We now have exports in both architecture code in in common code,
which causes a link failure when symbol versioning is eanbled, on
four architectures:
kernel/exit.o: In function `__crc_abort':
exit.c:(*ABS*+0xc0e2ec8b): multiple definition of `__crc_abort'
This removes the four architecture specific exports and only
leaves the export next to the __weak symbol.
Fixes: mmotm ("kernel/exit.c: export abort() to modules")
Signed-off-by: Arnd Bergmann <[email protected]>
---
Andrew, can you apply this to -mm on top of the other patch?
---
arch/arc/kernel/traps.c | 1 -
arch/arm/kernel/traps.c | 1 -
arch/m32r/kernel/traps.c | 1 -
arch/unicore32/kernel/traps.c | 1 -
4 files changed, 4 deletions(-)
diff --git a/arch/arc/kernel/traps.c b/arch/arc/kernel/traps.c
index 51a55b06cb2a..133a4dae41fe 100644
--- a/arch/arc/kernel/traps.c
+++ b/arch/arc/kernel/traps.c
@@ -169,4 +169,3 @@ void abort(void)
{
__asm__ __volatile__("trap_s 5\n");
}
-EXPORT_SYMBOL(abort);
diff --git a/arch/arm/kernel/traps.c b/arch/arm/kernel/traps.c
index e344bdd2e5ac..5e3633c24e63 100644
--- a/arch/arm/kernel/traps.c
+++ b/arch/arm/kernel/traps.c
@@ -793,7 +793,6 @@ void abort(void)
/* if that doesn't kill us, halt */
panic("Oops failed to kill thread");
}
-EXPORT_SYMBOL(abort);
void __init trap_init(void)
{
diff --git a/arch/m32r/kernel/traps.c b/arch/m32r/kernel/traps.c
index cb79fba79d43..b88a8dd14933 100644
--- a/arch/m32r/kernel/traps.c
+++ b/arch/m32r/kernel/traps.c
@@ -122,7 +122,6 @@ void abort(void)
/* if that doesn't kill us, halt */
panic("Oops failed to kill thread");
}
-EXPORT_SYMBOL(abort);
void __init trap_init(void)
{
diff --git a/arch/unicore32/kernel/traps.c b/arch/unicore32/kernel/traps.c
index 5f25b39f04d4..c4ac6043ebb0 100644
--- a/arch/unicore32/kernel/traps.c
+++ b/arch/unicore32/kernel/traps.c
@@ -298,7 +298,6 @@ void abort(void)
/* if that doesn't kill us, halt */
panic("Oops failed to kill thread");
}
-EXPORT_SYMBOL(abort);
void __init trap_init(void)
{
--
2.9.0
On Tue, Jan 02, 2018 at 11:31:12AM +0100, Arnd Bergmann wrote:
> We now have exports in both architecture code in in common code,
> which causes a link failure when symbol versioning is eanbled, on
> four architectures:
>
> kernel/exit.o: In function `__crc_abort':
> exit.c:(*ABS*+0xc0e2ec8b): multiple definition of `__crc_abort'
>
> This removes the four architecture specific exports and only
> leaves the export next to the __weak symbol.
>
> Fixes: mmotm ("kernel/exit.c: export abort() to modules")
> Signed-off-by: Arnd Bergmann <[email protected]>
For arm:
Acked-by: Russell King <[email protected]>
> ---
> Andrew, can you apply this to -mm on top of the other patch?
> ---
> arch/arc/kernel/traps.c | 1 -
> arch/arm/kernel/traps.c | 1 -
> arch/m32r/kernel/traps.c | 1 -
> arch/unicore32/kernel/traps.c | 1 -
> 4 files changed, 4 deletions(-)
>
> diff --git a/arch/arc/kernel/traps.c b/arch/arc/kernel/traps.c
> index 51a55b06cb2a..133a4dae41fe 100644
> --- a/arch/arc/kernel/traps.c
> +++ b/arch/arc/kernel/traps.c
> @@ -169,4 +169,3 @@ void abort(void)
> {
> __asm__ __volatile__("trap_s 5\n");
> }
> -EXPORT_SYMBOL(abort);
> diff --git a/arch/arm/kernel/traps.c b/arch/arm/kernel/traps.c
> index e344bdd2e5ac..5e3633c24e63 100644
> --- a/arch/arm/kernel/traps.c
> +++ b/arch/arm/kernel/traps.c
> @@ -793,7 +793,6 @@ void abort(void)
> /* if that doesn't kill us, halt */
> panic("Oops failed to kill thread");
> }
> -EXPORT_SYMBOL(abort);
>
> void __init trap_init(void)
> {
> diff --git a/arch/m32r/kernel/traps.c b/arch/m32r/kernel/traps.c
> index cb79fba79d43..b88a8dd14933 100644
> --- a/arch/m32r/kernel/traps.c
> +++ b/arch/m32r/kernel/traps.c
> @@ -122,7 +122,6 @@ void abort(void)
> /* if that doesn't kill us, halt */
> panic("Oops failed to kill thread");
> }
> -EXPORT_SYMBOL(abort);
>
> void __init trap_init(void)
> {
> diff --git a/arch/unicore32/kernel/traps.c b/arch/unicore32/kernel/traps.c
> index 5f25b39f04d4..c4ac6043ebb0 100644
> --- a/arch/unicore32/kernel/traps.c
> +++ b/arch/unicore32/kernel/traps.c
> @@ -298,7 +298,6 @@ void abort(void)
> /* if that doesn't kill us, halt */
> panic("Oops failed to kill thread");
> }
> -EXPORT_SYMBOL(abort);
>
> void __init trap_init(void)
> {
> --
> 2.9.0
>
--
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 8.8Mbps down 630kbps up
According to speedtest.net: 8.21Mbps down 510kbps up
Hi Arnd,
On Tue, Jan 02, 2018 at 11:31:12AM +0100, Arnd Bergmann wrote:
> We now have exports in both architecture code in in common code,
> which causes a link failure when symbol versioning is eanbled, on
> four architectures:
>
> kernel/exit.o: In function `__crc_abort':
> exit.c:(*ABS*+0xc0e2ec8b): multiple definition of `__crc_abort'
>
> This removes the four architecture specific exports and only
> leaves the export next to the __weak symbol.
