commit a9545779ee upstream.
Use kvm_pfn_t, a.k.a. u64, for the local 'pfn' variable when retrieving
a so called "remapped" hva/pfn pair. In theory, the hva could resolve to
a pfn in high memory on a 32-bit kernel.
This bug was inadvertantly exposed by commit bd2fae8da7 ("KVM: do not
assume PTE is writable after follow_pfn"), which added an error PFN value
to the mix, causing gcc to comlain about overflowing the unsigned long.
arch/x86/kvm/../../../virt/kvm/kvm_main.c: In function ‘hva_to_pfn_remapped’:
include/linux/kvm_host.h:89:30: error: conversion from ‘long long unsigned int’
to ‘long unsigned int’ changes value from
‘9218868437227405314’ to ‘2’ [-Werror=overflow]
89 | #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
| ^
virt/kvm/kvm_main.c:1935:9: note: in expansion of macro ‘KVM_PFN_ERR_RO_FAULT’
Cc: stable@vger.kernel.org
Fixes: add6a0cd1c ("KVM: MMU: try to fix up page faults before giving up")
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20210208201940.1258328-1-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9fd6dad126 upstream.
Currently, the follow_pfn function is exported for modules but
follow_pte is not. However, follow_pfn is very easy to misuse,
because it does not provide protections (so most of its callers
assume the page is writable!) and because it returns after having
already unlocked the page table lock.
Provide instead a simplified version of follow_pte that does
not have the pmdpp and range arguments. The older version
survives as follow_invalidate_pte() for use by fs/dax.c.
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit bd2fae8da7 upstream.
In order to convert an HVA to a PFN, KVM usually tries to use
the get_user_pages family of functinso. This however is not
possible for VM_IO vmas; in that case, KVM instead uses follow_pfn.
In doing this however KVM loses the information on whether the
PFN is writable. That is usually not a problem because the main
use of VM_IO vmas with KVM is for BARs in PCI device assignment,
however it is a bug. To fix it, use follow_pte and check pte_write
while under the protection of the PTE lock. The information can
be used to fail hva_to_pfn_remapped or passed back to the
caller via *writable.
Usage of follow_pfn was introduced in commit add6a0cd1c ("KVM: MMU: try to fix
up page faults before giving up", 2016-07-05); however, even older version
have the same issue, all the way back to commit 2e2e3738af ("KVM:
Handle vma regions with no backing page", 2008-07-20), as they also did
not check whether the PFN was writable.
Fixes: 2e2e3738af ("KVM: Handle vma regions with no backing page")
Reported-by: David Stevens <stevensd@google.com>
Cc: 3pvd@google.com
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 139bc8a614 upstream.
The use of a tagged address could be pretty confusing for the
whole memslot infrastructure as well as the MMU notifiers.
Forbid it altogether, as it never quite worked the first place.
Cc: stable@vger.kernel.org
Reported-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 88bf56d04b upstream.
In kvm_mmu_notifier_invalidate_range_start(), tlbs_dirty is used as:
need_tlb_flush |= kvm->tlbs_dirty;
with need_tlb_flush's type being int and tlbs_dirty's type being long.
It means that tlbs_dirty is always used as int and the higher 32 bits
is useless. We need to check tlbs_dirty in a correct way and this
change checks it directly without propagating it to need_tlb_flush.
Note: it's _extremely_ unlikely this neglecting of higher 32 bits can
cause problems in practice. It would require encountering tlbs_dirty
on a 4 billion count boundary, and KVM would need to be using shadow
paging or be running a nested guest.
Cc: stable@vger.kernel.org
Fixes: a4ee1ca4a3 ("KVM: MMU: delay flush all tlbs on sync_page path")
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
Message-Id: <20201217154118.16497-1-jiangshanlai@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 23bde34771 upstream.
It was recently reported that if GICR_TYPER is accessed before the RD base
address is set, we'll suffer from the unset @rdreg dereferencing. Oops...
gpa_t last_rdist_typer = rdreg->base + GICR_TYPER +
(rdreg->free_index - 1) * KVM_VGIC_V3_REDIST_SIZE;
It's "expected" that users will access registers in the redistributor if
the RD has been properly configured (e.g., the RD base address is set). But
it hasn't yet been covered by the existing documentation.
Per discussion on the list [1], the reporting of the GICR_TYPER.Last bit
for userspace never actually worked. And it's difficult for us to emulate
it correctly given that userspace has the flexibility to access it any
time. Let's just drop the reporting of the Last bit for userspace for now
(userspace should have full knowledge about it anyway) and it at least
prevents kernel from panic ;-)
[1] https://lore.kernel.org/kvmarm/c20865a267e44d1e2c0d52ce4e012263@kernel.org/
Fixes: ba7b3f1275 ("KVM: arm/arm64: Revisit Redistributor TYPER last bit computation")
Reported-by: Keqian Zhu <zhukeqian1@huawei.com>
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Link: https://lore.kernel.org/r/20201117151629.1738-1-yuzenghui@huawei.com
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1de111b51b upstream.
According to the SMCCC spec[1](7.5.2 Discovery) the
ARM_SMCCC_ARCH_WORKAROUND_1 function id only returns 0, 1, and
SMCCC_RET_NOT_SUPPORTED.
0 is "workaround required and safe to call this function"
1 is "workaround not required but safe to call this function"
SMCCC_RET_NOT_SUPPORTED is "might be vulnerable or might not be, who knows, I give up!"
SMCCC_RET_NOT_SUPPORTED might as well mean "workaround required, except
calling this function may not work because it isn't implemented in some
cases". Wonderful. We map this SMC call to
0 is SPECTRE_MITIGATED
1 is SPECTRE_UNAFFECTED
SMCCC_RET_NOT_SUPPORTED is SPECTRE_VULNERABLE
For KVM hypercalls (hvc), we've implemented this function id to return
SMCCC_RET_NOT_SUPPORTED, 0, and SMCCC_RET_NOT_REQUIRED. One of those
isn't supposed to be there. Per the code we call
arm64_get_spectre_v2_state() to figure out what to return for this
feature discovery call.
0 is SPECTRE_MITIGATED
SMCCC_RET_NOT_REQUIRED is SPECTRE_UNAFFECTED
SMCCC_RET_NOT_SUPPORTED is SPECTRE_VULNERABLE
Let's clean this up so that KVM tells the guest this mapping:
0 is SPECTRE_MITIGATED
1 is SPECTRE_UNAFFECTED
SMCCC_RET_NOT_SUPPORTED is SPECTRE_VULNERABLE
Note: SMCCC_RET_NOT_AFFECTED is 1 but isn't part of the SMCCC spec
Fixes: c118bbb527 ("arm64: KVM: Propagate full Spectre v2 workaround state to KVM guests")
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Acked-by: Marc Zyngier <maz@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Cc: Andre Przywara <andre.przywara@arm.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Link: https://developer.arm.com/documentation/den0028/latest [1]
Link: https://lore.kernel.org/r/20201023154751.1973872-1-swboyd@chromium.org
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 91a2c34b7d ]
VFIO allows a device driver to resolve a fault by mapping a MMIO
range. This can be subsequently result in user_mem_abort() to
try and compute a huge mapping based on the MMIO pfn, which is
a sure recipe for things to go wrong.
Instead, force a PTE mapping when the pfn faulted in has a device
mapping.
Fixes: 6d674e28f6 ("KVM: arm/arm64: Properly handle faulting of device mappings")
Suggested-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Santosh Shukla <sashukla@nvidia.com>
[maz: rewritten commit message]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/1603711447-11998-2-git-send-email-sashukla@nvidia.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit c4ad98e4b7 upstream.
KVM currently assumes that an instruction abort can never be a write.
This is in general true, except when the abort is triggered by
a S1PTW on instruction fetch that tries to update the S1 page tables
(to set AF, for example).
This can happen if the page tables have been paged out and brought
back in without seeing a direct write to them (they are thus marked
read only), and the fault handling code will make the PT executable(!)
instead of writable. The guest gets stuck forever.
In these conditions, the permission fault must be considered as
a write so that the Stage-1 update can take place. This is essentially
the I-side equivalent of the problem fixed by 60e21a0ef5 ("arm64: KVM:
Take S1 walks into account when determining S2 write faults").
Update kvm_is_write_fault() to return true on IABT+S1PTW, and introduce
kvm_vcpu_trap_is_exec_fault() that only return true when no faulting
on a S1 fault. Additionally, kvm_vcpu_dabt_iss1tw() is renamed to
kvm_vcpu_abt_iss1tw(), as the above makes it plain that it isn't
specific to data abort.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Will Deacon <will@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200915104218.1284701-2-maz@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 57bdb436ce ]
If we're going to fail out the vgic_add_lpi(), let's make sure the
allocated vgic_irq memory is also freed. Though it seems that both
cases are unlikely to fail.
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200414030349.625-3-yuzenghui@huawei.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 969ce8b526 ]
It's likely that the vcpu fails to handle all virtual interrupts if
userspace decides to destroy it, leaving the pending ones stay in the
ap_list. If the un-handled one is a LPI, its vgic_irq structure will
be eventually leaked because of an extra refcount increment in
vgic_queue_irq_unlock().
This was detected by kmemleak on almost every guest destroy, the
backtrace is as follows:
unreferenced object 0xffff80725aed5500 (size 128):
comm "CPU 5/KVM", pid 40711, jiffies 4298024754 (age 166366.512s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 08 01 a9 73 6d 80 ff ff ...........sm...
c8 61 ee a9 00 20 ff ff 28 1e 55 81 6c 80 ff ff .a... ..(.U.l...
backtrace:
[<000000004bcaa122>] kmem_cache_alloc_trace+0x2dc/0x418
[<0000000069c7dabb>] vgic_add_lpi+0x88/0x418
[<00000000bfefd5c5>] vgic_its_cmd_handle_mapi+0x4dc/0x588
[<00000000cf993975>] vgic_its_process_commands.part.5+0x484/0x1198
[<000000004bd3f8e3>] vgic_its_process_commands+0x50/0x80
[<00000000b9a65b2b>] vgic_mmio_write_its_cwriter+0xac/0x108
[<0000000009641ebb>] dispatch_mmio_write+0xd0/0x188
[<000000008f79d288>] __kvm_io_bus_write+0x134/0x240
[<00000000882f39ac>] kvm_io_bus_write+0xe0/0x150
[<0000000078197602>] io_mem_abort+0x484/0x7b8
[<0000000060954e3c>] kvm_handle_guest_abort+0x4cc/0xa58
[<00000000e0d0cd65>] handle_exit+0x24c/0x770
[<00000000b44a7fad>] kvm_arch_vcpu_ioctl_run+0x460/0x1988
[<0000000025fb897c>] kvm_vcpu_ioctl+0x4f8/0xee0
[<000000003271e317>] do_vfs_ioctl+0x160/0xcd8
[<00000000e7f39607>] ksys_ioctl+0x98/0xd8
Fix it by retiring all pending LPIs in the ap_list on the destroy path.
p.s. I can also reproduce it on a normal guest shutdown. It is because
userspace still send LPIs to vcpu (through KVM_SIGNAL_MSI ioctl) while
the guest is being shutdown and unable to handle it. A little strange
though and haven't dig further...
