Skip to contentSkip to conversation

drm/panic: Fixes and graphical logo

PATCH In review Geert Uytterhoeven 4 participants 25 messages Jun 13, 2024 – Jun 24, 2024

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

v2 cover drm/panic: Fixes and graphical logo

Hi all,

If drm/panic is enabled, a user-friendly message is shown on screen when a kernel panic occurs, together with an ASCII art penguin logo. Of course we can do better ;-)
Hence this patch series extends drm/panic to draw the monochrome graphical boot logo, when available, preceded by the customary fixes.

Changes compared to v1:
- Rebase against today's drm-misc-next, where drm_panic is broken on all current drivers due to an uninitialized pointer dereference. Presumably this was only tested with an out-of-tree driver change?
- New fixes [1/7], [3/7], and [4/7],
- New cleanup [5/7],
- Inline trivial draw_logo_mono().

This has been tested with rcar-du.

Thanks for your comments!

Geert Uytterhoeven (7):
drm/panic: Fix uninitialized drm_scanout_buffer.set_pixel() crash drm/panic: Fix off-by-one logo size checks
lib/fonts: Fix visiblity of SUN12x22 and TER16x32 if DRM_PANIC drm/panic: Spelling s/formater/formatter/
drm/panic: Convert to drm_fb_clip_offset()
drm/panic: Rename logo to logo_ascii
drm/panic: Add support for drawing a monochrome graphical logo

 drivers/gpu/drm/Kconfig     |  2 +-
 drivers/gpu/drm/drm_panic.c | 74 +++++++++++++++++++++++++++++++------
 drivers/video/logo/Kconfig  |  2 +
 lib/fonts/Kconfig           |  6 ++-
 4 files changed, 70 insertions(+), 14 deletions(-)
Signature · 12 lines
2.34.1

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds
#1 lore raw

PATCH v2 1/7 drm/panic: Fix uninitialized drm_scanout_buffer.set_pixel() crash

No implementations of drm_plane_helper_funcs.get_scanout_buffer() fill in the optional drm_scanout_buffer.set_pixel() member. Hence the member may contain non-zero garbage, causing a crash when deferencing it during drm panic.

Fix this by pre-initializing the drm_scanout_buffer object before calling drm_plane_helper_funcs.get_scanout_buffer().

Fixes 24d07f114e4ec760 ("drm/panic: Add a set_pixel() callback to drm_scanout_buffer")

---
v2:
- New.

---
 drivers/gpu/drm/drm_panic.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/drm_panic.c +1 -1
index 293d4dcbc80da7ba..fc04ed4e0b399f55 100644
@@ -582,7 +582,7 @@ static void draw_panic_dispatch(struct drm_scanout_buffer *sb)
 
 static void draw_panic_plane(struct drm_plane *plane)
 {
-	struct drm_scanout_buffer sb;
+	struct drm_scanout_buffer sb = { };
 	int ret;
 	unsigned long flags;
 
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 1/7
#2 lore raw

Re: [PATCH v2 1/7] drm/panic: Fix uninitialized drm_scanout_buffer.set_pixel() crash

Quoted 7 lines · Geert Uytterhoeven

No implementations of drm_plane_helper_funcs.get_scanout_buffer() fill in the optional drm_scanout_buffer.set_pixel() member. Hence the member may contain non-zero garbage, causing a crash when deferencing it during drm panic.

Fix this by pre-initializing the drm_scanout_buffer object before calling drm_plane_helper_funcs.get_scanout_buffer().

Good catch, I don't know how I didn't hit this bug before. Thanks for the fix.

Quoted 23 lines
Fixes 24d07f114e4ec760 ("drm/panic: Add a set_pixel() callback to drm_scanout_buffer")

---
v2:
- New.

---
  drivers/gpu/drm/drm_panic.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/drm_panic.c +1 -1
index 293d4dcbc80da7ba..fc04ed4e0b399f55 100644
@@ -582,7 +582,7 @@ static void draw_panic_dispatch(struct drm_scanout_buffer *sb)
 
  static void draw_panic_plane(struct drm_plane *plane)
  {
-	struct drm_scanout_buffer sb;
+	struct drm_scanout_buffer sb = { };
  	int ret;
  	unsigned long flags;
 
#3 lore raw

PATCH v2 2/7 drm/panic: Fix off-by-one logo size checks

In reply to Geert Uytterhoeven

Logos that are either just as wide or just as high as the display work fine.

Fixes bf9fb17c6672868d ("drm/panic: Add a drm panic handler")

---
v2:
- Rebased.

---
 drivers/gpu/drm/drm_panic.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/drm_panic.c +1 -1
index fc04ed4e0b399f55..814ef5c20c08ee42 100644
@@ -472,7 +472,7 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 	drm_panic_fill(sb, &r_screen, bg_color);
 
 	if ((r_msg.x1 >= drm_rect_width(&r_logo) || r_msg.y1 >= drm_rect_height(&r_logo)) &&
-	    drm_rect_width(&r_logo) < sb->width && drm_rect_height(&r_logo) < sb->height) {
+	    drm_rect_width(&r_logo) <= sb->width && drm_rect_height(&r_logo) <= sb->height) {
 		draw_txt_rectangle(sb, font, logo, logo_lines, false, &r_logo, fg_color);
 	}
 	draw_txt_rectangle(sb, font, panic_msg, msg_lines, true, &r_msg, fg_color);
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 2/7
#4 lore raw

Re: [PATCH v2 2/7] drm/panic: Fix off-by-one logo size checks

Quoted 2 lines · Geert Uytterhoeven

Logos that are either just as wide or just as high as the display work fine.

Sure, that looks good to me.

Quoted 23 lines
Fixes bf9fb17c6672868d ("drm/panic: Add a drm panic handler")

---
v2:
- Rebased.

---
  drivers/gpu/drm/drm_panic.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/drm_panic.c +1 -1
index fc04ed4e0b399f55..814ef5c20c08ee42 100644
@@ -472,7 +472,7 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
  	drm_panic_fill(sb, &r_screen, bg_color);
 
  	if ((r_msg.x1 >= drm_rect_width(&r_logo) || r_msg.y1 >= drm_rect_height(&r_logo)) &&
-	    drm_rect_width(&r_logo) < sb->width && drm_rect_height(&r_logo) < sb->height) {
+	    drm_rect_width(&r_logo) <= sb->width && drm_rect_height(&r_logo) <= sb->height) {
  		draw_txt_rectangle(sb, font, logo, logo_lines, false, &r_logo, fg_color);
  	}
  	draw_txt_rectangle(sb, font, panic_msg, msg_lines, true, &r_msg, fg_color);
#5 lore raw

PATCH v2 3/7 lib/fonts: Fix visiblity of SUN12x22 and TER16x32 if DRM_PANIC

In reply to Geert Uytterhoeven

When CONFIG_FONTS ("Select compiled-in fonts") is not enabled, the user should not be asked about any fonts. However, when CONFIG_DRM_PANIC is enabled, the user is still asked about the Sparc console 12x22 and Terminus 16x32 fonts.

