On 03/18/10 16:34, Renà Bolldorf wrote:On 03/18/10 23:58, Randy Dunlap wrote:On 03/18/10 15:08, Renï Bolldorf wrote:Signed-off-by: Rene Bolldorf<xsecute@xxxxxxxxxxxxxx>
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index ddd55dc..a1f7d93 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -2773,7 +2773,7 @@ int pci_resource_bar(struct pci_dev *dev, int
resno, enum pci_bar_type *type)
/* Some architectures require additional programming to enable VGA */
static arch_set_vga_state_t arch_set_vga_state;
-void __init pci_register_set_vga_state(arch_set_vga_state_t func)
+void pci_register_set_vga_state(arch_set_vga_state_t func)
{
arch_set_vga_state = func; /* NULL disables */
}
--
Hi,
Please put the offending/warning messages in the patch description.
Include your analysis if it isn't obvious.
thanks,
Hi,
the warning message was:
WARNING: drivers/pci/built-in.o(__ksymtab+0x190): Section mismatch in
reference from the variable __ksymtab_pci_register_set_vga_state to the
function .init.text:pci_register_set_vga_state()
The symbol pci_register_set_vga_state is exported and annotated __init
Fix this by removing the __init annotation of pci_register_set_vga_state
or drop the export.
$fgrep "pci_register_set_vga_state" `find .`
$./arch/x86/kernel/apic/x2apic_uv_x.c:
pci_register_set_vga_state(uv_set_vga_state);
$./include/linux/pci.h:extern void
pci_register_set_vga_state(arch_set_vga_state_t func);
$./drivers/pci/pci.c:void
pci_register_set_vga_state(arch_set_vga_state_t func)
$./drivers/pci/pci.c:EXPORT_SYMBOL(pci_register_set_vga_state);
this is a volatile function that set's a spec. state and is only used
from uv_system_init(arch/x86/kernel/apic/x2apic_uv_x.c), the macro can
also be removed.
Thanks. Yes, we shouldn't export __init functions, so the patch makes
sense.
BTW, you need to cc: one or more maintainers so that someone will merge
your patches. They don't scan the mailing lists looking for patches to merge.