Reviewed-by: James Morse <james.morse@arm.com>
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
[maz: moved the distributor deallocation down to avoid an UAF splat]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200414030349.625-2-yuzenghui@huawei.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 0bda9498dd ]
In kvm_vgic_dist_init() called from kvm_vgic_map_resources(), if
dist->vgic_model is invalid, dist->spis will be freed without set
dist->spis = NULL. And in vgicv2 resources clean up path,
__kvm_vgic_destroy() will be called to free allocated resources.
And dist->spis will be freed again in clean up chain because we
forget to set dist->spis = NULL in kvm_vgic_dist_init() failed
path. So double free would happen.
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Link: https://lore.kernel.org/r/1574923128-19956-1-git-send-email-linmiaohe@huawei.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit f65886606c upstream.
when kmalloc() fails in kvm_io_bus_unregister_dev(), before removing
the bus, we should iterate over all other devices linked to it and call
kvm_iodevice_destructor() for them
Fixes: 90db10434b ("KVM: kvm_io_bus_unregister_dev() should never fail")
Cc: stable@vger.kernel.org
Reported-and-tested-by: syzbot+f196caa45793d6374707@syzkaller.appspotmail.com
Link: https://syzkaller.appspot.com/bug?extid=f196caa45793d6374707
Signed-off-by: Rustam Kovhaev <rkovhaev@gmail.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20200907185535.233114-1-rkovhaev@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3fb884ffe9 upstream.
For the obscure cases where PMD and PUD are the same size
(64kB pages with 42bit VA, for example, which results in only
two levels of page tables), we can't map anything as a PUD,
because there is... erm... no PUD to speak of. Everything is
either a PMD or a PTE.
So let's only try and map a PUD when its size is different from
that of a PMD.
Cc: stable@vger.kernel.org
Fixes: b8e0ba7c8b ("KVM: arm64: Add support for creating PUD hugepages at stage 2")
Reported-by: Gavin Shan <gshan@redhat.com>
Reported-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Alexandru Elisei <alexandru.elisei@arm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Tested-by: Gavin Shan <gshan@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Alexandru Elisei <alexandru.elisei@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b5331379bc upstream.
When an MMU notifier call results in unmapping a range that spans multiple
PGDs, we end up calling into cond_resched_lock() when crossing a PGD boundary,
since this avoids running into RCU stalls during VM teardown. Unfortunately,
if the VM is destroyed as a result of OOM, then blocking is not permitted
and the call to the scheduler triggers the following BUG():
| BUG: sleeping function called from invalid context at arch/arm64/kvm/mmu.c:394
| in_atomic(): 1, irqs_disabled(): 0, non_block: 1, pid: 36, name: oom_reaper
| INFO: lockdep is turned off.
| CPU: 3 PID: 36 Comm: oom_reaper Not tainted 5.8.0 #1
| Hardware name: QEMU QEMU Virtual Machine, BIOS 0.0.0 02/06/2015
| Call trace:
| dump_backtrace+0x0/0x284
| show_stack+0x1c/0x28
| dump_stack+0xf0/0x1a4
| ___might_sleep+0x2bc/0x2cc
| unmap_stage2_range+0x160/0x1ac
| kvm_unmap_hva_range+0x1a0/0x1c8
| kvm_mmu_notifier_invalidate_range_start+0x8c/0xf8
| __mmu_notifier_invalidate_range_start+0x218/0x31c
| mmu_notifier_invalidate_range_start_nonblock+0x78/0xb0
| __oom_reap_task_mm+0x128/0x268
| oom_reap_task+0xac/0x298
| oom_reaper+0x178/0x17c
| kthread+0x1e4/0x1fc
| ret_from_fork+0x10/0x30
Use the new 'flags' argument to kvm_unmap_hva_range() to ensure that we
only reschedule if MMU_NOTIFIER_RANGE_BLOCKABLE is set in the notifier
flags.
Cc: <stable@vger.kernel.org>
Fixes: 8b3405e345 ("kvm: arm/arm64: Fix locking for kvm_free_stage2_pgd")
Cc: Marc Zyngier <maz@kernel.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Message-Id: <20200811102725.7121-3-will@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit fdfe7cbd58 upstream.
The 'flags' field of 'struct mmu_notifier_range' is used to indicate
whether invalidate_range_{start,end}() are permitted to block. In the
case of kvm_mmu_notifier_invalidate_range_start(), this field is not
forwarded on to the architecture-specific implementation of
kvm_unmap_hva_range() and therefore the backend cannot sensibly decide
whether or not to block.
Add an extra 'flags' parameter to kvm_unmap_hva_range() so that
architectures are aware as to whether or not they are permitted to block.
Cc: <stable@vger.kernel.org>
Cc: Marc Zyngier <maz@kernel.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Message-Id: <20200811102725.7121-2-will@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b757b47a2f upstream.
If a stage-2 page-table contains an executable, read-only mapping at the
pte level (e.g. due to dirty logging being enabled), a subsequent write
fault to the same page which tries to install a larger block mapping
(e.g. due to dirty logging having been disabled) will erroneously inherit
the exec permission and consequently skip I-cache invalidation for the
rest of the block.
