mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Mark Brown <broonie@kernel.org>
To: Marc Zyngier <maz@kernel.org>, Joey Gouly <joey.gouly@arm.com>,
	 Catalin Marinas <catalin.marinas@arm.com>,
	 Suzuki K Poulose <suzuki.poulose@arm.com>,
	Will Deacon <will@kernel.org>,
	 Paolo Bonzini <pbonzini@redhat.com>,
	Jonathan Corbet <corbet@lwn.net>,  Shuah Khan <shuah@kernel.org>,
	Oliver Upton <oupton@kernel.org>
Cc: Dave Martin <Dave.Martin@arm.com>, Fuad Tabba <tabba@google.com>,
	 Mark Rutland <mark.rutland@arm.com>,
	Ben Horgan <ben.horgan@arm.com>,
	 Jean-Philippe Brucker <jpb@kernel.org>,
	 linux-arm-kernel@lists.infradead.org, kvmarm@lists.linux.dev,
	 linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
	 linux-doc@vger.kernel.org, linux-kselftest@vger.kernel.org,
	 Peter Maydell <peter.maydell@linaro.org>,
	 Eric Auger <eric.auger@redhat.com>,
	Mark Brown <broonie@kernel.org>
Subject: [PATCH v13 03/32] arm64/fpsimd: Configure all ZCR/SMCR bits when loading task state
Date: Mon, 20 Jul 2026 00:07:30 +0100	[thread overview]
Message-ID: <20260720-kvm-arm64-sme-v13-3-d9abd3ffa245@kernel.org> (raw)
In-Reply-To: <20260720-kvm-arm64-sme-v13-0-d9abd3ffa245@kernel.org>

Currently, when loading a task's SVE and/or SME state, we configure
ZCR_ELx.LEN and/or SMCR_ELx.LEN with read-modify-write sequences,
preserving all other bits of the registers (e.g. SMCR_ELx.{FA64,ZT0}).
This relies on all the other bits already holding the expected value
for the task.

It would be simpler and more robust to configure all the bits without
a read-modify-write sequence:

 - Doing so will remove the need to configure these registers during
   feature detection and when returning from idle, simplifying the code
   and removing some redundant writes to the registers.

 - Doing so will permit KVM to clobber other bits of the registers when
   no task state is bound. Along with other changes (e.g. when saving
   state), this will make it possible for KVM to expose a different
   configuration to guests (e.g. disabling ZT0 for vCPUs, even if ZT0
   is exposed to userspace).

While SMCR_ELx.LEN and ZCR_ELx.LEN are self synchronizing SMCR_ELx.FA64
and SMCR_ELx.EZT0 are not so we add an isb() if those have changed.
This could be further optimised by considering if the SME configuration
of the loaded state means we will rely on the newly configured values
but for clarity this is not done here, it can be improved in future.

Signed-off-by: Mark Brown <broonie@kernel.org>
---
 arch/arm64/include/asm/fpsimd.h |  2 -
 arch/arm64/kernel/cpufeature.c  |  2 -
 arch/arm64/kernel/fpsimd.c      | 84 +++++++++++++++++++----------------------
 3 files changed, 38 insertions(+), 50 deletions(-)

diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h
index e13a85320ab6..59f3c4a9e390 100644
--- a/arch/arm64/include/asm/fpsimd.h
+++ b/arch/arm64/include/asm/fpsimd.h
@@ -360,8 +360,6 @@ struct arm64_cpu_capabilities;
 extern void cpu_enable_fpsimd(const struct arm64_cpu_capabilities *__unused);
 extern void cpu_enable_sve(const struct arm64_cpu_capabilities *__unused);
 extern void cpu_enable_sme(const struct arm64_cpu_capabilities *__unused);
-extern void cpu_enable_sme2(const struct arm64_cpu_capabilities *__unused);
-extern void cpu_enable_fa64(const struct arm64_cpu_capabilities *__unused);
 extern void cpu_enable_fpmr(const struct arm64_cpu_capabilities *__unused);
 
