/* The FPU state configuration data for kernel and user space */ struct fpu_state_config fpu_kernel_cfg __ro_after_init; struct fpu_state_config fpu_user_cfg __ro_after_init; struct vcpu_fpu_config guest_default_cfg __ro_after_init;
/* * Represents the initial FPU state. It's mostly (but not completely) zeroes, * depending on the FPU hardware format:
*/ struct fpstate init_fpstate __ro_after_init;
/* * Track FPU initialization and kernel-mode usage. 'true' means the FPU is * initialized and is not currently being used by the kernel:
*/
DEFINE_PER_CPU(bool, kernel_fpu_allowed);
/* * Track which context is using the FPU on the CPU:
*/
DEFINE_PER_CPU(struct fpu *, fpu_fpregs_owner_ctx);
/* * Can we use the FPU in kernel mode with the * whole "kernel_fpu_begin/end()" sequence?
*/ bool irq_fpu_usable(void)
{ if (WARN_ON_ONCE(in_nmi())) returnfalse;
/* * Return false in the following cases: * * - FPU is not yet initialized. This can happen only when the call is * coming from CPU onlining, for example for microcode checksumming. * - The kernel is already using the FPU, either because of explicit * nesting (which should never be done), or because of implicit * nesting when a hardirq interrupted a kernel-mode FPU section. * * The single boolean check below handles both cases:
*/ if (!this_cpu_read(kernel_fpu_allowed)) returnfalse;
/* * When not in NMI or hard interrupt context, FPU can be used in: * * - Task context except from within fpregs_lock()'ed critical * regions. * * - Soft interrupt processing context which cannot happen * while in a fpregs_lock()'ed critical region.
*/ if (!in_hardirq()) returntrue;
/* * In hard interrupt context it's safe when soft interrupts * are enabled, which means the interrupt did not hit in * a fpregs_lock()'ed critical region.
*/ return !softirq_count();
}
EXPORT_SYMBOL(irq_fpu_usable);
/* * Track AVX512 state use because it is known to slow the max clock * speed of the core.
*/ staticvoid update_avx_timestamp(struct fpu *fpu)
{
if (fpu->fpstate->regs.xsave.header.xfeatures & AVX512_TRACKING_MASK)
fpu->avx512_timestamp = jiffies;
}
/* * Save the FPU register state in fpu->fpstate->regs. The register state is * preserved. * * Must be called with fpregs_lock() held. * * The legacy FNSAVE instruction clears all FPU state unconditionally, so * register state has to be reloaded. That might be a pointless exercise * when the FPU is going to be used by another task right after that. But * this only affects 20+ years old 32bit systems and avoids conditionals all * over the place. * * FXSAVE and all XSAVE variants preserve the FPU register state.
*/ void save_fpregs_to_fpstate(struct fpu *fpu)
{ if (likely(use_xsave())) {
os_xsave(fpu->fpstate);
update_avx_timestamp(fpu); return;
}
if (likely(use_fxsr())) {
fxsave(&fpu->fpstate->regs.fxsave); return;
}
/* * Legacy FPU register saving, FNSAVE always clears FPU registers, * so we have to reload them from the memory state.
*/ asmvolatile("fnsave %[fp]; fwait" : [fp] "=m" (fpu->fpstate->regs.fsave));
frstor(&fpu->fpstate->regs.fsave);
}
void restore_fpregs_from_fpstate(struct fpstate *fpstate, u64 mask)
{ /* * AMD K7/K8 and later CPUs up to Zen don't save/restore * FDP/FIP/FOP unless an exception is pending. Clear the x87 state * here by setting it to fixed values. "m" is a random variable * that should be in L1.
*/ if (unlikely(static_cpu_has_bug(X86_BUG_FXSAVE_LEAK))) { asmvolatile( "fnclex\n\t" "emms\n\t" "fildl %[addr]"/* set F?P to defined value */
: : [addr] "m" (*fpstate));
}
if (use_xsave()) { /* * Dynamically enabled features are enabled in XCR0, but * usage requires also that the corresponding bits in XFD * are cleared. If the bits are set then using a related * instruction will raise #NM. This allows to do the * allocation of the larger FPU buffer lazy from #NM or if * the task has no permission to kill it which would happen * via #UD if the feature is disabled in XCR0. * * XFD state is following the same life time rules as * XSTATE and to restore state correctly XFD has to be * updated before XRSTORS otherwise the component would * stay in or go into init state even if the bits are set * in fpstate::regs::xsave::xfeatures.
