Re: [fsverity-utils PATCH v2 2/2] Allow to build and run sign/digest on Windows

From: Luca Boccassi <luca.boccassi@gmail.com>
To: Eric Biggers <ebiggers@kernel.org>
Cc: linux-fscrypt@vger.kernel.org
Subject: Re: [fsverity-utils PATCH v2 2/2] Allow to build and run sign/digest on Windows
Date: Thu, 17 Dec 2020 19:12:06 +0000	[thread overview]
Message-ID: <73c70f8bcb6632ab3e161d9b0263bc1563e96b34.camel@gmail.com> (raw)
In-Reply-To: <X9ur8imAGcnv7Xx6@sol.localdomain>

[-- Attachment #1: Type: text/plain, Size: 8201 bytes --]

On Thu, 2020-12-17 at 11:05 -0800, Eric Biggers wrote:
> On Thu, Dec 17, 2020 at 06:44:38PM +0000, Luca Boccassi wrote:
> > On Thu, 2020-12-17 at 10:32 -0800, Eric Biggers wrote:
> > > On Thu, Dec 17, 2020 at 02:47:49PM +0000, luca.boccassi@gmail.com wrote:
> > > > From: Luca Boccassi <luca.boccassi@microsoft.com>
> > > > 
> > > > Add some minimal compat type defs, and stub out the enable/measure
> > > > sources. Also add a way to handle the fact that mingw adds a
> > > > .exe extension automatically in the Makefile install rules, and
> > > > that there is not pkg-config and the libcrypto linker flag is
> > > > different.
> > > > 
> > > > Signed-off-by: Luca Boccassi <luca.boccassi@microsoft.com>
> > > > ---
> > > > v2: rework the stubbing out to detect mingw in the Makefile and remove
> > > >     sources from compilation, instead of ifdefs.
> > > >     add a new common/win32_defs.h for the compat definitions.
> > > >     define strerror_r using strerror_s.
> > > > 
> > > >     To compile with mingw:
> > > >       make CC=x86_64-w64-mingw32-gcc-8.3-win32
> > > >     note that the openssl headers and a win32 libcrypto.dll need
> > > >     to be available in the default search paths, and otherwise have
> > > >     to be specified as expected via CPPFLAGS/LDFLAGS
> > > > 
> > > 
> > > I got some warnings and an error when compiling:
> > > 
> > > $ make CC=x86_64-w64-mingw32-gcc
> > >   CC       lib/compute_digest.o
> > >   CC       lib/hash_algs.o
> > >   CC       lib/sign_digest.o
> > >   CC       lib/utils.o
> > > lib/utils.c: In function ‘xmalloc’:
> > > lib/utils.c:25:25: warning: unknown conversion type character ‘l’ in format [-Wformat=]
> > >    25 |   libfsverity_error_msg("out of memory (tried to allocate %" SIZET_PF " bytes)",
> > >       |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > In file included from lib/../common/win32_defs.h:24,
> > >                  from lib/../common/common_defs.h:18,
> > >                  from lib/lib_private.h:15,
> > >                  from lib/utils.c:14:
> > > /usr/x86_64-w64-mingw32/include/inttypes.h:36:18: note: format string is defined here
> > >    36 | #define PRIu64 "llu"
> > >       |                  ^
> > > lib/utils.c:25:25: warning: too many arguments for format [-Wformat-extra-args]
> > >    25 |   libfsverity_error_msg("out of memory (tried to allocate %" SIZET_PF " bytes)",
> > >       |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > >   AR       libfsverity.a
> > >   CC       lib/compute_digest.shlib.o
> > >   CC       lib/hash_algs.shlib.o
> > >   CC       lib/sign_digest.shlib.o
> > >   CC       lib/utils.shlib.o
> > > lib/utils.c: In function ‘xmalloc’:
> > > lib/utils.c:25:25: warning: unknown conversion type character ‘l’ in format [-Wformat=]
> > >    25 |   libfsverity_error_msg("out of memory (tried to allocate %" SIZET_PF " bytes)",
> > >       |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > In file included from lib/../common/win32_defs.h:24,
> > >                  from lib/../common/common_defs.h:18,
> > >                  from lib/lib_private.h:15,
> > >                  from lib/utils.c:14:
> > > /usr/x86_64-w64-mingw32/include/inttypes.h:36:18: note: format string is defined here
> > >    36 | #define PRIu64 "llu"
> > >       |                  ^
> > > lib/utils.c:25:25: warning: too many arguments for format [-Wformat-extra-args]
> > >    25 |   libfsverity_error_msg("out of memory (tried to allocate %" SIZET_PF " bytes)",
> > >       |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > >   CCLD     libfsverity.so.0
> > > /usr/lib/gcc/x86_64-w64-mingw32/10.2.0/../../../../x86_64-w64-mingw32/bin/ld: cannot find -l:libcrypto.dll
> > > collect2: error: ld returned 1 exit status
> > > make: *** [Makefile:137: libfsverity.so.0] Error 1
> > > 
> > > 
> > > 
> > > This is on Arch Linux with mingw-w64-gcc and mingw-w64-openssl installed.
> > > 
> > > So there's something wrong with the SIZET_PF format string, and also
> > > -l:libcrypto.dll isn't correct; it should be just -lcrypto like it is on Linux.
> > > (MinGW knows to look for a .dll file.)
> > > 
> > > - Eric
> > 
> > Mmh I don't get any warnings on Debian - gcc-mingw-w64-x86-64 8.3.0-
> > 6+21.3~deb10u1 - any idea how to fix it?
> > 
> > And on -lcrypto, it didn't use to work before the refactor - but now it
> > does. I have no clue what was happening. Will change it back in v3.
> > 
> 
> Apparently the definition of _GNU_SOURCE in lib/utils.c changes the printf
> implementation that is used from Microsoft's to MinGW's, but the use of
> __attribute__((format(printf))) is generating warnings assuming that Microsoft's
> printf implementation is used.  Always defining _GNU_SOURCE and then using
> __attribute__((format(gnu_printf))) might be the way to go:
> 
> diff --git a/Makefile b/Makefile
> index cc62818..44aee92 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -52,7 +52,7 @@ override CFLAGS := -Wall -Wundef				\
>  	$(call cc-option,-Wvla)					\
>  	$(CFLAGS)
>  
> -override CPPFLAGS := -Iinclude -D_FILE_OFFSET_BITS=64 $(CPPFLAGS)
> +override CPPFLAGS := -Iinclude -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE $(CPPFLAGS)
>  
>  ifneq ($(V),1)
>  QUIET_CC        = @echo '  CC      ' $@;
> diff --git a/common/win32_defs.h b/common/win32_defs.h
> index e13938a..4edb17f 100644
> --- a/common/win32_defs.h
> +++ b/common/win32_defs.h
> @@ -37,6 +37,11 @@
>  #  define __cold
>  #endif
>  
> +#ifndef __printf
> +#  define __printf(fmt_idx, vargs_idx) \
> +	__attribute__((format(gnu_printf, fmt_idx, vargs_idx)))
> +#endif
> +
>  typedef __signed__ char __s8;
>  typedef unsigned char __u8;
>  typedef __signed__ short __s16;
> diff --git a/lib/utils.c b/lib/utils.c
> index 8bb4413..55a4045 100644
> --- a/lib/utils.c
> +++ b/lib/utils.c
> @@ -9,8 +9,6 @@
>   * https://opensource.org/licenses/MIT.
>   */
>  
> -#define _GNU_SOURCE /* for asprintf() and strerror_r() */
> -
>  #include "lib_private.h"
>  
>  #include <stdio.h>

