Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-14 Thread Rusty Russell 
James Hogan  writes:
> On 14/03/13 04:00, Rusty Russell wrote:
>> From: Rusty Russell 
>> Subject: CONFIG_SYMBOL_PREFIX: cleanup.
...
> Reviewed-by: James Hogan 
> Tested-by: James Hogan  (metag)

Thanks.

> The only other special case of symbol prefixing I'm aware of is in
> scripts/genksyms/genksyms.c. It makes the decision at runtime based
> on the --arch=$ARCH argument, and is the only use of the arch argument:
>
>>  if ((strcmp(arch, "h8300") == 0) || (strcmp(arch, "blackfin") == 0) ||
>>  (strcmp(arch, "metag") == 0))
>>  mod_prefix = "_";
>
> Does the patch below look reasonable in addition to your patch?
> (Note: I'm not sure if genksyms is only used internally or whether
> it's API should be kept stable?).
>
> Thanks
> James
>
> Subject: [PATCH] genksyms: pass symbol-prefix instead of arch

Agreed, I've applied this.

Thanks,
Rusty.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-14 Thread James Hogan 
On 14/03/13 04:00, Rusty Russell wrote:
> From: Rusty Russell 
> Subject: CONFIG_SYMBOL_PREFIX: cleanup.
> 
> We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
> "_".  But Al Viro broke this in "consolidate cond_syscall and
> SYSCALL_ALIAS declarations" (in linux-next), and he's not the first to
> do so.
> 
> Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
> prefix it so something.  So various places define helpers which are
> defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:
> 
> 1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
> 2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
> 3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
> 4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
> 5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
> 6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
> 7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
>CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
>for pasting.
> 
> (arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).
> 
> Let's solve this properly:
> 1) No more generic prefix, just CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.
> 2) Make linux/export.h usable from asm.
> 3) Define VMLINUX_SYMBOL() and VMLINUX_SYMBOL_STR().
> 4) Make everyone use them.
> 
> Signed-off-by: Rusty Russell 

Reviewed-by: James Hogan 
Tested-by: James Hogan  (metag)

The only other special case of symbol prefixing I'm aware of is in
scripts/genksyms/genksyms.c. It makes the decision at runtime based
on the --arch=$ARCH argument, and is the only use of the arch argument:

>   if ((strcmp(arch, "h8300") == 0) || (strcmp(arch, "blackfin") == 0) ||
>   (strcmp(arch, "metag") == 0))
>   mod_prefix = "_";

Does the patch below look reasonable in addition to your patch?
(Note: I'm not sure if genksyms is only used internally or whether
it's API should be kept stable?).

Thanks
James

Subject: [PATCH] genksyms: pass symbol-prefix instead of arch

Pass symbol-prefix to genksyms instead of arch, so that the decision
what symbol prefix to use is kept in one place.

Basically genksyms used to take a -a $ARCH argument and it used that to
determine whether to add an underscore symbol prefix. It's now changed
to take a -s $SYMBOL_PREFIX argument so that the caller decides whether
a symbol prefix is required. The build system then uses
CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX to determine whether to pass the
argument.

Signed-off-by: James Hogan 
---
 scripts/Makefile.build  |  3 ++-
 scripts/genksyms/genksyms.c | 18 +++---
 2 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/scripts/Makefile.build b/scripts/Makefile.build
index 0e801c3..d5d859c 100644
--- a/scripts/Makefile.build
+++ b/scripts/Makefile.build
@@ -211,7 +211,8 @@ $(obj)/%.i: $(src)/%.c FORCE
 
 cmd_gensymtypes =   \
 $(CPP) -D__GENKSYMS__ $(c_flags) $< |   \
-$(GENKSYMS) $(if $(1), -T $(2)) -a $(ARCH)  \
+$(GENKSYMS) $(if $(1), -T $(2)) \
+ $(patsubst y,-s _,$(CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX)) \
  $(if $(KBUILD_PRESERVE),-p)\
  -r $(firstword $(wildcard $(2:.symtypes=.symref) /dev/null))
 
diff --git a/scripts/genksyms/genksyms.c b/scripts/genksyms/genksyms.c
index d25e4a1..88632df 100644
--- a/scripts/genksyms/genksyms.c
+++ b/scripts/genksyms/genksyms.c
@@ -45,7 +45,6 @@ int in_source_file;
 
 static int flag_debug, flag_dump_defs, flag_reference, flag_dump_types,
   flag_preserve, flag_warnings;
-static const char *arch = "";
 static const char *mod_prefix = "";
 
 static int errors;
@@ -731,7 +730,7 @@ static void genksyms_usage(void)
 {
fputs("Usage:\n" "genksyms [-adDTwqhV] > /path/to/.tmp_obj.ver\n" "\n"
 #ifdef __GNU_LIBRARY__
- "  -a, --archSelect architecture\n"
+ "  -s, --symbol-prefix   Select symbol prefix\n"
  "  -d, --debug   Increment the debug level (repeatable)\n"
  "  -D, --dumpDump expanded symbol defs (for debugging 
only)\n"
  "  -r, --reference file  Read reference symbols from a file\n"
@@ -742,7 +741,7 @@ static void genksyms_usage(void)
  "  -h, --helpPrint this message\n"
  "  -V, --version Print the release version\n"
 #else  /* __GNU_LIBRARY__ */
- "  -aSelect architecture\n"
+ "  -sSelect symbol prefix\n"
  "  -dIncrement the debug level (repeatable)\n"
  "  -DDump expanded symbol defs (for debugging 
only)\n"
  "  -r file   Read reference symbols from a file\n"
@@

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-13 Thread Rusty Russell 
Sam Ravnborg  writes:
>> actually, y'know, proof-read it.
> Hmm..

