linux-kernel - Re: Commit "locking/drm: Kill mutex trickery" causes hangs

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

Date:   Sun, 30 Oct 2016 17:09:41 -0700 (PDT)
From:   Hugh Dickins <hughd@...gle.com>
To:     Mike Krinkin <krinkin.m.u@...il.com>
cc:     peterz@...radead.org, jason.low2@....com, mingo@...nel.org,
        tglx@...utronix.de, chris@...is-wilson.co.uk,
        linux-kernel@...r.kernel.org
Subject: Re: Commit "locking/drm: Kill mutex trickery" causes hangs

On Mon, 31 Oct 2016, Mike Krinkin wrote:
> 
> i faced system hangs with recent linux-next versions, bisect points at the
> commit 3ab7c086d5ec72585ef0 ("locking/drm: Kill mutex trickery"), bisect log
> attached. System just hangs after few minutes when i compile kernel with -j4
> and watch some video simultaneously.
[...]
> also lspci -vvv output:
[...]
> 	Kernel driver in use: i915
> 	Kernel modules: i915

Yes, that's hit me too, on mmotm on i915.  i915_gem_shrinker_lock()
is broken: but copy the pattern from msm_gem_shrinker_lock() and it's
okay - patch below.  Well, okay-ish: I'm reluctant to sign off on that
as more than a quick fix for i915 linux-next users, since the unlock
variable and those _gem_shrinker_lock() wrappers should just be deleted
(if the mutex trickery is indeed to be killed).

And I'm still left with a "sleeping function called from invalid context"
warning, which seems easier to live with: I've not looked to see whether
that's a consequence of the mutex trickery killage or something else.

[   12.887922] BUG: sleeping function called from invalid context at drivers/base/power/runtime.c:956
[   12.887925] in_atomic(): 1, irqs_disabled(): 0, pid: 787, name: X
[   12.887927] 1 lock held by X/787:
[   12.887928]  #0: 
[   12.887929]  (
[   12.887930] &dev->struct_mutex
[   12.887931] ){+.+.+.}
[   12.887932] , at: 
[   12.887937] [<ffffffff813e0ccb>] i915_mutex_lock_interruptible+0x23/0x26
[   12.887939] Preemption disabled at:
[   12.887943] [<ffffffff813d67c0>] i915_gem_execbuffer_relocate_entry+0x5fb/0x70f
[   12.887947] CPU: 2 PID: 787 Comm: X Not tainted 4.9.0-rc2-mm1 #5
[   12.887948] Hardware name: LENOVO 4174EH1/4174EH1, BIOS 8CET51WW (1.31 ) 11/29/2011
[   12.887950] Call Trace:
[   12.887955]  dump_stack+0x67/0x90
[   12.887958]  ? i915_gem_execbuffer_relocate_entry+0x5fb/0x70f
[   12.887961]  ___might_sleep+0x223/0x23a
[   12.887963]  __might_sleep+0x6d/0x81
[   12.887966]  __pm_runtime_resume+0x35/0x7a
[   12.887970]  intel_runtime_pm_get+0x20/0x7f
[   12.887973]  aliasing_gtt_bind_vma+0x4d/0xb1
[   12.887975]  i915_vma_bind+0x67/0xbd
[   12.887977]  i915_gem_execbuffer_relocate_entry+0xc6/0x70f
[   12.887981]  ? _raw_spin_unlock_irq+0x27/0x45
[   12.887984]  i915_gem_execbuffer_relocate_vma+0x128/0x1dd
[   12.887987]  ? nommu_map_sg+0x9e/0xca
[   12.887990]  ? __i915_vma_do_pin+0x3da/0x421
[   12.887994]  ? i915_gem_execbuffer_reserve_vma.isra.34+0xbc/0x189
[   12.887996]  ? i915_gem_execbuffer_reserve.isra.35+0x32f/0x3da
[   12.887999]  i915_gem_do_execbuffer.isra.36+0x64c/0x10a9
[   12.888002]  i915_gem_execbuffer2+0x15d/0x203
[   12.888005]  drm_ioctl+0x25a/0x38b
[   12.888007]  ? i915_gem_execbuffer+0x2d3/0x2d3
[   12.888011]  vfs_ioctl+0x1c/0x33
[   12.888014]  do_vfs_ioctl+0x5c5/0x601
[   12.888016]  ? __fget+0x17e/0x18f
[   12.888019]  ? expand_files+0x23e/0x23e
[   12.888021]  SyS_ioctl+0x38/0x60
[   12.888023]  entry_SYSCALL_64_fastpath+0x18/0xad

--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -229,8 +229,9 @@ unsigned long i915_gem_shrink_all(struct
 static bool i915_gem_shrinker_lock(struct drm_device *dev, bool *unlock)
 {
 	if (!mutex_trylock(&dev->struct_mutex))
-		*unlock = false;
+		return false;
 
+	*unlock = true;
 	return true;
 }