arm64: mte: allow async MTE to be upgraded to sync on a per-CPU basis
Subsystems (2)
PATCH v3 1/1 arm64: mte: allow async MTE to be upgraded to sync on a per-CPU basis
On some CPUs the performance of MTE in synchronous mode is the same as that of asynchronous mode. This makes it worthwhile to enable synchronous mode on those CPUs when asynchronous mode is requested, in order to gain the error detection benefits of synchronous mode without the performance downsides. Therefore, make it possible for user programs to opt into upgrading to synchronous mode on those CPUs via a new prctl flag. The flag is orthogonal to the existing TCF modes in order to accommodate upgrading from other TCF modes in the future.
The feature is controlled on a per-CPU basis via sysfs.
---
v3:
- drop the device tree support
- add documentation
- add static_assert to ensure no overlap with real HW bits
- move per-CPU variable initialization to mte.c
- use smp_call_function_single instead of stop_machine
v2:
- make it an opt-in behavior
- change the format of the device tree node
- also allow controlling the feature via sysfs
.../arm64/memory-tagging-extension.rst | 20 ++++++ arch/arm64/include/asm/processor.h | 8 +++ arch/arm64/kernel/mte.c | 71 ++++++++++++++++++- arch/arm64/kernel/process.c | 15 +++- include/uapi/linux/prctl.h | 2 + 5 files changed, 113 insertions(+), 3 deletions(-)
Documentation/arm64/memory-tagging-extension.rst +20 -0
@@ -120,6 +120,25 @@ in the ``PR_MTE_TAG_MASK`` bit-field. interface provides an include mask. An include mask of ``0`` (exclusion mask ``0xffff``) results in the CPU always generating tag ``0``. +Upgrading to stricter tag checking modes +---------------------------------------- + +On some CPUs the performance of MTE in stricter tag checking modes +is the same as that of less strict tag checking modes. This makes it +worthwhile to enable stricter checks on those CPUs when a less strict +checking mode is requested, in order to gain the error detection +benefits of the stricter checks without the performance downsides. To +opt into upgrading to a stricter checking mode on those CPUs, the user +can set the ``PR_MTE_DYNAMIC_TCF`` flag bit in the ``flags`` argument +to the ``prctl(PR_SET_TAGGED_ADDR_CTRL, flags, 0, 0, 0)`` system call. + +This feature is currently only supported for upgrading from +asynchronous mode. To configure a CPU to upgrade from asynchronous mode +to synchronous mode, a privileged user may write the value ``1`` to +``/sys/devices/system/cpu/cpu<N>/mte_upgrade_async``, and to disable +upgrading they may write the value ``2``. By default the feature is +disabled on all CPUs. + Initial process state ---------------------
@@ -128,6 +147,7 @@ On ``execve()``, the new process has the following configuration: - ``PR_TAGGED_ADDR_ENABLE`` set to 0 (disabled) - Tag checking mode set to ``PR_MTE_TCF_NONE`` - ``PR_MTE_TAG_MASK`` set to 0 (all tags excluded) +- ``PR_MTE_DYNAMIC_TCF`` set to 0 (disabled) - ``PSTATE.TCO`` set to 0 - ``PROT_MTE`` not set on any of the initial memory maps
arch/arm64/include/asm/processor.h +8 -0
@@ -159,6 +159,13 @@ struct thread_struct {#define SCTLR_USER_MASK \ (SCTLR_ELx_ENIA | SCTLR_ELx_ENIB | SCTLR_ELx_ENDA | SCTLR_ELx_ENDB | \ SCTLR_EL1_TCF0_MASK) +#define SCTLR_USER_DYNAMIC_TCF (1ULL << 63) + +/* +* Make sure that SCTLR_USER_DYNAMIC_TCF doesn't overlap with any of the real +* per-task hardware bits. +*/ +static_assert ((SCTLR_USER_MASK &SCTLR_USER_DYNAMIC_TCF ) ==0 );static inline void arch_thread_struct_whitelist (unsigned long *offset,unsigned long *size)
@@ -251,6 +258,7 @@ extern void release_thread(struct task_struct *);unsigned long get_wchan (struct task_struct *p); +void update_sctlr_el1 (u64 sctlr);void set_task_sctlr_el1 (u64 sctlr);/* Thread switching */
arch/arm64/kernel/mte.c +70 -1
104 lines not on this page · raw message
arch/arm64/kernel/process.c +13 -2
@@ -527,8 +527,19 @@ static void erratum_1418040_thread_switch(struct task_struct *prev,write_sysreg (val, cntkctl_el1); } -static void update_sctlr_el1 (u64 sctlr) +DECLARE_PER_CPU_READ_MOSTLY (u64, mte_upgrade_async); + +void update_sctlr_el1 (u64 sctlr) { +if (sctlr &SCTLR_USER_DYNAMIC_TCF ) { + sctlr &= ~SCTLR_USER_DYNAMIC_TCF ; + +if ((sctlr &SCTLR_EL1_TCF0_MASK ) ==SCTLR_EL1_TCF0_ASYNC ) { + sctlr &= ~SCTLR_EL1_TCF0_MASK ; + sctlr |=__this_cpu_read (mte_upgrade_async); + } + } +/* * EnIA must not be cleared while in the kernel as this is necessary for * in-kernel PAC. It will be cleared on kernel exit if needed.