Fix this by moving the "|| DRM_PANIC" to where it belongs. Split the dependency in two rules to improve readability.

Fixes b94605a3889b9084 ("lib/fonts: Allow to select fonts for drm_panic")

---
v2:
- New.

---
 lib/fonts/Kconfig | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
lib/fonts/Kconfig +4 -2
index befcb463f7381d1a..3ac26bdbc3ff01a3 100644
@@ -105,7 +105,8 @@ config FONT_SUN8x16
 
 config FONT_SUN12x22
 	bool "Sparc console 12x22 font (not supported by all drivers)"
-	depends on (FRAMEBUFFER_CONSOLE && (!SPARC && FONTS || SPARC)) || DRM_PANIC
+	depends on FRAMEBUFFER_CONSOLE || DRM_PANIC
+	depends on !SPARC && FONTS
 	help
 	  This is the high resolution console font for Sun machines with very
 	  big letters (like the letters used in the SPARC PROM). If the
@@ -113,7 +114,8 @@ config FONT_SUN12x22
 
 config FONT_TER16x32
 	bool "Terminus 16x32 font (not supported by all drivers)"
-	depends on (FRAMEBUFFER_CONSOLE && (!SPARC && FONTS || SPARC)) || DRM_PANIC
+	depends on FRAMEBUFFER_CONSOLE || DRM_PANIC
+	depends on !SPARC && FONTS || SPARC
 	help
 	  Terminus Font is a clean, fixed width bitmap font, designed
 	  for long (8 and more hours per day) work with computers.
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 3/7
#6 lore raw

Re: [PATCH v2 3/7] lib/fonts: Fix visiblity of SUN12x22 and TER16x32 if DRM_PANIC

Quoted 7 lines · Geert Uytterhoeven

When CONFIG_FONTS ("Select compiled-in fonts") is not enabled, the user should not be asked about any fonts. However, when CONFIG_DRM_PANIC is enabled, the user is still asked about the Sparc console 12x22 and Terminus 16x32 fonts.

Fix this by moving the "|| DRM_PANIC" to where it belongs. Split the dependency in two rules to improve readability.

Sorry I think I misunderstood the SPARC && FONTS condition. Your fix is much clearer, thanks.

Quoted 34 lines
Fixes b94605a3889b9084 ("lib/fonts: Allow to select fonts for drm_panic")

---
v2:
- New.

---
  lib/fonts/Kconfig | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
lib/fonts/Kconfig +4 -2
index befcb463f7381d1a..3ac26bdbc3ff01a3 100644
@@ -105,7 +105,8 @@ config FONT_SUN8x16
 
  config FONT_SUN12x22
  	bool "Sparc console 12x22 font (not supported by all drivers)"
-	depends on (FRAMEBUFFER_CONSOLE && (!SPARC && FONTS || SPARC)) || DRM_PANIC
+	depends on FRAMEBUFFER_CONSOLE || DRM_PANIC
+	depends on !SPARC && FONTS
  	help
  	  This is the high resolution console font for Sun machines with very
  	  big letters (like the letters used in the SPARC PROM). If the
@@ -113,7 +114,8 @@ config FONT_SUN12x22
 
  config FONT_TER16x32
  	bool "Terminus 16x32 font (not supported by all drivers)"
-	depends on (FRAMEBUFFER_CONSOLE && (!SPARC && FONTS || SPARC)) || DRM_PANIC
+	depends on FRAMEBUFFER_CONSOLE || DRM_PANIC
+	depends on !SPARC && FONTS || SPARC
  	help
  	  Terminus Font is a clean, fixed width bitmap font, designed
  	  for long (8 and more hours per day) work with computers.
#7 lore raw

PATCH v2 4/7 drm/panic: Spelling s/formater/formatter/

In reply to Geert Uytterhoeven

Fix a misspelling of "formatter".

Fixes 54034bebb22fd4be ("drm/panic: Add a kmsg panic screen")

---
v2:
- New.

---
 drivers/gpu/drm/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/Kconfig +1 -1
index a9972ce05d7e6fe4..e3c51009d9b476b3 100644
@@ -138,7 +138,7 @@ config DRM_PANIC_DEBUG
 	  If in doubt, say "N".
 
 config DRM_PANIC_SCREEN
-	string "Panic screen formater"
+	string "Panic screen formatter"
 	default "user"
 	depends on DRM_PANIC
 	help
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 4/7
#8 lore raw

Re: [PATCH v2 4/7] drm/panic: Spelling s/formater/formatter/

Quoted 1 line · Geert Uytterhoeven

Fix a misspelling of "formatter".

It's probably because the same word has only one t in my native language. Thanks for the fix.

Quoted 23 lines
Fixes 54034bebb22fd4be ("drm/panic: Add a kmsg panic screen")

---
v2:
- New.

---
  drivers/gpu/drm/Kconfig | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/Kconfig +1 -1
index a9972ce05d7e6fe4..e3c51009d9b476b3 100644
@@ -138,7 +138,7 @@ config DRM_PANIC_DEBUG
  	  If in doubt, say "N".
 
  config DRM_PANIC_SCREEN
-	string "Panic screen formater"
+	string "Panic screen formatter"
  	default "user"
  	depends on DRM_PANIC
  	help
#9 lore raw

PATCH v2 5/7 drm/panic: Convert to drm_fb_clip_offset()

In reply to Geert Uytterhoeven

Use the drm_fb_clip_offset() helper instead of open-coding the same operation.

---
DRM_PANIC already selects DRM_KMS_HELPER.

v2:
- New.

