Paul,
Looking at a snippet of the test output:
Service Provider read_consumed: 8
Service Provider command: BR_NOOP
Service Provider command: BR_FAILED_REPLY <-------------- the txn
failed as expected.
Manager read_consumed: 8
Manager command: BR_NOOP
Manager command: BR_TRANSACTION_COMPLETE
not ok 3
<--------- but for some reason didn't exit(103)
# Failed test at ./test line 56.
It looks like the test script expects test_binder to fail with
exit(103) after processing the Server Provider commands. It's not
clear why it didn't, since the return of a BR_FAILED_REPLY for that
transaction should have executed this code (line 392 of
test_binder.c):
if (cmd == BR_FAILED_REPLY ||
cmd == BR_DEAD_REPLY ||
cmd == BR_DEAD_BINDER) {
fprintf(stderr,
"Failed response from Service Provider using Managers FD\n");
exit(103);
}
Could this be an issue with the test? At least it doesn't look like a
transaction is succeeding when it shouldn't.
-Todd
On Tue, Mar 19, 2019 at 5:15 PM Todd Kjos <tkjos@xxxxxxxxxx> wrote:
>
> [...]
>
> > > Is there a public dashboard where I can take a look at those binder failures?
> >
> > Not really. I send test results to a not-yet-publicized mailing list,
> > but there is more detail in the GitHub issue below (my last comment
> > has the verbose test output):
> >
> > * https://github.com/SELinuxProject/selinux-kernel/issues/46
> >
>
> Ok, so it looks like something was introduced that causes binder to be
> too permissive (test 3 transaction succeeded when failure is
> expected). I don't know of any recent binder changes that could have
> caused that.
>
> It will take me a while to set up this test environment. Is this easy
> for you to run? Any chance of bisecting or at least trying a few
> versions to narrow it down? Here's a list of the recent patchset -- it
> would be useful to know which caused it (or if none of them did):
>
> 9e98c678c2d6a Linux 5.1-rc1
> ...
> 26528be6720bb binder: fix handling of misaligned binder object
> bde4a19fc04f5 binder: use userspace pointer as base of buffer space
> c41358a5f5217 binder: remove user_buffer_offset
> db6b0b810bf94 binder: avoid kernel vm_area for buffer fixups
> 7a67a39320dfb binder: add function to copy binder object from buffer
> 8ced0c6231ead binder: add functions to copy to/from binder buffers
> 1a7c3d9bb7a92 binder: create userspace-to-binder-buffer copy function
> ...
> 1c163f4c7b3f6 (tag: v5.0) Linux 5.0
>
> Thanks,
>
> -Todd
