On Tue, Feb 8, 2022 at 10:44 PM Jeff Mahoney <jeffm(a)suse.com>
wrote:
>
> Hi Richard -
>
> On 5/19/21 16:00, Richard Guy Briggs wrote:
> > The openat2(2) syscall was added in kernel v5.6 with commit fddb5d430ad9
> > ("open: introduce openat2(2) syscall")
> >
> > Add the openat2(2) syscall to the audit syscall classifier.
> >
> > Link:
https://github.com/linux-audit/audit-kernel/issues/67
> > Signed-off-by: Richard Guy Briggs <rgb(a)redhat.com>
> > Link:
https://lore.kernel.org/r/f5f1a4d8699613f8c02ce762807228c841c2e26f.162136...
> > ---
>
> [...]
>
> > diff --git a/kernel/auditsc.c b/kernel/auditsc.c
> > index d775ea16505b..3f59ab209dfd 100644
> > --- a/kernel/auditsc.c
> > +++ b/kernel/auditsc.c
> > @@ -76,6 +76,7 @@
> > #include <linux/fsnotify_backend.h>
> > #include <uapi/linux/limits.h>
> > #include <uapi/linux/netfilter/nf_tables.h>
> > +#include <uapi/linux/openat2.h>
> >
> > #include "audit.h"
> >
> > @@ -196,6 +197,8 @@ static int audit_match_perm(struct audit_context *ctx, int
mask)
> > return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0]
== SYS_BIND);
> > case AUDITSC_EXECVE:
> > return mask & AUDIT_PERM_EXEC;
> > + case AUDITSC_OPENAT2:
> > + return mask & ACC_MODE((u32)((struct open_how
*)ctx->argv[2])->flags);
> > default:
> > return 0;
> > }
>
> ctx->argv[2] holds a userspace pointer and can't be dereferenced like this.
>
> I'm getting oopses, like so:
> BUG: unable to handle page fault for address: 00007fff961bbe70
Thanks Jeff.
Yes, this is obviously the wrong thing to being doing; I remember
checking to make sure we placed the audit_openat2_how() hook after the
open_how was copied from userspace, but I missed the argv dereference
in the syscall exit path when reviewing the code.
Richard, as we are already copying the open_how info into
audit_context::openat2 safely, the obvious fix is to convert
audit_match_perm() to use the previously copied value instead of argv.
If you can't submit a patch for this today please let me know.
I haven't heard anything from Richard so I put together a patch which
should fix the problem (link below). It's currently untested, but
I've got a kernel building now with the patch ...