*/
xfd_update_state(fpstate);
/* * Restoring state always needs to modify all features * which are in @mask even if the current task cannot use * extended features. * * So fpstate->xfeatures cannot be used here, because then * a feature for which the task has no permission but was * used by the previous task would not go into init state.
*/
mask = fpu_kernel_cfg.max_features & mask;
/* * KVM sets the FP+SSE bits in the XSAVE header when copying FPU state * to userspace, even when XSAVE is unsupported, so that restoring FPU * state on a different CPU that does support XSAVE can cleanly load * the incoming state using its natural XSAVE. In other words, KVM's * uABI size may be larger than this host's default size. Conversely, * the default size should never be larger than KVM's base uABI size; * all features that can expand the uABI size must be opt-in.
*/
gfpu->uabi_size = sizeof(struct kvm_xsave); if (WARN_ON_ONCE(fpu_user_cfg.default_size > gfpu->uabi_size))
gfpu->uabi_size = fpu_user_cfg.default_size;
/* * fpu_enable_guest_xfd_features - Check xfeatures against guest perm and enable * @guest_fpu: Pointer to the guest FPU container * @xfeatures: Features requested by guest CPUID * * Enable all dynamic xfeatures according to guest perm and requested CPUID. * * Return: 0 on success, error code otherwise
*/ int fpu_enable_guest_xfd_features(struct fpu_guest *guest_fpu, u64 xfeatures)
{
lockdep_assert_preemption_enabled();
/* Nothing to do if all requested features are already enabled. */
xfeatures &= ~guest_fpu->xfeatures; if (!xfeatures) return 0;
/** * fpu_sync_guest_vmexit_xfd_state - Synchronize XFD MSR and software state * * Must be invoked from KVM after a VMEXIT before enabling interrupts when * XFD write emulation is disabled. This is required because the guest can * freely modify XFD and the state at VMEXIT is not guaranteed to be the * same as the state on VMENTER. So software state has to be updated before * any operation which depends on it can take place. * * Note: It can be invoked unconditionally even when write emulation is * enabled for the price of a then pointless MSR read.
*/ void fpu_sync_guest_vmexit_xfd_state(void)
{ struct fpstate *fpstate = x86_task_fpu(current)->fpstate;
if (!cur_fps->is_confidential) { /* Includes XFD update */
restore_fpregs_from_fpstate(cur_fps, XFEATURE_MASK_FPSTATE);
} else { /* * XSTATE is restored by firmware from encrypted * memory. Make sure XFD state is correct while * running with guest fpstate
*/
xfd_update_state(cur_fps);
}
if (!cpu_feature_enabled(X86_FEATURE_XSAVE)) { if (ustate->xsave.header.xfeatures & ~XFEATURE_MASK_FPSSE) return -EINVAL; if (ustate->fxsave.mxcsr & ~mxcsr_feature_mask) return -EINVAL;
memcpy(&kstate->regs.fxsave, &ustate->fxsave, sizeof(ustate->fxsave)); return 0;
}
if (ustate->xsave.header.xfeatures & ~xcr0) return -EINVAL;
/* * Nullify @vpkru to preserve its current value if PKRU's bit isn't set * in the header. KVM's odd ABI is to leave PKRU untouched in this * case (all other components are eventually re-initialized).
*/ if (!(ustate->xsave.header.xfeatures & XFEATURE_MASK_PKRU))
vpkru = NULL;
/* Put sane initial values into the control registers. */ if (likely(kfpu_mask & KFPU_MXCSR) && boot_cpu_has(X86_FEATURE_XMM))
ldmxcsr(MXCSR_DEFAULT);
if (unlikely(kfpu_mask & KFPU_387) && boot_cpu_has(X86_FEATURE_FPU)) asmvolatile ("fninit");
}
EXPORT_SYMBOL_GPL(kernel_fpu_begin_mask);
void kernel_fpu_end(void)
{ /* Toggle kernel_fpu_allowed back to true: */
WARN_ON_FPU(this_cpu_read(kernel_fpu_allowed));
this_cpu_write(kernel_fpu_allowed, true);
if (!irqs_disabled())
fpregs_unlock();
}
EXPORT_SYMBOL_GPL(kernel_fpu_end);
/* * Sync the FPU register state to current's memory register state when the * current task owns the FPU. The hardware register state is preserved.