---
 drivers/gpu/drm/drm_panic.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/drm_panic.c +2 -2
index 814ef5c20c08ee42..5b0acf8c86e402a8 100644
@@ -285,7 +285,7 @@ static void drm_panic_blit(struct drm_scanout_buffer *sb, struct drm_rect *clip,
 		return drm_panic_blit_pixel(sb, clip, sbuf8, spitch, fg_color);
 
 	map = sb->map[0];
-	iosys_map_incr(&map, clip->y1 * sb->pitch[0] + clip->x1 * sb->format->cpp[0]);
+	iosys_map_incr(&map, drm_fb_clip_offset(sb->pitch[0], sb->format, clip));
 
 	switch (sb->format->cpp[0]) {
 	case 2:
@@ -373,7 +373,7 @@ static void drm_panic_fill(struct drm_scanout_buffer *sb, struct drm_rect *clip,
 		return drm_panic_fill_pixel(sb, clip, color);
 
 	map = sb->map[0];
-	iosys_map_incr(&map, clip->y1 * sb->pitch[0] + clip->x1 * sb->format->cpp[0]);
+	iosys_map_incr(&map, drm_fb_clip_offset(sb->pitch[0], sb->format, clip));
 
 	switch (sb->format->cpp[0]) {
 	case 2:
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 5/7
#10 lore raw

Re: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()

Quoted 2 lines · Geert Uytterhoeven

Use the drm_fb_clip_offset() helper instead of open-coding the same operation.

Thanks, it's a nice cleanup.

Quoted 33 lines

---
DRM_PANIC already selects DRM_KMS_HELPER.

v2:
- New.

---
  drivers/gpu/drm/drm_panic.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/drm_panic.c +2 -2
index 814ef5c20c08ee42..5b0acf8c86e402a8 100644
@@ -285,7 +285,7 @@ static void drm_panic_blit(struct drm_scanout_buffer *sb, struct drm_rect *clip,
  		return drm_panic_blit_pixel(sb, clip, sbuf8, spitch, fg_color);
 
  	map = sb->map[0];
-	iosys_map_incr(&map, clip->y1 * sb->pitch[0] + clip->x1 * sb->format->cpp[0]);
+	iosys_map_incr(&map, drm_fb_clip_offset(sb->pitch[0], sb->format, clip));
 
  	switch (sb->format->cpp[0]) {
  	case 2:
@@ -373,7 +373,7 @@ static void drm_panic_fill(struct drm_scanout_buffer *sb, struct drm_rect *clip,
  		return drm_panic_fill_pixel(sb, clip, color);
 
  	map = sb->map[0];
-	iosys_map_incr(&map, clip->y1 * sb->pitch[0] + clip->x1 * sb->format->cpp[0]);
+	iosys_map_incr(&map, drm_fb_clip_offset(sb->pitch[0], sb->format, clip));
 
  	switch (sb->format->cpp[0]) {
  	case 2:
#11 lore raw

Re: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()

In reply to Geert Uytterhoeven
kernel test robot noticed the following build errors:

Hi Geert,

kernel test robot noticed the following build errors:

[auto build test ERROR on drm-misc/drm-misc-next]
[cannot apply to linus/master v6.10-rc3 next-20240613]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Geert-Uytterhoeven/drm-panic-Fix-uninitialized-drm_scanout_buffer-set_pixel-crash/20240614-032053
base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
patch link:    https://lore.kernel.org/r/3121082eb4beb461773ebb6f656ed9b4286967ee.1718305355.git.geert%2Brenesas%40glider.be
patch subject: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()
config: x86_64-randconfig-003-20240615 (https://download.01.org/0day-ci/archive/20240615/[email protected]/config)
compiler: gcc-13 (Ubuntu 13.2.0-4ubuntu3) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240615/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

>> depmod: ERROR: Cycle detected: drm -> drm_kms_helper -> drm

depmod: ERROR: Found 2 modules in dependency cycles!
make[3]: *** [scripts/Makefile.modinst:128: depmod] Error 1 shuffle=844234264 make[3]: Target '__modinst' not remade because of errors. make[2]: *** [Makefile:1842: modules_install] Error 2 shuffle=844234264 make[1]: *** [Makefile:240: __sub-make] Error 2 shuffle=844234264 make[1]: Target 'modules_install' not remade because of errors.
make: *** [Makefile:240: __sub-make] Error 2 shuffle=844234264
make: Target 'modules_install' not remade because of errors.

Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#12 lore raw

Re: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()

Quoted 21 lines · kernel test robot

kernel test robot noticed the following build errors:

[auto build test ERROR on drm-misc/drm-misc-next]
[cannot apply to linus/master v6.10-rc3 next-20240613]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Geert-Uytterhoeven/drm-panic-Fix-uninitialized-drm_scanout_buffer-set_pixel-crash/20240614-032053
base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
patch link:    https://lore.kernel.org/r/3121082eb4beb461773ebb6f656ed9b4286967ee.1718305355.git.geert%2Brenesas%40glider.be
patch subject: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()
config: x86_64-randconfig-003-20240615 (https://download.01.org/0day-ci/archive/20240615/[email protected]/config)
compiler: gcc-13 (Ubuntu 13.2.0-4ubuntu3) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240615/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

Quoted 4 lines

All errors (new ones prefixed by >>):

Quoted 1 line
Quoted 1 line

depmod: ERROR: Cycle detected: drm -> drm_kms_helper -> drm

depmod: ERROR: Found 2 modules in dependency cycles!

Oops, so DRM core cannot call any of the helpers, and DRM_PANIC selecting DRM_KMS_HELPER was wrong in the first place?

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
#13 lore raw

Re: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()

Quoted 22 lines · Geert Uytterhoeven
Quoted 21 lines · kernel test robot

kernel test robot noticed the following build errors:

[auto build test ERROR on drm-misc/drm-misc-next]
[cannot apply to linus/master v6.10-rc3 next-20240613]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Geert-Uytterhoeven/drm-panic-Fix-uninitialized-drm_scanout_buffer-set_pixel-crash/20240614-032053
base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
patch link:    https://lore.kernel.org/r/3121082eb4beb461773ebb6f656ed9b4286967ee.1718305355.git.geert%2Brenesas%40glider.be
patch subject: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()
config: x86_64-randconfig-003-20240615 (https://download.01.org/0day-ci/archive/20240615/[email protected]/config)
compiler: gcc-13 (Ubuntu 13.2.0-4ubuntu3) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240615/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

Quoted 7 lines
Quoted 4 lines

All errors (new ones prefixed by >>):

Quoted 1 line
Quoted 1 line

depmod: ERROR: Cycle detected: drm -> drm_kms_helper -> drm

depmod: ERROR: Found 2 modules in dependency cycles!

Oops, so DRM core cannot call any of the helpers, and DRM_PANIC selecting DRM_KMS_HELPER was wrong in the first place?

