KVM: arm64: Fix softirq masking in FPSIMD register saving sequence

Stable commit 28b82be094e2 ("KVM: arm64: Fix kernel BUG() due to bad backport of FPSIMD/SVE/SME fix") fixed a kernel BUG() caused by a bad backport of upstream commit fbc7e61195e2 ("KVM: arm64: Unconditionally save+flush host FPSIMD/SVE/SME state") by ensuring that softirqs are disabled/enabled across the fpsimd register save operation.

Unfortunately, although this fixes the original issue, it can now lead to deadlock when re-enabling softirqs causes pending softirqs to be handled with locks already held:

 | BUG: spinlock recursion on CPU#7, CPU 3/KVM/57616
 |  lock: 0xffff3045ef850240, .magic: dead4ead, .owner: CPU 3/KVM/57616, .owner_cpu: 7
 | CPU: 7 PID: 57616 Comm: CPU 3/KVM Tainted: G           O       6.1.152 #1
 | Hardware name: SoftIron SoftIron Platform Mainboard/SoftIron Platform Mainboard, BIOS 1.31 May 11 2023
 | Call trace:
 |  dump_backtrace+0xe4/0x110
 |  show_stack+0x20/0x30
 |  dump_stack_lvl+0x6c/0x88
 |  dump_stack+0x18/0x34
 |  spin_dump+0x98/0xac
 |  do_raw_spin_lock+0x70/0x128
 |  _raw_spin_lock+0x18/0x28
 |  raw_spin_rq_lock_nested+0x18/0x28
 |  update_blocked_averages+0x70/0x550
 |  run_rebalance_domains+0x50/0x70
 |  handle_softirqs+0x198/0x328
 |  __do_softirq+0x1c/0x28
 |  ____do_softirq+0x18/0x28
 |  call_on_irq_stack+0x30/0x48
 |  do_softirq_own_stack+0x24/0x30
 |  do_softirq+0x74/0x90
 |  __local_bh_enable_ip+0x64/0x80
 |  fpsimd_save_and_flush_cpu_state+0x5c/0x68
 |  kvm_arch_vcpu_put_fp+0x4c/0x88
 |  kvm_arch_vcpu_put+0x28/0x88
 |  kvm_sched_out+0x38/0x58
 |  __schedule+0x55c/0x6c8
 |  schedule+0x60/0xa8

Take a tiny step towards the upstream fix in 9b19700e623f ("arm64:
fpsimd: Drop unneeded 'busy' flag") by additionally disabling hardirqs while saving the fpsimd registers.

---
 arch/arm64/kernel/fpsimd.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

index d0d836448a76..83827384982e 100644
@@ -1873,13 +1873,17 @@ static void fpsimd_flush_cpu_state(void)
  */
 void fpsimd_save_and_flush_cpu_state(void)
 {
+	unsigned long flags;
+
 	if (!system_supports_fpsimd())
 		return;
 	WARN_ON(preemptible());
-	get_cpu_fpsimd_context();
+	local_irq_save(flags);
+	__get_cpu_fpsimd_context();
 	fpsimd_save();
 	fpsimd_flush_cpu_state();
-	put_cpu_fpsimd_context();
+	__put_cpu_fpsimd_context();
+	local_irq_restore(flags);
 }
 
 #ifdef CONFIG_KERNEL_MODE_NEON

2.51.0.618.g983fd99d29-goog

This is a note to let you know that I've just added the patch titled

KVM: arm64: Fix softirq masking in FPSIMD register saving sequence

to the 6.6-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
kvm-arm64-fix-softirq-masking-in-fpsimd-register-saving-sequence.patch
and it can be found in the queue-6.6 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