>> +   $(KERNELRELEASE) "$(patsubst 
>> y,_,$(CONFIG_HAVE_SYMBOL_PREFIX_UNDERSCORE))"
>>  
>> +config HAVE_UNDERSCORE_SYMBOL_PREFIX
>
> HAVE_UNDERSCORE_... or HAVE_SYMBOL_... confusion.
> I prefer the HAVE_SYMBOL_... variant but no strong feelings..

Thanks, fixed.  (HAVE_UNDERSCORE_ won by grep majority vote, but I don't
really care either).

>> + *
>> + * If you think the above arrogance just encourages more people to add
>> + * random crap to this file, you're not alone.
> Kill this.

??  It went in without my Ack, and it's a particularly offensive turd.

I can replace it with a civilized comment, instead of being snarky
though...

>>  /* Some toolchains use a `_' prefix for all user symbols. */
>> -#ifdef CONFIG_SYMBOL_PREFIX
>> -#define MODULE_SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
>> +#ifdef CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX
>> +#define __VMLINUX_SYMBOL(x) _##x
>> +#define __VMLINUX_SYMBOL_STR(x) "_" #x
>> +#define VMLINUX_SYMBOL_PREFIX_STR "_"
>>  #else
>> -#define MODULE_SYMBOL_PREFIX ""
>> +#define __VMLINUX_SYMBOL(x) x
>> +#define __VMLINUX_SYMBOL_STR(x) #x
>> +#define VMLINUX_SYMBOL_PREFIX_STR ""
>>  #endif
>
> We know the prefix is an underscore. No benefits from defining
> VMLINUX_SYMBOL_PREFIX_STR. The config name even syas so.
> Skipping the above give us only one way to check
> for the prefix - today we mix the two.

After testing, I think you're right.  The two users are both slightly
clearer after this (though mtd is a tiny bit less optimal).

Note: James, your patch hunk now looks like this:

  - buf_printf(b, "\t{ %#8x, \"%s\" },\n", s->crc, s->name);
  + buf_printf(b, "\t{ %#8x, VMLINUX_SYMBOL_STR(%s) },\n",
  +s->crc, s->name);

Here's version 3:

From: Rusty Russell 
Subject: CONFIG_SYMBOL_PREFIX: cleanup.

We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
"_".  But Al Viro broke this in "consolidate cond_syscall and
SYSCALL_ALIAS declarations" (in linux-next), and he's not the first to
do so.

Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
prefix it so something.  So various places define helpers which are
defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:

1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
   CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
   for pasting.

(arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).

Let's solve this properly:
1) No more generic prefix, just CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.
2) Make linux/export.h usable from asm.
3) Define VMLINUX_SYMBOL() and VMLINUX_SYMBOL_STR().
4) Make everyone use them.

Signed-off-by: Rusty Russell 

diff --git a/Makefile b/Makefile
index a05ea42..0b09ba5 100644
--- a/Makefile
+++ b/Makefile
@@ -1398,7 +1398,7 @@ quiet_cmd_rmfiles = $(if $(wildcard $(rm-files)),CLEAN   
$(wildcard $(rm-files))
 # Run depmod only if we have System.map and depmod is executable
 quiet_cmd_depmod = DEPMOD  $(KERNELRELEASE)
   cmd_depmod = $(CONFIG_SHELL) $(srctree)/scripts/depmod.sh $(DEPMOD) \
-   $(KERNELRELEASE) "$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))"
+   $(KERNELRELEASE) "$(patsubst 
y,_,$(CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX))"
 
 # Create temporary dir for module support files
 # clean it up only when building all modules
diff --git a/arch/Kconfig b/arch/Kconfig
index 1455579..7b433a4 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -384,6 +384,12 @@ config MODULES_USE_ELF_REL
  Modules only use ELF REL relocations.  Modules with ELF RELA
  relocations will give an error.
 
+config HAVE_UNDERSCORE_SYMBOL_PREFIX
+   bool
+   help
+ Some architectures generate an _ in front of C symbols; things like
+ module loading and assembly files need to know about this.
+
 #
 # ABI hall of shame
 #
diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index c3f2e0b..453ebe4 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -1,7 +1,3 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
-
 config MMU
def_bool n
 
@@ -33,6 +29,7 @@ config BLACKFIN
select ARCH_HAVE_CUSTOM_GPIO_H
select ARCH_WANT_OPTIONAL_GPIOLIB
select HAVE_UID16
+   select HAVE_UNDERSCORE_SYMBOL_PREFIX
select VIRT_TO_BUS
select ARCH_WANT_IPC_PARSE_VERSION
select HAVE_GENERIC_HARDIRQS
diff --git a/arch/h8300/Kconfig b/arch/h8300/Kconfig
index 79250de..303e4f9 100644
--- a/arch/h8300/Kconfig
+++ b/arch/h8300/Kconfig
@@ -12,10 +12,7 @@ config H8300

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-13 Thread Sam Ravnborg 
On Wed, Mar 13, 2013 at 09:21:01AM +, James Hogan wrote:
> Hi Sam,
> 
> On 13/03/13 06:31, Sam Ravnborg wrote:
> >>  /* Some toolchains use a `_' prefix for all user symbols. */
> >> -#ifdef CONFIG_SYMBOL_PREFIX
> >> -#define MODULE_SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
> >> +#ifdef CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX
> >> +#define __VMLINUX_SYMBOL(x) _##x
> >> +#define __VMLINUX_SYMBOL_STR(x) "_" #x
> >> +#define VMLINUX_SYMBOL_PREFIX_STR "_"
> >>  #else
> >> -#define MODULE_SYMBOL_PREFIX ""
> >> +#define __VMLINUX_SYMBOL(x) x
> >> +#define __VMLINUX_SYMBOL_STR(x) #x
> >> +#define VMLINUX_SYMBOL_PREFIX_STR ""
> >>  #endif
> > 
> > We know the prefix is an underscore. No benefits from defining
> > VMLINUX_SYMBOL_PREFIX_STR.
> 
> scripts/modpost.c uses the actual prefix string though:
> 
> > if (memcmp(symname, MODULE_SYMBOL_PREFIX,
> >strlen(MODULE_SYMBOL_PREFIX)) == 0) {
> > mod->unres =
> >   alloc_symbol(symname +
> >strlen(MODULE_SYMBOL_PREFIX),
> >ELF_ST_BIND(sym->st_info) == STB_WEAK,
> >mod->unres);
> 