Q: So how does this work with DRM_PANIC calling
drm_fb_helper_emergency_disable()?
A: drm_fb_helper_emergency_disable() is a dummy if
!CONFIG_DRM_FBDEV_EMULATION, so I guess no one tried to build a failing randconfig with CONFIG_DRM_FBDEV_EMULATION=y yet.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
maintainer
#14 lore raw

Re: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()

Quoted 29 lines · Geert Uytterhoeven
Quoted 27 lines · Geert Uytterhoeven
Quoted 25 lines · kernel test robot

kernel test robot noticed the following build errors:

[auto build test ERROR on drm-misc/drm-misc-next]
[cannot apply to linus/master v6.10-rc3 next-20240613]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Geert-Uytterhoeven/drm-panic-Fix-uninitialized-drm_scanout_buffer-set_pixel-crash/20240614-032053
base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
patch link:    https://lore.kernel.org/r/3121082eb4beb461773ebb6f656ed9b4286967ee.1718305355.git.geert%2Brenesas%40glider.be
patch subject: [PATCH v2 5/7] drm/panic: Convert to drm_fb_clip_offset()
config: x86_64-randconfig-003-20240615 (https://download.01.org/0day-ci/archive/20240615/[email protected]/config)
compiler: gcc-13 (Ubuntu 13.2.0-4ubuntu3) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240615/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

Quoted 1 line
Quoted 1 line

depmod: ERROR: Cycle detected: drm -> drm_kms_helper -> drm

depmod: ERROR: Found 2 modules in dependency cycles!

Quoted 8 lines
Quoted 2 lines

Oops, so DRM core cannot call any of the helpers, and DRM_PANIC selecting DRM_KMS_HELPER was wrong in the first place?

Q: So how does this work with DRM_PANIC calling
drm_fb_helper_emergency_disable()?
A: drm_fb_helper_emergency_disable() is a dummy if
!CONFIG_DRM_FBDEV_EMULATION, so I guess no one tried to build a failing randconfig with CONFIG_DRM_FBDEV_EMULATION=y yet.

drm_fb_helper_emergency_disable() is part of the series
https://patchwork.freedesktop.org/series/132720/
which, after discussing it on IRC with sima and Javier, is not a good solution, and is abandoned.

I think the "select DRM_KMS_HELPER" is a leftover from earlier version of drm_panic, that used the color conversion function from drm_kms_helper, but that has changed in v10 and later.

drm_panic is called from the drm_core code, so in fact it can't depends on the kms helper.

I don't see a good solution to workaround this circular dependency, maybe depends on DRM_KMS_HELPER being built-in ? (so that means drm_core will also be built-in). But that means platform that build drm core as module, won't be able to use drm_panic.

There are a few of them in the kernel tree:
rg -l CONFIG_DRM=m
arch/powerpc/configs/85xx/stx_gp3_defconfig
arch/powerpc/configs/ppc6xx_defconfig
arch/powerpc/configs/skiroot_defconfig
arch/powerpc/configs/pmac32_defconfig
arch/mips/configs/ci20_defconfig
arch/arc/configs/axs101_defconfig
arch/arc/configs/axs103_smp_defconfig
arch/riscv/configs/defconfig
arch/parisc/configs/generic-32bit_defconfig
arch/arm/configs/davinci_all_defconfig
arch/arm/configs/omap2plus_defconfig
arch/arm/configs/pxa_defconfig
arch/arm64/configs/defconfig

Quoted 5 lines

Gr{oetje,eeting}s,

Geert

Signature · 1 line
Jocelyn
#15 lore raw

PATCH v2 6/7 drm/panic: Rename logo to logo_ascii

In reply to Geert Uytterhoeven

Rename variables related to the ASCII logo, to prepare for the advent of support for graphical logos.

---
v2:
- Rebased.

---
 drivers/gpu/drm/drm_panic.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)
drivers/gpu/drm/drm_panic.c +6 -5
index 5b0acf8c86e402a8..f7e22b69bb25d3be 100644
@@ -78,7 +78,7 @@ static struct drm_panic_line panic_msg[] = {
 	PANIC_LINE("Please reboot your computer."),
 };
 
-static const struct drm_panic_line logo[] = {
+static const struct drm_panic_line logo_ascii[] = {
 	PANIC_LINE("     .--.        _"),
 	PANIC_LINE("    |o_o |      | |"),
 	PANIC_LINE("    |:_/ |      | |"),
@@ -447,7 +447,7 @@ static void draw_txt_rectangle(struct drm_scanout_buffer *sb,
 static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 {
 	size_t msg_lines = ARRAY_SIZE(panic_msg);
-	size_t logo_lines = ARRAY_SIZE(logo);
+	size_t logo_ascii_lines = ARRAY_SIZE(logo_ascii);
 	u32 fg_color = convert_from_xrgb8888(CONFIG_DRM_PANIC_FOREGROUND_COLOR, sb->format->format);
 	u32 bg_color = convert_from_xrgb8888(CONFIG_DRM_PANIC_BACKGROUND_COLOR, sb->format->format);
 	const struct font_desc *font = get_default_font(sb->width, sb->height, NULL, NULL);
@@ -459,8 +459,8 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 	r_screen = DRM_RECT_INIT(0, 0, sb->width, sb->height);
 
 	r_logo = DRM_RECT_INIT(0, 0,
-			       get_max_line_len(logo, logo_lines) * font->width,
-			       logo_lines * font->height);
+			       get_max_line_len(logo_ascii, logo_ascii_lines) * font->width,
+			       logo_ascii_lines * font->height);
 	r_msg = DRM_RECT_INIT(0, 0,
 			      min(get_max_line_len(panic_msg, msg_lines) * font->width, sb->width),
 			      min(msg_lines * font->height, sb->height));
@@ -473,7 +473,8 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 
 	if ((r_msg.x1 >= drm_rect_width(&r_logo) || r_msg.y1 >= drm_rect_height(&r_logo)) &&
 	    drm_rect_width(&r_logo) <= sb->width && drm_rect_height(&r_logo) <= sb->height) {
-		draw_txt_rectangle(sb, font, logo, logo_lines, false, &r_logo, fg_color);
+		draw_txt_rectangle(sb, font, logo_ascii, logo_ascii_lines, false, &r_logo,
+				   fg_color);
 	}
 	draw_txt_rectangle(sb, font, panic_msg, msg_lines, true, &r_msg, fg_color);
 }
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 6/7
#16 lore raw

Re: [PATCH v2 6/7] drm/panic: Rename logo to logo_ascii

Quoted 2 lines · Geert Uytterhoeven

Rename variables related to the ASCII logo, to prepare for the advent of support for graphical logos.

Thanks, that looks good to me.

Quoted 52 lines

---
v2:
- Rebased.

