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] [day] [month] [year] [list]
Message-ID: <202504071303.BF0B5846@keescook>
Date: Mon, 7 Apr 2025 13:09:36 -0700
From: Kees Cook <kees@...nel.org>
To: "Gustavo A. R. Silva" <gustavoars@...nel.org>
Cc: Lyude Paul <lyude@...hat.com>, Danilo Krummrich <dakr@...nel.org>,
	David Airlie <airlied@...il.com>, Simona Vetter <simona@...ll.ch>,
	dri-devel@...ts.freedesktop.org, nouveau@...ts.freedesktop.org,
	linux-kernel@...r.kernel.org, linux-hardening@...r.kernel.org
Subject: Re: [PATCH v2][next] drm/nouveau: fifo: Avoid
 -Wflex-array-member-not-at-end warning

On Fri, Apr 04, 2025 at 09:58:03AM -0600, Gustavo A. R. Silva wrote:
> -Wflex-array-member-not-at-end was introduced in GCC-14, and we are
> getting ready to enable it, globally.
> 
> Adjust heap allocation to account for the flex-array-in-the-middle
> issue, and refactor the rest of the code accordingly.
> 
> So, with these changes, fix the following warning:
> 
> drivers/gpu/drm/nouveau/nvif/fifo.c:29:42: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
> 
> Signed-off-by: Gustavo A. R. Silva <gustavoars@...nel.org>
> ---
> Changes in v2:
>  - Adjust heap allocation instead of using the DEFINE_RAW_FLEX() helper.
>  - Link: https://lore.kernel.org/lkml/202504041254.6e26LBdj-lkp@intel.com/
> 
> v1:
>  - Link: https://lore.kernel.org/linux-hardening/Z-7IQcWNePAMQEM0@kspp/
> 
>  drivers/gpu/drm/nouveau/nvif/fifo.c | 35 +++++++++++++++--------------
>  1 file changed, 18 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/gpu/drm/nouveau/nvif/fifo.c b/drivers/gpu/drm/nouveau/nvif/fifo.c
> index a463289962b2..2c887dffa45d 100644
> --- a/drivers/gpu/drm/nouveau/nvif/fifo.c
> +++ b/drivers/gpu/drm/nouveau/nvif/fifo.c
> @@ -25,33 +25,34 @@ static int
>  nvif_fifo_runlists(struct nvif_device *device)
>  {
>  	struct nvif_object *object = &device->object;
> -	struct {
> -		struct nv_device_info_v1 m;
> -		struct {
> -			struct nv_device_info_v1_data runlists;
> -			struct nv_device_info_v1_data runlist[64];
> -		} v;
> -	} *a;
> +	struct nv_device_info_v1_data *runlists;
> +	struct nv_device_info_v1_data *runlist;
> +	struct nv_device_info_v1 *a;
> +	const u8 runlist_cnt = 64;
>  	int ret, i;
>  
>  	if (device->runlist)
>  		return 0;
>  
> -	if (!(a = kmalloc(sizeof(*a), GFP_KERNEL)))
> +	a = kmalloc(struct_size(a, data, runlist_cnt + 1), GFP_KERNEL);

Hrm, a pre-existing issue, but this is dumping quite a few uninitialized
bytes into this allocation, as struct nv_device_info_v1::pad is never
initialized and runlists->data is never initialized. It looks like these
bytes are being written out to the video card? Or some kind of ioctl?

This should really be using kzalloc() or making sure stuff isn't
uninitialized.

> +	if (!a)
>  		return -ENOMEM;
> -	a->m.version = 1;
> -	a->m.count = sizeof(a->v) / sizeof(a->v.runlists);
> -	a->v.runlists.mthd = NV_DEVICE_HOST_RUNLISTS;
> -	for (i = 0; i < ARRAY_SIZE(a->v.runlist); i++) {
> -		a->v.runlist[i].mthd = NV_DEVICE_HOST_RUNLIST_ENGINES;
> -		a->v.runlist[i].data = i;
> +
> +	runlists = &a->data[0];
> +	runlist = &a->data[1];
> +	a->version = 1;
> +	a->count = runlist_cnt + 1;

Seems struct nv_device_info_v1::count is the counted_by for struct
nv_device_info_v1::data?

> +	runlists->mthd = NV_DEVICE_HOST_RUNLISTS;
> +	for (i = 0; i < runlist_cnt; i++) {
> +		runlist[i].mthd = NV_DEVICE_HOST_RUNLIST_ENGINES;
> +		runlist[i].data = i;
>  	}
>  
>  	ret = nvif_object_mthd(object, NV_DEVICE_V0_INFO, a, sizeof(*a));
>  	if (ret)
>  		goto done;
>  
> -	device->runlists = fls64(a->v.runlists.data);
> +	device->runlists = fls64(runlists->data);

I assume nvif_object_mthd() writes runlists->data?

>  	device->runlist = kcalloc(device->runlists, sizeof(*device->runlist),
>  				  GFP_KERNEL);
>  	if (!device->runlist) {
> @@ -60,8 +61,8 @@ nvif_fifo_runlists(struct nvif_device *device)
>  	}
>  
>  	for (i = 0; i < device->runlists; i++) {
> -		if (a->v.runlist[i].mthd != NV_DEVICE_INFO_INVALID)
> -			device->runlist[i].engines = a->v.runlist[i].data;
> +		if (runlist[i].mthd != NV_DEVICE_INFO_INVALID)
> +			device->runlist[i].engines = runlist[i].data;
>  	}
>  
>  done:
> -- 
> 2.43.0
> 

-- 
Kees Cook

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