Re: evdev-induced panic (devfs / destroy_dev race?)

On Thu, Oct 23, 2025 at 07:41:48AM -0500, Kyle Evans wrote:
> Hi,
> 
> Not sure if anyone else has noticed this, but I seem to have scared up an evdev panic:
> 
> Fatal trap 9: general protection fault while in kernel mode
> cpuid = 1; apic id = 01
> instruction pointer = 0x20:0xffffffff80b98b5b
> stack pointer           = 0x28:0xfffffe01a62b4b60
> frame pointer           = 0x28:0xfffffe01a62b4c00
> code segment        = base 0x0, limit 0xfffff, type 0x1b
>             = DPL 0, pres 1, long 1, def32 0, gran 1
> processor eflags    = interrupt enabled, resume, IOPL = 0
> current process     = 7339 (moused)
> rdi: fffff80001b55128 rsi: 0000000000000001 rdx: 0000000000000000
> rcx: fffffe0186e5b570  r8: fffff80353686cd0  r9: fffffe01a62b5000
> rax: fffff800027d7780 rbx: fffff80001b55128 rbp: fffffe01a62b4c00
> r10: 0000000000002af8 r11: 000000000000298c r12: fffff80353686780
> r13: deadc0dedeadc0c0 r14: 0000000000000001 r15: 0000000000000000
> trap number     = 9
> panic: general protection fault
> cpuid = 1
> time = 1761098215
> KDB: stack backtrace:
> db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe01a62b48e0
> vpanic() at vpanic+0x136/frame 0xfffffe01a62b4a10
> panic() at panic+0x43/frame 0xfffffe01a62b4a70
> trap_fatal() at trap_fatal+0x68/frame 0xfffffe01a62b4a90
> calltrap() at calltrap+0x8/frame 0xfffffe01a62b4a90
> --- trap 0x9, rip = 0xffffffff80b98b5b, rsp = 0xfffffe01a62b4b60, rbp = 0xfffffe01a62b4c00 ---
> _sx_xlock_hard() at _sx_xlock_hard+0x18b/frame 0xfffffe01a62b4c00
> _sx_xlock() at _sx_xlock+0xac/frame 0xfffffe01a62b4c40
> evdev_dtor() at evdev_dtor+0x5c/frame 0xfffffe01a62b4c70
> devfs_destroy_cdevpriv() at devfs_destroy_cdevpriv+0xab/frame 0xfffffe01a62b4c90
> devfs_close_f() at devfs_close_f+0x63/frame 0xfffffe01a62b4cc0
> _fdrop() at _fdrop+0x1a/frame 0xfffffe01a62b4ce0
> closef() at closef+0x1e3/frame 0xfffffe01a62b4d70
> closefp_impl() at closefp_impl+0x71/frame 0xfffffe01a62b4db0
> closefp_hl() at closefp_hl+0x70/frame 0xfffffe01a62b4e00
> amd64_syscall() at amd64_syscall+0x169/frame 0xfffffe01a62b4f30
> fast_syscall_common() at fast_syscall_common+0xf8/frame 0xfffffe01a62b4f30
> --- syscall (6, FreeBSD ELF64, close), rip = 0x2b5f166d5eaa, rsp = 0x2b5f133b49c8, rbp = 0x2b5f133b4ab0 ---
> KDB: enter: panic
> 
> This was seemingly the result of removing my mouse's USB dongle.  Presumably detach revoked the client and
> woke it up, which then triggered the above close() from moused to race with destroy_dev() for invoking the
> cdevpriv dtor.

Do you have core dump?

> 
> I spent a few minutes thinking about it and didn't really come to a good idea of what the fix might be,
> though I suspect there's nothing evdev can really do about it at the moment and we might need to somehow
> coordinate this in destroy_dev().

destroy_dev() already does the following;
- clears si_devsw so no new thread could enter devfs cdev methods
- drain until all threads in cdev methods return from them
- destruct cdevprivs

At this point no new cdevpriv structures should be created because no new
threads are allowed into cdev methods.  I would say that there is something
else causing the problem, not a lack of coordination with destroy_dev().