---
  drivers/gpu/drm/drm_panic.c | 11 ++++++-----
  1 file changed, 6 insertions(+), 5 deletions(-)
drivers/gpu/drm/drm_panic.c +6 -5
index 5b0acf8c86e402a8..f7e22b69bb25d3be 100644
@@ -78,7 +78,7 @@ static struct drm_panic_line panic_msg[] = {
  	PANIC_LINE("Please reboot your computer."),
  };
 
-static const struct drm_panic_line logo[] = {
+static const struct drm_panic_line logo_ascii[] = {
  	PANIC_LINE("     .--.        _"),
  	PANIC_LINE("    |o_o |      | |"),
  	PANIC_LINE("    |:_/ |      | |"),
@@ -447,7 +447,7 @@ static void draw_txt_rectangle(struct drm_scanout_buffer *sb,
  static void draw_panic_static_user(struct drm_scanout_buffer *sb)
  {
  	size_t msg_lines = ARRAY_SIZE(panic_msg);
-	size_t logo_lines = ARRAY_SIZE(logo);
+	size_t logo_ascii_lines = ARRAY_SIZE(logo_ascii);
  	u32 fg_color = convert_from_xrgb8888(CONFIG_DRM_PANIC_FOREGROUND_COLOR, sb->format->format);
  	u32 bg_color = convert_from_xrgb8888(CONFIG_DRM_PANIC_BACKGROUND_COLOR, sb->format->format);
  	const struct font_desc *font = get_default_font(sb->width, sb->height, NULL, NULL);
@@ -459,8 +459,8 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
  	r_screen = DRM_RECT_INIT(0, 0, sb->width, sb->height);
 
  	r_logo = DRM_RECT_INIT(0, 0,
-			       get_max_line_len(logo, logo_lines) * font->width,
-			       logo_lines * font->height);
+			       get_max_line_len(logo_ascii, logo_ascii_lines) * font->width,
+			       logo_ascii_lines * font->height);
  	r_msg = DRM_RECT_INIT(0, 0,
  			      min(get_max_line_len(panic_msg, msg_lines) * font->width, sb->width),
  			      min(msg_lines * font->height, sb->height));
@@ -473,7 +473,8 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 
  	if ((r_msg.x1 >= drm_rect_width(&r_logo) || r_msg.y1 >= drm_rect_height(&r_logo)) &&
  	    drm_rect_width(&r_logo) <= sb->width && drm_rect_height(&r_logo) <= sb->height) {
-		draw_txt_rectangle(sb, font, logo, logo_lines, false, &r_logo, fg_color);
+		draw_txt_rectangle(sb, font, logo_ascii, logo_ascii_lines, false, &r_logo,
+				   fg_color);
  	}
  	draw_txt_rectangle(sb, font, panic_msg, msg_lines, true, &r_msg, fg_color);
  }
#17 lore raw

PATCH v2 7/7 drm/panic: Add support for drawing a monochrome graphical logo

In reply to Geert Uytterhoeven

Re-use the existing support for boot-up logos to draw a monochrome graphical logo in the DRM panic handler. When no suitable graphical logo is available, the code falls back to the ASCII art penguin logo.

Note that all graphical boot-up logos are freed during late kernel initialization, hence a copy must be made for later use.

---
v2:
- Rebased,
- Inline trivial draw_logo_mono().

---
 drivers/gpu/drm/drm_panic.c | 65 +++++++++++++++++++++++++++++++++----
 drivers/video/logo/Kconfig  |  2 ++
 2 files changed, 60 insertions(+), 7 deletions(-)
drivers/gpu/drm/drm_panic.c +58 -7
index f7e22b69bb25d3be..af30f243b2802ad7 100644
@@ -7,11 +7,15 @@
  */
 
 #include <linux/font.h>
+#include <linux/init.h>
 #include <linux/iosys-map.h>
 #include <linux/kdebug.h>
 #include <linux/kmsg_dump.h>
+#include <linux/linux_logo.h>
 #include <linux/list.h>
+#include <linux/math.h>
 #include <linux/module.h>
+#include <linux/overflow.h>
 #include <linux/printk.h>
 #include <linux/types.h>
 
@@ -88,6 +92,42 @@ static const struct drm_panic_line logo_ascii[] = {
 	PANIC_LINE(" \\___)=(___/"),
 };
 
+#ifdef CONFIG_LOGO
+static const struct linux_logo *logo_mono;
+
+static int drm_panic_setup_logo(void)
+{
+	const struct linux_logo *logo = fb_find_logo(1);
+	const unsigned char *logo_data;
+	struct linux_logo *logo_dup;
+
+	if (!logo || logo->type != LINUX_LOGO_MONO)
+		return 0;
+
+	/* The logo is __init, so we must make a copy for later use */
+	logo_data = kmemdup(logo->data,
+			    size_mul(DIV_ROUND_UP(logo->width, BITS_PER_BYTE), logo->height),
+			    GFP_KERNEL);
+	if (!logo_data)
+		return -ENOMEM;
+
+	logo_dup = kmemdup(logo, sizeof(*logo), GFP_KERNEL);
+	if (!logo_dup) {
+		kfree(logo_data);
+		return -ENOMEM;
+	}
+
+	logo_dup->data = logo_data;
+	logo_mono = logo_dup;
+
+	return 0;
+}
+
+device_initcall(drm_panic_setup_logo);
+#else
+#define logo_mono	((const struct linux_logo *)NULL)
+#endif
+
 /*
  * Color conversion
  */
@@ -452,15 +492,22 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 	u32 bg_color = convert_from_xrgb8888(CONFIG_DRM_PANIC_BACKGROUND_COLOR, sb->format->format);
 	const struct font_desc *font = get_default_font(sb->width, sb->height, NULL, NULL);
 	struct drm_rect r_screen, r_logo, r_msg;
+	unsigned int logo_width, logo_height;
 
 	if (!font)
 		return;
 
 	r_screen = DRM_RECT_INIT(0, 0, sb->width, sb->height);
 
-	r_logo = DRM_RECT_INIT(0, 0,
-			       get_max_line_len(logo_ascii, logo_ascii_lines) * font->width,
-			       logo_ascii_lines * font->height);
+	if (logo_mono) {
+		logo_width = logo_mono->width;
+		logo_height = logo_mono->height;
+	} else {
+		logo_width = get_max_line_len(logo_ascii, logo_ascii_lines) * font->width;
+		logo_height = logo_ascii_lines * font->height;
+	}
+
+	r_logo = DRM_RECT_INIT(0, 0, logo_width, logo_height);
 	r_msg = DRM_RECT_INIT(0, 0,
 			      min(get_max_line_len(panic_msg, msg_lines) * font->width, sb->width),
 			      min(msg_lines * font->height, sb->height));
@@ -471,10 +518,14 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
 	/* Fill with the background color, and draw text on top */
 	drm_panic_fill(sb, &r_screen, bg_color);
 
