xref: /freebsd-src/share/man/man9/cpu_machdep.9 (revision 9c87cbbcaaedbc4e07d5c9d0248bf76f72531f57)

.\"-
.\" SPDX-License-Identifier: BSD-2-Clause
.\"
.\" Copyright (c) 2024 (holder)
.\"
.\" This software was developed by SRI International, the University of
.\" Cambridge Computer Laboratory (Department of Computer Science and
.\" Technology), and Capabilities Limited under Defense Advanced Research
.\" Projects Agency (DARPA) Contract No. FA8750-24-C-B047 ("DEC").
.\"
.Dd January 3, 2025
.Dt cpu_machdep 9
.Os
.Sh NAME
.Nm cpu_machdep ,
.Nm cpu_copy_thread ,
.Nm cpu_exec_vmspace_reuse ,
.Nm cpu_exit ,
.Nm cpu_fetch_syscall_args ,
.Nm cpu_fork ,
.Nm cpu_fork_kthread_handler ,
.Nm cpu_idle ,
.Nm cpu_idle_wakeup ,
.Nm cpu_procctl ,
.Nm cpu_set_syscall_retval ,
.Nm cpu_set_upcall ,
.Nm cpu_set_user_tls ,
.Nm cpu_switch ,
.Nm cpu_sync_core ,
.Nm cpu_thread_alloc ,
.Nm cpu_thread_clean ,
.Nm cpu_thread_exit ,
.Nm cpu_thread_free ,
.Nm cpu_throw
.Nd machine-dependent interfaces to handle CPU and thread state
.Sh SYNOPSIS
.In sys/proc.h
.In sys/ptrace.h
.Ft void
.Fn cpu_copy_thread "struct thread *td" "struct thread *td0"
.Ft bool
.Fn cpu_exec_vmspace_reuse "struct proc *p" "struct vm_map *map"
.Ft void
.Fn cpu_exit "struct thread *td"
.Ft int
.Fn cpu_fetch_syscall_args "struct thread *td"
.Ft void
.Fo cpu_fork
.Fa "struct thread *td1" "struct proc *p2" "struct thread *td2" "int flags"
.Fc
.Ft void
.Fo cpu_fork_kthread_handler
.Fa "struct thread *td" "void (*func)(void *)" "void *arg"
.Fc
.Ft void
.Fn cpu_idle "int busy"
.Ft int
.Fn cpu_idle_wakeup "int cpu"
.Ft int
.Fo cpu_procctl
.Fa "struct thread *td" "int idtype" "id_t id" "int com" "void *data"
.Fc
.Ft int
.Fn cpu_ptrace "struct thread *_td" "int req" "void *addr" "int data"
.Ft void
.Fn cpu_set_syscall_retval "struct thread *td" "int error"
.Ft int
.Fo cpu_set_upcall
.Fa "struct thread *td" "void (*entry)(void *)" "void *arg" "stack_t *stack"
.Fc
.Ft int
.Fn cpu_set_user_tls "struct thread *td" "void *tls_base"
.Ft void
.Fn cpu_switch "struct thread *old" "struct thread *new" "struct mtx *mtx"
.Ft void
.Fn cpu_sync_core "void"
.Ft void
.Fn cpu_thread_alloc "struct thread *td"
.Ft void
.Fn cpu_thread_clean "struct thread *td"
.Ft void
.Fn cpu_thread_exit "struct thread *td"
.Ft void
.Fn cpu_thread_free "struct thread *td"
.Ft void
.Fn cpu_throw "struct thread *old" "struct thread *new"
.Sh DESCRIPTION
These functions provide architecture-specific implementations of
machine-independent abstractions.
.Pp
.Fn cpu_exec_vmspace_reuse
returns true if
.Fn exec_new_vmspace
can reuse an existing
.Vt struct vmspace
.Pq Fa map
for the process
.Fa p
during
.Xr execve 2 .
This is only invoked if
.Fa map
is not shared with any other consumers.
If this returns false,
.Fn exec_new_vmspace
will create a new
.Vt struct vmspace .
.Pp
.Fn cpu_exit
releases machine-dependent resources other than the address space for the
process containing
.Fa td
during process exit.
.Pp
.Fn cpu_fork
copies and updates machine-dependent state
(for example, the pcb and user registers) from the forking thread
.Fa td1
in an existing process to the new thread
.Fa td2
in the new process
.Fa p2 .
This function must set up the new thread's kernel stack and pcb so that
.Fa td2
calls
.Fn fork_exit
when it begins execution passing a pointer to
.Fn fork_return
as the
.Fa callout
argument and
.Fa td2
as the
.Fa arg
argument.
.Pp
.Fn cpu_fork_kthread_handler
adjusts a new thread's initial pcb and/or kernel stack to pass
.Fa func
and
.Fa arg
as the
.Fa callout
and
.Fa arg
arguments to
.Fn fork_exit .
This must be called before a new thread is scheduled to run and is
used to set the
.Dq main
function for kernel threads.
.Pp
.Fn cpu_copy_thread
copies machine-dependent state (for example, the pcb and user registers) from
.Fa td
to
.Fa td0
when creating a new thread in the same process.
This function must set up the new thread's kernel stack and pcb so that
.Fa td0
calls
.Fn fork_exit
when it begins execution passing a pointer to
.Fn fork_return
as the
.Fa callout
argument and
.Fa td0
as the
.Fa arg
argument.
.Pp
.Fn cpu_set_upcall
updates a new thread's initial user register state to call
.Fa entry
with
.Fa arg
as the sole argument using the user stack described in
.Fa stack .
.Pp
.Fn cpu_set_user_tls
sets a new thread's initial user thread pointer register to
reference the user TLS base pointer
.Fa tls_base .
