[Intel-gfx] [PATCH] drm/i915: Rip out GET_SPRITE_COLORKEY ioctl
Ville Syrjälä
ville.syrjala at linux.intel.com
Fri Mar 27 10:55:30 PDT 2015
On Fri, Mar 27, 2015 at 07:40:43PM +0200, Tommi Rantala wrote:
> 2015-03-27 18:42 GMT+02:00 Jani Nikula <jani.nikula at linux.intel.com>:
> > On Fri, 27 Mar 2015, Daniel Vetter <daniel at ffwll.ch> wrote:
> >> On Fri, Mar 27, 2015 at 09:10:02AM +0100, Daniel Vetter wrote:
> >>> It's completely unused and Tommi noticed that the #define is borked
> >>> since forever. I've done a git search in userspace and only found
> >>> broken definitions and no users anywhere.
> >>>
> >>> Cc: Tommi Rantala <tt.rantala at gmail.com>
> >>> Signed-off-by: Daniel Vetter <daniel.vetter at intel.com>
> >>
> >> Hm Tommi discovered oopses in there, so I guess this should be
> >> cherry-picked to -fixes+cc: stable too? Jani?
> >
> > My OCD really wants to know why this blows up. The get/set functions
> > look so similar that it feels like the set should fail just the same...
> > Tommi, did you try just the set part of your test program [1]?
>
> Yes, both the set and get ioctls crash:
>
> [ 20.868660] BUG: unable to handle kernel NULL pointer dereference
> at (null)
> [ 20.876527] IP: [< (null)>] (null)
> [ 20.881573] PGD c4f7d067 PUD c2a6b067 PMD 0
> [ 20.885866] Oops: 0010 [#1] SMP KASAN
> [ 20.889549] CPU: 1 PID: 2207 Comm: main Not tainted 4.0.0-rc5+ #89
> [ 20.902805] task: ffff8800c4fad380 ti: ffff8800c2b98000 task.ti:
> ffff8800c2b98000
> [ 20.910257] RIP: 0010:[<0000000000000000>] [< (null)>]
> (null)
> [ 20.917722] RSP: 0018:ffff8800c2b9fd30 EFLAGS: 00010246
> [ 20.923012] RAX: ffffed002e87c961 RBX: ffff88017463d000 RCX: 0000000000000006
> [ 20.930116] RDX: dffffc0000000000 RSI: ffff8800c2b9fdd8 RDI: ffff8801743e4800
> [ 20.937214] RBP: ffff8800c2b9fd68 R08: 0000000000000000 R09: 0000000000000000
> [ 20.944318] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800c2b9fdd8
> [ 20.951416] R13: ffff8801743e48d8 R14: 00000000fffffffe R15: ffff8801743e4800
> [ 20.958524] FS: 00007f7139b3a700(0000) GS:ffff880175e00000(0000)
> knlGS:0000000000000000
> [ 20.966575] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 20.972300] CR2: 0000000000000000 CR3: 00000000c2a67000 CR4: 00000000000406e0
> [ 20.979407] Stack:
> [ 20.981414] ffffffff81b4a11d ffff8800c2b9fd68 ffff88017463d000
> ffff8800c4c50c00
> [ 20.988838] 0000000000000014 fffffffffffffff2 ffffffff8271c3e0
> ffff8800c2b9fe88
> [ 20.996238] ffffffff818acbbc ffff8800c2b9fe18 ffffffff8165d7c2
> ffffffff8165d660
> [ 21.003658] Call Trace:
> [ 21.006110] [<ffffffff81b4a11d>] ? intel_sprite_set_colorkey+0xad/0xf0
> [ 21.012695] [<ffffffff818acbbc>] drm_ioctl+0x27c/0x890
> [ 21.017904] [<ffffffff8165d7c2>] ? avc_has_perm+0x182/0x320
> [ 21.023544] [<ffffffff8165d660>] ? avc_has_perm+0x20/0x320
> [ 21.029098] [<ffffffff81b4a070>] ? intel_pre_disable_primary+0x90/0x90
> [ 21.035690] [<ffffffff8165ffac>] ? inode_has_perm.isra.28+0x7c/0xa0
> [ 21.042023] [<ffffffff812f8caf>] do_vfs_ioctl+0x3cf/0x720
> [ 21.047488] [<ffffffff81660caa>] ? selinux_file_ioctl+0x6a/0x130
> [ 21.053558] [<ffffffff812f9081>] SyS_ioctl+0x81/0xa0
> [ 21.058595] [<ffffffff825e08b2>] system_call_fastpath+0x12/0x17
> [ 21.064580] Code: Bad RIP value.
