mbox series

[00/99] fbdev: Fix memory leak in option parsing

Message ID 20230306160016.4459-1-tzimmermann@suse.de (mailing list archive)
Headers show
Series fbdev: Fix memory leak in option parsing | expand

Message

Thomas Zimmermann March 6, 2023, 3:58 p.m. UTC
Introduce struct option_iter and helpers to parse command-line
options with comma-separated key-value pairs. Then convert fbdev
drivers to the new interface. Fixes a memory leak in the parsing of
the video= option.

Before commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to
caller; clarify ownership"), a call to fb_get_options() either
returned an internal string or a duplicated string; hence ownership of
the string's memory buffer was not well defined, but depended on how
users specified the video= option on the kernel command line. For
global settings, the caller owned the returned memory and for per-driver
settings, fb_get_options() owned the memory. As calling drivers were
unable to detect the case, the memory was leaked.

Commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to caller;
clarify ownership") changed sematics to caller-owned strings. Drivers
still leaked the memory, but at least ownership was clear.

This patchset fixes the memory leak and changes string ownership back
to fb_get_options(). Patch 1 introduces struct option_iter and a few
helpers. The interface takes an option string, such as video=, in the
common form value1,key2:value2,value3 etc and returns the individial
comma-separated pairs. Various modules use this pattern, so the code
is located under lib/.

Patches 2 to 98 go through fbdev drivers and convert them to the new
interface. This often requires a number of cleanups. A driver would
typically refer to the option string's video mode. Such strings are now
copied to driver-allocated memory so that drivers don't refer directly
to the option string's memory. The option iterator then replaces manual
parsing loops based on strsep(",").

Patch 99 finally changes the ownership of the option string to be
internal to fb_get_option(); thereby fixing the memory leak. The option
iterator holds its own copy of the string and is not affected by the
change.

Most fbdev drivers only support to parse option strings if they are
built-in. I assume that's because of the original fuzzy semantics of
fb_get_options(). A later patchset could change the driver to respect
video= settings in any configuration.

