| Message ID | 20180912073046.26475-1-tomi.valkeinen@ti.com |
|---|---|
| State | New |
| Headers | show |
| Series | fbdev/omapfb: fix omapfb_memory_read infoleak | expand |
On Wed, Sep 12, 2018 at 12:30 AM, Tomi Valkeinen <tomi.valkeinen@ti.com> wrote: > OMAPFB_MEMORY_READ ioctl reads pixels from the LCD's memory and copies > them to a userspace buffer. The code has two issues: > > - The user provided width and height could be large enough to overflow > the calculations > - The copy_to_user() can copy uninitialized memory to the userspace, > which might contain sensitive kernel information. > > Fix these by limiting the width & height parameters, and only copying > the amount of data that we actually received from the LCD. > > Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com> > Reported-by: Jann Horn <jannh@google.com> > Cc: stable@vger.kernel.org > Cc: security@kernel.org > Cc: Will Deacon <will.deacon@arm.com> > Cc: Jann Horn <jannh@google.com> > Cc: Tony Lindgren <tony@atomide.com> > --- > drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c b/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c > index ef69273074ba..a3edb20ea4c3 100644 > --- a/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c > +++ b/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c > @@ -496,6 +496,9 @@ static int omapfb_memory_read(struct fb_info *fbi, > if (!access_ok(VERIFY_WRITE, mr->buffer, mr->buffer_size)) > return -EFAULT; > > + if (mr->w > 4096 || mr->h > 4096) > + return -EINVAL; > + > if (mr->w * mr->h * 3 > mr->buffer_size) > return -EINVAL; alternatively, replace the above two tests with: alloc_size = array3_size(mr->w, mr->h, 3); if (alloc_size > mr->buffer_size) return -EINVAL; buf = vmalloc(alloc_size); ... > > @@ -509,7 +512,7 @@ static int omapfb_memory_read(struct fb_info *fbi, > mr->x, mr->y, mr->w, mr->h); > > if (r > 0) { > - if (copy_to_user(mr->buffer, buf, mr->buffer_size)) > + if (copy_to_user(mr->buffer, buf, r)) But yes, this seems correct regardless: userspace was being overwritten beyond "r", potentially. -Kees > r = -EFAULT; > } > > -- > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. > Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki > -- Kees Cook Pixel Security
[ added dri-devel@lists.freedesktop.org to Cc: ] On 09/12/2018 09:30 AM, Tomi Valkeinen wrote: > OMAPFB_MEMORY_READ ioctl reads pixels from the LCD's memory and copies > them to a userspace buffer. The code has two issues: > > - The user provided width and height could be large enough to overflow > the calculations > - The copy_to_user() can copy uninitialized memory to the userspace, > which might contain sensitive kernel information. > > Fix these by limiting the width & height parameters, and only copying > the amount of data that we actually received from the LCD. > > Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com> > Reported-by: Jann Horn <jannh@google.com> > Cc: stable@vger.kernel.org > Cc: security@kernel.org > Cc: Will Deacon <will.deacon@arm.com> > Cc: Jann Horn <jannh@google.com> > Cc: Tony Lindgren <tony@atomide.com> Patch queued for 4.19, thanks. > --- > drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c b/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c > index ef69273074ba..a3edb20ea4c3 100644 > --- a/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c > +++ b/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c > @@ -496,6 +496,9 @@ static int omapfb_memory_read(struct fb_info *fbi, > if (!access_ok(VERIFY_WRITE, mr->buffer, mr->buffer_size)) > return -EFAULT; > > + if (mr->w > 4096 || mr->h > 4096) > + return -EINVAL; > + > if (mr->w * mr->h * 3 > mr->buffer_size) > return -EINVAL; > > @@ -509,7 +512,7 @@ static int omapfb_memory_read(struct fb_info *fbi, > mr->x, mr->y, mr->w, mr->h); > > if (r > 0) { > - if (copy_to_user(mr->buffer, buf, mr->buffer_size)) > + if (copy_to_user(mr->buffer, buf, r)) > r = -EFAULT; > } Best regards, -- Bartlomiej Zolnierkiewicz Samsung R&D Institute Poland Samsung Electronics
diff --git a/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c b/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c index ef69273074ba..a3edb20ea4c3 100644 --- a/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c +++ b/drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c @@ -496,6 +496,9 @@ static int omapfb_memory_read(struct fb_info *fbi, if (!access_ok(VERIFY_WRITE, mr->buffer, mr->buffer_size)) return -EFAULT; + if (mr->w > 4096 || mr->h > 4096) + return -EINVAL; + if (mr->w * mr->h * 3 > mr->buffer_size) return -EINVAL; @@ -509,7 +512,7 @@ static int omapfb_memory_read(struct fb_info *fbi, mr->x, mr->y, mr->w, mr->h); if (r > 0) { - if (copy_to_user(mr->buffer, buf, mr->buffer_size)) + if (copy_to_user(mr->buffer, buf, r)) r = -EFAULT; }
OMAPFB_MEMORY_READ ioctl reads pixels from the LCD's memory and copies them to a userspace buffer. The code has two issues: - The user provided width and height could be large enough to overflow the calculations - The copy_to_user() can copy uninitialized memory to the userspace, which might contain sensitive kernel information. Fix these by limiting the width & height parameters, and only copying the amount of data that we actually received from the LCD. Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com> Reported-by: Jann Horn <jannh@google.com> Cc: stable@vger.kernel.org Cc: security@kernel.org Cc: Will Deacon <will.deacon@arm.com> Cc: Jann Horn <jannh@google.com> Cc: Tony Lindgren <tony@atomide.com> --- drivers/video/fbdev/omap2/omapfb/omapfb-ioctl.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) -- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki