From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 3F8E63009EE; Wed, 25 Mar 2026 09:03:50 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1774429430; cv=none; b=F3O5MD4B05+3UWs6Fky2xbC5XdkySeLqz8CyddLJF+xdk+Q0D7+/sFu4zfLvkzGlg6pP6qIpCtGyVC3l0oxqzGrBJHRa7m+WjMEUvz1UXCTGsP4ZSbK+lKBzdSzGMiKBbZ61Y1RCwTx/bcV6BCjhmSesafKdDPm0zxzNGDnVTYU= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1774429430; c=relaxed/simple; bh=w/GTjS8YC8uH+Sy4A1EDV8UGyt/H8PZtAnk1cljNo8w=; h=Date:Message-ID:From:To:Cc:Subject:In-Reply-To:References: MIME-Version:Content-Type; b=V8m2+w1/2n/gLPP7U78UyRjRHp1p7mkDRl3UkWewqrJs9bc5DtbFXbA23Dttat/lXZByS+Rxl9AX8hP2fku8ezIaNlhJAaAVv/8L6Eu8QJqd77vfefMjNpECMAmQevVoy2PVmoWwtrwsHOb6ais2oLLocpwLB12w1Rka8x2iZTU= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b=fLn3NICR; arc=none smtp.client-ip=10.30.226.201 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b="fLn3NICR" Received: by smtp.kernel.org (Postfix) with ESMTPSA id C50B2C4CEF7; Wed, 25 Mar 2026 09:03:49 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1774429430; bh=w/GTjS8YC8uH+Sy4A1EDV8UGyt/H8PZtAnk1cljNo8w=; h=Date:From:To:Cc:Subject:In-Reply-To:References:From; b=fLn3NICRRfLAwA8fdLwt9a3S1jWOlTsyhgvm9TKxtZQRkOTPn8eHM0Xie9VDcCkZ/ 8GJrVhP3m4gxmWaHeYlqNHI+OKhIvY/oJIVcpXXModAzVFaBKxCcb5IdkMSHsXYcg8 XO313s7oi20EAwUjxewUYW3Wta5AzC082RHyrjwFrm+YIr98GK8PK9KLX9+mDX3XC4 vGyt6W2oHGh8pDpJDbHCyrvtvQTgvJx8WHlZqYIYMiRGk3+fPTmKI2yMJ3yNGO6Oj+ NzaBbOgn605Hf2tAPp7Y+JBWc7/JLkhdGfhSc8tn0x/LKgFlIWJt4A7uE6w4llufPk RTldLE+GWj4pg== Received: from sofa.misterjones.org ([185.219.108.64] helo=goblin-girl.misterjones.org) by disco-boy.misterjones.org with esmtpsa (TLS1.3) tls TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 (Exim 4.98.2) (envelope-from ) id 1w5K9T-00000005Trk-21XU; Wed, 25 Mar 2026 09:03:47 +0000 Date: Wed, 25 Mar 2026 09:03:47 +0000 Message-ID: <86o6kc46x8.wl-maz@kernel.org> From: Marc Zyngier To: Wei-Lin Chang Cc: kvm@vger.kernel.org, linux-kselftest@vger.kernel.org, linux-arm-kernel@lists.infradead.org, kvmarm@lists.linux.dev, linux-kernel@vger.kernel.org, Paolo Bonzini , Shuah Khan , Oliver Upton , Joey Gouly , Suzuki K Poulose , Zenghui Yu , Catalin Marinas , Will Deacon Subject: Re: [PATCH 1/3] KVM: arm64: selftests: Add library functions for NV In-Reply-To: <20260325003620.2214766-2-weilin.chang@arm.com> References: <20260325003620.2214766-1-weilin.chang@arm.com> <20260325003620.2214766-2-weilin.chang@arm.com> User-Agent: Wanderlust/2.15.9 (Almost Unreal) SEMI-EPG/1.14.7 (Harue) FLIM-LB/1.14.9 (=?UTF-8?B?R29qxY0=?=) APEL-LB/10.8 EasyPG/1.0.0 Emacs/30.1 (aarch64-unknown-linux-gnu) MULE/6.0 (HANACHIRUSATO) Precedence: bulk X-Mailing-List: kvm@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 (generated by SEMI-EPG 1.14.7 - "Harue") Content-Type: text/plain; charset=US-ASCII X-SA-Exim-Connect-IP: 185.219.108.64 X-SA-Exim-Rcpt-To: weilin.chang@arm.com, kvm@vger.kernel.org, linux-kselftest@vger.kernel.org, linux-arm-kernel@lists.infradead.org, kvmarm@lists.linux.dev, linux-kernel@vger.kernel.org, pbonzini@redhat.com, shuah@kernel.org, oupton@kernel.org, joey.gouly@arm.com, suzuki.poulose@arm.com, yuzenghui@huawei.com, catalin.marinas@arm.com, will@kernel.org X-SA-Exim-Mail-From: maz@kernel.org X-SA-Exim-Scanned: No (on disco-boy.misterjones.org); SAEximRunCond expanded to