Allowing ourself to know this is "_" then we could make the above code look
something like:

if (have_symbol_prefix_underscore) {
if (symname[0] == '_') {
mod->unres =
...

But I do not feel strong about this part.

It is much more important to clean up so we use the
same definition all over the kernel.

Sam
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-13 Thread James Hogan 
Hi Sam,

On 13/03/13 06:31, Sam Ravnborg wrote:
>>  /* Some toolchains use a `_' prefix for all user symbols. */
>> -#ifdef CONFIG_SYMBOL_PREFIX
>> -#define MODULE_SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
>> +#ifdef CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX
>> +#define __VMLINUX_SYMBOL(x) _##x
>> +#define __VMLINUX_SYMBOL_STR(x) "_" #x
>> +#define VMLINUX_SYMBOL_PREFIX_STR "_"
>>  #else
>> -#define MODULE_SYMBOL_PREFIX ""
>> +#define __VMLINUX_SYMBOL(x) x
>> +#define __VMLINUX_SYMBOL_STR(x) #x
>> +#define VMLINUX_SYMBOL_PREFIX_STR ""
>>  #endif
> 
> We know the prefix is an underscore. No benefits from defining
> VMLINUX_SYMBOL_PREFIX_STR.

scripts/modpost.c uses the actual prefix string though:

>   if (memcmp(symname, MODULE_SYMBOL_PREFIX,
>  strlen(MODULE_SYMBOL_PREFIX)) == 0) {
>   mod->unres =
> alloc_symbol(symname +
>  strlen(MODULE_SYMBOL_PREFIX),
>  ELF_ST_BIND(sym->st_info) == STB_WEAK,
>  mod->unres);

See also my "module: fix symbol versioning with symbol prefixes" patch
for an additional use in modpost.

Having a MODULE_SYMBOL_PREFIX/VMLINUX_SYMBOL_PREFIX_STR seems like the
simplest way to keep this code clean and readable without special cases.

Cheers
James

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-12 Thread Sam Ravnborg 
> actually, y'know, proof-read it.
Hmm..


> +   $(KERNELRELEASE) "$(patsubst 
> y,_,$(CONFIG_HAVE_SYMBOL_PREFIX_UNDERSCORE))"
>  
> +config HAVE_UNDERSCORE_SYMBOL_PREFIX

HAVE_UNDERSCORE_... or HAVE_SYMBOL_... confusion.
I prefer the HAVE_SYMBOL_... variant but no strong feelings..

> + *
> + * If you think the above arrogance just encourages more people to add
> + * random crap to this file, you're not alone.
Kill this.

>  /* Some toolchains use a `_' prefix for all user symbols. */
> -#ifdef CONFIG_SYMBOL_PREFIX
> -#define MODULE_SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
> +#ifdef CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX
> +#define __VMLINUX_SYMBOL(x) _##x
> +#define __VMLINUX_SYMBOL_STR(x) "_" #x
> +#define VMLINUX_SYMBOL_PREFIX_STR "_"
>  #else
> -#define MODULE_SYMBOL_PREFIX ""
> +#define __VMLINUX_SYMBOL(x) x
> +#define __VMLINUX_SYMBOL_STR(x) #x
> +#define VMLINUX_SYMBOL_PREFIX_STR ""
>  #endif

We know the prefix is an underscore. No benefits from defining
VMLINUX_SYMBOL_PREFIX_STR. The config name even syas so.
Skipping the above give us only one way to check
for the prefix - today we mix the two.

Sam

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-12 Thread Rusty Russell 
Stephen Rothwell  writes:

> Hi Rusty,
>
> Looks partly better.   You seem to be using
> CONFIG_SYMBOL_PREFIX_UNDERSCORE but selecting
> CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.  One trivial comment below.
>
> Maybe this was an unfinished version of the patch?

Indeed.  It was crap now I've applied James' copious fixes, and
actually, y'know, proof-read it.

This is what I'll push into my fixes/ branch now, and if it survives a
day in linux-next, will push to Linus.

Thanks,
Rusty.

v2: Rename CONFIG_SYMBOL_PREFIX_UNDERSCORE to
CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX, which is defined in
arch/Kconfig and selected by the 3 archs which need it.

v3: Clean up remnants of v1, s/VMLINUX_SYMBOL_NAME/VMLINUX_SYMBOL_STR/
in description, note that Al's commit is in linux-next only,
remove whitespace-only hunk.

CONFIG_SYMBOL_PREFIX: cleanup.

We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
"_".  But Al Viro broke this in "consolidate cond_syscall and
SYSCALL_ALIAS declarations" (in linux-next), and he's not the first to
do so.

Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
prefix it so something.  So various places define helpers which are
defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:

1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
   CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
   for pasting.

(arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).

Let's solve this properly:
1) No more generic prefix, just CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.
2) Make linux/export.h usable from asm.
3) Define VMLINUX_SYMBOL, VMLINUX_SYMBOL_STR and VMLINUX_SYMBOL_PREFIX_STR.
4) Make everyone use them.

