Implement argument_ptr () syscall for handling process arguments

Makefile

@@ -4,4 +4,4 @@ include make/apps.mk
 include make/kernel.mk
 include make/dist.mk
 include make/docs.mk
-include make/libc.mk
+include make/libmsl.mk

amd64/link.ld

@@ -6,6 +6,8 @@ PHDRS {
   text PT_LOAD;
   rodata PT_LOAD;
   data PT_LOAD;
+  bss PT_LOAD;
+  tls PT_TLS;
 }
 
 SECTIONS {
@@ -13,32 +15,53 @@ SECTIONS {
 
   .text : {
     *(.text .text.*)
+    *(.ltext .ltext.*)
   } :text
   
-  . = ALIGN(CONSTANT(MAXPAGESIZE));
+  . = ALIGN(0x1000);
   
   .rodata : {
     *(.rodata .rodata.*)
   } :rodata
 
-  .note.gnu.build-id : {
+  . = ALIGN(0x1000);
-    *(.note.gnu.build-id)
-  } :rodata
-  
-  . = ALIGN(CONSTANT(MAXPAGESIZE));
 
   .data : {
     *(.data .data.*)
+    *(.ldata .ldata.*)
   } :data
    
+  . = ALIGN(0x1000);
+
   __bss_start = .;
 
   .bss : {
     *(.bss .bss.*)
-  } :data
+    *(.lbss .lbss.*)
+  } :bss
 
   __bss_end = .;
   
+  . = ALIGN(0x1000);
+
+  __tdata_start = .;
+
+  .tdata : {
+    *(.tdata .tdata.*)
+  } :tls
+  
+  __tdata_end = .;
+  
+  __tbss_start = .;
+
+  .tbss : {
+    *(.tbss .tbss.*)
+  } :tls
+  
+  __tbss_end = .;
+  
+  __tls_size = __tbss_end - __tdata_start;
+
   /DISCARD/ : {
     *(.eh_frame*)
     *(.note .note.*)

aux/bochsrc.txt

@@ -1,4 +1,4 @@
-cpu: model=p4_prescott_celeron_336
+cpu: model=p4_prescott_celeron_336, ips=200000000
 
 memory: guest=4096 host=2048
 
@@ -9,6 +9,7 @@ ata0: enabled=1
 ata0-master: type=cdrom, path=mop3.iso, status=inserted
 com1: enabled=1, mode=file, dev=bochs-com1.txt
 pci: enabled=1, chipset=i440fx
+clock: sync=realtime, time0=local
 
 boot: cdrom
 

include/m/syscall_defs.h

@@ -7,7 +7,10 @@
 #define SYS_UNMAP         4
 #define SYS_CLONE         5
 #define SYS_SCHED         6
-#define SYS_CREATE_MEM    7
+#define SYS_MUTEX_CREATE  7
-#define SYS_UNLINK_MEM    8
+#define SYS_MUTEX_DELETE  8
+#define SYS_MUTEX_LOCK    9
+#define SYS_MUTEX_UNLOCK  10
+#define SYS_ARGUMENT_PTR  11
 
 #endif // _M_SYSCALL_DEFS_H

init/init.c

@@ -1,25 +1,46 @@
-#include <alloc/liballoc.h>
 #include <limits.h>
-#include <m/status.h>
+#include <proc/local.h>
-#include <m/system.h>
+#include <proc/proc.h>
 #include <stddef.h>
 #include <stdint.h>
 #include <string/string.h>
 
-#define EXAMPLE 1
+#define MUTEX 2000
 
-#if EXAMPLE == 1
+LOCAL volatile char letter = 'c';
-void app_main (void) {
-  test ('a');
-  test ('a');
-  test ('a');
 
-  int* xs = malloc (1024 * sizeof (*xs));
+void app_proc (void) {
-  memset (xs, 123, 1024 * sizeof (*xs));
+  char arg_letter = (char)(uintptr_t)argument_ptr ();
-  free (xs);
 
-  test ('a');
+  letter = arg_letter;
-  test ('a');
+
-  test ('a');
+  for (;;) {
+    mutex_lock (MUTEX);
+
+    for (int i = 0; i < 3; i++)
+      test (letter);
+
+    mutex_unlock (MUTEX);
+  }
+
+  process_quit ();
+}
+
+void app_main (void) {
+  mutex_create (MUTEX);
+
+  letter = 'a';
+
+  process_spawn (&app_proc, (void*)'a');
+  process_spawn (&app_proc, (void*)'b');
+  process_spawn (&app_proc, (void*)'c');
+
+  for (;;) {
+    mutex_lock (MUTEX);
+
+    for (int i = 0; i < 3; i++)
+      test (letter);
+
+    mutex_unlock (MUTEX);
+  }
 }
-#endif

init/src.mk

@@ -1,3 +1,3 @@
-S += init.S
+c += init.c
 
 o += init.o

kernel/amd64/apic.c

@@ -4,7 +4,7 @@
 #include <amd64/msr.h>
 #include <libk/std.h>
 #include <limine/requests.h>
-#include <sync/rw_spin_lock.h>
+#include <sync/spin_lock.h>
 #include <sys/debug.h>
 #include <sys/mm.h>
 #include <sys/spin.h>
@@ -38,7 +38,7 @@
 
 struct ioapic {
   struct acpi_madt_ioapic table_data;
-  rw_spin_lock_t lock;
+  spin_lock_t lock;
   uintptr_t mmio_base;
 };
 
@@ -59,10 +59,10 @@ static spin_lock_t lapic_calibration_lock = SPIN_LOCK_INIT;
 static uint32_t amd64_ioapic_read (struct ioapic* ioapic, uint32_t reg) {
   spin_lock_ctx_t ctxioar;
 
-  rw_spin_read_lock (&ioapic->lock, &ctxioar);
+  spin_lock (&ioapic->lock, &ctxioar);
   *(volatile uint32_t*)ioapic->mmio_base = reg;
   uint32_t ret = *(volatile uint32_t*)(ioapic->mmio_base + 0x10);
-  rw_spin_read_unlock (&ioapic->lock, &ctxioar);
+  spin_unlock (&ioapic->lock, &ctxioar);
   return ret;
 }
 
@@ -70,10 +70,10 @@ static uint32_t amd64_ioapic_read (struct ioapic* ioapic, uint32_t reg) {
 static void amd64_ioapic_write (struct ioapic* ioapic, uint32_t reg, uint32_t value) {
   spin_lock_ctx_t ctxioaw;
 
-  rw_spin_write_lock (&ioapic->lock, &ctxioaw);
+  spin_lock (&ioapic->lock, &ctxioaw);
   *(volatile uint32_t*)ioapic->mmio_base = reg;
   *(volatile uint32_t*)(ioapic->mmio_base + 0x10) = value;
-  rw_spin_write_unlock (&ioapic->lock, &ctxioaw);
+  spin_unlock (&ioapic->lock, &ctxioaw);
 }
 
 /* Find an IOAPIC corresposting to provided IRQ */
@@ -82,8 +82,6 @@ static struct ioapic* amd64_ioapic_find (uint32_t irq) {
 
   for (size_t i = 0; i < ioapic_entries; i++) {
     ioapic = &ioapics[i];
-    /* uint32_t version = amd64_ioapic_read ((uintptr_t)hhdm->offset +
-     * (uintptr_t)ioapic->table_data.address, 1); */
     uint32_t version = amd64_ioapic_read (ioapic, 1);
     uint32_t max = ((version >> 16) & 0xFF);
 
@@ -162,9 +160,9 @@ void amd64_ioapic_init (void) {
       struct acpi_madt_ioapic* ioapic_table_data = (struct acpi_madt_ioapic*)current;
       mm_map_kernel_page ((uintptr_t)ioapic_table_data->address,
                           (uintptr_t)hhdm->offset + (uintptr_t)ioapic_table_data->address,
-                          MM_PG_PRESENT | MM_PG_RW | MM_PD_RELOAD);
+                          MM_PG_PRESENT | MM_PG_RW);
       ioapics[ioapic_entries++] = (struct ioapic){
-          .lock = RW_SPIN_LOCK_INIT,
+          .lock = SPIN_LOCK_INIT,
           .table_data = *ioapic_table_data,
           .mmio_base = ((uintptr_t)hhdm->offset + (uintptr_t)ioapic_table_data->address),
       };
@@ -233,7 +231,7 @@ static uint32_t amd64_lapic_calibrate (uint32_t us) {
 static void amd64_lapic_start (uint32_t ticks) {
   amd64_lapic_write (LAPIC_DCR, DIVIDER_VALUE);
   amd64_lapic_write (LAPIC_TIMICT, ticks);
-  amd64_lapic_write (LAPIC_LVTTR, SCHED_PREEMPT_TIMER | (1 << 17) | (1 << 16));
+  amd64_lapic_write (LAPIC_LVTTR, SCHED_PREEMPT_TIMER | (1 << 17));
 }
 
 /*
@@ -248,8 +246,7 @@ void amd64_lapic_init (uint32_t us) {
   uintptr_t lapic_paddr = amd64_rdmsr (MSR_APIC_BASE) & 0xFFFFF000;
   thiscpu->lapic_mmio_base = lapic_paddr + (uintptr_t)hhdm->offset;
 
-  mm_map_kernel_page (lapic_paddr, thiscpu->lapic_mmio_base,
+  mm_map_kernel_page (lapic_paddr, thiscpu->lapic_mmio_base, MM_PG_PRESENT | MM_PG_RW);
-                      MM_PG_PRESENT | MM_PG_RW | MM_PD_LOCK | MM_PD_RELOAD);
 
   amd64_lapic_write (LAPIC_SIVR, 0xFF | (1 << 8));
 
@@ -270,5 +267,5 @@ void amd64_lapic_ipi (uint32_t lapic_id, uint32_t vec) {
   }
 
   amd64_lapic_write (LAPIC_ICR + 0x10, (lapic_id << 24));
-  amd64_lapic_write (LAPIC_ICR, vec);
+  amd64_lapic_write (LAPIC_ICR, vec | (1 << 14));
 }

kernel/amd64/bootmain.c

@@ -9,6 +9,7 @@
 #include <irq/irq.h>
 #include <libk/std.h>
 #include <limine/limine.h>
+#include <limine/requests.h>
 #include <mm/liballoc.h>
 #include <mm/pmm.h>
 #include <proc/proc.h>
@@ -29,7 +30,9 @@ ALIGNED (16) static uint8_t uacpi_memory_buffer[UACPI_MEMORY_BUFFER_MAX];
  * the necessary platform-dependent subsystems/drivers and jump into the init app.
  */
 void bootmain (void) {
-  struct cpu* bsp_cpu = cpu_make ();
+  struct limine_mp_response* mp = limine_mp_request.response;
+
+  struct cpu* bsp_cpu = cpu_make (mp->bsp_lapic_id);
 
   amd64_init (bsp_cpu, false);
   syscall_init ();
@@ -46,8 +49,6 @@ void bootmain (void) {
 
   smp_init ();
 
-  mm_init2 ();
-
   proc_init ();
 
   for (;;)

kernel/amd64/hpet.c

@@ -129,8 +129,7 @@ void amd64_hpet_init (void) {
   hpet_paddr = (uintptr_t)hpet->address.address;
 
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-  mm_map_kernel_page (hpet_paddr, (uintptr_t)hhdm->offset + hpet_paddr,
+  mm_map_kernel_page (hpet_paddr, (uintptr_t)hhdm->offset + hpet_paddr, MM_PG_PRESENT | MM_PG_RW);
-                      MM_PG_PRESENT | MM_PG_RW | MM_PD_RELOAD);
 
   uint64_t caps = amd64_hpet_read64 (HPET_GCIDR);
   hpet_32bits = (caps & (1 << 13)) ? 0 : 1;

kernel/amd64/intr.c

@@ -165,10 +165,21 @@ static void amd64_intr_exception (struct saved_regs* regs) {
 
 /* Handle incoming interrupt, dispatch IRQ handlers. */
 void amd64_intr_handler (void* stack_ptr) {
+  spin_lock_ctx_t ctxcpu, ctxpr;
+
   amd64_load_kernel_cr3 ();
 
   struct saved_regs* regs = stack_ptr;
 
+  spin_lock (&thiscpu->lock, &ctxcpu);
+  struct proc* proc_current = thiscpu->proc_current;
+  spin_lock (&proc_current->lock, &ctxpr);
+
+  memcpy (&proc_current->pdata.regs, regs, sizeof (struct saved_regs));
+
+  spin_unlock (&proc_current->lock, &ctxpr);
+  spin_unlock (&thiscpu->lock, &ctxcpu);
+
   if (regs->trap <= 31) {
     amd64_intr_exception (regs);
   } else {
@@ -177,13 +188,7 @@ void amd64_intr_handler (void* stack_ptr) {
     struct irq* irq = irq_find (regs->trap);
 
     if (irq != NULL) {
-      if ((irq->flags & IRQ_INTERRUPT_SAFE))
-        __asm__ volatile ("sti");
-
       irq->func (irq->arg, stack_ptr);
-
-      if ((irq->flags & IRQ_INTERRUPT_SAFE))
-        __asm__ volatile ("cli");
     }
   }
 }
@@ -214,15 +219,3 @@ void irq_save (spin_lock_ctx_t* ctx) { *ctx = amd64_irq_save_flags (); }
 
 /* Restore interrupt state */
 void irq_restore (spin_lock_ctx_t* ctx) { amd64_irq_restore_flags (*ctx); }
-
-/* Map custom IRQ mappings to legacy IRQs */
-uint32_t amd64_resolve_irq (uint32_t irq) {
-  static const uint32_t mappings[] = {
-      [SCHED_PREEMPT_TIMER] = 0,
-      [TLB_SHOOTDOWN] = 6,
-      [CPU_REQUEST_SCHED] = 3,
-      [CPU_SPURIOUS] = 5,
-  };
-
-  return mappings[irq];
-}

kernel/amd64/intr.h

@@ -32,7 +32,6 @@ struct saved_regs {
 } PACKED;
 
 void amd64_load_idt (void);
-uint32_t amd64_resolve_irq (uint32_t irq);
 void amd64_intr_init (void);
 
 #endif // _KERNEL_AMD64_INTR_H

kernel/amd64/mm.c

@@ -22,10 +22,12 @@ struct pg_index {
 } PACKED;
 
 /* Kernel page directory */
-static struct pd kernel_pd = {.lock = SPIN_LOCK_INIT};
+static struct pd kernel_pd;
-static spin_lock_ctx_t ctxkpd;
+static spin_lock_t kernel_pd_lock;
-/* Lock needed to sync between map/unmap operations and TLB shootdown */
+
-static spin_lock_t mm_lock = SPIN_LOCK_INIT;
+void mm_kernel_lock (spin_lock_ctx_t* ctx) { spin_lock (&kernel_pd_lock, ctx); }
+
+void mm_kernel_unlock (spin_lock_ctx_t* ctx) { spin_lock (&kernel_pd_lock, ctx); }
 
 /* Get current value of CR3 register */
 static uintptr_t amd64_current_cr3 (void) {
@@ -112,15 +114,7 @@ static void amd64_reload_cr3 (void) {
 
 /* Map physical address to virtual address with flags. TLB needs to be flushed afterwards. */
 void mm_map_page (struct pd* pd, uintptr_t paddr, uintptr_t vaddr, uint32_t flags) {
-  spin_lock_ctx_t ctxmm, ctxpd;
-
-  spin_lock (&mm_lock, &ctxmm);
-
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-  bool do_reload = false;
-
-  if (flags & MM_PD_LOCK)
-    spin_lock (&pd->lock, &ctxpd);
 
   uint64_t amd64_flags = amd64_mm_resolve_flags (flags);
 
@@ -129,69 +123,50 @@ void mm_map_page (struct pd* pd, uintptr_t paddr, uintptr_t vaddr, uint32_t flag
 
   uint64_t* pml3 = amd64_mm_next_table (pml4, pg_index.pml4, true);
   if (pml3 == NULL)
-    goto done;
+    return;
 
   uint64_t* pml2 = amd64_mm_next_table (pml3, pg_index.pml3, true);
   if (pml2 == NULL)
-    goto done;
+    return;
 
   uint64_t* pml1 = amd64_mm_next_table (pml2, pg_index.pml2, true);
   if (pml1 == NULL)
-    goto done;
+    return;
 
   uint64_t* pte = &pml1[pg_index.pml1];
 
   *pte = ((paddr & ~0xFFFULL) | (amd64_flags & 0x7ULL));
-  do_reload = true;
-
-done:
-  if (do_reload && (flags & MM_PD_RELOAD))
-    amd64_reload_cr3 ();
-
-  if (flags & MM_PD_LOCK)
-    spin_unlock (&pd->lock, &ctxpd);
-
-  spin_unlock (&mm_lock, &ctxmm);
 }
 
 /* Map a page into kernel page directory */
 void mm_map_kernel_page (uintptr_t paddr, uintptr_t vaddr, uint32_t flags) {
   mm_map_page (&kernel_pd, paddr, vaddr, flags);
+  amd64_reload_cr3 ();
 }
 
