Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
MIME-Version: 1.0
References: <20220420095720.3331609-1-kherbst@redhat.com>
In-Reply-To: <20220420095720.3331609-1-kherbst@redhat.com>
From:   Karol Herbst <kherbst@redhat.com>
Date:   Mon, 2 May 2022 16:39:32 +0200
Message-ID: <CACO55tv6rKETHhx=1yJ7Gf91sq1cDHyNgZxEdaN48PFBRdKu2g@mail.gmail.com>
Subject: Re: [PATCH v2] drm/i915: Fix race in __i915_vma_remove_closed
To:     LKML <linux-kernel@vger.kernel.org>
Cc:     Intel Graphics Development <intel-gfx@lists.freedesktop.org>,
        dri-devel <dri-devel@lists.freedesktop.org>,
        Joonas Lahtinen <joonas.lahtinen@linux.intel.com>,
        Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

friendly ping.

I am not even myself completely convinced that this is the correct
patch and it might just workaround some issues, but list_debug.c does
check if a list was already deleted and throws an error if it was and
this patch indeed fixes this one issue as multiple threads could enter
__i915_vma_remove_closed on the same vma.


On Wed, Apr 20, 2022 at 11:57 AM Karol Herbst <kherbst@redhat.com> wrote:
>
> i915_vma_reopen checked if the vma is closed before without taking the
> lock. So multiple threads could attempt removing the vma.
>
> Instead the lock needs to be taken before actually checking.
>
> v2: move struct declaration
>
> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: intel-gfx@lists.freedesktop.org
> Cc: dri-devel@lists.freedesktop.org
> Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/5732
> Signed-off-by: Karol Herbst <kherbst@redhat.com>
> ---
>  drivers/gpu/drm/i915/i915_vma.c | 11 +++++++----
>  1 file changed, 7 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
> index 162e8d83691b..2efdad2b43fa 100644
> --- a/drivers/gpu/drm/i915/i915_vma.c
> +++ b/drivers/gpu/drm/i915/i915_vma.c
> @@ -1615,17 +1615,17 @@ void i915_vma_close(struct i915_vma *vma)
>
>  static void __i915_vma_remove_closed(struct i915_vma *vma)
>  {
> -       struct intel_gt *gt = vma->vm->gt;
> -
> -       spin_lock_irq(&gt->closed_lock);
>         list_del_init(&vma->closed_link);
> -       spin_unlock_irq(&gt->closed_lock);
>  }
>
>  void i915_vma_reopen(struct i915_vma *vma)
>  {
> +       struct intel_gt *gt = vma->vm->gt;
> +
> +       spin_lock_irq(&gt->closed_lock);
>         if (i915_vma_is_closed(vma))
>                 __i915_vma_remove_closed(vma);
> +       spin_unlock_irq(&gt->closed_lock);
>  }
>
>  static void force_unbind(struct i915_vma *vma)
> @@ -1641,6 +1641,7 @@ static void force_unbind(struct i915_vma *vma)
>  static void release_references(struct i915_vma *vma, bool vm_ddestroy)
>  {
>         struct drm_i915_gem_object *obj = vma->obj;
> +       struct intel_gt *gt = vma->vm->gt;
>
>         GEM_BUG_ON(i915_vma_is_active(vma));
>
> @@ -1651,7 +1652,9 @@ static void release_references(struct i915_vma *vma, bool vm_ddestroy)
>
>         spin_unlock(&obj->vma.lock);
>
> +       spin_lock_irq(&gt->closed_lock);
>         __i915_vma_remove_closed(vma);
> +       spin_unlock_irq(&gt->closed_lock);
>
>         if (vm_ddestroy)
>                 i915_vm_resv_put(vma->vm);
> --
> 2.35.1
>