From nobody Fri Apr 26 23:38:02 2024 Delivered-To: importer@patchew.org Received-SPF: pass (zoho.com: domain of gnu.org designates 208.118.235.17 as permitted sender) client-ip=208.118.235.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Authentication-Results: mx.zohomail.com; spf=pass (zoho.com: domain of gnu.org designates 208.118.235.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=fail(p=none dis=none) header.from=linaro.org Return-Path: Received: from lists.gnu.org (lists.gnu.org [208.118.235.17]) by mx.zohomail.com with SMTPS id 1544110101000361.14437644713496; Thu, 6 Dec 2018 07:28:21 -0800 (PST) Received: from localhost ([::1]:41430 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gUvZb-0008W2-K7 for importer@patchew.org; Thu, 06 Dec 2018 10:28:19 -0500 Received: from eggs.gnu.org ([2001:4830:134:3::10]:47321) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gUvX5-0006CT-NO for qemu-devel@nongnu.org; Thu, 06 Dec 2018 10:25:48 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gUvM3-0002fT-7m for qemu-devel@nongnu.org; Thu, 06 Dec 2018 10:14:24 -0500 Received: from orth.archaic.org.uk ([2001:8b0:1d0::2]:53272) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1gUvLx-0002OU-TX; Thu, 06 Dec 2018 10:14:14 -0500 Received: from pm215 by orth.archaic.org.uk with local (Exim 4.89) (envelope-from ) id 1gUvLl-0004rU-TJ; Thu, 06 Dec 2018 15:14:01 +0000 From: Peter Maydell To: qemu-arm@nongnu.org, qemu-devel@nongnu.org Date: Thu, 6 Dec 2018 15:14:01 +0000 Message-Id: <20181206151401.13455-1-peter.maydell@linaro.org> X-Mailer: git-send-email 2.19.2 MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 2001:8b0:1d0::2 Subject: [Qemu-devel] [RFC] arm: Allow system registers for KVM guests to be changed by QEMU code X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Dongjiu Geng , =?UTF-8?q?Alex=20Benn=C3=A9e?= , patches@linaro.org Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: "Qemu-devel" Content-Type: text/plain; charset="utf-8" At the moment the Arm implementations of kvm_arch_{get,put}_registers() don't support having QEMU change the values of system registers (aka coprocessor registers for AArch32). This is because although kvm_arch_get_registers() calls write_list_to_cpustate() to update the CPU state struct fields (so QEMU code can read the values in the usual way), kvm_arch_put_registers() does not call write_cpustate_to_list(), meaning that any changes to the CPU state struct fields will not be passed back to KVM. The rationale for this design is documented in a comment in the AArch32 kvm_arch_put_registers() -- writing the values in the cpregs list into the CPU state struct is "lossy" because the write of a register might not succeed, and so if we blindly copy the CPU state values back again we will incorrectly change register values for the guest. The assumption was that no QEMU code would need to write to the registers. However, when we implemented debug support for KVM guests, we broke that assumption: the code to handle "set the guest up to take a breakpoint exception" does so by updating various guest registers including ESR_EL1. Support this by making kvm_arch_put_registers() synchronize CPU state back into the list. We sync only those registers where the initial write succeeds, which should be sufficient. Signed-off-by: Peter Maydell Reviewed-by: Alex Benn=C3=A9e Tested-by: Alex Benn=C3=A9e Tested-by: Dongjiu Geng --- I think this patch: * should be necessary for the current KVM debug code to be able to actually set the ESR_EL1 for the guest when it wants to deliver a breakpoint exception to it * will also be necessary for Dongjiu's patchset to inject external aborts on memory errors I have tagged it "RFC" because I don't have a setup to test that; I've just given it a quick smoke test that it runs a VM OK. Please test this and check whether it really does fix the bugs I think it does :-) Opinions welcome on whether the "try the write-and-read-back" approach in write_cpustate_to_list() is too hacky and it would be better to actually record whether write_list_to_cpustate() succeeded for each register. (That would require another array.) --- target/arm/cpu.h | 9 ++++++++- target/arm/helper.c | 27 +++++++++++++++++++++++++-- target/arm/kvm32.c | 20 ++------------------ target/arm/kvm64.c | 2 ++ target/arm/machine.c | 2 +- 5 files changed, 38 insertions(+), 22 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index 2a73fed9a01..c0c111378ea 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -2321,18 +2321,25 @@ bool write_list_to_cpustate(ARMCPU *cpu); /** * write_cpustate_to_list: * @cpu: ARMCPU + * @kvm_sync: true if this is for syncing back to KVM * * For each register listed in the ARMCPU cpreg_indexes list, write * its value from the ARMCPUState structure into the cpreg_values list. * This is used to copy info from TCG's working data structures into * KVM or for outbound migration. * + * @kvm_sync is true if we are doing this in order to sync the + * register state back to KVM. In this case we will only update + * values in the list if the previous list->cpustate sync actually + * successfully wrote the CPU state. Otherwise we will keep the value + * that is in the list. + * * Returns: true if all register values were read correctly, * false if some register was unknown or could not be read. * Note that we do not stop early on failure -- we will attempt * reading all registers in the list. */ -bool write_cpustate_to_list(ARMCPU *cpu); +bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); =20 #define ARM_CPUID_TI915T 0x54029152 #define ARM_CPUID_TI925T 0x54029252 diff --git a/target/arm/helper.c b/target/arm/helper.c index 0da1424f72d..bc2969eb4d8 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -263,7 +263,7 @@ static bool raw_accessors_invalid(const ARMCPRegInfo *r= i) return true; } =20 -bool write_cpustate_to_list(ARMCPU *cpu) +bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync) { /* Write the coprocessor state from cpu->env to the (index,value) list= . */ int i; @@ -272,6 +272,7 @@ bool write_cpustate_to_list(ARMCPU *cpu) for (i =3D 0; i < cpu->cpreg_array_len; i++) { uint32_t regidx =3D kvm_to_cpreg_id(cpu->cpreg_indexes[i]); const ARMCPRegInfo *ri; + uint64_t newval; =20 ri =3D get_arm_cp_reginfo(cpu->cp_regs, regidx); if (!ri) { @@ -281,7 +282,29 @@ bool write_cpustate_to_list(ARMCPU *cpu) if (ri->type & ARM_CP_NO_RAW) { continue; } - cpu->cpreg_values[i] =3D read_raw_cp_reg(&cpu->env, ri); + + newval =3D read_raw_cp_reg(&cpu->env, ri); + if (kvm_sync) { + /* + * Only sync if the previous list->cpustate sync succeeded. + * Rather than tracking the success/failure state for every + * item in the list, we just recheck "does the raw write we mu= st + * have made in write_list_to_cpustate() read back OK" here. + */ + uint64_t oldval =3D cpu->cpreg_values[i]; + + if (oldval =3D=3D newval) { + continue; + } + + write_raw_cp_reg(&cpu->env, ri, oldval); + if (read_raw_cp_reg(&cpu->env, ri) !=3D oldval) { + continue; + } + + write_raw_cp_reg(&cpu->env, ri, newval); + } + cpu->cpreg_values[i] =3D newval; } return ok; } diff --git a/target/arm/kvm32.c b/target/arm/kvm32.c index bd51eb43c86..a75e04cc8f3 100644 --- a/target/arm/kvm32.c +++ b/target/arm/kvm32.c @@ -387,24 +387,8 @@ int kvm_arch_put_registers(CPUState *cs, int level) return ret; } =20 - /* Note that we do not call write_cpustate_to_list() - * here, so we are only writing the tuple list back to - * KVM. This is safe because nothing can change the - * CPUARMState cp15 fields (in particular gdb accesses cannot) - * and so there are no changes to sync. In fact syncing would - * be wrong at this point: for a constant register where TCG and - * KVM disagree about its value, the preceding write_list_to_cpustate() - * would not have had any effect on the CPUARMState value (since the - * register is read-only), and a write_cpustate_to_list() here would - * then try to write the TCG value back into KVM -- this would either - * fail or incorrectly change the value the guest sees. - * - * If we ever want to allow the user to modify cp15 registers via - * the gdb stub, we would need to be more clever here (for instance - * tracking the set of registers kvm_arch_get_registers() successfully - * managed to update the CPUARMState with, and only allowing those - * to be written back up into the kernel). - */ + write_cpustate_to_list(cpu, true); + if (!write_list_to_kvmstate(cpu, level)) { return EINVAL; } diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index 0a502091e76..d8ac978d7c3 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -834,6 +834,8 @@ int kvm_arch_put_registers(CPUState *cs, int level) return ret; } =20 + write_cpustate_to_list(cpu, true); + if (!write_list_to_kvmstate(cpu, level)) { return EINVAL; } diff --git a/target/arm/machine.c b/target/arm/machine.c index 7a22ebc2098..0dd0157f4d4 100644 --- a/target/arm/machine.c +++ b/target/arm/machine.c @@ -626,7 +626,7 @@ static int cpu_pre_save(void *opaque) abort(); } } else { - if (!write_cpustate_to_list(cpu)) { + if (!write_cpustate_to_list(cpu, false)) { /* This should never fail. */ abort(); } --=20 2.19.2