.Pp
.Fn cpu_fetch_syscall_args
fetches the current system call arguments for the native FreeBSD ABI from the
current thread's user register state and/or user stack.
The arguments are saved in the
.Fa td_sa
member of
.Fa td .
.Pp
.Fn cpu_set_syscall_retval
updates the user register state for
.Fa td
to store system call error and return values.
If
.Fa error
is 0,
indicate success and return the two values in
.Fa td_retval .
If
.Fa error
is
.Dv ERESTART,
adjust the user PC to re-invoke the current system call after returning
to user mode.
If
.Fa error
is
.Dv EJUSTRETURN ,
leave the current user register state unchanged.
For any other value of
.Fa error ,
indicate error and return
.Fa error
as the error code.
.Pp
.Fn cpu_idle
waits for the next interrupt to occur on the current CPU.
If an architecture supports low power idling,
this function should place the CPU into a low power state while waiting.
.Fa busy
is a hint from the scheduler.
If
.Fa busy
is non-zero,
the scheduler expects a short sleep,
so the CPU should prefer low-latency over maximum power savings.
If
.Fa busy
is zero,
the CPU should maximumize power savings including deferring unnecessary
clock interrupts via
.Fn cpu_idleclock .
.Pp
.Fn cpu_idle_wakeup
awakens the idle CPU with the ID
.Fa cpu
from a low-power state.
.Pp
.Fn cpu_procctl
handles any machine-dependent
.Xr procctl 2
requests.
.Pp
.Fn cpu_ptrace
handles any machine-dependent
.Xr ptrace 2
requests.
.Pp
.Fn cpu_switch
switches the current CPU between threads by swapping register state.
This function saves the current CPU register state in the pcb of
.Fa old
and loads register values from the pcb of
.Fa new
before returning.
While the pcb generally contains caller-save kernel register state,
it can also contain user registers that are not saved in the trapframe.
.Pp
After saving the current CPU register state of
.Fa old ,
.Fn cpu_switch
stores
.Fa mtx
in the
.Fa td_lock
member of
.Fa old
transferring ownership of the old thread.
No data belonging to
.Fa old
can be accessed after that store.
Specifically, the old thread's kernel stack must not be accessed after
this point.
.Pp
When
.Dv SCHED_ULE
is being used,
this function must wait (via spinning) for the
.Fa td_lock
member of
.Fa new
to change to a value not equal to
.Va &blocked_lock
before loading register values from
.Fa new
or accessing its kernel stack.
.Pp
From the caller's perspective,
.Fn cpu_switch
returns when
.Fa old
is rescheduled in the future,
possibly on a different CPU.
However, the implementation of
.Fn cpu_switch
returns immediately on the same CPU into the previously-saved context of
.Fa new .
.Pp
.Fn cpu_throw
is similar to
.Fn cpu_switch
but does not save any state for
.Fa old
or write to the old thread's
.Fa td_lock
member.
.Pp
.Fn cpu_sync_core
ensures that all possible speculation and out-of-order execution is
serialized on the current CPU.
Note that this is called from an IPI handler so only has to handle
additional serialization beyond that provided by handling an IPI.
.Ss Thread Object Lifecycle
These functions support the management of machine-dependent thread
state in conjunction with a thread object's lifecycle.
.Pp
The general model is that a thread object is allocated each time a
new kernel thread is created either by system calls like
.Xr fork 2
or
.Xr thr_new 2
or when kernel-only threads are created via
.Xr kproc_create 9 ,
.Xr kproc_kthread_add 9 ,
or
.Xr kthread_add 9 .
When a kernel thread exits,
the thread object is freed.
However, there is one special case to support an optimization where each
free process object caches a thread object.
When a process exits, the last thread object is not freed but remains
attached to the process.
When the process object is later reused for a new process in
.Xr fork 2 ,
the kernel recycles that last thread object and uses it as the initial
thread in the new process.
When a thread is recycled, some of the steps in the thread allocation
and free cycle are skipped as an optimization.
.Pp
.Fn cpu_thread_alloc
initializes machine-dependent fields in
.Fa td
after allocating a new kernel stack.
This function typically sets the
.Fa td_pcb
and initial
.Fa td_frame
pointers.
.Fn cpu_thread_alloc
is called both when allocating a new thread object and
when a recycled thread allocates a new kernel stack.
Note that this function is
.Em not
called if a recycled thread reuses its existing kernel stack.
.Pp
.Fn cpu_thread_clean
releases any machine-dependent resources for the last thread in a
process during
.Xr wait 2 .
The thread is a candidate for recycling so should be reset to run as a
new thread in case it is recycled by a future
.Xr fork 2 .
.Pp
.Fn cpu_thread_exit
cleans any machine-dependent state in
.Fa td
while it is exiting.
This is called by the exiting thread so cannot free state needed during
in-kernel execution.
.Pp
.Fn cpu_thread_free
releases any machine-dependent state in
.Fa td
when it is being freed.
This is called for any thread that was not the last thread in a process
once it has finished execution.
.Sh SEE ALSO
.Xr fork 2 ,
.Xr procctl 2 ,
.Xr ptrace 2 ,
.Xr thr_new 2 ,
.Xr wait 2 ,
.Xr kproc_create 9 ,
.Xr kproc_kthread_add 9 ,
.Xr kthread_add 9 ,
.Xr mi_switch 9
.Sh AUTHORS
This manual page was
developed by SRI International, the University of Cambridge Computer
Laboratory (Department of Computer Science and Technology), and
Capabilities Limited under contract
.Pq FA8750-24-C-B047
.Pq Do DEC Dc .