summaryrefslogtreecommitdiff
path: root/arch/arm64/kernel/fpsimd.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2019-07-08 19:54:55 +0300
committerLinus Torvalds <torvalds@linux-foundation.org>2019-07-08 19:54:55 +0300
commitdfd437a257924484b144ee750e60affc95562c6d (patch)
treed692781730ed72743b8abbc42a04a176cb2f1d6d /arch/arm64/kernel/fpsimd.c
parent0ecfebd2b52404ae0c54a878c872bb93363ada36 (diff)
parent0c61efd322b75ed3143e3d130ebecbebf561adf5 (diff)
downloadlinux-dfd437a257924484b144ee750e60affc95562c6d.tar.xz
Merge tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Pull arm64 updates from Catalin Marinas: - arm64 support for syscall emulation via PTRACE_SYSEMU{,_SINGLESTEP} - Wire up VM_FLUSH_RESET_PERMS for arm64, allowing the core code to manage the permissions of executable vmalloc regions more strictly - Slight performance improvement by keeping softirqs enabled while touching the FPSIMD/SVE state (kernel_neon_begin/end) - Expose a couple of ARMv8.5 features to user (HWCAP): CondM (new XAFLAG and AXFLAG instructions for floating point comparison flags manipulation) and FRINT (rounding floating point numbers to integers) - Re-instate ARM64_PSEUDO_NMI support which was previously marked as BROKEN due to some bugs (now fixed) - Improve parking of stopped CPUs and implement an arm64-specific panic_smp_self_stop() to avoid warning on not being able to stop secondary CPUs during panic - perf: enable the ARM Statistical Profiling Extensions (SPE) on ACPI platforms - perf: DDR performance monitor support for iMX8QXP - cache_line_size() can now be set from DT or ACPI/PPTT if provided to cope with a system cache info not exposed via the CPUID registers - Avoid warning on hardware cache line size greater than ARCH_DMA_MINALIGN if the system is fully coherent - arm64 do_page_fault() and hugetlb cleanups - Refactor set_pte_at() to avoid redundant READ_ONCE(*ptep) - Ignore ACPI 5.1 FADTs reported as 5.0 (infer from the 'arm_boot_flags' introduced in 5.1) - CONFIG_RANDOMIZE_BASE now enabled in defconfig - Allow the selection of ARM64_MODULE_PLTS, currently only done via RANDOMIZE_BASE (and an erratum workaround), allowing modules to spill over into the vmalloc area - Make ZONE_DMA32 configurable * tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (54 commits) perf: arm_spe: Enable ACPI/Platform automatic module loading arm_pmu: acpi: spe: Add initial MADT/SPE probing ACPI/PPTT: Add function to return ACPI 6.3 Identical tokens ACPI/PPTT: Modify node flag detection to find last IDENTICAL x86/entry: Simplify _TIF_SYSCALL_EMU handling arm64: rename dump_instr as dump_kernel_instr arm64/mm: Drop [PTE|PMD]_TYPE_FAULT arm64: Implement panic_smp_self_stop() arm64: Improve parking of stopped CPUs arm64: Expose FRINT capabilities to userspace arm64: Expose ARMv8.5 CondM capability to userspace arm64: defconfig: enable CONFIG_RANDOMIZE_BASE arm64: ARM64_MODULES_PLTS must depend on MODULES arm64: bpf: do not allocate executable memory arm64/kprobes: set VM_FLUSH_RESET_PERMS on kprobe instruction pages arm64/mm: wire up CONFIG_ARCH_HAS_SET_DIRECT_MAP arm64: module: create module allocations without exec permissions arm64: Allow user selection of ARM64_MODULE_PLTS acpi/arm64: ignore 5.1 FADTs that are reported as 5.0 arm64: Allow selecting Pseudo-NMI again ...
Diffstat (limited to 'arch/arm64/kernel/fpsimd.c')
-rw-r--r--arch/arm64/kernel/fpsimd.c139
1 files changed, 96 insertions, 43 deletions
diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c
index 0cfcf5c237c5..eec4776ae5f0 100644
--- a/arch/arm64/kernel/fpsimd.c
+++ b/arch/arm64/kernel/fpsimd.c
@@ -82,7 +82,8 @@
* To prevent this from racing with the manipulation of the task's FPSIMD state
* from task context and thereby corrupting the state, it is necessary to
* protect any manipulation of a task's fpsimd_state or TIF_FOREIGN_FPSTATE
- * flag with local_bh_disable() unless softirqs are already masked.
+ * flag with {, __}get_cpu_fpsimd_context(). This will still allow softirqs to
+ * run but prevent them to use FPSIMD.
*
* For a certain task, the sequence may look something like this:
* - the task gets scheduled in; if both the task's fpsimd_cpu field
@@ -145,6 +146,56 @@ extern void __percpu *efi_sve_state;
#endif /* ! CONFIG_ARM64_SVE */
+DEFINE_PER_CPU(bool, fpsimd_context_busy);
+EXPORT_PER_CPU_SYMBOL(fpsimd_context_busy);
+
+static void __get_cpu_fpsimd_context(void)
+{
+ bool busy = __this_cpu_xchg(fpsimd_context_busy, true);
+
+ WARN_ON(busy);
+}
+
+/*
+ * Claim ownership of the CPU FPSIMD context for use by the calling context.
+ *
+ * The caller may freely manipulate the FPSIMD context metadata until
+ * put_cpu_fpsimd_context() is called.
+ *
+ * The double-underscore version must only be called if you know the task
+ * can't be preempted.
+ */
+static void get_cpu_fpsimd_context(void)
+{
+ preempt_disable();
+ __get_cpu_fpsimd_context();
+}
+
+static void __put_cpu_fpsimd_context(void)
+{
+ bool busy = __this_cpu_xchg(fpsimd_context_busy, false);
+
+ WARN_ON(!busy); /* No matching get_cpu_fpsimd_context()? */
+}
+
+/*
+ * Release the CPU FPSIMD context.
+ *
+ * Must be called from a context in which get_cpu_fpsimd_context() was
+ * previously called, with no call to put_cpu_fpsimd_context() in the
+ * meantime.
+ */
+static void put_cpu_fpsimd_context(void)
+{
+ __put_cpu_fpsimd_context();
+ preempt_enable();
+}
+
+static bool have_cpu_fpsimd_context(void)
+{
+ return !preemptible() && __this_cpu_read(fpsimd_context_busy);
+}
+
/*
* Call __sve_free() directly only if you know task can't be scheduled
* or preempted.
@@ -215,12 +266,10 @@ static void sve_free(struct task_struct *task)
* This function should be called only when the FPSIMD/SVE state in
* thread_struct is known to be up to date, when preparing to enter
* userspace.
- *
- * Softirqs (and preemption) must be disabled.
*/
static void task_fpsimd_load(void)
{
- WARN_ON(!in_softirq() && !irqs_disabled());
+ WARN_ON(!have_cpu_fpsimd_context());
if (system_supports_sve() && test_thread_flag(TIF_SVE))
sve_load_state(sve_pffr(&current->thread),
@@ -233,16 +282,14 @@ static void task_fpsimd_load(void)
/*
* Ensure FPSIMD/SVE storage in memory for the loaded context is up to
* date with respect to the CPU registers.
- *
- * Softirqs (and preemption) must be disabled.
*/
-void fpsimd_save(void)
+static void fpsimd_save(void)
{
struct fpsimd_last_state_struct const *last =
this_cpu_ptr(&fpsimd_last_state);
/* set by fpsimd_bind_task_to_cpu() or fpsimd_bind_state_to_cpu() */
- WARN_ON(!in_softirq() && !irqs_disabled());
+ WARN_ON(!have_cpu_fpsimd_context());
if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) {
if (system_supports_sve() && test_thread_flag(TIF_SVE)) {
@@ -364,7 +411,8 @@ static __uint128_t arm64_cpu_to_le128(__uint128_t x)
* task->thread.sve_state.
*
* Task can be a non-runnable task, or current. In the latter case,
- * softirqs (and preemption) must be disabled.
+ * the caller must have ownership of the cpu FPSIMD context before calling
+ * this function.
* task->thread.sve_state must point to at least sve_state_size(task)
* bytes of allocated kernel memory.
* task->thread.uw.fpsimd_state must be up to date before calling this
@@ -393,7 +441,8 @@ static void fpsimd_to_sve(struct task_struct *task)
* task->thread.uw.fpsimd_state.
*
* Task can be a non-runnable task, or current. In the latter case,
- * softirqs (and preemption) must be disabled.
+ * the caller must have ownership of the cpu FPSIMD context before calling
+ * this function.
* task->thread.sve_state must point to at least sve_state_size(task)
* bytes of allocated kernel memory.
* task->thread.sve_state must be up to date before calling this function.
@@ -557,7 +606,7 @@ int sve_set_vector_length(struct task_struct *task,
* non-SVE thread.