 extern void fpsimd_suspend_exit(void);
diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index 9a22df0c5120..0609dce1989e 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -2992,7 +2992,6 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
 		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
 		.capability = ARM64_SME_FA64,
 		.matches = has_cpuid_feature,
-		.cpu_enable = cpu_enable_fa64,
 		ARM64_CPUID_FIELDS(ID_AA64SMFR0_EL1, FA64, IMP)
 	},
 	{
@@ -3000,7 +2999,6 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
 		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
 		.capability = ARM64_SME2,
 		.matches = has_cpuid_feature,
-		.cpu_enable = cpu_enable_sme2,
 		ARM64_CPUID_FIELDS(ID_AA64PFR1_EL1, SME, SME2)
 	},
 #endif /* CONFIG_ARM64_SME */
diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c
index 78c9d8dab545..b868db2982ee 100644
--- a/arch/arm64/kernel/fpsimd.c
+++ b/arch/arm64/kernel/fpsimd.c
@@ -277,6 +277,27 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type,
 	task->thread.vl_onexec[type] = vl;
 }
 
+static unsigned long task_zcr(const struct task_struct *task)
+{
+	unsigned long vq = sve_vq_from_vl(task_get_sve_vl(task));
+	unsigned long zcr = vq - 1;
+
+	return zcr;
+}
+
+static unsigned long task_smcr(const struct task_struct *task)
+{
+	unsigned long vq = sve_vq_from_vl(task_get_sme_vl(task));
+	unsigned long smcr = vq - 1;
+
+	if (system_supports_fa64())
+		smcr |= SMCR_ELx_FA64;
+	if (system_supports_sme2())
+		smcr |= SMCR_ELx_EZT0;
+
+	return smcr;
+}
+
 /*
  * TIF_SME controls whether a task can use SME without trapping while
  * in userspace, when TIF_SME is set then we must have storage
@@ -377,10 +398,8 @@ static void task_fpsimd_load(void)
 			if (!thread_sm_enabled(&current->thread))
 				WARN_ON_ONCE(!test_and_set_thread_flag(TIF_SVE));
 
-			if (test_thread_flag(TIF_SVE)) {
-				unsigned long vq = sve_vq_from_vl(task_get_sve_vl(current));
-				sysreg_clear_set_s(SYS_ZCR_EL1, ZCR_ELx_LEN, vq - 1);
-			}
+			if (system_supports_sve())
+				sysreg_cond_update_s(SYS_ZCR_EL1, task_zcr(current));
 
 			restore_sve_regs = true;
 			restore_ffr = true;
@@ -402,12 +421,17 @@ static void task_fpsimd_load(void)
 
 	/* Restore SME, override SVE register configuration if needed */
 	if (system_supports_sme()) {
-		unsigned long sme_vl = task_get_sme_vl(current);
-
-		/* Ensure VL is set up for restoring data */
 		if (test_thread_flag(TIF_SME)) {
-			unsigned long vq = sve_vq_from_vl(sme_vl);
-			sysreg_clear_set_s(SYS_SMCR_EL1, SMCR_ELx_LEN, vq - 1);
+			u64 old_smcr = read_sysreg_s(SYS_SMCR_EL1);
+			u64 new_smcr = task_smcr(current);
+
+			sysreg_cond_update_s(SYS_SMCR_EL1, new_smcr);
+
+			/* FA64 and EZT0 are not self synchronising */
+			old_smcr &= SMCR_ELx_FA64 | SMCR_ELx_EZT0;
+			new_smcr &= SMCR_ELx_FA64 | SMCR_ELx_EZT0;
+			if (old_smcr != new_smcr)
+				isb();
 		}
 
 		write_sysreg_s(current->thread.svcr, SYS_SVCR);
@@ -1217,26 +1241,6 @@ void cpu_enable_sme(const struct arm64_cpu_capabilities *__always_unused p)
 	isb();
 }
 
