Packaging: support >4K code length without LZMA

scripts/static-pie-gen.py

@@ -69,6 +69,7 @@
 # Since we append a little-endian 8-byte nonnegative integer, we can practically ensure that the last byte is zero.
 code_raw = memory_bin[:-8]
 code_raw += (len(code_raw) + 8 - loader_fdict['entrypoint_offset']).to_bytes(8, byteorder='little')
+code_raw_len = len(code_raw)
 code_raw_b91 = base91.encode(code_raw, use_rle=True).decode('ascii')
 code_raw_b91_len = len(code_raw_b91)
 code_raw_b91 = '"' + code_raw_b91 + '"'
@@ -121,7 +122,7 @@
 
 # template
 template_candidates = [template_path]
-if lang_name in ["C", "Rust"] and "x86_64" in target_name and "short" in template_path and len(code_raw) <= 4096 - 256:
+if lang_name in ["C", "Rust"] and "x86_64" in target_name and "short" in template_path:
     template_candidates.append(template_path.replace("short", "shorter"))
 
 # exports
@@ -149,6 +150,7 @@
         "$$$$binary_raw_base91$$$$": code_raw_b91,
         "$$$$binary_raw_base91_len$$$$": str(code_raw_b91_len),
         "$$$$binary_base91_chunked$$$$": r,
+        "$$$$code_raw_len$$$$": str(code_raw_len),
         "$$$$min_len_4096$$$$": str(min(len(code_b91)+1, 4096)),
         "$$$$entrypoint_offset$$$$": str(loader_fdict['entrypoint_offset']),
         "$$$$exports_cpp$$$$": exports_cpp

scripts/templates/static-pie-prestub-amd64-print.py

@@ -7,7 +7,7 @@
 prestub = bytearray(prestub)
 if len(prestub) > 0 and prestub[-1] == 0:
     prestub = prestub[:-1]
-    asciz = True
+    asciz = '--no-asciz' not in sys.argv
 else:
     asciz = False
 

scripts/templates/static-pie-prestub-amd64-shorter-c.asm

@@ -4,7 +4,7 @@
 ; (prestub: the code that runs before the stub and sets the stage)
 ;
 ; build: nasm -f bin -O9 static-pie-prestub-amd64-shorter-c.asm -o static-pie-prestub-amd64-shorter-c.bin
-; note: after building with the above command, run static-pie-prestub-amd64-print.py static-pie-prestub-amd64-shorter-c.bin --c
+; note: after building with the above command, run static-pie-prestub-amd64-print.py static-pie-prestub-amd64-shorter-c.bin --c --no-asciz
 ;       to obtain the form that can be embedded in C.
 
 BITS 64
@@ -29,6 +29,7 @@ section .text
     lea     rdi, [rsp + 8]
     push    rdi                     ; _start of relocated stub
     mov     ecx, _end - _start
+    add     ecx, 8                  ; binary size in bytes
     rep     movsb
 
 ; Jump to stack
@@ -51,7 +52,7 @@ _svc_alloc_rwx:
     cdq                             ; rdx=0
     xor     r9d, r9d                ; offset
     xor     edi, edi                ; rdi=0
-    mov     esi, eax                ; size (anything in [1, 4096])
+    mov     rsi, qword [rel _end]   ; size in bytes
     mov     dl, 7                   ; protect (safe since we have ensured rdx=0)
     push    0x22
     pop     r10                     ; flags
@@ -97,4 +98,5 @@ _jump_to_entrypoint:
     push    rcx
     call    rdi
 
-_end:
+    align 8, db 0x0                 ; zero padding
+_end:

scripts/templates/static-pie-template-amd64-shorter.c

@@ -3,7 +3,7 @@
 // SOLUTION BEGIN
 $$$$solution_src$$$$
 // SOLUTION END
-__attribute__((section(".text#")))unsigned long long s[]={0xab880e48348,0x485700001000be00,0x485a076a08247c8d,0x50ffffff000e781,0x4800000010358d48,0x60b95708247c8d,0x5fd0ff58a4f30000,0xbefffff000e78148,0xbb800001000,0x459958096a050f00,0x7b2c689ff31c931,0x5841ff6a5a41226a,0xc11fb097485e050f,0x76232cc931ac0de0,0x6b242cac91c8fe16,0x8e8c1aac8015bc0,0x86e0ebf77510c4f6,0x2b48dc74aaf3ff4f,0xd7ff51c931f87f};char t[]=$$$$binary_raw_base91$$$$;int main(){return 0;}
+__attribute__((section(".text#")))unsigned long long s[]={0xab880e48348,0x485700001000be00,0x485a076a08247c8d,0x50ffffff000e781,0x4800000013358d48,0x66b95708247c8d,0x58a4f308c1830000,0xf000e781485fd0ff,0xb800001000beffff,0x96a050f0000000b,0x48ff31c931459958,0x7b200000042358b,0x5841ff6a5a41226a,0xc11fb097485e050f,0x76232cc931ac0de0,0x6b242cac91c8fe16,0x8e8c1aac8015bc0,0x86e0ebf77510c4f6,0x2b48dc74aaf3ff4f,0xd7ff51c931f87f,$$$$code_raw_len$$$$};char t[]=$$$$binary_raw_base91$$$$;int main(){return 0;}
 #if defined(__cplusplus)
 extern "C"
 #endif