[PATCH v5 3/4] KVM: arm64: Register ptdump with debugfs on guest creation

[Sebastian Ene]

On Tue, Feb 13, 2024 at 12:56:20AM +0000, Oliver Upton wrote:
> On Wed, Feb 07, 2024 at 02:48:32PM +0000, Sebastian Ene wrote:

Hello Oliver,

> > While arch/*/mem/ptdump handles the kernel pagetable dumping code,
> > introduce KVM/ptdump which shows the guest stage-2 pagetables. The
> > separation is necessary because most of the definitions from the
> > stage-2 pagetable reside in the KVM path and we will be invoking
> > functionality **specific** to KVM.
> > 
> > When a guest is created, register a new file entry under the guest
> > debugfs dir which allows userspace to show the contents of the guest
> > stage-2 pagetables when accessed.
> > 
> > Signed-off-by: Sebastian Ene <sebastianene at google.com>
> > ---
> >  arch/arm64/kvm/Kconfig      | 13 ++++++
> >  arch/arm64/kvm/Makefile     |  1 +
> >  arch/arm64/kvm/debug.c      |  7 ++++
> >  arch/arm64/kvm/kvm_ptdump.h | 20 ++++++++++
> >  arch/arm64/kvm/ptdump.c     | 79 +++++++++++++++++++++++++++++++++++++
> >  5 files changed, 120 insertions(+)
> >  create mode 100644 arch/arm64/kvm/kvm_ptdump.h
> >  create mode 100644 arch/arm64/kvm/ptdump.c
> > 
> > diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
> > index 6c3c8ca73e7f..28097dd72174 100644
> > --- a/arch/arm64/kvm/Kconfig
> > +++ b/arch/arm64/kvm/Kconfig
> > @@ -68,4 +68,17 @@ config PROTECTED_NVHE_STACKTRACE
> >  
> >  	  If unsure, or not using protected nVHE (pKVM), say N.
> >  
> > +config PTDUMP_STAGE2_DEBUGFS
> > +       bool "Present the stage-2 pagetables to debugfs"
> > +       depends on PTDUMP_DEBUGFS && KVM
> > +       default n
> > +       help
> > +         Say Y here if you want to show the stage-2 kernel pagetables
> > +         layout in a debugfs file. This information is only useful for kernel developers
> > +         who are working in architecture specific areas of the kernel.
> > +         It is probably not a good idea to enable this feature in a production
> > +         kernel.
> > +
> > +         If in doubt, say N.
> > +
> >  endif # VIRTUALIZATION
> > diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
> > index c0c050e53157..190eac17538c 100644
> > --- a/arch/arm64/kvm/Makefile
> > +++ b/arch/arm64/kvm/Makefile
> > @@ -23,6 +23,7 @@ kvm-y += arm.o mmu.o mmio.o psci.o hypercalls.o pvtime.o \
> >  	 vgic/vgic-its.o vgic/vgic-debug.o
> >  
> >  kvm-$(CONFIG_HW_PERF_EVENTS)  += pmu-emul.o pmu.o
> > +kvm-$(CONFIG_PTDUMP_STAGE2_DEBUGFS) += ptdump.o
> >  
> >  always-y := hyp_constants.h hyp-constants.s
> >  
> > diff --git a/arch/arm64/kvm/debug.c b/arch/arm64/kvm/debug.c
> > index 8725291cb00a..aef52836cd90 100644
> > --- a/arch/arm64/kvm/debug.c
> > +++ b/arch/arm64/kvm/debug.c
> > @@ -14,6 +14,8 @@
> >  #include <asm/kvm_arm.h>
> >  #include <asm/kvm_emulate.h>
> >  
> > +#include <kvm_ptdump.h>
> > +
> >  #include "trace.h"
> >  
> >  /* These are the bits of MDSCR_EL1 we may manipulate */
> > @@ -342,3 +344,8 @@ void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu)
> >  	vcpu_clear_flag(vcpu, DEBUG_STATE_SAVE_SPE);
> >  	vcpu_clear_flag(vcpu, DEBUG_STATE_SAVE_TRBE);
> >  }
> > +
> > +int kvm_arch_create_vm_debugfs(struct kvm *kvm)
> > +{
> > +	return kvm_ptdump_guest_register(kvm);
> > +}
> > diff --git a/arch/arm64/kvm/kvm_ptdump.h b/arch/arm64/kvm/kvm_ptdump.h
> > new file mode 100644
> > index 000000000000..a7c00a28481b
> > --- /dev/null
> > +++ b/arch/arm64/kvm/kvm_ptdump.h
> > @@ -0,0 +1,20 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (C) Google, 2023
> > + * Author: Sebastian Ene <sebastianene at google.com>
> > + */
> > +
> > +#ifndef __KVM_PTDUMP_H
> > +#define __KVM_PTDUMP_H
> > +
> > +#include <linux/kvm_host.h>
> > +#include <asm/ptdump.h>
> > +
> > +
> > +#ifdef CONFIG_PTDUMP_STAGE2_DEBUGFS
> > +int kvm_ptdump_guest_register(struct kvm *kvm);
> > +#else
> > +static inline int kvm_ptdump_guest_register(struct kvm *kvm) { return 0; }
> > +#endif /* CONFIG_PTDUMP_STAGE2_DEBUGFS */
> > +
> > +#endif /* __KVM_PTDUMP_H */
> > diff --git a/arch/arm64/kvm/ptdump.c b/arch/arm64/kvm/ptdump.c
> > new file mode 100644
> > index 000000000000..a4e984da8aa7
> > --- /dev/null
> > +++ b/arch/arm64/kvm/ptdump.c
> > @@ -0,0 +1,79 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +//
> > +// Debug helper used to dump the stage-2 pagetables of the system and their
> > +// associated permissions.
> > +//
> > +// Copyright (C) Google, 2023
> > +// Author: Sebastian Ene <sebastianene at google.com>
> 
> Same comment as last time about ... the comment :)
> 
> Should be of the form
> 
> /*
>  *
>  */
> 