-	if ((r_msg.x1 >= drm_rect_width(&r_logo) || r_msg.y1 >= drm_rect_height(&r_logo)) &&
-	    drm_rect_width(&r_logo) <= sb->width && drm_rect_height(&r_logo) <= sb->height) {
-		draw_txt_rectangle(sb, font, logo_ascii, logo_ascii_lines, false, &r_logo,
-				   fg_color);
+	if ((r_msg.x1 >= logo_width || r_msg.y1 >= logo_height) &&
+	    logo_width <= sb->width && logo_height <= sb->height) {
+		if (logo_mono)
+			drm_panic_blit(sb, &r_logo, logo_mono->data, DIV_ROUND_UP(logo_width, 8),
+				       fg_color);
+		else
+			draw_txt_rectangle(sb, font, logo_ascii, logo_ascii_lines, false, &r_logo,
+					   fg_color);
 	}
 	draw_txt_rectangle(sb, font, panic_msg, msg_lines, true, &r_msg, fg_color);
 }
drivers/video/logo/Kconfig +2 -0
index b7d94d1dd1585a84..ce6bb753522d215d 100644
@@ -8,6 +8,8 @@ menuconfig LOGO
 	depends on FB_CORE || SGI_NEWPORT_CONSOLE
 	help
 	  Enable and select frame buffer bootup logos.
+	  Monochrome logos will also be used by the DRM panic handler, if
+	  enabled.
 
 if LOGO
 
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 7/7
#18 lore raw

Re: [PATCH v2 7/7] drm/panic: Add support for drawing a monochrome graphical logo

Quoted 6 lines · Geert Uytterhoeven

Re-use the existing support for boot-up logos to draw a monochrome graphical logo in the DRM panic handler. When no suitable graphical logo is available, the code falls back to the ASCII art penguin logo.

Note that all graphical boot-up logos are freed during late kernel initialization, hence a copy must be made for later use.

Would it be possible to have the logo not in the __init section if DRM_PANIC is set ?

The patch looks good to me anyway.

Quoted 132 lines

---
v2:
- Rebased,
- Inline trivial draw_logo_mono().

---
  drivers/gpu/drm/drm_panic.c | 65 +++++++++++++++++++++++++++++++++----
  drivers/video/logo/Kconfig  |  2 ++
  2 files changed, 60 insertions(+), 7 deletions(-)
drivers/gpu/drm/drm_panic.c +58 -7
index f7e22b69bb25d3be..af30f243b2802ad7 100644
@@ -7,11 +7,15 @@
   */
 
  #include <linux/font.h>
+#include <linux/init.h>
  #include <linux/iosys-map.h>
  #include <linux/kdebug.h>
  #include <linux/kmsg_dump.h>
+#include <linux/linux_logo.h>
  #include <linux/list.h>
+#include <linux/math.h>
  #include <linux/module.h>
+#include <linux/overflow.h>
  #include <linux/printk.h>
  #include <linux/types.h>
 
@@ -88,6 +92,42 @@ static const struct drm_panic_line logo_ascii[] = {
  	PANIC_LINE(" \\___)=(___/"),
  };
 
+#ifdef CONFIG_LOGO
+static const struct linux_logo *logo_mono;
+
+static int drm_panic_setup_logo(void)
+{
+	const struct linux_logo *logo = fb_find_logo(1);
+	const unsigned char *logo_data;
+	struct linux_logo *logo_dup;
+
+	if (!logo || logo->type != LINUX_LOGO_MONO)
+		return 0;
+
+	/* The logo is __init, so we must make a copy for later use */
+	logo_data = kmemdup(logo->data,
+			    size_mul(DIV_ROUND_UP(logo->width, BITS_PER_BYTE), logo->height),
+			    GFP_KERNEL);
+	if (!logo_data)
+		return -ENOMEM;
+
+	logo_dup = kmemdup(logo, sizeof(*logo), GFP_KERNEL);
+	if (!logo_dup) {
+		kfree(logo_data);
+		return -ENOMEM;
+	}
+
+	logo_dup->data = logo_data;
+	logo_mono = logo_dup;
+
+	return 0;
+}
+
+device_initcall(drm_panic_setup_logo);
+#else
+#define logo_mono	((const struct linux_logo *)NULL)
+#endif
+
  /*
   * Color conversion
   */
@@ -452,15 +492,22 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
  	u32 bg_color = convert_from_xrgb8888(CONFIG_DRM_PANIC_BACKGROUND_COLOR, sb->format->format);
  	const struct font_desc *font = get_default_font(sb->width, sb->height, NULL, NULL);
  	struct drm_rect r_screen, r_logo, r_msg;
+	unsigned int logo_width, logo_height;
 
  	if (!font)
  		return;
 
  	r_screen = DRM_RECT_INIT(0, 0, sb->width, sb->height);
 
-	r_logo = DRM_RECT_INIT(0, 0,
-			       get_max_line_len(logo_ascii, logo_ascii_lines) * font->width,
-			       logo_ascii_lines * font->height);
+	if (logo_mono) {
+		logo_width = logo_mono->width;
+		logo_height = logo_mono->height;
+	} else {
+		logo_width = get_max_line_len(logo_ascii, logo_ascii_lines) * font->width;
+		logo_height = logo_ascii_lines * font->height;
+	}
+
+	r_logo = DRM_RECT_INIT(0, 0, logo_width, logo_height);
  	r_msg = DRM_RECT_INIT(0, 0,
  			      min(get_max_line_len(panic_msg, msg_lines) * font->width, sb->width),
  			      min(msg_lines * font->height, sb->height));
@@ -471,10 +518,14 @@ static void draw_panic_static_user(struct drm_scanout_buffer *sb)
  	/* Fill with the background color, and draw text on top */
  	drm_panic_fill(sb, &r_screen, bg_color);
 