@@ -659,7 +670,7 @@ long set_tagged_addr_ctrl(struct task_struct *task, unsigned long arg)return -EINVAL ;if (system_supports_mte ()) - valid_mask |=PR_MTE_TCF_MASK |PR_MTE_TAG_MASK ; + valid_mask |=PR_MTE_TCF_MASK |PR_MTE_TAG_MASK |PR_MTE_DYNAMIC_TCF ;if (arg & ~valid_mask)return -EINVAL ;
include/uapi/linux/prctl.h +2 -0
@@ -242,6 +242,8 @@ struct prctl_mm_map {/* MTE tag inclusion mask */ # define PR_MTE_TAG_SHIFT 3# define PR_MTE_TAG_MASK (0xffffUL << PR_MTE_TAG_SHIFT) +/* Enable dynamic upgrading of MTE tag check fault mode */ +# define PR_MTE_DYNAMIC_TCF (1UL << 19)/* Control reclaim behavior when allocating memory */ #define PR_SET_IO_FLUSHER 57
Signature · 7 lines
2.32.0.272.g935e593368-goog _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Catalin Marinas
maintainerQuoted 16 lines · Peter Collingbourne
Documentation/arm64/memory-tagging-extension.rst +8
-0@@ -120,6 +120,25 @@ in the ``PR_MTE_TAG_MASK`` bit-field. interface provides an include mask. An include mask of ``0`` (exclusion mask ``0xffff``) results in the CPU always generating tag ``0``. +Upgrading to stricter tag checking modes +---------------------------------------- + +On some CPUs the performance of MTE in stricter tag checking modes +is the same as that of less strict tag checking modes. This makes it +worthwhile to enable stricter checks on those CPUs when a less strict +checking mode is requested, in order to gain the error detection +benefits of the stricter checks without the performance downsides. To
Documentation/arm64/memory-tagging-extension.rst in the ``PR_MTE_TAG_MASK`` bit-field. · line 140
Quoted 10 lines
Documentation/arm64/memory-tagging-extension.rst +10
-0+opt into upgrading to a stricter checking mode on those CPUs, the user +can set the ``PR_MTE_DYNAMIC_TCF`` flag bit in the ``flags`` argument +to the ``prctl(PR_SET_TAGGED_ADDR_CTRL, flags, 0, 0, 0)`` system call. + +This feature is currently only supported for upgrading from +asynchronous mode. To configure a CPU to upgrade from asynchronous mode +to synchronous mode, a privileged user may write the value ``1`` to +``/sys/devices/system/cpu/cpu<N>/mte_upgrade_async``, and to disable +upgrading they may write the value ``2``. By default the feature is +disabled on all CPUs.
Why not 0 to disable? This should be the default. I'd keep 2 for upgrading to the new asymmetric mode in v8.7 (reads sync, writes async).
Quoted 11 lines
arch/arm64/kernel/process.c +2
-1@@ -527,8 +527,19 @@ static void erratum_1418040_thread_switch(struct task_struct *prev,write_sysreg (val, cntkctl_el1); } -static void update_sctlr_el1 (u64 sctlr) +DECLARE_PER_CPU_READ_MOSTLY (u64, mte_upgrade_async); +
arch/arm64/kernel/process.c static void erratum_1418040_thread_switch(struct task_struct *prev, · line 541
Quoted 10 lines
arch/arm64/kernel/process.c +9
-0+void update_sctlr_el1 (u64 sctlr) { +if (sctlr &SCTLR_USER_DYNAMIC_TCF ) { + sctlr &= ~SCTLR_USER_DYNAMIC_TCF ; + +if ((sctlr &SCTLR_EL1_TCF0_MASK ) ==SCTLR_EL1_TCF0_ASYNC ) { + sctlr &= ~SCTLR_EL1_TCF0_MASK ; + sctlr |=__this_cpu_read (mte_upgrade_async); + } + }
I replied to your v2 already. I'd prefer this to be handled in mte.c.
