linux-kernel - Re: [PATCH] drm/i915: change node clearing from memset to initialization

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 for Android: free password hash cracker in your pocket

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

Date:   Sat, 16 Apr 2022 11:33:34 -0700
From:   Joe Perches <joe@...ches.com>
To:     Tom Rix <trix@...hat.com>, jani.nikula@...ux.intel.com,
        joonas.lahtinen@...ux.intel.com, rodrigo.vivi@...el.com,
        tvrtko.ursulin@...ux.intel.com, airlied@...ux.ie, daniel@...ll.ch
Cc:     intel-gfx@...ts.freedesktop.org, dri-devel@...ts.freedesktop.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] drm/i915: change node clearing from memset to
 initialization

On Sat, 2022-04-16 at 13:23 -0400, Tom Rix wrote:
> In insert_mappable_node(), the parameter node is
> cleared late in node's use with memset.
> insert_mappable_node() is a singleton, called only
> from i915_gem_gtt_prepare() which itself is only
> called by i915_gem_gtt_pread() and
> i915_gem_gtt_pwrite_fast() where the definition of
> node originates.
> 
> Instead of using memset, initialize node to 0 at it's
> definitions.

trivia: /it's/its/

Only reason _not_ to do this is memset is guaranteed to
zero any padding that might go to userspace.

But it doesn't seem there is any padding anyway nor is
the struct available to userspace.

So this seems fine though it might increase overall code
size a tiny bit.

I do have a caveat: see below:

> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
[]
> @@ -328,7 +327,6 @@ static struct i915_vma *i915_gem_gtt_prepare(struct drm_i915_gem_object *obj,
>  		goto err_ww;
>  	} else if (!IS_ERR(vma)) {
>  		node->start = i915_ggtt_offset(vma);
> -		node->flags = 0;

Why is this unneeded?

from: drm_mm_insert_node_in_range which can set node->flags

		__set_bit(DRM_MM_NODE_ALLOCATED_BIT, &node->flags);