 /* Unmap a virtual address. TLB needs to be flushed afterwards */
-void mm_unmap_page (struct pd* pd, uintptr_t vaddr, uint32_t flags) {
+void mm_unmap_page (struct pd* pd, uintptr_t vaddr) {
-  spin_lock_ctx_t ctxmm, ctxpd;
-
-  spin_lock (&mm_lock, &ctxmm);
-
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-  bool do_reload = false;
-
-  if (flags & MM_PD_LOCK)
-    spin_lock (&pd->lock, &ctxpd);
 
   uint64_t* pml4 = (uint64_t*)(pd->cr3_paddr + (uintptr_t)hhdm->offset);
   struct pg_index pg_index = amd64_mm_page_index (vaddr);
 
   uint64_t* pml3 = amd64_mm_next_table (pml4, pg_index.pml4, false);
   if (pml3 == NULL)
-    goto done;
+    return;
 
   uint64_t* pml2 = amd64_mm_next_table (pml3, pg_index.pml3, false);
   if (pml2 == NULL)
-    goto done;
+    return;
 
   uint64_t* pml1 = amd64_mm_next_table (pml2, pg_index.pml2, false);
   if (pml1 == NULL)
-    goto done;
+    return;
 
   uint64_t* pte = &pml1[pg_index.pml1];
 
-  if ((*pte) & AMD64_PG_PRESENT) {
+  if ((*pte) & AMD64_PG_PRESENT)
     *pte = 0;
-    do_reload = true;
-  }
 
   if (amd64_mm_is_table_empty (pml1)) {
     uintptr_t pml1_phys = pml2[pg_index.pml2] & ~0xFFFULL;
@@ -210,28 +185,14 @@ void mm_unmap_page (struct pd* pd, uintptr_t vaddr, uint32_t flags) {
       }
     }
   }
-
-done:
-  if (do_reload && (flags & MM_PD_RELOAD))
-    amd64_reload_cr3 ();
-
-  if (flags & MM_PD_LOCK)
-    spin_unlock (&pd->lock, &ctxpd);
-
-  spin_unlock (&mm_lock, &ctxmm);
 }
 
 /* Unmap a page from kernel page directory */
-void mm_unmap_kernel_page (uintptr_t vaddr, uint32_t flags) {
+void mm_unmap_kernel_page (uintptr_t vaddr) {
-  mm_unmap_page (&kernel_pd, vaddr, flags);
+  mm_unmap_page (&kernel_pd, vaddr);
+  amd64_reload_cr3 ();
 }
 
-/* Lock kernel page directory */
-void mm_lock_kernel (void) { spin_lock (&kernel_pd.lock, &ctxkpd); }
-
-/* Unlock kernel page directory */
-void mm_unlock_kernel (void) { spin_unlock (&kernel_pd.lock, &ctxkpd); }
-
 /* Allocate a userspace-ready page directory */
 uintptr_t mm_alloc_user_pd_phys (void) {
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
@@ -250,26 +211,10 @@ uintptr_t mm_alloc_user_pd_phys (void) {
   return cr3;
 }
 
-/* Reload after map/unmap operation was performed. This function does the TLB shootdown. */
+bool mm_validate (struct pd* pd, uintptr_t vaddr) {
-void mm_reload (void) {
-  struct limine_mp_response* mp = limine_mp_request.response;
-
-  for (size_t i = 0; i < mp->cpu_count; i++) {
-    amd64_lapic_ipi (mp->cpus[i]->lapic_id, TLB_SHOOTDOWN);
-  }
-}
-
-bool mm_validate (struct pd* pd, uintptr_t vaddr, uint32_t flags) {
-  spin_lock_ctx_t ctxmm, ctxpd;
-
-  spin_lock (&mm_lock, &ctxmm);
-
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
   bool ret = false;
 
-  if (flags & MM_PD_LOCK)
-    spin_lock (&pd->lock, &ctxpd);
-
   uint64_t* pml4 = (uint64_t*)(pd->cr3_paddr + (uintptr_t)hhdm->offset);
   struct pg_index pg_index = amd64_mm_page_index (vaddr);
 
@@ -289,45 +234,26 @@ bool mm_validate (struct pd* pd, uintptr_t vaddr, uint32_t flags) {
   ret = (pte & AMD64_PG_PRESENT) != 0;
 
 done:
-  if (flags & MM_PD_LOCK)
-    spin_unlock (&pd->lock, &ctxpd);
-
-  spin_unlock (&mm_lock, &ctxmm);
-
   return ret;
 }
 
-bool mm_validate_buffer (struct pd* pd, uintptr_t vaddr, size_t size, uint32_t flags) {
+bool mm_validate_buffer (struct pd* pd, uintptr_t vaddr, size_t size) {
   bool ok = true;
-  spin_lock_ctx_t ctxpd;
-
-  if (flags & MM_PD_LOCK)
-    spin_lock (&pd->lock, &ctxpd);
 
   for (size_t i = 0; i < size; i++) {
-    ok = mm_validate (pd, vaddr + i, 0);
+    ok = mm_validate (pd, vaddr + i);
     if (!ok)
       goto done;
   }
 
 done:
-  if (flags & MM_PD_LOCK)
-    spin_unlock (&pd->lock, &ctxpd);
-
   return ok;
 }
 
-uintptr_t mm_p2v (struct pd* pd, uintptr_t paddr, uint32_t flags) {
+uintptr_t mm_p2v (struct pd* pd, uintptr_t paddr) {
-  spin_lock_ctx_t ctxmm, ctxpd;
-
-  spin_lock (&mm_lock, &ctxmm);
-
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
   uintptr_t ret = 0;
 
-  if (flags & MM_PD_LOCK)
-    spin_lock (&pd->lock, &ctxpd);
-
   uint64_t* pml4 = (uint64_t*)(pd->cr3_paddr + (uintptr_t)hhdm->offset);
 
   for (size_t i4 = 0; i4 < 512; i4++) {
@@ -358,25 +284,13 @@ uintptr_t mm_p2v (struct pd* pd, uintptr_t paddr, uint32_t flags) {
   }
 
 done:
-  if (flags & MM_PD_LOCK)
-    spin_unlock (&pd->lock, &ctxpd);
-
-  spin_unlock (&mm_lock, &ctxmm);
-
   return ret;
 }
 
-uintptr_t mm_v2p (struct pd* pd, uintptr_t vaddr, uint32_t flags) {
+uintptr_t mm_v2p (struct pd* pd, uintptr_t vaddr) {
-  spin_lock_ctx_t ctxmm, ctxpd;
-
-  spin_lock (&mm_lock, &ctxmm);
-
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
   uintptr_t ret = 0;
 
-  if (flags & MM_PD_LOCK)
-    spin_lock (&pd->lock, &ctxpd);
-
   uint64_t* pml4 = (uint64_t*)(pd->cr3_paddr + (uintptr_t)hhdm->offset);
   struct pg_index pg_index = amd64_mm_page_index (vaddr);
 
@@ -400,27 +314,8 @@ uintptr_t mm_v2p (struct pd* pd, uintptr_t vaddr, uint32_t flags) {
   ret = ((pte & ~0xFFFULL) | (vaddr & 0xFFFULL));
 
 done:
-  if (flags & MM_PD_LOCK)
-    spin_unlock (&pd->lock, &ctxpd);
-
-  spin_unlock (&mm_lock, &ctxmm);
-
   return ret;
 }
 
-/* TLB shootdown IRQ handler */
-static void amd64_tlb_shootdown_irq (void* arg, void* regs) {
-  (void)arg, (void)regs;
-
-  amd64_reload_cr3 ();
-  DEBUG ("cpu %u TLB shootdown\n", thiscpu->id);
-}
-
-/* Continue initializing memory management subsystem for AMD64 after the essential parts were
- * initialized */
-void mm_init2 (void) {
-  irq_attach (&amd64_tlb_shootdown_irq, NULL, TLB_SHOOTDOWN, IRQ_INTERRUPT_SAFE);
-}
-
 /* Initialize essentials for the AMD64 memory management subsystem */
 void mm_init (void) { kernel_pd.cr3_paddr = amd64_current_cr3 (); }

kernel/amd64/mm.h

@@ -7,12 +7,9 @@
 #define PAGE_SIZE 4096
 
 struct pd {
-  spin_lock_t lock;
   uintptr_t cr3_paddr;
-  atomic_int refs;
 };
 
 void amd64_load_kernel_cr3 (void);
-void mm_init2 (void);
 
 #endif // _KERNEL_AMD64_MM_H

kernel/amd64/proc.c

@@ -1,6 +1,7 @@
 #include <amd64/gdt.h>
 #include <amd64/proc.h>
 #include <aux/elf.h>
+#include <libk/align.h>
 #include <libk/list.h>
 #include <libk/rbtree.h>
 #include <libk/std.h>
@@ -8,18 +9,18 @@
 #include <limine/requests.h>
 #include <mm/liballoc.h>
 #include <mm/pmm.h>
+#include <proc/mutex.h>
 #include <proc/proc.h>
+#include <proc/procgroup.h>
 #include <proc/resource.h>
-#include <sync/rw_spin_lock.h>
 #include <sync/spin_lock.h>
 #include <sys/debug.h>
+#include <sys/proc.h>
 
-/* 0 is kpproc */
+static atomic_int pids = 0;
-static atomic_int pids = 1;
 
 struct proc* proc_from_elf (uint8_t* elf_contents) {
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-  int rid;
 
   struct proc* proc = malloc (sizeof (*proc));
   if (proc == NULL)
@@ -31,67 +32,18 @@ struct proc* proc_from_elf (uint8_t* elf_contents) {
   atomic_store (&proc->state, PROC_READY);
   proc->pid = atomic_fetch_add (&pids, 1);
 
-  proc->resources = malloc (sizeof (*proc->resources));
+  proc->procgroup = procgroup_create ();
-  if (proc->resources == NULL) {
+  if (proc->procgroup == NULL) {
     free (proc);
     return NULL;
   }
-  proc->resources->tree = NULL;
+  procgroup_attach (proc->procgroup, proc);
-  proc->resources->lock = RW_SPIN_LOCK_INIT;
-  proc->resources->refs = 1;
-  proc->resources->sys_rids = 0;
 
-  proc->pd = malloc (sizeof (*proc->pd));
+  uintptr_t kstack_paddr = pmm_alloc (KSTACK_SIZE / PAGE_SIZE);
-  if (proc->pd == NULL) {
+  proc->pdata.kernel_stack = kstack_paddr + (uintptr_t)hhdm->offset + KSTACK_SIZE;
-    free (proc->resources);
-    free (proc);
-    return NULL;
-  }
 
-  proc->pd->lock = SPIN_LOCK_INIT;
+  procgroup_map (proc->procgroup, PROC_USTACK_TOP - USTACK_SIZE, USTACK_SIZE / PAGE_SIZE,
-  proc->pd->refs = 1;
+                 MM_PG_USER | MM_PG_PRESENT | MM_PG_RW, NULL);
-  proc->pd->cr3_paddr = mm_alloc_user_pd_phys ();
-  if (proc->pd->cr3_paddr == 0) {
-    free (proc->pd);
-    free (proc->resources);
-    free (proc);
-    return NULL;
-  }
-
-  struct proc_resource_mem_init kstk_mem_init = {.pages = KSTACK_SIZE / PAGE_SIZE,
-                                                 .managed = false};
-  rid = atomic_fetch_add (&proc->resources->sys_rids, 1);
-  struct proc_resource* kstk_r =
-      proc_create_resource (proc, rid, PR_MEM, RV_PRIVATE, (void*)&kstk_mem_init);
-  if (kstk_r == NULL) {
-    pmm_free (proc->pd->cr3_paddr, 1);
-    free (proc->pd);
-    free (proc->resources);
-    free (proc);
-    return NULL;
-  }
-
-  proc->pdata.kernel_stack = kstk_r->u.mem.paddr + (uintptr_t)hhdm->offset + KSTACK_SIZE;
-
-  struct proc_resource_mem_init ustk_mem_init = {.pages = USTACK_SIZE / PAGE_SIZE,
-                                                 .managed = false};
-  rid = atomic_fetch_add (&proc->resources->sys_rids, 1);
-  struct proc_resource* ustk_r =
-      proc_create_resource (proc, rid, PR_MEM, RV_PRIVATE, (void*)&ustk_mem_init);
-  if (ustk_r == NULL) {
-    kstk_r->ops.cleanup (proc, kstk_r);
-    free (kstk_r);
-    pmm_free (proc->pd->cr3_paddr, 1);
-    free (proc->pd);
-    free (proc->resources);
-    free (proc);
-    return NULL;
-  }
-
-  proc->pdata.user_stack = ustk_r->u.mem.paddr;
-
-  proc_map (proc, proc->pdata.user_stack, PROC_USTACK_TOP - USTACK_SIZE, USTACK_SIZE / PAGE_SIZE,
-            MM_PG_USER | MM_PG_PRESENT | MM_PG_RW);
 
   proc->flags |= PROC_USTK_PREALLOC;
 
@@ -106,11 +58,10 @@ struct proc* proc_from_elf (uint8_t* elf_contents) {
   return proc;
 }
 
-struct proc* proc_clone (struct proc* proto, uintptr_t vstack_top, size_t stack_size,
+struct proc* proc_clone (struct proc* proto, uintptr_t vstack_top, uintptr_t entry,
-                         uintptr_t entry) {
+                         uintptr_t argument_ptr) {
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-  spin_lock_ctx_t ctxprt, ctxrs;
+  spin_lock_ctx_t ctxprt;
-  int rid;
 
   struct proc* proc = malloc (sizeof (*proc));
   if (proc == NULL)
@@ -124,54 +75,13 @@ struct proc* proc_clone (struct proc* proto, uintptr_t vstack_top, size_t stack_
 
   spin_lock (&proto->lock, &ctxprt);
 
-  proc->pd = proto->pd;
+  proc->procgroup = proto->procgroup;
-  proc->mappings = proto->mappings;
+  procgroup_attach (proc->procgroup, proc);
-  atomic_fetch_add (&proto->pd->refs, 1);
-
-  proc->resources = proto->resources;
-
-  rw_spin_write_lock (&proc->resources->lock, &ctxrs);
-
-  atomic_fetch_add (&proc->resources->refs, 1);
-
-  struct rb_node_link* rnode;
-  rbtree_first (&proc->resources->tree, rnode);
-  while (rnode) {
-    struct rb_node_link* next;
-    rbtree_next (rnode, next);
-
-    struct proc_resource* resource =
-        rbtree_entry (rnode, struct proc_resource, local_resource_tree_link);
-    atomic_fetch_add (&resource->refs, 1);
-
-    rnode = next;
-  }
-
-  rw_spin_write_unlock (&proc->resources->lock, &ctxrs);
 
   spin_unlock (&proto->lock, &ctxprt);
 
-  uintptr_t vstack_bottom = vstack_top - stack_size;
+  uintptr_t kstack_paddr = pmm_alloc (KSTACK_SIZE / PAGE_SIZE);
-
+  proc->pdata.kernel_stack = kstack_paddr + (uintptr_t)hhdm->offset + KSTACK_SIZE;
-  uintptr_t pstack_bottom = mm_v2p (proc->pd, vstack_bottom, MM_PD_LOCK);
-  if (pstack_bottom == 0) {
-    free (proc);
-    return NULL;
-  }
-
-  struct proc_resource_mem_init kstk_mem_init = {.pages = KSTACK_SIZE / PAGE_SIZE,
-                                                 .managed = false};
-  rid = atomic_fetch_add (&proc->resources->sys_rids, 1);
-  struct proc_resource* kstk_r =
-      proc_create_resource (proc, rid, PR_MEM, RV_PRIVATE, (void*)&kstk_mem_init);
-  if (kstk_r == NULL) {
-    free (proc);
-    return NULL;
-  }
-
-  proc->pdata.kernel_stack = kstk_r->u.mem.paddr + (uintptr_t)hhdm->offset + KSTACK_SIZE;
-
-  proc->pdata.user_stack = pstack_bottom + stack_size;
 
   proc->pdata.regs.ss = GDT_UDATA | 0x03;
   proc->pdata.regs.rsp = (uint64_t)vstack_top;
@@ -179,40 +89,50 @@ struct proc* proc_clone (struct proc* proto, uintptr_t vstack_top, size_t stack_
   proc->pdata.regs.cs = GDT_UCODE | 0x03;
   proc->pdata.regs.rip = (uint64_t)entry;
 
+  proc->uvaddr_argument = argument_ptr;
+
+  proc_init_tls (proc);
+
   return proc;
 }
 
 void proc_cleanup (struct proc* proc) {
-  struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
+  proc_sqs_cleanup (proc);
-  spin_lock_ctx_t ctxprpd;
+  proc_mutexes_cleanup (proc);
 
-  proc_cleanup_resources (proc);
+  pmm_free (proc->pdata.kernel_stack, KSTACK_SIZE / PAGE_SIZE);
+  procgroup_unmap (proc->procgroup, proc->pdata.tls_vaddr, proc->procgroup->tls.tls_tmpl_pages);
 
-  if (atomic_fetch_sub (&proc->pd->refs, 1) == 1) {
+  procgroup_detach (proc->procgroup, proc);
-    DEBUG ("PID %d Free virtual address space\n", proc->pid);
-    struct list_node_link *mapping_link, *mapping_link_tmp;
-    spin_lock (&proc->pd->lock, &ctxprpd);
-
-    list_foreach (proc->mappings, mapping_link, mapping_link_tmp) {
-      struct proc_mapping* mapping =
-          list_entry (mapping_link, struct proc_mapping, proc_mappings_link);
-
-      list_remove (proc->mappings, mapping_link);
-      free (mapping);
-    }
-
-    pmm_free (proc->pd->cr3_paddr, 1);
-    spin_unlock (&proc->pd->lock, &ctxprpd);
-    free (proc->pd);
-  }
-
-  pmm_free (proc->pdata.kernel_stack - (uintptr_t)hhdm->offset - KSTACK_SIZE,
-            KSTACK_SIZE / PAGE_SIZE);
-
-  if ((proc->flags & PROC_USTK_PREALLOC))
-    pmm_free (proc->pdata.user_stack, USTACK_SIZE / PAGE_SIZE);
-
-  DEBUG ("PID %d Free stacks\n", proc->pid);
 