false On Wed, 25 Mar 2026 00:36:18 +0000, Wei-Lin Chang wrote: > > The API is designed for userspace to first call prepare_{l2_stack, > hyp_state, eret_destination, nested_sync_handler}, with a function > supplied to prepare_eret_destination() to be run in L2. Then run_l2() > can be called in L1 to run the given function in L2. > > Signed-off-by: Wei-Lin Chang > --- > tools/testing/selftests/kvm/Makefile.kvm | 2 + > .../selftests/kvm/include/arm64/nested.h | 18 ++++++ > .../testing/selftests/kvm/lib/arm64/nested.c | 61 +++++++++++++++++++ > .../selftests/kvm/lib/arm64/nested_asm.S | 35 +++++++++++ > 4 files changed, 116 insertions(+) > create mode 100644 tools/testing/selftests/kvm/include/arm64/nested.h > create mode 100644 tools/testing/selftests/kvm/lib/arm64/nested.c > create mode 100644 tools/testing/selftests/kvm/lib/arm64/nested_asm.S > > diff --git a/tools/testing/selftests/kvm/Makefile.kvm b/tools/testing/selftests/kvm/Makefile.kvm > index 98da9fa4b8b7..5e681e8e0cd7 100644 > --- a/tools/testing/selftests/kvm/Makefile.kvm > +++ b/tools/testing/selftests/kvm/Makefile.kvm > @@ -34,6 +34,8 @@ LIBKVM_arm64 += lib/arm64/gic.c > LIBKVM_arm64 += lib/arm64/gic_v3.c > LIBKVM_arm64 += lib/arm64/gic_v3_its.c > LIBKVM_arm64 += lib/arm64/handlers.S > +LIBKVM_arm64 += lib/arm64/nested.c > +LIBKVM_arm64 += lib/arm64/nested_asm.S > LIBKVM_arm64 += lib/arm64/processor.c > LIBKVM_arm64 += lib/arm64/spinlock.c > LIBKVM_arm64 += lib/arm64/ucall.c > diff --git a/tools/testing/selftests/kvm/include/arm64/nested.h b/tools/testing/selftests/kvm/include/arm64/nested.h > new file mode 100644 > index 000000000000..739ff2ee0161 > --- /dev/null > +++ b/tools/testing/selftests/kvm/include/arm64/nested.h > @@ -0,0 +1,18 @@ > +/* SPDX-License-Identifier: GPL-2.0-only */ > +/* > + * ARM64 Nested virtualization defines > + */ > + > +#ifndef SELFTEST_KVM_NESTED_H > +#define SELFTEST_KVM_NESTED_H > + > +void prepare_l2_stack(struct kvm_vm *vm, struct kvm_vcpu *vcpu); > +void prepare_hyp_state(struct kvm_vm *vm, struct kvm_vcpu *vcpu); > +void prepare_eret_destination(struct kvm_vm *vm, struct kvm_vcpu *vcpu, void *l2_pc); > +void prepare_nested_sync_handler(struct kvm_vm *vm, struct kvm_vcpu *vcpu); > + > +void run_l2(void); > +void after_hvc(void); > +void do_hvc(void); > + > +#endif /* SELFTEST_KVM_NESTED_H */ > diff --git a/tools/testing/selftests/kvm/lib/arm64/nested.c b/tools/testing/selftests/kvm/lib/arm64/nested.c > new file mode 100644 > index 000000000000..111d02f44cfe > --- /dev/null > +++ b/tools/testing/selftests/kvm/lib/arm64/nested.c > @@ -0,0 +1,61 @@ > +// SPDX-License-Identifier: GPL-2.0 > +/* > + * ARM64 Nested virtualization helpers > + */ > + > +#include "kvm_util.h" > +#include "nested.h" > +#include "processor.h" > +#include "test_util.h" > + > +#include > + > +static void hvc_handler(struct ex_regs *regs) > +{ > + GUEST_ASSERT_EQ(get_current_el(), 2); > + GUEST_PRINTF("hvc handler\n"); > + regs->pstate = PSR_MODE_EL2h | PSR_D_BIT | PSR_A_BIT | PSR_I_BIT | PSR_F_BIT; > + regs->pc = (u64)after_hvc; > +} > + > +void prepare_l2_stack(struct kvm_vm *vm, struct kvm_vcpu *vcpu) > +{ > + size_t l2_stack_size; > + uint64_t l2_stack_paddr; > + > + l2_stack_size = vm->page_size == 4096 ? DEFAULT_STACK_PGS * vm->page_size : > + vm->page_size; Please use symbolic constants. Also, this