Re: [PATCH v4] thunderbolt: Fix tb->lock deadlock during hot-unplug on AMD USB4 routers

[Mario Limonciello <mario.limonciello@amd.com>]

On 8/31/26 08:06, Mika Westerberg wrote:
> Hi,
> 
> On Mon, Aug 31, 2026 at 07:55:16AM -0500, Mario Limonciello wrote:
>>>> +static void tb_domain_reset_interface(struct tb *tb)
>>>> +{
>>>> +	guard(mutex)(&tb->lock);
>>>> +	__tb_domain_reset_interface_locked(tb);
>>>> +}
>>>> +
>>>>    /**
>>>>     * tb_domain_disconnect_xdomain_paths() - Disable DMA paths for XDomain
>>>>     * @tb: Domain disabling the DMA paths
>>>> @@ -835,7 +849,7 @@ int tb_domain_disconnect_xdomain_paths(struct tb *tb, struct tb_xdomain *xd,
>>>>    	if (ret)
>>>>    		return ret;
>>>> -	if (tb->nhi->quirks & QUIRK_RESET_DMA_ON_TEARDOWN)
>>>> +	if (!xd->is_unplugged)
>>>
>>> If I read this right, if you unplug a tree (say a router and then after
>>> that router there is inter-domain link) with this check the reset does not
>>> happen and AMD system still hangs?
>>>
>>
>> It should be a different problem.  The quirk fixed a problem in the USB4
>> router (router wasn't functional when problem occurred).
>>
>> This fixes a deadlock in the kernel that system gets stuck while shutting
>> down.
> 
> Right, but now since this checks !xd->is_unplugged and only then does the
> reset, if you unplug the whole chain including the XDomain connection (i.e.
> the unplug happens say on host router downstream port) then we mark
> everything below that with is_unplugged=1 so now this function will skip
> the quirk and the DMA will get stuck, no?
> 
> Because..
> 
>>
>>>>    		tb_domain_reset_interface(tb);
>>>>    	return 0;
>>>> diff --git a/drivers/thunderbolt/tb.c b/drivers/thunderbolt/tb.c
>>>> index b7cc6894a598..89dfb3381345 100644
>>>> --- a/drivers/thunderbolt/tb.c
>>>> +++ b/drivers/thunderbolt/tb.c
>>>> @@ -2489,6 +2489,7 @@ static void tb_handle_hotplug(struct work_struct *work)
>>>>    			tb_xdomain_remove(xd);
>>>>    			port->xdomain = NULL;
>>>>    			__tb_disconnect_xdomain_paths(tb, xd, -1, -1, -1, -1);
>>>> +			__tb_domain_reset_interface_locked(tb);
> 
> ... this is not called when the unplug happens elsewhere (like before the
> XDomain link).

Ah yeah.

Juan - take a closer look at this, I think you're gonna have to spin it 
again.