Ensure that exec permission is only inherited by write faults when the
new mapping is of the same size as the existing one. A subsequent
instruction abort will result in I-cache invalidation for the entire
block mapping.
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Tested-by: Quentin Perret <qperret@google.com>
Reviewed-by: Quentin Perret <qperret@google.com>
Cc: Marc Zyngier <maz@kernel.org>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200723101714.15873-1-will@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ef3e40a7ea upstream.
When using the PtrAuth feature in a guest, we need to save the host's
keys before allowing the guest to program them. For that, we dump
them in a per-CPU data structure (the so called host context).
But both call sites that do this are in preemptible context,
which may end up in disaster should the vcpu thread get preempted
before reentering the guest.
Instead, save the keys eagerly on each vcpu_load(). This has an
increased overhead, but is at least safe.
Cc: stable@vger.kernel.org
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 0370964dd3 upstream.
On a VHE system, the EL1 state is left in the CPU most of the time,
and only syncronized back to memory when vcpu_put() is called (most
of the time on preemption).
Which means that when injecting an exception, we'd better have a way
to either:
(1) write directly to the EL1 sysregs
(2) synchronize the state back to memory, and do the changes there
For an AArch64, we already do (1), so we are safe. Unfortunately,
doing the same thing for AArch32 would be pretty invasive. Instead,
we can easily implement (2) by calling the put/load architectural
backends, and keep preemption disabled. We can then reload the
state back into EL1.
Cc: stable@vger.kernel.org
Reported-by: James Morse <james.morse@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e649b3f018 upstream.
Commit b1394e745b ("KVM: x86: fix APIC page invalidation") tried
to fix inappropriate APIC page invalidation by re-introducing arch
specific kvm_arch_mmu_notifier_invalidate_range() and calling it from
kvm_mmu_notifier_invalidate_range_start. However, the patch left a
possible race where the VMCS APIC address cache is updated *before*
it is unmapped:
(Invalidator) kvm_mmu_notifier_invalidate_range_start()
(Invalidator) kvm_make_all_cpus_request(kvm, KVM_REQ_APIC_PAGE_RELOAD)
(KVM VCPU) vcpu_enter_guest()
(KVM VCPU) kvm_vcpu_reload_apic_access_page()
(Invalidator) actually unmap page
Because of the above race, there can be a mismatch between the
host physical address stored in the APIC_ACCESS_PAGE VMCS field and
the host physical address stored in the EPT entry for the APIC GPA
(0xfee0000). When this happens, the processor will not trap APIC
accesses, and will instead show the raw contents of the APIC-access page.
Because Windows OS periodically checks for unexpected modifications to
the LAPIC register, this will show up as a BSOD crash with BugCheck
CRITICAL_STRUCTURE_CORRUPTION (109) we are currently seeing in
https://bugzilla.redhat.com/show_bug.cgi?id=1751017.
The root cause of the issue is that kvm_arch_mmu_notifier_invalidate_range()
cannot guarantee that no additional references are taken to the pages in
the range before kvm_mmu_notifier_invalidate_range_end(). Fortunately,
this case is supported by the MMU notifier API, as documented in
include/linux/mmu_notifier.h:
* If the subsystem
* can't guarantee that no additional references are taken to
* the pages in the range, it has to implement the
* invalidate_range() notifier to remove any references taken
* after invalidate_range_start().
The fix therefore is to reload the APIC-access page field in the VMCS
from kvm_mmu_notifier_invalidate_range() instead of ..._range_start().
Cc: stable@vger.kernel.org
Fixes: b1394e745b ("KVM: x86: fix APIC page invalidation")
Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=197951
Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com>
Message-Id: <20200606042627.61070-1-eiichi.tsukata@nutanix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 9a50ebbffa ]
When a guest tries to read the active state of its interrupts,
we currently just return whatever state we have in memory. This
means that if such an interrupt lives in a List Register on another
CPU, we fail to obsertve the latest active state for this interrupt.
In order to remedy this, stop all the other vcpus so that they exit
and we can observe the most recent value for the state. This is
similar to what we are doing for the write side of the same
registers, and results in new MMIO handlers for userspace (which
do not need to stop the guest, as it is supposed to be stopped
already).
Reported-by: Julien Grall <julien@xen.org>
Reviewed-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 0225fd5e0a upstream.
In the unlikely event that a 32bit vcpu traps into the hypervisor
on an instruction that is located right at the end of the 32bit
range, the emulation of that instruction is going to increment
PC past the 32bit range. This isn't great, as userspace can then
observe this value and get a bit confused.
Conversly, userspace can do things like (in the context of a 64bit
guest that is capable of 32bit EL0) setting PSTATE to AArch64-EL0,
set PC to a 64bit value, change PSTATE to AArch32-USR, and observe
that PC hasn't been truncated. More confusion.
Fix both by:
- truncating PC increments for 32bit guests
- sanitizing all 32bit regs every time a core reg is changed by
userspace, and that PSTATE indicates a 32bit mode.
Cc: stable@vger.kernel.org
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1c32ca5dc6 upstream.
When deciding whether a guest has to be stopped we check whether this
is a private interrupt or not. Unfortunately, there's an off-by-one bug
here, and we fail to recognize a whole range of interrupts as being
global (GICv2 SPIs 32-63).
Fix the condition from > to be >=.
Cc: stable@vger.kernel.org
Fixes: abd7229626 ("KVM: arm/arm64: Simplify active_change_prepare and plug race")
Reported-by: André Przywara <andre.przywara@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit fcfbc61754 upstream.
