linuxolator: amd64 Linux Test Project failures

[Scot Hetzel]

On 12/29/06, Divacky Roman <xdivac02 at stud.fit.vutbr.cz> wrote:
> On Fri, Dec 29, 2006 at 11:11:00AM -0600, Scot Hetzel wrote:
> > I'm seeing similar failures on these functions when running the
> > ltp-20061222 tests on amd64:
> >
> > linux_[access, chdir, chmod, chown, chroot, creat, lchown, link,
> > lstat, mkdir, mkdirat, mknod, rename, rmdir, stat, statfs, symlink,
> > truncate, unlink]
> >
> > The problem is that the ltp tests are passing in a negative value for
> > the path into the functions, and is expection a return value of
> > EFAULT.  Instead they are returning with (i.e access03):
> >
> > access((char *)-1,R_OK) failed with errno 2 : No such file or
> > directory but expected 14 (EFAULT)
> >
> > I added a printf before the LCONVPATHEXIST macro:
> >
> >        printf(ARGS(access, "Checking if path [%p/%p] exists"), (char
> > *) -1, &args->path);
> >        LCONVPATHEXIST(td, args->path, &path);
> >
> > and it resulted in the following output:
> >
> > Dec 29 03:32:32 hp010 kernel: linux(1259): access(Checking if path
> > [0xffffffffffffffff/0xffffffffa31f2be0] exists)
>
> forget my last mail.. the flu strikes badly..
> anyway - remove the & and post the results I think I know what might
> cause the bug....
>

I just added another %p to the format, here are my results with all of
the debugging printfs.

Dec 29 15:32:38 hp010 kernel: linux(1807): access(Checking if path
[0xffffffffffffffff/0xffffffffa3bc8be0/0x28069000] exists) <-
&args-path, args-path
Dec 29 15:32:38 hp010 kernel: kern_alternate_path: error = 0, path =
0xffffffffa3bc8980/0x28069000  <- value of error after the copyinstr
Dec 29 15:32:38 hp010 kernel: linux(1807): __func__(Path checked: error = 22)
Dec 29 15:32:38 hp010 kernel: linux(1807): access(Checking
kern_access: path = 0xffffffffa3bc8b18)

Scot
-- 
DISCLAIMER:
No electrons were mamed while sending this message. Only slightly bruised.