Thread (4 messages) 4 messages, 2 authors, 2023-10-31

Re: Several kmemleak reports + "refcount_t: underflow; use-after-free" at boot when OF_UNITTEST + OF_OVERLAY is set (Kernel v6.6-rc6, PowerMac G5 11,2)

From: Rob Herring <robh@kernel.org>
Date: 2023-10-30 16:27:02
Also in: linux-devicetree

On Wed, Oct 18, 2023 at 4:38 PM Erhard Furtner [off-list ref] wrote:
Greetings!

Getting this at every boot on my G5 with kernel v6.6-rc6 with OF_UNITTEST and OF_OVERLAY selected:

[...]
### dt-test ### EXPECT \ : OF: ERROR: of_node_release() detected bad
of_node_put() on /testcase-data/refcount-node ### dt-test ### pass
of_unittest_lifecycle():3189 OF: ERROR: of_node_release() detected bad
of_node_put() on /testcase-data/refcount-node ### dt-test ### EXPECT / : OF:
ERROR: of_node_release() detected bad of_node_put() on
/testcase-data/refcount-node ### dt-test ### EXPECT \ : ------------[ cut here
]------------ ### dt-test ### EXPECT \ : WARNING: <<all>> ### dt-test ###
EXPECT \ : refcount_t: underflow; use-after-free. ### dt-test ### EXPECT \ :
The test tells you to expect a use-after-free...
---[ end trace <<int>> ]--- ### dt-test ### pass of_unittest_lifecycle():3209
------------[ cut here ]------------
refcount_t: underflow; use-after-free.
Then you get a use-after-free. Looks like it is working as designed.

I believe it's the same with kmemleak.

Note that running DT unittests also taints the kernel. That's because
they are not meant to be run on a production system.

Rob
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help