From [email protected] Mon Oct 6 11:58:17 2025
From: Will Deacon <[email protected]>
Date: Fri, 3 Oct 2025 19:40:54 +0100
Subject: KVM: arm64: Fix softirq masking in FPSIMD register saving sequence
To: [email protected]
Cc: [email protected], [email protected], Will Deacon <[email protected]>, Ard Biesheuvel <[email protected]>, Lee Jones <[email protected]>, Sasha Levin <[email protected]>, Greg Kroah-Hartman <[email protected]>, Kenneth Van Alstyne <[email protected]>
Message-ID: <[email protected]>

From: Will Deacon <[email protected]>

Stable commit 28b82be094e2 ("KVM: arm64: Fix kernel BUG() due to bad backport of FPSIMD/SVE/SME fix") fixed a kernel BUG() caused by a bad backport of upstream commit fbc7e61195e2 ("KVM: arm64: Unconditionally save+flush host FPSIMD/SVE/SME state") by ensuring that softirqs are disabled/enabled across the fpsimd register save operation.

Unfortunately, although this fixes the original issue, it can now lead to deadlock when re-enabling softirqs causes pending softirqs to be handled with locks already held:

 | BUG: spinlock recursion on CPU#7, CPU 3/KVM/57616
 |  lock: 0xffff3045ef850240, .magic: dead4ead, .owner: CPU 3/KVM/57616, .owner_cpu: 7
 | CPU: 7 PID: 57616 Comm: CPU 3/KVM Tainted: G           O       6.1.152 #1
 | Hardware name: SoftIron SoftIron Platform Mainboard/SoftIron Platform Mainboard, BIOS 1.31 May 11 2023
 | Call trace:
 |  dump_backtrace+0xe4/0x110
 |  show_stack+0x20/0x30
 |  dump_stack_lvl+0x6c/0x88
 |  dump_stack+0x18/0x34
 |  spin_dump+0x98/0xac
 |  do_raw_spin_lock+0x70/0x128
 |  _raw_spin_lock+0x18/0x28
 |  raw_spin_rq_lock_nested+0x18/0x28
 |  update_blocked_averages+0x70/0x550
 |  run_rebalance_domains+0x50/0x70
 |  handle_softirqs+0x198/0x328
 |  __do_softirq+0x1c/0x28
 |  ____do_softirq+0x18/0x28
 |  call_on_irq_stack+0x30/0x48
 |  do_softirq_own_stack+0x24/0x30
 |  do_softirq+0x74/0x90
 |  __local_bh_enable_ip+0x64/0x80
 |  fpsimd_save_and_flush_cpu_state+0x5c/0x68
 |  kvm_arch_vcpu_put_fp+0x4c/0x88
 |  kvm_arch_vcpu_put+0x28/0x88
 |  kvm_sched_out+0x38/0x58
 |  __schedule+0x55c/0x6c8
 |  schedule+0x60/0xa8

Take a tiny step towards the upstream fix in 9b19700e623f ("arm64:
fpsimd: Drop unneeded 'busy' flag") by additionally disabling hardirqs while saving the fpsimd registers.

---
 arch/arm64/kernel/fpsimd.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

index d0d836448a76..83827384982e 100644
@@ -1873,13 +1873,17 @@ static void fpsimd_flush_cpu_state(void)
  */
 void fpsimd_save_and_flush_cpu_state(void)
 {
+	unsigned long flags;
+
 	if (!system_supports_fpsimd())
 		return;
 	WARN_ON(preemptible());
-	get_cpu_fpsimd_context();
+	local_irq_save(flags);
+	__get_cpu_fpsimd_context();
 	fpsimd_save();
 	fpsimd_flush_cpu_state();
-	put_cpu_fpsimd_context();
+	__put_cpu_fpsimd_context();
+	local_irq_restore(flags);
 }
 
 #ifdef CONFIG_KERNEL_MODE_NEON

2.51.0.618.g983fd99d29-goog



Patches currently in stable-queue which might be from [email protected] are

queue-6.6/kvm-arm64-fix-softirq-masking-in-fpsimd-register-saving-sequence.patch