Thomas Zimmermann (99):
  lib: Add option iterator
  fbdev/68328fb: Remove trailing whitespaces
  fbdev/68328fb: Remove unused option string
  fbdev/acornfb: Only init fb_info once
  fbdev/acornfb: Parse option string with struct option_iter
  fbdev/amifb: Duplicate video-mode option string
  fbdev/amifb: Parse option string with struct option_iter
  fbdev/arkfb: Duplicate video-mode option string
  fbdev/atafb: Duplicate video-mode option string
  fbdev/atafb: Parse option string with struct option_iter
  fbdev/aty: Duplicate video-mode option string
  fbdev/aty: Parse option string with struct option_iter
  fbdev/au1100fb: Parse option string with struct option_iter
  fbdev/au1200fb: Parse option string with struct option_iter
  fbdev/cirrusfb: Duplicate video-mode option string
  fbdev/cirrusfb: Parse option string with struct option_iter
  fbdev/controlfb: Remove trailing whitespaces
  fbdev/controlfb: Parse option string with struct option_iter
  fbdev/cyber2000fb: Parse option string with struct option_iter
  fbdev/efifb: Parse option string with struct option_iter
  fbdev/fm2fb: Parse option string with struct option_iter
  fbdev/fsl-diu-fb: Duplicate video-mode option string
  fbdev/fsl-diu-fb: Parse option string with struct option_iter
  fbdev/gbefb: Duplicate video-mode option string
  fbdev/gbefb: Parse option string with struct option_iter
  fbdev/geode: Duplicate video-mode option string
  fbdev/geode: Parse option string with struct option_iter
  fbdev/grvga: Duplicate video-mode option string
  fbdev/grvga: Parse option string with struct option_iter
  fbdev/gxt4500: Duplicate video-mode option string
  fbdev/hyperv_fb: Duplicate video-mode option string
  fbdev/i740fb: Duplicate video-mode option string
  fbdev/i740fb: Parse option string with struct option_iter
  fbdev/i810: Duplicate video-mode option string
  fbdev/i810: Parse option string with struct option_iter
  fbdev/imsttfb: Parse option string with struct option_iter
  fbdev/intelfb: Duplicate video-mode option string
  fbdev/intelfb: Parse option string with struct option_iter
  fbdev/imxfb: Duplicate video-mode option string
  fbdev/imxfb: Parse option string with struct option_iter
  fbdev/kyrofb: Duplicate video-mode option string
  fbdev/kyrofb: Parse option string with struct option_iter
  fbdev/macfb: Remove trailing whitespaces
  fbdev/macfb: Parse option string with struct option_iter
  fbdev/matroxfb: Parse option string with struct option_iter
  fbdev/mx3fb: Duplicate video-mode option string
  fbdev/mx3fb: Parse option string with struct option_iter
  fbdev/neofb: Duplicate video-mode option string
  fbdev/neofb: Parse option string with struct option_iter
  fbdev/nvidiafb: Duplicate video-mode option string
  fbdev/nvidiafb: Parse option string with struct option_iter
  fbdev/ocfb: Duplicate video-mode option string
  fbdev/ocfb: Parse option string with struct option_iter
  fbdev/omapfb: Parse option string with struct option_iter
  fbdev/platinumfb: Remove trailing whitespaces
  fbdev/platinumfb: Parse option string with struct option_iter
  fbdev/pm2fb: Duplicate video-mode option string
  fbdev/pm2fb: Parse option string with struct option_iter
  fbdev/pm3fb: Duplicate video-mode option string
  fbdev/pm3fb: Parse option string with struct option_iter
  fbdev/ps3fb: Duplicate video-mode option string
  fbdev/ps3fb: Parse option string with struct option_iter
  fbdev/pvr2fb: Duplicate video-mode option string
  fbdev/pvr2fb: Parse option string with struct option_iter
  fbdev/pxafb: Parse option string with struct option_iter
  fbdev/rivafb: Duplicate video-mode option string
  fbdev/rivafb: Parse option string with struct option_iter
  fbdev/s3fb: Duplicate video-mode option string
  fbdev/s3fb: Parse option string with struct option_iter
  fbdev/savagefb: Duplicate video-mode option string
  fbdev/savagefb: Parse option string with struct option_iter
  fbdev/sisfb: Constify mode string
  fbdev/sisfb: Parse option string with struct option_iter
  fbdev/skeletonfb: Parse option string with struct option_iter
  fbdev/sm712fb: Duplicate video-mode option string
  fbdev/sstfb: Duplicate video-mode option string
  fbdev/sstfb: Parse option string with struct option_iter
  fbdev/stifb: Remove trailing whitespaces
  fbdev/sti: Constify option string
  fbdev/tdfxfb: Duplicate video-mode option string
  fbdev/tdfxfb: Parse option string with struct option_iter
  fbdev/tgafb: Duplicate video-mode option string
  fbdev/tgafb: Parse option string with struct option_iter
  fbdev/tmiofb: Remove unused option string
  fbdev/tridentfb: Duplicate video-mode option string
  fbdev/tridentfb: Parse option string with struct option_iter
  fbdev/uvesafb: Duplicate video-mode option string
  fbdev/uvesafb: Parse option string with struct option_iter
  fbdev/valkyriefb: Remove trailing whitespaces
  fbdev/valkyriefb: Parse option string with struct option_iter
  fbdev/vermilion: Remove unused option string
  fbdev/vesafb: Parse option string with struct option_iter
  fbdev/vfb: Remove trailing whitespaces
  fbdev/vfb: Duplicate video-mode option string
  fbdev/vfb: Parse option string with struct option_iter
  fbdev/viafb: Parse option string with struct option_iter
  fbdev/vt8623fb: Duplicate video-mode option string
  staging/sm750fb: Parse option string with struct option_iter
  fbdev: Constify option strings

