On Thu, Apr 13, 2017 at 04:50:46PM +0100, Suzuki K. Poulose wrote: > On Thu, Apr 13, 2017 at 10:17:54AM +0100, Suzuki K Poulose wrote: > > On 12/04/17 19:43, Marc Zyngier wrote: > > > On 12/04/17 17:19, Andrey Konovalov wrote: > > > > > > Hi Andrey, > > > > > > > Apparently this wasn't fixed, I've got this report again on > > > > linux-next-c4e7b35a3 (Apr 11), which includes 8b3405e34 "kvm: > > > > arm/arm64: Fix locking for kvm_free_stage2_pgd". > > > > > > This looks like a different bug. > > > > > > > > > > > I now have a way to reproduce it, so I can test proposed patches. I > > > > don't have a simple C reproducer though. > > > > > > > > The bug happens when the following syzkaller program is executed: > > > > > > > > mmap(&(0x7f0000000000/0xc000)=nil, (0xc000), 0x3, 0x32, > > > > 0xffffffffffffffff, 0x0) > > > > unshare(0x400) > > > > perf_event_open(&(0x7f000002f000-0x78)={0x1, 0x78, 0x0, 0x0, 0x0, 0x0, > > > > 0x0, 0x6, 0x0, 0x0, 0xd34, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, > > > > 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, 0x0, 0xffffffff, > > > > 0xffffffffffffffff, 0x0) > > > > r0 = openat$kvm(0xffffffffffffff9c, > > > > &(0x7f000000c000-0x9)="2f6465762f6b766d00", 0x0, 0x0) > > > > ioctl$TIOCSBRK(0xffffffffffffffff, 0x5427) > > > > r1 = ioctl$KVM_CREATE_VM(r0, 0xae01, 0x0) > > > > syz_kvm_setup_cpu$arm64(r1, 0xffffffffffffffff, > > > > &(0x7f0000dc6000/0x18000)=nil, &(0x7f000000c000)=[{0x0, > > > > &(0x7f000000c000)="5ba3c16f533efbed09f8221253c73763327fadce2371813b45dd7f7982f84a873e4ae89a6c2bd1af83a6024c36a1ff518318", > > > > 0x32}], 0x1, 0x0, &(0x7f000000d000-0x10)=[@featur2={0x1, 0x3}], 0x1) > > > > > > Is that the only thing the program does? Or is there anything running in > > > parallel? > > > > > > > ================================================================== > > > > BUG: KASAN: use-after-free in arch_spin_is_locked > > > > include/linux/compiler.h:254 [inline] > > > > BUG: KASAN: use-after-free in unmap_stage2_range+0x990/0x9a8 > > > > arch/arm64/kvm/../../../arch/arm/kvm/mmu.c:295 > > > > Read of size 8 at addr ffff800004476730 by task syz-executor/13106 > > > > > > > > CPU: 1 PID: 13106 Comm: syz-executor Not tainted > > > > 4.11.0-rc6-next-20170411-xc2-11025-gc4e7b35a33d4-dirty #5 > > > > Hardware name: Hardkernel ODROID-C2 (DT) > > > > Call trace: > > > > [<ffff20000808fd08>] dump_backtrace+0x0/0x440 > > > > arch/arm64/kernel/traps.c:505 > > > > [<ffff2000080903c0>] show_stack+0x20/0x30 arch/arm64/kernel/traps.c:228 > > > > [<ffff2000088df030>] __dump_stack lib/dump_stack.c:16 [inline] > > > > [<ffff2000088df030>] dump_stack+0x110/0x168 lib/dump_stack.c:52 > > > > [<ffff200008406db8>] print_address_description+0x60/0x248 > > > > mm/kasan/report.c:252 > > > > [<ffff2000084072c8>] kasan_report_error mm/kasan/report.c:351 [inline] > > > > [<ffff2000084072c8>] kasan_report+0x218/0x300 mm/kasan/report.c:408 > > > > [<ffff200008407428>] __asan_report_load8_noabort+0x18/0x20 > > > > mm/kasan/report.c:429 > > > > [<ffff2000080db1b8>] arch_spin_is_locked include/linux/compiler.h:254 > > > > [inline] > > > > > > This is the assert on the spinlock, and the memory is gone. > > > > > > > [<ffff2000080db1b8>] unmap_stage2_range+0x990/0x9a8 > > > > arch/arm64/kvm/../../../arch/arm/kvm/mmu.c:295 > > > > [<ffff2000080db248>] kvm_free_stage2_pgd.part.16+0x30/0x98 > > > > arch/arm64/kvm/../../../arch/arm/kvm/mmu.c:842 > > > > [<ffff2000080ddfb8>] kvm_free_stage2_pgd > > > > arch/arm64/kvm/../../../arch/arm/kvm/mmu.c:838 [inline] > > > > > > But we've taken than lock here. There's only a handful of instructions > > > in between, and the memory can only go away if there is something > > > messing with us in parallel. > > > > > > > [<ffff2000080ddfb8>] kvm_arch_flush_shadow_all+0x40/0x58 > > > > arch/arm64/kvm/../../../arch/arm/kvm/mmu.c:1895 > > > > [<ffff2000080c379c>] kvm_mmu_notifier_release+0x154/0x1d0 > > > > arch/arm64/kvm/../../../virt/kvm/kvm_main.c:472 > > > > [<ffff2000083f2b60>] __mmu_notifier_release+0x1c0/0x3e0 > > > > mm/mmu_notifier.c:75 > > > > [<ffff2000083a1fb4>] mmu_notifier_release > > > > include/linux/mmu_notifier.h:235 [inline] > > > > [<ffff2000083a1fb4>] exit_mmap+0x21c/0x288 mm/mmap.c:2941 > > > > [<ffff20000810ecd4>] __mmput kernel/fork.c:888 [inline] > > > > [<ffff20000810ecd4>] mmput+0xdc/0x2e0 kernel/fork.c:910 > > > > [<ffff20000811fda8>] exit_mm kernel/exit.c:557 [inline] > > > > [<ffff20000811fda8>] do_exit+0x648/0x2020 kernel/exit.c:865 > > > > [<ffff2000081218b4>] do_group_exit+0xdc/0x260 kernel/exit.c:982 > > > > [<ffff20000813adf0>] get_signal+0x358/0xf58 kernel/signal.c:2318 > > > > [<ffff20000808de98>] do_signal+0x170/0xc10 > > > > arch/arm64/kernel/signal.c:370 > > > > [<ffff20000808edb4>] do_notify_resume+0xe4/0x120 > > > > arch/arm64/kernel/signal.c:421 > > > > [<ffff200008083e68>] work_pending+0x8/0x14 > > > > > > So we're being serviced with a signal. Do you know if this signal is > > > generated by your syzkaller program? We could be racing between do_exit > > > triggered by a fatal signal (this trace) and the closing of the two file > > > descriptors (vcpu and vm). > > > > > > Paolo: does this look possible to you? I can't see what locking we have > > > that could prevent this race. > > > > On a quick look, I see two issues: > > > > 1) It looks like the mmu_notifier->ops.release could be called twice for a > > notifier, > > from mmu_notifier_unregister() and exit_mmap()->mmu_notifier_release(), > > which is > > causing the problem as above. > > > > This could possibly be avoided by swapping the order of the following > > operations > > in themmu_notifier_unregister(): > > > > a) Invoke ops->release under src_read_lock() > > b) Delete the notifier from the list. > > > > which can prevent mmu_notifier_release() calling the ops->release() again, > > before > > we reach (b). > > > > > > 2) The core KVM code does an mmgrab()/mmdrop on the current->mm to pin the > > mm_struct. But > > this doesn't prevent the "real_address user space" from being destroyed. > > Since KVM > > actually depends on the user pages and page tables, it should > > really/also(?) use > > mmget()/mmput() (See Documentation/vm/active_mm.txt). I understand that > > mmget() shouldn't > > be used for pinning unbounded amount of time. But since we do it from > > within the same > > process context (like say threads), we should be safe to do so. >
Option 2 doesn't work, as it creates a circular dependency with exit_mmap vs kvm_destory_vm due to the mmap on VCPU (which prevents KVM from getting dropped). After a couple of trials at resolving this issue, we have something better to resolve it, below. ----8>---- kvm: Fix mmu_notifier release race The KVM uses mmu_notifier (wherever available) to keep track of the changes to the mm of the guest. The guest shadow page tables are released when the VM exits via mmu_notifier->ops.release(). There is a rare chance that the mmu_notifier->release could be called more than once via two different paths, which could end up in use-after-free of kvm instance. e.g: thread A thread B ------- -------------- get_signal-> kvm_destroy_vm()-> do_exit-> mmu_notifier_unregister-> exit_mm-> kvm_arch_flush_shadow_all()-> exit_mmap-> spin_lock(&kvm->mmu_lock) mmu_notifier_release-> .... kvm_arch_flush_shadow_all()-> ..... ... spin_lock(&kvm->mmu_lock) ..... spin_unlock(&kvm->mmu_lock) kvm_arch_free_kvm() *** use after free of kvm *** This patch attempts to solve the problem by holding a reference to the KVM for the mmu_notifier, which is dropped only from notifier->ops.release(). This will ensure that the KVM struct is available till we reach the kvm_mmu_notifier_release, and the kvm_destroy_vm is called only from/after it. So, we can unregister the notifier with no_release option and hence avoiding the race above. However, we need to make sure that the KVM is freed only after the mmu_notifier has finished processing the notifier due to the following possible path of execution : mmu_notifier_release -> kvm_mmu_notifier_release -> kvm_put_kvm -> kvm_destroy_vm -> kvm_arch_free_kvm Reported-by: andreyk...