I get the following error when I try to compile 2.5.8
init/main.o: In function `start_kernel':
init/main.o(.text.init+0x5e2): undefined reference to
`setup_per_cpu_areas'
Looking at the code it looks like someone got confused ;)
around the ifdefs.I'm under the assumption that setup_per_cpu_areas()
does nothing on a uniprocessor. The patch compile fine on
my PC.
--- init/main.c Tue Apr 16 00:33:04 2002
+++ init/main.c_new Tue Apr 16 00:32:51 2002
@@ -344,7 +344,9 @@
lock_kernel();
printk(linux_banner);
setup_arch(&command_line);
+#ifdef CONFIG_SMP
setup_per_cpu_areas();
+#endif
printk("Kernel command line: %s\n", saved_command_line);
parse_options(command_line);
trap_init();
Cheers
-Bongani
On Mon, 2002-04-15 at 18:53, Bongani wrote:
> I get the following error when I try to compile 2.5.8
> init/main.o: In function `start_kernel':
> init/main.o(.text.init+0x5e2): undefined reference to
> `setup_per_cpu_areas'
>
> Looking at the code it looks like someone got confused ;)
> around the ifdefs.I'm under the assumption that setup_per_cpu_areas()
> does nothing on a uniprocessor. The patch compile fine on
> my PC.
A better approach would probably be to define setup_per_cpu_areas to
nothing when CONFIG_SMP is not set so as not to have #ifdefs in the code
itself ... for example,
diff -urN linux-2.5.8/init/main.c linux/init/main.c
--- linux-2.5.8/init/main.c Sun Apr 14 15:18:46 2002
+++ linux/init/main.c Mon Apr 15 18:41:54 2002
@@ -272,6 +272,8 @@
#define smp_init() do { } while (0)
#endif
+#define setup_per_cpu_areas() do { } while(0)
+
#else
#ifdef __GENERIC_PER_CPU
Robert Love
On Tue, 2002-04-16 at 00:45, Robert Love wrote:
> On Mon, 2002-04-15 at 18:53, Bongani wrote:
> > I get the following error when I try to compile 2.5.8
> > init/main.o: In function `start_kernel':
> > init/main.o(.text.init+0x5e2): undefined reference to
> > `setup_per_cpu_areas'
> >
> > Looking at the code it looks like someone got confused ;)
> > around the ifdefs.I'm under the assumption that setup_per_cpu_areas()
> > does nothing on a uniprocessor. The patch compile fine on
> > my PC.
>
> A better approach would probably be to define setup_per_cpu_areas to
> nothing when CONFIG_SMP is not set so as not to have #ifdefs in the code
> itself ... for example,
>
> diff -urN linux-2.5.8/init/main.c linux/init/main.c
> --- linux-2.5.8/init/main.c Sun Apr 14 15:18:46 2002
> +++ linux/init/main.c Mon Apr 15 18:41:54 2002
> @@ -272,6 +272,8 @@
> #define smp_init() do { } while (0)
> #endif
>
> +#define setup_per_cpu_areas() do { } while(0)
> +
> #else
>
> #ifdef __GENERIC_PER_CPU
>
Does this also look cleaner ?
--- init/main.c Tue Apr 16 01:31:29 2002
+++ init/main.c_new Tue Apr 16 01:30:13 2002
@@ -272,6 +272,8 @@
#define smp_init() do { } while (0)
#endif
+#define setup_per_cpu_areas() do { } while(0)
+
#else
#ifdef __GENERIC_PER_CPU
@@ -297,9 +299,9 @@
}
}
#else
-static inline void setup_per_cpu_areas(void)
-{
-}
+
+#define setup_per_cpu_areas() do { } while(0)
+
#endif /* !__GENERIC_PER_CPU */
/* Called by boot processor to activate the rest. */
On Mon, 2002-04-15 at 19:34, Bongani wrote:
> Does this also look cleaner ?
> -static inline void setup_per_cpu_areas(void)
> -{
> -}
> +
> +#define setup_per_cpu_areas() do { } while(0)
> +
Personally yes, but others would disagree.
In fact, if we use a define setup_per_cpu_areas can not be used outside
of this compilation unit. Right now this looks to be the case, but if
something other than init/main.c wanted to use setup_per_cpu_areas we
would need to make the code an actual function or put the define in a
header file.
Since either case should optimize away, maybe we should make it a static
inline in both cases, since that is the authors original preference ...
Robert Love
On Tue, 2002-04-16 at 01:24, Robert Love wrote:
> On Mon, 2002-04-15 at 19:34, Bongani wrote:
>
> > Does this also look cleaner ?
>
> > -static inline void setup_per_cpu_areas(void)
> > -{
> > -}
> > +
> > +#define setup_per_cpu_areas() do { } while(0)
> > +
>
> Personally yes, but others would disagree.
>
> In fact, if we use a define setup_per_cpu_areas can not be used outside
> of this compilation unit. Right now this looks to be the case, but if
> something other than init/main.c wanted to use setup_per_cpu_areas we
> would need to make the code an actual function or put the define in a
> header file.
>
> Since either case should optimize away, maybe we should make it a static
> inline in both cases, since that is the authors original preference ...
>
> Robert Love
>
Ok, so this patch should be fine then
--- init/main.c Tue Apr 16 06:52:20 2002
+++ init/main.c_new Tue Apr 16 06:55:17 2002
@@ -272,6 +272,10 @@
#define smp_init() do { } while (0)
#endif
+static inline void setup_per_cpu_areas(void)
+{
+}
+
#else
#ifdef __GENERIC_PER_CPU
@@ -297,9 +301,11 @@
}
}
#else
+
static inline void setup_per_cpu_areas(void)
{
}
+
#endif /* !__GENERIC_PER_CPU */
/* Called by boot processor to activate the rest. */
On Tue, 2002-04-16 at 00:57, Bongani wrote:
> Ok, so this patch should be fine then
I likee.
> --- init/main.c Tue Apr 16 06:52:20 2002
> +++ init/main.c_new Tue Apr 16 06:55:17 2002
> @@ -272,6 +272,10 @@
> #define smp_init() do { } while (0)
> #endif
>
> +static inline void setup_per_cpu_areas(void)
> +{
> +}
> +
> #else
>
> #ifdef __GENERIC_PER_CPU
> @@ -297,9 +301,11 @@
> }
> }
> #else
> +
> static inline void setup_per_cpu_areas(void)
> {
> }
> +
> #endif /* !__GENERIC_PER_CPU */
>
> /* Called by boot processor to activate the rest. */
Robert Love