drm/i915/fbdev: Serialise early hotplug events with async fbdev config
authorChris Wilson <chris@chris-wilson.co.uk>
Sat, 25 Nov 2017 19:41:55 +0000 (19:41 +0000)
committerChris Wilson <chris@chris-wilson.co.uk>
Sun, 26 Nov 2017 12:44:59 +0000 (12:44 +0000)
commitad88d7fc6c032ddfb32b8d496a070ab71de3a64f
treee127133bbd175fd960d81aba7d0a6fc395867778
parent53fa54a6ef4fa5a261468d13dc0f54739adaa456
drm/i915/fbdev: Serialise early hotplug events with async fbdev config

As both the hotplug event and fbdev configuration run asynchronously, it
is possible for them to run concurrently. If configuration fails, we were
freeing the fbdev causing a use-after-free in the hotplug event.

<7>[ 3069.935211] [drm:intel_fb_initial_config [i915]] Not using firmware configuration
<7>[ 3069.935225] [drm:drm_setup_crtcs] looking for cmdline mode on connector 77
<7>[ 3069.935229] [drm:drm_setup_crtcs] looking for preferred mode on connector 77 0
<7>[ 3069.935233] [drm:drm_setup_crtcs] found mode 3200x1800
<7>[ 3069.935236] [drm:drm_setup_crtcs] picking CRTCs for 8192x8192 config
<7>[ 3069.935253] [drm:drm_setup_crtcs] desired mode 3200x1800 set on crtc 43 (0,0)
<7>[ 3069.935323] [drm:intelfb_create [i915]] no BIOS fb, allocating a new one
<4>[ 3069.967737] general protection fault: 0000 [#1] PREEMPT SMP
<0>[ 3069.977453] ---------------------------------
<4>[ 3069.977457] Modules linked in: i915(+) vgem snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic x86_pkg_temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm r8169 mei_me mii prime_numbers mei i2c_hid pinctrl_geminilake pinctrl_intel [last unloaded: i915]
<4>[ 3069.977492] CPU: 1 PID: 15414 Comm: kworker/1:0 Tainted: G     U          4.14.0-CI-CI_DRM_3388+ #1
<4>[ 3069.977497] Hardware name: Intel Corp. Geminilake/GLK RVP1 DDR4 (05), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017
<4>[ 3069.977508] Workqueue: events output_poll_execute
<4>[ 3069.977512] task: ffff880177734e40 task.stack: ffffc90001fe4000
<4>[ 3069.977519] RIP: 0010:__lock_acquire+0x109/0x1b60
<4>[ 3069.977523] RSP: 0018:ffffc90001fe7bb0 EFLAGS: 00010002
<4>[ 3069.977526] RAX: 6b6b6b6b6b6b6b6b RBX: 0000000000000282 RCX: 0000000000000000
<4>[ 3069.977530] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff880170d4efd0
<4>[ 3069.977534] RBP: ffffc90001fe7c70 R08: 0000000000000001 R09: 0000000000000000
<4>[ 3069.977538] R10: 0000000000000000 R11: ffffffff81899609 R12: ffff880170d4efd0
<4>[ 3069.977542] R13: ffff880177734e40 R14: 0000000000000001 R15: 0000000000000000
<4>[ 3069.977547] FS:  0000000000000000(0000) GS:ffff88017fc80000(0000) knlGS:0000000000000000
<4>[ 3069.977551] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
<4>[ 3069.977555] CR2: 00007f7e8b7bcf04 CR3: 0000000003e0f000 CR4: 00000000003406e0
<4>[ 3069.977559] Call Trace:
<4>[ 3069.977565]  ? mark_held_locks+0x64/0x90
<4>[ 3069.977571]  ? _raw_spin_unlock_irq+0x24/0x50
<4>[ 3069.977575]  ? _raw_spin_unlock_irq+0x24/0x50
<4>[ 3069.977579]  ? trace_hardirqs_on_caller+0xde/0x1c0
<4>[ 3069.977583]  ? _raw_spin_unlock_irq+0x2f/0x50
<4>[ 3069.977588]  ? finish_task_switch+0xa5/0x210
<4>[ 3069.977592]  ? lock_acquire+0xaf/0x200
<4>[ 3069.977596]  lock_acquire+0xaf/0x200
<4>[ 3069.977600]  ? __mutex_lock+0x5e9/0x9b0
<4>[ 3069.977604]  _raw_spin_lock+0x2a/0x40
<4>[ 3069.977608]  ? __mutex_lock+0x5e9/0x9b0
<4>[ 3069.977612]  __mutex_lock+0x5e9/0x9b0
<4>[ 3069.977616]  ? drm_fb_helper_hotplug_event.part.19+0x16/0xa0
<4>[ 3069.977621]  ? drm_fb_helper_hotplug_event.part.19+0x16/0xa0
<4>[ 3069.977625]  drm_fb_helper_hotplug_event.part.19+0x16/0xa0
<4>[ 3069.977630]  output_poll_execute+0x8d/0x180
<4>[ 3069.977635]  process_one_work+0x22e/0x660
<4>[ 3069.977640]  worker_thread+0x48/0x3a0
<4>[ 3069.977644]  ? _raw_spin_unlock_irqrestore+0x4c/0x60
<4>[ 3069.977649]  kthread+0x102/0x140
<4>[ 3069.977653]  ? process_one_work+0x660/0x660
<4>[ 3069.977657]  ? kthread_create_on_node+0x40/0x40
<4>[ 3069.977662]  ret_from_fork+0x27/0x40
<4>[ 3069.977666] Code: 8d 62 f8 c3 49 81 3c 24 e0 fa 3c 82 41 be 00 00 00 00 45 0f 45 f0 83 fe 01 77 86 89 f0 49 8b 44 c4 08 48 85 c0 0f 84 76 ff ff ff <f0> ff 80 38 01 00 00 8b 1d 62 f9 e8 01 45 8b 85 b8 08 00 00 85
<1>[ 3069.977707] RIP: __lock_acquire+0x109/0x1b60 RSP: ffffc90001fe7bb0
<4>[ 3069.977712] ---[ end trace 4ad012eb3af62df7 ]---

In order to keep the dev_priv->ifbdev alive after failure, we have to
avoid the free and leave it empty until we unload the module (which is
less than ideal, but a necessary evil for simplicity). Then we can use
intel_fbdev_sync() to serialise the hotplug event with the configuration.
The serialisation between the two was removed in commit 934458c2c95d
("Revert "drm/i915: Fix races on fbdev""), but the use after free is much
older, commit 366e39b4d2c5 ("drm/i915: Tear down fbdev if initialization
fails")

Fixes: 366e39b4d2c5 ("drm/i915: Tear down fbdev if initialization fails")
Fixes: 934458c2c95d ("Revert "drm/i915: Fix races on fbdev"")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lukas Wunner <lukas@wunner.de>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Reviewed-by: Lukas Wunner <lukas@wunner.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20171125194155.355-1-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/intel_fbdev.c