Signed-off-by: Rusty Russell 
Reviewed-by: James Hogan 
Tested-by: James Hogan  (on metag)

diff --git a/Makefile b/Makefile
index a05ea42..10fb6c7 100644
--- a/Makefile
+++ b/Makefile
@@ -1398,7 +1398,7 @@ quiet_cmd_rmfiles = $(if $(wildcard $(rm-files)),CLEAN   
$(wildcard $(rm-files))
 # Run depmod only if we have System.map and depmod is executable
 quiet_cmd_depmod = DEPMOD  $(KERNELRELEASE)
   cmd_depmod = $(CONFIG_SHELL) $(srctree)/scripts/depmod.sh $(DEPMOD) \
-   $(KERNELRELEASE) "$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))"
+   $(KERNELRELEASE) "$(patsubst 
y,_,$(CONFIG_HAVE_SYMBOL_PREFIX_UNDERSCORE))"
 
 # Create temporary dir for module support files
 # clean it up only when building all modules
diff --git a/arch/Kconfig b/arch/Kconfig
index 5a1779c..12ba8f9 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -391,6 +391,12 @@ config MODULES_USE_ELF_REL
  Modules only use ELF REL relocations.  Modules with ELF RELA
  relocations will give an error.
 
+config HAVE_UNDERSCORE_SYMBOL_PREFIX
+   bool
+   help
+ Some architectures generate an _ in front of C symbols; things like
+ module loading and assembly files need to know about this.
+
 #
 # ABI hall of shame
 #
diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 600494c..314ee6a 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -1,7 +1,3 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
-
 config MMU
def_bool n
 
@@ -33,6 +29,7 @@ config BLACKFIN
select ARCH_HAVE_CUSTOM_GPIO_H
select ARCH_WANT_OPTIONAL_GPIOLIB
select HAVE_UID16
+   select HAVE_UNDERSCORE_SYMBOL_PREFIX
select HAVE_VIRT_TO_BUS
select ARCH_WANT_IPC_PARSE_VERSION
select HAVE_GENERIC_HARDIRQS
diff --git a/arch/h8300/Kconfig b/arch/h8300/Kconfig
index ae8551e..2333cf6 100644
--- a/arch/h8300/Kconfig
+++ b/arch/h8300/Kconfig
@@ -12,10 +12,7 @@ config H8300
select MODULES_USE_ELF_RELA
select OLD_SIGSUSPEND3
select OLD_SIGACTION
-
-config SYMBOL_PREFIX
-   string
-   default "_"
+   select HAVE_UNDERSCORE_SYMBOL_PREFIX
 
 config MMU
bool
diff --git a/arch/metag/Kconfig b/arch/metag/Kconfig
index afc8973..2099617 100644
--- a/arch/metag/Kconfig
+++ b/arch/metag/Kconfig
@@ -1,7 +1,3 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
-
 config METAG
def_bool y
select EMBEDDED
@@ -27,6 +23,7 @@ config METAG
select HAVE_MOD_ARCH_SPECIFIC
select HAVE_PERF_EVENTS
select HAVE_SYSCALL_TRACEPOINTS
+   select HAVE_UNDERSCORE_SYMBOL_PREFIX
select IRQ_DOMAIN
select MODULES_USE_ELF_RELA
select OF
diff --git a/drivers/mtd/chips/gen_probe.c b/drivers/mtd/chips/gen_probe.c
index 3b9a284..00d20ad 100644
--- a/drivers/mtd/chips/gen_probe.c
+++ b/drivers/mtd/chips/gen_probe.c
@@ -204,1

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-12 Thread James Hogan 
Hi Rusty,

On 12/03/13 04:48, Rusty Russell wrote:
> v2: Rename CONFIG_SYMBOL_PREFIX_UNDERSCORE to 
> CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX,
> which is defined in arch/Kconfig and selected by the 3 archs which need 
> it.

Sorry I didn't get a chance to try your patch yesterday.

> Subject: CONFIG_SYMBOL_PREFIX: cleanup.
> 
> We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
> "_".  But Al Viro broke this in "consolidate cond_syscall and
> SYSCALL_ALIAS declarations", and he's not the first to do so.
> 
> Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
> prefix it so something.  So various places define helpers which are
> defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:
> 
> 1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
> 2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
> 3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
> 4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
> 5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
> 6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
> 7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
>CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
>for pasting.
> 
> (arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).
> 
> Let's solve this properly:
> 1) No more generic prefix, just CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.
> 2) Make linux/export.h usable from asm.
> 3) Define VMLINUX_SYMBOL, VMLINUX_SYMBOL_NAME and VMLINUX_SYMBOL_PREFIX_STR.

You don't seem to define VMLINUX_SYMBOL_NAME

> 4) Make everyone use them.
> 
> Signed-off-by: Rusty Russell 

After a few modifications detailed below (with a fixup patch at the
end for convenience) it seems to work no worse than before for metag
(module symbol versioning was apparently already broken, I need to
look into that).

Reviewed-by: James Hogan 
Tested-by: James Hogan  (on metag)

Cheers
James

> 
> diff --git a/Makefile b/Makefile
> index a05ea42..9dc948d 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -1398,7 +1398,7 @@ quiet_cmd_rmfiles = $(if $(wildcard $(rm-files)),CLEAN  
>  $(wildcard $(rm-files))
>  # Run depmod only if we have System.map and depmod is executable
>  quiet_cmd_depmod = DEPMOD  $(KERNELRELEASE)
>cmd_depmod = $(CONFIG_SHELL) $(srctree)/scripts/depmod.sh $(DEPMOD) \
> -   $(KERNELRELEASE) "$(patsubst 
> "%",%,$(CONFIG_SYMBOL_PREFIX))"
> +   $(KERNELRELEASE) "$(patsubst 
> y,_,$(CONFIG_SYMBOL_PREFIX_UNDERSCORE))"

should this be CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX now?