Will it not be better to remove the abort() as they are exactly same
like the weak function. That was the original plan we had when this
weak function was done. Like in the attached patch..
--
Regards
Sudip
On Tue, Jan 02, 2018 at 11:34:45AM +0000, Sudip Mukherjee wrote:
> Hi Arnd,
>
> On Tue, Jan 02, 2018 at 11:31:12AM +0100, Arnd Bergmann wrote:
> > We now have exports in both architecture code in in common code,
> > which causes a link failure when symbol versioning is eanbled, on
> > four architectures:
> >
> > kernel/exit.o: In function `__crc_abort':
> > exit.c:(*ABS*+0xc0e2ec8b): multiple definition of `__crc_abort'
> >
> > This removes the four architecture specific exports and only
> > leaves the export next to the __weak symbol.
>
> Will it not be better to remove the abort() as they are exactly same
> like the weak function. That was the original plan we had when this
> weak function was done. Like in the attached patch..
If that was the plan, why wasn't it done - why did we end up with a
build-breaking patch merged in -mm?
--
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 8.8Mbps down 630kbps up
According to speedtest.net: 8.21Mbps down 510kbps up
On Tue, Jan 02, 2018 at 11:38:29AM +0000, Russell King - ARM Linux wrote:
> On Tue, Jan 02, 2018 at 11:34:45AM +0000, Sudip Mukherjee wrote:
> > Hi Arnd,
> >
> > On Tue, Jan 02, 2018 at 11:31:12AM +0100, Arnd Bergmann wrote:
> > > We now have exports in both architecture code in in common code,
> > > which causes a link failure when symbol versioning is eanbled, on
> > > four architectures:
> > >
> > > kernel/exit.o: In function `__crc_abort':
> > > exit.c:(*ABS*+0xc0e2ec8b): multiple definition of `__crc_abort'
> > >
> > > This removes the four architecture specific exports and only
> > > leaves the export next to the __weak symbol.
> >
> > Will it not be better to remove the abort() as they are exactly same
> > like the weak function. That was the original plan we had when this
> > weak function was done. Like in the attached patch..
>
> If that was the plan, why wasn't it done - why did we end up with a
> build-breaking patch merged in -mm?
The proposed patch (which I attached in the earlier mail) was breaking
the build unless weak abort() is exported. So, ideally the export of
weak abort() and the proposed patch should have been part of a series.
But afaik, the export has been done to fix another build failure reported
by Vineet.
Andrew, I think the best way now is to apply the patch sent by Arnd,
and I will send a separate patch to remove abort() after proper testing.
--
Regards
Sudip
On 01/02/2018 02:33 AM, Arnd Bergmann wrote:
> We now have exports in both architecture code in in common code,
> which causes a link failure when symbol versioning is eanbled, on
> four architectures:
>
> kernel/exit.o: In function `__crc_abort':
> exit.c:(*ABS*+0xc0e2ec8b): multiple definition of `__crc_abort'
>
> This removes the four architecture specific exports and only
> leaves the export next to the __weak symbol.
>
> Fixes: mmotm ("kernel/exit.c: export abort() to modules")
> Signed-off-by: Arnd Bergmann <[email protected]>
> ---
> Andrew, can you apply this to -mm on top of the other patch?
> ---
> arch/arc/kernel/traps.c | 1 -
> arch/arm/kernel/traps.c | 1 -
> arch/m32r/kernel/traps.c | 1 -
> arch/unicore32/kernel/traps.c | 1 -
> 4 files changed, 4 deletions(-)
>
> diff --git a/arch/arc/kernel/traps.c b/arch/arc/kernel/traps.c
> index 51a55b06cb2a..133a4dae41fe 100644
> --- a/arch/arc/kernel/traps.c
> +++ b/arch/arc/kernel/traps.c
> @@ -169,4 +169,3 @@ void abort(void)
> {
> __asm__ __volatile__("trap_s 5\n");
> }
> -EXPORT_SYMBOL(abort);
FWIW, this hunk did not yet hit mainline. I've removed it from my patch in
for-curr and re-pushed !
Thx,
-Vineet