Re: [PATCH v2] clk: Disable KUNIT_UML_PCI

[Brian Masney <bmasney@redhat.com>]

Hi Peng,

On Mon, Dec 22, 2025 at 03:06:33PM +0800, Peng Fan (OSS) wrote:
> From: Peng Fan <peng.fan@nxp.com>
> 
> commit 031cdd3bc3f3 ("kunit: Enable PCI on UML without triggering WARN()")
> enables KUNIT_UML_PCI, but clk driver could not work with it.
> Deselect KUNIT_UML_PCI to avoid the failure. Dump as below:
> WARNING: CPU: 0 PID: 227 at lib/logic_iomem.c:141 __raw_readl+0xac/0xe0
> CPU: 0 UID: 0 PID: 227 Comm: kunit_try_catch Tainted: G
> Tainted: [N]=TEST
> Stack:
>  a0883d00 00000001 00000000 ffffff00
>  603ef142 60044832 6002598b 00000000
>  00000000 600211b3 00000001 00000000
> Call Trace:
>  [<6032534c>] ? __raw_readl+0xac/0xe0
>  [<60044832>] ? dump_stack_lvl+0x57/0x73
>  [<6002598b>] ? _printk+0x0/0x61
>  [<600211b3>] ? __warn.cold+0x61/0xeb
>  [<600212cc>] ? warn_slowpath_fmt+0x8f/0x9c
>  [<6002123d>] ? warn_slowpath_fmt+0x0/0x9c
>  [<6032534c>] ? __raw_readl+0xac/0xe0
>  [<6002123d>] ? warn_slowpath_fmt+0x0/0x9c
>  [<6029e2ad>] ? clk_gate_endisable+0xcd/0x110
>  [<6029e315>] ? clk_gate_enable+0x15/0x20
>  [<6028795e>] ? clk_core_enable+0x6e/0xf0
>  [<60289f1f>] ? clk_enable+0x4f/0xa0
>  [<602a06af>] ? clk_gate_test_enable+0xbf/0x360
>  [<60053df9>] ? os_nsecs+0x29/0x40
>  [<600cd300>] ? ktime_get_ts64+0x0/0x130
>  [<600816c0>] ? to_kthread+0x0/0x50
>  [<602507bb>] ? kunit_try_run_case+0x7b/0x100
>  [<600816c0>] ? to_kthread+0x0/0x50
>  [<60252aa0>] ? kunit_generic_run_threadfn_adapter+0x0/0x30
>  [<60252ab2>] ? kunit_generic_run_threadfn_adapter+0x12/0x30
>  [<60082091>] ? kthread+0xf1/0x270
>  [<60047591>] ? new_thread_handler+0x41/0x60
> ---[ end trace 0000000000000000 ]---
> 
> Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
> Signed-off-by: Peng Fan <peng.fan@nxp.com>
> ---
> Changes in v2:
> - Drop fixes tag
> - Add R-b
> - Remove "CONFIG_UML_PCI_OVER_VIRTIO=n"
> ---
>  drivers/clk/.kunitconfig | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/clk/.kunitconfig b/drivers/clk/.kunitconfig
> index 08e26137f3d9..8a0ea41934a2 100644
> --- a/drivers/clk/.kunitconfig
> +++ b/drivers/clk/.kunitconfig
> @@ -1,4 +1,5 @@
>  CONFIG_KUNIT=y
> +CONFIG_KUNIT_UML_PCI=n
>  CONFIG_OF=y
>  CONFIG_OF_OVERLAY=y
>  CONFIG_COMMON_CLK=y
> @@ -6,4 +7,3 @@ CONFIG_CLK_KUNIT_TEST=y
>  CONFIG_CLK_FIXED_RATE_KUNIT_TEST=y
>  CONFIG_CLK_GATE_KUNIT_TEST=y
>  CONFIG_CLK_FD_KUNIT_TEST=y
> -CONFIG_UML_PCI_OVER_VIRTIO=n

I talked to Stephen at Linux Plumbers 1.5 weeks ago about this
particular issue. My recollection is that he felt that that the issue
ultimately should be fixed in UML / kunit. I'm not entirely sure how
to do that though. He said that the tests run properly when you run
them natively on arm64.

If I was going to take a stab at this, personally I would stub out
the writel as described at 
Documentation/dev-tools/kunit/api/functionredirection.rst.

So, add a KUNIT_STATIC_STUB_REDIRECT to clk_gate_readl() and
clk_gate_writel() in drivers/clk/clk-gate.c. The clk gate tests can
then provide their own mock readl/writel.

I don't know if that'll get accepted.

Brian