> diff --git a/arch/h8300/Kconfig b/arch/h8300/Kconfig
> index ae8551e..ba043e3 100644
> --- a/arch/h8300/Kconfig
> +++ b/arch/h8300/Kconfig
> @@ -12,10 +12,10 @@ config H8300
>   select MODULES_USE_ELF_RELA
>   select OLD_SIGSUSPEND3
>   select OLD_SIGACTION
> + select HAVE_UNDERSCORE_SYMBOL_PREFIX
>  
> -config SYMBOL_PREFIX
> - string
> - default "_"
> +config SYMBOL_PREFIX_UNDERSCORE
> + def_bool y

Not needed anymore I think


> diff --git a/arch/metag/Kconfig b/arch/metag/Kconfig
> index afc8973..88272ce 100644
> --- a/arch/metag/Kconfig
> +++ b/arch/metag/Kconfig
> @@ -1,6 +1,5 @@
> -config SYMBOL_PREFIX
> - string
> - default "_"
> +config SYMBOL_PREFIX_UNDERSCORE
> + def_bool y

Same again.


> diff --git a/include/asm-generic/unistd.h b/include/asm-generic/unistd.h
> index 4077b5d..80122d4 100644
> --- a/include/asm-generic/unistd.h
> +++ b/include/asm-generic/unistd.h
> @@ -1,4 +1,5 @@
>  #include 
> +#include 
>  
>  /*
>   * These are required system calls, we should
> @@ -17,12 +18,7 @@
>   * but it doesn't work on all toolchains, so we just do it by hand
>   */
>  #ifndef cond_syscall
> -#ifdef CONFIG_SYMBOL_PREFIX
> -#define __SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
> -#else
> -#define __SYMBOL_PREFIX
> -#endif
> -#define cond_syscall(x) asm(".weak\t" __SYMBOL_PREFIX #x "\n\t" \
> - ".set\t" __SYMBOL_PREFIX #x "," \
> - __SYMBOL_PREFIX "sys_ni_syscall")
> +#define cond_syscall(x) asm(".weak\t" SYMBOL_NAME_STR(x) "\n\t"  \
> + ".set\t" SYMBOL_NAME_STR(x) "," \
> + SYMBOL_NAME_STR(sys_ni_syscall))

s/SYMBOL_NAME_STR/VMLINUX_SYMBOL_STR/



> diff --git a/include/linux/export.h b/include/linux/export.h
> index 696c0f4..0080e93 100644
> --- a/include/linux/export.h
> +++ b/include/linux/export.h
> @@ -7,15 +7,27 @@
>   *
>   * If you feel the need to add #include  to this file
>   * then you are doing something wrong and should go away silently.
> + *
> + * If you think the above arrogance just encourages more people to add
> + * random crap to this file, you're not alone.

:-)

>   */
>  
>  /* Some toolchains use a `_' prefix for all user symbols. */
> -#ifdef CONFIG_SYMBOL_PREFIX
> -#define MODULE_SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
> +#ifdef CONFIG_SYMBOL_PREFIX_UNDERSCORE

CONFIG_HAVE_UNDERSCORE_SY

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-12 Thread Stephen Rothwell 
Hi Rusty,

Looks partly better.   You seem to be using
CONFIG_SYMBOL_PREFIX_UNDERSCORE but selecting
CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.  One trivial comment below.

Maybe this was an unfinished version of the patch?

On Tue, 12 Mar 2013 15:18:15 +1030 Rusty Russell  wrote:
>
> diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
> index 600494c..314ee6a 100644
> --- a/arch/blackfin/Kconfig
> +++ b/arch/blackfin/Kconfig
> @@ -45,6 +42,7 @@ config BLACKFIN
>   select HAVE_MOD_ARCH_SPECIFIC
>   select MODULES_USE_ELF_RELA
>  
> +
>  config GENERIC_CSUM
>   def_bool y

Extra blank line added.

-- 
Cheers,
Stephen Rothwells...@canb.auug.org.au


pgpM7PAMQmqYH.pgp
Description: PGP signature

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-12 Thread Rusty Russell 
Stephen Rothwell  writes:

> Hi Rusty,
>
> On Mon, 11 Mar 2013 17:05:17 +1030 Rusty Russell  
> wrote:
>>
>> diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
>> index 600494c..cd0f7c7 100644
>> --- a/arch/blackfin/Kconfig
>> +++ b/arch/blackfin/Kconfig
>> @@ -1,6 +1,5 @@
>> -config SYMBOL_PREFIX
>> -string
>> -default "_"
>> +config SYMBOL_PREFIX_UNDERSCORE
>> +def_bool y
>
> Please, if you are going to do this, just have
>
> config SYMBOL_PREFIX_UNDERSCORE
>   bool
>
> in one Kconfig file (probably arch/Kconfig) and then select it from the
> appropriate architecture Kconfig files.  You could even put a help text
> on it to explain why a new architecture may need to select it.
>
> -- 
> Cheers,
> Stephen Rothwells...@canb.auug.org.au

OK.

v2: Rename CONFIG_SYMBOL_PREFIX_UNDERSCORE to 
CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX,
which is defined in arch/Kconfig and selected by the 3 archs which need it.


Subject: CONFIG_SYMBOL_PREFIX: cleanup.

We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
"_".  But Al Viro broke this in "consolidate cond_syscall and
SYSCALL_ALIAS declarations", and he's not the first to do so.

Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
prefix it so something.  So various places define helpers which are
defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:

1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
   CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
   for pasting.

(arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).

Let's solve this properly:
1) No more generic prefix, just CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.
2) Make linux/export.h usable from asm.
3) Define VMLINUX_SYMBOL, VMLINUX_SYMBOL_NAME and VMLINUX_SYMBOL_PREFIX_STR.
4) Make everyone use them.

Signed-off-by: Rusty Russell 

diff --git a/Makefile b/Makefile
index a05ea42..9dc948d 100644
--- a/Makefile
+++ b/Makefile
@@ -1398,7 +1398,7 @@ quiet_cmd_rmfiles = $(if $(wildcard $(rm-files)),CLEAN   
$(wildcard $(rm-files))
 # Run depmod only if we have System.map and depmod is executable
 quiet_cmd_depmod = DEPMOD  $(KERNELRELEASE)
   cmd_depmod = $(CONFIG_SHELL) $(srctree)/scripts/depmod.sh $(DEPMOD) \
-   $(KERNELRELEASE) "$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))"
+   $(KERNELRELEASE) "$(patsubst 
y,_,$(CONFIG_SYMBOL_PREFIX_UNDERSCORE))"
 
 # Create temporary dir for module support files
 # clean it up only when building all modules
diff --git a/arch/Kconfig b/arch/Kconfig
index 5a1779c..12ba8f9 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -391,6 +391,12 @@ config MODULES_USE_ELF_REL
  Modules only use ELF REL relocations.  Modules with ELF RELA
  relocations will give an error.
 
+config HAVE_UNDERSCORE_SYMBOL_PREFIX
+   bool
+   help
+ Some architectures generate an _ in front of C symbols; things like
+ module loading and assembly files need to know about this.
+
 #
 # ABI hall of shame
 #
diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 600494c..314ee6a 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -1,7 +1,3 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
-
 config MMU
def_bool n
 
@@ -33,6 +29,7 @@ config BLACKFIN
select ARCH_HAVE_CUSTOM_GPIO_H
select ARCH_WANT_OPTIONAL_GPIOLIB
select HAVE_UID16
+   select HAVE_UNDERSCORE_SYMBOL_PREFIX
select HAVE_VIRT_TO_BUS
select ARCH_WANT_IPC_PARSE_VERSION
select HAVE_GENERIC_HARDIRQS
@@ -45,6 +42,7 @@ config BLACKFIN
select HAVE_MOD_ARCH_SPECIFIC
select MODULES_USE_ELF_RELA
 
+
 config GENERIC_CSUM
def_bool y
 
diff --git a/arch/h8300/Kconfig b/arch/h8300/Kconfig
index ae8551e..ba043e3 100644
--- a/arch/h8300/Kconfig
+++ b/arch/h8300/Kconfig
@@ -12,10 +12,10 @@ config H8300
select MODULES_USE_ELF_RELA
select OLD_SIGSUSPEND3
select OLD_SIGACTION
+   select HAVE_UNDERSCORE_SYMBOL_PREFIX
 
-config SYMBOL_PREFIX
-   string
-   default "_"
+config SYMBOL_PREFIX_UNDERSCORE
+   def_bool y
 
 config MMU
bool
diff --git a/arch/metag/Kconfig b/arch/metag/Kconfig
index afc8973..88272ce 100644
--- a/arch/metag/Kconfig
+++ b/arch/metag/Kconfig
@@ -1,6 +1,5 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
+config SYMBOL_PREFIX_UNDERSCORE
+   def_bool y
 
 config METAG
def_bool y
@@ -27,6 +26,7 @@ config METAG
select HAVE_MOD_ARCH_SPECIFIC
select HAVE_PERF_EVENTS
select HAVE_SYSCALL_TRACEPOINTS
+   sele

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-11 Thread Stephen Rothwell 
Hi Rusty,

On Mon, 11 Mar 2013 17:05:17 +1030 Rusty Russell  wrote:
>
> diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
> index 600494c..cd0f7c7 100644
> --- a/arch/blackfin/Kconfig
> +++ b/arch/blackfin/Kconfig
> @@ -1,6 +1,5 @@
> -config SYMBOL_PREFIX
> - string
> - default "_"
> +config SYMBOL_PREFIX_UNDERSCORE
> + def_bool y

Please, if you are going to do this, just have

config SYMBOL_PREFIX_UNDERSCORE
bool

in one Kconfig file (probably arch/Kconfig) and then select it from the
appropriate architecture Kconfig files.  You could even put a help text
on it to explain why a new architecture may need to select it.

-- 
Cheers,
Stephen Rothwells...@canb.auug.org.au


pgpfOWnRjJYGd.pgp
Description: PGP signature

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-10 Thread Rusty Russell 
James Hogan  writes:

> Hi Rusty,
>
> On 08/03/13 00:03, Rusty Russell wrote:
>> James Hogan  writes:
>>> Also the definition of SYMBOL_PREFIX in  is removed as
>>> it conflicts, isn't used anywhere, and is defined as a string so differs
>>> from the assembly definition.
>> 
>> So now, if CONFIG_SYMBOL_PREFIX, SYMBOL_PREFIX is defined on the cmdline
>> as a string.  Otherwise it's empty (not the empty string?):
>
> No, SYMBOL_PREFIX is now defined as a non-string, same as asm files

Ah, I misread this.  It's stripping the quotes, not adding them:

_sym_flags = -DSYMBOL_PREFIX=$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))

> Does that make sense? It's all a bit messy unfortunately (hence RFC).

Yes, let's clear it up once and for all.  Does this work?

CONFIG_SYMBOL_PREFIX: cleanup.

We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
"_".  But Al Viro broke this in "consolidate cond_syscall and
SYSCALL_ALIAS declarations", and he's not the first to do so.

Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
prefix it so something.  So various places define helpers which are
defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:

1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
   CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
   for pasting.

(arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).

Let's solve this properly:
1) No more generic prefix, just CONFIG_SYMBOL_PREFIX_UNDERSCORE.
2) Make linux/export.h usable from asm.
3) Define VMLINUX_SYMBOL, VMLINUX_SYMBOL_NAME and VMLINUX_SYMBOL_PREFIX_STR.
4) Make everyone use them.

Signed-off-by: Rusty Russell 

diff --git a/Makefile b/Makefile
index 5bd9f77..b2e3e1b 100644
--- a/Makefile
+++ b/Makefile
@@ -1398,7 +1398,7 @@ quiet_cmd_rmfiles = $(if $(wildcard $(rm-files)),CLEAN   
$(wildcard $(rm-files))
 # Run depmod only if we have System.map and depmod is executable
 quiet_cmd_depmod = DEPMOD  $(KERNELRELEASE)
   cmd_depmod = $(CONFIG_SHELL) $(srctree)/scripts/depmod.sh $(DEPMOD) \
-   $(KERNELRELEASE) "$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))"
+   $(KERNELRELEASE) "$(patsubst 
y,_,$(CONFIG_SYMBOL_PREFIX_UNDERSCORE))"
 
 # Create temporary dir for module support files
 # clean it up only when building all modules
diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 600494c..cd0f7c7 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -1,6 +1,5 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
+config SYMBOL_PREFIX_UNDERSCORE
+   def_bool y
 
 config MMU
def_bool n
diff --git a/arch/h8300/Kconfig b/arch/h8300/Kconfig
index ae8551e..ec9924d 100644
--- a/arch/h8300/Kconfig
+++ b/arch/h8300/Kconfig
@@ -13,9 +13,8 @@ config H8300
select OLD_SIGSUSPEND3
select OLD_SIGACTION
 
-config SYMBOL_PREFIX
-   string
-   default "_"
+config SYMBOL_PREFIX_UNDERSCORE
+   def_bool y
 
 config MMU
bool
diff --git a/arch/metag/Kconfig b/arch/metag/Kconfig
index afc8973..e8719b4 100644
--- a/arch/metag/Kconfig
+++ b/arch/metag/Kconfig
@@ -1,6 +1,5 @@
-config SYMBOL_PREFIX
-   string
-   default "_"
+config SYMBOL_PREFIX_UNDERSCORE
+   def_bool y
 
 config METAG
def_bool y
diff --git a/drivers/mtd/chips/gen_probe.c b/drivers/mtd/chips/gen_probe.c
index 3b9a284..00d20ad 100644
--- a/drivers/mtd/chips/gen_probe.c
+++ b/drivers/mtd/chips/gen_probe.c
@@ -204,14 +204,14 @@ static inline struct mtd_info *cfi_cmdset_unknown(struct 
map_info *map,
struct cfi_private *cfi = map->fldrv_priv;
__u16 type = primary?cfi->cfiq->P_ID:cfi->cfiq->A_ID;
 #ifdef CONFIG_MODULES
-   char probename[16+sizeof(MODULE_SYMBOL_PREFIX)];
+   char probename[16+sizeof(VMLINUX_SYMBOL_PREFIX_STR)];
cfi_cmdset_fn_t *probe_function;
 
-   sprintf(probename, MODULE_SYMBOL_PREFIX "cfi_cmdset_%4.4X", type);
+   sprintf(probename, VMLINUX_SYMBOL_PREFIX_STR "cfi_cmdset_%4.4X", type);
 
probe_function = __symbol_get(probename);
if (!probe_function) {
-   request_module(probename + sizeof(MODULE_SYMBOL_PREFIX) - 1);
+   request_module(probename + sizeof(VMLINUX_SYMBOL_PREFIX_STR)-1);
probe_function = __symbol_get(probename);
}
 
diff --git a/include/asm-generic/unistd.h b/include/asm-generic/unistd.h
index 4077b5d..80122d4 100644
--- a/include/asm-generic/unistd.h
+++ b/include/asm-generic/unistd.h
@@ -1,4 +1,5 @@
 #include 
+#include 
 
 /*
  * These are required system calls, we should
@@ -17,12 +18,7 @@
  * but it doesn't work on all toolchains, so we

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-08 Thread James Hogan 
Hi Rusty,

On 08/03/13 00:03, Rusty Russell wrote:
> James Hogan  writes:
>> Also the definition of SYMBOL_PREFIX in  is removed as
>> it conflicts, isn't used anywhere, and is defined as a string so differs
>> from the assembly definition.
> 
> So now, if CONFIG_SYMBOL_PREFIX, SYMBOL_PREFIX is defined on the cmdline
> as a string.  Otherwise it's empty (not the empty string?):

No, SYMBOL_PREFIX is now defined as a non-string, same as asm files, but
the now unused definition in linux/kernel.h did define it as a string as
it used CONFIG_SYMBOL_PREFIX which is a string. When I said "and is
defined as a string" I was referring to the one in linux/kernel.h that
this removes.

Does that make sense? It's all a bit messy unfortunately (hence RFC).