+  /* clean the process */
   free (proc);
 }
+
+void proc_init_tls (struct proc* proc) {
+  struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
+
+  if (proc->procgroup->tls.tls_tmpl == NULL)
+    return;
+
+  size_t tls_size = proc->procgroup->tls.tls_tmpl_size;
+  size_t pages = proc->procgroup->tls.tls_tmpl_pages;
+
+  uintptr_t tls_paddr;
+  uint32_t flags = MM_PG_USER | MM_PG_PRESENT | MM_PG_RW;
+
+  uintptr_t tls_vaddr = procgroup_map (proc->procgroup, 0, pages, flags, &tls_paddr);
+
+  uintptr_t k_tls_addr = (uintptr_t)hhdm->offset + tls_paddr;
+
+  memset ((void*)k_tls_addr, 0, pages * PAGE_SIZE);
+  memcpy ((void*)k_tls_addr, (void*)proc->procgroup->tls.tls_tmpl, tls_size);
+
+  uintptr_t ktcb = k_tls_addr + tls_size;
+  uintptr_t utcb = tls_vaddr + tls_size;
+
+  *(uintptr_t*)ktcb = utcb;
+
+  proc->pdata.fs_base = utcb;
+  proc->pdata.tls_vaddr = tls_vaddr;
+}

kernel/amd64/proc.h

@@ -4,17 +4,19 @@
 #include <amd64/intr.h>
 #include <libk/std.h>
 
-/// Top of userspace process' stack
+/* Top of userspace process' stack */
 #define PROC_USTACK_TOP 0x00007FFFFFFFF000ULL
-/// Size of userspace process' stack
+/* Size of userspace process' stack */
 #define USTACK_SIZE (256 * PAGE_SIZE)
+/* proc_map () base address */
+#define PROC_MAP_BASE 0x0000700000000000
 
-/// Platform-dependent process data
+/* Platform-dependent process data */
 struct proc_platformdata {
   struct saved_regs regs;
-  uintptr_t user_stack;
   uintptr_t kernel_stack;
-  uint64_t gs_base;
+  uint64_t fs_base;
+  uintptr_t tls_vaddr;
 };
 
 #endif // _KERNEL_AMD64_PROC_H

kernel/amd64/procgroup.h

@@ -0,0 +1,13 @@
+#ifndef _KERNEL_AMD64_PROCGRPUP_H
+#define _KERNEL_AMD64_PROCGRPUP_H
+
+#include <libk/std.h>
+
+struct procgroup_tls {
+  uint8_t* tls_tmpl;
+  size_t tls_tmpl_size;
+  size_t tls_tmpl_total_size;
+  size_t tls_tmpl_pages;
+};
+
+#endif // _KERNEL_AMD64_PROCGRPUP_H

kernel/amd64/sched1.c

@@ -3,14 +3,21 @@
 #include <amd64/sched.h>
 #include <libk/std.h>
 #include <proc/proc.h>
+#include <sync/spin_lock.h>
 #include <sys/mm.h>
 #include <sys/smp.h>
 
-void do_sched (struct proc* proc) {
+void do_sched (struct proc* proc, spin_lock_t* cpu_lock, spin_lock_ctx_t* ctxcpu) {
-  __asm__ volatile ("cli");
+  spin_lock_ctx_t ctxpr;
+
+  spin_lock (&proc->lock, &ctxpr);
 
   thiscpu->tss.rsp0 = proc->pdata.kernel_stack;
   thiscpu->syscall_kernel_stack = proc->pdata.kernel_stack;
+  amd64_wrmsr (MSR_FS_BASE, proc->pdata.fs_base);
 
-  amd64_do_sched ((void*)&proc->pdata.regs, (void*)proc->pd->cr3_paddr);
+  spin_unlock (&proc->lock, &ctxpr);
+  spin_unlock (cpu_lock, ctxcpu);
+
+  amd64_do_sched ((void*)&proc->pdata.regs, (void*)proc->procgroup->pd.cr3_paddr);
 }

kernel/amd64/smp.c

@@ -8,8 +8,10 @@
 #include <libk/string.h>
 #include <limine/requests.h>
 #include <mm/liballoc.h>
+#include <proc/proc.h>
 #include <sync/spin_lock.h>
 #include <sys/debug.h>
+#include <sys/sched.h>
 #include <sys/smp.h>
 #include <sys/syscall.h>
 
@@ -21,7 +23,7 @@ static struct cpu cpus[CPUS_MAX];
 static atomic_int cpu_init_count;
 
 /// Allocate a CPU structure
-struct cpu* cpu_make (void) {
+struct cpu* cpu_make (uint64_t lapic_id) {
   int id = atomic_fetch_add (&cpu_counter, 1);
 
   struct cpu* cpu = &cpus[id];
@@ -29,6 +31,7 @@ struct cpu* cpu_make (void) {
   memset (cpu, 0, sizeof (*cpu));
   cpu->lock = SPIN_LOCK_INIT;
   cpu->id = id;
+  cpu->lapic_id = lapic_id;
 
   amd64_wrmsr (MSR_GS_BASE, (uint64_t)cpu);
 
@@ -41,21 +44,36 @@ struct cpu* cpu_get (void) {
 }
 
 void cpu_request_sched (struct cpu* cpu) {
-  struct limine_mp_response* mp = limine_mp_request.response;
+  if (cpu == thiscpu) {
+    proc_sched ();
+    return;
+  }
 
-  for (size_t i = 0; i < mp->cpu_count; i++) {
+  amd64_lapic_ipi (cpu->lapic_id, CPU_REQUEST_SCHED);
-    if (cpu->id == i) {
+}
-      amd64_lapic_ipi (mp->cpus[i]->lapic_id, CPU_REQUEST_SCHED);
+
-      break;
+struct cpu* cpu_find_lightest (void) {
+  struct cpu* cpu = &cpus[0];
+
+  int load = atomic_load (&cpu->proc_run_q_count);
+
+  for (unsigned int i = 1; i < cpu_counter; i++) {
+    struct cpu* new_cpu = &cpus[i];
+    int new_load = atomic_load (&new_cpu->proc_run_q_count);
+    if (new_load < load) {
+      load = new_load;
+      cpu = new_cpu;
     }
   }
+
+  return cpu;
 }
 
 /// Bootstrap code for non-BSP CPUs
 static void amd64_smp_bootstrap (struct limine_mp_info* mp_info) {
   amd64_load_kernel_cr3 ();
 
-  struct cpu* cpu = cpu_make ();
+  struct cpu* cpu = cpu_make (mp_info->lapic_id);
 
   amd64_init (cpu, true); /* gdt + idt */
   syscall_init ();
@@ -64,12 +82,14 @@ static void amd64_smp_bootstrap (struct limine_mp_info* mp_info) {
 
   DEBUG ("CPU %u is online!\n", thiscpu->id);
 
-  __asm__ volatile ("sti");
-
   atomic_fetch_sub (&cpu_init_count, 1);
 
-  for (;;)
+  struct proc* spin_proc = proc_spawn_rd ("spin.exe");
-    ;
+  proc_register (spin_proc, thiscpu);
+
+  spin_lock_ctx_t ctxcpu;
+  spin_lock (&spin_proc->cpu->lock, &ctxcpu);
+  do_sched (spin_proc, &spin_proc->cpu->lock, &ctxcpu);
 }
 
 /// Initialize SMP subsystem for AMD64. Start AP CPUs
@@ -81,7 +101,7 @@ void smp_init (void) {
   cpu_init_count = mp->cpu_count - 1; /* Don't include BSP */
 
   for (size_t i = 0; i < mp->cpu_count; i++) {
-    if (mp->cpus[i]->lapic_id != thiscpu->id) {
+    if (mp->cpus[i]->lapic_id != thiscpu->lapic_id) {
       DEBUG ("Trying CPU %u\n", mp->cpus[i]->lapic_id);
       mp->cpus[i]->goto_address = &amd64_smp_bootstrap;
     }

kernel/amd64/smp.h

@@ -2,6 +2,7 @@
 #define _KERNEL_AMD64_SMP_H
 
 #include <amd64/gdt.h>
+#include <amd64/intr.h>
 #include <amd64/tss.h>
 #include <aux/compiler.h>
 #include <libk/rbtree.h>
@@ -23,17 +24,20 @@ struct cpu {
 
   uintptr_t lapic_mmio_base;
   uint64_t lapic_ticks;
+  uint64_t lapic_id;
   uint32_t id;
 
   spin_lock_t lock;
 
-  struct rb_node_link* proc_run_q;
+  struct list_node_link* proc_run_q;
   struct proc* proc_current;
+  atomic_int proc_run_q_count;
 };
 
-struct cpu* cpu_make (void);
+struct cpu* cpu_make (uint64_t lapic_id);
 struct cpu* cpu_get (void);
 void cpu_request_sched (struct cpu* cpu);
+struct cpu* cpu_find_lightest (void);
 
 #define thiscpu (cpu_get ())
 

kernel/amd64/syscall.c

@@ -3,6 +3,7 @@
 #include <amd64/mm.h>
 #include <amd64/msr-index.h>
 #include <amd64/msr.h>
+#include <libk/string.h>
 #include <m/status.h>
 #include <m/syscall_defs.h>
 #include <proc/proc.h>
@@ -12,26 +13,29 @@
 
 extern void amd64_syscall_entry (void);
 
-int amd64_syscall_dispatch (void* stack_ptr) {
+uintptr_t amd64_syscall_dispatch (void* stack_ptr) {
-  amd64_load_kernel_cr3 ();
+  spin_lock_ctx_t ctxcpu, ctxpr;
 
+  amd64_load_kernel_cr3 ();
   struct saved_regs* regs = stack_ptr;
 
+  spin_lock (&thiscpu->lock, &ctxcpu);
+  struct proc* caller = thiscpu->proc_current;
+  spin_lock (&caller->lock, &ctxpr);
+
+  memcpy (&caller->pdata.regs, regs, sizeof (struct saved_regs));
+
+  spin_unlock (&caller->lock, &ctxpr);
+  spin_unlock (&thiscpu->lock, &ctxcpu);
+
   int syscall_num = regs->rax;
   syscall_handler_func_t func = syscall_find_handler (syscall_num);
 
-  if (func == NULL)
+  if (func == NULL) {
     return -ST_SYSCALL_NOT_FOUND;
+  }
 
-  struct proc* caller = thiscpu->proc_current;
+  return func (caller, regs, regs->rdi, regs->rsi, regs->rdx, regs->r10, regs->r8, regs->r9);
-
-  __asm__ volatile ("sti");
-
-  int result = func (caller, regs, regs->rdi, regs->rsi, regs->rdx, regs->r10, regs->r8, regs->r9);
-
-  __asm__ volatile ("cli");
-
-  return result;
 }
 
 void syscall_init (void) {

kernel/amd64/syscallentry.S

@@ -4,6 +4,8 @@
 
 .global amd64_syscall_entry
 amd64_syscall_entry:
+  cli
+
   movq %rsp, %gs:0
   movq %gs:8, %rsp
  

kernel/irq/irq.c

@@ -2,7 +2,7 @@
 #include <libk/list.h>
 #include <libk/std.h>
 #include <mm/liballoc.h>
-#include <sync/rw_spin_lock.h>
+#include <sync/spin_lock.h>
 #include <sys/debug.h>
 
 #if defined(__x86_64__)
@@ -12,9 +12,9 @@
 
 struct irq* irq_table[0x100];
 
-static rw_spin_lock_t irqs_lock;
+static spin_lock_t irqs_lock = SPIN_LOCK_INIT;
 
-bool irq_attach (void (*func) (void*, void*), void* arg, uint32_t irq_num, uint32_t flags) {
+bool irq_attach (void (*func) (void*, void*), void* arg, uint32_t irq_num) {
   spin_lock_ctx_t ctxiqa;
 
   struct irq* irq = malloc (sizeof (*irq));
@@ -25,16 +25,10 @@ bool irq_attach (void (*func) (void*, void*), void* arg, uint32_t irq_num, uint3
   irq->func = func;
   irq->arg = arg;
   irq->irq_num = irq_num;
-  irq->flags = flags;
 
-  rw_spin_write_lock (&irqs_lock, &ctxiqa);
+  spin_lock (&irqs_lock, &ctxiqa);
   irq_table[irq_num] = irq;
-  rw_spin_write_unlock (&irqs_lock, &ctxiqa);
+  spin_unlock (&irqs_lock, &ctxiqa);
-
-#if defined(__x86_64__)
-  uint8_t resolution = amd64_resolve_irq (irq_num);
-  amd64_ioapic_route_irq (irq_num, resolution, 0, amd64_lapic_id ());
-#endif
 
   return true;
 }
@@ -42,11 +36,11 @@ bool irq_attach (void (*func) (void*, void*), void* arg, uint32_t irq_num, uint3
 struct irq* irq_find (uint32_t irq_num) {
   spin_lock_ctx_t ctxiqa;
 
-  rw_spin_read_lock (&irqs_lock, &ctxiqa);
+  spin_lock (&irqs_lock, &ctxiqa);
 
   struct irq* irq = irq_table[irq_num];
 
-  rw_spin_read_unlock (&irqs_lock, &ctxiqa);
+  spin_unlock (&irqs_lock, &ctxiqa);
 
   return irq;
 }

kernel/irq/irq.h

@@ -4,9 +4,6 @@
 #include <libk/list.h>
 #include <libk/std.h>
 
-#define IRQ_INTERRUPT_SAFE   (1 << 0)
-#define IRQ_INTERRUPT_UNSAFE (1 << 1)
-
 typedef void (*irq_func_t) (void* arg, void* regs);
 
 struct irq {
@@ -15,10 +12,9 @@ struct irq {
   irq_func_t func;
   void* arg;
   uint32_t irq_num;
-  uint32_t flags;
 };
 
-bool irq_attach (irq_func_t, void* arg, uint32_t irq_num, uint32_t flags);
+bool irq_attach (irq_func_t, void* arg, uint32_t irq_num);
 struct irq* irq_find (uint32_t irq_num);
 
 #endif // _KERNEL_IRQ_IRQ_H

kernel/proc/kpproc_fb.h

@@ -1,20 +0,0 @@
-#ifndef _KERNEL_PROC_KPPROC_FB_H
-#define _KERNEL_PROC_KPPROC_FB_H
-
-#include <aux/compiler.h>
-#include <libk/std.h>
-
-/* data to expose as a kpproc resource */
-struct kpproc_fb {
-  uintptr_t paddr;
-  uint64_t w, h, pitch;
-  uint16_t bpp;
-  uint8_t red_mask_size;
-  uint8_t red_mask_shift;
-  uint8_t green_mask_size;
-  uint8_t green_mask_shift;
-  uint8_t blue_mask_size;
-  uint8_t blue_mask_shift;
-};
-
-#endif // _KERNEL_PROC_KPPROC_FB_H

kernel/proc/locks.txt

@@ -0,0 +1,10 @@
+Lock hierarchy for process scheduling:
+
+1. proc_tree_lock
+2. cpu->lock
+3. procgroup->lock
+4. proc->lock
+5. sq->lock
+
+1. procgroup_tree_lock
+2. procgroup->lock

kernel/proc/mem.c

@@ -1,33 +0,0 @@
-#include <libk/std.h>
-#include <mm/pmm.h>
-#include <proc/mem.h>
-#include <proc/proc.h>
-#include <proc/resource.h>
-#include <sync/spin_lock.h>
-
-bool proc_create_resource_mem (struct proc_resource_mem* mem, struct proc_resource_mem_init* init) {
-  if (init->pages == 0)
-    return false;
-
-  if (init->managed) {
-    mem->paddr = init->paddr;
-    mem->managed = true;
-  } else {
-    uintptr_t paddr = pmm_alloc (init->pages);
-    if (paddr == PMM_ALLOC_ERR)
-      return false;
-    mem->paddr = paddr;
-    mem->managed = false;
-  }
-
-  mem->pages = mem->alive_pages = init->pages;
-
-  return true;
-}
-
-void proc_cleanup_resource_mem (struct proc* proc, struct proc_resource* resource) {
-  (void)proc;
-
-  if (!resource->u.mem.managed)
-    pmm_free (resource->u.mem.paddr, resource->u.mem.pages);
-}

kernel/proc/mem.h

@@ -1,27 +0,0 @@
-#ifndef _KERNEL_PROC_MEM_H
-#define _KERNEL_PROC_MEM_H
-
-#include <libk/std.h>
-
-struct proc;
-struct proc_resource;
-
-struct proc_resource_mem {
-  struct proc_resource* resource;
-
-  uintptr_t paddr;
-  size_t pages;
-  ptrdiff_t alive_pages;
-  bool managed;
-};
-
-struct proc_resource_mem_init {
-  uintptr_t paddr;
-  size_t pages;
-  bool managed;
-};
-
-bool proc_create_resource_mem (struct proc_resource_mem* mem, struct proc_resource_mem_init* init);
-void proc_cleanup_resource_mem (struct proc* proc, struct proc_resource* resource);
-
-#endif // _KERNEL_PROC_MEM_H

kernel/proc/mutex.c

@@ -2,41 +2,95 @@
 #include <libk/rbtree.h>
 #include <libk/std.h>
 #include <libk/string.h>
+#include <mm/liballoc.h>
 #include <proc/mutex.h>
 #include <proc/proc.h>
+#include <proc/suspension_q.h>
 #include <sync/spin_lock.h>
 #include <sys/debug.h>
+#include <sys/smp.h>
+#include <sys/spin_lock.h>
 
