From: james.morse@arm.com (James Morse)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH v5 22/23] arm64: KVM: Allow mapping of vectors outside of the RAM region
Date: Wed, 14 Mar 2018 11:40:59 +0000 [thread overview]
Message-ID: <5AA90A4B.2020900@arm.com> (raw)
In-Reply-To: <20180301155538.26860-23-marc.zyngier@arm.com>
Hi Marc,
On 01/03/18 15:55, Marc Zyngier wrote:
> We're now ready to map our vectors in weird and wonderful locations.
> On enabling ARM64_HARDEN_EL2_VECTORS, a vector slots gets allocated
> if this hasn't been already done via ARM64_HARDEN_BRANCH_PREDICTOR
> and gets mapped outside of the normal RAM region, next to the
> idmap.
>
> That way, being able to obtain VBAR_EL2 doesn't reveal the mapping
> of the rest of the hypervisor code.
> diff --git a/Documentation/arm64/memory.txt b/Documentation/arm64/memory.txt
> index c58cc5dbe667..c5dab30d3389 100644
> --- a/Documentation/arm64/memory.txt
> +++ b/Documentation/arm64/memory.txt
> @@ -90,7 +90,8 @@ When using KVM without the Virtualization Host Extensions, the
> hypervisor maps kernel pages in EL2 at a fixed (and potentially
> random) offset from the linear mapping. See the kern_hyp_va macro and
> kvm_update_va_mask function for more details. MMIO devices such as
> -GICv2 gets mapped next to the HYP idmap page.
> +GICv2 gets mapped next to the HYP idmap page, as do vectors when
> +ARM64_HARDEN_EL2_VECTORS is selected for particular CPUs.
>
> When using KVM with the Virtualization Host Extensions, no additional
> mappings are created, since the host kernel runs directly in EL2.
> diff --git a/arch/arm64/include/asm/kvm_mmu.h b/arch/arm64/include/asm/kvm_mmu.h
> index 3da9e5aea936..433d13d0c271 100644
> --- a/arch/arm64/include/asm/kvm_mmu.h
> +++ b/arch/arm64/include/asm/kvm_mmu.h
> @@ -360,33 +360,90 @@ static inline unsigned int kvm_get_vmid_bits(void)
> return (cpuid_feature_extract_unsigned_field(reg, ID_AA64MMFR1_VMIDBITS_SHIFT) == 2) ? 16 : 8;
> }
>
> -#ifdef CONFIG_HARDEN_BRANCH_PREDICTOR
> +#if (defined(CONFIG_HARDEN_BRANCH_PREDICTOR) || \
> + defined(CONFIG_HARDEN_EL2_VECTORS))
> +/*
> + * EL2 vectors can be mapped and rerouted in a number of ways,
> + * depending on the kernel configuration and CPU present:
> + *
> + * - If the CPU has the ARM64_HARDEN_BRANCH_PREDICTOR cap, the
> + * hardening sequence is placed in one of the vector slots, which is
> + * executed before jumping to the real vectors.
> + *
> + * - If the CPU has both the ARM64_HARDEN_EL2_VECTORS and BP
> + * hardening, the slot containing the hardening sequence is mapped
> + * next to the idmap page, and executed before jumping to the real
> + * vectors.
> + *
> + * - If the CPU only has ARM64_HARDEN_EL2_VECTORS, then an empty slot
> + * is selected, mapped next to the idmap page, and executed before
> + * jumping to the real vectors.
> + * Note that ARM64_HARDEN_EL2_VECTORS is somewhat incompatible with
> + * VHE, as we don't have hypervisor-specific mappings. If the system
> + * is VHE and yet selects this capability, it will be ignored.
Silently? This isn't a problem as the CPUs you enable this for don't have VHE.
Is it worth a warning? If we did ever need to support it, we can pull the same
trick the arch code uses, using a fixmap entry for the vectors.