-void cpu_enable_sme2(const struct arm64_cpu_capabilities *__always_unused p)
-{
-	/* This must be enabled after SME */
-	BUILD_BUG_ON(ARM64_SME2 <= ARM64_SME);
-
-	/* Allow use of ZT0 */
-	write_sysreg_s(read_sysreg_s(SYS_SMCR_EL1) | SMCR_ELx_EZT0_MASK,
-		       SYS_SMCR_EL1);
-}
-
-void cpu_enable_fa64(const struct arm64_cpu_capabilities *__always_unused p)
-{
-	/* This must be enabled after SME */
-	BUILD_BUG_ON(ARM64_SME_FA64 <= ARM64_SME);
-
-	/* Allow use of FA64 */
-	write_sysreg_s(read_sysreg_s(SYS_SMCR_EL1) | SMCR_ELx_FA64_MASK,
-		       SYS_SMCR_EL1);
-}
-
 void __init sme_setup(void)
 {
 	struct vl_info *info = &vl_info[ARM64_VEC_SME];
@@ -1283,20 +1287,10 @@ void __init sme_setup(void)
 
 void fpsimd_suspend_exit(void)
 {
-	u64 smcr = 0;
-
-	if (system_supports_sve())
-		write_sysreg_s(0, SYS_ZCR_EL1);
-
-	if (system_supports_sme()) {
-		if (system_supports_fa64())
-			smcr |= SMCR_ELx_FA64;
-		if (system_supports_sme2())
-			smcr |= SMCR_ELx_EZT0;
+	if (!system_supports_sme())
+		return;
 
-		write_sysreg_s(smcr, SYS_SMCR_EL1);
-		write_sysreg_s(0, SYS_SMPRI_EL1);
-	}
+	write_sysreg_s(0, SYS_SMPRI_EL1);
 }
 
 /*
@@ -1338,8 +1332,7 @@ void do_sve_acc(unsigned long esr, struct pt_regs *regs)
 	 * any effective streaming mode SVE state.
 	 */
 	if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) {
-		unsigned long vq = sve_vq_from_vl(task_get_sve_vl(current));
-		sysreg_clear_set_s(SYS_ZCR_EL1, ZCR_ELx_LEN, vq - 1);
+		sysreg_cond_update_s(SYS_ZCR_EL1, task_zcr(current));
 		sve_flush_live();
 		fpsimd_bind_task_to_cpu();
 	} else {
@@ -1474,8 +1467,7 @@ void do_sme_acc(unsigned long esr, struct pt_regs *regs)
 		WARN_ON(1);
 
 	if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) {
-		unsigned long vq = sve_vq_from_vl(task_get_sme_vl(current));
-		sysreg_clear_set_s(SYS_SMCR_EL1, SMCR_ELx_LEN, vq - 1);
+		sysreg_cond_update_s(SYS_SMCR_EL1, task_smcr(current));
 