When reading/writing using the guest/host cache, check for a bad hva
before checking for a NULL memslot, which triggers the slow path for
handing cross-page accesses. Because the memslot is nullified on error
by __kvm_gfn_to_hva_cache_init(), if the bad hva is encountered after
crossing into a new page, then the kvm_{read,write}_guest() slow path
could potentially write/access the first chunk prior to detecting the
bad hva.
Arguably, performing a partial access is semantically correct from an
architectural perspective, but that behavior is certainly not intended.
In the original implementation, memslot was not explicitly nullified
and therefore the partial access behavior varied based on whether the
memslot itself was null, or if the hva was simply bad. The current
behavior was introduced as a seemingly unintentional side effect in
commit f1b9dd5eb8 ("kvm: Disallow wraparound in
kvm_gfn_to_hva_cache_init"), which justified the change with "since some
callers don't check the return code from this function, it sit seems
prudent to clear ghc->memslot in the event of an error".
Regardless of intent, the partial access is dependent on _not_ checking
the result of the cache initialization, which is arguably a bug in its
own right, at best simply weird.
Fixes: 8f964525a1 ("KVM: Allow cross page reads and writes from cached translations.")
Cc: Jim Mattson <jmattson@google.com>
Cc: Andrew Honig <ahonig@google.com>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 4a267aa707 upstream.
According to the ARM ARM, registers CNT{P,V}_TVAL_EL0 have bits [63:32]
RES0 [1]. When reading the register, the value is truncated to the least
significant 32 bits [2], and on writes, TimerValue is treated as a signed
32-bit integer [1, 2].
When the guest behaves correctly and writes 32-bit values, treating TVAL
as an unsigned 64 bit register works as expected. However, things start
to break down when the guest writes larger values, because
(u64)0x1_ffff_ffff = 8589934591. but (s32)0x1_ffff_ffff = -1, and the
former will cause the timer interrupt to be asserted in the future, but
the latter will cause it to be asserted now. Let's treat TVAL as a
signed 32-bit register on writes, to match the behaviour described in
the architecture, and the behaviour experimentally exhibited by the
virtual timer on a non-vhe host.
[1] Arm DDI 0487E.a, section D13.8.18
[2] Arm DDI 0487E.a, section D11.2.4
Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
[maz: replaced the read-side mask with lower_32_bits]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Fixes: 8fa7616248 ("KVM: arm/arm64: arch_timer: Fix CNTP_TVAL calculation")
Link: https://lore.kernel.org/r/20200127103652.2326-1-alexandru.elisei@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit aa76829171 upstream.
At the moment a SW_INCR counter always overflows on 32-bit
boundary, independently on whether the n+1th counter is
programmed as CHAIN.
Check whether the SW_INCR counter is a 64b counter and if so,
implement the 64b logic.
Fixes: 80f393a23b ("KVM: arm/arm64: Support chained PMU counters")
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200124142535.29386-4-eric.auger@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3837407c1a upstream.
The specification says PMSWINC increments PMEVCNTR<n>_EL1 by 1
if PMEVCNTR<n>_EL0 is enabled and configured to count SW_INCR.
For PMEVCNTR<n>_EL0 to be enabled, we need both PMCNTENSET to
be set for the corresponding event counter but we also need
the PMCR.E bit to be set.
Fixes: 7a0adc7064 ("arm64: KVM: Add access handler for PMSWINC register")
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200124142535.29386-2-eric.auger@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 21aecdbd7f upstream.
KVM's inject_abt64() injects an external-abort into an aarch64 guest.
The KVM_CAP_ARM_INJECT_EXT_DABT is intended to do exactly this, but
for an aarch32 guest inject_abt32() injects an implementation-defined
exception, 'Lockdown fault'.
Change this to external abort. For non-LPAE we now get the documented:
| Unhandled fault: external abort on non-linefetch (0x008) at 0x9c800f00
and for LPAE:
| Unhandled fault: synchronous external abort (0x210) at 0x9c800f00
Fixes: 74a64a9816 ("KVM: arm/arm64: Unify 32bit fault injection")
Reported-by: Beata Michalska <beata.michalska@linaro.org>
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200121123356.203000-3-james.morse@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 018f22f95e upstream.
Beata reports that KVM_SET_VCPU_EVENTS doesn't inject the expected
exception to a non-LPAE aarch32 guest.
The host intends to inject DFSR.FS=0x14 "IMPLEMENTATION DEFINED fault
(Lockdown fault)", but the guest receives DFSR.FS=0x04 "Fault on
instruction cache maintenance". This fault is hooked by
do_translation_fault() since ARMv6, which goes on to silently 'handle'
the exception, and restart the faulting instruction.
It turns out, when TTBCR.EAE is clear DFSR is split, and FS[4] has
to shuffle up to DFSR[10].
As KVM only does this in one place, fix up the static values. We
now get the expected:
| Unhandled fault: lock abort (0x404) at 0x9c800f00
Fixes: 74a64a9816 ("KVM: arm/arm64: Unify 32bit fault injection")
Reported-by: Beata Michalska <beata.michalska@linaro.org>
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200121123356.203000-2-james.morse@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit cf2d23e0ba upstream.
kvm_test_age_hva() is called upon mmu_notifier_test_young(), but wrong
address range has been passed to handle_hva_to_gpa(). With the wrong
address range, no young bits will be checked in handle_hva_to_gpa().
It means zero is always returned from mmu_notifier_test_young().
This fixes the issue by passing correct address range to the underly
function handle_hva_to_gpa(), so that the hardware young (access) bit
will be visited.
