Consider structures, unions and enums defined in the source file as
internal and do not expand them. This way, changes to e.g. struct
serial_private in drivers/tty/serial/8250_pci.c will not affect the
checksum of the pciserial_* exports.
Enum constants can be used as array sizes; if the enum itself does not
appear in the symbol expansion, a change in the enum constant will go
unnoticed. Example patch that changes the ABI but does not change the
checksum with current genksyms:
| enum e {
| E1,
| E2,
|+ E3,
| E_MAX
| };
|
| struct s {
| int a[E_MAX];
| }
|
| int f(struct s *s) { ... }
| EXPORT_SYMBOL(f)
Therefore, remember the value of each enum constant and
expand each occurence to <constant> <value>. The value is not actually
computed, but instead an expression in the form
(last explicitly assigned value) + N
is used. This avoids having to parse and semantically understand whole
of C.
Note: The changes won't take effect until the lexer and parser are
rebuilt by the next patch.
Signed-off-by: Michal Marek <mmarek@suse.cz>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
FreeBSD does not like <malloc.h> when __STDC__ is defined, use the standard
<stdlib.h> instead.
Signed-off-by: Arnaud Lacombe <lacombar@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
We are having two kinds of problems with genksyms today: fake checksum
changes without actual ABI changes, and changes which we would rather like
to ignore (such as an additional field at the end of a structure that
modules are not supposed to touch, for example).
I have thought about ways to improve genksyms and compute checksums
differently to avoid those problems, but in the end I don't see a
fundamentally better way. So here are some genksyms patches for at least
making the checksums more easily manageable, if we cannot fully fix them.
In addition to the bugfixes (the first two patches), this allows genksyms
to track checksum changes and report why a checksum changed (third patch),
and to selectively ignore changes (fourth patch).
This patch:
Gcc __attribute__ definitions may occur repeatedly, e.g.,
static int foo __attribute__((__used__))
__attribute__((aligned (16)));
The genksyms parser does not understand this, and generates a syntax error.
Fix this case.
Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Recently the __extension__ keyword has been introduced in the kernel.
Teach genksyms about this keyword so it can generate correct CRC for
exported symbols that uses a symbol marked __extension__.
For now only the typedef variant:
__extension__ typedef ...
is supported.
Later we may add more variants as needed.
This patch contains the actual source file changes. The
following patch will hold modifications to the generated
files (*_shipped) and only after the second patch the fix
has effect.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
This is a one-line change to parse.y.
To take advantage of this the scripts/genksyms/*_shipped files needs to
be rebuild - this is the next patch.
When a .c file contains:
DEFINE_PER_CPU(struct foo_s *, bar);
the .cpp output looks like:
__attribute__((__section__(".data.percpu"))) __typeof__(struct foo_s *) per_cpu__bar;
With the existing parse.y, the value inside the paranthesis of
__typeof__() does not evaluate as a type_specifier and therefore
per_cpu__bar does not get assigned a type for genksyms which results in
the EXPORT_PER_CPU_SYMBOL() not generating a CRC value.
I have compared the Modules.symvers with and without this
patch and for ia64's defconfig, the only change is:
Before 0x00000000 per_cpu____sn_nodepda vmlinux
After 0x9d3f3faa per_cpu____sn_nodepda vmlinux
per_cpu____sn_nodepda was the original source of my problems.
Signed-off-by: Robin Holt <holt@sgi.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