-bool proc_create_resource_mutex (struct proc_mutex* mutex) {
+void proc_mutexes_cleanup (struct proc* proc) {
-  memset (mutex, 0, sizeof (*mutex));
+  spin_lock_ctx_t ctxpg, ctxrs;
 
-  return true;
+  spin_lock (&proc->procgroup->lock, &ctxpg);
+
+  struct rb_node_link* rnode;
+  rbtree_first (&proc->procgroup->resource_tree, rnode);
+
+  while (rnode) {
+    struct rb_node_link* next;
+    rbtree_next (rnode, next);
+
+    struct proc_resource* resource = rbtree_entry (rnode, struct proc_resource, resource_tree_link);
+
+    rnode = next;
+
+    spin_lock (&resource->lock, &ctxrs);
+
+    if (resource->type != PR_MUTEX) {
+      spin_unlock (&resource->lock, &ctxrs);
+      continue;
     }
 
-void proc_cleanup_resource_mutex (struct proc* proc, struct proc_resource* resource) {
+    if (resource->u.mutex.owner == proc && resource->u.mutex.locked) {
+      spin_unlock (&resource->lock, &ctxrs);
+
+      proc_mutex_unlock (proc, &resource->u.mutex);
+    }
+  }
+
+  spin_unlock (&proc->procgroup->lock, &ctxpg);
+}
+
+bool proc_cleanup_resource_mutex (struct proc_resource* resource) {
   struct proc_mutex* mutex = &resource->u.mutex;
+  spin_lock_ctx_t ctxmt, ctxsq;
 
-  proc_mutex_unlock (proc, mutex);
+  spin_lock (&mutex->resource->lock, &ctxmt);
+  spin_lock (&mutex->suspension_q.lock, &ctxsq);
+
+  bool reschedule = PROC_NO_RESCHEDULE;
+
+  while (mutex->suspension_q.proc_list != NULL) {
+    struct list_node_link* node = mutex->suspension_q.proc_list;
+    struct proc_sq_entry* sq_entry = list_entry (node, struct proc_sq_entry, sq_link);
+    struct proc* suspended_proc = sq_entry->proc;
+
+    /* we will relock during resume */
+    spin_unlock (&mutex->suspension_q.lock, &ctxsq);
+    spin_unlock (&mutex->resource->lock, &ctxmt);
+
+    reschedule = reschedule || proc_sq_resume (suspended_proc, sq_entry);
+
+    /* reacquire */
+    spin_lock (&mutex->resource->lock, &ctxmt);
+    spin_lock (&mutex->suspension_q.lock, &ctxsq);
   }
 
-void proc_mutex_lock (struct proc* proc, struct proc_mutex* mutex) {
+  mutex->locked = false;
+  mutex->owner = NULL;
+
+  spin_unlock (&mutex->suspension_q.lock, &ctxsq);
+  spin_unlock (&mutex->resource->lock, &ctxmt);
+
+  return reschedule;
+}
+
+bool proc_mutex_lock (struct proc* proc, struct proc_mutex* mutex) {
   spin_lock_ctx_t ctxmt;
 
-try:
   spin_lock (&mutex->resource->lock, &ctxmt);
 
   if (!mutex->locked || mutex->owner == proc) {
     mutex->locked = true;
     mutex->owner = proc;
     spin_unlock (&mutex->resource->lock, &ctxmt);
-    return;
+    return PROC_NO_RESCHEDULE;
   }
 
-  spin_unlock (&mutex->resource->lock, &ctxmt);
+  return proc_sq_suspend (proc, &mutex->suspension_q, &mutex->resource->lock, &ctxmt);
-
-  proc_suspend (proc, &mutex->suspension_q);
-
-  goto try;
 }
 
 bool proc_mutex_unlock (struct proc* proc, struct proc_mutex* mutex) {
@@ -46,26 +100,24 @@ bool proc_mutex_unlock (struct proc* proc, struct proc_mutex* mutex) {
 
   if (mutex->owner != proc) {
     spin_unlock (&mutex->resource->lock, &ctxmt);
-    return false;
+    return PROC_NO_RESCHEDULE;
   }
 
   spin_lock (&mutex->suspension_q.lock, &ctxsq);
 
-  struct proc* resumed_proc = NULL;
+  struct list_node_link* node = mutex->suspension_q.proc_list;
-  struct rb_node_link* node;
-  rbtree_first (&mutex->suspension_q.proc_tree, node);
 
   if (node) {
-    resumed_proc = rbtree_entry (node, struct proc, suspension_link);
+    struct proc_sq_entry* sq_entry = list_entry (node, struct proc_sq_entry, sq_link);
+    struct proc* resumed_proc = sq_entry->proc;
+
     mutex->owner = resumed_proc;
     mutex->locked = true;
 
     spin_unlock (&mutex->suspension_q.lock, &ctxsq);
     spin_unlock (&mutex->resource->lock, &ctxmt);
 
-    proc_resume (resumed_proc);
+    return proc_sq_resume (resumed_proc, sq_entry);
-
-    return true;
   }
 
   mutex->locked = false;
@@ -74,5 +126,5 @@ bool proc_mutex_unlock (struct proc* proc, struct proc_mutex* mutex) {
   spin_unlock (&mutex->suspension_q.lock, &ctxsq);
   spin_unlock (&mutex->resource->lock, &ctxmt);
 
-  return true;
+  return PROC_NEED_RESCHEDULE;
 }

kernel/proc/mutex.h

@@ -15,9 +15,9 @@ struct proc_mutex {
   struct proc* owner;
 };
 
-bool proc_create_resource_mutex (struct proc_mutex* mutex);
+bool proc_cleanup_resource_mutex (struct proc_resource* resource);
-void proc_cleanup_resource_mutex (struct proc* proc, struct proc_resource* resource);
+bool proc_mutex_lock (struct proc* proc, struct proc_mutex* mutex);
-void proc_mutex_lock (struct proc* proc, struct proc_mutex* mutex);
 bool proc_mutex_unlock (struct proc* proc, struct proc_mutex* mutex);
+void proc_mutexes_cleanup (struct proc* proc);
 
 #endif // _KERNEL_PROC_MUTEX_H

kernel/proc/proc.c

@@ -9,11 +9,10 @@
 #include <limine/requests.h>
 #include <mm/liballoc.h>
 #include <mm/pmm.h>
-#include <proc/kpproc_fb.h>
 #include <proc/proc.h>
+#include <proc/procgroup.h>
 #include <proc/resource.h>
 #include <rd/rd.h>
-#include <sync/rw_spin_lock.h>
 #include <sync/spin_lock.h>
 #include <sys/debug.h>
 #include <sys/mm.h>
@@ -28,117 +27,17 @@
 
 #define SCHED_REAP_FREQ 10
 
-/*
- * Lock hierachy:
- * - proc_tree_lock
- *   - cpu->lock
- *     - proc->lock
- *       - suspension_q->lock
- */
-
 static struct rb_node_link* proc_tree = NULL;
-static rw_spin_lock_t proc_tree_lock = RW_SPIN_LOCK_INIT;
+static spin_lock_t proc_tree_lock = SPIN_LOCK_INIT;
 
 static atomic_int sched_cycles = 0;
 
-/* kernel pseudo process */
-static struct proc kpproc;
-
 static bool proc_check_elf (uint8_t* elf) {
   if (!((elf[0] == 0x7F) && (elf[1] == 'E') && (elf[2] == 'L') && (elf[3] == 'F')))
     return false;
   return true;
 }
 
-bool proc_map (struct proc* proc, uintptr_t start_paddr, uintptr_t start_vaddr, size_t pages,
-               uint32_t flags) {
-  spin_lock_ctx_t ctxprpd;
-  struct proc_mapping* mapping = malloc (sizeof (*mapping));
-
-  if (mapping == NULL)
-    return false;
-
-  mapping->paddr = start_paddr;
-  mapping->vaddr = start_vaddr;
-  mapping->size = pages * PAGE_SIZE;
-
-  flags &= ~(MM_PD_LOCK | MM_PD_RELOAD); /* clear LOCK flag if present, because we lock manualy */
-
-  spin_lock (&proc->pd->lock, &ctxprpd);
-
-  list_append (proc->mappings, &mapping->proc_mappings_link);
-
-  for (uintptr_t vpage = start_vaddr, ppage = start_paddr; vpage < start_vaddr + pages * PAGE_SIZE;
-       vpage += PAGE_SIZE, ppage += PAGE_SIZE) {
-    mm_map_page (proc->pd, ppage, vpage, flags);
-  }
-
-  spin_unlock (&proc->pd->lock, &ctxprpd);
-
-  return true;
-}
-
-bool proc_unmap (struct proc* proc, uintptr_t start_vaddr, size_t pages) {
-  size_t unmap_size = pages * PAGE_SIZE;
-  uintptr_t end_vaddr = start_vaddr + unmap_size;
-  struct list_node_link *mapping_link, *mapping_link_tmp;
-  bool used_tail_mapping = false;
-  spin_lock_ctx_t ctxprpd;
-
-  struct proc_mapping* tail_mapping = malloc (sizeof (*tail_mapping));
-  if (tail_mapping == NULL)
-    return false;
-
-  spin_lock (&proc->pd->lock, &ctxprpd);
-
-  list_foreach (proc->mappings, mapping_link, mapping_link_tmp) {
-    struct proc_mapping* mapping =
-        list_entry (mapping_link, struct proc_mapping, proc_mappings_link);
-
-    uintptr_t m_end = mapping->vaddr + mapping->size;
-
-    /* check overlap */
-    if ((start_vaddr < m_end) && (end_vaddr > mapping->vaddr)) {
-      /* split in the middle */
-      if ((start_vaddr > mapping->vaddr) && (end_vaddr < m_end)) {
-        tail_mapping->vaddr = end_vaddr;
-        tail_mapping->paddr = mapping->paddr + (end_vaddr - mapping->vaddr);
-        tail_mapping->size = m_end - end_vaddr;
-
-        mapping->size = start_vaddr - mapping->vaddr;
-
-        list_insert_after (proc->mappings, &mapping->proc_mappings_link,
-                           &tail_mapping->proc_mappings_link);
-
-        used_tail_mapping = true;
-
-        break;
-      } else if ((start_vaddr <= mapping->vaddr) && (end_vaddr < m_end)) { /* shrink left */
-        size_t diff = end_vaddr - mapping->vaddr;
-        mapping->vaddr += diff;
-        mapping->paddr += diff;
-        mapping->size -= diff;
-      } else if ((start_vaddr > mapping->vaddr) && (end_vaddr >= m_end)) { /* shrink right */
-        mapping->size = start_vaddr - mapping->vaddr;
-      } else { /* full overlap */
-        list_remove (proc->mappings, &mapping->proc_mappings_link);
-        free (mapping);
-      }
-    }
-  }
-
-  if (!used_tail_mapping)
-    free (tail_mapping);
-
-  for (uintptr_t vpage = start_vaddr; vpage < end_vaddr; vpage += PAGE_SIZE) {
-    mm_unmap_page (proc->pd, vpage, 0);
-  }
-
-  spin_unlock (&proc->pd->lock, &ctxprpd);
-
-  return true;
-}
-
 struct elf_aux proc_load_segments (struct proc* proc, uint8_t* elf) {
   struct elf_aux aux;
 
@@ -163,25 +62,37 @@ struct elf_aux proc_load_segments (struct proc* proc, uint8_t* elf) {
 
       size_t blks = div_align_up (phdr->p_memsz + off, PAGE_SIZE);
 
-      struct proc_resource_mem_init mem_init = {.pages = blks};
-      int rid = atomic_fetch_add (&proc->resources->sys_rids, 1);
-      struct proc_resource* r =
-          proc_create_resource (proc, rid, PR_MEM, RV_PRIVATE, (void*)&mem_init);
-      if (r == NULL) {
-        DEBUG ("pmm oom error while loading ELF segments! (tried to alloc %zu blks)\n", blks);
-      }
-
-      uintptr_t p_addr = r->u.mem.paddr;
-
-      memset ((void*)((uintptr_t)hhdm->offset + p_addr), 0, blks * PAGE_SIZE);
-      memcpy ((void*)((uintptr_t)hhdm->offset + p_addr + off),
-              (void*)((uintptr_t)elf + phdr->p_offset), phdr->p_filesz);
-
       uint32_t pg_flags = MM_PG_USER | MM_PG_PRESENT;
       if (phdr->p_flags & PF_W)
         pg_flags |= MM_PG_RW;
 
-      proc_map (proc, p_addr, v_addr, blks, pg_flags);
+      uintptr_t p_addr;
+      procgroup_map (proc->procgroup, v_addr, blks, pg_flags, &p_addr);
+
+      memset ((void*)((uintptr_t)hhdm->offset + p_addr), 0, blks * PAGE_SIZE);
+      memcpy ((void*)((uintptr_t)hhdm->offset + p_addr + off),
+              (void*)((uintptr_t)elf + phdr->p_offset), phdr->p_filesz);
+    } break;
+    case PT_TLS: {
+#if defined(__x86_64__)
+      if (phdr->p_memsz > 0) {
+        size_t tls_align = phdr->p_align ? phdr->p_align : sizeof (uintptr_t);
+        size_t tls_size = align_up (phdr->p_memsz, tls_align);
+        size_t tls_total_needed = tls_size + sizeof (uintptr_t);
+        size_t blks = div_align_up (tls_total_needed, PAGE_SIZE);
+        proc->procgroup->tls.tls_tmpl_pages = blks;
+        proc->procgroup->tls.tls_tmpl_size = tls_size;
+        proc->procgroup->tls.tls_tmpl_total_size = tls_total_needed;
+
+        proc->procgroup->tls.tls_tmpl = malloc (blks * PAGE_SIZE);
+        memset (proc->procgroup->tls.tls_tmpl, 0, blks * PAGE_SIZE);
+
+        memcpy (proc->procgroup->tls.tls_tmpl, (void*)((uintptr_t)elf + phdr->p_offset),
+                phdr->p_filesz);
+
+        proc_init_tls (proc);
+      }
+#endif
     } break;
     }
   }
@@ -189,11 +100,10 @@ struct elf_aux proc_load_segments (struct proc* proc, uint8_t* elf) {
   return aux;
 }
 
-static struct proc* proc_spawn_rd (char* name) {
+struct proc* proc_spawn_rd (char* name) {
   struct rd_file* rd_file = rd_get_file (name);
 
   bool ok = proc_check_elf (rd_file->content);
-  DEBUG ("ELF magic %s\n", (ok ? "OK" : "BAD"));
 
   if (!ok)
     return NULL;
@@ -205,61 +115,61 @@ struct proc* proc_find_pid (int pid) {
   spin_lock_ctx_t ctxprtr;
   struct proc* proc = NULL;
 
-  rw_spin_read_lock (&proc_tree_lock, &ctxprtr);
+  spin_lock (&proc_tree_lock, &ctxprtr);
   rbtree_find (struct proc, &proc_tree, pid, proc, proc_tree_link, pid);
-  rw_spin_read_unlock (&proc_tree_lock, &ctxprtr);
+  spin_unlock (&proc_tree_lock, &ctxprtr);
 
   return proc;
 }
 
-void proc_register (struct proc* proc, struct cpu* cpu) {
+void proc_register (struct proc* proc, struct cpu* cpu1) {
   spin_lock_ctx_t ctxcpu, ctxprtr;
 
-  proc->cpu = cpu;
+  proc->cpu = cpu1 != NULL ? cpu1 : cpu_find_lightest ();
 
+  struct cpu* cpu = proc->cpu;
+
+  spin_lock (&proc_tree_lock, &ctxprtr);
   spin_lock (&cpu->lock, &ctxcpu);
-  rbtree_insert (struct proc, &cpu->proc_run_q, &proc->cpu_run_q_link, cpu_run_q_link, pid);
 
+  rbtree_insert (struct proc, &proc_tree, &proc->proc_tree_link, proc_tree_link, pid);
+
+  atomic_fetch_add (&cpu->proc_run_q_count, 1);
+  list_append (cpu->proc_run_q, &proc->cpu_run_q_link);
   if (cpu->proc_current == NULL)
     cpu->proc_current = proc;
 
+  spin_unlock (&proc_tree_lock, &ctxprtr);
   spin_unlock (&cpu->lock, &ctxcpu);
-
-  rw_spin_write_lock (&proc_tree_lock, &ctxprtr);
-  rbtree_insert (struct proc, &proc_tree, &proc->proc_tree_link, proc_tree_link, pid);
-  rw_spin_write_unlock (&proc_tree_lock, &ctxprtr);
 }
 
 /* caller holds cpu->lock */
 static struct proc* proc_find_sched (struct cpu* cpu) {
-  struct rb_node_link* node = NULL;
+  if (!cpu->proc_run_q)
-  struct proc* current = cpu->proc_current;
-  struct proc* proc = NULL;
-
-  if (current)
-    rbtree_next (&current->cpu_run_q_link, node);
-
-  if (!node)
-    rbtree_first (&cpu->proc_run_q, node);
-
-  if (!node)
     return NULL;
 