> + */
> #include <asm/mmu.h>
>
> +extern void *__kvm_bp_vect_base;
> +extern int __kvm_harden_el2_vector_slot;
> +
> static inline void *kvm_get_hyp_vector(void)
> {
> struct bp_hardening_data *data = arm64_get_bp_hardening_data();
> - void *vect = kvm_ksym_ref(__kvm_hyp_vector);
> + int slot = -1;
> +
> + if (cpus_have_const_cap(ARM64_HARDEN_BRANCH_PREDICTOR) && data->fn)
> + slot = data->hyp_vectors_slot;
> +
> + if (cpus_have_const_cap(ARM64_HARDEN_EL2_VECTORS) &&
> + !has_vhe() && slot == -1)
> + slot = __kvm_harden_el2_vector_slot;
>
> - if (data->fn) {
> - vect = __bp_harden_hyp_vecs_start +
> - data->hyp_vectors_slot * SZ_2K;
> + if (slot != -1) {
> + void *vect;
>
> if (!has_vhe())
> - vect = lm_alias(vect);
> + vect = __kvm_bp_vect_base;
> + else
> + vect = __bp_harden_hyp_vecs_start;
> + vect += slot * SZ_2K;
> +
> + return vect;
> }
>
> - vect = kern_hyp_va(vect);
> - return vect;
> + return kern_hyp_va(kvm_ksym_ref(__kvm_hyp_vector));
> }
>
> +/* This is only called on a !VHE system */
> static inline int kvm_map_vectors(void)
> {
> - return create_hyp_mappings(kvm_ksym_ref(__bp_harden_hyp_vecs_start),
> - kvm_ksym_ref(__bp_harden_hyp_vecs_end),
> - PAGE_HYP_EXEC);
> -}
> + phys_addr_t vect_pa = virt_to_phys(__bp_harden_hyp_vecs_start);
> + unsigned long size = __bp_harden_hyp_vecs_end - __bp_harden_hyp_vecs_start;
> +
> + if (cpus_have_const_cap(ARM64_HARDEN_BRANCH_PREDICTOR)) {
> + int ret;
> +
> + ret = create_hyp_mappings(kvm_ksym_ref(__bp_harden_hyp_vecs_start),
> + kvm_ksym_ref(__bp_harden_hyp_vecs_end),
> + PAGE_HYP_EXEC);
We don't have to do this for the regular vectors, as they are part of the
__hyp_text. How come these aren't?
The existing Makefile depends on KVM to build these. How come it isn't under
arch/arm64/kvm?
> + if (ret)
> + return ret;
> +
> + __kvm_bp_vect_base = kvm_ksym_ref(__bp_harden_hyp_vecs_start);
> + __kvm_bp_vect_base = kern_hyp_va(__kvm_bp_vect_base);
> + }
> +
> + if (cpus_have_const_cap(ARM64_HARDEN_EL2_VECTORS)) {
> + __kvm_harden_el2_vector_slot = atomic_inc_return(&arm64_el2_vector_last_slot);
> + BUG_ON(__kvm_harden_el2_vector_slot >= BP_HARDEN_EL2_SLOTS);
> + return create_hyp_exec_mappings(vect_pa, size,
> + &__kvm_bp_vect_base);
> + }
>
> + return 0;
> +}
> #else
> static inline void *kvm_get_hyp_vector(void)
> {
> diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile
> index b87541360f43..e7fc471c91a6 100644
> --- a/arch/arm64/kernel/Makefile
> +++ b/arch/arm64/kernel/Makefile
> @@ -54,8 +54,8 @@ arm64-reloc-test-y := reloc_test_core.o reloc_test_syms.o
> arm64-obj-$(CONFIG_CRASH_DUMP) += crash_dump.o
> arm64-obj-$(CONFIG_ARM_SDE_INTERFACE) += sdei.o
>
> -ifeq ($(CONFIG_KVM),y)
> -arm64-obj-$(CONFIG_HARDEN_BRANCH_PREDICTOR) += bpi.o
> +ifneq ($(filter y,$(CONFIG_HARDEN_BRANCH_PREDICTOR) $(CONFIG_HARDEN_EL2_VECTORS)),)
> +arm64-obj-$(CONFIG_KVM) += bpi.o
> endif
Isn't Kconfig 'select'ing a hidden-option the usual way this is done?