-	if ((r_msg.x1 >= drm_rect_width(&r_logo) || r_msg.y1 >= drm_rect_height(&r_logo)) &&
-	    drm_rect_width(&r_logo) <= sb->width && drm_rect_height(&r_logo) <= sb->height) {
-		draw_txt_rectangle(sb, font, logo_ascii, logo_ascii_lines, false, &r_logo,
-				   fg_color);
+	if ((r_msg.x1 >= logo_width || r_msg.y1 >= logo_height) &&
+	    logo_width <= sb->width && logo_height <= sb->height) {
+		if (logo_mono)
+			drm_panic_blit(sb, &r_logo, logo_mono->data, DIV_ROUND_UP(logo_width, 8),
+				       fg_color);
+		else
+			draw_txt_rectangle(sb, font, logo_ascii, logo_ascii_lines, false, &r_logo,
+					   fg_color);
  	}
  	draw_txt_rectangle(sb, font, panic_msg, msg_lines, true, &r_msg, fg_color);
  }
drivers/video/logo/Kconfig +2 -0
index b7d94d1dd1585a84..ce6bb753522d215d 100644
@@ -8,6 +8,8 @@ menuconfig LOGO
  	depends on FB_CORE || SGI_NEWPORT_CONSOLE
  	help
  	  Enable and select frame buffer bootup logos.
+	  Monochrome logos will also be used by the DRM panic handler, if
+	  enabled.
 
  if LOGO
 
#19 lore raw

Re: [PATCH v2 7/7] drm/panic: Add support for drawing a monochrome graphical logo

Hi Jocelyn,

Quoted 10 lines · Jocelyn Falempe
Quoted 6 lines · Geert Uytterhoeven

Re-use the existing support for boot-up logos to draw a monochrome graphical logo in the DRM panic handler. When no suitable graphical logo is available, the code falls back to the ASCII art penguin logo.

Note that all graphical boot-up logos are freed during late kernel initialization, hence a copy must be made for later use.

Would it be possible to have the logo not in the __init section if DRM_PANIC is set ?

That would be rather complicated. The C source files for the logos (there can be multiple) are generated by drivers/video/logo/pnmtologo.c.

Quoted 3 lines

The patch looks good to me anyway.

Thanks!

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
maintainer
#20 lore raw

Re: [PATCH v2 7/7] drm/panic: Add support for drawing a monochrome graphical logo

In reply to Geert Uytterhoeven
Quoted 66 lines · Geert Uytterhoeven

Re-use the existing support for boot-up logos to draw a monochrome graphical logo in the DRM panic handler. When no suitable graphical logo is available, the code falls back to the ASCII art penguin logo.

Note that all graphical boot-up logos are freed during late kernel initialization, hence a copy must be made for later use.

---
v2:
- Rebased,
- Inline trivial draw_logo_mono().

---
  drivers/gpu/drm/drm_panic.c | 65 +++++++++++++++++++++++++++++++++----
  drivers/video/logo/Kconfig  |  2 ++
  2 files changed, 60 insertions(+), 7 deletions(-)
drivers/gpu/drm/drm_panic.c +29 -0
index f7e22b69bb25d3be..af30f243b2802ad7 100644
@@ -7,11 +7,15 @@
   */
 
  #include <linux/font.h>
+#include <linux/init.h>
  #include <linux/iosys-map.h>
  #include <linux/kdebug.h>
  #include <linux/kmsg_dump.h>
+#include <linux/linux_logo.h>
  #include <linux/list.h>
+#include <linux/math.h>
  #include <linux/module.h>
+#include <linux/overflow.h>
  #include <linux/printk.h>
  #include <linux/types.h>
 
@@ -88,6 +92,42 @@ static const struct drm_panic_line logo_ascii[] = {
  	PANIC_LINE(" \\___)=(___/"),
  };
 
+#ifdef CONFIG_LOGO
+static const struct linux_logo *logo_mono;
+
+static int drm_panic_setup_logo(void)
+{
+	const struct linux_logo *logo = fb_find_logo(1);
+	const unsigned char *logo_data;
+	struct linux_logo *logo_dup;
+
+	if (!logo || logo->type != LINUX_LOGO_MONO)
+		return 0;
+
+	/* The logo is __init, so we must make a copy for later use */
+	logo_data = kmemdup(logo->data,
+			    size_mul(DIV_ROUND_UP(logo->width, BITS_PER_BYTE), logo->height),
+			    GFP_KERNEL);
+	if (!logo_data)
+		return -ENOMEM;
+
+	logo_dup = kmemdup(logo, sizeof(*logo), GFP_KERNEL);
+	if (!logo_dup) {
+		kfree(logo_data);
+		return -ENOMEM;
+	}
+

drivers/gpu/drm/drm_panic.c static const struct drm_panic_line logo_ascii[] = { · line 129

Quoted 10 lines
drivers/gpu/drm/drm_panic.c +10 -0
+	logo_dup->data = logo_data;
+	logo_mono = logo_dup;
+
+	return 0;
+}
+
+device_initcall(drm_panic_setup_logo);
+#else
+#define logo_mono	((const struct linux_logo *)NULL)
+#endif

I didn't notice that the first time, but the core drm can be built as a module. That means this will leak memory each time the module is removed.

But to solve the circular dependency between drm_kms_helper and drm_panic, one solution is to depends on drm core to be built-in. In this case there won't be a leak.
So depending on how we solve the circular dependency, it can be acceptable.

Signature · 1 line
Jocelyn
#21 lore raw

Re: [PATCH v2 7/7] drm/panic: Add support for drawing a monochrome graphical logo

Hi Jocelyn,

Quoted 47 lines · Jocelyn Falempe
Quoted 46 lines · Geert Uytterhoeven

Re-use the existing support for boot-up logos to draw a monochrome graphical logo in the DRM panic handler. When no suitable graphical logo is available, the code falls back to the ASCII art penguin logo.

Note that all graphical boot-up logos are freed during late kernel initialization, hence a copy must be made for later use.

drivers/gpu/drm/drm_panic.c +31 -0
 
      PANIC_LINE(" \\___)=(___/"),
  };
 
+#ifdef CONFIG_LOGO
+static const struct linux_logo *logo_mono;
+
+static int drm_panic_setup_logo(void)
+{
+     const struct linux_logo *logo = fb_find_logo(1);
+     const unsigned char *logo_data;
+     struct linux_logo *logo_dup;
+
+     if (!logo || logo->type != LINUX_LOGO_MONO)
+             return 0;
+
+     /* The logo is __init, so we must make a copy for later use */
+     logo_data = kmemdup(logo->data,
+                         size_mul(DIV_ROUND_UP(logo->width, BITS_PER_BYTE), logo->height),
+                         GFP_KERNEL);
+     if (!logo_data)
+             return -ENOMEM;
+
+     logo_dup = kmemdup(logo, sizeof(*logo), GFP_KERNEL);
+     if (!logo_dup) {
+             kfree(logo_data);
+             return -ENOMEM;
+     }
+
+     logo_dup->data = logo_data;
+     logo_mono = logo_dup;
+
+     return 0;
+}
+
Quoted 8 lines

drivers/gpu/drm/drm_panic.c static const struct drm_panic_line logo_ascii[] = { · line 129

Quoted 4 lines
drivers/gpu/drm/drm_panic.c +4 -0
+device_initcall(drm_panic_setup_logo);
+#else
+#define logo_mono    ((const struct linux_logo *)NULL)
+#endif

I didn't notice that the first time, but the core drm can be built as a module.
That means this will leak memory each time the module is removed.

