Implement system call sched_get/set_affinity

kernel/src/sched/priority_scheduler.rs

@@ -1,8 +1,9 @@
 // SPDX-License-Identifier: MPL-2.0
 
+use core::sync::atomic::Ordering;
+
 use ostd::{
     cpu::{num_cpus, CpuId, CpuSet, PinCurrentCpu},
-    sync::PreemptDisabled,
     task::{
         scheduler::{
             info::CommonSchedInfo, inject_scheduler, EnqueueFlags, LocalRunQueue, Scheduler,
@@ -254,11 +255,11 @@ impl PreemptSchedInfo for Thread {
     const LOWEST_TASK_PRIORITY: Priority = Priority::new(PriorityRange::new(PriorityRange::MAX));
 
     fn priority(&self) -> Priority {
-        self.priority()
+        self.atomic_priority().load(Ordering::Relaxed)
     }
 
-    fn cpu_affinity(&self) -> SpinLockGuard<CpuSet, PreemptDisabled> {
+    fn cpu_affinity(&self) -> CpuSet {
-        self.lock_cpu_affinity()
+        self.atomic_cpu_affinity().load()
     }
 }
 
@@ -268,7 +269,7 @@ trait PreemptSchedInfo {
 
     fn priority(&self) -> Priority;
 
-    fn cpu_affinity(&self) -> SpinLockGuard<CpuSet, PreemptDisabled>;
+    fn cpu_affinity(&self) -> CpuSet;
 
     fn is_real_time(&self) -> bool {
         self.priority() < Self::REAL_TIME_TASK_PRIORITY

kernel/src/syscall/arch/riscv.rs

@@ -79,7 +79,7 @@ use crate::syscall::{
     rt_sigpending::sys_rt_sigpending,
     rt_sigprocmask::sys_rt_sigprocmask,
     rt_sigsuspend::sys_rt_sigsuspend,
-    sched_getaffinity::sys_sched_getaffinity,
+    sched_affinity::{sys_sched_getaffinity, sys_sched_setaffinity},
     sched_yield::sys_sched_yield,
     semctl::sys_semctl,
     semget::sys_semget,
@@ -191,6 +191,7 @@ impl_syscall_nums_and_dispatch_fn! {
     SYS_SETITIMER = 103          => sys_setitimer(args[..3]);
     SYS_TIMER_CREATE = 107       => sys_timer_create(args[..3]);
     SYS_TIMER_DELETE = 111       => sys_timer_delete(args[..1]);
+    SYS_SCHED_SETAFFINITY = 122  => sys_sched_setaffinity(args[..3]);
     SYS_SCHED_GETAFFINITY = 123  => sys_sched_getaffinity(args[..3]);
     SYS_SCHED_YIELD = 124        => sys_sched_yield(args[..0]);
     SYS_KILL = 129               => sys_kill(args[..2]);

kernel/src/syscall/arch/x86.rs

@@ -86,7 +86,7 @@ use crate::syscall::{
     rt_sigprocmask::sys_rt_sigprocmask,
     rt_sigreturn::sys_rt_sigreturn,
     rt_sigsuspend::sys_rt_sigsuspend,
-    sched_getaffinity::sys_sched_getaffinity,
+    sched_affinity::{sys_sched_getaffinity, sys_sched_setaffinity},
     sched_yield::sys_sched_yield,
     select::sys_select,
     semctl::sys_semctl,
@@ -263,6 +263,7 @@ impl_syscall_nums_and_dispatch_fn! {
     SYS_GETTID = 186           => sys_gettid(args[..0]);
     SYS_TIME = 201             => sys_time(args[..1]);
     SYS_FUTEX = 202            => sys_futex(args[..6]);
+    SYS_SCHED_SETAFFINITY = 203 => sys_sched_setaffinity(args[..3]);
     SYS_SCHED_GETAFFINITY = 204 => sys_sched_getaffinity(args[..3]);
     SYS_EPOLL_CREATE = 213     => sys_epoll_create(args[..1]);
     SYS_GETDENTS64 = 217       => sys_getdents64(args[..3]);

