Re: [PATCH bpf-next v1 12/13] bpf: lsm: Add selftests for BPF_PROG_TYPE_LSM

[Andrii Nakryiko <andrii.nakryiko@gmail.com>]

On Fri, Jan 3, 2020 at 4:09 PM KP Singh <kpsingh@chromium.org> wrote:
>
> On 23-Dez 22:49, Andrii Nakryiko wrote:
> > On Fri, Dec 20, 2019 at 7:42 AM KP Singh <kpsingh@chromium.org> wrote:
> > >
> > > From: KP Singh <kpsingh@google.com>
> > >
> > > * Load a BPF program that audits mprotect calls
> > > * Attach the program to the "file_mprotect" LSM hook
> > > * Verify if the program is actually loading by reading
> > >   securityfs
> > > * Initialize the perf events buffer and poll for audit events
> > > * Do an mprotect on some memory allocated on the heap
> > > * Verify if the audit event was received
> > >
> > > Signed-off-by: KP Singh <kpsingh@google.com>
> > > ---
> > >  MAINTAINERS                                   |   2 +
> > >  .../bpf/prog_tests/lsm_mprotect_audit.c       | 129 ++++++++++++++++++
> > >  .../selftests/bpf/progs/lsm_mprotect_audit.c  |  58 ++++++++
> > >  3 files changed, 189 insertions(+)
> > >  create mode 100644 tools/testing/selftests/bpf/prog_tests/lsm_mprotect_audit.c
> > >  create mode 100644 tools/testing/selftests/bpf/progs/lsm_mprotect_audit.c
> > >
> >
> > [...]
> >
> > > +/*
> > > + * Define some of the structs used in the BPF program.
> > > + * Only the field names and their sizes need to be the
> > > + * same as the kernel type, the order is irrelevant.
> > > + */
> > > +struct mm_struct {
> > > +       unsigned long start_brk, brk, start_stack;
> > > +};
> > > +
> > > +struct vm_area_struct {
> > > +       unsigned long start_brk, brk, start_stack;
> > > +       unsigned long vm_start, vm_end;
> > > +       struct mm_struct *vm_mm;
> > > +       unsigned long vm_flags;
> > > +};
> > > +
> > > +BPF_TRACE_3("lsm/file_mprotect", mprotect_audit,
> > > +           struct vm_area_struct *, vma,
> > > +           unsigned long, reqprot, unsigned long, prot)
> > > +{
> > > +       struct mprotect_audit_log audit_log = {};
> > > +       int is_heap = 0;
> > > +
> > > +       __builtin_preserve_access_index(({
> >
> > you don't need __builtin_preserve_access_index, if you mark
> > vm_area_struct and mm_struct with
> > __attribute__((preserve_access_index)
>
> Cool, updated!
>
> >
> > > +               is_heap = (vma->vm_start >= vma->vm_mm->start_brk &&
> > > +                                    vma->vm_end <= vma->vm_mm->brk);
> > > +       }));
> > > +
> > > +       audit_log.magic = MPROTECT_AUDIT_MAGIC;
> > > +       audit_log.is_heap = is_heap;
> > > +       bpf_lsm_event_output(&perf_buf_map, BPF_F_CURRENT_CPU, &audit_log,
> > > +                            sizeof(audit_log));
> >
> > You test would be much simpler if you use global variables to pass
> > data back to userspace, instead of using perf buffer.
> >
> > Also please see fentry_fexit.c test for example of using BPF skeleton
> > to shorten and simpify userspace part of test.
>
> Thanks for the skeleton work!
>
> This makes using global variables easier and the tests are indeed much
> simpler, I have updated it for the next revision.
>
> One follow up question regarding global variables, let's say I have
> the following global variable defined in the BPF program:
>
> struct result_info {
>         __u32 count;
> };
>
> struct result_info result = {
>         .count = 0,
> };
>
> The defintion of result_info needs to be included before the .skel.h
> as it's not automatically generated or maybe I am missing a
> trick here?
>
> For now, I have defined this in a header which gets included both in
> the program and the test.

Yes, ideally all common types should be shared in a common header. My
initial implementation actually supported dumping out all the types in
a generated skeleton, but that was inconvenient in a lot of cases, so
I dropped that.

>
> - KP
>
> >
> > > +       return 0;
> > > +}
> > > --
> > > 2.20.1
> > >