I get the following warning with the mingw build now:

lib/utils.c: In function ‘xmalloc’:
lib/utils.c:23:25: warning: format ‘%u’ expects argument of type ‘unsigned int’, but argument 2 has type ‘size_t’ {aka ‘long long unsigned int’} [-Wformat=]
   libfsverity_error_msg("out of memory (tried to allocate %" SIZET_PF " bytes)",
                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
           size);
           ~~~~           
In file included from lib/../common/win32_defs.h:24,
                 from lib/../common/common_defs.h:18,
                 from lib/lib_private.h:15,
                 from lib/utils.c:12:
/usr/share/mingw-w64/include/inttypes.h:94:20: note: format string is defined here
 #define PRIu64 "I64u"
  AR       libfsverity.a
  CC       lib/sign_digest.shlib.o
  CC       lib/compute_digest.shlib.o
  CC       lib/hash_algs.shlib.o
  CC       lib/utils.shlib.o
lib/utils.c: In function ‘xmalloc’:
lib/utils.c:23:25: warning: format ‘%u’ expects argument of type ‘unsigned int’, but argument 2 has type ‘size_t’ {aka ‘long long unsigned int’} [-Wformat=]
   libfsverity_error_msg("out of memory (tried to allocate %" SIZET_PF " bytes)",
                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
           size);
           ~~~~           
In file included from lib/../common/win32_defs.h:24,
                 from lib/../common/common_defs.h:18,
                 from lib/lib_private.h:15,
                 from lib/utils.c:12:
/usr/share/mingw-w64/include/inttypes.h:94:20: note: format string is defined here
 #define PRIu64 "I64u"

But, honestly, it seems harmless to me. If someone on Windows is trying
to get a digest and don't have memory to do it, they'll have bigger
problems to worry about than knowing how much it was requested. I'll
send a v3 with your suggested changes. As far as I can read online,
handling %zu in a cross compatible way is like the number one
annoyance.

-- 
Kind regards,
Luca Boccassi

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]