 Documentation/core-api/kernel-api.rst        |   9 ++
 drivers/staging/sm750fb/sm750.c              |  45 +++---
 drivers/video/fbdev/68328fb.c                |  24 +--
 drivers/video/fbdev/acornfb.c                |  25 ++-
 drivers/video/fbdev/amifb.c                  |  30 ++--
 drivers/video/fbdev/arkfb.c                  |  15 +-
 drivers/video/fbdev/atafb.c                  |  25 ++-
 drivers/video/fbdev/aty/aty128fb.c           |  28 +++-
 drivers/video/fbdev/aty/atyfb_base.c         |  29 +++-
 drivers/video/fbdev/aty/radeon_base.c        |  32 ++--
 drivers/video/fbdev/au1100fb.c               |  13 +-
 drivers/video/fbdev/au1200fb.c               |  15 +-
 drivers/video/fbdev/cirrusfb.c               |  37 +++--
 drivers/video/fbdev/controlfb.c              |  49 +++---
 drivers/video/fbdev/core/fb_cmdline.c        |  13 +-
 drivers/video/fbdev/core/modedb.c            |   8 +-
 drivers/video/fbdev/cyber2000fb.c            |  19 +--
 drivers/video/fbdev/efifb.c                  |  48 +++---
 drivers/video/fbdev/ep93xx-fb.c              |   2 +-
 drivers/video/fbdev/fm2fb.c                  |  16 +-
 drivers/video/fbdev/fsl-diu-fb.c             |  30 ++--
 drivers/video/fbdev/gbefb.c                  |  25 +--
 drivers/video/fbdev/geode/gx1fb_core.c       |  18 +--
 drivers/video/fbdev/geode/gxfb_core.c        |  25 +--
 drivers/video/fbdev/geode/lxfb_core.c        |  29 ++--
 drivers/video/fbdev/grvga.c                  |  32 +++-
 drivers/video/fbdev/gxt4500.c                |  18 ++-
 drivers/video/fbdev/hyperv_fb.c              |  18 ++-
 drivers/video/fbdev/i740fb.c                 |  32 ++--
 drivers/video/fbdev/i810/i810_main.c         |  30 +++-
 drivers/video/fbdev/imsttfb.c                |  18 ++-
 drivers/video/fbdev/imxfb.c                  |  23 ++-
 drivers/video/fbdev/intelfb/intelfbdrv.c     |  29 +++-
 drivers/video/fbdev/kyro/fbdev.c             |  27 ++--
 drivers/video/fbdev/macfb.c                  |  28 ++--
 drivers/video/fbdev/matrox/matroxfb_base.c   |  21 +--
 drivers/video/fbdev/mx3fb.c                  |  27 +++-
 drivers/video/fbdev/neofb.c                  |  32 ++--
 drivers/video/fbdev/nvidia/nvidia.c          |  32 ++--
 drivers/video/fbdev/ocfb.c                   |  25 ++-
 drivers/video/fbdev/omap/omapfb_main.c       |  15 +-
 drivers/video/fbdev/platinumfb.c             |  46 +++---
 drivers/video/fbdev/pm2fb.c                  |  31 ++--
 drivers/video/fbdev/pm3fb.c                  |  33 ++--
 drivers/video/fbdev/ps3fb.c                  |  32 ++--
 drivers/video/fbdev/pvr2fb.c                 |  29 ++--
 drivers/video/fbdev/pxafb.c                  |  20 ++-
 drivers/video/fbdev/riva/fbdev.c             |  32 ++--
 drivers/video/fbdev/s3fb.c                   |  33 ++--
 drivers/video/fbdev/savage/savagefb_driver.c |  26 ++-
 drivers/video/fbdev/sis/sis_main.c           |  26 +--
 drivers/video/fbdev/skeletonfb.c             |  17 +-
 drivers/video/fbdev/sm712fb.c                |  19 ++-
 drivers/video/fbdev/sstfb.c                  |  31 ++--
 drivers/video/fbdev/stifb.c                  | 162 +++++++++----------
 drivers/video/fbdev/tdfxfb.c                 |  27 ++--
 drivers/video/fbdev/tgafb.c                  |  35 ++--
 drivers/video/fbdev/tmiofb.c                 |  24 +--
 drivers/video/fbdev/tridentfb.c              |  33 ++--
 drivers/video/fbdev/uvesafb.c                |  27 ++--
 drivers/video/fbdev/valkyriefb.c             |  32 ++--
 drivers/video/fbdev/vermilion/vermilion.c    |   7 +-
 drivers/video/fbdev/vesafb.c                 |  18 ++-
 drivers/video/fbdev/vfb.c                    |  41 +++--
 drivers/video/fbdev/via/viafbdev.c           |  17 +-
 drivers/video/fbdev/vt8623fb.c               |  16 +-
 include/linux/cmdline.h                      |  29 ++++
 include/linux/fb.h                           |   2 +-
 lib/Makefile                                 |   2 +-
 lib/cmdline_iter.c                           |  97 +++++++++++
 70 files changed, 1252 insertions(+), 708 deletions(-)
 create mode 100644 include/linux/cmdline.h
 create mode 100644 lib/cmdline_iter.c

