On Mon, 14 Jun 2021 at 10:59, Greg KH <greg@xxxxxxxxx> wrote:
> On Mon, Jun 14, 2021 at 10:54:00AM +0200, Andrew Zaborowski wrote:
> > Hi Greg,
> >
> > On Tue, 25 May 2021 at 13:53, Greg KH <greg@xxxxxxxxx> wrote:
> > > On Tue, May 25, 2021 at 01:36:27PM +0200, Andrew Zaborowski wrote:
> > > > From: Jarkko Sakkinen <jarkko@xxxxxxxxxx>
> > > >
> > > > BUG_ON() should not be used in the kernel code, unless there are
> > > > exceptional reasons to do so. Replace BUG_ON() with WARN() and
> > > > return.
> > > >
> > > > Cc: stable@xxxxxxxxxxxxxxx
> > > > Fixes: b3811d36a3e7 ("KEYS: checking the input id parameters before finding asymmetric key")
> > > > Signed-off-by: Jarkko Sakkinen <jarkko@xxxxxxxxxx>
> > > > ---
> > > > No changes from original submission by Jarkko.
> > > >
> > > > crypto/asymmetric_keys/asymmetric_type.c | 5 ++++-
> > > > 1 file changed, 4 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/crypto/asymmetric_keys/asymmetric_type.c b/crypto/asymmetric_keys/asymmetric_type.c
> > > > index ad8af3d70ac..a00bed3e04d 100644
> > > > --- a/crypto/asymmetric_keys/asymmetric_type.c
> > > > +++ b/crypto/asymmetric_keys/asymmetric_type.c
> > > > @@ -54,7 +54,10 @@ struct key *find_asymmetric_key(struct key *keyring,
> > > > char *req, *p;
> > > > int len;
> > > >
> > > > - BUG_ON(!id_0 && !id_1);
> > > > + if (!id_0 && !id_1) {
> > > > + WARN(1, "All ID's are NULL\n");
> > >
> > > You still just rebooted a machine (panic-on-warn is commonly set).
> > >
> > > Please just handle this properly, print an error message with dev_err()
> > > or pr_err() and move on, don't crash things.
> >
> > Like Eric Biggers said, a panic is probably what you want here since
> > this would be a basic bug, if you even want to check this.
>
> Ok, then keep the BUG_ON(), no change needed.
>
> > You can't
> > be looking for a key if you don't have any of the identifiers. There
> > a 4 current callers, 2 that have checks right before the call and 2
> > where this could be triggered by a bug in an ASN.1 parser or
> > corruption.
> >
> > What's the right way to get this change merged? There's clearly no
> > need to coordinate with whoever would merge 2/2 of the series.
>
> Why do you need to change this if BUG_ON() is the correct thing to do
> here?
At this function's level the situation is easily recoverable, I guess
the idea is to allow the kernel to continue in the "continue at any
cost" type scenario where you'd have panic_on_* set false.
Best regards
