XHCI works on real hardware!

kernel/amd64/mm.c

@@ -58,7 +58,7 @@ static struct pg_index mm_page_index (uint64_t vaddr) {
 /* Walk paging tables and allocate necessary structures along the way */
 static uint64_t* mm_next_table (uint64_t* table, uint64_t entry_idx, bool alloc) {
   uint64_t entry = table[entry_idx];
-  physaddr_t paddr;
+  uintptr_t paddr;
 
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
 
@@ -194,7 +194,7 @@ void mm_unmap_kernel_page (uintptr_t vaddr) {
 uintptr_t mm_alloc_user_pd_phys (void) {
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
 
-  physaddr_t cr3 = pmm_alloc (1);
+  uintptr_t cr3 = pmm_alloc (1);
   if (cr3 == PMM_ALLOC_ERR)
     return 0;
 

kernel/amd64/spin_lock.c

@@ -1,7 +1,7 @@
 #include <sys/spin_lock.h>
 
 /// Relax the spinlock using AMD64 pause instruction
-void spin_lock_relax (void) { __asm__ volatile ("pause"); }
+void spin_lock_relax (void) { __asm__ volatile ("pause" ::: "memory"); }
 
 void spin_lock_save_flags (uint64_t* flags) {
   __asm__ volatile ("pushfq; cli; popq %0" : "=rm"(*flags)::"memory");

kernel/amd64/src.mk

@@ -12,7 +12,8 @@ c += amd64/bootmain.c \
 		 amd64/sched1.c \
 		 amd64/proc.c \
 		 amd64/syscall.c \
-		 amd64/gdt.c
+		 amd64/gdt.c \
+		 amd64/stall.c
 
 S += amd64/intr_stub.S \
 		 amd64/spin.S \
@@ -39,4 +40,5 @@ o += amd64/bootmain.o \
 		 amd64/syscall.o \
 		 amd64/syscallentry.o \
 		 amd64/gdt.o \
-		 amd64/sse.o
+		 amd64/sse.o \
+		 amd64/stall.o

kernel/amd64/stall.c

@@ -0,0 +1,31 @@
+#include <libk/std.h>
+#include <sys/spin_lock.h>
+#include <sys/stall.h>
+
+static uint64_t stall_read_tsc (void) {
+  uint32_t lo, hi;
+  __asm__ volatile ("rdtsc" : "=a"(lo), "=d"(hi));
+  return ((uint64_t)hi << 32) | lo;
+}
+
+static uint64_t stall_get_tsc_freq_hz (void) {
+  uint32_t eax, ebx, ecx, edx;
+  __asm__ volatile ("cpuid" : "=a"(eax), "=b"(ebx), "=c"(ecx), "=d"(edx) : "a"(0x15));
+
+  if (eax == 0 || ebx == 0 || ecx == 0)
+    return 2500000000ULL;
+
+  return (uint64_t)ecx * ebx / eax;
+}
+
+void stall_ms (uint64_t ms) {
+  uint64_t freq_hz = stall_get_tsc_freq_hz ();
+
+  uint64_t cycles = freq_hz / 1000;
+  uint64_t wait_cycles = ms * cycles;
+
+  uint64_t now = stall_read_tsc ();
+
+  while ((stall_read_tsc () - now) < wait_cycles)
+    spin_lock_relax ();
+}