-  struct rb_node_link* first = node;
+  struct list_node_link *current, *start;
+
+  if (cpu->proc_current)
+    current = cpu->proc_current->cpu_run_q_link.next;
+  else
+    current = cpu->proc_run_q;
+
+  if (!current)
+    current = cpu->proc_run_q;
+
+  start = current;
+
   do {
-    proc = rbtree_entry (node, struct proc, cpu_run_q_link);
+    struct proc* proc = list_entry (current, struct proc, cpu_run_q_link);
 
     if (atomic_load (&proc->state) == PROC_READY)
       return proc;
 
-    rbtree_next (node, node);
+    current = current->next ? current->next : cpu->proc_run_q;
+  } while (current != start);
 
-    if (!node)
+  return NULL;
-      rbtree_first (&cpu->proc_run_q, node);
-
-  } while (node != first);
-
-  return ((atomic_load (&current->state) == PROC_READY) ? current : NULL);
 }
 
 static void proc_reap (void) {
@@ -268,7 +178,7 @@ static void proc_reap (void) {
   spin_lock_ctx_t ctxprtr;
   spin_lock_ctx_t ctxpr;
 
-  rw_spin_write_lock (&proc_tree_lock, &ctxprtr);
+  spin_lock (&proc_tree_lock, &ctxprtr);
 
   struct rb_node_link* node;
   rbtree_first (&proc_tree, node);
@@ -281,15 +191,14 @@ static void proc_reap (void) {
     if (atomic_load (&proc->state) == PROC_DEAD) {
       spin_lock (&proc->lock, &ctxpr);
       rbtree_delete (&proc_tree, &proc->proc_tree_link);
-      spin_unlock (&proc->lock, &ctxpr);
-
       list_append (reap_list, &proc->reap_link);
+      spin_unlock (&proc->lock, &ctxpr);
     }
 
     node = next;
   }
 
-  rw_spin_write_unlock (&proc_tree_lock, &ctxprtr);
+  spin_unlock (&proc_tree_lock, &ctxprtr);
 
   struct list_node_link *reap_link, *reap_link_tmp;
   list_foreach (reap_list, reap_link, reap_link_tmp) {
@@ -301,8 +210,8 @@ static void proc_reap (void) {
   }
 }
 
-void proc_sched (void* regs) {
+void proc_sched (void) {
-  spin_lock_ctx_t ctxcpu, ctxpr;
+  spin_lock_ctx_t ctxcpu;
 
   int s_cycles = atomic_fetch_add (&sched_cycles, 1);
 
@@ -314,21 +223,12 @@ void proc_sched (void* regs) {
 
   spin_lock (&cpu->lock, &ctxcpu);
 
-  struct proc* prev = cpu->proc_current;
-
-  if (prev != NULL) {
-    spin_lock (&prev->lock, &ctxpr);
-    prev->pdata.regs = *(struct saved_regs*)regs;
-    spin_unlock (&prev->lock, &ctxpr);
-  }
-
   next = proc_find_sched (cpu);
 
   if (next) {
     cpu->proc_current = next;
-    spin_unlock (&cpu->lock, &ctxcpu);
 
-    do_sched (next);
+    do_sched (next, &cpu->lock, &ctxcpu);
   } else {
     cpu->proc_current = NULL;
     spin_unlock (&cpu->lock, &ctxcpu);
@@ -343,11 +243,13 @@ void proc_kill (struct proc* proc) {
 
   spin_lock (&proc->lock, &ctxpr);
   atomic_store (&proc->state, PROC_DEAD);
+  proc->cpu = NULL;
   spin_unlock (&proc->lock, &ctxpr);
 
   spin_lock (&cpu->lock, &ctxcpu);
 
-  rbtree_delete (&cpu->proc_run_q, &proc->cpu_run_q_link);
+  list_remove (cpu->proc_run_q, &proc->cpu_run_q_link);
+  atomic_fetch_sub (&cpu->proc_run_q_count, 1);
   if (cpu->proc_current == proc)
     cpu->proc_current = NULL;
 
@@ -358,118 +260,24 @@ void proc_kill (struct proc* proc) {
   cpu_request_sched (cpu);
 }
 
-void proc_suspend (struct proc* proc, struct proc_suspension_q* sq) {
-  spin_lock_ctx_t ctxpr, ctxcpu, ctxsq;
-  struct cpu* cpu = proc->cpu;
-
-  spin_lock (&proc->lock, &ctxpr);
-  atomic_store (&proc->state, PROC_SUSPENDED);
-  proc->suspension_q = sq;
-  spin_unlock (&proc->lock, &ctxpr);
-
-  /* remove from run q */
-  spin_lock (&cpu->lock, &ctxcpu);
-
-  rbtree_delete (&cpu->proc_run_q, &proc->cpu_run_q_link);
-  if (cpu->proc_current == proc)
-    cpu->proc_current = NULL;
-  spin_unlock (&cpu->lock, &ctxcpu);
-
-  spin_lock (&sq->lock, &ctxsq);
-  rbtree_insert (struct proc, &sq->proc_tree, &proc->suspension_link, suspension_link, pid);
-  spin_unlock (&sq->lock, &ctxsq);
-
-  cpu_request_sched (cpu);
-}
-
-void proc_resume (struct proc* proc) {
-  spin_lock_ctx_t ctxsq, ctxpr, ctxcpu;
-  struct cpu* cpu = proc->cpu;
-  struct proc_suspension_q* sq = proc->suspension_q;
-
-  spin_lock (&sq->lock, &ctxsq);
-  rbtree_delete (&sq->proc_tree, &proc->suspension_link);
-  spin_unlock (&sq->lock, &ctxsq);
-
-  spin_lock (&proc->lock, &ctxpr);
-  proc->suspension_q = NULL;
-  atomic_store (&proc->state, PROC_READY);
-  spin_unlock (&proc->lock, &ctxpr);
-
-  spin_lock (&cpu->lock, &ctxcpu);
-  rbtree_insert (struct proc, &cpu->proc_run_q, &proc->cpu_run_q_link, cpu_run_q_link, pid);
-  spin_unlock (&cpu->lock, &ctxcpu);
-
-  cpu_request_sched (cpu);
-}
-
 static void proc_irq_sched (void* arg, void* regs) {
   (void)arg;
-  proc_sched (regs);
+  proc_sched ();
-}
-
-static void proc_kpproc_init (void) {
-  struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-
-  memset (&kpproc, 0, sizeof (kpproc));
-
-  kpproc.lock = SPIN_LOCK_INIT;
-  kpproc.state = PROC_PSEUDO;
-  kpproc.pid = 0;
-
-  kpproc.resources = malloc (sizeof (*kpproc.resources));
-  kpproc.resources->tree = NULL;
-  kpproc.resources->lock = RW_SPIN_LOCK_INIT;
-  kpproc.resources->refs = 1;
-  kpproc.resources->sys_rids = 0;
-
-  kpproc.pd = mm_get_kernel_pd ();
-  kpproc.cpu = thiscpu;
-
-  rbtree_insert (struct proc, &proc_tree, &kpproc.proc_tree_link, proc_tree_link, pid);
-
-  /* prepare kernel resources */
-  {
-    /* frame buffer */
-
-    struct limine_framebuffer_response* fb = limine_framebuffer_request.response;
-    struct kpproc_fb fb_info = {
-        .paddr = (uintptr_t)fb->framebuffers[0]->address - (uintptr_t)hhdm->offset,
-        .w = fb->framebuffers[0]->width,
-        .h = fb->framebuffers[0]->height,
-        .pitch = fb->framebuffers[0]->pitch,
-        .bpp = fb->framebuffers[0]->bpp,
-        .red_mask_size = fb->framebuffers[0]->red_mask_size,
-        .red_mask_shift = fb->framebuffers[0]->red_mask_shift,
-        .green_mask_size = fb->framebuffers[0]->green_mask_size,
-        .green_mask_shift = fb->framebuffers[0]->green_mask_shift,
-        .blue_mask_size = fb->framebuffers[0]->blue_mask_size,
-        .blue_mask_shift = fb->framebuffers[0]->blue_mask_shift,
-    };
-
-    DEBUG ("Framebuffer address %p\n", fb_info.paddr);
-
-    size_t pages = align_up (sizeof (fb_info), PAGE_SIZE) / PAGE_SIZE;
-    uintptr_t fb_info_memblk_paddr = pmm_alloc (pages);
-    memcpy ((struct kpproc_fb*)((uintptr_t)hhdm->offset + fb_info_memblk_paddr), &fb_info,
-            sizeof (fb_info));
-
-    struct proc_resource_mem_init mem_init = {
-        .pages = pages, .paddr = fb_info_memblk_paddr, .managed = true};
-    proc_create_resource (&kpproc, 0, PR_MEM, RV_PUBLIC, &mem_init);
-  }
 }
 
 void proc_init (void) {
 #if defined(__x86_64__)
-  irq_attach (&proc_irq_sched, NULL, SCHED_PREEMPT_TIMER, IRQ_INTERRUPT_SAFE);
+  irq_attach (&proc_irq_sched, NULL, SCHED_PREEMPT_TIMER);
-  irq_attach (&proc_irq_sched, NULL, CPU_REQUEST_SCHED, IRQ_INTERRUPT_SAFE);
+  irq_attach (&proc_irq_sched, NULL, CPU_REQUEST_SCHED);
 #endif
 
-  proc_kpproc_init ();
+  struct proc* spin_proc = proc_spawn_rd ("spin.exe");
+  proc_register (spin_proc, thiscpu);
 
   struct proc* init = proc_spawn_rd ("init.exe");
-  proc_register (init, thiscpu);
+  proc_register (init, NULL);
 
-  do_sched (init);
+  spin_lock_ctx_t ctxcpu;
+  spin_lock (&spin_proc->cpu->lock, &ctxcpu);
+  do_sched (spin_proc, &spin_proc->cpu->lock, &ctxcpu);
 }

kernel/proc/proc.h

@@ -6,9 +6,9 @@
 #include <libk/list.h>
 #include <libk/rbtree.h>
 #include <libk/std.h>
+#include <proc/procgroup.h>
 #include <proc/resource.h>
 #include <proc/suspension_q.h>
-#include <sync/rw_spin_lock.h>
 #include <sync/spin_lock.h>
 #include <sys/mm.h>
 
@@ -17,59 +17,41 @@
 #include <amd64/proc.h> /* USTACK_SIZE */
 #endif
 
+#define PROC_NEED_RESCHEDULE true
+#define PROC_NO_RESCHEDULE   false
+
 /* process states */
 #define PROC_READY     0
 #define PROC_DEAD      1
 #define PROC_SUSPENDED 2
-#define PROC_PSEUDO    3
 
+/* process flags */
 #define PROC_USTK_PREALLOC (1 << 0)
 
 struct cpu;
 
-struct proc_mapping {
-  struct list_node_link proc_mappings_link;
-
-  uintptr_t paddr;
-  uintptr_t vaddr;
-  size_t size;
-};
-
-struct proc_resources {
-  atomic_int refs;
-  atomic_int sys_rids;
-  struct rb_node_link* tree;
-  rw_spin_lock_t lock;
-};
-
 struct proc {
   int pid;
   struct rb_node_link proc_tree_link;
-  struct rb_node_link cpu_run_q_link;
+  struct rb_node_link procgroup_memb_tree_link;
-  struct rb_node_link suspension_link;
+  struct list_node_link cpu_run_q_link;
   struct list_node_link reap_link;
-
+  struct list_node_link* sq_entries;
-  struct list_node_link* mappings; /* pd.lock implicitly protects this field */
+  struct procgroup* procgroup;
   struct proc_platformdata pdata;
   uint32_t flags;
-  struct pd* pd;
   spin_lock_t lock;
   struct cpu* cpu;
   atomic_int state;
-  struct proc_suspension_q* suspension_q;
+  uintptr_t uvaddr_argument;
-  struct proc_resources* resources;
 };
 
-void proc_suspend (struct proc* proc, struct proc_suspension_q* sq);
+void proc_sched (void);
-void proc_resume (struct proc* proc);
-void proc_sched (void* regs);
 void proc_kill (struct proc* proc);
-bool proc_map (struct proc* proc, uintptr_t start_paddr, uintptr_t start_vaddr, size_t pages,
-               uint32_t flags);
-bool proc_unmap (struct proc* proc, uintptr_t start_vaddr, size_t pages);
 struct elf_aux proc_load_segments (struct proc* proc, uint8_t* elf);
 void proc_register (struct proc* proc, struct cpu* cpu);
 struct proc* proc_find_pid (int pid);
+struct proc* proc_spawn_rd (char* name);
 void proc_init (void);
 
 #endif // _KERNEL_PROC_PROC_H

kernel/proc/procgroup.c

@@ -0,0 +1,218 @@
+#include <libk/rbtree.h>
+#include <libk/std.h>
+#include <mm/liballoc.h>
+#include <mm/pmm.h>
+#include <proc/proc.h>
+#include <proc/procgroup.h>
+#include <sync/spin_lock.h>
+#include <sys/debug.h>
+#include <sys/mm.h>
+
+static struct rb_node_link* procgroup_tree = NULL;
+static spin_lock_t procgroup_tree_lock = SPIN_LOCK_INIT;
+static atomic_int pgids = 0;
+
+uintptr_t procgroup_map (struct procgroup* procgroup, uintptr_t vaddr, size_t pages, uint32_t flags,
+                         uintptr_t* out_paddr) {
+  spin_lock_ctx_t ctxpg;
+
+  spin_lock (&procgroup->lock, &ctxpg);
+
+  vaddr = (vaddr == 0) ? procgroup->map_base : vaddr;
+
+  struct proc_mapping* mapping = malloc (sizeof (*mapping));
+
+  if (mapping == NULL) {
+    spin_unlock (&procgroup->lock, &ctxpg);
+    return 0;
+  }
+
+  uintptr_t paddr = pmm_alloc (pages);
+
+  if (paddr == PMM_ALLOC_ERR) {
+    free (mapping);
+    spin_unlock (&procgroup->lock, &ctxpg);
+    return 0;
+  }
+
+  if (out_paddr != NULL)
+    *out_paddr = paddr;
+
+  mapping->paddr = paddr;
+  mapping->vaddr = vaddr;
+  mapping->size = pages * PAGE_SIZE;
+
+  procgroup->map_base += pages * PAGE_SIZE;
+
+  list_append (procgroup->mappings, &mapping->proc_mappings_link);
+
+  for (uintptr_t vpage = vaddr, ppage = paddr; vpage < vaddr + pages * PAGE_SIZE;
+       vpage += PAGE_SIZE, ppage += PAGE_SIZE) {
+    mm_map_page (&procgroup->pd, ppage, vpage, flags);
+  }
+
+  spin_unlock (&procgroup->lock, &ctxpg);
+
+  return vaddr;
+}
+
+bool procgroup_unmap (struct procgroup* procgroup, uintptr_t start_vaddr, size_t pages) {
+  size_t unmap_size = pages * PAGE_SIZE;
+  uintptr_t end_vaddr = start_vaddr + unmap_size;
+
+  struct list_node_link *mapping_link, *mapping_link_tmp;
+
+  bool used_tail_mapping = false;
+  spin_lock_ctx_t ctxpg;
+
+  struct proc_mapping* tail_mapping = malloc (sizeof (*tail_mapping));
+  if (tail_mapping == NULL)
+    return false;
+
+  spin_lock (&procgroup->lock, &ctxpg);
+
+  list_foreach (procgroup->mappings, mapping_link, mapping_link_tmp) {
+    struct proc_mapping* mapping =
+        list_entry (mapping_link, struct proc_mapping, proc_mappings_link);
+
+    uintptr_t m_start = mapping->vaddr;
+    uintptr_t m_end = mapping->vaddr + mapping->size;
+
+    /* check overlap */
+    if ((start_vaddr < m_end) && (end_vaddr > mapping->vaddr)) {
+      uintptr_t free_vstart = (start_vaddr > m_start) ? start_vaddr : m_start;
+      uintptr_t free_vend = (end_vaddr < m_end) ? end_vaddr : m_end;
+      size_t free_size = free_vend - free_vstart;
+
+      uintptr_t ppage_to_free = mapping->paddr + (free_vstart - m_start);
+      pmm_free (ppage_to_free, free_size / PAGE_SIZE);
+
+      /* split in the middle */
+      if ((start_vaddr > m_start) && (end_vaddr < m_end)) {
+        tail_mapping->vaddr = end_vaddr;
+        tail_mapping->paddr = mapping->paddr + (end_vaddr - m_start);
+        tail_mapping->size = m_end - end_vaddr;
+
+        mapping->size = start_vaddr - m_start;
+
+        list_insert_after (procgroup->mappings, &mapping->proc_mappings_link,
+                           &tail_mapping->proc_mappings_link);
+
+        used_tail_mapping = true;
+
+        break;
+      } else if ((start_vaddr <= m_start) && (end_vaddr < m_end)) { /* shrink left */
+        size_t diff = end_vaddr - m_start;
+        mapping->vaddr += diff;
+        mapping->paddr += diff;
+        mapping->size -= diff;
+      } else if ((start_vaddr > m_start) && (end_vaddr >= m_end)) { /* shrink right */
+        mapping->size = start_vaddr - m_start;
+      } else { /* full overlap */
+        list_remove (procgroup->mappings, &mapping->proc_mappings_link);
+        free (mapping);
+      }
+    }
+  }
+
+  if (!used_tail_mapping)
+    free (tail_mapping);
+
+  for (uintptr_t vpage = start_vaddr; vpage < end_vaddr; vpage += PAGE_SIZE) {
+    mm_unmap_page (&procgroup->pd, vpage);
+  }
+
+  spin_unlock (&procgroup->lock, &ctxpg);
+
+  return true;
+}
+
+struct procgroup* procgroup_create (void) {
+  spin_lock_ctx_t ctxpgtr;
+
+  struct procgroup* procgroup = malloc (sizeof (*procgroup));
+  if (procgroup == NULL) {
+    return NULL;
+  }
+
+  procgroup->refs = 0;
+  procgroup->memb_proc_tree = NULL;
+  procgroup->lock = SPIN_LOCK_INIT;
+  procgroup->pgid = atomic_fetch_add (&pgids, 1);
+  procgroup->pd.cr3_paddr = mm_alloc_user_pd_phys ();
+  procgroup->map_base = PROC_MAP_BASE;
+
+  spin_lock (&procgroup_tree_lock, &ctxpgtr);
+  rbtree_insert (struct procgroup, &procgroup_tree, &procgroup->procgroup_tree_link,
+                 procgroup_tree_link, pgid);
+  spin_unlock (&procgroup_tree_lock, &ctxpgtr);
+
+  return procgroup;
+}
+
+void procgroup_attach (struct procgroup* procgroup, struct proc* proc) {
+  spin_lock_ctx_t ctxpg, ctxpr;
+
+  spin_lock (&procgroup->lock, &ctxpg);
+  spin_lock (&proc->lock, &ctxpr);
+
+  rbtree_insert (struct proc, &procgroup->memb_proc_tree, &proc->procgroup_memb_tree_link,
+                 procgroup_memb_tree_link, pid);
+  atomic_fetch_add (&procgroup->refs, 1);
+
+  spin_unlock (&proc->lock, &ctxpr);
+  spin_unlock (&procgroup->lock, &ctxpg);
+}
+
+void procgroup_detach (struct procgroup* procgroup, struct proc* proc) {
+  spin_lock_ctx_t ctxpg, ctxpr, ctxpgtr;
+
+  spin_lock (&procgroup->lock, &ctxpg);
+  spin_lock (&proc->lock, &ctxpr);
+
+  rbtree_delete (&procgroup->memb_proc_tree, &proc->procgroup_memb_tree_link);
+  int refs = atomic_fetch_sub (&procgroup->refs, 1);
+
+  spin_unlock (&proc->lock, &ctxpr);
+  spin_unlock (&procgroup->lock, &ctxpg);
+
+  if (refs == 1) {
+    spin_lock (&procgroup_tree_lock, &ctxpgtr);
+    spin_lock (&procgroup->lock, &ctxpg);
+
+    rbtree_delete (&procgroup_tree, &procgroup->procgroup_tree_link);
+
+    spin_unlock (&procgroup->lock, &ctxpg);
+    spin_unlock (&procgroup_tree_lock, &ctxpgtr);
+
+    /* delete resources */
+    struct rb_node_link* rnode;
+    rbtree_first (&procgroup->resource_tree, rnode);
+    while (rnode) {
+      struct rb_node_link* next;
+      rbtree_next (rnode, next);
+
+      struct proc_resource* resource =
+          rbtree_entry (rnode, struct proc_resource, resource_tree_link);
+
+      rnode = next;
+
+      proc_delete_resource (resource);
+    }
+
+    struct list_node_link *mapping_link, *mapping_link_tmp;
+    list_foreach (procgroup->mappings, mapping_link, mapping_link_tmp) {
+      struct proc_mapping* mapping =
+          list_entry (mapping_link, struct proc_mapping, proc_mappings_link);
+
+      pmm_free (mapping->paddr, mapping->size / PAGE_SIZE);
+      free (mapping);
+    }
+
+    pmm_free (procgroup->pd.cr3_paddr, 1);
+
+    free (procgroup->tls.tls_tmpl);
+
+    free (procgroup);
+  }
+}