looks wrong if the default stack size is 32k and the page size is 16k. You probably want to express a stack size directly, rather than a number of pages. > + l2_stack_paddr = __vm_phy_pages_alloc(vm, l2_stack_size / vm->page_size, > + 0, 0, false); > + vcpu_set_reg(vcpu, ARM64_CORE_REG(sp_el1), l2_stack_paddr + l2_stack_size); > +} > + > +void prepare_hyp_state(struct kvm_vm *vm, struct kvm_vcpu *vcpu) > +{ > + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_HCR_EL2), HCR_EL2_RW); Surely the E2H value matters. Or are you planning to only run this on configuration that hardcode E2H==0? That'd be pretty limiting. > +} > + > +void prepare_eret_destination(struct kvm_vm *vm, struct kvm_vcpu *vcpu, void *l2_pc) > +{ > + vm_paddr_t do_hvc_paddr = addr_gva2gpa(vm, (vm_vaddr_t)do_hvc); > + vm_paddr_t l2_pc_paddr = addr_gva2gpa(vm, (vm_vaddr_t)l2_pc); > + > + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_SPSR_EL2), PSR_MODE_EL1h | > + PSR_D_BIT | > + PSR_A_BIT | > + PSR_I_BIT | > + PSR_F_BIT); > + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_ELR_EL2), l2_pc_paddr); > + /* HACK: use TPIDR_EL2 to pass address, see run_l2() in nested_asm.S */ > + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_TPIDR_EL2), do_hvc_paddr); > +} > + > +void prepare_nested_sync_handler(struct kvm_vm *vm, struct kvm_vcpu *vcpu) > +{ > + if (!vm->handlers) { > + vm_init_descriptor_tables(vm); > + vcpu_init_descriptor_tables(vcpu); > + } > + vm_install_sync_handler(vm, VECTOR_SYNC_LOWER_64, > + ESR_ELx_EC_HVC64, hvc_handler); > +} > diff --git a/tools/testing/selftests/kvm/lib/arm64/nested_asm.S b/tools/testing/selftests/kvm/lib/arm64/nested_asm.S > new file mode 100644 > index 000000000000..4ecf2d510a6f > --- /dev/null > +++ b/tools/testing/selftests/kvm/lib/arm64/nested_asm.S > @@ -0,0 +1,35 @@ > +/* SPDX-License-Identifier: GPL-2.0 */ > +/* > + * ARM64 Nested virtualization assembly helpers > + */ > + > +.globl run_l2 > +.globl after_hvc > +.globl do_hvc > +run_l2: > + /* > + * At this point TPIDR_EL2 will contain the gpa of do_hvc from > + * prepare_eret_destination(). gpa of do_hvc have to be passed in > + * because we want L2 to issue an hvc after it returns from the user > + * passed function. In order for that to happen the lr must be > + * controlled, which at this point holds the value of the address of > + * the next instruction after this run_l2() call, which is not useful > + * for L2. Additionally, L1 can't translate gva into gpa, so we can't > + * calculate it here. > + * > + * So first save lr, then move TPIDR_EL2 to lr so when the user supplied > + * L2 function returns, L2 jumps to do_hvc and let the L1 hvc handler > + * take control. This implies we expect the L2 code to preserve lr and > + * calls a regular ret in the end, which is true for normal C functions. > + * The hvc handler will jump back to after_hvc when finished, and lr > + * will be restored and we can return run_l2(). > + */ > + stp x29, lr, [sp, #-16]! > + mrs x0, tpidr_el2 > + mov lr, x0 > + eret > +after_hvc: > + ldp x29, lr, [sp], #16 > + ret > +do_hvc: > + hvc #0 This probably works for a single instruction L2 guest, but not having any save/restore of the L2 context makes it hard to build anything on top of this. Thanks, M. -- Without deviation from the norm, progress is not possible.