Signature · 6 lines
Catalin _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Catalin Marinas
maintainerQuoted 22 lines · Catalin Marinas
Quoted 21 lines · Peter Collingbourne
Documentation/arm64/memory-tagging-extension.rst +13
-0@@ -120,6 +120,25 @@ in the ``PR_MTE_TAG_MASK`` bit-field. interface provides an include mask. An include mask of ``0`` (exclusion mask ``0xffff``) results in the CPU always generating tag ``0``. +Upgrading to stricter tag checking modes +---------------------------------------- + +On some CPUs the performance of MTE in stricter tag checking modes +is the same as that of less strict tag checking modes. This makes it +worthwhile to enable stricter checks on those CPUs when a less strict +checking mode is requested, in order to gain the error detection +benefits of the stricter checks without the performance downsides. To +opt into upgrading to a stricter checking mode on those CPUs, the user +can set the ``PR_MTE_DYNAMIC_TCF`` flag bit in the ``flags`` argument +to the ``prctl(PR_SET_TAGGED_ADDR_CTRL, flags, 0, 0, 0)`` system call. + +This feature is currently only supported for upgrading from
Quoted 8 lines
Documentation/arm64/memory-tagging-extension.rst in the ``PR_MTE_TAG_MASK`` bit-field. · line 140
Quoted 5 lines
Documentation/arm64/memory-tagging-extension.rst +5
-0+asynchronous mode. To configure a CPU to upgrade from asynchronous mode +to synchronous mode, a privileged user may write the value ``1`` to +``/sys/devices/system/cpu/cpu<N>/mte_upgrade_async``, and to disable +upgrading they may write the value ``2``. By default the feature is +disabled on all CPUs.Why not 0 to disable? This should be the default. I'd keep 2 for upgrading to the new asymmetric mode in v8.7 (reads sync, writes async).
Ah, I think I get it, you wanted to use the TCF values. Fine by me but I'd actually still keep 0 for disable (which is the default value I guess) with the 1 (or 3 later) for the upgrade. We may also add an mte_upgrade_asym file at some point.
Signature · 6 lines
Catalin _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Quoted 33 lines · Catalin Marinas
Quoted 22 lines · Catalin Marinas
Quoted 21 lines · Peter Collingbourne
Documentation/arm64/memory-tagging-extension.rst +13
-0@@ -120,6 +120,25 @@ in the ``PR_MTE_TAG_MASK`` bit-field. interface provides an include mask. An include mask of ``0`` (exclusion mask ``0xffff``) results in the CPU always generating tag ``0``. +Upgrading to stricter tag checking modes +---------------------------------------- + +On some CPUs the performance of MTE in stricter tag checking modes +is the same as that of less strict tag checking modes. This makes it +worthwhile to enable stricter checks on those CPUs when a less strict +checking mode is requested, in order to gain the error detection +benefits of the stricter checks without the performance downsides. To +opt into upgrading to a stricter checking mode on those CPUs, the user +can set the ``PR_MTE_DYNAMIC_TCF`` flag bit in the ``flags`` argument +to the ``prctl(PR_SET_TAGGED_ADDR_CTRL, flags, 0, 0, 0)`` system call. + +This feature is currently only supported for upgrading fromQuoted 8 lines
Documentation/arm64/memory-tagging-extension.rst in the ``PR_MTE_TAG_MASK`` bit-field. · line 140
Quoted 5 lines
Documentation/arm64/memory-tagging-extension.rst +5
-0+asynchronous mode. To configure a CPU to upgrade from asynchronous mode +to synchronous mode, a privileged user may write the value ``1`` to +``/sys/devices/system/cpu/cpu<N>/mte_upgrade_async``, and to disable +upgrading they may write the value ``2``. By default the feature is +disabled on all CPUs.Why not 0 to disable? This should be the default. I'd keep 2 for upgrading to the new asymmetric mode in v8.7 (reads sync, writes async).
Quoted 4 lines
Ah, I think I get it, you wanted to use the TCF values. Fine by me but I'd actually still keep 0 for disable (which is the default value I guess) with the 1 (or 3 later) for the upgrade. We may also add an
mte_upgrade_asymfile at some point.
So you want the settings to be:
0 - disable upgrading
1 - upgrade to sync
2 - upgrade to async
3 - upgrade to asym
Where 2 is disallowed? I guess that makes sense. So for mte_upgrade_asym we would only allow writing 0 or 1 and for mte_upgrade_none we would allow writing any value.
Peter
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Catalin Marinas
maintainerQuoted 12 lines · Peter Collingbourne
arch/arm64/kernel/mte.c +12
-0+static ssize_t mte_upgrade_async_store (struct device *dev, +struct device_attribute *attr, +const char *buf,size_t count) +{ +ssize_t ret; +u32 val; +u64 tcf; + + ret =kstrtou32 (buf,0 , &val); +if (ret <0 ) +return ret; +
arch/arm64/kernel/mte.c int mte_ptrace_copy_tags(struct task_struct *child, long request, · line 502
Quoted 10 lines
arch/arm64/kernel/mte.c +10
-0+ tcf = ((u64 )val) <<SCTLR_EL1_TCF0_SHIFT ; +if (tcf !=SCTLR_EL1_TCF0_NONE && tcf !=SCTLR_EL1_TCF0_SYNC && + tcf !=SCTLR_EL1_TCF0_ASYNC ) +return -EINVAL ; + +device_lock (dev); +per_cpu (mte_upgrade_async, dev->id) = tcf; + +if (cpu_online (dev->id)) + ret =smp_call_function_single (dev->id, sync_sctlr,NULL ,0 );
Forgot about this. I get it now, you only need to kick the CPU that is getting changed. This looks fine.
Signature · 6 lines
Catalin _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel