Hi, Linus,

On Thu, 2025-12-04 at 09:35 -0800, Linus Torvalds wrote:
> On Tue, 2 Dec 2025 at 23:23, Dave Airlie <[email protected]> wrote:
> > 
> > This is the main drm pull request for 6.19.
> 
> I have pulled this and pushed out the result. It "works" for me on my
> Xe setup, but I do get a new refcount warning, which looks like
> something might be seriously wrong.
> 
> (Side note: I'm also running with
> 
>  
> https://lore.kernel.org/all/[email protected]/
> 
> fixed up for the merge and applied to make my dual 6k monitor setup
> happy).
> 
> I get this:
> 
>   refcount_t: underflow; use-after-free.
>   WARNING: lib/refcount.c:28 at 0x0, CPU#43: gst-plugin-scan/2770
>   Modules linked in: rfcomm sunrpc nf_conntrack_netbios_ns ...
>   CPU: 43 UID: 60578 PID: 2770 Comm: gst-plugin-scan Not tainted
> 6.18.0-06709-gab59a11f1240 #25 PREEMPTLAZY
>   Hardware name: Gigabyte Technology Co., Ltd. TRX40 AORUS
> MASTER/TRX40 AORUS MASTER, BIOS F7 09/07/2022
>   RIP: 0010:refcount_warn_saturate+0x42/0x70
>   ...
>   Call Trace:
>    xe_exec_queue_destroy+0x1be/0x1d0 [xe]
>    xe_vm_close_and_put+0x3b8/0x770 [xe]
>    xe_vm_destroy_ioctl+0x107/0x180 [xe]
>    drm_ioctl_kernel+0xa3/0xd0 [drm]
>    drm_ioctl+0x30f/0x470 [drm]
>    xe_drm_ioctl+0x48/0x70 [xe]
>    __se_sys_ioctl+0x6d/0xb0
>    do_syscall_64+0x6a/0x260
>    entry_SYSCALL_64_after_hwframe+0x4b/0x53
> 
> which hopefully gives somebody an idea.

Looks like a backmerge gone wrong and duplicating a put.

We'll pick that up and try to provide a fix ASAP.

Thanks,
Thomas


> 
> Added Xe people involved with recent commits to the participants.
> 
> Anybody?
> 
>              Linus

Reply via email to