Comments

Geert Uytterhoeven March 7, 2023, 7:53 a.m. UTC | #1
Hi Thomas,

On Mon, Mar 6, 2023 at 5:00 PM Thomas Zimmermann <tzimmermann@suse.de> wrote:
> Introduce struct option_iter and helpers to parse command-line
> options with comma-separated key-value pairs. Then convert fbdev
> drivers to the new interface. Fixes a memory leak in the parsing of
> the video= option.
>
> Before commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to
> caller; clarify ownership"), a call to fb_get_options() either
> returned an internal string or a duplicated string; hence ownership of
> the string's memory buffer was not well defined, but depended on how
> users specified the video= option on the kernel command line. For
> global settings, the caller owned the returned memory and for per-driver
> settings, fb_get_options() owned the memory. As calling drivers were
> unable to detect the case, the memory was leaked.
>
> Commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to caller;
> clarify ownership") changed sematics to caller-owned strings. Drivers
> still leaked the memory, but at least ownership was clear.
>
> This patchset fixes the memory leak and changes string ownership back
> to fb_get_options(). Patch 1 introduces struct option_iter and a few
> helpers. The interface takes an option string, such as video=, in the
> common form value1,key2:value2,value3 etc and returns the individial
> comma-separated pairs. Various modules use this pattern, so the code
> is located under lib/.
>
> Patches 2 to 98 go through fbdev drivers and convert them to the new
> interface. This often requires a number of cleanups. A driver would
> typically refer to the option string's video mode. Such strings are now
> copied to driver-allocated memory so that drivers don't refer directly
> to the option string's memory. The option iterator then replaces manual
> parsing loops based on strsep(",").

Thanks for your series!

Unfortunately I cannot say I'm thrilled about this: you are replacing
a single small dynamic memory leak by 36 larger static memory leaks.
Am I missing something?

Gr{oetje,eeting}s,

                        Geert
Thomas Zimmermann March 7, 2023, 8:23 a.m. UTC | #2
Hi Geert

Am 07.03.23 um 08:53 schrieb Geert Uytterhoeven:
> Hi Thomas,
> 
> On Mon, Mar 6, 2023 at 5:00 PM Thomas Zimmermann <tzimmermann@suse.de> wrote:
>> Introduce struct option_iter and helpers to parse command-line
>> options with comma-separated key-value pairs. Then convert fbdev
>> drivers to the new interface. Fixes a memory leak in the parsing of
>> the video= option.
>>
>> Before commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to
>> caller; clarify ownership"), a call to fb_get_options() either
>> returned an internal string or a duplicated string; hence ownership of
>> the string's memory buffer was not well defined, but depended on how
>> users specified the video= option on the kernel command line. For
>> global settings, the caller owned the returned memory and for per-driver
>> settings, fb_get_options() owned the memory. As calling drivers were
>> unable to detect the case, the memory was leaked.
>>
>> Commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to caller;
>> clarify ownership") changed sematics to caller-owned strings. Drivers
>> still leaked the memory, but at least ownership was clear.
>>
>> This patchset fixes the memory leak and changes string ownership back
>> to fb_get_options(). Patch 1 introduces struct option_iter and a few
>> helpers. The interface takes an option string, such as video=, in the
>> common form value1,key2:value2,value3 etc and returns the individial
>> comma-separated pairs. Various modules use this pattern, so the code
>> is located under lib/.
>>
>> Patches 2 to 98 go through fbdev drivers and convert them to the new
>> interface. This often requires a number of cleanups. A driver would
>> typically refer to the option string's video mode. Such strings are now
>> copied to driver-allocated memory so that drivers don't refer directly
>> to the option string's memory. The option iterator then replaces manual
>> parsing loops based on strsep(",").
> 
> Thanks for your series!
> 
> Unfortunately I cannot say I'm thrilled about this: you are replacing
> a single small dynamic memory leak by 36 larger static memory leaks.

