linux-kernel - Re: [PATCH] drm/panthor: Fix kernel panic on partial unmap of a GPU VA region

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20251017144836.2746973b@fedora>
Date: Fri, 17 Oct 2025 14:48:36 +0200
From: Boris Brezillon <boris.brezillon@...labora.com>
To: Liviu Dudau <liviu.dudau@....com>
Cc: Akash Goel <akash.goel@....com>, steven.price@....com,
 dri-devel@...ts.freedesktop.org, linux-kernel@...r.kernel.org,
 maarten.lankhorst@...ux.intel.com, mripard@...nel.org, tzimmermann@...e.de,
 airlied@...il.com, daniel@...ll.ch, nd@....com
Subject: Re: [PATCH] drm/panthor: Fix kernel panic on partial unmap of a GPU
 VA region

On Fri, 17 Oct 2025 13:30:53 +0100
Liviu Dudau <liviu.dudau@....com> wrote:

> On Fri, Oct 17, 2025 at 11:29:22AM +0100, Akash Goel wrote:
> > This commit address a kernel panic issue that can happen if Userspace
> > tries to partially unmap a GPU virtual region (aka drm_gpuva).
> > The VM_BIND interface allows partial unmapping of a BO.
> > 
> > Panthor driver pre-allocates memory for the new drm_gpuva structures
> > that would be needed for the map/unmap operation, done using drm_gpuvm
> > layer. It expected that only one new drm_gpuva would be needed on umap
> > but a partial unmap can require 2 new drm_gpuva and that's why it
> > ended up doing a NULL pointer dereference causing a kernel panic.
> > 
> > Following dump was seen when partial unmap was exercised.
> >  Unable to handle kernel NULL pointer dereference at virtual address 0000000000000078
> >  Mem abort info:
> >    ESR = 0x0000000096000046
> >    EC = 0x25: DABT (current EL), IL = 32 bits
> >    SET = 0, FnV = 0
> >    EA = 0, S1PTW = 0
> >    FSC = 0x06: level 2 translation fault
> >  Data abort info:
> >    ISV = 0, ISS = 0x00000046, ISS2 = 0x00000000
> >    CM = 0, WnR = 1, TnD = 0, TagAccess = 0
> >    GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
> >  user pgtable: 4k pages, 48-bit VAs, pgdp=000000088a863000
> >  [000000000000078] pgd=080000088a842003, p4d=080000088a842003, pud=0800000884bf5003, pmd=0000000000000000
> >  Internal error: Oops: 0000000096000046 [#1] PREEMPT SMP
> >  <snip>
> >  pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> >  pc : panthor_gpuva_sm_step_remap+0xe4/0x330 [panthor]
> >  lr : panthor_gpuva_sm_step_remap+0x6c/0x330 [panthor]
> >  sp : ffff800085d43970
> >  x29: ffff800085d43970 x28: ffff00080363e440 x27: ffff0008090c6000
> >  x26: 0000000000000030 x25: ffff800085d439f8 x24: ffff00080d402000
> >  x23: ffff800085d43b60 x22: ffff800085d439e0 x21: ffff00080abdb180
> >  x20: 0000000000000000 x19: 0000000000000000 x18: 0000000000000010
> >  x17: 6e656c202c303030 x16: 3666666666646466 x15: 393d61766f69202c
> >  x14: 312d3d7361203a70 x13: 303030323d6e656c x12: ffff80008324bf58
> >  x11: 0000000000000003 x10: 0000000000000002 x9 : ffff8000801a6a9c
> >  x8 : ffff00080360b300 x7 : 0000000000000000 x6 : 000000088aa35fc7
> >  x5 : fff1000080000000 x4 : ffff8000842ddd30 x3 : 0000000000000001
> >  x2 : 0000000100000000 x1 : 0000000000000001 x0 : 0000000000000078
> >  Call trace:
> >   panthor_gpuva_sm_step_remap+0xe4/0x330 [panthor]
> >   op_remap_cb.isra.22+0x50/0x80
> >   __drm_gpuvm_sm_unmap+0x10c/0x1c8
> >   drm_gpuvm_sm_unmap+0x40/0x60
> >   panthor_vm_exec_op+0xb4/0x3d0 [panthor]
> >   panthor_vm_bind_exec_sync_op+0x154/0x278 [panthor]
> >   panthor_ioctl_vm_bind+0x160/0x4a0 [panthor]
> >   drm_ioctl_kernel+0xbc/0x138
> >   drm_ioctl+0x240/0x500
> >   __arm64_sys_ioctl+0xb0/0xf8
> >   invoke_syscall+0x4c/0x110
> >   el0_svc_common.constprop.1+0x98/0xf8
> >   do_el0_svc+0x24/0x38
> >   el0_svc+0x40/0xf8
> >   el0t_64_sync_handler+0xa0/0xc8
> >   el0t_64_sync+0x174/0x178
> > 
> > Signed-off-by: Akash Goel <akash.goel@....com>  
> 
> Do we need a Fixes tag here?

We definitely want a Fixes tag, yep. Thanks for spotting that.

> 
> Reviewed-by: Liviu Dudau <liviu.dudau@....com>
> 
> Best regards,
> Liviu
> 
> > ---
> >  drivers/gpu/drm/panthor/panthor_mmu.c | 10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> > index 6dec4354e378..7870e7dbaa5d 100644
> > --- a/drivers/gpu/drm/panthor/panthor_mmu.c
> > +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
> > @@ -1175,10 +1175,14 @@ panthor_vm_op_ctx_prealloc_vmas(struct panthor_vm_op_ctx *op_ctx)
> >  		break;
> >  
> >  	case DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP:
> > -		/* Partial unmaps might trigger a remap with either a prev or a next VA,
> > -		 * but not both.
> > +		/* Two VMAs can be needed for an unmap, as an unmap can happen
> > +		 * in the middle of a drm_gpuva, requiring a remap with both
> > +		 * prev & next VA. Or an unmap can span more than one drm_gpuva
> > +		 * where the first and last ones are covered partially, requring
> > +		 * a remap for the first with a prev VA and remap for the last
> > +		 * with a next VA.
> >  		 */
> > -		vma_count = 1;
> > +		vma_count = 2;
> >  		break;
> >  
> >  	default:
> > -- 
> > 2.25.1
> >   
>