On Sun, Mar 08, 2026 at 12:10:08PM +0000, Justin Roberts wrote:
>
> Yes, that fixed things, with the caveat the first time I booted with the
> custom kernel, I had a hang the second time I ran vulkaninfo. However
> I've been unable to replicated since then despite cold boots and warm boots.
Here is a related trace to your photo. Encountered a few days ago
running piglit with a kernel based on not yet committed 6.18 drm.
At the moment I suspect a lock needs to be added to idr.
kernel: protection fault trap, code=0
Stopped at idr_tree_SPLAY+0xbc: cmpq 0x10(%r9),%r8
ddb{0}> tr
idr_tree_SPLAY(ffffffff82b94430,fffffd838b1ca1e0) at idr_tree_SPLAY+0xbc
idr_alloc(ffffffff82b94430,0,8000,10000,5) at idr_alloc+0xc1
amdgpu_pasid_alloc(10) at amdgpu_pasid_alloc+0x5d
amdgpu_driver_open_kms(ffff800000227058,ffff8000027e4200) at
amdgpu_driver_open_kms+0xac
drm_file_alloc(ffff8000001b44c0) at drm_file_alloc+0x251
drmopen(85700,3,2000,ffff80004aed17c0) at drmopen+0x12c
spec_open_clone(ffff80004ae81b18) at spec_open_clone+0x166
spec_open(ffff80004ae81b18) at spec_open+0x198
VOP_OPEN(fffffd839b504cb8,3,fffffd840d1f7dd8,ffff80004aed17c0) at VOP_OPEN+0x57
vn_open(ffff80004ae81d30,3,0) at vn_open+0x2ea
doopenat(ffff80004aed17c0,ffffff9c,759dd604a4f0,2,0,ffff80004ae81ed0) at doopen
at+0x221
syscall(ffff80004ae81f50) at syscall+0x5f9
Xsyscall() at Xsyscall+0x128
end of kernel
end trace frame: 0x759dd604a4e0, count: -13
rdi 0xffffffff82b94430 amdgpu_pasid_ida
rsi 0xfffffd838b1ca1e0
rbp 0xffff80004ae81880
rbx 0x10000 __ALIGN_SIZE+0xf000
rdx 0xfffffd838b1ca6c0
rcx 0xffff80004ae81850
rax 0xffff80004ae81850
r8 0x8000 __ALIGN_SIZE+0x7000
r9 0xd4b6bf2bc0cddc0d
r10 0xd4b6bf2bc0cddb2d
r11 0x4ac1077fb6b7ea7c
r12 0xfffffd838b1ca1e0
r13 0x8000 __ALIGN_SIZE+0x7000
r14 0x10000 __ALIGN_SIZE+0xf000
r15 0xffffffff82b94430 amdgpu_pasid_ida
rip 0xffffffff813e042c idr_tree_SPLAY+0xbc
cs 0x8
rflags 0x10282 __ALIGN_SIZE+0xf282
rsp 0xffff80004ae81850
ss 0x10
