Better boot.S and link script

[dinosaure]

/cc @TheLortex may be a nicer boot.S and link script.
However, a bug persists. I think this layout solve the issue about bss nicely.
The remaining bug is about the Sys.argv.(0) and its usage. Indeed, a main.ml such as:

let () = Format.printf "Hello %s!" Sys.argv.(0)

But this code does not show anything:

let () = Format.printf "Hello %S!" Sys.argv.(0)

[dinosaure]

To go further, this code only shows Hello :

let () =
  Format.printf "Hello %!" ;
  Format.printf "%S%!" Sys.argv.(0) ;
  Format.printf "!\n%!"

The problem is definitely about the usage of Sys.argv.(0).

[dinosaure]

It seems that string_to_caml_string does something wrong.

[dinosaure]

I bisected where the Raspberry Pi 4 fails and it seems that:

String.unsafe_blit str 0 res 1 len ;

fails (only on Raspberry Pi 4). I'm not sure why but I will try to continue to introspect what is going on - note that I tried this code:

let string_to_caml_string str =
  let str = String.escaped str in
  let l = String.length str in
  let res = Bytes.make (l + 2) '\"' in
  Bytes.blit_string str 0 res 1 l;
  Bytes.unsafe_to_string res

let fn0 str =
  let _str = String.escaped str in
  Format.printf "[+][+] 000\n%!" ;
;;

let fn1 str =
  let str = String.escaped str in
  Format.printf "[+][+] 000\n%!" ;
  let len = String.length str in
  Format.printf "[+][+] 001\n%!" ;
  let _res = Bytes.make (len + 2) '\"' in
  Format.printf "[+][+] 002\n%!" ;
;;

let fn2 str =
  let str = String.escaped str in
  Format.printf "[+][+] 000\n%!" ;
  let len = String.length str in
  Format.printf "[+][+] 001\n%!" ;
  let res = Bytes.make (len + 2) '\"' in
  Format.printf "[+][+] 002\n%!" ;
  Bytes.blit_string str 0 res 1 len ;
  Format.printf "[+][+] 003\n%!" ;
;;

let () =
  Format.printf "[+] 001\n%!" ;
  let _res = Bytes.create 0x100 in
  Format.printf "[+] 002\n%!" ;
  let _str = String.escaped "\000\001\000" in
  Format.printf "[+] 003\n%!" ;
  let _str = String.length _str in
  Format.printf "[+] 004\n%!" ;
  let _str = "Hello World!" in
  Format.printf "[+] 005\n%!" ;
  let _len = String.length _str in
  Format.printf "[+] 006\n%!" ;
  Bytes.blit_string _str 0 (Bytes.create 0x100) 0 _len ;
  Format.printf "[+] 007\n%!" ;
  fn0 "gi(l)braltar" ;
  Format.printf "[+] 008\n%!" ;
  let _res = Bytes.make (_len + 2) '\"' in
  Format.printf "[+] 009\n%!" ;
  Bytes.blit Bytes.empty 0 _res 0 0 ;
  Format.printf "[+] 010\n%!" ;
  fn1 "gi(l)braltar" ;
  Format.printf "[+] 011\n%!" ;
  fn2 "gi(l)braltar" ;
  Format.printf "[+] 012\n%!" ;
  Format.printf ">>> %s.\n%!" (string_to_caml_string "gi(l)braltar") ;
;;

And the output is:

 _____ _ _ _           _ _           
|   __|_| | |_ ___ ___| | |_ ___ ___ 
|  |  | | | . |  _| .'| |  _| .'|  _|
|_____|_|_|___|_| |__,|_|_| |__,|_|  
RPi4: Memory map: 768 MB addressable:
RPi4:   reserved @ (0x0 - 0x7ffff)
RPi4:       text @ (0x80000 - 0xe91ef)
RPi4:     rodata @ (0xe91f0 - 0xf091f)
RPi4:       data @ (0xf0920 - 0x13970f)
RPi4:       heap >= 0x13a000 < stack < 0x30000000
[+] 001
[+] 002
[+] 003
[+] 004
[+] 005
[+] 006
[+] 007
[+][+] 000
[+] 008
[+] 009
[+] 010
[+][+] 000
[+][+] 001
[+][+] 002
[+] 011
[+][+] 000
[+][+] 001
[+][+] 002

[dinosaure]

The bug is about the implementation of memcpy, see mirage/ocaml-solo5#100 for more details. I will propose on this PR a simpler implementation of memcpy/memmove (so we will start to diverge from ocaml-freestanding/nolibc).

[palainp]

According to the man page, the memcpy function should return a pointer to the dest address (return v_dst;). The returned value does not appear to be used here but at least mirage/mirage-crypto@f351442 uses it to ensure aligned memory usage and it may fail there. This is the same for memmove below.

[dinosaure]

Ok, I will propose a PR to fix that, thanks for this looking.

[palainp]

And btw, there may be some patch that has corrected the pb after this merge because I am currently able to run the example code completely with the original memcpy and memmove functions.

 _____ _ _ _           _ _                                                                                  
|   __|_| | |_ ___ ___| | |_ ___ ___                                                                        
|  |  | | | . |  _| .'| |  _| .'|  _|                                                                       
|_____|_|_|___|_| |__,|_|_| |__,|_|                                                                         
EL:1                                                                                                        
Interrupts: up                                                                                              
MMU: ON                                                                                                     
RPi4: Memory map: 752 MB addressable:                                                                       
RPi4:   reserved @ (0x0 - 0xffffffffffffffff)                                                                              
RPi4:       text @ (0x0 - 0x6518f)                                                                          
RPi4:     rodata @ (0x65190 - 0x6b6bf)                                                                      
RPi4:       data @ (0x6b6c0 - 0x1030c3f)                                                                    
RPi4:       heap >= 0x1031000 < stack < 0x2f000000                                                          
[+] 001                                                                                                     
[+] 002                                                                                                     
[+] 003                                                                                                     
[+] 004                                                                                                     
[+] 005                                                                                                     
[+] 006                                                                                                     
[+] 007                                                                                                     
[+][+] 000                                                                                                  
[+] 008                                                                                                     
[+] 009                                                                                                     
[+] 010                                                                                                     
[+][+] 000                                                                                                  
[+][+] 001                                                                                                  
[+][+] 002                                                                                                  
[+] 011                                                                                                     
[+][+] 000                                                                                                  
[+][+] 001                                                                                                  
[+][+] 002                                                                                                  
[+][+] 003                                                                                                  
[+] 012                                                                                                     
>>> "gi(l)braltar".                                                                                         

Maybe it is better to keep the same code here as the nolibc in ocaml-freestanding? The memcpy seems to really take care of 32b alignement and permit to copy 4B at a time which can be faster than 1B copy for longs arrays.

[dinosaure]

Hmhmm can you open a new issue for that to keep our experiments 👍