Re: [PATCH] s390/kvm: avoid hypfs error message

[Christian Borntraeger]

Am 20.06.22 um 11:25 schrieb Juergen Gross:
> On 20.06.22 11:19, Heiko Carstens wrote:
> > On Mon, Jun 20, 2022 at 09:18:37AM +0200, Christian Borntraeger wrote:
> > > Am 20.06.22 um 08:03 schrieb Juergen Gross:
> > > > Ping?
> > > >
> > > > On 07.06.22 14:33, Juergen Gross wrote:
> > > > > When booting under KVM the following error messages are issued:
> > > > >
> > > > > hypfs.7f5705: The hardware system does not support hypfs
> > > > > hypfs.7a79f0: Initialization of hypfs failed with rc=-61
> > > > >
> > > > > While being documented, they can easily be avoided by bailing out of
> > > > > hypfs_init() early in case of running as a KVM guest.
> > > > >
> > > > > Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
> > > > > ---
> > > > >    arch/s390/hypfs/inode.c | 3 +++
> > > > >    1 file changed, 3 insertions(+)
> > > > >
> > > > > diff --git a/arch/s390/hypfs/inode.c b/arch/s390/hypfs/inode.c
> > > > > index 5c97f48cea91..bdf078f3c641 100644
> > > > > --- a/arch/s390/hypfs/inode.c
> > > > > +++ b/arch/s390/hypfs/inode.c
> > > > > @@ -464,6 +464,9 @@ static int __init hypfs_init(void)
> > > > >    {
> > > > >        int rc;
> > > > > +    if (MACHINE_IS_KVM)
> > > > > +        return -ENODATA;
> > > > > +
> > > > >        hypfs_dbfs_init();
> > > > >        if (hypfs_diag_init()) {
> > >
> > > In case KVM implements hypfs this check would then be wrong.
> > > Question to people on CC/TO.
> > > Would it be an option to still check with KVM but avoid the error message.
> > > So basically changing hypfs_diag_init and fail_dbfs_exit to check
> > > for KVM on error?
> > > Or is this worse?
> >
> > I'd say just move the pr_err("Initialization of hypfs failed with...")
> > one label above to fail_hypfs_diag_exit. Then we still get the message
> > that the hardware system doesn't support hypfs, which seems to be
> > wanted, and the error message only appears for an error.
> >
> > Even though I personally dislike printing everything to the console
> > this seems to be what is/was preferred. So let's keep that.
>
> Works for me.
>
> Would you be fine with additionally:
>
> @@ __init int hypfs_diag_init(void)
>           int rc;
>
>           if (diag204_probe()) {
> -                pr_err("The hardware system does not support hypfs\n");
> +                pr_info("The hardware system does not support hypfs\n");
>                   return -ENODATA;
>           }
>
> As this not really an error.

Yes, makes sense.