linux-kernel - Re: [PATCH v6] staging: rtl8723bs: Add error handling for sd

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: <aa78d490-f0c7-4977-ae25-fe15d78b8d13@stanley.mountain>
Date: Mon, 7 Apr 2025 13:36:44 +0300
From: Dan Carpenter <dan.carpenter@...aro.org>
To: Wentao Liang <vulab@...as.ac.cn>
Cc: gregkh@...uxfoundation.org, philipp.g.hortmann@...il.com,
	linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org,
	stable@...r.kernel.org
Subject: Re: [PATCH v6] staging: rtl8723bs: Add error handling for sd_read()

On Mon, Apr 07, 2025 at 06:03:18PM +0800, Wentao Liang wrote:
> The sdio_read32() calls sd_read(), but does not handle the error if
> sd_read() fails. This could lead to subsequent operations processing
> invalid data. A proper implementation can be found in sdio_readN().
> 
> Add error handling for the sd_read() to free tmpbuf and return error
> code if sd_read() fails. This ensure that the memcpy() is only performed
> when the read operation is successful.
> 
> Fixes: 554c0a3abf21 ("staging: Add rtl8723bs sdio wifi driver")
> Cc: stable@...r.kernel.org # v4.12+
> Signed-off-by: Wentao Liang <vulab@...as.ac.cn>
> ---
> v6: Fix improper code to propagate error code
> v5: Fix error code
> v4: Add change log and fix error code
> v3: Add Cc flag
> v2: Change code to initialize val
> 
>  drivers/staging/rtl8723bs/hal/sdio_ops.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/rtl8723bs/hal/sdio_ops.c b/drivers/staging/rtl8723bs/hal/sdio_ops.c
> index 21e9f1858745..eb21c7e55949 100644
> --- a/drivers/staging/rtl8723bs/hal/sdio_ops.c
> +++ b/drivers/staging/rtl8723bs/hal/sdio_ops.c
> @@ -185,7 +185,12 @@ static u32 sdio_read32(struct intf_hdl *intfhdl, u32 addr)
>  			return SDIO_ERR_VAL32;
>  
>  		ftaddr &= ~(u16)0x3;
> -		sd_read(intfhdl, ftaddr, 8, tmpbuf);
> +		err = sd_read(intfhdl, ftaddr, 8, tmpbuf);
> +		if (err) {
> +			kfree(tmpbuf);
> +			return (u32)err;

Heh.

So the fundamental problem is that non of the callers check for errors.

To be honest, I had expected you to just return zero, but I don't like to
give out the answers to students.  I hadn't even known that SDIO_ERR_VAL32
was an option.  It's still a garbage value but it's kind of a predictable
garbage value and, whatever, it seemed fine to me.  It wasn't fine to Greg
so, yeah, you have to re-write it.  But now this is again not fine to me
(or Greg when he gets around to checking his email).

The bug here is that if you pull out the hardware while doing a read
then it returns whatever was in the kmalloc().  In other words it's an
information leak.

I think you could make an argument that returnnig zero is a good solution.
It fixes the information leak.  It's not a a horrible random value like
"(u32)-EINVAL".

The other option would be to go through all the callers and add error
handling.  So for this this function you would have to pass a pointer to
u32 *val and return zero on success or negative on failure.

regards,
dan carpenter