Re: [lockdep] drm/i915: possible circular locking dependency in i915 driver init

From: Masami Hiramatsu
Date: Mon Aug 15 2016 - 06:16:38 EST

Next message: Jiri Olsa: "[PATCH] x86/smp: Fix __max_logical_packages value setup"
Previous message: Maarten Lankhorst: "Re: [PATCH 2/2] drm/fence: allow fence waiting to be interrupted by userspace"
In reply to: Daniel Vetter: "Re: [lockdep] drm/i915: possible circular locking dependency in i915 driver init"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Mon, 15 Aug 2016 08:39:29 +0200
Daniel Vetter <daniel@xxxxxxxx> wrote:

> On Sun, Aug 14, 2016 at 11:01:35PM +0900, Masami Hiramatsu wrote:
> > Hello,
> >
> > I've found a suspicious circular locking dependency in i915 by lockdep.
> > It seems main driver initialization thread and sub fbdev configuration
> > thread take locks in different order implicitly. Please check it.
> >
> > The lockdep report is here.
>
> Should be fixed in 4.8-rc2, please confirm.

Thanks! I've confirmed this has fixed on 4.8-rc2.

Regards,

> -Daniel
>
> >
> > [ 4.254984] ======================================================
> > [ 4.254984] [ INFO: possible circular locking dependency detected ]
> > [ 4.254985] 4.8.0-rc1+ #8 Not tainted
> > [ 4.254985] -------------------------------------------------------
> > [ 4.254986] kworker/u16:1/72 is trying to acquire lock:
> > [ 4.255001] (&dev->mode_config.mutex){+.+.+.}, at: [<ffffffffc02eafb0>] drm_modeset_lock_all+0x40/0x130 [drm]
> > [ 4.255001]
> > but task is already holding lock:
> > [ 4.255005] ((fb_notifier_list).rwsem){++++.+}, at: [<ffffffff980b2ea5>] __blocking_notifier_call_chain+0x35/0x70
> > [ 4.255005]
> > which lock already depends on the new lock.
> >
> > [ 4.255005]
> > the existing dependency chain (in reverse order) is:
> > [ 4.255006]
> > -> #1 ((fb_notifier_list).rwsem){++++.+}:
> > [ 4.255009] [<ffffffff980e7a00>] lock_acquire+0x100/0x1f0
> > [ 4.255011] [<ffffffff988b29a9>] down_write+0x49/0x80
> > [ 4.255013] [<ffffffff980b3031>] blocking_notifier_chain_register+0x21/0xb0
> > [ 4.255015] [<ffffffff984c93c8>] fb_register_client+0x18/0x20
> > [ 4.255016] [<ffffffff984c8c1e>] backlight_device_register+0x12e/0x250
> > [ 4.255046] [<ffffffffc04e3842>] intel_backlight_device_register+0xa2/0x160 [i915]
> > [ 4.255070] [<ffffffffc04b173e>] intel_connector_register+0xe/0x10 [i915]
> > [ 4.255093] [<ffffffffc04cd5db>] intel_dp_connector_register+0x1b/0x80 [i915]
> > [ 4.255102] [<ffffffffc02d943a>] drm_connector_register+0x4a/0x80 [drm]
> > [ 4.255110] [<ffffffffc02ddbe6>] drm_modeset_register_all+0x1c6/0x260 [drm]
> > [ 4.255116] [<ffffffffc02d4ba2>] drm_dev_register+0xc2/0xd0 [drm]
> > [ 4.255135] [<ffffffffc04212f1>] i915_driver_load+0x771/0x1460 [i915]
> > [ 4.255153] [<ffffffffc042c63f>] i915_pci_probe+0x4f/0x70 [i915]
> > [ 4.255155] [<ffffffff98497905>] local_pci_probe+0x45/0xa0
> > [ 4.255157] [<ffffffff98498d11>] pci_device_probe+0xe1/0x130
> > [ 4.255159] [<ffffffff985ae5e4>] driver_probe_device+0x224/0x430
> > [ 4.255160] [<ffffffff985ae8d3>] __driver_attach+0xe3/0xf0
> > [ 4.255160] [<ffffffff985ac166>] bus_for_each_dev+0x66/0xa0
> > [ 4.255161] [<ffffffff985adcee>] driver_attach+0x1e/0x20
> > [ 4.255162] [<ffffffff985ad7d0>] bus_add_driver+0x200/0x270
> > [ 4.255164] [<ffffffff985af520>] driver_register+0x60/0xe0
> > [ 4.255165] [<ffffffff98497230>] __pci_register_driver+0x60/0x70
> > [ 4.255182] [<ffffffffc035f05b>] i915_init+0x5b/0x62 [i915]
> > [ 4.255183] [<ffffffff9800217d>] do_one_initcall+0x3d/0x160
> > [ 4.255186] [<ffffffff981c6d8d>] do_init_module+0x5f/0x1f8
> > [ 4.255187] [<ffffffff9812fbdc>] load_module+0x246c/0x2bd0
> > [ 4.255188] [<ffffffff9813057c>] SYSC_finit_module+0xbc/0xf0
> > [ 4.255190] [<ffffffff981305ce>] SyS_finit_module+0xe/0x10
> > [ 4.255191] [<ffffffff988b5680>] entry_SYSCALL_64_fastpath+0x23/0xc1
> > [ 4.255192]
> > -> #0 (&dev->mode_config.mutex){+.+.+.}:
> > [ 4.255194] [<ffffffff980e732c>] __lock_acquire+0x125c/0x14b0
> > [ 4.255195] [<ffffffff980e7a00>] lock_acquire+0x100/0x1f0
> > [ 4.255197] [<ffffffff988b0889>] mutex_lock_nested+0x69/0x3b0
> > [ 4.255206] [<ffffffffc02eafb0>] drm_modeset_lock_all+0x40/0x130 [drm]
> > [ 4.255212] [<ffffffffc03f877b>] drm_fb_helper_restore_fbdev_mode_unlocked+0x2b/0x80 [drm_kms_helper]
> > [ 4.255217] [<ffffffffc03f87fd>] drm_fb_helper_set_par+0x2d/0x50 [drm_kms_helper]
> > [ 4.255240] [<ffffffffc04c21da>] intel_fbdev_set_par+0x1a/0x60 [i915]
> > [ 4.255242] [<ffffffff984c2f2f>] fbcon_init+0x57f/0x600
> > [ 4.255244] [<ffffffff98555026>] visual_init+0xd6/0x130
> > [ 4.255245] [<ffffffff98556831>] do_bind_con_driver+0x1c1/0x3a0
> > [ 4.255246] [<ffffffff98556d16>] do_take_over_console+0x116/0x180
> > [ 4.255247] [<ffffffff984beb67>] do_fbcon_takeover+0x57/0xb0
> > [ 4.255249] [<ffffffff984c39a8>] fbcon_event_notify+0x658/0x750
> > [ 4.255250] [<ffffffff980b2c1a>] notifier_call_chain+0x4a/0x70
> > [ 4.255251] [<ffffffff980b2ebd>] __blocking_notifier_call_chain+0x4d/0x70
> > [ 4.255253] [<ffffffff980b2ef6>] blocking_notifier_call_chain+0x16/0x20
> > [ 4.255254] [<ffffffff984c940b>] fb_notifier_call_chain+0x1b/0x20
> > [ 4.255255] [<ffffffff984cb6c1>] register_framebuffer+0x251/0x330
> > [ 4.255259] [<ffffffffc03f8a7f>] drm_fb_helper_initial_config+0x25f/0x3f0 [drm_kms_helper]
> > [ 4.255282] [<ffffffffc04c3498>] intel_fbdev_initial_config+0x18/0x30 [i915]
> > [ 4.255283] [<ffffffff980b4aa7>] async_run_entry_fn+0x37/0x150
> > [ 4.255284] [<ffffffff980aa9c8>] process_one_work+0x1e8/0x710
> > [ 4.255285] [<ffffffff980aaf3b>] worker_thread+0x4b/0x4f0
> > [ 4.255286] [<ffffffff980b193f>] kthread+0xef/0x110
> > [ 4.255287] [<ffffffff988b58cf>] ret_from_fork+0x1f/0x40
> > [ 4.255287]
> > other info that might help us debug this:
> >
> > [ 4.255288] Possible unsafe locking scenario:
> >
> > [ 4.255288] CPU0 CPU1
> > [ 4.255288] ---- ----
> > [ 4.255289] lock((fb_notifier_list).rwsem);
> > [ 4.255290] lock(&dev->mode_config.mutex);
> > [ 4.255290] lock((fb_notifier_list).rwsem);
> > [ 4.255291] lock(&dev->mode_config.mutex);
> > [ 4.255291]
> > *** DEADLOCK ***
> >
> > And i915 related kconfigs are here:
> > ----
> > CONFIG_DRM_I915=m
> > CONFIG_DRM_I915_PRELIMINARY_HW_SUPPORT=y
> > CONFIG_DRM_I915_USERPTR=y
> > CONFIG_DRM_I915_GVT=y
> > # CONFIG_DRM_I915_WERROR is not set
> > # CONFIG_DRM_I915_DEBUG is not set
> > ----
> >
> > I've hit this on my DELL XPS15 which has skylake CPU/GPU.
> >
> > Thanks,
> >
> >
> > --
> > Masami Hiramatsu <mhiramat@xxxxxxxxxx>
> > _______________________________________________
> > dri-devel mailing list
> > dri-devel@xxxxxxxxxxxxxxxxxxxxx
> > https://lists.freedesktop.org/mailman/listinfo/dri-devel
>
> --
> Daniel Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch

--
Masami Hiramatsu <mhiramat@xxxxxxxxxx>

Next message: Jiri Olsa: "[PATCH] x86/smp: Fix __max_logical_packages value setup"
Previous message: Maarten Lankhorst: "Re: [PATCH 2/2] drm/fence: allow fence waiting to be interrupted by userspace"
In reply to: Daniel Vetter: "Re: [lockdep] drm/i915: possible circular locking dependency in i915 driver init"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]