netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Re: [PATCH v2] wifi: mt76: Fix strscpy buffer overflow in mt76_connac2_load_patch
       [not found] ` <20251205161202.48409-1-mikhail.v.gavrilov@gmail.com>
@ 2025-12-23 21:54   ` Nathan Chancellor
  0 siblings, 0 replies; only message in thread
From: Nathan Chancellor @ 2025-12-23 21:54 UTC (permalink / raw)
  To: Mikhail Gavrilov
  Cc: Mario Limonciello, Felix Fietkau, Lorenzo Bianconi,
	linux-wireless, linux-mediatek, linux-kernel, Johannes Berg,
	Andrew Lunn, David S. Miller, Eric Dumazet, Jakub Kicinski,
	Paolo Abeni, netdev

+ netdev and wireless/networking maintainers

On Fri, Dec 05, 2025 at 09:12:02PM +0500, Mikhail Gavrilov wrote:
> Commit f804a5895eba ("wifi: mt76: Strip whitespace from build ddate") introduced
> a kernel panic/WARN on systems using MediaTek MT7921e (and potentially others
> using mt76_connac_lib) due to an incorrect buffer size calculation.
> 
> The error logged is:
> "strnlen: detected buffer overflow: 17 byte read of buffer size 16"
> 
> This occurs because the field 'hdr->build_date' is a fixed-size array of 16 bytes.
> The patch allocated a 17-byte local buffer 'build_date' but used 'sizeof(build_date)'
> (17) as the read limit for strscpy, causing Fortify Source to correctly detect
> an attempt to read 17 bytes from the 16-byte source field.
> 
> To fix this, replace strscpy with memcpy, which is appropriate for raw data
> copying, and explicitly use the size of the source field (sizeof(hdr->build_date) = 16)
> to limit the read, followed by manual null termination.
> 
> Fixes: f804a5895eba ("wifi: mt76: Strip whitespace from build ddate")
> Signed-off-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>

I got bit by this regression when installing v6.19-rc2 on my new test
machine, which has an MT7925 (RZ717) chip in it. I don't see this in
either Felix's or the main wireless tree yet but I do understand it is
the end of the year with breaks and such (along with Johannes not
actually being on CC since he is not in the output of get_maintainers.pl
for drivers/net/wireless/mediatek/mt76). If there is not going to be a
wireless pull soon, can this be applied to net directly so that it gets
to Linus soon? It was rather annoying to do a bisect for a regression
that already has a pending fix.

Tested-by: Nathan Chancellor <nathan@kernel.org>

> ---
>  drivers/net/wireless/mediatek/mt76/mt76_connac_mcu.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/wireless/mediatek/mt76/mt76_connac_mcu.c b/drivers/net/wireless/mediatek/mt76/mt76_connac_mcu.c
> index ea99167765b0..d2c4c65ec464 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt76_connac_mcu.c
> +++ b/drivers/net/wireless/mediatek/mt76/mt76_connac_mcu.c
> @@ -3125,8 +3125,11 @@ int mt76_connac2_load_patch(struct mt76_dev *dev, const char *fw_name)
>  	}
>  
>  	hdr = (const void *)fw->data;
> -	strscpy(build_date, hdr->build_date, sizeof(build_date));
> -	build_date[16] = '\0';
> +	/* hdr->build_date is 16 bytes. Copy exactly 16 bytes to the 17-byte buffer,
> +	 * and then add the null terminator at index 16.
> +	 */
> +	memcpy(build_date, hdr->build_date, sizeof(hdr->build_date));
> +	build_date[sizeof(hdr->build_date)] = '\0';
>  	strim(build_date);
>  	dev_info(dev->dev, "HW/SW Version: 0x%x, Build Time: %.16s\n",
>  		 be32_to_cpu(hdr->hw_sw_ver), build_date);
> -- 
> 2.52.0
> 

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2025-12-23 21:54 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <CABXGCsMeAZyNJ-Axt_CUCXgyieWPV3rrcLpWsveMPT8R0YPGnQ@mail.gmail.com>
     [not found] ` <20251205161202.48409-1-mikhail.v.gavrilov@gmail.com>
2025-12-23 21:54   ` [PATCH v2] wifi: mt76: Fix strscpy buffer overflow in mt76_connac2_load_patch Nathan Chancellor

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).