*/ void fpu_sync_fpstate(struct fpu *fpu)
{
WARN_ON_FPU(fpu != x86_task_fpu(current));
fpregs_lock();
trace_x86_fpu_before_save(fpu);
if (!test_thread_flag(TIF_NEED_FPU_LOAD))
save_fpregs_to_fpstate(fpu);
trace_x86_fpu_after_save(fpu);
fpregs_unlock();
}
staticinlineunsignedint init_fpstate_copy_size(void)
{ if (!use_xsave()) return fpu_kernel_cfg.default_size;
/* XSAVE(S) just needs the legacy and the xstate header part */ returnsizeof(init_fpstate.regs.xsave);
}
/* * Used in two places: * 1) Early boot to setup init_fpstate for non XSAVE systems * 2) fpu_alloc_guest_fpstate() which is invoked from KVM
*/ void fpstate_init_user(struct fpstate *fpstate)
{ if (!cpu_feature_enabled(X86_FEATURE_FPU)) {
fpstate_init_soft(&fpstate->regs.soft); return;
}
if (cpu_feature_enabled(X86_FEATURE_FXSR))
fpstate_init_fxstate(fpstate); else
fpstate_init_fstate(fpstate);
}
staticvoid __fpstate_reset(struct fpstate *fpstate)
{ /* * Supervisor features (and thus sizes) may diverge between guest * FPUs and host FPUs, as some supervisor features are supported * for guests despite not being utilized by the host. User * features and sizes are always identical, which allows for * common guest and userspace ABI. * * For the host, set XFD to the kernel's desired initialization * value. For guests, set XFD to its architectural RESET value.
*/ if (fpstate->is_guest) {
fpstate->size = guest_default_cfg.size;
fpstate->xfeatures = guest_default_cfg.features;
fpstate->xfd = 0;
} else {
fpstate->size = fpu_kernel_cfg.default_size;
fpstate->xfeatures = fpu_kernel_cfg.default_features;
fpstate->xfd = init_fpstate.xfd;
}
void fpstate_reset(struct fpu *fpu)
{ /* Set the fpstate pointer to the default fpstate */
fpu->fpstate = &fpu->__fpstate;
__fpstate_reset(fpu->fpstate);
/* Initialize the permission related info in fpu */
fpu->perm.__state_perm = fpu_kernel_cfg.default_features;
fpu->perm.__state_size = fpu_kernel_cfg.default_size;
fpu->perm.__user_state_size = fpu_user_cfg.default_size;
fpu->guest_perm.__state_perm = guest_default_cfg.features;
fpu->guest_perm.__state_size = guest_default_cfg.size; /* * User features and sizes are always identical between host and * guest FPUs, which allows for common guest and userspace ABI.
*/
fpu->guest_perm.__user_state_size = fpu_user_cfg.default_size;
}
spin_lock_irq(¤t->sighand->siglock); /* Fork also inherits the permissions of the parent */
dst_fpu->perm = src_fpu->perm;
dst_fpu->guest_perm = src_fpu->guest_perm;
spin_unlock_irq(¤t->sighand->siglock);
}
}
/* A passed ssp of zero will not cause any update */ staticint update_fpu_shstk(struct task_struct *dst, unsignedlong ssp)
{ #ifdef CONFIG_X86_USER_SHADOW_STACK struct cet_user_state *xstate;
/* If ssp update is not needed. */ if (!ssp) return 0;
/* * If there is a non-zero ssp, then 'dst' must be configured with a shadow * stack and the fpu state should be up to date since it was just copied * from the parent in fpu_clone(). So there must be a valid non-init CET * state location in the buffer.
*/ if (WARN_ON_ONCE(!xstate)) return 1;
xstate->user_ssp = (u64)ssp; #endif return 0;
}
/* Clone current's FPU state on fork */ int fpu_clone(struct task_struct *dst, u64 clone_flags, bool minimal, unsignedlong ssp)
{ /* * We allocate the new FPU structure right after the end of the task struct. * task allocation size already took this into account. * * This is safe because task_struct size is a multiple of cacheline size, * thus x86_task_fpu() will always be cacheline aligned as well.
*/ struct fpu *dst_fpu = (void *)dst + sizeof(*dst);
/* The new task's FPU state cannot be valid in the hardware. */
dst_fpu->last_cpu = -1;
fpstate_reset(dst_fpu);
if (!cpu_feature_enabled(X86_FEATURE_FPU)) return 0;
/* * Enforce reload for user space tasks and prevent kernel threads * from trying to save the FPU registers on context switch.
*/
set_tsk_thread_flag(dst, TIF_NEED_FPU_LOAD);
/* * No FPU state inheritance for kernel threads and IO * worker threads.
*/ if (minimal) { /* Clear out the minimal state */
memcpy(&dst_fpu->fpstate->regs, &init_fpstate.regs,
init_fpstate_copy_size()); return 0;
}
/* * If a new feature is added, ensure all dynamic features are * caller-saved from here!