While I hadn't considered a modular DRM core, there is no memory leak:
after the logos are freed (from late_initcall_sync()), fb_find_logo() returns NULL. This does mean there won't be a graphical logo on panic, though.

Quoted 4 lines

But to solve the circular dependency between drm_kms_helper and drm_panic, one solution is to depends on drm core to be built-in. In this case there won't be a leak.
So depending on how we solve the circular dependency, it can be acceptable.

So far there is no reason to select DRM_KMS_HELPER, right?

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
maintainer
#22 lore raw

Re: [PATCH v2 7/7] drm/panic: Add support for drawing a monochrome graphical logo

Quoted 59 lines · Geert Uytterhoeven

Hi Jocelyn,

Quoted 47 lines · Jocelyn Falempe
Quoted 46 lines · Geert Uytterhoeven

Re-use the existing support for boot-up logos to draw a monochrome graphical logo in the DRM panic handler. When no suitable graphical logo is available, the code falls back to the ASCII art penguin logo.

Note that all graphical boot-up logos are freed during late kernel initialization, hence a copy must be made for later use.

drivers/gpu/drm/drm_panic.c +31 -0
 
       PANIC_LINE(" \\___)=(___/"),
   };
 
+#ifdef CONFIG_LOGO
+static const struct linux_logo *logo_mono;
+
+static int drm_panic_setup_logo(void)
+{
+     const struct linux_logo *logo = fb_find_logo(1);
+     const unsigned char *logo_data;
+     struct linux_logo *logo_dup;
+
+     if (!logo || logo->type != LINUX_LOGO_MONO)
+             return 0;
+
+     /* The logo is __init, so we must make a copy for later use */
+     logo_data = kmemdup(logo->data,
+                         size_mul(DIV_ROUND_UP(logo->width, BITS_PER_BYTE), logo->height),
+                         GFP_KERNEL);
+     if (!logo_data)
+             return -ENOMEM;
+
+     logo_dup = kmemdup(logo, sizeof(*logo), GFP_KERNEL);
+     if (!logo_dup) {
+             kfree(logo_data);
+             return -ENOMEM;
+     }
+
+     logo_dup->data = logo_data;
+     logo_mono = logo_dup;
+
+     return 0;
+}
+
Quoted 8 lines

drivers/gpu/drm/drm_panic.c static const struct drm_panic_line logo_ascii[] = { · line 129

Quoted 4 lines
drivers/gpu/drm/drm_panic.c +4 -0
+device_initcall(drm_panic_setup_logo);
+#else
+#define logo_mono    ((const struct linux_logo *)NULL)
+#endif

I didn't notice that the first time, but the core drm can be built as a module.
That means this will leak memory each time the module is removed.

Quoted 4 lines

While I hadn't considered a modular DRM core, there is no memory leak:
after the logos are freed (from late_initcall_sync()), fb_find_logo() returns NULL. This does mean there won't be a graphical logo on panic, though.

Yes, you're right, thanks for the precision.

Quoted 7 lines
Quoted 5 lines

But to solve the circular dependency between drm_kms_helper and drm_panic, one solution is to depends on drm core to be built-in. In this case there won't be a leak.
So depending on how we solve the circular dependency, it can be acceptable.

So far there is no reason to select DRM_KMS_HELPER, right?

The current version doesn't need DRM_KMS_HELPER.
But for example, it uses struct drm_rect, and a few functions (drm_rect_width()) that are defined in .h, but other drm_rect_* functions are defined in DRM_KMS_HELPER. And as you pointed out in your patch, it duplicates the drm_fb_clip_offset(). So I'm not sure if it can avoid the dependency on DRM_KMS_HELPER in the long term.

Quoted 5 lines

Gr{oetje,eeting}s,

Geert

maintainer
#23 lore raw
In reply to Geert Uytterhoeven

Hi,

I want to push at least the first patch that is an important fix.
But if there are no objections, I can push the whole series except patch 5 "drm/panic: Convert to drm_fb_clip_offset()" which causes some build issue.

Best regards,

Signature · 1 line
Jocelyn
Quoted 35 lines · Geert Uytterhoeven

Hi all,

If drm/panic is enabled, a user-friendly message is shown on screen when a kernel panic occurs, together with an ASCII art penguin logo. Of course we can do better ;-)
Hence this patch series extends drm/panic to draw the monochrome graphical boot logo, when available, preceded by the customary fixes.

Changes compared to v1:
- Rebase against today's drm-misc-next, where drm_panic is broken on all current drivers due to an uninitialized pointer dereference. Presumably this was only tested with an out-of-tree driver change?
- New fixes [1/7], [3/7], and [4/7],
- New cleanup [5/7],
- Inline trivial draw_logo_mono().

This has been tested with rcar-du.

Thanks for your comments!

Geert Uytterhoeven (7):
drm/panic: Fix uninitialized drm_scanout_buffer.set_pixel() crash drm/panic: Fix off-by-one logo size checks
lib/fonts: Fix visiblity of SUN12x22 and TER16x32 if DRM_PANIC drm/panic: Spelling s/formater/formatter/
drm/panic: Convert to drm_fb_clip_offset()
drm/panic: Rename logo to logo_ascii
drm/panic: Add support for drawing a monochrome graphical logo

  drivers/gpu/drm/Kconfig     |  2 +-
  drivers/gpu/drm/drm_panic.c | 74 +++++++++++++++++++++++++++++++------
  drivers/video/logo/Kconfig  |  2 +
  lib/fonts/Kconfig           |  6 ++-
  4 files changed, 70 insertions(+), 14 deletions(-)
maintainer
#24 lore raw
Quoted 4 lines · Jocelyn Falempe

Hi,

I want to push at least the first patch that is an important fix.
But if there are no objections, I can push the whole series except patch 5 "drm/panic: Convert to drm_fb_clip_offset()" which causes some build issue.

I just pushed them to drm-misc-next.
Thanks all.

Quoted 3 lines

Best regards,