How do I get the address of a C global variable through the Haskell FFI?

Question

For example, in C, I have something like this declared:

extern char _binary_res_txt_start[];

It came from this command:

ld -r -b binary -o binary.o res.txt

How can I get a Ptr in Haskell that points there? If it were a function, I'd do something like foreign import ccall "&exp" a_exp :: FunPtr (Double -> Double), but I don't think that works for variables.

Do you really need to load the data from an actual `.o` file for some reason, or is just the point that you want the data hard-baked into the executable in the end? In that case, there are better routes. — leftaroundabout, Apr 30 '20 at 13:58
There is a way to link a function from an object file `foreign import ccall "&exp" a_exp :: FunPtr (Double -> Double)` but how to link a data block? — mxvd, Apr 30 '20 at 14:16
@mxvd I incorporated your comment into the question, and also tried to simplify to get to the core of it. Let me know if I interpreted everything correctly. — Joseph Sible-Reinstate Monica, Apr 30 '20 at 15:40

Joseph Sible-Reinstate Monica · Accepted Answer · 2020-05-01T14:50:34.793

1

The foreign import syntax works for variables too, as long as you use &. You can then use peekCString or packCString to read it (or peekCStringLen or packCStringLen if it's not null-terminated but you know its length). Here's a full example:

printf 'foo\0bar' > res.txt
ld -r -b binary -o binary.o res.txt

extern char _binary_res_txt_start[];
extern char _binary_res_txt_end[];

import Foreign.C.Types
import Foreign.C.String
import Foreign.Ptr

foreign import ccall "&_binary_res_txt_start" txt_start :: Ptr CChar
foreign import ccall "&_binary_res_txt_end" txt_end :: Ptr CChar

main = do
  str <- peekCStringLen (txt_start, txt_end `minusPtr` txt_start)
  print str

"foo\NULbar"

Note that while ld -b binary also produces a _binary_res_txt_size symbol, it's no longer useful now that ASLR is a thing. See Why doesn't a linked binary file's _size symbol work correctly? for details on that.

edited May 01 '20 at 14:50

answered Apr 30 '20 at 15:38

Joseph Sible-Reinstate Monica

45,431
5
48
98

` Not in scope: type constructor or class ‘CChar’` where to find it? – mxvd May 01 '20 at 11:30
@mxvd [`import Foreign.C.Types`](https://hackage.haskell.org/package/base-4.12.0.0/docs/Foreign-C-Types.html#t:CChar) – Joseph Sible-Reinstate Monica May 01 '20 at 11:45
`import GHC.Ptr` `import Foreign.C.Types` `foreign import ccall safe "static &_binary_res_txt_start" txt_start :: Ptr СChar` `foreign import ccall safe "static &_binary_res_txt_size" txt_size :: Ptr Int` How can I now show the text on the screen? Text in UTF-8 `putStrLn` or `print` how to get a string from an address? – mxvd May 01 '20 at 12:15
@mxvd See the last sentence of my answer. [`peekCStringLen`](https://hackage.haskell.org/package/base-4.12.0.0/docs/Foreign-C-String.html#v:peekCStringLen) from `Foreign.C.String` can do that. – Joseph Sible-Reinstate Monica May 01 '20 at 12:18
@mxvd I just added a full example. – Joseph Sible-Reinstate Monica May 01 '20 at 13:16
`foreign import ccall "&_binary_res_txt_size" txt_size :: Ptr Int` 'size' is a value and not a pointer `foreign import ccall "_binary_res_txt_size" txt_size :: Int` – mxvd May 01 '20 at 13:19
@mxvd `&size` is a pointer even if `size` isn't. And your way doesn't work. – Joseph Sible-Reinstate Monica May 01 '20 at 13:29
`len <- peek txt_size str <- peekCStringLen (txt_start, fromIntegral len)` This code generates `Segmentation fault` at runtime – mxvd May 01 '20 at 13:31
@mxvd The standard says "It should be noted that a static foreign declaration that does not import an address (i.e., where `&` is not used in the specification of the external entity) always refers to a C function, even if the Haskell type is non-functional. For example, `foreign import ccall foo :: CInt` refers to a pure C function foo with no arguments that returns an integer value." – Joseph Sible-Reinstate Monica May 01 '20 at 13:32
@mxvd Try the exact code I posted without making any changes. – Joseph Sible-Reinstate Monica May 01 '20 at 13:33
The object file contains a string pointer and a length (not a length pointer) – mxvd May 01 '20 at 13:35
@mxvd I know it's not a pointer. Please try my exact code anyway and see what happens. – Joseph Sible-Reinstate Monica May 01 '20 at 13:37
I try the exact code you posted without making any changes. This code generates Segmentation fault at runtime – mxvd May 01 '20 at 13:41
@mxvd Please copy and paste the exact C declarations of `_binary_res_txt_start` and `_binary_res_txt_size` from the C code, then. I suspect they don't mean what you think they do. – Joseph Sible-Reinstate Monica May 01 '20 at 13:45
the object file is generated in this way `cat > ress.txt` `some text ` `ld -r -b binary -o binary.o res.txt` `objdump -x binary.o` – mxvd May 01 '20 at 13:50
@mxvd Okay, I see the problem now. `ld -b binary` stores the size in a really strange way. Fix coming shortly. – Joseph Sible-Reinstate Monica May 01 '20 at 14:17
@mxvd You looked too soon. I just now added the fix. – Joseph Sible-Reinstate Monica May 01 '20 at 14:39

How do I get the address of a C global variable through the Haskell FFI?

1 Answers1