* Re: ntfs.ko needs unknown symbol end_iomem (under uml)
[not found] ` <1098367622.10371.13.camel@imp.csi.cam.ac.uk>
@ 2004-10-21 14:44 ` Yura Pakhuchiy
0 siblings, 0 replies; only message in thread
From: Yura Pakhuchiy @ 2004-10-21 14:44 UTC (permalink / raw)
To: linux-kernel; +Cc: Anton Altaparmakov
Hi,
Anton Altaparmakov wrote:
> On Thu, 2004-10-21 at 14:53, Yura Pakhuchiy wrote:
>
>>Hi Anton,
>>
>>When I compile ntfs as module for UML I receive during build:
>>*** Warning: "end_iomem" [fs/ntfs/ntfs.ko] undefined!
>>
>>It's ntfs bug or uml bug? Or am I doing something wrong?
>
>
> UML is broken. It defines VMALLOC_START to ((end_iomem +
> VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)) but it clearly does not export
> end_iomem to modules which means modules cannot use VMALLOC_START.
>
> But ntfs uses VMALLOC_START to determine if a pointer is in kmalloc()-ed
> memory or vmalloc()-ed memory so if you want to build it as a module you
> need to fix uml to export end_iomem, i.e. by for example adding:
>
> EXPORT_SYMBOL(end_iomem);
>
> after the end_iomem definition in arch/um/kernel/um_arch.c and the
> reconfiguring, recbuilding the kernel.
>
> You may wish to report this to LKML / the UML maintainer (after you have
> made sure that this does actually fix it).
>
> Best regards,
>
> Anton
I didn't find UML maintainer in MAINTAINERS, so I send this to
linux-kernel@vger.kernel.org.
I added line that Anton suggested and it's fixed above bug. Patch below.
Best regards,
Yura
--- ntfs-2.6-devel/arch/um/kernel/um_arch.c 2004-10-19 20:50:05.000000000 +0300
+++ ntfs-2.6-yura/arch/um/kernel/um_arch.c 2004-10-21 17:14:18.000000000 +0300
@@ -300,6 +300,7 @@ static void __init uml_postsetup(void)
/* Set during early boot */
unsigned long brk_start;
unsigned long end_iomem;
+EXPORT_SYMBOL(end_iomem);
#define MIN_VMALLOC (32 * 1024 * 1024)
^ permalink raw reply [flat|nested] only message in thread