kernel/proc/procgroup.h

@@ -0,0 +1,43 @@
+#ifndef _KERNEL_PROC_PROCGROUP_H
+#define _KERNEL_PROC_PROCGROUP_H
+
+#include <libk/list.h>
+#include <libk/rbtree.h>
+#include <libk/std.h>
+#include <proc/resource.h>
+#include <sync/spin_lock.h>
+#include <sys/mm.h>
+#include <sys/procgroup.h>
+
+struct proc;
+
+struct proc_mapping {
+  struct list_node_link proc_mappings_link;
+
+  uintptr_t paddr;
+  uintptr_t vaddr;
+  size_t size;
+};
+
+struct procgroup {
+  int pgid;
+  struct rb_node_link procgroup_tree_link;
+  struct rb_node_link* memb_proc_tree;
+  spin_lock_t lock;
+  atomic_int refs;
+  struct rb_node_link* resource_tree;
+  atomic_int sys_rids;
+  struct pd pd;
+  struct list_node_link* mappings;
+  uintptr_t map_base;
+  struct procgroup_tls tls;
+};
+
+struct procgroup* procgroup_create (void);
+void procgroup_attach (struct procgroup* procgroup, struct proc* proc);
+void procgroup_detach (struct procgroup* procgroup, struct proc* proc);
+uintptr_t procgroup_map (struct procgroup* procgroup, uintptr_t vaddr, size_t pages, uint32_t flags,
+                         uintptr_t* out_paddr);
+bool procgroup_unmap (struct procgroup* procgroup, uintptr_t start_vaddr, size_t pages);
+
+#endif // _KERNEL_PROC_PROCGROUP_H

kernel/proc/resource.c

@@ -7,169 +7,53 @@
 #include <mm/pmm.h>
 #include <proc/mutex.h>
 #include <proc/proc.h>
+#include <proc/procgroup.h>
 #include <proc/resource.h>
 #include <sync/spin_lock.h>
 #include <sys/debug.h>
 
-static struct rb_node_link* resource_tree = NULL;
+struct proc_resource* proc_find_resource (struct procgroup* procgroup, int rid) {
-static rw_spin_lock_t resource_tree_lock = RW_SPIN_LOCK_INIT;
+  spin_lock_ctx_t ctxpg;
-
-void proc_cleanup_resources (struct proc* proc) {
-  spin_lock_ctx_t ctxrs;
-
-  rw_spin_write_lock (&proc->resources->lock, &ctxrs);
-
-  struct rb_node_link* rnode;
-  rbtree_first (&proc->resources->tree, rnode);
-  while (rnode) {
-    struct rb_node_link* next;
-    rbtree_next (rnode, next);
-
-    struct proc_resource* resource =
-        rbtree_entry (rnode, struct proc_resource, local_resource_tree_link);
-
-    rnode = next;
-
-    proc_drop_resource (proc, resource, false);
-  }
-
-  rw_spin_write_unlock (&proc->resources->lock, &ctxrs);
-
-  if (atomic_fetch_sub (&proc->resources->refs, 1) == 1) {
-    free (proc->resources);
-  }
-}
-
-void proc_drop_resource (struct proc* proc, struct proc_resource* resource, bool lock) {
-  spin_lock_ctx_t ctxrs;
-
-  DEBUG ("resource=%p created_by=%d vis=%d type=%d rid=%d refs=%d\n", resource,
-         resource->created_by_pid, resource->visibility, resource->type, resource->rid,
-         atomic_load (&resource->refs));
-
-  if (atomic_fetch_sub (&resource->refs, 1) == 1) {
-    switch (resource->visibility) {
-    case RV_PRIVATE: {
-      if (lock)
-        rw_spin_write_lock (&proc->resources->lock, &ctxrs);
-
-      rbtree_delete (&proc->resources->tree, &resource->local_resource_tree_link);
-
-      if (lock)
-        rw_spin_write_unlock (&proc->resources->lock, &ctxrs);
-    } break;
-    case RV_PUBLIC: {
-      if (lock)
-        rw_spin_write_lock (&resource_tree_lock, &ctxrs);
-
-      rbtree_delete (&resource_tree, &resource->global_resource_tree_link);
-
-      if (lock)
-        rw_spin_write_unlock (&resource_tree_lock, &ctxrs);
-    } break;
-    default: {
-      assert (0);
-    } break;
-    }
-
-    resource->ops.cleanup (proc, resource);
-    free (resource);
-  }
-}
-
-struct proc_resource* proc_find_resource (struct proc* proc, int rid, int vis) {
   struct proc_resource* resource = NULL;
-  spin_lock_ctx_t ctxrs;
 
-  switch (vis) {
+  spin_lock (&procgroup->lock, &ctxpg);
-  case RV_PRIVATE: {
+  rbtree_find (struct proc_resource, &procgroup->resource_tree, rid, resource, resource_tree_link,
-    /* User wants to create a private resource, so search locally */
-    rw_spin_read_lock (&proc->resources->lock, &ctxrs);
-    rbtree_find (struct proc_resource, &proc->resources->tree, rid, resource,
-                 local_resource_tree_link, rid);
-    rw_spin_read_unlock (&proc->resources->lock, &ctxrs);
-  } break;
-  case RV_PUBLIC: {
-    /* User wants to create a public resource, so search globally */
-    rw_spin_read_lock (&resource_tree_lock, &ctxrs);
-    rbtree_find (struct proc_resource, &resource_tree, rid, resource, global_resource_tree_link,
                rid);
-    rw_spin_read_unlock (&resource_tree_lock, &ctxrs);
+  spin_unlock (&procgroup->lock, &ctxpg);
-  } break;
-  default: {
-    assert (0);
-  } break;
-  }
 
   return resource;
 }
 
-struct proc_resource* proc_create_resource (struct proc* proc, int rid, int type, int vis,
+struct proc_resource* proc_create_resource_mutex (struct procgroup* procgroup, int rid) {
-                                            void* data) {
+  spin_lock_ctx_t ctxpg;
-  spin_lock_ctx_t ctxrs;
+  struct proc_resource* resource;
 
-  /* Check if resource RID already exists */
+  resource = proc_find_resource (procgroup, rid);
-  struct proc_resource* resource_check = proc_find_resource (proc, rid, vis);
+  if (resource != NULL)
+    return resource;
 
-  /* Resource was found either way, so it already exists */
+  resource = malloc (sizeof (*resource));
-  if (resource_check != NULL)
-    return NULL;
-
-  /* create the resource */
-
-  struct proc_resource* resource = malloc (sizeof (*resource));
   if (resource == NULL)
     return NULL;
 
   memset (resource, 0, sizeof (*resource));
-
   resource->lock = SPIN_LOCK_INIT;
-  resource->type = type;
-  resource->refs = 1;
-  resource->rid = rid;
-  resource->visibility = vis;
-  resource->created_by_pid = proc->pid;
-
-  switch (resource->type) {
-  case PR_MEM: {
-    struct proc_resource_mem_init* mem_init = data;
-    proc_create_resource_mem (&resource->u.mem, mem_init);
-    resource->ops.cleanup = &proc_cleanup_resource_mem;
-    resource->u.mem.resource = resource;
-    DEBUG ("PR_MEM resource=%p created_by=%d, type=%d rid=%d paddr=%p, pages=%zu\n", resource,
-           resource->created_by_pid, resource->type, resource->rid, resource->u.mem.paddr,
-           resource->u.mem.pages);
-  } break;
-  case PR_MUTEX: {
-    proc_create_resource_mutex (&resource->u.mutex);
   resource->ops.cleanup = &proc_cleanup_resource_mutex;
   resource->u.mutex.resource = resource;
-    DEBUG ("PR_MUTEX resource=%p created_by=%d type=%d rid=%d\n", resource,
+  resource->rid = rid;
-           resource->created_by_pid, resource->type, resource->rid);
+  resource->type = PR_MUTEX;
-  } break;
-  default: {
-    free (resource);
-    return NULL;
-  } break;
-  }
 
-  switch (resource->visibility) {
+  spin_lock (&procgroup->lock, &ctxpg);
-  case RV_PRIVATE: {
+  rbtree_insert (struct proc_resource, &procgroup->resource_tree, &resource->resource_tree_link,
-    rw_spin_write_lock (&proc->resources->lock, &ctxrs);
+                 resource_tree_link, rid);
-    rbtree_insert (struct proc_resource, &proc->resources->tree,
+  spin_unlock (&procgroup->lock, &ctxpg);
-                   &resource->local_resource_tree_link, local_resource_tree_link, rid);
-    rw_spin_write_unlock (&proc->resources->lock, &ctxrs);
-  } break;
-  case RV_PUBLIC: {
-    rw_spin_write_lock (&resource_tree_lock, &ctxrs);
-    rbtree_insert (struct proc_resource, &resource_tree, &resource->global_resource_tree_link,
-                   global_resource_tree_link, rid);
-    rw_spin_write_unlock (&resource_tree_lock, &ctxrs);
-  } break;
-  default: {
-    assert (0);
-  } break;
-  }
 
   return resource;
 }
+
+bool proc_delete_resource (struct proc_resource* resource) {
+  bool reschedule = resource->ops.cleanup (resource);
+  free (resource);
+
+  return reschedule;
+}

kernel/proc/resource.h

@@ -4,40 +4,29 @@
 #include <libk/list.h>
 #include <libk/rbtree.h>
 #include <libk/std.h>
-#include <proc/mem.h>
 #include <proc/mutex.h>
 #include <sync/spin_lock.h>
 
-#define PR_MEM   0
 #define PR_MUTEX 1
 
-#define RV_PRIVATE 0
-#define RV_PUBLIC  1
-
 struct proc;
+struct procgroup;
 
 struct proc_resource {
   int type;
   int rid;
-  int visibility;
   spin_lock_t lock;
-  atomic_int refs;
+  struct rb_node_link resource_tree_link;
-  struct rb_node_link global_resource_tree_link;
-  struct rb_node_link local_resource_tree_link;
   union {
-    struct proc_resource_mem mem;
     struct proc_mutex mutex;
   } u;
   struct {
-    void (*cleanup) (struct proc* proc, struct proc_resource* resource);
+    bool (*cleanup) (struct proc_resource* resource);
   } ops;
-  int created_by_pid;
 };
 
-struct proc_resource* proc_create_resource (struct proc* proc, int rid, int type, int vis,
+struct proc_resource* proc_find_resource (struct procgroup* procgroup, int rid);
-                                            void* data);
+struct proc_resource* proc_create_resource_mutex (struct procgroup* procgroup, int rid);
-struct proc_resource* proc_find_resource (struct proc* proc, int rid, int vis);
+bool proc_delete_resource (struct proc_resource* resource);
-void proc_drop_resource (struct proc* proc, struct proc_resource* resource, bool lock);
-void proc_cleanup_resources (struct proc* proc);
 
 #endif // _KERNEL_PROC_RESOURCE_H

kernel/proc/src.mk

@@ -1,9 +1,11 @@
 c += proc/proc.c \
 		 proc/resource.c \
 		 proc/mutex.c \
-		 proc/mem.c
+		 proc/procgroup.c \
+		 proc/suspension_q.c
 
 o += proc/proc.o \
 		 proc/resource.o \
 		 proc/mutex.o \
-		 proc/mem.o
+		 proc/procgroup.o \
+		 proc/suspension_q.o

kernel/proc/suspension_q.c

@@ -0,0 +1,111 @@
+#include <libk/list.h>
+#include <libk/std.h>
+#include <mm/liballoc.h>
+#include <proc/proc.h>
+#include <proc/resource.h>
+#include <proc/suspension_q.h>
+#include <sync/spin_lock.h>
+#include <sys/smp.h>
+#include <sys/spin_lock.h>
+
+bool proc_sq_suspend (struct proc* proc, struct proc_suspension_q* sq, spin_lock_t* resource_lock,
+                      spin_lock_ctx_t* ctxrl) {
+  spin_lock_ctx_t ctxpr, ctxcpu, ctxsq;
+  struct cpu* cpu = proc->cpu;
+
+  struct proc_sq_entry* sq_entry = malloc (sizeof (*sq_entry));
+  if (!sq_entry) {
+    spin_unlock (resource_lock, ctxrl);
+    return PROC_NO_RESCHEDULE;
+  }
+
+  sq_entry->proc = proc;
+  sq_entry->sq = sq;
+
+  spin_lock (&cpu->lock, &ctxcpu);
+  spin_lock (&proc->lock, &ctxpr);
+  spin_lock (&sq->lock, &ctxsq);
+
+  spin_unlock (resource_lock, ctxrl);
+
+  atomic_store (&proc->state, PROC_SUSPENDED);
+
+  /* append to sq's list */
+  list_append (sq->proc_list, &sq_entry->sq_link);
+
+  /* append to proc's list */
+  list_append (proc->sq_entries, &sq_entry->proc_link);
+
+  list_remove (cpu->proc_run_q, &proc->cpu_run_q_link);
+  atomic_fetch_sub (&cpu->proc_run_q_count, 1);
+
+  if (cpu->proc_current == proc)
+    cpu->proc_current = NULL;
+
+  proc->cpu = NULL;
+
+  spin_unlock (&sq->lock, &ctxsq);
+  spin_unlock (&proc->lock, &ctxpr);
+  spin_unlock (&cpu->lock, &ctxcpu);
+
+  return PROC_NEED_RESCHEDULE;
+}
+
+bool proc_sq_resume (struct proc* proc, struct proc_sq_entry* sq_entry) {
+  spin_lock_ctx_t ctxsq, ctxpr, ctxcpu;
+  struct cpu* cpu = cpu_find_lightest ();
+  struct proc_suspension_q* sq = sq_entry->sq;
+
+  spin_lock (&cpu->lock, &ctxcpu);
+  spin_lock (&proc->lock, &ctxpr);
+  spin_lock (&sq->lock, &ctxsq);
+
+  /* remove from sq's list */
+  list_remove (sq->proc_list, &sq_entry->sq_link);
+
+  /* remove from proc's list */
+  list_remove (proc->sq_entries, &sq_entry->proc_link);
+
+  proc->cpu = cpu;
+
+  if (proc->sq_entries == NULL)
+    atomic_store (&proc->state, PROC_READY);
+
+  list_append (cpu->proc_run_q, &proc->cpu_run_q_link);
+  atomic_fetch_add (&cpu->proc_run_q_count, 1);
+
+  spin_unlock (&sq->lock, &ctxsq);
+  spin_unlock (&proc->lock, &ctxpr);
+  spin_unlock (&cpu->lock, &ctxcpu);
+
+  free (sq_entry);
+
+  return PROC_NEED_RESCHEDULE;
+}
+
+void proc_sqs_cleanup (struct proc* proc) {
+  spin_lock_ctx_t ctxsq, ctxpr;
+
+  spin_lock (&proc->lock, &ctxpr);
+
+  /* clean suspension queue entries */
+  struct list_node_link *sq_link, *sq_link_tmp;
+  list_foreach (proc->sq_entries, sq_link, sq_link_tmp) {
+    struct proc_sq_entry* sq_entry = list_entry (sq_link, struct proc_sq_entry, proc_link);
+    struct proc_suspension_q* sq = sq_entry->sq;
+
+    spin_lock (&sq->lock, &ctxsq);
+
+    /* remove from sq's list */
+    list_remove (sq->proc_list, &sq_entry->sq_link);
+
+    /* remove from proc's list */
+    list_remove (proc->sq_entries, &sq_entry->proc_link);
+
+    spin_unlock (&sq->lock, &ctxsq);
+
+    free (sq_entry);
+  }
+
+  spin_unlock (&proc->lock, &ctxpr);
+}

kernel/proc/suspension_q.h

@@ -1,12 +1,26 @@
 #ifndef _KERNEL_PROC_SUSPENTION_Q_H
 #define _KERNEL_PROC_SUSPENTION_Q_H
 