@google.com Cc: Mark Rutland <mark.rutl...@arm.com> Cc: Paolo Bonzin <pbonz...@redhat.com> Cc: Radim Krčmář <rkrc...@redhat.com> Cc: Marc Zyngier <marc.zyng...@arm.com> Cc: Christoffer Dall <christoffer.d...@linaro.org> Cc: andreyk...@google.com Cc: Marc Zyngier <marc.zyng...@arm.com> Tested-by: Mark Rutland <mark.rutl...@arm.com> Signed-off-by: Suzuki K Poulose <suzuki.poul...@arm.com> --- include/linux/kvm_host.h | 1 + virt/kvm/kvm_main.c | 59 ++++++++++++++++++++++++++++++++++++++++++------ 2 files changed, 53 insertions(+), 7 deletions(-) diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h index d025074..561e968 100644 --- a/include/linux/kvm_host.h +++ b/include/linux/kvm_host.h @@ -424,6 +424,7 @@ struct kvm { struct mmu_notifier mmu_notifier; unsigned long mmu_notifier_seq; long mmu_notifier_count; + struct rcu_head mmu_notifier_rcu; #endif long tlbs_dirty; struct list_head devices; diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index 88257b3..2c3fdd4 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -471,6 +471,7 @@ static void kvm_mmu_notifier_release(struct mmu_notifier *mn, idx = srcu_read_lock(&kvm->srcu); kvm_arch_flush_shadow_all(kvm); srcu_read_unlock(&kvm->srcu, idx); + kvm_put_kvm(kvm); } static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { @@ -486,8 +487,46 @@ static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { static int kvm_init_mmu_notifier(struct kvm *kvm) { + int rc; kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; - return mmu_notifier_register(&kvm->mmu_notifier, current->mm); + rc = mmu_notifier_register(&kvm->mmu_notifier, current->mm); + /* + * We hold a reference to KVM here to make sure that the KVM + * doesn't get free'd before ops->release() completes. + */ + if (!rc) + kvm_get_kvm(kvm); + return rc; +} + +static void kvm_free_vm_rcu(struct rcu_head *rcu) +{ + struct kvm *kvm = container_of(rcu, struct kvm, mmu_notifier_rcu); + kvm_arch_free_vm(kvm); +} + +static void kvm_flush_shadow_mmu(struct kvm *kvm) +{ + /* + * We hold a reference to kvm instance for mmu_notifier and is + * only released when ops->release() is called via exit_mmap path. + * So, when we reach here ops->release() has been called already, which + * flushes the shadow page tables. Hence there is no need to call the + * release() again when we unregister the notifier. However, we need + * to delay freeing up the kvm until the release() completes, since + * we could reach here via : + * kvm_mmu_notifier_release() -> kvm_put_kvm() -> kvm_destroy_vm() + */ + mmu_notifier_unregister_no_release(&kvm->mmu_notifier, kvm->mm); +} + +static void kvm_free_vm(struct kvm *kvm) +{ + /* + * Wait until the mmu_notifier has finished the release(). + * See comments above in kvm_flush_shadow_mmu. + */ + mmu_notifier_call_srcu(&kvm->mmu_notifier_rcu, kvm_free_vm_rcu); } #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ @@ -497,6 +536,16 @@ static int kvm_init_mmu_notifier(struct kvm *kvm) return 0; } +static void kvm_flush_shadow_mmu(struct kvm *kvm) +{ + kvm_arch_flush_shadow_all(kvm); +} + +static void kvm_free_vm(struct kvm *kvm) +{ + kvm_arch_free_vm(kvm); +} + #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ static struct kvm_memslots *kvm_alloc_memslots(void) @@ -733,18 +782,14 @@ static void kvm_destroy_vm(struct kvm *kvm) kvm->buses[i] = NULL; } kvm_coalesced_mmio_free(kvm); -#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) - mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); -#else - kvm_arch_flush_shadow_all(kvm); -#endif + kvm_flush_shadow_mmu(kvm); kvm_arch_destroy_vm(kvm); kvm_destroy_devices(kvm); for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) kvm_free_memslots(kvm, kvm->memslots[i]); cleanup_srcu_struct(&kvm->irq_srcu); cleanup_srcu_struct(&kvm->srcu); - kvm_arch_free_vm(kvm); + kvm_free_vm(kvm); preempt_notifier_dec(); hardware_disable_all(); mmdrop(mm); -- 2.7.4 _______________________________________________ kvmarm mailing list kvmarm@lists.cs.columbia.edu https://lists.cs.columbia.edu/mailman/listinfo/kvmarm