Fixes: 35307b9a5f ("arm/arm64: KVM: Implement Stage-2 page aging")
Signed-off-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20200121055659.19560-1-gshan@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 8c58be3449 upstream.
Saving/restoring an unmapped collection is a valid scenario. For
example this happens if a MAPTI command was sent, featuring an
unmapped collection. At the moment the CTE fails to be restored.
Only compare against the number of online vcpus if the rdist
base is set.
Fixes: ea1ad53e1e ("KVM: arm64: vgic-its: Collection table save/restore")
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Zenghui Yu <yuzenghui@huawei.com>
Link: https://lore.kernel.org/r/20191213094237.19627-1-eric.auger@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 42cde48b2d ]
Avoid the "writable" check in __gfn_to_hva_many(), which will always fail
on read-only memslots due to gfn_to_hva() assuming writes. Functionally,
this allows x86 to create large mappings for read-only memslots that
are backed by HugeTLB mappings.
Note, the changelog for commit 05da45583d ("KVM: MMU: large page
support") states "If the largepage contains write-protected pages, a
large pte is not used.", but "write-protected" refers to pages that are
temporarily read-only, e.g. read-only memslots didn't even exist at the
time.
Fixes: 4d8b81abc4 ("KVM: introduce readonly memslot")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
[Redone using kvm_vcpu_gfn_to_memslot_prot. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f9b84e1922 ]
Use kvm_vcpu_gfn_to_hva() when retrieving the host page size so that the
correct set of memslots is used when handling x86 page faults in SMM.
Fixes: 54bf36aac5 ("KVM: x86: use vcpu-specific functions to read/write/translate GFNs")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 736c291c9f ]
Convert a plethora of parameters and variables in the MMU and page fault
flows from type gva_t to gpa_t to properly handle TDP on 32-bit KVM.
Thanks to PSE and PAE paging, 32-bit kernels can access 64-bit physical
addresses. When TDP is enabled, the fault address is a guest physical
address and thus can be a 64-bit value, even when both KVM and its guest
are using 32-bit virtual addressing, e.g. VMX's VMCS.GUEST_PHYSICAL is a
64-bit field, not a natural width field.
Using a gva_t for the fault address means KVM will incorrectly drop the
upper 32-bits of the GPA. Ditto for gva_to_gpa() when it is used to
translate L2 GPAs to L1 GPAs.
Opportunistically rename variables and parameters to better reflect the
dual address modes, e.g. use "cr2_or_gpa" for fault addresses and plain
"addr" instead of "vaddr" when the address may be either a GVA or an L2
GPA. Similarly, use "gpa" in the nonpaging_page_fault() flows to avoid
a confusing "gpa_t gva" declaration; this also sets the stage for a
future patch to combing nonpaging_page_fault() and tdp_page_fault() with
minimal churn.
Sprinkle in a few comments to document flows where an address is known
to be a GVA and thus can be safely truncated to a 32-bit value. Add
WARNs in kvm_handle_page_fault() and FNAME(gva_to_gpa_nested)() to help
document such cases and detect bugs.
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 917248144d upstream.
__kvm_map_gfn()'s call to gfn_to_pfn_memslot() is
* relatively expensive
* in certain cases (such as when done from atomic context) cannot be called
Stashing gfn-to-pfn mapping should help with both cases.
This is part of CVE-2019-3016.
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1eff70a9ab upstream.
kvm_vcpu_(un)map operates on gfns from any current address space.
In certain cases we want to make sure we are not mapping SMRAM
and for that we can use kvm_(un)map_gfn() that we are introducing
in this patch.
This is part of CVE-2019-3016.
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b6ae256afd upstream.
On AArch64 you can do a sign-extended load to either a 32-bit or 64-bit
register, and we should only sign extend the register up to the width of
the register as specified in the operation (by using the 32-bit Wn or
64-bit Xn register specifier).
As it turns out, the architecture provides this decoding information in
the SF ("Sixty-Four" -- how cute...) bit.
Let's take advantage of this with the usual 32-bit/64-bit header file
dance and do the right thing on AArch64 hosts.
Signed-off-by: Christoffer Dall <christoffer.dall@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20191212195055.5541-1-christoffer.dall@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1cfbb484de upstream.
Confusingly, there are three SPSR layouts that a kernel may need to deal
with:
(1) An AArch64 SPSR_ELx view of an AArch64 pstate
(2) An AArch64 SPSR_ELx view of an AArch32 pstate
(3) An AArch32 SPSR_* view of an AArch32 pstate
When the KVM AArch32 support code deals with SPSR_{EL2,HYP}, it's either
dealing with #2 or #3 consistently. On arm64 the PSR_AA32_* definitions
match the AArch64 SPSR_ELx view, and on arm the PSR_AA32_* definitions
match the AArch32 SPSR_* view.
However, when we inject an exception into an AArch32 guest, we have to
synthesize the AArch32 SPSR_* that the guest will see. Thus, an AArch64
host needs to synthesize layout #3 from layout #2.
This patch adds a new host_spsr_to_spsr32() helper for this, and makes
use of it in the KVM AArch32 support code. For arm64 we need to shuffle
the DIT bit around, and remove the SS bit, while for arm we can use the
value as-is.
I've open-coded the bit manipulation for now to avoid having to rework
the existing PSR_* definitions into PSR64_AA32_* and PSR32_AA32_*
definitions. I hope to perform a more thorough refactoring in future so
that we can handle pstate view manipulation more consistently across the
kernel tree.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Alexandru Elisei <alexandru.elisei@arm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200108134324.46500-4-mark.rutland@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3c2483f154 upstream.