*/
BUILD_BUG_ON(XFEATURE_MASK_USER_DYNAMIC != XFEATURE_MASK_XTILE_DATA);
/* * Save the default portion of the current FPU state into the * clone. Assume all dynamic features to be defined as caller- * saved, which enables skipping both the expansion of fpstate * and the copying of any dynamic state. * * Do not use memcpy() when TIF_NEED_FPU_LOAD is set because * copying is not valid when current uses non-default states.
*/
fpregs_lock(); if (test_thread_flag(TIF_NEED_FPU_LOAD))
fpregs_restore_userregs();
save_fpregs_to_fpstate(dst_fpu);
fpregs_unlock(); if (!(clone_flags & CLONE_THREAD))
fpu_inherit_perms(dst_fpu);
/* * Children never inherit PASID state. * Force it to have its init value:
*/ if (use_xsave())
dst_fpu->fpstate->regs.xsave.header.xfeatures &= ~XFEATURE_MASK_PASID;
/* * Update shadow stack pointer, in case it changed during clone.
*/ if (update_fpu_shstk(dst, ssp)) return 1;
trace_x86_fpu_copy_dst(dst_fpu);
return 0;
}
/* * While struct fpu is no longer part of struct thread_struct, it is still * allocated after struct task_struct in the "task_struct" kmem cache. But * since FPU is expected to be part of struct thread_struct, we have to * adjust for it here.
*/ void fpu_thread_struct_whitelist(unsignedlong *offset, unsignedlong *size)
{ /* The allocation follows struct task_struct. */
*offset = sizeof(struct task_struct) - offsetof(struct task_struct, thread);
*offset += offsetof(struct fpu, __fpstate.regs);
*size = fpu_kernel_cfg.default_size;
}
/* * Drops current FPU state: deactivates the fpregs and * the fpstate. NOTE: it still leaves previous contents * in the fpregs in the eager-FPU case. * * This function can be used in cases where we know that * a state-restore is coming: either an explicit one, * or a reschedule.
*/ void fpu__drop(struct task_struct *tsk)
{ struct fpu *fpu;
if (test_tsk_thread_flag(tsk, TIF_NEED_FPU_LOAD)) return;
fpu = x86_task_fpu(tsk);
preempt_disable();
if (fpu == x86_task_fpu(current)) { /* Ignore delayed exceptions from user space */ asmvolatile("1: fwait\n" "2:\n"
_ASM_EXTABLE(1b, 2b));
fpregs_deactivate(fpu);
}
trace_x86_fpu_dropped(fpu);
preempt_enable();
}
/* * Clear FPU registers by setting them up from the init fpstate. * Caller must do fpregs_[un]lock() around it.
*/ staticinlinevoid restore_fpregs_from_init_fpstate(u64 features_mask)
{ if (use_xsave())
os_xrstor(&init_fpstate, features_mask); elseif (use_fxsr())
fxrstor(&init_fpstate.regs.fxsave); else
frstor(&init_fpstate.regs.fsave);
pkru_write_default();
}
/* * Reset current->fpu memory state to the init values.
*/ staticvoid fpu_reset_fpstate_regs(void)
{ struct fpu *fpu = x86_task_fpu(current);
fpregs_lock();
__fpu_invalidate_fpregs_state(fpu); /* * This does not change the actual hardware registers. It just * resets the memory image and sets TIF_NEED_FPU_LOAD so a * subsequent return to usermode will reload the registers from the * task's memory image. * * Do not use fpstate_init() here. Just copy init_fpstate which has * the correct content already except for PKRU. * * PKRU handling does not rely on the xstate when restoring for * user space as PKRU is eagerly written in switch_to() and * flush_thread().
*/
memcpy(&fpu->fpstate->regs, &init_fpstate.regs, init_fpstate_copy_size());
set_thread_flag(TIF_NEED_FPU_LOAD);
fpregs_unlock();
}
/* * Reset current's user FPU states to the init states. current's * supervisor states, if any, are not modified by this function. The * caller guarantees that the XSTATE header in memory is intact.
*/ void fpu__clear_user_states(struct fpu *fpu)
{
WARN_ON_FPU(fpu != x86_task_fpu(current));
fpregs_lock(); if (!cpu_feature_enabled(X86_FEATURE_FPU)) {
fpu_reset_fpstate_regs();
fpregs_unlock(); return;
}
/* * Ensure that current's supervisor states are loaded into their * corresponding registers.
