i2p/haiku-beta6
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

kernel/smp: Move the per-CPU SMP message mailbox into cpu_ent.

Browse Source 
Putting all the per-CPU messages in one array means they'll all
be on the same cache lines. The cpu_ent structure is cache-line-aligned,
so this means they'll be on separate cache lines.
This commit is contained in:
Augustin Cavalier
2026-04-27 15:14:14 -04:00
parent f0b91b00ac
commit 747a839b95
2 changed files with 14 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
headers/private/kernel/cpu.h
+1
View File
@@ -64,6 +64,7 @@ typedef struct CACHE_LINE_ALIGN cpu_ent {
bigtime_t last_kernel_time;
bigtime_t last_user_time;
struct smp_msg* cpu_msg;
int32 ici_counter;
// used in the kernel debugger
src/system/kernel/smp.cpp
+13 -14
View File
@@ -97,8 +97,6 @@ static struct smp_msg* sFreeMessages = NULL;
static int32 sFreeMessageCount = 0;
static rw_spinlock sFreeMessageSpinlock = B_RW_SPINLOCK_INITIALIZER;
static struct smp_msg* sCPUMessages[SMP_MAX_CPUS] = { NULL, };
static struct smp_msg* sBroadcastMessages = NULL;
static rw_spinlock sBroadcastMessageSpinlock = B_RW_SPINLOCK_INITIALIZER;
static int32 sBroadcastMessageCounter;
@@ -214,7 +212,7 @@ update_lock_held(spinlock* lock)
#endif // B_DEBUG_SPINLOCK_CONTENTION
int
static int
dump_ici_messages(int argc, char** argv)
{
// count broadcast messages
@@ -239,14 +237,14 @@ dump_ici_messages(int argc, char** argv)
// count per-CPU messages
for (int32 i = 0; i < sNumCPUs; i++) {
count = 0;
message = sCPUMessages[i];
message = gCPU[i].cpu_msg;
while (message != NULL) {
count++;
message = message->next;
}
kprintf("CPU %" B_PRId32 " messages: %" B_PRId32 ", first: %p\n", i,
count, sCPUMessages[i]);
count, gCPU[i].cpu_msg);
}
return 0;
@@ -733,12 +731,12 @@ return_free_message(struct smp_msg* msg)
static void
prepend_message(struct smp_msg*& listHead, struct smp_msg* msg)
prepend_message(struct smp_msg*& mailbox, struct smp_msg* msg)
{
while (true) {
struct smp_msg* next = atomic_pointer_get(&listHead);
struct smp_msg* next = atomic_pointer_get(&mailbox);
msg->next = next;
if (atomic_pointer_test_and_set(&listHead, msg, next) == next)
if (atomic_pointer_test_and_set(&mailbox, msg, next) == next)
break;
cpu_pause();
}
@@ -751,15 +749,16 @@ check_for_message(int currentCPU, mailbox_source& sourceMailbox)
if (!sICIEnabled)
return NULL;
struct smp_msg* msg = atomic_pointer_get(&sCPUMessages[currentCPU]);
struct smp_msg** mailbox = &gCPU[currentCPU].cpu_msg;
struct smp_msg* msg = atomic_pointer_get(mailbox);
if (msg != NULL) {
// since only this CPU ever dequeues, we can just use atomics
while (true) {
if (atomic_pointer_test_and_set(&sCPUMessages[currentCPU], msg->next, msg) == msg)
if (atomic_pointer_test_and_set(mailbox, msg->next, msg) == msg)
break;
cpu_pause();
msg = atomic_pointer_get(&sCPUMessages[currentCPU]);
msg = atomic_pointer_get(mailbox);
ASSERT(msg != NULL);
}
@@ -779,7 +778,7 @@ check_for_message(int currentCPU, mailbox_source& sourceMailbox)
continue;
}
// mark it so we wont try to process this one again
// mark it so we won't try to process this one again
msg->proc_bitmap.ClearBitAtomic(currentCPU);
atomic_add(&gCPU[currentCPU].ici_counter, 1);
@@ -1029,7 +1028,7 @@ smp_send_ici(int32 targetCPU, int32 message, addr_t data, addr_t data2,
msg->done = 0;
// stick it in the appropriate cpu's mailbox
prepend_message(sCPUMessages[targetCPU], msg);
prepend_message(gCPU[targetCPU].cpu_msg, msg);
arch_smp_send_ici(targetCPU);
@@ -1171,7 +1170,7 @@ smp_multicast_ici(const CPUSet& cpuMask, int32 message, addr_t data,
if ((!self && targetCPUs == 1) || (self && targetCPUs == 2)) {
// stick it in the appropriate cpu's mailbox
prepend_message(sCPUMessages[firstNonCurrentCPU], msg);
prepend_message(gCPU[firstNonCurrentCPU].cpu_msg, msg);
arch_smp_send_ici(firstNonCurrentCPU);