-#include <libk/rbtree.h>
+#include <libk/list.h>
 #include <sync/spin_lock.h>
 
+struct proc;
+
 struct proc_suspension_q {
-  struct rb_node_link* proc_tree;
+  struct list_node_link* proc_list;
   spin_lock_t lock;
 };
 
+struct proc_sq_entry {
+  struct list_node_link sq_link;
+  struct list_node_link proc_link;
+  struct proc* proc;
+  struct proc_suspension_q* sq;
+};
+
+void proc_sqs_cleanup (struct proc* proc);
+bool proc_sq_suspend (struct proc* proc, struct proc_suspension_q* sq, spin_lock_t* resource_lock,
+                      spin_lock_ctx_t* ctxrl);
+bool proc_sq_resume (struct proc* proc, struct proc_sq_entry* sq_entry);
+
 #endif // _KERNEL_PROC_SUSPENTION_Q_H

kernel/sync/rw_spin_lock.c

@@ -1,67 +0,0 @@
-#include <libk/assert.h>
-#include <libk/std.h>
-#include <sync/rw_spin_lock.h>
-#include <sys/debug.h>
-#include <sys/irq.h>
-#include <sys/spin_lock.h>
-
-#define WRITER_WAIT (1U << 31)
-#define READER_MASK (~WRITER_WAIT)
-
-void rw_spin_read_lock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx) {
-  uint32_t value;
-
-  irq_save (ctx);
-
-  for (;;) {
-    value = atomic_load_explicit (rw, memory_order_relaxed);
-
-    if ((value & WRITER_WAIT) == 0) {
-      if (atomic_compare_exchange_weak_explicit (rw, &value, value + 1, memory_order_acquire,
-                                                 memory_order_relaxed)) {
-        return;
-      }
-    }
-
-    spin_lock_relax ();
-  }
-}
-
-void rw_spin_read_unlock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx) {
-  uint32_t old = atomic_fetch_sub_explicit (rw, 1, memory_order_release);
-  assert ((old & READER_MASK) > 0);
-  irq_restore (ctx);
-}
-
-void rw_spin_write_lock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx) {
-  uint32_t value;
-
-  irq_save (ctx);
-
-  /* announce writer */
-  for (;;) {
-    value = atomic_load_explicit (rw, memory_order_relaxed);
-
-    if ((value & WRITER_WAIT) == 0) {
-      if (atomic_compare_exchange_weak_explicit (rw, &value, (value | WRITER_WAIT),
-                                                 memory_order_acquire, memory_order_relaxed))
-        break;
-    } else {
-      spin_lock_relax ();
-    }
-  }
-
-  /* wait for readers */
-  for (;;) {
-    value = atomic_load_explicit (rw, memory_order_acquire);
-    if ((value & READER_MASK) == 0)
-      return;
-
-    spin_lock_relax ();
-  }
-}
-
-void rw_spin_write_unlock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx) {
-  atomic_store_explicit (rw, 0, memory_order_release);
-  irq_restore (ctx);
-}

kernel/sync/rw_spin_lock.h

@@ -1,17 +0,0 @@
-#ifndef _KERNEL_SYNC_RW_SPIN_LOCK_H
-#define _KERNEL_SYNC_RW_SPIN_LOCK_H
-
-#include <libk/std.h>
-#include <sync/spin_lock.h>
-#include <sys/spin_lock.h>
-
-#define RW_SPIN_LOCK_INIT 0
-
-typedef _Atomic (uint32_t) rw_spin_lock_t;
-
-void rw_spin_read_lock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx);
-void rw_spin_read_unlock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx);
-void rw_spin_write_lock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx);
-void rw_spin_write_unlock (rw_spin_lock_t* rw, spin_lock_ctx_t* ctx);
-
-#endif // _KERNEL_SYNC_RW_SPIN_LOCK_H

kernel/sync/src.mk

@@ -1,5 +1,3 @@
-c += sync/spin_lock.c \
+c += sync/spin_lock.c
-		 sync/rw_spin_lock.c
 
-o += sync/spin_lock.o \
+o += sync/spin_lock.o
-		 sync/rw_spin_lock.o

kernel/sys/mm.h

@@ -2,6 +2,7 @@
 #define _KERNEL_SYS_MM_H
 
 #include <libk/std.h>
+#include <sync/spin_lock.h>
 
 #if defined(__x86_64__)
 #include <amd64/mm.h>
@@ -10,21 +11,18 @@
 #define MM_PG_PRESENT (1 << 0)
 #define MM_PG_RW      (1 << 1)
 #define MM_PG_USER    (1 << 2)
-#define MM_PD_RELOAD  (1 << 30)
-#define MM_PD_LOCK    (1 << 31)
 
 uintptr_t mm_alloc_user_pd_phys (void);
-void mm_reload (void);
+void mm_kernel_lock (spin_lock_ctx_t* ctx);
+void mm_kernel_unlock (spin_lock_ctx_t* ctx);
 void mm_map_page (struct pd* pd, uintptr_t paddr, uintptr_t vaddr, uint32_t flags);
 void mm_map_kernel_page (uintptr_t paddr, uintptr_t vaddr, uint32_t flags);
-void mm_unmap_page (struct pd* pd, uintptr_t vaddr, uint32_t flags);
+void mm_unmap_page (struct pd* pd, uintptr_t vaddr);
-void mm_unmap_kernel_page (uintptr_t vaddr, uint32_t flags);
+void mm_unmap_kernel_page (uintptr_t vaddr);
-void mm_lock_kernel (void);
+bool mm_validate (struct pd* pd, uintptr_t vaddr);
-void mm_unlock_kernel (void);
+bool mm_validate_buffer (struct pd* pd, uintptr_t vaddr, size_t size);
-bool mm_validate (struct pd* pd, uintptr_t vaddr, uint32_t flags);
+uintptr_t mm_p2v (struct pd* pd, uintptr_t paddr);
-bool mm_validate_buffer (struct pd* pd, uintptr_t vaddr, size_t size, uint32_t flags);
+uintptr_t mm_v2p (struct pd* pd, uintptr_t vaddr);
-uintptr_t mm_p2v (struct pd* pd, uintptr_t paddr, uint32_t flags);
-uintptr_t mm_v2p (struct pd* pd, uintptr_t vaddr, uint32_t flags);
 struct pd* mm_get_kernel_pd (void);
 void mm_init (void);
 

kernel/sys/proc.h

@@ -6,8 +6,9 @@
 struct proc;
 
 struct proc* proc_from_elf (uint8_t* elf_contents);
-struct proc* proc_clone (struct proc* proto, uintptr_t vstack_top, size_t stack_size,
+struct proc* proc_clone (struct proc* proto, uintptr_t vstack_top, uintptr_t entry,
-                         uintptr_t entry);
+                         uintptr_t argument_ptr);
 void proc_cleanup (struct proc* proc);
+void proc_init_tls (struct proc* proc);
 
 #endif // _KERNEL_SYS_PROC_H

kernel/sys/procgroup.h

@@ -0,0 +1,8 @@
+#ifndef _KERNEL_SYS_PROCGROUP_H
+#define _KERNEL_SYS_PROCGROUP_H
+
+#if defined(__x86_64__)
+#include <amd64/procgroup.h>
+#endif
+
+#endif // _KERNEL_SYS_PROCGROUP_H

kernel/sys/sched.h

@@ -4,6 +4,6 @@
 #include <libk/std.h>
 #include <proc/proc.h>
 
-void do_sched (struct proc* proc);
+void do_sched (struct proc* proc, spin_lock_t* cpu_lock, spin_lock_ctx_t* ctxcpu);
 
 #endif // _KERNEL_SYS_SCHED_H

kernel/syscall/syscall.c

@@ -5,9 +5,9 @@
 #include <m/status.h>
 #include <m/syscall_defs.h>
 #include <mm/pmm.h>
-#include <proc/mem.h>
 #include <proc/mutex.h>
 #include <proc/proc.h>
+#include <proc/procgroup.h>
 #include <proc/resource.h>
 #include <sync/spin_lock.h>
 #include <sys/debug.h>
@@ -16,23 +16,26 @@
 #include <syscall/syscall.h>
 
 #define DEFINE_SYSCALL(name)                                                                       \
-  int name (struct proc* UNUSED proc, void* UNUSED regs, uintptr_t UNUSED a1, uintptr_t UNUSED a2, \
+  uintptr_t name (struct proc* UNUSED proc, void* UNUSED regs, uintptr_t UNUSED a1,                \
-            uintptr_t UNUSED a3, uintptr_t UNUSED a4, uintptr_t UNUSED a5, uintptr_t UNUSED a6)
+                  uintptr_t UNUSED a2, uintptr_t UNUSED a3, uintptr_t UNUSED a4,                   \
+                  uintptr_t UNUSED a5, uintptr_t UNUSED a6)
+
+#define SYSRESULT(x) ((uintptr_t)(x))
 
 static void* sys_get_user_buffer (struct proc* proc, uintptr_t uvaddr, size_t size) {
   struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
-  spin_lock_ctx_t ctxprpd;
+  spin_lock_ctx_t ctxpg;
 
-  spin_lock (&proc->pd->lock, &ctxprpd);
+  spin_lock (&proc->procgroup->lock, &ctxpg);
 
-  if (!mm_validate_buffer (proc->pd, (uintptr_t)uvaddr, size, 0)) {
+  if (!mm_validate_buffer (&proc->procgroup->pd, (uintptr_t)uvaddr, size)) {
-    spin_unlock (&proc->pd->lock, &ctxprpd);
+    spin_unlock (&proc->procgroup->lock, &ctxpg);
     return NULL;
   }
 
-  uintptr_t out_paddr = mm_v2p (proc->pd, uvaddr, 0);
+  uintptr_t out_paddr = mm_v2p (&proc->procgroup->pd, uvaddr);
 
-  spin_unlock (&proc->pd->lock, &ctxprpd);
+  spin_unlock (&proc->procgroup->lock, &ctxpg);
 
   uintptr_t out_kvaddr = (uintptr_t)hhdm->offset + out_paddr;
 
@@ -42,50 +45,26 @@ static void* sys_get_user_buffer (struct proc* proc, uintptr_t uvaddr, size_t si
 /* int quit (void) */
 DEFINE_SYSCALL (sys_quit) {
   proc_kill (proc);
-  return ST_OK;
+  return SYSRESULT (ST_OK);
 }
 
 /* int test (void) */
 DEFINE_SYSCALL (sys_test) {
   char c = (char)a1;
   DEBUG ("test syscall from %d! %c\n", proc->pid, c);
-  return ST_OK;
+  return SYSRESULT (ST_OK);
 }
 
-/* int map (int mem_rid, int vis, uintptr_t vaddr, uint32_t flags) */
+/* int map (uintptr_t vaddr, size_t pages, uint32_t flags) */
 DEFINE_SYSCALL (sys_map) {
-  spin_lock_ctx_t ctxrs;
+  uintptr_t vaddr = a1;
-
+  size_t pages = (size_t)a2;
-  int mem_rid = (int)a1;
+  uint32_t flags = (uint32_t)a3;
-  int vis = (int)a2;
-  uintptr_t vaddr = a3;
-  uint32_t flags = (uint32_t)a4;
 
   if (vaddr % PAGE_SIZE != 0)
-    return -ST_UNALIGNED;
+    return SYSRESULT (-ST_UNALIGNED);
 
-  if (!(vis == RV_PUBLIC || vis == RV_PRIVATE))
+  return SYSRESULT (procgroup_map (proc->procgroup, vaddr, pages, flags, NULL));
-    return -ST_BAD_RESOURCE;
-
-  struct proc_resource* mem_resource = proc_find_resource (proc, mem_rid, vis);
-
-  if (mem_resource == NULL) {
-    return -ST_NOT_FOUND;
-  }
-
-  spin_lock (&mem_resource->lock, &ctxrs);
-
-  if (mem_resource->type != PR_MEM) {
-    spin_unlock (&mem_resource->lock, &ctxrs);
-    return -ST_BAD_RESOURCE;
-  }
-
-  uintptr_t paddr = mem_resource->u.mem.paddr;
-  size_t pages = mem_resource->u.mem.pages;
-
-  spin_unlock (&mem_resource->lock, &ctxrs);
-
-  return proc_map (proc, paddr, vaddr, pages, flags) ? ST_OK : -ST_OOM_ERROR;
 }
 
 /* int unmap (uintptr_t vaddr, size_t pages) */
@@ -94,93 +73,94 @@ DEFINE_SYSCALL (sys_unmap) {
   size_t pages = (size_t)a2;
 
   if (vaddr % PAGE_SIZE != 0)
-    return -ST_UNALIGNED;
+    return SYSRESULT (-ST_UNALIGNED);
 
-  return proc_unmap (proc, vaddr, pages) ? ST_OK : -ST_OOM_ERROR;
+  return SYSRESULT (procgroup_unmap (proc->procgroup, vaddr, pages));
 }
 
-/* int create_mem (int rid, int vis, size_t pages) */
+/* int clone (uintptr_t vstack_top, void* entry, void* argument_ptr) */
-DEFINE_SYSCALL (sys_create_mem) {
-  int rid = (int)a1;
-  int vis = (int)a2;
-  size_t pages = (size_t)a3;
-
-  if (!(vis == RV_PUBLIC || vis == RV_PRIVATE))
-    return -ST_BAD_RESOURCE;
-
-  if (pages == 0)
-    return ST_OK;
-
-  int rid1 = rid < 0 ? atomic_fetch_add (&proc->resources->sys_rids, 1) : rid;
-
-  struct proc_resource_mem_init mem_init = {.managed = false, .pages = pages};
-  struct proc_resource* mem_resource = proc_create_resource (proc, rid1, PR_MEM, vis, &mem_init);
-
-  if (mem_resource == NULL)
-    return -ST_OOM_ERROR;
-
-  return mem_resource->rid;
-}
-
-/* int unlink_mem (int rid, int vis, size_t pages) */
-DEFINE_SYSCALL (sys_unlink_mem) {
-  spin_lock_ctx_t ctxrs;
-
-  int rid = (int)a1;
-  int vis = (int)a2;
-  size_t pages = (size_t)a3;
-
-  if (!(vis == RV_PUBLIC || vis == RV_PRIVATE))
-    return -ST_BAD_RESOURCE;
-
-  struct proc_resource* mem_resource = proc_find_resource (proc, rid, vis);
-
-  if (mem_resource == NULL)
-    return -ST_NOT_FOUND;
-
-  spin_lock (&mem_resource->lock, &ctxrs);
-
-  if (mem_resource->type != PR_MEM) {
-    spin_unlock (&mem_resource->lock, &ctxrs);
-    return -ST_BAD_RESOURCE;
-  }
-
-  mem_resource->u.mem.alive_pages -= pages;
-  if (mem_resource->u.mem.alive_pages < 0) {
-    spin_unlock (&mem_resource->lock, &ctxrs);
-    proc_drop_resource (proc, mem_resource, true);
-  }
-
-  return ST_OK;
-}
-
-/* int clone (uintptr_t vstack_top, size_t stack_size, void* entry) */
 DEFINE_SYSCALL (sys_clone) {
   uintptr_t vstack_top = a1;
-  size_t stack_size = (size_t)a2;
+  uintptr_t entry = a2;
-  uintptr_t entry = a3;
+  uintptr_t argument_ptr = a3;
 
