On Sat, Nov 20, 2021 at 9:43 AM Jacob Kroon via Gcc-patches
<[email protected]> wrote:
>
> On 11/20/21 09:32, Jakub Jelinek wrote:
> > On Sat, Nov 20, 2021 at 12:24:21AM -0800, Andrew Pinski via Gcc-patches
> > wrote:
> >> On Sat, Nov 20, 2021 at 12:18 AM Jacob Kroon via Gcc-patches
> >> <[email protected]> wrote:
> >>>
> >>> cc1/cc1plus both include a checksum of the object files, archives and
> >>> options used during linking. Unless the host binutils has been built
> >>> with --enable-deterministic-archives, the archives will have different
> >>> checksums each build due to changes in timestamps of the containing
> >>> object files, and thus the checksum that is embedded in cc1/cc1plus will
> >>> also change.
> >>>
> >>> Fix this by passing "D" to ar/ranlib when creating the archives.
> >>
> >> How portable is the D option? That does it work with Mac OS X's ar;
> >> what about AIX's ar; what about Solaris's ar, etc?
> >> GNU binutils is not the only ar which is supported for the host.
> >
> > It isn't portable and even if ar/ranlib do support that option, not all
> > users will want it, so forcing it upon them unconditionally is IMO not a
> > good idea. If anything, configure needs to check if those options are
> > supported and there needs to be preferrably non-default gcc configure
> > option to request it.
> >
>
> Aha, I see. Then perhaps an easier and more portable solution for
> improving the reproducibility would be to make it optional whether or
> not to include the archives when calculating the checksums.
>
> Or perhaps by checksumming the containing object files instead of the
> archives themselves ?
>
> Would something like that be acceptable ?
At SUSE we are carrying patches to not include the checksum at all
but instead use the linker generated build-id as checksum for PCH.
I suppose when we add the configure option to disable PCH support
we could also elide checksum processing (or as in the patch simply
always use all-zeros).
Attached for reference.
Richard.
>
> Jacob
Use the binaries build-id as checksum for PCH purposes.
diff --git a/gcc/c-family/c-pch.c b/gcc/c-family/c-pch.c
index 9c0bd0b631d..de762517fc3 100644
--- a/gcc/c-family/c-pch.c
+++ b/gcc/c-family/c-pch.c
@@ -65,6 +65,66 @@ static FILE *pch_outfile;
static const char *get_ident (void);
+#if _GNU_SOURCE
+#include <link.h>
+
+#define ALIGN(val, align) (((val) + (align) - 1) & ~((align) - 1))
+
+static int
+get_build_id_1 (struct dl_phdr_info *info, size_t, void *data)
+{
+ for (unsigned i = 0; i < info->dlpi_phnum; ++i)
+ {
+ if (info->dlpi_phdr[i].p_type != PT_NOTE)
+ continue;
+ ElfW(Nhdr) *nhdr
+ = (ElfW(Nhdr) *)(info->dlpi_addr + info->dlpi_phdr[i].p_vaddr);
+ ptrdiff_t size = info->dlpi_phdr[i].p_filesz;
+ ptrdiff_t align = info->dlpi_phdr[i].p_align;
+ if (align != 8)
+ align = 4;
+ while (size >= (ptrdiff_t)sizeof (ElfW(Nhdr)))
+ {
+ if (nhdr->n_type == NT_GNU_BUILD_ID
+ && nhdr->n_namesz == 4
+ && strncmp ((char *)nhdr
+ + sizeof (ElfW(Nhdr)),
+ "GNU", 4) == 0
+ && nhdr->n_descsz >= 16)
+ {
+ memcpy (data,
+ (char *)nhdr
+ + ALIGN (sizeof (ElfW(Nhdr))
+ + nhdr->n_namesz, align), 16);
+ return 1;
+ }
+ size_t offset = (ALIGN (sizeof (ElfW(Nhdr))
+ + nhdr->n_namesz, align)
+ + ALIGN(nhdr->n_descsz, align));
+ nhdr = (ElfW(Nhdr) *)((char *)nhdr + offset);
+ size -= offset;
+ }
+ }
+
+ return 0;
+}
+
+static const unsigned char *
+get_build_id ()
+{
+ static unsigned char build_id[16];
+ if (!dl_iterate_phdr (get_build_id_1, build_id))
+ return NULL;
+ return build_id;
+}
+#else
+static const unsigned char *
+get_build_id ()
+{
+ return NULL;
+}
+#endif
+
/* Compute an appropriate 8-byte magic number for the PCH file, so that
utilities like file(1) can identify it, and so that GCC can quickly
ignore non-PCH files and PCH files that are of a completely different
@@ -120,8 +180,11 @@ pch_init (void)
v.pch_init = &pch_init;
target_validity = targetm.get_pch_validity (&v.target_data_length);
+ const unsigned char *chksum = get_build_id ();
+ if (!chksum)
+ chksum = executable_checksum;
if (fwrite (partial_pch, IDENT_LENGTH, 1, f) != 1
- || fwrite (executable_checksum, 16, 1, f) != 1
+ || fwrite (chksum, 16, 1, f) != 1
|| fwrite (&v, sizeof (v), 1, f) != 1
|| fwrite (target_validity, v.target_data_length, 1, f) != 1)
fatal_error (input_location, "cannot write to %s: %m", pch_file);
@@ -232,7 +295,10 @@ c_common_valid_pch (cpp_reader *pfile, const char *name, int fd)
cpp_warning (pfile, CPP_W_INVALID_PCH, "%s: not a PCH file", name);
return 2;
}
- if (memcmp (ident + IDENT_LENGTH, executable_checksum, 16) != 0)
+ const unsigned char *chksum = get_build_id ();
+ if (!chksum)
+ chksum = executable_checksum;
+ if (memcmp (ident + IDENT_LENGTH, chksum, 16) != 0)
{
cpp_warning (pfile, CPP_W_INVALID_PCH,
"%s: created by a different GCC executable", name);
diff --git a/gcc/genchecksum.c b/gcc/genchecksum.c
index 09fbb63fa93..ec8b3281d53 100644
--- a/gcc/genchecksum.c
+++ b/gcc/genchecksum.c
@@ -113,8 +113,13 @@ main (int argc, char ** argv)
puts ("#include \"config.h\"");
puts ("#include \"system.h\"");
fputs ("EXPORTED_CONST unsigned char executable_checksum[16] = { ", stdout);
+#if _GNU_SOURCE
+ for (i = 0; i < 16; i++)
+ printf ("0x%02x%s", 0, i == 15 ? " };\n" : ", ");
+#else
for (i = 0; i < 16; i++)
printf ("0x%02x%s", result[i], i == 15 ? " };\n" : ", ");
+#endif
return 0;
}