 		fpsimd_bind_task_to_cpu();
 	} else {

-- 
2.47.3


  parent reply	other threads:[~2026-07-19 23:10 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-07-19 23:07 [PATCH v13 00/32] KVM: arm64: Implement support for SME Mark Brown
2026-07-19 23:07 ` [PATCH v13 01/32] arm64/sysreg: Define full value read/modify/write helpers Mark Brown
2026-07-19 23:07 ` [PATCH v13 02/32] arm64/fpsimd: Ensure all of ZCR_EL1 is initialised from idle Mark Brown
2026-07-23 12:58   ` Mark Rutland
2026-07-24 15:16     ` Mark Rutland
2026-07-19 23:07 ` Mark Brown [this message]
2026-07-19 23:07 ` [PATCH v13 04/32] arm64/fpsimd: Decide to save ZT0 and streaming mode FFR at bind time Mark Brown
2026-07-19 23:07 ` [PATCH v13 05/32] arm64/sve: Factor virtualizable VL discovery out of SVE specific code Mark Brown
2026-07-19 23:07 ` [PATCH v13 06/32] arm64/fpsimd: Determine maximum virtualisable SME vector length Mark Brown
2026-07-19 23:07 ` [PATCH v13 07/32] KVM: arm64: Remove bitrotted comment on handle_sve() Mark Brown
2026-07-19 23:07 ` [PATCH v13 08/32] KVM: arm64: Handle FEAT_IDST for guest accesses to hidden registers Mark Brown
2026-07-19 23:07 ` [PATCH v13 09/32] KVM: arm64: Pull ctxt_has_ helpers to start of sysreg-sr.h Mark Brown
2026-07-19 23:07 ` [PATCH v13 10/32] KVM: arm64: Rename SVE finalization constants to be more general Mark Brown
2026-07-19 23:07 ` [PATCH v13 11/32] KVM: arm64: Remove special case for FP state loading from ZCR_EL2 traps Mark Brown
2026-07-19 23:07 ` [PATCH v13 12/32] KVM: arm64: Define internal features for SME Mark Brown
2026-07-19 23:07 ` [PATCH v13 13/32] KVM: arm64: Rename sve_state_reg_region Mark Brown
2026-07-19 23:07 ` [PATCH v13 14/32] KVM: arm64: Store vector lengths in an array Mark Brown
2026-07-19 23:07 ` [PATCH v13 15/32] KVM: arm64: Factor SVE code out of fpsimd_lazy_switch_to_host() Mark Brown
2026-07-19 23:07 ` [PATCH v13 16/32] KVM: arm64: Document the KVM ABI for SME Mark Brown
2026-07-19 23:07 ` [PATCH v13 17/32] KVM: arm64: Implement SME vector length configuration Mark Brown
2026-07-19 23:07 ` [PATCH v13 18/32] KVM: arm64: Support SME control registers Mark Brown
2026-07-19 23:07 ` [PATCH v13 19/32] KVM: arm64: Support TPIDR2_EL0 Mark Brown
2026-07-19 23:07 ` [PATCH v13 20/32] KVM: arm64: Support SME identification registers for guests Mark Brown
2026-07-19 23:07 ` [PATCH v13 21/32] KVM: arm64: Support SME priority registers Mark Brown
2026-07-19 23:07 ` [PATCH v13 22/32] KVM: arm64: Support userspace access to streaming mode Z and P registers Mark Brown
2026-07-19 23:07 ` [PATCH v13 23/32] KVM: arm64: Flush register state on writes to SVCR.SM and SVCR.ZA Mark Brown
2026-07-19 23:07 ` [PATCH v13 24/32] KVM: arm64: Expose SME specific state to userspace Mark Brown
2026-07-19 23:07 ` [PATCH v13 25/32] KVM: arm64: Context switch SME state for guests Mark Brown
2026-07-19 23:07 ` [PATCH v13 26/32] KVM: arm64: Handle SME exceptions Mark Brown
2026-07-19 23:07 ` [PATCH v13 27/32] KVM: arm64: Expose SME to nested guests Mark Brown
2026-07-19 23:07 ` [PATCH v13 28/32] KVM: arm64: Provide interface for configuring and enabling SME for guests Mark Brown
2026-07-19 23:07 ` [PATCH v13 29/32] KVM: arm64: selftests: Remove spurious check for single bit safe values Mark Brown
2026-07-19 23:07 ` [PATCH v13 30/32] KVM: arm64: selftests: Skip impossible invalid value tests Mark Brown
2026-07-19 23:07 ` [PATCH v13 31/32] KVM: arm64: selftests: Add SME system registers to get-reg-list Mark Brown
2026-07-19 23:07 ` [PATCH v13 32/32] KVM: arm64: selftests: Add SME to set_id_regs test Mark Brown

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=20260720-kvm-arm64-sme-v13-3-d9abd3ffa245@kernel.org \
    --to=broonie@kernel.org \
    --cc=Dave.Martin@arm.com \
    --cc=ben.horgan@arm.com \
    --cc=catalin.marinas@arm.com \
    --cc=corbet@lwn.net \
    --cc=eric.auger@redhat.com \
    --cc=joey.gouly@arm.com \
    --cc=jpb@kernel.org \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.linux.dev \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=maz@kernel.org \
    --cc=oupton@kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=shuah@kernel.org \
    --cc=suzuki.poulose@arm.com \
    --cc=tabba@google.com \
    --cc=will@kernel.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

all inboxes | Powered by JetHome®