When KVM injects an exception into a guest, it generates the CPSR value
from scratch, configuring CPSR.{M,A,I,T,E}, and setting all other
bits to zero.
This isn't correct, as the architecture specifies that some CPSR bits
are (conditionally) cleared or set upon an exception, and others are
unchanged from the original context.
This patch adds logic to match the architectural behaviour. To make this
simple to follow/audit/extend, documentation references are provided,
and bits are configured in order of their layout in SPSR_EL2. This
layout can be seen in the diagram on ARM DDI 0487E.a page C5-426.
Note that this code is used by both arm and arm64, and is intended to
fuction with the SPSR_EL2 and SPSR_HYP layouts.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Alexandru Elisei <alexandru.elisei@arm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200108134324.46500-3-mark.rutland@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 6d674e28f6 upstream.
A device mapping is normally always mapped at Stage-2, since there
is very little gain in having it faulted in.
Nonetheless, it is possible to end-up in a situation where the device
mapping has been removed from Stage-2 (userspace munmaped the VFIO
region, and the MMU notifier did its job), but present in a userspace
mapping (userpace has mapped it back at the same address). In such
a situation, the device mapping will be demand-paged as the guest
performs memory accesses.
This requires to be careful when dealing with mapping size, cache
management, and to handle potential execution of a device mapping.
Reported-by: Alexandru Elisei <alexandru.elisei@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Tested-by: Alexandru Elisei <alexandru.elisei@arm.com>
Reviewed-by: James Morse <james.morse@arm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20191211165651.7889-2-maz@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ca185b2609 upstream.
It's possible that two LPIs locate in the same "byte_offset" but target
two different vcpus, where their pending status are indicated by two
different pending tables. In such a scenario, using last_byte_offset
optimization will lead KVM relying on the wrong pending table entry.
Let us use last_ptr instead, which can be treated as a byte index into
a pending table and also, can be vcpu specific.
Fixes: 280771252c ("KVM: arm64: vgic-v3: KVM_DEV_ARM_VGIC_SAVE_PENDING_TABLES")
Cc: stable@vger.kernel.org
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Eric Auger <eric.auger@redhat.com>
Link: https://lore.kernel.org/r/20191029071919.177-4-yuzenghui@huawei.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add a comment explaining the rational behind having both
no_compat open and ioctl callbacks to fend off compat tasks.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
On a system without KVM_COMPAT, we prevent IOCTLs from being issued
by a compat task. Although this prevents most silly things from
happening, it can still confuse a 32bit userspace that is able
to open the kvm device (the qemu test suite seems to be pretty
mad with this behaviour).
Take a more radical approach and return a -ENODEV to the compat
task.
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
VT-d posted interrupts, DAX/ZONE_DEVICE,
module unload/reload.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iQEcBAABAgAGBQJdyrEsAAoJEL/70l94x66DIOkH/Asqrh4o4pwfRHWE+9rnM6PI
j8oFi7Q4eRXJnP4zEMnMbb6xD/BfSH1tWEcPcYgIxD/t0DFx8F92/xsETAJ/Qc5n
CWpmnhMkJqERlV+GSRuBqnheMo0CEH1Ab1QZKhh5U3//pK3OtGY9WyydJHWcquTh
bGh2pnxwVZOtIIEmclUUfKjyR2Fu8hJLnQwzWgYZ27UK7J2pLmiiTX0vwQG359Iq
sDn9ND33pCBW5e/D2mzccRjOJEvzwrumewM1sRDsoAYLJzUjg9+xD83vZDa1d7R6
gajCDFWVJbPoLvUY+DgsZBwMMlogElimJMT/Zft3ERbCsYJbFvcmwp4JzyxDxQ4=
=J6KN
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull kvm fixes from Paolo Bonzini:
"Fix unwinding of KVM_CREATE_VM failure, VT-d posted interrupts,
DAX/ZONE_DEVICE, and module unload/reload"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: MMU: Do not treat ZONE_DEVICE pages as being reserved
KVM: VMX: Introduce pi_is_pir_empty() helper
KVM: VMX: Do not change PID.NDST when loading a blocked vCPU
KVM: VMX: Consider PID.PIR to determine if vCPU has pending interrupts
KVM: VMX: Fix comment to specify PID.ON instead of PIR.ON
KVM: X86: Fix initialization of MSR lists
KVM: fix placement of refcount initialization
KVM: Fix NULL-ptr deref after kvm_create_vm fails
Explicitly exempt ZONE_DEVICE pages from kvm_is_reserved_pfn() and
instead manually handle ZONE_DEVICE on a case-by-case basis. For things
like page refcounts, KVM needs to treat ZONE_DEVICE pages like normal
pages, e.g. put pages grabbed via gup(). But for flows such as setting
A/D bits or shifting refcounts for transparent huge pages, KVM needs to
to avoid processing ZONE_DEVICE pages as the flows in question lack the
underlying machinery for proper handling of ZONE_DEVICE pages.
This fixes a hang reported by Adam Borowski[*] in dev_pagemap_cleanup()
when running a KVM guest backed with /dev/dax memory, as KVM straight up
doesn't put any references to ZONE_DEVICE pages acquired by gup().
Note, Dan Williams proposed an alternative solution of doing put_page()
on ZONE_DEVICE pages immediately after gup() in order to simplify the
auditing needed to ensure is_zone_device_page() is called if and only if
the backing device is pinned (via gup()). But that approach would break
kvm_vcpu_{un}map() as KVM requires the page to be pinned from map() 'til
unmap() when accessing guest memory, unlike KVM's secondary MMU, which
coordinates with mmu_notifier invalidations to avoid creating stale
page references, i.e. doesn't rely on pages being pinned.