That's fair enough.

> Am I missing something?

The current size of the videomode buffers is ridiculously large. I just 
needed something that could hold the string. A long mode description 
might look like

   1920x1080MR-32@120ime

which has 21 characters. 32-byte buffers would probably be more than enough.

I think it should also be possible to do a simple kstrdup() on the given 
videomode string and free the copy in the module's _fini function.

That also brings up the question of these MODULE ifdefs. Almost all of 
the fbdev drivers only parse the command-line option if they are not 
build as a module. Do you know why? Because of the awkward semantics of 
the old fb_get_options()?  I think this should be changed so that they 
always respect the video= parameter.

Best regards
Thomas

> 
> Gr{oetje,eeting}s,
> 
>                          Geert
>
Geert Uytterhoeven March 7, 2023, 8:57 a.m. UTC | #3
Hi Thomas,

On Tue, Mar 7, 2023 at 9:23 AM Thomas Zimmermann <tzimmermann@suse.de> wrote:
> Am 07.03.23 um 08:53 schrieb Geert Uytterhoeven:
> > On Mon, Mar 6, 2023 at 5:00 PM Thomas Zimmermann <tzimmermann@suse.de> wrote:
> >> Introduce struct option_iter and helpers to parse command-line
> >> options with comma-separated key-value pairs. Then convert fbdev
> >> drivers to the new interface. Fixes a memory leak in the parsing of
> >> the video= option.
> >>
> >> Before commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to
> >> caller; clarify ownership"), a call to fb_get_options() either
> >> returned an internal string or a duplicated string; hence ownership of
> >> the string's memory buffer was not well defined, but depended on how
> >> users specified the video= option on the kernel command line. For
> >> global settings, the caller owned the returned memory and for per-driver
> >> settings, fb_get_options() owned the memory. As calling drivers were
> >> unable to detect the case, the memory was leaked.
> >>
> >> Commit 73ce73c30ba9 ("fbdev: Transfer video= option strings to caller;
> >> clarify ownership") changed sematics to caller-owned strings. Drivers
> >> still leaked the memory, but at least ownership was clear.
> >>
> >> This patchset fixes the memory leak and changes string ownership back
> >> to fb_get_options(). Patch 1 introduces struct option_iter and a few
> >> helpers. The interface takes an option string, such as video=, in the
> >> common form value1,key2:value2,value3 etc and returns the individial
> >> comma-separated pairs. Various modules use this pattern, so the code
> >> is located under lib/.
> >>
> >> Patches 2 to 98 go through fbdev drivers and convert them to the new
> >> interface. This often requires a number of cleanups. A driver would
> >> typically refer to the option string's video mode. Such strings are now
> >> copied to driver-allocated memory so that drivers don't refer directly
> >> to the option string's memory. The option iterator then replaces manual
> >> parsing loops based on strsep(",").
> >
> > Thanks for your series!
> >
> > Unfortunately I cannot say I'm thrilled about this: you are replacing
> > a single small dynamic memory leak by 36 larger static memory leaks.
>
> That's fair enough.
>
> > Am I missing something?
>
> The current size of the videomode buffers is ridiculously large. I just
> needed something that could hold the string. A long mode description
> might look like
>
>    1920x1080MR-32@120ime
>
> which has 21 characters. 32-byte buffers would probably be more than enough.

But there are a few exceptions...

> I think it should also be possible to do a simple kstrdup() on the given
> videomode string and free the copy in the module's _fini function.

That sounds like the sanest approach to me.

> That also brings up the question of these MODULE ifdefs. Almost all of
> the fbdev drivers only parse the command-line option if they are not
> build as a module. Do you know why? Because of the awkward semantics of
> the old fb_get_options()?

That's just historical: to get to see anything on the console (on
the graphics hardware without VGA text mode that fbdev was originally
developed for), you needed to have your main fbdev driver builtin.
Drivers for secondary displays could be loadable modules, and using
fbset for those offered more flexibility than a module parameter.

> I think this should be changed so that they
> always respect the video= parameter.

I agree that makes sense.

Gr{oetje,eeting}s,

                        Geert