On Wed, Jul 29, 2020 at 11:29:43PM +0200, Daniel Borkmann wrote: > On 7/29/20 7:28 PM, Song Liu wrote: > > On Tue, Jul 28, 2020 at 8:37 AM Jean-Philippe Brucker > > <jean-philippe@xxxxxxxxxx> wrote: > > > > > > When a tracing BPF program attempts to read memory without using the > > > bpf_probe_read() helper, the verifier marks the load instruction with > > > the BPF_PROBE_MEM flag. Since the arm64 JIT does not currently recognize > > > this flag it falls back to the interpreter. > > > > > > Add support for BPF_PROBE_MEM, by appending an exception table to the > > > BPF program. If the load instruction causes a data abort, the fixup > > > infrastructure finds the exception table and fixes up the fault, by > > > clearing the destination register and jumping over the faulting > > > instruction. > > > > > > To keep the compact exception table entry format, inspect the pc in > > > fixup_exception(). A more generic solution would add a "handler" field > > > to the table entry, like on x86 and s390. > > > > > > Signed-off-by: Jean-Philippe Brucker <jean-philippe@xxxxxxxxxx> > > > > This patch looks good to me. > > > > Acked-by: Song Liu <songliubraving@xxxxxx> > > +1, applied, thanks a lot! > > > It is possible to add a selftest for this? I thought about this a > > little bit, but > > didn't get a good idea. > > Why not adding a test_verifier.c test case which calls into bpf_get_current_task() > to fetch pointer to current and then read out some field via BPF_PROBE_MEM which > should then succeed on x86/s390x/arm64 but be skipped on the other archs? Jean-Philippe, > could you look into following up with such test case(s)? Sure I'll take a look. Ilya also added a selftests to trigger exceptions in https://lore.kernel.org/bpf/20200715233301.933201-5-iii@xxxxxxxxxxxxx/ It's useful but I think it relies on the verifier not mandating NULL checks for next-level pointers (they are ptr_ instead of ptr_or_null_), which might change in the future. So I'm wondering if we can deliberately access an invalid pointer with the help of bpf_test_run, and check that the result is zero. Thanks, Jean