[*] http://lkml.kernel.org/r/20190919115547.GA17963@angband.pl
Reported-by: Adam Borowski <kilobyte@angband.pl>
Analyzed-by: David Hildenbrand <david@redhat.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Cc: stable@vger.kernel.org
Fixes: 3565fce3a6 ("mm, x86: get_user_pages() for dax mappings")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reported by syzkaller:
=============================
WARNING: suspicious RCU usage
-----------------------------
./include/linux/kvm_host.h:536 suspicious rcu_dereference_check() usage!
other info that might help us debug this:
rcu_scheduler_active = 2, debug_locks = 1
no locks held by repro_11/12688.
stack backtrace:
Call Trace:
dump_stack+0x7d/0xc5
lockdep_rcu_suspicious+0x123/0x170
kvm_dev_ioctl+0x9a9/0x1260 [kvm]
do_vfs_ioctl+0x1a1/0xfb0
ksys_ioctl+0x6d/0x80
__x64_sys_ioctl+0x73/0xb0
do_syscall_64+0x108/0xaa0
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Commit a97b0e773e (kvm: call kvm_arch_destroy_vm if vm creation fails)
sets users_count to 1 before kvm_arch_init_vm(), however, if kvm_arch_init_vm()
fails, we need to decrease this count. By moving it earlier, we can push
the decrease to out_err_no_arch_destroy_vm without introducing yet another
error label.
syzkaller source: https://syzkaller.appspot.com/x/repro.c?x=15209b84e00000
Reported-by: syzbot+75475908cd0910f141ee@syzkaller.appspotmail.com
Fixes: a97b0e773e ("kvm: call kvm_arch_destroy_vm if vm creation fails")
Cc: Jim Mattson <jmattson@google.com>
Analyzed-by: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reported by syzkaller:
kasan: CONFIG_KASAN_INLINE enabled
kasan: GPF could be caused by NULL-ptr deref or user memory access
general protection fault: 0000 [#1] PREEMPT SMP KASAN
CPU: 0 PID: 14727 Comm: syz-executor.3 Not tainted 5.4.0-rc4+ #0
RIP: 0010:kvm_coalesced_mmio_init+0x5d/0x110 arch/x86/kvm/../../../virt/kvm/coalesced_mmio.c:121
Call Trace:
kvm_dev_ioctl_create_vm arch/x86/kvm/../../../virt/kvm/kvm_main.c:3446 [inline]
kvm_dev_ioctl+0x781/0x1490 arch/x86/kvm/../../../virt/kvm/kvm_main.c:3494
vfs_ioctl fs/ioctl.c:46 [inline]
file_ioctl fs/ioctl.c:509 [inline]
do_vfs_ioctl+0x196/0x1150 fs/ioctl.c:696
ksys_ioctl+0x62/0x90 fs/ioctl.c:713
__do_sys_ioctl fs/ioctl.c:720 [inline]
__se_sys_ioctl fs/ioctl.c:718 [inline]
__x64_sys_ioctl+0x6e/0xb0 fs/ioctl.c:718
do_syscall_64+0xca/0x5d0 arch/x86/entry/common.c:290
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Commit 9121923c45 ("kvm: Allocate memslots and buses before calling kvm_arch_init_vm")
moves memslots and buses allocations around, however, if kvm->srcu/irq_srcu fails
initialization, NULL will be returned instead of error code, NULL will not be intercepted
in kvm_dev_ioctl_create_vm() and be dereferenced by kvm_coalesced_mmio_init(), this patch
fixes it.
Moving the initialization is required anyway to avoid an incorrect synchronize_srcu that
was also reported by syzkaller:
wait_for_completion+0x29c/0x440 kernel/sched/completion.c:136
__synchronize_srcu+0x197/0x250 kernel/rcu/srcutree.c:921
synchronize_srcu_expedited kernel/rcu/srcutree.c:946 [inline]
synchronize_srcu+0x239/0x3e8 kernel/rcu/srcutree.c:997
kvm_page_track_unregister_notifier+0xe7/0x130 arch/x86/kvm/page_track.c:212
kvm_mmu_uninit_vm+0x1e/0x30 arch/x86/kvm/mmu.c:5828
kvm_arch_destroy_vm+0x4a2/0x5f0 arch/x86/kvm/x86.c:9579
kvm_create_vm arch/x86/kvm/../../../virt/kvm/kvm_main.c:702 [inline]
so do it.
Reported-by: syzbot+89a8060879fa0bd2db4f@syzkaller.appspotmail.com
Reported-by: syzbot+e27e7027eb2b80e44225@syzkaller.appspotmail.com
Fixes: 9121923c45 ("kvm: Allocate memslots and buses before calling kvm_arch_init_vm")
Cc: Jim Mattson <jmattson@google.com>
Cc: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The page table pages corresponding to broken down large pages are zapped in
FIFO order, so that the large page can potentially be recovered, if it is
not longer being used for execution. This removes the performance penalty
for walking deeper EPT page tables.
By default, one large page will last about one hour once the guest
reaches a steady state.
Signed-off-by: Junaid Shahid <junaids@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add a function to create a kernel thread associated with a given VM. In
particular, it ensures that the worker thread inherits the priority and
cgroups of the calling thread.
Signed-off-by: Junaid Shahid <junaids@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
