linux-kernel - Re: [PATCH] dma: idxd: cdev: Fix uninitialized use of sva in idxd_cdev

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]

Message-ID: <875xjb6c07.fsf@intel.com>
Date: Thu, 10 Apr 2025 12:40:40 -0700
From: Vinicius Costa Gomes <vinicius.gomes@...el.com>
To: Purva Yeshi <purvayeshi550@...il.com>, dave.jiang@...el.com,
 vkoul@...nel.org
Cc: dmaengine@...r.kernel.org, linux-kernel@...r.kernel.org, Purva Yeshi
 <purvayeshi550@...il.com>
Subject: Re: [PATCH] dma: idxd: cdev: Fix uninitialized use of sva in
 idxd_cdev_open

Purva Yeshi <purvayeshi550@...il.com> writes:

> Fix Smatch-detected issue:
> drivers/dma/idxd/cdev.c:321 idxd_cdev_open() error:
> uninitialized symbol 'sva'.
>
> 'sva' pointer may be used uninitialized in error handling paths.
> Specifically, if PASID support is enabled and iommu_sva_bind_device()
> returns an error, the code jumps to the cleanup label and attempts to
> call iommu_sva_unbind_device(sva) without ensuring that sva was
> successfully assigned. This triggers a Smatch warning about an
> uninitialized symbol.
>
> Initialize sva to NULL at declaration and add a check using
> IS_ERR_OR_NULL() before unbinding the device. This ensures the
> function does not use an invalid or uninitialized pointer during
> cleanup.
>
> Signed-off-by: Purva Yeshi <purvayeshi550@...il.com>
> ---
>  drivers/dma/idxd/cdev.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/dma/idxd/cdev.c b/drivers/dma/idxd/cdev.c
> index ff94ee892339..7bd031a60894 100644
> --- a/drivers/dma/idxd/cdev.c
> +++ b/drivers/dma/idxd/cdev.c
> @@ -222,7 +222,7 @@ static int idxd_cdev_open(struct inode *inode, struct file *filp)
>  	struct idxd_wq *wq;
>  	struct device *dev, *fdev;
>  	int rc = 0;
> -	struct iommu_sva *sva;
> +	struct iommu_sva *sva = NULL;
>  	unsigned int pasid;
>  	struct idxd_cdev *idxd_cdev;
>  
> @@ -317,7 +317,7 @@ static int idxd_cdev_open(struct inode *inode, struct file *filp)
>  	if (device_user_pasid_enabled(idxd))
>  		idxd_xa_pasid_remove(ctx);
>  failed_get_pasid:
> -	if (device_user_pasid_enabled(idxd))
> +	if (device_user_pasid_enabled(idxd) && !IS_ERR_OR_NULL(sva))

Optional: I would change this to only checking for the validity of
'sva', the other condition would be true if 'sva' is valid.

But for consistency with the condition above, I am not opposed to the
way this patch is written:

Acked-by: Vinicius Costa Gomes <vinicius.gomes@...el.com>


Cheers,
-- 
Vinicius