kernel/src/syscall/mod.rs

@@ -93,7 +93,7 @@ mod rt_sigpending;
 mod rt_sigprocmask;
 mod rt_sigreturn;
 mod rt_sigsuspend;
-mod sched_getaffinity;
+mod sched_affinity;
 mod sched_yield;
 mod select;
 mod semctl;

kernel/src/syscall/sched_affinity.rs

@@ -0,0 +1,130 @@
+// SPDX-License-Identifier: MPL-2.0
+
+use core::{cmp, mem};
+
+use ostd::cpu::{num_cpus, CpuId, CpuSet};
+
+use super::SyscallReturn;
+use crate::{prelude::*, process::posix_thread::thread_table, thread::Tid};
+
+pub fn sys_sched_getaffinity(
+    tid: Tid,
+    cpuset_size: usize,
+    cpu_set_ptr: Vaddr,
+    ctx: &Context,
+) -> Result<SyscallReturn> {
+    let cpu_set = match tid {
+        0 => ctx.thread.atomic_cpu_affinity().load(),
+        _ => match thread_table::get_thread(tid) {
+            Some(thread) => thread.atomic_cpu_affinity().load(),
+            None => return Err(Error::with_message(Errno::ESRCH, "thread does not exist")),
+        },
+    };
+
+    let bytes_written = write_cpu_set_to(ctx.get_user_space(), &cpu_set, cpuset_size, cpu_set_ptr)?;
+
+    Ok(SyscallReturn::Return(bytes_written as isize))
+}
+
+// TODO: The manual page of `sched_setaffinity` says that if the thread is not
+// running on the CPU specified in the affinity mask, it would be migrated to
+// one of the CPUs specified in the mask. We currently do not support this
+// feature as the scheduler is not ready for migration yet.
+pub fn sys_sched_setaffinity(
+    tid: Tid,
+    cpuset_size: usize,
+    cpu_set_ptr: Vaddr,
+    ctx: &Context,
+) -> Result<SyscallReturn> {
+    let user_cpu_set = read_cpu_set_from(ctx.get_user_space(), cpuset_size, cpu_set_ptr)?;
+
+    match tid {
+        0 => ctx.thread.atomic_cpu_affinity().store(&user_cpu_set),
+        _ => match thread_table::get_thread(tid) {
+            Some(thread) => {
+                thread.atomic_cpu_affinity().store(&user_cpu_set);
+            }
+            None => return Err(Error::with_message(Errno::ESRCH, "thread does not exist")),
+        },
+    }
+
+    Ok(SyscallReturn::Return(0))
+}
+
+// Linux uses `DECLARE_BITMAP` for `cpu_set_t`, inside which each part is a
+// `long`. We use the same scheme to ensure byte endianness compatibility.
+type Part = u64;
+const SIZE_OF_PART: usize = mem::size_of::<Part>();
+const CPUS_IN_PART: usize = SIZE_OF_PART * 8;
+
+fn read_cpu_set_from(
+    uspace: CurrentUserSpace,
+    cpuset_size: usize,
+    cpu_set_ptr: Vaddr,
+) -> Result<CpuSet> {
+    if cpuset_size == 0 {
+        return Err(Error::with_message(Errno::EINVAL, "invalid cpuset size"));
+    }
+
+    let num_cpus = num_cpus();
+
+    let mut ret_set = CpuSet::new_empty();
+
+    let nr_parts_to_read = cmp::min(cpuset_size / SIZE_OF_PART, num_cpus.div_ceil(CPUS_IN_PART));
+    for part_id in 0..nr_parts_to_read {
+        let user_part: Part = uspace.read_val(cpu_set_ptr + part_id * SIZE_OF_PART)?;
+        for bit_id in 0..CPUS_IN_PART {
+            if user_part & (1 << bit_id) != 0 {
+                // If the CPU ID is invalid, just ignore it.
+                let Ok(cpu_id) = CpuId::try_from(part_id * CPUS_IN_PART + bit_id) else {
+                    continue;
+                };
+                ret_set.add(cpu_id);
+            }
+        }
+    }
+
+    if ret_set.is_empty() {
+        return Err(Error::with_message(Errno::EINVAL, "empty cpuset"));
+    }
+
+    Ok(ret_set)
+}
+
+// Returns the number of bytes written.
+fn write_cpu_set_to(
+    uspace: CurrentUserSpace,
+    cpu_set: &CpuSet,
+    cpuset_size: usize,
+    cpu_set_ptr: Vaddr,
+) -> Result<usize> {
+    if cpuset_size == 0 {
+        return Err(Error::with_message(Errno::EINVAL, "invalid cpuset size"));
+    }
+
+    let num_cpus = num_cpus();
+
+    let nr_parts_to_write = cmp::min(cpuset_size / SIZE_OF_PART, num_cpus.div_ceil(CPUS_IN_PART));
+    let mut user_part: Part = 0;
+    let mut part_idx = 0;
+    for cpu_id in cpu_set.iter() {
+        let id = cpu_id.as_usize();
+        while part_idx < cmp::min(id / CPUS_IN_PART, nr_parts_to_write) {
+            uspace.write_val(cpu_set_ptr + part_idx * SIZE_OF_PART, &user_part)?;
+            user_part = 0;
+            part_idx += 1;
+        }
+        if part_idx >= nr_parts_to_write {
+            break;
+        }
+        user_part |= 1 << (id % CPUS_IN_PART);
+    }
+
+    while part_idx < nr_parts_to_write {
+        uspace.write_val(cpu_set_ptr + part_idx * SIZE_OF_PART, &user_part)?;
+        user_part = 0;
+        part_idx += 1;
+    }
+
+    Ok(nr_parts_to_write * SIZE_OF_PART)
+}