I forgot to update this, sorry. Let me fix it.

> > +#include <linux/debugfs.h>
> > +#include <linux/kvm_host.h>
> > +#include <linux/seq_file.h>
> > +
> > +#include <asm/kvm_pkvm.h>
> 
> is this needed?
>

We only need  <asm/kvm_pgtable.h> so I will update this.

> > +#include <kvm_ptdump.h>
> > +
> > +
> > +static int kvm_ptdump_guest_open(struct inode *inode, struct file *file);
> > +static int kvm_ptdump_guest_show(struct seq_file *m, void *);
> 
> can you structure the file in a way to avoid forward declarations?
> 

Sounds good, let me drop those.

> > +static const struct file_operations kvm_ptdump_guest_fops = {
> > +	.open		= kvm_ptdump_guest_open,
> > +	.read		= seq_read,
> > +	.llseek		= seq_lseek,
> > +	.release	= single_release,
> > +};
> > +
> > +static int kvm_ptdump_guest_open(struct inode *inode, struct file *file)
> > +{
> > +	return single_open(file, kvm_ptdump_guest_show, inode->i_private);
> > +}
> > +
> 
> Shouldn't we take a reference on the KVM struct at open to avoid UAF?
> 
> 	struct kvm *kvm = inode->i_private;
> 
> 	if (!kvm_get_kvm_safe(kvm))
> 		return -ENOENT;
> 
> Then you can do a put on it at close().
> 

Thanks, I though that the kvm_destroy_vm_debugfs will keep spinning if
there are opened paths to the debugfs entry, but I guess nothing prevents
that from happening and the kvm struct can be removed behind our back.

> > +static int kvm_ptdump_visitor(const struct kvm_pgtable_visit_ctx *ctx,
> > +			      enum kvm_pgtable_walk_flags visit)
> > +{
> > +	struct pg_state *st = ctx->arg;
> > +	struct ptdump_state *pt_st = &st->ptdump;
> > +
> > +	note_page(pt_st, ctx->addr, ctx->level, ctx->old);
> > +	return 0;
> > +}
> > +
> > +static int kvm_ptdump_show_common(struct seq_file *m,
> > +				  struct kvm_pgtable *pgtable,
> > +				  struct pg_state *parser_state)
> > +{
> > +	struct kvm_pgtable_walker walker = (struct kvm_pgtable_walker) {
> > +		.cb     = kvm_ptdump_visitor,
> > +		.arg	= parser_state,
> > +		.flags	= KVM_PGTABLE_WALK_LEAF,
> > +	};
> > +
> > +	return kvm_pgtable_walk(pgtable, 0, BIT(pgtable->ia_bits), &walker);
> > +}
> > +
> > +static int kvm_ptdump_guest_show(struct seq_file *m, void *)
> > +{
> > +	struct kvm *guest_kvm = m->private;
> > +	struct kvm_s2_mmu *mmu = &guest_kvm->arch.mmu;
> > +	struct pg_state parser_state = {0};
> > +	int ret;
> > +
> > +	write_lock(&guest_kvm->mmu_lock);
> > +	ret = kvm_ptdump_show_common(m, mmu->pgt, &parser_state);
> > +	write_unlock(&guest_kvm->mmu_lock);
> > +
> > +	return ret;
> > +}
> > +
> > +int kvm_ptdump_guest_register(struct kvm *kvm)
> > +{
> > +	struct dentry *parent;
> > +
> > +	parent = debugfs_create_file("stage2_page_tables", 0400,
> > +				     kvm->debugfs_dentry, kvm,
> > +				     &kvm_ptdump_guest_fops);
> > +	if (IS_ERR(parent))
> > +		return PTR_ERR(parent);
> 
> This makes the otherwise benign debugfs failure into something fatal for
> VM creation, no?
> 
> From the documentation on debugfs_create_file():
> 
>  * NOTE: it's expected that most callers should _ignore_ the errors returned
>  * by this function. Other debugfs functions handle the fact that the "dentry"
>  * passed to them could be an error and they don't crash in that case.
>  * Drivers should generally work fine even if debugfs fails to init anyway.
> 
> The fact that kvm_arch_create_vm_debugfs() has a return value is a bit
> of an anti-pattern to begin with.
>

Ack, I will ignore the retun code then.

> -- 
> Thanks,
> Oliver

Thanks,
Sebastian