Cheers
James

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-07 Thread Rusty Russell 
James Hogan  writes:
> The commit "consolidate cond_syscall and SYSCALL_ALIAS declarations"
> broke the build on blackfin and metag due to the following code:
>
>   #ifndef SYMBOL_NAME
>   #ifdef CONFIG_SYMBOL_PREFIX
>   #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
>   #else
>   #define SYMBOL_NAME(x) x
>   #endif
>   #endif
>   #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))
>
> __stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get
> lines like this in the assembly output:
>
>   .weak CONFIG_SYMBOL_PREFIXsys_quotactl
>   .set
>   CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall
>
> This is fixed by defining SYMBOL_PREFIX from the command line for c
> files in addition to assembly for architectures that set
> CONFIG_SYMBOL_PREFIX (scripts/Makefile.lib), and defining __SYMBOL_NAME
> as:
>
>   #define __SYMBOL_NAME(x) __stringify(SYMBOL_PREFIX) #x
>
> We first have to ensure SYMBOL_PREFIX is defined (which avoids polluting
> the command lines for architectures that don't use symbol prefixes).
> Also the definition of SYMBOL_PREFIX in  is removed as
> it conflicts, isn't used anywhere, and is defined as a string so differs
> from the assembly definition.

So now, if CONFIG_SYMBOL_PREFIX, SYMBOL_PREFIX is defined on the cmdline
as a string.  Otherwise it's empty (not the empty string?):

> +/* This helps us to avoid #ifdef SYMBOL_PREFIX */
> +#ifndef SYMBOL_PREFIX
> +#define SYMBOL_PREFIX
>  #endif
> -#define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))
> +
> +#define __SYMBOL_NAME(x) __stringify(SYMBOL_PREFIX) #x

And why are you __stringify()ing a string?

>  #ifndef cond_syscall
>  #define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
> diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
> index 07125e6..f1cce6a 100644
> --- a/scripts/Makefile.lib
> +++ b/scripts/Makefile.lib
> @@ -123,6 +123,7 @@ ifdef CONFIG_SYMBOL_PREFIX
>  _sym_flags = -DSYMBOL_PREFIX=$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))
>  _cpp_flags += $(_sym_flags)
>  _a_flags += $(_sym_flags)
> +_c_flags += $(_sym_flags)
>  endif

Confused,
Rusty.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[RFC -next] linux/linkage.h: fix symbol prefix handling

2013-03-07 Thread James Hogan 
The commit "consolidate cond_syscall and SYSCALL_ALIAS declarations"
broke the build on blackfin and metag due to the following code:

  #ifndef SYMBOL_NAME
  #ifdef CONFIG_SYMBOL_PREFIX
  #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
  #else
  #define SYMBOL_NAME(x) x
  #endif
  #endif
  #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))

__stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get
lines like this in the assembly output:

  .weak CONFIG_SYMBOL_PREFIXsys_quotactl
  .set
  CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall

This is fixed by defining SYMBOL_PREFIX from the command line for c
files in addition to assembly for architectures that set
CONFIG_SYMBOL_PREFIX (scripts/Makefile.lib), and defining __SYMBOL_NAME
as:

  #define __SYMBOL_NAME(x) __stringify(SYMBOL_PREFIX) #x

We first have to ensure SYMBOL_PREFIX is defined (which avoids polluting
the command lines for architectures that don't use symbol prefixes).
Also the definition of SYMBOL_PREFIX in  is removed as
it conflicts, isn't used anywhere, and is defined as a string so differs
from the assembly definition.

Signed-off-by: James Hogan 
---
This is just one possible way of fixing it. It felt the cleanest out of
the ways I tried, but feel free to suggest something better!

 include/linux/kernel.h  |  7 ---
 include/linux/linkage.h | 12 +---
 scripts/Makefile.lib|  1 +
 3 files changed, 6 insertions(+), 14 deletions(-)

diff --git a/include/linux/kernel.h b/include/linux/kernel.h
index 80d3687..e13e992 100644
--- a/include/linux/kernel.h
+++ b/include/linux/kernel.h
@@ -723,13 +723,6 @@ static inline void ftrace_dump(enum ftrace_dump_mode 
oops_dump_mode) { }
 /* Trap pasters of __FUNCTION__ at compile-time */
 #define __FUNCTION__ (__func__)
 
-/* This helps us to avoid #ifdef CONFIG_SYMBOL_PREFIX */
-#ifdef CONFIG_SYMBOL_PREFIX
-#define SYMBOL_PREFIX CONFIG_SYMBOL_PREFIX
-#else
-#define SYMBOL_PREFIX ""
-#endif
-
 /* Rebuild everything on CONFIG_FTRACE_MCOUNT_RECORD */
 #ifdef CONFIG_FTRACE_MCOUNT_RECORD
 # define REBUILD_DUE_TO_FTRACE_MCOUNT_RECORD
diff --git a/include/linux/linkage.h b/include/linux/linkage.h
index 829d66c..acb869b 100644
--- a/include/linux/linkage.h
+++ b/include/linux/linkage.h
@@ -15,14 +15,12 @@
 #define asmlinkage CPP_ASMLINKAGE
 #endif
 
-#ifndef SYMBOL_NAME
-#ifdef CONFIG_SYMBOL_PREFIX
-#define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
-#else
-#define SYMBOL_NAME(x) x
-#endif
+/* This helps us to avoid #ifdef SYMBOL_PREFIX */
+#ifndef SYMBOL_PREFIX
+#define SYMBOL_PREFIX
 #endif
-#define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))
+
+#define __SYMBOL_NAME(x) __stringify(SYMBOL_PREFIX) #x
 
 #ifndef cond_syscall
 #define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
index 07125e6..f1cce6a 100644
--- a/scripts/Makefile.lib
+++ b/scripts/Makefile.lib
@@ -123,6 +123,7 @@ ifdef CONFIG_SYMBOL_PREFIX
 _sym_flags = -DSYMBOL_PREFIX=$(patsubst "%",%,$(CONFIG_SYMBOL_PREFIX))
 _cpp_flags += $(_sym_flags)
 _a_flags += $(_sym_flags)
+_c_flags += $(_sym_flags)
 endif
 
 
-- 
1.8.1.2


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/