[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAKwvOdkXmEe46cG9Hn837215ghWA7UNKtg7ZQM8CpQcEQnoWfg@mail.gmail.com>
Date: Thu, 31 Aug 2023 14:41:31 -0700
From: Nick Desaulniers <ndesaulniers@...gle.com>
To: Helge Deller <deller@....de>
Cc: linux-fbdev@...r.kernel.org,
dri-devel <dri-devel@...ts.freedesktop.org>,
clang-built-linux <llvm@...ts.linux.dev>,
LKML <linux-kernel@...r.kernel.org>,
Andrew Morton <akpm@...ux-foundation.org>,
Nathan Chancellor <nathan@...nel.org>,
Kees Cook <keescook@...omium.org>,
Justin Stitt <justinstitt@...gle.com>
Subject: Re: truncation in drivers/video/fbdev/neofb.c
On Thu, Aug 31, 2023 at 2:23 PM Helge Deller <deller@....de> wrote:
>
> * Helge Deller <deller@....de>:
> > On 8/29/23 18:45, Nick Desaulniers wrote:
> > > A recent change in clang made it better about spotting snprintf that
> > > will result in truncation. Nathan reported the following instances:
> > >
> > > drivers/video/fbdev/neofb.c:1959:3: warning: 'snprintf' will always be
> > > truncated; specified size is 16, but format string expands to at least
> > > 17 [-Wfortify-source]
>
> FYI, I've added the patch below to the fbdev for-next git tree.
>
> Helge
>
> From: Helge Deller <deller@....de>
> Subject: [PATCH] fbdev: neofb: Shorten Neomagic product name in info struct
>
> Avoid those compiler warnings:
> neofb.c:1959:3: warning: 'snprintf' will always be truncated;
> specified size is 16, but format string expands to at least 17 [-Wfortify-source]
>
> Signed-off-by: Helge Deller <deller@....de>
> Reported-by: Nathan Chancellor <nathan@...nel.org>
> Reported-by: Nick Desaulniers <ndesaulniers@...gle.com>
> Link: https://lore.kernel.org/all/CAKwvOdn0xoVWjQ6ufM_rojtKb0f1i1hW-J_xYGfKDNFdHwaeHQ@mail.gmail.com/
> Link: https://github.com/ClangBuiltLinux/linux/issues/1923
This indeed makes the warning go away, but that's more so due to the
use of strscpy now rather than snprintf. That alone is a good change
but we still have definite truncation. See below:
>
> diff --git a/drivers/video/fbdev/neofb.c b/drivers/video/fbdev/neofb.c
> index d2f622b4c372..b905fe93b525 100644
> --- a/drivers/video/fbdev/neofb.c
> +++ b/drivers/video/fbdev/neofb.c
> @@ -1948,49 +1948,40 @@ static struct fb_info *neo_alloc_fb_info(struct pci_dev *dev,
>
> switch (info->fix.accel) {
> case FB_ACCEL_NEOMAGIC_NM2070:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 128");
> + strscpy(info->fix.id, "MagicGraph128", sizeof(info->fix.id));
> break;
> case FB_ACCEL_NEOMAGIC_NM2090:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 128V");
> + strscpy(info->fix.id, "MagicGraph128V", sizeof(info->fix.id));
> break;
> case FB_ACCEL_NEOMAGIC_NM2093:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 128ZV");
> + strscpy(info->fix.id, "MagicGraph128ZV", sizeof(info->fix.id));
> break;
> case FB_ACCEL_NEOMAGIC_NM2097:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 128ZV+");
> + strscpy(info->fix.id, "Mag.Graph128ZV+", sizeof(info->fix.id));
I see the Mag.Graph change here; I'm curious why MagicGraph256XL+
didn't need that, too?
MagicGraph128ZV+
MagicGraph256XL+
those look like the same number of characters; are they? Does this one
need the `.` change?
```
diff --git a/drivers/video/fbdev/neofb.c b/drivers/video/fbdev/neofb.c
index b905fe93b525..6b7836f7f809 100644
--- a/drivers/video/fbdev/neofb.c
+++ b/drivers/video/fbdev/neofb.c
@@ -1957,7 +1957,7 @@ static struct fb_info *neo_alloc_fb_info(struct
pci_dev *dev,
strscpy(info->fix.id, "MagicGraph128ZV", sizeof(info->fix.id));
break;
case FB_ACCEL_NEOMAGIC_NM2097:
- strscpy(info->fix.id, "Mag.Graph128ZV+", sizeof(info->fix.id));
+ strscpy(info->fix.id, "MagicGraph128ZV+", sizeof(info->fix.id));
break;
case FB_ACCEL_NEOMAGIC_NM2160:
strscpy(info->fix.id, "MagicGraph128XD", sizeof(info->fix.id));
```
Builds without error for me. Though I guess the compiler doesn't know
what strscpy is semantically.
Ah
MagicGraph128ZV+
MagicGraph256XL+
are both 17 characters. (16 literal characters + trailing NUL)
So MagicGraph256XL+ (and MagicGraph256AV+) below will fail to copy the
NUL from the source, then overwrite the `+` with NUL IIUC.
There's no ambiguity for MagicGraph256XL+, but now MagicGraph256AV+
would look like MagicGraph256AV which is another possible variant.
Then perhaps MagicGraph256XL+ MagicGraph256AV+ and below might need to change.
Sorry, initially it looked like a simple fix and I was ready to sign
off on it; I hope I'm mistaken. Thank you for the patch, but do
consider sending it to the ML for review before committing.
> break;
> case FB_ACCEL_NEOMAGIC_NM2160:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 128XD");
> + strscpy(info->fix.id, "MagicGraph128XD", sizeof(info->fix.id));
> break;
> case FB_ACCEL_NEOMAGIC_NM2200:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 256AV");
> + strscpy(info->fix.id, "MagicGraph256AV", sizeof(info->fix.id));
> info->flags |= FBINFO_HWACCEL_IMAGEBLIT |
> FBINFO_HWACCEL_COPYAREA |
> FBINFO_HWACCEL_FILLRECT;
> break;
> case FB_ACCEL_NEOMAGIC_NM2230:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 256AV+");
> + strscpy(info->fix.id, "MagicGraph256AV+", sizeof(info->fix.id));
> info->flags |= FBINFO_HWACCEL_IMAGEBLIT |
> FBINFO_HWACCEL_COPYAREA |
> FBINFO_HWACCEL_FILLRECT;
> break;
> case FB_ACCEL_NEOMAGIC_NM2360:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 256ZX");
> + strscpy(info->fix.id, "MagicGraph256ZX", sizeof(info->fix.id));
> info->flags |= FBINFO_HWACCEL_IMAGEBLIT |
> FBINFO_HWACCEL_COPYAREA |
> FBINFO_HWACCEL_FILLRECT;
> break;
> case FB_ACCEL_NEOMAGIC_NM2380:
> - snprintf(info->fix.id, sizeof(info->fix.id),
> - "MagicGraph 256XL+");
> + strscpy(info->fix.id, "MagicGraph256XL+", sizeof(info->fix.id));
> info->flags |= FBINFO_HWACCEL_IMAGEBLIT |
> FBINFO_HWACCEL_COPYAREA |
> FBINFO_HWACCEL_FILLRECT;
--
Thanks,
~Nick Desaulniers
Powered by blists - more mailing lists