On Thu, 11 Apr 2024 at 23:05, Guenter Roeck linux@roeck-us.net wrote:
On Thu, Apr 11, 2024 at 06:00:25PM +0300, Dan Carpenter wrote:
On Thu, Apr 11, 2024 at 08:20:55PM +0530, Naresh Kamboju wrote:
I use to notice kernel panic while running kunit tests now I have noticed this
Unable to handle kernel paging request at virtual address KASAN: null-ptr-deref in range pc : kunit_test_null_dereference (lib/kunit/kunit-test.c:119) lr : kunit_generic_run_threadfn_adapter (lib/kunit/try-catch.c:31)
The kunit tests run to completion and the system is stable. Kernel did not panic.
[ Snip ]
<0>[ 76.808597] Internal error: Oops: 0000000096000005 [#1] PREEMPT SMP <4>[ 76.809876] Modules linked in: <4>[ 76.812055] CPU: 1 PID: 567 Comm: kunit_try_catch Tainted: G B N 6.9.0-rc3-next-20240410 #1 <4>[ 76.812987] Hardware name: linux,dummy-virt (DT) <4>[ 76.814123] pstate: 12400009 (nzcV daif +PAN -UAO +TCO -DIT -SSBS BTYPE=--) <4>[ 76.814947] pc : kunit_test_null_dereference (lib/kunit/kunit-test.c:119)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
This is a new intentional NULL dereferencer that was added yesterday.
Maybe these should have a big printk, "Intentional NULL dereference coming up!\n".
Can the backtrace be suppressed, similar to the warnings suppression I am working on ?
I'd like to do that going forward. Of course, this isn't a warning, so it doesn't work as-is (and it'd be harder to pass things like the function name through), but it seems like a worthwhile feature to have going forward. We did have some similar stuff for trapping KASAN errors as a part of the KASAN tests a while ago: that's also something to look at.
I have been playing with the warning suppression with the fortify test, and that seems to be working well.
Cheers, -- David