fix meaningless log statement
  libfc: Differentiate echange timer cancellation debug statements
  libfc: Remove extra space in fc_exch_timer_cancel definition
  fcoe: fix the link error status block sparse warnings
  fcoe: Fix smatch warning in fcoe_fdmi_info function
  libfc: Reject PLOGI from nodes with incompatible role
  [SCSI] enable destruction of blocked devices which fail LUN scanning
  [SCSI] Fix race between starved list and device removal
  [SCSI] megaraid_sas: fix a bug for 64 bit arches
  [SCSI] scsi_debug: reduce duplication between prot_verify_read and prot_verify_write
  [SCSI] scsi_debug: simplify offset calculation for dif_storep
  [SCSI] scsi_debug: invalidate protection info for unmapped region
  [SCSI] scsi_debug: fix NULL pointer dereference with parameters dif=0 dix=1
  [SCSI] scsi_debug: fix incorrectly nested kmap_atomic()
  [SCSI] scsi_debug: fix invalid address passed to kunmap_atomic()
  [SCSI] mpt3sas: Bump driver version to v02.100.00.00
  [SCSI] mpt3sas: when async scanning is enabled then while scanning, devices are removed but their transport layer entries are not removed
  ...

11 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 13 Jul 2013 22:37:57 +0000 (15:37 -0700)] 
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Thomas Gleixner:
 "Fix a potential deadlock versus hrtimers"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched: Fix HRTICK

11 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 13 Jul 2013 22:37:30 +0000 (15:37 -0700)] 
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq updates from Thomas Gleixner:
 - core fix for missing round up in the generic irq chip implementation
 - new irq chip for MOXA SoCs
 - a few fixes and cleanups in the irqchip drivers

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip: Add support for MOXA ART SoCs
  genirq: generic chip: Use DIV_ROUND_UP to calculate numchips
  irqchip: nvic: Fix wrong num_ct argument for irq_alloc_domain_generic_chips()
  irqchip: sun4i: Staticize sun4i_irq_ack()
  irqchip: vt8500: Staticize local symbols

11 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 13 Jul 2013 22:36:09 +0000 (15:36 -0700)] 
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer updates from Thomas Gleixner:
 - watchdog fixes for full dynticks
 - improved debug output for full dynticks
 - remove an obsolete full dynticks check
 - two ARM SoC clocksource drivers for sharing across SoCs
 - tick broadcast fix for CPU hotplug

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick: broadcast: Check broadcast mode on CPU hotplug
  clocksource: arm_global_timer: Add ARM global timer support
  clocksource: Add Marvell Orion SoC timer
  nohz: Remove obsolete check for full dynticks CPUs to be RCU nocbs
  watchdog: Boot-disable by default on full dynticks
  watchdog: Rename confusing state variable
  watchdog: Register / unregister watchdog kthreads on sysctl control
  nohz: Warn if the machine can not perform nohz_full

11 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 13 Jul 2013 22:35:47 +0000 (15:35 -0700)] 
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Thomas Gleixner:
 - fix for do_div() abuse on x86
 - locking fix in perf core
 - a pile of (build) fixes and cleanups in perf tools

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (27 commits)
  perf/x86: Fix incorrect use of do_div() in NMI warning
  perf: Fix perf_lock_task_context() vs RCU
  perf: Remove WARN_ON_ONCE() check in __perf_event_enable() for valid scenario
  perf: Clone child context from parent context pmu
  perf script: Fix broken include in Context.xs
  perf tools: Fix -ldw/-lelf link test when static linking
  perf tools: Revert regression in configuration of Python support
  perf tools: Fix perf version generation
  perf stat: Fix per-socket output bug for uncore events
  perf symbols: Fix vdso list searching
  perf evsel: Fix missing increment in sample parsing
  perf tools: Update symbol_conf.nr_events when processing attribute events
  perf tools: Fix new_term() missing free on error path
  perf tools: Fix parse_events_terms() segfault on error path
  perf evsel: Fix count parameter to read call in event_format__new
  perf tools: fix a typo of a Power7 event name
  perf tools: Fix -x/--exclude-other option for report command
  perf evlist: Enhance perf_evlist__start_workload()
  perf record: Remove -f/--force option
  perf record: Remove -A/--append option
  ...