> [ 21.067916] RIP [< (null)>] (null)
> [ 21.073048] RSP <ffff8800c2b9fd30>
> [ 21.076524] CR2: 0000000000000000
> [ 21.079863] ---[ end trace 161ba639126f6a45 ]---
>
>
> [ 274.286068] BUG: unable to handle kernel NULL pointer dereference
> at (null)
> [ 274.295149] IP: [< (null)>] (null)
> [ 274.300242] PGD 171999067 PUD 171b93067 PMD 0
> [ 274.304744] Oops: 0010 [#1] SMP KASAN
> [ 274.308460] CPU: 0 PID: 2202 Comm: main Not tainted 4.0.0-rc5+ #89
> [ 274.321856] task: ffff8801726914e0 ti: ffff880172928000 task.ti:
> ffff880172928000
> [ 274.329383] RIP: 0010:[<0000000000000000>] [< (null)>]
> (null)
> [ 274.336924] RSP: 0018:ffff88017292fd30 EFLAGS: 00010246
> [ 274.342267] RAX: ffffed002e7bc362 RBX: ffff88017442f000 RCX: 0000000000000007
> [ 274.349446] RDX: 0000000000000000 RSI: ffff88017292fdd8 RDI: ffff880173de1800
> [ 274.356624] RBP: ffff88017292fd68 R08: 0000000000000000 R09: 0000000000000000
> [ 274.363803] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
> [ 274.370979] R13: ffff880173de18d8 R14: ffff88017292fdd8 R15: ffff880173de1800
> [ 274.378157] FS: 00007f48d6b16700(0000) GS:ffff880175c00000(0000)
> knlGS:0000000000000000
> [ 274.386297] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 274.392078] CR2: 0000000000000000 CR3: 000000017188d000 CR4: 00000000000406f0
> [ 274.399257] Stack:
> [ 274.401280] ffffffff81b4a1f7 ffff88017292fd68 ffff88017442f000
> ffff880172cc7c00
> [ 274.408761] 0000000000000014 fffffffffffffff2 ffffffff8271c3c0
> ffff88017292fe88
> [ 274.416244] ffffffff818acbbc ffff88017292fe18 ffffffff8165d7c2
> ffffffff8165d660
> [ 274.423727] Call Trace:
> [ 274.426192] [<ffffffff81b4a1f7>] ? intel_sprite_get_colorkey+0x97/0xc0
> [ 274.432849] [<ffffffff818acbbc>] drm_ioctl+0x27c/0x890
> [ 274.438107] [<ffffffff8165d7c2>] ? avc_has_perm+0x182/0x320
> [ 274.443800] [<ffffffff8165d660>] ? avc_has_perm+0x20/0x320
> [ 274.449407] [<ffffffff81b4a160>] ? intel_sprite_set_colorkey+0xf0/0xf0
> [ 274.456065] [<ffffffff8165ffac>] ? inode_has_perm.isra.28+0x7c/0xa0
> [ 274.462462] [<ffffffff812f8caf>] do_vfs_ioctl+0x3cf/0x720
> [ 274.467984] [<ffffffff81660caa>] ? selinux_file_ioctl+0x6a/0x130
> [ 274.474115] [<ffffffff812f9081>] SyS_ioctl+0x81/0xa0
> [ 274.479199] [<ffffffff825e08b2>] system_call_fastpath+0x12/0x17
> [ 274.485240] Code: Bad RIP value.
> [ 274.488597] RIP [< (null)>] (null)
> [ 274.493776] RSP <ffff88017292fd30>
> [ 274.497283] CR2: 0000000000000000
>
>
> I debugged this a bit, and found that in intel_sprite_set_colorkey(),
> the "intel_plane->update_colorkey" function pointer is NULL, and in
> intel_sprite_get_colorkey(), the "intel_plane->get_colorkey" pointer
> is NULL. Hence the crash.
>
> If I got it right, the pointers are not set for the "primary" and
> "cursor" planes, as initialized in intel_primary_plane_create() and
> intel_cursor_plane_create().
Ah true. So my patch to kill the rmw stuff should actually fix that
crash. Although we should not accept these ioctls for the
primary/cursor planes. I'll toss in a patch for that.
--
Ville Syrjälä
Intel OTC
More information about the Intel-gfx
mailing list