kernel/src/syscall/sched_getaffinity.rs

@@ -1,70 +0,0 @@
-// SPDX-License-Identifier: MPL-2.0
-
-use core::{cmp, mem};
-
-use super::SyscallReturn;
-use crate::{
-    prelude::*,
-    process::{process_table, Pid},
-};
-
-fn get_num_cpus() -> usize {
-    // TODO: Properly determine the number of available CPUs
-    // This could be through a system configuration query.
-    1
-}
-
-pub fn sys_sched_getaffinity(
-    pid: Pid,
-    cpuset_size: usize,
-    cpu_set_ptr: Vaddr,
-    ctx: &Context,
-) -> Result<SyscallReturn> {
-    let num_cpus = get_num_cpus();
-
-    if cpuset_size < core::mem::size_of::<cpu_set_t>() {
-        return Err(Error::with_message(Errno::EINVAL, "invalid cpuset size"));
-    }
-
-    match pid {
-        0 => {
-            // TODO: Get the current thread's CPU affinity
-            // Placeholder for future implementation.
-        }
-        _ => {
-            match process_table::get_process(pid) {
-                Some(_process) => { /* Placeholder if process-specific logic needed */ }
-                None => return Err(Error::with_message(Errno::ESRCH, "process does not exist")),
-            }
-        }
-    }
-
-    let dummy_cpu_set = cpu_set_t::new(num_cpus);
-
-    ctx.get_user_space()
-        .write_val(cpu_set_ptr, &dummy_cpu_set)?;
-
-    Ok(SyscallReturn::Return(0))
-}
-
-const CPU_SETSIZE: usize = 1024; // Max number of CPU bits.
-const __NCPUBITS: usize = 8 * mem::size_of::<usize>();
-
-#[derive(Debug, Clone, Copy, Pod)]
-#[repr(C, packed)]
-struct cpu_set_t {
-    __bits: [usize; CPU_SETSIZE / __NCPUBITS],
-}
-
-impl cpu_set_t {
-    /// Creates a new cpu_set_t representing available CPUs.
-    fn new(num_cpus: usize) -> Self {
-        let mut bits = [0usize; CPU_SETSIZE / __NCPUBITS];
-
-        for cpu in 0..cmp::min(num_cpus, CPU_SETSIZE) {
-            bits[cpu / __NCPUBITS] |= 1 << (cpu % __NCPUBITS);
-        }
-
-        Self { __bits: bits }
-    }
-}

kernel/src/thread/mod.rs

@@ -4,7 +4,10 @@
 
 use core::sync::atomic::Ordering;
 
