Re: 3.0.1 compile error on Fedora 15

[Randy Dunlap <rdunlap@xenotime.net>]

On Mon, 08 Aug 2011 11:55:29 -0700 Ben Greear wrote:

> On 08/08/2011 11:52 AM, Randy Dunlap wrote:
> > On Mon, 08 Aug 2011 11:43:43 -0700 Ben Greear wrote:
> >
> >> [greearb@t60-ben linux-3.0.p4s]$ make
> >> make -C /home/greearb/git/linux-3.0.dev.y O=/home/greearb/kernel/2.6/linux-3.0.p4s/.
> >>     Using /home/greearb/git/linux-3.0.dev.y as source for kernel
> >>     GEN     /home/greearb/kernel/2.6/linux-3.0.p4s/Makefile
> >>     CHK     include/linux/version.h
> >>     CHK     include/generated/utsrelease.h
> >>     CALL    /home/greearb/git/linux-3.0.dev.y/scripts/checksyscalls.sh
> >>     CHK     include/generated/compile.h
> >>     CC      fs/binfmt_misc.o
> >> In file included from /home/greearb/git/linux-3.0.dev.y/arch/x86/include/asm/uaccess.h:570:0,
> >>                    from /home/greearb/git/linux-3.0.dev.y/include/linux/uaccess.h:5,
> >>                    from /home/greearb/git/linux-3.0.dev.y/include/linux/highmem.h:7,
> >>                    from /home/greearb/git/linux-3.0.dev.y/include/linux/pagemap.h:10,
> >>                    from /home/greearb/git/linux-3.0.dev.y/fs/binfmt_misc.c:26:
> >> /home/greearb/git/linux-3.0.dev.y/arch/x86/include/asm/uaccess_32.h: In function ‘parse_command.part.0’:
> >> /home/greearb/git/linux-3.0.dev.y/arch/x86/include/asm/uaccess_32.h:211:26: error: call to ‘copy_from_user_overflow’ declared with attribute error:
> >> copy_from_user() buffer size is not provably correct
> >> make[3]: *** [fs/binfmt_misc.o] Error 1
> >> make[2]: *** [fs] Error 2
> >> make[1]: *** [sub-make] Error 2
> >> make: *** [all] Error 2
> >> [greearb@t60-ben linux-3.0.p4s]$
> >>
> >>
> >> The code in question is below.  Is this missing an assignment to n
> >> in the else branch?
> >>
> >> static inline unsigned long __must_check copy_from_user(void *to,
> >>                                             const void __user *from,
> >>                                             unsigned long n)
> >> {
> >>           int sz = __compiletime_object_size(to);
> >>
> >>           if (likely(sz == -1 || sz>= n))
> >>                   n = _copy_from_user(to, from, n);
> >>           else
> >>                   copy_from_user_overflow();
> >>
> >>           return n;
> >> }
> >
> >
> > Do you have this kconfig option enabled?
> > CONFIG_DEBUG_STRICT_USER_COPY_CHECKS=y
> >
> > If so, you should get the error that you reported.
> > If not, then you need a patch like the one below (which has been posted
> > by other people several times -- might need to be added to -stable):
> 
> I do have that enabled.   I'm slightly confused by your response above:
> 
> Do I need to disable the check AND apply the patch below, or just disable
> the check, or just apply the patch and leave the check enabled?

Hm, I guess I'm confused also.  The patch only applies to linux-next,
not to 3.0.x.

The patch below is needed (in linux-next) to fix a build
error when CONFIG_DEBUG_STRICT_USER_COPY_CHECKS is not enabled.

When you do have CONFIG_DEBUG_STRICT_USER_COPY_CHECKS, the build error is
the correct result.  AFAIK.  So it looks to me like you should either
disable the kconfig option or fix the code in fs/binfmt_misc.c.
I would do the former -- depending on your use case.


> Thanks,
> Ben
> 
> >
> >
> > ---
> >   lib/Makefile |    2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > --- linux-next-20110728.orig/lib/Makefile
> > +++ linux-next-20110728/lib/Makefile
> > @@ -14,7 +14,7 @@ lib-y := ctype.o string.o vsprintf.o cmd
> >   	 proportions.o prio_heap.o ratelimit.o show_mem.o \
> >   	 is_single_threaded.o plist.o decompress.o find_next_bit.o
> >
> > -lib-$(CONFIG_ARCH_HAS_DEBUG_STRICT_USER_COPY_CHECKS) += usercopy.o
> > +obj-$(CONFIG_ARCH_HAS_DEBUG_STRICT_USER_COPY_CHECKS) += usercopy.o
> >   lib-$(CONFIG_MMU) += ioremap.o
> >   lib-$(CONFIG_SMP) += cpumask.o
> >
> 
> 
> -- 


---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***