lesson 3, boot sector with memory addressing

cfenollosa · cfenollosa · commit 8211b950b10e · 2014-10-05T11:28:07.000+02:00
diff --git a/03-bootsector-memory/README.md b/03-bootsector-memory/README.md
@@ -1,13 +1,13 @@
 *Concepts you may want to Google beforehand: memory offsets, pointers*
 
-The only goal of this lesson is to learn where the boot sector is stored
-
 Please open page 14 [of this document](
 http://www.cs.bham.ac.uk/~exr/lectures/opsys/10_11/lectures/os-dev.pdf)<sup>1</sup>
 and look at the figure with the memory layout.
 
-I could just go ahead and tell you that it starts at `0x7C00`, but it's
-better with an example.
+The only goal of this lesson is to learn where the boot sector is stored
+
+I could just go ahead and tell you that the BIOS places it at `0x7C00`, but an 
+example with wrong solutions will make things clearer.
 
 We want to print an X on screen. We will try 4 different strategies
 and see which ones work and why.
@@ -23,7 +23,7 @@ Then we will try to access `the_secret` in many different ways:
 1. `mov al, the_secret`
 2. `mov al, [the_secret]`
 3. `mov al, the_secret + 0x7C00`
-4. `mov al, 2d + 0x7C00`, where `2d` is the actual position of the X in the binary
+4. `mov al, 2d + 0x7C00`, where `2d` is the actual position of the 'X' byte in the binary
 
 Take a look at the code and read the comments.
 
@@ -33,6 +33,21 @@ the bytes following 1 and 2 are just random garbage.
 If you add or remove instructions, remember to compute the new offset of the X
 by counting the bytes, and replace `0x2d` with the new one.
 
+Please don't continue onto the next file unless you have 100% understood
+the boot sector offset and memory addressing.
+
+Now, since offsetting `0x7c00` everywhere is very inconvenient, assemblers let
+us define a "global offset" for every memory location, with the `org` command:
+
+```nasm
+[org 0x7c00]
+```
+
+Go ahead and open `boot_sect_memory_org.asm` and you will see the canonical
+way to print data with the boot sector, which is now attempt 2. Compile the code
+and run it, and you will see how the `org` command affects each previous solution.
+
+Read the comments for a full explanation of the changes with and without `org`
 
 -----
 
diff --git a/03-bootsector-memory/boot_sect_memory.asm b/03-bootsector-memory/boot_sect_memory.asm
@@ -10,7 +10,7 @@ int 0x10
 
 ; attempt 2
 ; It tries to print the memory address of 'the_secret' which is the correct approach.
-; However, BIOS starts loading at address 0x7c00
+; However, BIOS places our bootsector binary at address 0x7c00
 ; so we need to add that padding beforehand. We'll do that in attempt 3
 mov al, "2"
 int 0x10
@@ -19,7 +19,9 @@ int 0x10
 
 ; attempt 3
 ; Add the BIOS starting offset 0x7c00 to the memory address of the X
-; and then dereference the contents of that pointer
+; and then dereference the contents of that pointer.
+; We need the help of a different register 'bx' because 'mov al, [ax]' is illegal.
+; A register can't be used as source and destination for the same command.
 mov al, "3"
 int 0x10
 mov bx, the_secret
@@ -29,19 +31,21 @@ int 0x10
 
 ; attempt 4
 ; We try a shortcut since we know that the X is stored at byte 0x2d in our binary
+; That's smart but ineffective, we don't want to be recounting label offsets
+; every time we change the code
 mov al, "4"
 int 0x10
 mov al, [0x7c2d]
 int 0x10
 
 
-jmp $
-
+jmp $ ; infinite loop
 
 the_secret:
-    ; ASCII code 0x58 is stored just before the zero-padding
-    ; on this code that is at byte 0x2d (check it out using xdd)
+    ; ASCII code 0x58 ('X') is stored just before the zero-padding.
+    ; On this code that is at byte 0x2d (check it out using 'xdd file.bin')
     db "X"
 
+; zero padding and magic bios number
 times 510-($-$$) db 0
 dw 0xaa55
diff --git a/03-bootsector-memory/boot_sect_memory_org.asm b/03-bootsector-memory/boot_sect_memory_org.asm
@@ -0,0 +1,47 @@
+[org 0x7c00]
+mov ah, 0x0e
+
+; attempt 1
+; Will fail again regardless of 'org' because we are still addressing the pointer
+; and not the data it points to
+mov al, "1"
+int 0x10
+mov al, the_secret
+int 0x10
+
+; attempt 2
+; Having solved the memory offset problem with 'org', this is now the correct answer
+mov al, "2"
+int 0x10
+mov al, [the_secret]
+int 0x10
+
+; attempt 3
+; As you expected, we are adding 0x7c00 twice, so this is not going to work
+mov al, "3"
+int 0x10
+mov bx, the_secret
+add bx, 0x7c00
+mov al, [bx]
+int 0x10
+
+; attempt 4
+; This still works because there are no memory references to pointers, so
+; the 'org' mode never applies. Directly addressing memory by counting bytes
+; is always going to work, but it's inconvenient
+mov al, "4"
+int 0x10
+mov al, [0x7c2d]
+int 0x10
+
+
+jmp $ ; infinite loop
+
+the_secret:
+    ; ASCII code 0x58 ('X') is stored just before the zero-padding.
+    ; On this code that is at byte 0x2d (check it out using 'xdd file.bin')
+    db "X"
+
+; zero padding and magic bios number
+times 510-($-$$) db 0
+dw 0xaa55