-use ostd::{cpu::CpuSet, sync::PreemptDisabled, task::Task};
+use ostd::{
+    cpu::{AtomicCpuSet, CpuSet},
+    task::Task,
+};
 
 use self::status::{AtomicThreadStatus, ThreadStatus};
 use crate::{
@@ -34,8 +37,8 @@ pub struct Thread {
     status: AtomicThreadStatus,
     /// Thread priority
     priority: AtomicPriority,
-    /// Thread cpu affinity
+    /// Thread CPU affinity
-    cpu_affinity: SpinLock<CpuSet>,
+    cpu_affinity: AtomicCpuSet,
 }
 
 impl Thread {
@@ -52,7 +55,7 @@ impl Thread {
             data: Box::new(data),
             status: AtomicThreadStatus::new(status),
             priority: AtomicPriority::new(priority),
-            cpu_affinity: SpinLock::new(cpu_affinity),
+            cpu_affinity: AtomicCpuSet::new(cpu_affinity),
         }
     }
 
@@ -67,6 +70,11 @@ impl Thread {
             .cloned()
     }
 
+    /// Returns the task associated with this thread.
+    pub fn task(&self) -> Arc<Task> {
+        self.task.upgrade().unwrap()
+    }
+
     /// Gets the Thread from task's data.
     ///
     /// # Panics
@@ -116,14 +124,9 @@ impl Thread {
         self.priority.store(new_priority, Ordering::Relaxed)
     }
 
-    /// Acquires the lock of cpu affinity.
+    /// Returns the reference to the atomic CPU affinity.
-    pub fn lock_cpu_affinity(&self) -> SpinLockGuard<CpuSet, PreemptDisabled> {
+    pub fn atomic_cpu_affinity(&self) -> &AtomicCpuSet {
-        self.cpu_affinity.lock()
+        &self.cpu_affinity
-    }
-
-    /// Updates the cpu affinity with the new value.
-    pub fn set_cpu_affinity(&self, new_cpu_affinity: CpuSet) {
-        *self.cpu_affinity.lock() = new_cpu_affinity;
     }
 
     pub fn yield_now() {

ostd/src/cpu/set.rs

@@ -80,6 +80,11 @@ impl CpuSet {
             .sum()
     }
 
+    /// Returns true if the set is empty.
+    pub fn is_empty(&self) -> bool {
+        self.bits.iter().all(|part| *part == 0)
+    }
+
     /// Adds all CPUs to the set.
     pub fn add_all(&mut self) {
         self.bits.fill(!0);
@@ -92,6 +97,8 @@ impl CpuSet {
     }
 
     /// Iterates over the CPUs in the set.
+    ///
+    /// The order of the iteration is guaranteed to be in ascending order.
     pub fn iter(&self) -> impl Iterator<Item = CpuId> + '_ {
         self.bits.iter().enumerate().flat_map(|(part_idx, &part)| {
             (0..BITS_PER_PART).filter_map(move |bit_idx| {
@@ -165,9 +172,9 @@ impl AtomicCpuSet {
     ///
     /// This operation can only be done in the [`Ordering::Relaxed`] memory
     /// order. It cannot be used to synchronize anything between CPUs.
-    pub fn store(&self, value: CpuSet) {
+    pub fn store(&self, value: &CpuSet) {
-        for (part, new_part) in self.bits.iter().zip(value.bits) {
+        for (part, new_part) in self.bits.iter().zip(value.bits.iter()) {
-            part.store(new_part, Ordering::Relaxed);
+            part.store(*new_part, Ordering::Relaxed);
         }
     }
 
@@ -260,7 +267,7 @@ mod test {
                 }
             }
 
-            atomic_set.store(CpuSet::with_capacity_val(test_num_cpus, 0));
+            atomic_set.store(&CpuSet::with_capacity_val(test_num_cpus, 0));
 
             for cpu_id in test_all_iter() {
                 assert!(!atomic_set.contains(cpu_id, Ordering::Relaxed));