diff --git a/fox32os.def b/fox32os.def
index 1da9a31..2f5515e 100644
--- a/fox32os.def
+++ b/fox32os.def
@@ -13,3 +13,4 @@ end_current_task: jmp [0x00000838]
 
 ; memory jump table
 allocate_memory:  jmp [0x00000840]
+free_memory:      jmp [0x00000844]
diff --git a/kernel/allocator.asm b/kernel/allocator.asm
index 17a9282..6343438 100644
--- a/kernel/allocator.asm
+++ b/kernel/allocator.asm
@@ -1,35 +1,191 @@
 ; memory allocator routines
-; this is a very basic memory allocator, it doesn't even allow freeing
-; TODO: make this better
 
-const MEMORY_TOP: 0x02000000
+; block header fields:
+; data.32 size - size of block, NOT INCLUDING THE HEADER
+; data.32 prev - pointer to previous free block, or zero
+; data.32 next - pointer to next free block, or zero
+
+const HEADER_SIZE: 12
+const MEMORY_TOP:  0x02000000
+
+initialize_allocator:
+    push r0
+    push r1
+
+    mov [free_list_head], kernel_bottom
+
+    ; set the free block size to MEMORY_TOP - [free_list_head]
+    mov r0, [free_list_head]
+    mov r1, MEMORY_TOP
+    sub r1, r0
+    mov [r0], r1
+
+    ; mark this as the only free block
+    add r0, 4
+    mov [r0], 0
+    add r0, 4
+    mov [r0], 0
+
+    pop r1
+    pop r0
+    ret
 
 ; allocate a block of memory
 ; inputs:
 ; r0: size in bytes
 ; outputs:
-; r0: pointer to allocated block (or zero if no blocks free)
+; r0: pointer to allocated block (or zero if no blocks free of requested size)
 allocate_memory:
+    push r1
+    push r10
+    push r11
+
+    mov [block], [free_list_head]
+
+    ; r10: real_size = requested size + header size
+    mov r10, r0
+    add r10, HEADER_SIZE
+allocate_memory_while_block:
+    mov r0, [block]
+    call block_get_size
+    cmp r0, r10
+    ifgteq jmp allocate_memory_good_block
+
+    ; block = block->next
+    mov r0, [block]
+    call block_get_next
+    mov [block], r0
+
+    cmp [block], 0
+    ifnz jmp allocate_memory_while_block
+    ; if we reach this point, no good blocks were found
+    mov r0, 0
+
+    pop r11
+    pop r10
+    pop r1
+    ret
+allocate_memory_good_block:
+    mov r11, r10
+    add r11, HEADER_SIZE
+    add r11, 16
+    mov r0, [block]
+    call block_get_size
+    cmp r0, r11
+    ifgt jmp allocate_memory_good_block_carve
+
+    ; next = block->next
+    ; prev = block->prev
+    mov r0, [block]
+    call block_get_next
+    mov [next], r0
+    mov r0, [block]
+    call block_get_prev
+    mov [prev], r0
+
+    cmp [next], 0
+    ifnz call allocate_memory_good_block_set_nextprev
+
+    cmp [prev], 0
+    ifnz jmp allocate_memory_good_block_set_prevnext
+    mov [free_list_head], [next]
+allocate_memory_good_block_ret:
+    mov r0, [block]
+    add r0, HEADER_SIZE
+
+    pop r11
+    pop r10
+    pop r1
+    ret
+allocate_memory_good_block_carve:
+    ; block->size -= real_size
+    mov r0, [block]
+    call block_get_size
+    sub r0, r10
+    mov r1, r0
+    mov r0, [block]
+    call block_set_size
+
+    ; block += block->size
+    mov r0, [block]
+    call block_get_size
+    add [block], r0
+
+    ; block->size = real_size
+    mov r0, [block]
+    mov r1, r10
+    call block_set_size
+
+    jmp allocate_memory_good_block_ret
+allocate_memory_good_block_set_nextprev:
+    ; next->prev = prev
+    mov r0, [next]
+    mov r1, [prev]
+    call block_set_prev
+    ret
+allocate_memory_good_block_set_prevnext:
+    ; prev->next = next
+    mov r0, [prev]
+    mov r1, [next]
+    call block_set_next
+    jmp allocate_memory_good_block_ret
+block: data.32 0
+next:  data.32 0
+prev:  data.32 0
+block_get_size:
+    mov r0, [r0]
+    ret
+block_set_size:
+    mov [r0], r1
+    ret
+block_get_prev:
+    add r0, 4
+    mov r0, [r0]
+    ret
+block_set_prev:
+    add r0, 4
+    mov [r0], r1
+    ret
+block_get_next:
+    add r0, 8
+    mov r0, [r0]
+    ret
+block_set_next:
+    add r0, 8
+    mov [r0], r1
+    ret
+
+; free a block of memory
+; inputs:
+; r0: pointer to allocated block
+; outputs:
+; none
+free_memory:
+    push r0
     push r1
     push r2
 
-    mov r1, MEMORY_TOP
-    mov r2, [free_base]
-    sub r1, r2
-    cmp r1, r0
-    iflteq jmp allocate_memory_full
-    mov r1, [free_base]
-    add [free_base], r0
-    mov r0, r1
-
-    pop r2
-    pop r1
-    ret
-allocate_memory_full:
-    mov r0, 0
+    ; point to the header
+    sub r0, HEADER_SIZE
+    mov r2, r0
+
+    ; add it to the free list
+    mov r1, 0
+    call block_set_prev
+    mov r0, r2
+    mov r1, [free_list_head]
+    call block_set_next
+
+    cmp [free_list_head], 0
+    ifnz mov r1, r0
+    ifnz mov r0, [free_list_head]
+    ifnz call block_set_prev
+
+    mov [free_list_head], r2
 
     pop r2
     pop r1
+    pop r0
     ret
 
-free_base: data.32 kernel_bottom
+free_list_head: data.32 kernel_bottom
diff --git a/kernel/main.asm b/kernel/main.asm
index cbb2053..97ca2bf 100644
--- a/kernel/main.asm
+++ b/kernel/main.asm
@@ -31,6 +31,7 @@ jump_table:
     ; memory jump table
     org.pad 0x00000840
     data.32 allocate_memory
+    data.32 free_memory
 
     ; initialization code
 entry:
@@ -38,7 +39,10 @@ entry:
     mov r0, BACKGROUND_COLOR
     call fill_background
 
-draw_startup_text:
+    ; initialize the memory allocator
+    call initialize_allocator
+
+    ; draw the startup text
     mov r0, startup_str
     mov r1, 16
     mov r2, 464