11 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 13 Jul 2013 22:35:12 +0000 (15:35 -0700)] 
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull core locking updates from Thomas Gleixner:
 "Header cleanup as requested by Linus"

(This is the "don't include support for ww_mutex in a header file that
everybody wants, when almost nobody wants the ww part" change)

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  mutex: Move ww_mutex definitions to ww_mutex.h

11 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sat, 13 Jul 2013 22:00:26 +0000 (15:00 -0700)] 
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "This is our first set of fixes from arm-soc for 3.11.
   - A handful of build and warning fixes from Arnd
   - A collection of OMAP fixes
   - defconfig updates to make the default configs more useful for real
     use (and testing) out of the box on hardware

  And a couple of other small fixes.  Some of these have been recently
  applied but it's normally how we deal with fixes, with less bake time
  in -next needed"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (32 commits)
  arm: multi_v7_defconfig: Tweaks for omap and sunxi
  arm: multi_v7_defconfig: add i.MX options and NFS root
  ARM: omap2: add select of TI_PRIV_EDMA
  ARM: exynos: select PM_GENERIC_DOMAINS only when used
  ARM: ixp4xx: avoid circular header dependency
  ARM: OMAP: omap_common_late_init may be unused
  ARM: sti: move DEBUG_STI_UART into alphabetical order
  ARM: OMAP: build mach-omap code only if needed
  ARM: zynq: use DT_MACHINE_START
  ARM: omap5: omap5 has SCU and TWD
  ARM: OMAP2+: omap2plus_defconfig: Enable appended DTB support
  ARM: OMAP2+: Enable TI_EDMA in omap2plus_defconfig
  ARM: OMAP2+: omap2plus_defconfig: enable DRA752 thermal support by default
  ARM: OMAP2+: omap2plus_defconfig: enable TI bandgap driver
  ARM: OMAP2+: devices: remove duplicated include from devices.c
  ARM: OMAP3: igep0020: Set DSS pins in correct mux mode.
  ARM: OMAP2+: N900: enable N900-specific drivers even if device tree is enabled
  ARM: OMAP2+: Cocci spatch "ptr_ret.spatch"
  ARM: OMAP2+: Remove obsolete Makefile line
  ARM: OMAP5: Enable Cortex A15 errata 798181
  ...

11 years agoMerge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Sat, 13 Jul 2013 21:58:36 +0000 (14:58 -0700)] 
Merge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm

Pull ARM fixes from Russell King:
 "A few fixes for ARM, mostly just one liners with the exception of the
  missing section specification.  We decided not to rely on .previous to
  fix this but to explicitly state the section we want the code to be
  in."

* 'fixes' of git://git.linaro.org/people/rmk/linux-arm:
  ARM: 7778/1: smp_twd: twd_update_frequency need be run on all online CPUs
  ARM: 7782/1: Kconfig: Let ARM_ERRATA_364296 not depend on CONFIG_SMP
  ARM: mm: fix boot on SA1110 Assabet
  ARM: 7781/1: mmu: Add debug_ll_io_init() mappings to early mappings
  ARM: 7780/1: add missing linker section markup to head-common.S

11 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Sat, 13 Jul 2013 21:52:21 +0000 (14:52 -0700)] 
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS updates from Ralf Baechle:
 "MIPS updates:

   - All the things that didn't make 3.10.
   - Removes the Windriver PPMC platform.  Nobody will miss it.
   - Remove a workaround from kernel/irq/irqdomain.c which was there
     exclusivly for MIPS.  Patch by Grant Likely.
   - More small improvments for the SEAD 3 platform
   - Improvments on the BMIPS / SMP support for the BCM63xx series.
   - Various cleanups of dead leftovers.
   - Platform support for the Cavium Octeon-based EdgeRouter Lite.

  Two large KVM patchsets didn't make it for this pull request because
  their respective authors are vacationing"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus: (124 commits)
  MIPS: Kconfig: Add missing MODULES dependency to VPE_LOADER
  MIPS: BCM63xx: CLK: Add dummy clk_{set,round}_rate() functions
  MIPS: SEAD3: Disable L2 cache on SEAD-3.
  MIPS: BCM63xx: Enable second core SMP on BCM6328 if available
  MIPS: BCM63xx: Add SMP support to prom.c
  MIPS: define write{b,w,l,q}_relaxed
  MIPS: Expose missing pci_io{map,unmap} declarations
  MIPS: Malta: Update GCMP detection.
  Revert "MIPS: make CAC_ADDR and UNCAC_ADDR account for PHYS_OFFSET"
  MIPS: APSP: Remove <asm/kspd.h>
  SSB: Kconfig: Amend SSB_EMBEDDED dependencies
  MIPS: microMIPS: Fix improper definition of ISA exception bit.
  MIPS: Don't try to decode microMIPS branch instructions where they cannot exist.
  MIPS: Declare emulate_load_store_microMIPS as a static function.
  MIPS: Fix typos and cleanup comment
  MIPS: Cleanup indentation and whitespace
  MIPS: BMIPS: support booting from physical CPU other than 0
  MIPS: Only set cpu_has_mmips if SYS_SUPPORTS_MICROMIPS
  MIPS: GIC: Fix gic_set_affinity infinite loop
  MIPS: Don't save/restore OCTEON wide multiplier state on syscalls.
  ...

This page took 0.055841 seconds and 5 git commands to generate.