Thanks,
James
next prev parent reply other threads:[~2018-03-14 11:40 UTC|newest]
Thread overview: 50+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-03-01 15:55 [PATCH v5 00/23] KVM/arm64: Randomise EL2 mappings (variant 3a mitigation) Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 01/23] arm64: alternatives: Add dynamic patching feature Marc Zyngier
2018-03-07 18:09 ` Catalin Marinas
2018-03-01 15:55 ` [PATCH v5 02/23] arm64: insn: Add N immediate encoding Marc Zyngier
2018-03-07 18:09 ` Catalin Marinas
2018-03-01 15:55 ` [PATCH v5 03/23] arm64: insn: Add encoder for bitwise operations using literals Marc Zyngier
2018-03-07 18:10 ` Catalin Marinas
2018-03-12 14:44 ` Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 04/23] arm64: KVM: Dynamically patch the kernel/hyp VA mask Marc Zyngier
2018-03-07 18:10 ` Catalin Marinas
2018-03-01 15:55 ` [PATCH v5 05/23] arm64: cpufeatures: Drop the ARM64_HYP_OFFSET_LOW feature flag Marc Zyngier
2018-03-07 18:11 ` Catalin Marinas
2018-03-13 8:44 ` Suzuki K Poulose
2018-03-01 15:55 ` [PATCH v5 06/23] KVM: arm/arm64: Do not use kern_hyp_va() with kvm_vgic_global_state Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 07/23] KVM: arm/arm64: Demote HYP VA range display to being a debug feature Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 08/23] KVM: arm/arm64: Move ioremap calls to create_hyp_io_mappings Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 09/23] KVM: arm/arm64: Keep GICv2 HYP VAs in kvm_vgic_global_state Marc Zyngier
2018-03-13 9:35 ` Suzuki K Poulose
2018-03-13 11:40 ` Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 10/23] KVM: arm/arm64: Move HYP IO VAs to the "idmap" range Marc Zyngier
2018-03-09 18:59 ` James Morse
2018-03-12 14:02 ` Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 11/23] arm64; insn: Add encoder for the EXTR instruction Marc Zyngier
2018-03-07 18:12 ` Catalin Marinas
2018-03-01 15:55 ` [PATCH v5 12/23] arm64: insn: Allow ADD/SUB (immediate) with LSL #12 Marc Zyngier
2018-03-07 18:13 ` Catalin Marinas
2018-03-01 15:55 ` [PATCH v5 13/23] arm64: KVM: Dynamically compute the HYP VA mask Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 14/23] arm64: KVM: Introduce EL2 VA randomisation Marc Zyngier
2018-03-13 11:31 ` James Morse
2018-03-13 11:48 ` James Morse
2018-03-01 15:55 ` [PATCH v5 15/23] arm64: Update the KVM memory map documentation Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 16/23] arm64: KVM: Move vector offsetting from hyp-init.S to kvm_get_hyp_vector Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 17/23] arm64: KVM: Move stashing of x0/x1 into the vector code itself Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 18/23] arm64: KVM: Add epilogue branching to the vector code Marc Zyngier
2018-03-08 13:59 ` Catalin Marinas
2018-03-01 15:55 ` [PATCH v5 19/23] arm64: KVM: Allow far branches from vector slots to the main vectors Marc Zyngier
2018-03-08 13:59 ` Catalin Marinas
2018-03-12 18:27 ` James Morse
2018-03-12 19:43 ` Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 20/23] arm/arm64: KVM: Introduce EL2-specific executable mappings Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 21/23] arm64: Make BP hardening slot counter available Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 22/23] arm64: KVM: Allow mapping of vectors outside of the RAM region Marc Zyngier
2018-03-08 17:54 ` Andrew Jones
2018-03-13 10:30 ` Marc Zyngier
2018-03-13 11:14 ` Andrew Jones
2018-03-09 18:59 ` James Morse
2018-03-12 14:23 ` Marc Zyngier
2018-03-14 11:40 ` James Morse [this message]
2018-03-14 12:02 ` Marc Zyngier
2018-03-01 15:55 ` [PATCH v5 23/23] arm64: Enable ARM64_HARDEN_EL2_VECTORS on Cortex-A57 and A72 Marc Zyngier
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5AA90A4B.2020900@arm.com \
--to=james.morse@arm.com \
--cc=linux-arm-kernel@lists.infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).