-  struct cpu* cpu = proc->cpu;
+  struct proc* new = proc_clone (proc, vstack_top, entry, argument_ptr);
-
-  struct proc* new = proc_clone (proc, vstack_top, stack_size, entry);
-
-  DEBUG ("new=%p\n", new);
 
   if (new == NULL) {
-    return -ST_OOM_ERROR;
+    return SYSRESULT (-ST_OOM_ERROR);
   }
 
   int pid = new->pid;
 
-  proc_register (new, cpu);
+  proc_register (new, NULL);
 
-  return pid;
+  return SYSRESULT (pid);
 }
 
-/* int proc_sched (void) */
+/* void* argument_ptr (void) */
+DEFINE_SYSCALL (sys_argument_ptr) { return proc->uvaddr_argument; }
+
+/* int sched (void) */
 DEFINE_SYSCALL (sys_sched) {
-  proc_sched (regs);
+  proc_sched ();
-  return ST_OK;
+  return SYSRESULT (ST_OK);
+}
+
+/* int mutex_create (int mutex_rid) */
+DEFINE_SYSCALL (sys_mutex_create) {
+  int mutex_rid = (int)a1;
+
+  struct proc_resource* mutex_resource = proc_create_resource_mutex (proc->procgroup, mutex_rid);
+
+  if (mutex_resource == NULL)
+    return SYSRESULT (-ST_OOM_ERROR);
+
+  return SYSRESULT (mutex_resource->rid);
+}
+
+/* int mutex_delete (int mutex_rid) */
+DEFINE_SYSCALL (sys_mutex_delete) {
+  int mutex_rid = (int)a1;
+
+  struct proc_resource* mutex_resource = proc_find_resource (proc->procgroup, mutex_rid);
+
+  if (mutex_resource == NULL)
+    return SYSRESULT (-ST_NOT_FOUND);
+
+  if (proc_delete_resource (mutex_resource) == PROC_NEED_RESCHEDULE)
+    proc_sched ();
+
+  return SYSRESULT (ST_OK);
+}
+
+/* int mutex_lock (int mutex_rid) */
+DEFINE_SYSCALL (sys_mutex_lock) {
+  int mutex_rid = (int)a1;
+
+  struct proc_resource* mutex_resource = proc_find_resource (proc->procgroup, mutex_rid);
+
+  if (mutex_resource == NULL)
+    return SYSRESULT (-ST_NOT_FOUND);
+
+  if (proc_mutex_lock (proc, &mutex_resource->u.mutex) == PROC_NEED_RESCHEDULE)
+    proc_sched ();
+
+  return SYSRESULT (ST_OK);
+}
+
+/* int mutex_unlock (int mutex_rid) */
+DEFINE_SYSCALL (sys_mutex_unlock) {
+  int mutex_rid = (int)a1;
+
+  struct proc_resource* mutex_resource = proc_find_resource (proc->procgroup, mutex_rid);
+
+  if (mutex_resource == NULL)
+    return SYSRESULT (-ST_NOT_FOUND);
+
+  if (proc_mutex_unlock (proc, &mutex_resource->u.mutex) == PROC_NEED_RESCHEDULE)
+    proc_sched ();
+
+  return SYSRESULT (ST_OK);
 }
 
 static syscall_handler_func_t handler_table[] = {
@@ -189,9 +169,12 @@ static syscall_handler_func_t handler_table[] = {
     [SYS_MAP] = &sys_map,
     [SYS_UNMAP] = &sys_unmap,
     [SYS_CLONE] = &sys_clone,
+    [SYS_ARGUMENT_PTR] = &sys_argument_ptr,
     [SYS_SCHED] = &sys_sched,
-    [SYS_CREATE_MEM] = &sys_create_mem,
+    [SYS_MUTEX_CREATE] = &sys_mutex_create,
-    [SYS_UNLINK_MEM] = &sys_unlink_mem,
+    [SYS_MUTEX_DELETE] = &sys_mutex_delete,
+    [SYS_MUTEX_LOCK] = &sys_mutex_lock,
+    [SYS_MUTEX_UNLOCK] = &sys_mutex_unlock,
 };
 
 syscall_handler_func_t syscall_find_handler (int syscall_num) {

kernel/syscall/syscall.h

@@ -4,8 +4,9 @@
 #include <libk/std.h>
 #include <proc/proc.h>
 
-typedef int (*syscall_handler_func_t) (struct proc* proc, void* regs, uintptr_t a1, uintptr_t a2,
+typedef uintptr_t (*syscall_handler_func_t) (struct proc* proc, void* regs, uintptr_t a1,
-                                       uintptr_t a3, uintptr_t a4, uintptr_t a5, uintptr_t a6);
+                                             uintptr_t a2, uintptr_t a3, uintptr_t a4, uintptr_t a5,
+                                             uintptr_t a6);
 
 syscall_handler_func_t syscall_find_handler (int syscall_num);
 

libmsl/alloc/liballoc.c

@@ -4,39 +4,19 @@
 #include <alloc/liballoc.h>
 #include <m/system.h>
 
-static uintptr_t liballoc_map_base = PROC_MAP_BASE;
+#define LIBALLOC_MUTEX 500
-static int mem_rid_base = 1000000;
 
-void liballoc_init (void) {}
+void liballoc_init (void) { mutex_create (LIBALLOC_MUTEX); }
 
-int liballoc_lock (void) { return 0; }
+void liballoc_deinit (void) { mutex_delete (LIBALLOC_MUTEX); }
 
-int liballoc_unlock (void) { return 0; }
+int liballoc_lock (void) { return mutex_lock (LIBALLOC_MUTEX); }
 
-void* liballoc_alloc (int pages, int* mem_rid) {
+int liballoc_unlock (void) { return mutex_unlock (LIBALLOC_MUTEX); }
-  uintptr_t current_base = liballoc_map_base;
 
-  *mem_rid = create_mem (mem_rid_base++, RV_PRIVATE, pages);
+void* liballoc_alloc (int pages) { return map (0, pages, MAP_FLAGS | MAP_RW); }
-  if (*mem_rid < 0) {
-    return NULL;
-  }
 
-  if (map (*mem_rid, RV_PRIVATE, current_base, MAP_FLAGS | MAP_RW) < 0) {
+int liballoc_free (void* ptr, int pages) { return unmap ((uintptr_t)ptr, pages); }
-    unlink_mem (*mem_rid, RV_PRIVATE, pages);
-    return NULL;
-  }
-
-  uintptr_t old_base = current_base;
-  current_base += pages * PAGE_SIZE;
-
-  return (void*)old_base;
-}
-
-int liballoc_free (void* ptr, int pages, int mem_rid) {
-  unmap ((uintptr_t)ptr, pages);
-  unlink_mem (mem_rid, RV_PRIVATE, pages);
-  return 0;
-}
 
 /**  Durand's Ridiculously Amazing Super Duper Memory functions.  */
 
@@ -203,7 +183,6 @@ static struct boundary_tag* allocate_new_tag (unsigned int size) {
   unsigned int pages;
   unsigned int usage;
   struct boundary_tag* tag;
-  int mem_rid;
 
   // This is how much space is required.
   usage = size + sizeof (struct boundary_tag);
@@ -217,7 +196,7 @@ static struct boundary_tag* allocate_new_tag (unsigned int size) {
   if (pages < (unsigned int)l_pageCount)
     pages = l_pageCount;
 
-  tag = (struct boundary_tag*)liballoc_alloc (pages, &mem_rid);
+  tag = (struct boundary_tag*)liballoc_alloc (pages);
 
   if (tag == NULL)
     return NULL; // uh oh, we ran out of memory.
@@ -226,7 +205,6 @@ static struct boundary_tag* allocate_new_tag (unsigned int size) {
   tag->size = size;
   tag->real_size = pages * l_pageSize;
   tag->index = -1;
-  tag->mem_rid = mem_rid;
 
   tag->next = NULL;
   tag->prev = NULL;
@@ -349,7 +327,7 @@ void free (void* ptr) {
       if (pages < (unsigned int)l_pageCount)
         pages = l_pageCount;
 
-      liballoc_free (tag, pages, tag->mem_rid);
+      liballoc_free (tag, pages);
 
       liballoc_unlock ();
       return;

libmsl/alloc/liballoc.h

@@ -41,8 +41,6 @@ struct boundary_tag {
 
   struct boundary_tag* next; //< Linked list info.
   struct boundary_tag* prev; //< Linked list info.
-
-  int mem_rid;
 };
 
 /** This function is supposed to lock the memory data structures. It
@@ -69,7 +67,7 @@ extern int liballoc_unlock (void);
  * \return NULL if the pages were not allocated.
  * \return A pointer to the allocated memory.
  */
-extern void* liballoc_alloc (int pages, int* mem_rid);
+extern void* liballoc_alloc (int pages);
 
 /** This frees previously allocated memory. The void* parameter passed
  * to the function is the exact same value returned from a previous
@@ -79,7 +77,7 @@ extern void* liballoc_alloc (int pages, int* mem_rid);
  *
  * \return 0 if the memory was successfully freed.
  */
-extern int liballoc_free (void* ptr, int pages, int mem_rid);
+extern int liballoc_free (void* ptr, int pages);
 
 void* malloc (size_t);         //< The standard function.
 void* realloc (void*, size_t); //< The standard function.
@@ -87,6 +85,7 @@ void* calloc (size_t, size_t); //< The standard function.
 void free (void*);             //< The standard function.
 
 void liballoc_init (void);
+void liballoc_deinit (void);
 
 #ifdef __cplusplus
 }

libmsl/amd64/syscall.c

@@ -2,7 +2,7 @@
 #include <stddef.h>
 #include <stdint.h>
 
-int amd64_syscall (int syscall_num, uintptr_t a1, uintptr_t a2, uintptr_t a3, uintptr_t a4,
+uintptr_t amd64_syscall (int syscall_num, uintptr_t a1, uintptr_t a2, uintptr_t a3, uintptr_t a4,
                          uintptr_t a5, uintptr_t a6) {
   uint64_t result;
   __asm__ volatile ("movq %[a4], %%r10\n"
@@ -13,5 +13,5 @@ int amd64_syscall (int syscall_num, uintptr_t a1, uintptr_t a2, uintptr_t a3, ui
                     : "a"(syscall_num), "D"(a1), "S"(a2),
                       "d"(a3), [a4] "r"(a4), [a5] "r"(a5), [a6] "r"(a6)
                     : "r10", "r8", "r9", "r11", "rcx", "cc", "memory");
-  return (int)result;
+  return result;
 }

libmsl/amd64/syscall.h

@@ -3,7 +3,7 @@
 
 #include <stdint.h>
 
-int amd64_syscall (int syscall_num, uintptr_t a1, uintptr_t a2, uintptr_t a3, uintptr_t a4,
+uintptr_t amd64_syscall (int syscall_num, uintptr_t a1, uintptr_t a2, uintptr_t a3, uintptr_t a4,
                          uintptr_t a5, uintptr_t a6);
 
 #endif // _LIBMSL_AMD64_SYSCALL_H

libmsl/init/__premain.c

@@ -17,8 +17,7 @@ static void clear_bss (void) {
 void __premain (void) {
   clear_bss ();
   liballoc_init ();
-
   app_main ();
-
+  liballoc_deinit ();
   quit ();
 }

libmsl/m/system.c

@@ -9,24 +9,28 @@
 
 #define do_syscall(...) do_syscall1 (__VA_ARGS__, 0, 0, 0, 0, 0, 0)
 
-int quit (void) { return do_syscall (SYS_QUIT); }
+int quit (void) { return do_syscall (SYS_QUIT, 0); }
 
 int test (char c) { return do_syscall (SYS_TEST, c); }
 
-int map (int mem_rid, int vis, uintptr_t vaddr, uint32_t flags) {
+int sched (void) { return do_syscall (SYS_SCHED, 0); }
-  return do_syscall (SYS_MAP, mem_rid, vis, vaddr, flags);
+
+void* map (uintptr_t vaddr, size_t pages, uint32_t flags) {
+  return (void*)do_syscall (SYS_MAP, vaddr, pages, flags);
 }
 
 int unmap (uintptr_t vaddr, size_t pages) { return do_syscall (SYS_UNMAP, vaddr, pages); }
 
-int create_mem (int mem_rid, int vis, size_t pages) {
+int clone (uintptr_t vstack_top, void (*entry) (void), void* argument_ptr) {
-  return do_syscall (SYS_CREATE_MEM, mem_rid, vis, pages);
+  return do_syscall (SYS_CLONE, vstack_top, entry, argument_ptr);
 }
 
-int unlink_mem (int mem_rid, int vis, size_t pages) {
+int mutex_create (int mutex_rid) { return do_syscall (SYS_MUTEX_CREATE, mutex_rid); }
-  return do_syscall (SYS_UNLINK_MEM, mem_rid, vis, pages);
-}
 
-int clone (uintptr_t vstack_top, size_t stack_size, void (*entry) (void)) {
+int mutex_delete (int mutex_rid) { return do_syscall (SYS_MUTEX_DELETE, mutex_rid); }
-  return do_syscall (SYS_CLONE, vstack_top, stack_size, entry);
+
-}
+int mutex_lock (int mutex_rid) { return do_syscall (SYS_MUTEX_LOCK, mutex_rid); }
+
+int mutex_unlock (int mutex_rid) { return do_syscall (SYS_MUTEX_UNLOCK, mutex_rid); }
+
+void* argument_ptr (void) { return (void*)do_syscall (SYS_ARGUMENT_PTR, 0); }

libmsl/m/system.h

@@ -5,7 +5,6 @@
 #include <stdint.h>
 
 #if defined(__x86_64__)
-#define PROC_MAP_BASE 0x0000700000000000
 #define PAGE_SIZE 4096
 #endif
 
@@ -14,15 +13,16 @@
 #define MAP_USER    (1 << 2)
 #define MAP_FLAGS   (MAP_PRESENT | MAP_USER)
 
-#define RV_PRIVATE 0
-#define RV_PUBLIC  1
-
 int quit (void);
 int test (char c);
-int map (int mem_rid, int vis, uintptr_t vaddr, uint32_t flags);
+int sched (void);
+void* map (uintptr_t vaddr, size_t pages, uint32_t flags);
 int unmap (uintptr_t vaddr, size_t pages);
-int create_mem (int mem_rid, int vis, size_t pages);
+int clone (uintptr_t vstack_top, void (*entry) (void), void* argument_ptr);
-int unlink_mem (int mem_rid, int vis, size_t pages);
+int mutex_create (int mutex_rid);
-int clone (uintptr_t vstack_top, size_t stack_size, void (*entry) (void));
+int mutex_delete (int mutex_rid);
+int mutex_lock (int mutex_rid);
+int mutex_unlock (int mutex_rid);
+void* argument_ptr (void);
 
 #endif // _LIBMSL_M_SYSTEM_H

libmsl/proc/.gitignore

@@ -0,0 +1 @@
+*.o

libmsl/proc/local.h

@@ -0,0 +1,6 @@
+#ifndef _LIBMSL_PROC_TLS_H
+#define _LIBMSL_PROC_TLS_H
+
+#define LOCAL __thread
+
+#endif // _LIBMSL_PROC_TLS_H

libmsl/proc/proc.c

@@ -0,0 +1,19 @@
+#include <alloc/liballoc.h>
+#include <m/status.h>
+#include <m/system.h>
+#include <proc/proc.h>
+#include <stddef.h>
+#include <stdint.h>
+
+int process_spawn (process_func_t func, void* argument_ptr) {
+  void* stack = malloc (PROC_STACK_SIZE);
+  if (stack == NULL)
+    return -ST_OOM_ERROR;
+
+  uintptr_t top = (uintptr_t)stack + PROC_STACK_SIZE;
+  return clone (top, func, argument_ptr);
+}
+
+int process_quit (void) { return quit (); }
+
+void* process_argument (void) { return argument_ptr (); }

libmsl/proc/proc.h

@@ -0,0 +1,14 @@
+#ifndef _LIBMSL_PROC_PROC_H
+#define _LIBMSL_PROC_PROC_H
+
+#include <m/system.h>
+
+#define PROC_STACK_SIZE 256 * PAGE_SIZE
+
+typedef void (*process_func_t) (void);
+
+int process_spawn (process_func_t func, void* argument_ptr);
+int process_quit (void);
+void* process_argument (void);
+
+#endif // _LIBMSL_PROC_PROC_H

libmsl/proc/src.mk

@@ -0,0 +1,3 @@
+c += proc/proc.c
+
+o += proc/proc.o

libmsl/src.mk

@@ -3,3 +3,4 @@ include init/src.mk
 include m/src.mk
 include string/src.mk
 include alloc/src.mk
+include proc/src.mk

make/apps.mk

@@ -1,4 +1,4 @@
-apps := init
+apps := init spin
 
 all_apps:
 	@for d in $(apps); do make -C $$d platform=$(platform) all; done

make/libmsl.mk

@@ -7,4 +7,4 @@ clean_libmsl:
 format_libmsl:
 	make -C libmsl platform=$(platform) format
 
-.PHONY: all_libmsl clean_libmsl
+.PHONY: all_libmsl clean_libmsl format_libmsl

spin/.gitignore

@@ -0,0 +1,2 @@
+*.o
+*.exe

spin/Makefile

@@ -0,0 +1 @@
+include ../make/user.mk

spin/app.mk

@@ -0,0 +1 @@
+app := spin.exe

spin/spin.c

@@ -0,0 +1,4 @@
+void app_main (void) {
+  for (;;)
+    ;
+}

spin/src.mk

@@ -0,0 +1,3 @@
+c += spin.c
+
+o += spin.o