*/ if (xfeatures_mask_supervisor() &&
!fpregs_state_valid(fpu, smp_processor_id()))
os_xrstor_supervisor(fpu->fpstate);
/* Ensure XFD state is in sync before reloading XSTATE */
xfd_update_state(fpu->fpstate);
/* Reset user states in registers. */
restore_fpregs_from_init_fpstate(XFEATURE_MASK_USER_RESTORE);
/* * Now all FPU registers have their desired values. Inform the FPU * state machine that current's FPU registers are in the hardware * registers. The memory image does not need to be updated because * any operation relying on it has to save the registers first when * current's FPU is marked active.
*/
fpregs_mark_activate();
fpregs_unlock();
}
void fpu_flush_thread(void)
{
fpstate_reset(x86_task_fpu(current));
fpu_reset_fpstate_regs();
} /* * Load FPU context before returning to userspace.
*/ void switch_fpu_return(void)
{ if (!static_cpu_has(X86_FEATURE_FPU)) return;
void fpregs_lock_and_load(void)
{ /* * fpregs_lock() only disables preemption (mostly). So modifying state * in an interrupt could screw up some in progress fpregs operation. * Warn about it.
*/
WARN_ON_ONCE(!irq_fpu_usable());
WARN_ON_ONCE(current->flags & PF_KTHREAD);
fpregs_lock();
fpregs_assert_state_consistent();
if (test_thread_flag(TIF_NEED_FPU_LOAD))
fpregs_restore_userregs();
}
#ifdef CONFIG_X86_DEBUG_FPU /* * If current FPU state according to its tracking (loaded FPU context on this * CPU) is not valid then we must have TIF_NEED_FPU_LOAD set so the context is * loaded on return to userland.
*/ void fpregs_assert_state_consistent(void)
{ struct fpu *fpu = x86_task_fpu(current);
int fpu__exception_code(struct fpu *fpu, int trap_nr)
{ int err;
if (trap_nr == X86_TRAP_MF) { unsignedshort cwd, swd; /* * (~cwd & swd) will mask out exceptions that are not set to unmasked * status. 0x3f is the exception bits in these regs, 0x200 is the * C1 reg you need in case of a stack fault, 0x040 is the stack * fault bit. We should only be taking one exception at a time, * so if this combination doesn't produce any single exception, * then we have a bad program that isn't synchronizing its FPU usage * and it will suffer the consequences since we won't be able to * fully reproduce the context of the exception.
*/ if (boot_cpu_has(X86_FEATURE_FXSR)) {
cwd = fpu->fpstate->regs.fxsave.cwd;
swd = fpu->fpstate->regs.fxsave.swd;
} else {
cwd = (unsignedshort)fpu->fpstate->regs.fsave.cwd;
swd = (unsignedshort)fpu->fpstate->regs.fsave.swd;
}
err = swd & ~cwd;
} else { /* * The SIMD FPU exceptions are handled a little differently, as there * is only a single status/control register. Thus, to determine which * unmasked exception was caught we must mask the exception mask bits * at 0x1f80, and then use these to mask the exception bits at 0x3f.
*/ unsignedshort mxcsr = MXCSR_DEFAULT;
if (boot_cpu_has(X86_FEATURE_XMM))
mxcsr = fpu->fpstate->regs.fxsave.mxcsr;
err = ~(mxcsr >> 7) & mxcsr;
}
if (err & 0x001) { /* Invalid op */ /* * swd & 0x240 == 0x040: Stack Underflow * swd & 0x240 == 0x240: Stack Overflow * User must clear the SF bit (0x40) if set
*/ return FPE_FLTINV;
} elseif (err & 0x004) { /* Divide by Zero */ return FPE_FLTDIV;
} elseif (err & 0x008) { /* Overflow */ return FPE_FLTOVF;
} elseif (err & 0x012) { /* Denormal, Underflow */ return FPE_FLTUND;
} elseif (err & 0x020) { /* Precision */ return FPE_FLTRES;
}
/* * If we're using IRQ 13, or supposedly even some trap * X86_TRAP_MF implementations, it's possible * we get a spurious trap, which is not an error.
*/ return 0;
}
/* * Initialize register state that may prevent from entering low-power idle. * This function will be invoked from the cpuidle driver only when needed.
*/
noinstr void fpu_idle_fpregs(void)
{ /* Note: AMX_TILE being enabled implies XGETBV1 support */ if (cpu_feature_enabled(X86_FEATURE_AMX_TILE) &&
(xfeatures_in_use() & XFEATURE_MASK_XTILE)) {
tile_release();
__this_cpu_write(fpu_fpregs_owner_ctx, NULL);
}
}
Messung V0.5
¤ Dauer der Verarbeitung: 0.18 Sekunden
(vorverarbeitet)
¤
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.