*/
if (task == current) {
- local_bh_disable();
+ get_cpu_fpsimd_context();
fpsimd_save();
}
@@ -567,7 +616,7 @@ int sve_set_vector_length(struct task_struct *task,
sve_to_fpsimd(task);
if (task == current)
- local_bh_enable();
+ put_cpu_fpsimd_context();
/*
* Force reallocation of task SVE state to the correct size
@@ -880,7 +929,7 @@ asmlinkage void do_sve_acc(unsigned int esr, struct pt_regs *regs)
sve_alloc(current);
- local_bh_disable();
+ get_cpu_fpsimd_context();
fpsimd_save();
@@ -891,7 +940,7 @@ asmlinkage void do_sve_acc(unsigned int esr, struct pt_regs *regs)
if (test_and_set_thread_flag(TIF_SVE))
WARN_ON(1); /* SVE access shouldn't have trapped */
- local_bh_enable();
+ put_cpu_fpsimd_context();
}
/*
@@ -935,6 +984,8 @@ void fpsimd_thread_switch(struct task_struct *next)
if (!system_supports_fpsimd())
return;
+ __get_cpu_fpsimd_context();
+
/* Save unsaved fpsimd state, if any: */
fpsimd_save();
@@ -949,6 +1000,8 @@ void fpsimd_thread_switch(struct task_struct *next)
update_tsk_thread_flag(next, TIF_FOREIGN_FPSTATE,
wrong_task || wrong_cpu);
+
+ __put_cpu_fpsimd_context();
}
void fpsimd_flush_thread(void)
@@ -958,7 +1011,7 @@ void fpsimd_flush_thread(void)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ get_cpu_fpsimd_context();
fpsimd_flush_task_state(current);
memset(&current->thread.uw.fpsimd_state, 0,
@@ -999,7 +1052,7 @@ void fpsimd_flush_thread(void)
current->thread.sve_vl_onexec = 0;
}
- local_bh_enable();
+ put_cpu_fpsimd_context();
}
/*
@@ -1011,9 +1064,9 @@ void fpsimd_preserve_current_state(void)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ get_cpu_fpsimd_context();
fpsimd_save();
- local_bh_enable();
+ put_cpu_fpsimd_context();
}
/*
@@ -1030,7 +1083,8 @@ void fpsimd_signal_preserve_current_state(void)
/*
* Associate current's FPSIMD context with this cpu
- * Preemption must be disabled when calling this function.
+ * The caller must have ownership of the cpu FPSIMD context before calling
+ * this function.
*/
void fpsimd_bind_task_to_cpu(void)
{
@@ -1076,14 +1130,14 @@ void fpsimd_restore_current_state(void)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ get_cpu_fpsimd_context();
if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE)) {
task_fpsimd_load();
fpsimd_bind_task_to_cpu();
}
- local_bh_enable();
+ put_cpu_fpsimd_context();
}
/*
@@ -1096,7 +1150,7 @@ void fpsimd_update_current_state(struct user_fpsimd_state const *state)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ get_cpu_fpsimd_context();
current->thread.uw.fpsimd_state = *state;
if (system_supports_sve() && test_thread_flag(TIF_SVE))
@@ -1107,7 +1161,7 @@ void fpsimd_update_current_state(struct user_fpsimd_state const *state)
clear_thread_flag(TIF_FOREIGN_FPSTATE);
- local_bh_enable();
+ put_cpu_fpsimd_context();
}
/*
@@ -1133,18 +1187,29 @@ void fpsimd_flush_task_state(struct task_struct *t)
/*
* Invalidate any task's FPSIMD state that is present on this cpu.
- * This function must be called with softirqs disabled.
+ * The FPSIMD context should be acquired with get_cpu_fpsimd_context()
+ * before calling this function.
*/
-void fpsimd_flush_cpu_state(void)
+static void fpsimd_flush_cpu_state(void)
{
__this_cpu_write(fpsimd_last_state.st, NULL);
set_thread_flag(TIF_FOREIGN_FPSTATE);
}
-#ifdef CONFIG_KERNEL_MODE_NEON
+/*
+ * Save the FPSIMD state to memory and invalidate cpu view.
+ * This function must be called with preemption disabled.
+ */
+void fpsimd_save_and_flush_cpu_state(void)
+{
+ WARN_ON(preemptible());
+ __get_cpu_fpsimd_context();
+ fpsimd_save();
+ fpsimd_flush_cpu_state();
+ __put_cpu_fpsimd_context();
+}
-DEFINE_PER_CPU(bool, kernel_neon_busy);
-EXPORT_PER_CPU_SYMBOL(kernel_neon_busy);
+#ifdef CONFIG_KERNEL_MODE_NEON
/*
* Kernel-side NEON support functions
@@ -1170,19 +1235,13 @@ void kernel_neon_begin(void)
BUG_ON(!may_use_simd());
- local_bh_disable();
-
- __this_cpu_write(kernel_neon_busy, true);
+ get_cpu_fpsimd_context();
/* Save unsaved fpsimd state, if any: */
fpsimd_save();
/* Invalidate any task state remaining in the fpsimd regs: */
fpsimd_flush_cpu_state();
-
- preempt_disable();
-
- local_bh_enable();
}
EXPORT_SYMBOL(kernel_neon_begin);
@@ -1197,15 +1256,10 @@ EXPORT_SYMBOL(kernel_neon_begin);
*/
void kernel_neon_end(void)
{
- bool busy;
-
if (!system_supports_fpsimd())
return;
- busy = __this_cpu_xchg(kernel_neon_busy, false);
- WARN_ON(!busy); /* No matching kernel_neon_begin()? */
-
- preempt_enable();
+ put_cpu_fpsimd_context();
}
EXPORT_SYMBOL(kernel_neon_end);
@@ -1297,8 +1351,7 @@ static int fpsimd_cpu_pm_notifier(struct notifier_block *self,
{
switch (cmd) {
case CPU_PM_ENTER:
- fpsimd_save();
- fpsimd_flush_cpu_state();
+ fpsimd_save_and_flush_cpu_state();
break;
case CPU_PM_EXIT:
break;