Re: [PATCH] gss_krb5: Fix memleak in krb5_make_rc4_seq_num

From: Ard Biesheuvel
Date: Mon Aug 31 2020 - 11:27:54 EST


On Sat, 29 Aug 2020 at 18:43, J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote:
>
> This code is rarely if ever used, and there are pending patches to
> remove it completely, so I don't think it's worth trying to fix a rare
> memory leak at this point.
>
> --b.
>

FYI I just submitted v3 of my series removing this code to the
linux-crypto list, and so hopefully it will disappear in v5.10


> On Thu, Aug 27, 2020 at 04:02:50PM +0800, Dinghao Liu wrote:
> > When kmalloc() fails, cipher should be freed
> > just like when krb5_rc4_setup_seq_key() fails.
> >
> > Fixes: e7afe6c1d486b ("sunrpc: fix 4 more call sites that were using stack memory with a scatterlist")
> > Signed-off-by: Dinghao Liu <dinghao.liu@xxxxxxxxxx>
> > ---
> > net/sunrpc/auth_gss/gss_krb5_seqnum.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/sunrpc/auth_gss/gss_krb5_seqnum.c b/net/sunrpc/auth_gss/gss_krb5_seqnum.c
> > index 507105127095..88ca58d11082 100644
> > --- a/net/sunrpc/auth_gss/gss_krb5_seqnum.c
> > +++ b/net/sunrpc/auth_gss/gss_krb5_seqnum.c
> > @@ -53,8 +53,10 @@ krb5_make_rc4_seq_num(struct krb5_ctx *kctx, int direction, s32 seqnum,
> > return PTR_ERR(cipher);
> >
> > plain = kmalloc(8, GFP_NOFS);
> > - if (!plain)
> > - return -ENOMEM;
> > + if (!plain) {
> > + code = -ENOMEM;
> > + goto out;
> > + }
> >
> > plain[0] = (unsigned char) ((seqnum >> 24) & 0xff);
> > plain[1] = (unsigned char) ((seqnum >> 16) & 0xff);
> > --
> > 2.17.1