Skip to contentSkip to conversation

drm/fbdev-dma: Only set smem_start is enable per module option

PATCH In review Thomas Zimmermann 3 participants 5 messages Jun 17, 2024 – Jun 18, 2024

Subsystems (2)
lore
Share & tools
lore

Thomas Zimmermann

author maintainer
#0 lore raw

PATCH 1/1 drm/fbdev-dma: Only set smem_start is enable per module option

Only export struct fb_info.fix.smem_start if that is required by the user and the memory does not come from vmalloc().

Setting struct fb_info.fix.smem_start breaks systems where DMA memory is backed by vmalloc address space. An example error is shown below.

[    3.536043] ------------[ cut here ]------------
[    3.540716] virt_to_phys used for non-linear address: 000000007fc4f540 (0xffff800086001000)
[    3.552628] WARNING: CPU: 4 PID: 61 at arch/arm64/mm/physaddr.c:12 __virt_to_phys+0x68/0x98
[    3.565455] Modules linked in:
[    3.568525] CPU: 4 PID: 61 Comm: kworker/u12:5 Not tainted 6.6.23-06226-g4986cc3e1b75-dirty #250
[    3.577310] Hardware name: NXP i.MX95 19X19 board (DT)
[    3.582452] Workqueue: events_unbound deferred_probe_work_func
[    3.588291] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[    3.595233] pc : __virt_to_phys+0x68/0x98
[    3.599246] lr : __virt_to_phys+0x68/0x98
[    3.603276] sp : ffff800083603990
[    3.677939] Call trace:
[    3.680393]  __virt_to_phys+0x68/0x98
[    3.684067]  drm_fbdev_dma_helper_fb_probe+0x138/0x238
[    3.689214]  __drm_fb_helper_initial_config_and_unlock+0x2b0/0x4c0
[    3.695385]  drm_fb_helper_initial_config+0x4c/0x68
[    3.700264]  drm_fbdev_dma_client_hotplug+0x8c/0xe0
[    3.705161]  drm_client_register+0x60/0xb0
[    3.709269]  drm_fbdev_dma_setup+0x94/0x148

Additionally, DMA memory is assumed to by contiguous in physical address space, which is not guaranteed by vmalloc().

Resolve this by checking the module flag drm_leak_fbdev_smem when DRM allocated the instance of struct fb_info. Fbdev-dma then only sets smem_start only if required (via FBINFO_HIDE_SMEM_START). Also guarantee that the framebuffer is not located in vmalloc address space.

Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
---
 drivers/gpu/drm/drm_fb_helper.c | 6 +++---
 drivers/gpu/drm/drm_fbdev_dma.c | 5 ++++-
 2 files changed, 7 insertions(+), 4 deletions(-)
drivers/gpu/drm/drm_fb_helper.c +3 -3
index e2e19f49342e1..18565ec684517 100644
@@ -517,6 +517,9 @@ struct fb_info *drm_fb_helper_alloc_info(struct drm_fb_helper *fb_helper)
 	if (!info)
 		return ERR_PTR(-ENOMEM);
 
+	if (!drm_leak_fbdev_smem)
+		info->flags |= FBINFO_HIDE_SMEM_START;
+
 	ret = fb_alloc_cmap(&info->cmap, 256, 0);
 	if (ret)
 		goto err_release;
@@ -1853,9 +1856,6 @@ __drm_fb_helper_initial_config_and_unlock(struct drm_fb_helper *fb_helper)
 	info = fb_helper->info;
 	info->var.pixclock = 0;
 
-	if (!drm_leak_fbdev_smem)
-		info->flags |= FBINFO_HIDE_SMEM_START;
-
 	/* Need to drop locks to avoid recursive deadlock in
 	 * register_framebuffer. This is ok because the only thing left to do is
 	 * register the fbdev emulation instance in kernel_fb_helper_list. */
drivers/gpu/drm/drm_fbdev_dma.c +4 -1
index 97ef6300d47e0..4b0527ddeb5a7 100644
@@ -137,7 +137,10 @@ static int drm_fbdev_dma_helper_fb_probe(struct drm_fb_helper *fb_helper,
 		info->flags |= FBINFO_READS_FAST; /* signal caching */
 	info->screen_size = sizes->surface_height * fb->pitches[0];
 	info->screen_buffer = map.vaddr;
-	info->fix.smem_start = page_to_phys(virt_to_page(info->screen_buffer));
+	if (!(info->flags & FBINFO_HIDE_SMEM_START)) {
+		if (!drm_WARN_ON(dev, is_vmalloc_addr(info->screen_buffer)))
+			info->fix.smem_start = page_to_phys(virt_to_page(info->screen_buffer));
+	}
 	info->fix.smem_len = info->screen_size;
 
 	/* deferred I/O */
Signature · 1 line
2.45.2
Tested-by on v1 1/1
#1 lore raw

Hi Thomas,

Quoted 36 lines · Thomas Zimmermann

Only export struct fb_info.fix.smem_start if that is required by the user and the memory does not come from vmalloc().

Setting struct fb_info.fix.smem_start breaks systems where DMA memory is backed by vmalloc address space. An example error is shown below.

[    3.536043] ------------[ cut here ]------------
[    3.540716] virt_to_phys used for non-linear address: 000000007fc4f540 (0xffff800086001000)
[    3.552628] WARNING: CPU: 4 PID: 61 at arch/arm64/mm/physaddr.c:12 __virt_to_phys+0x68/0x98
[    3.565455] Modules linked in:
[    3.568525] CPU: 4 PID: 61 Comm: kworker/u12:5 Not tainted 6.6.23-06226-g4986cc3e1b75-dirty #250
[    3.577310] Hardware name: NXP i.MX95 19X19 board (DT)
[    3.582452] Workqueue: events_unbound deferred_probe_work_func
[    3.588291] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[    3.595233] pc : __virt_to_phys+0x68/0x98
[    3.599246] lr : __virt_to_phys+0x68/0x98
[    3.603276] sp : ffff800083603990
[    3.677939] Call trace:
[    3.680393]  __virt_to_phys+0x68/0x98
[    3.684067]  drm_fbdev_dma_helper_fb_probe+0x138/0x238
[    3.689214]  __drm_fb_helper_initial_config_and_unlock+0x2b0/0x4c0
[    3.695385]  drm_fb_helper_initial_config+0x4c/0x68
[    3.700264]  drm_fbdev_dma_client_hotplug+0x8c/0xe0
[    3.705161]  drm_client_register+0x60/0xb0
[    3.709269]  drm_fbdev_dma_setup+0x94/0x148

Additionally, DMA memory is assumed to by contiguous in physical address space, which is not guaranteed by vmalloc().

Resolve this by checking the module flag drm_leak_fbdev_smem when DRM allocated the instance of struct fb_info. Fbdev-dma then only sets smem_start only if required (via FBINFO_HIDE_SMEM_START). Also guarantee that the framebuffer is not located in vmalloc address space.

Quoted 4 lines
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")

Thanks, this fixes the issue I was seeing on R-Car Gen3/Gen4 with rcar-du.

No regressions on R-Car Gen2 (rcar-du) and R-Mobile A1 (shmobile) which didn't shown the warning in the first place.

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

author maintainer
#2 lore raw

Hi

Am 17.06.24 um 19:42 schrieb Geert Uytterhoeven:

Quoted 44 lines

    Hi Thomas,

Quoted 36 lines · Thomas Zimmermann

Only export struct fb_info.fix.smem_start if that is required by the user and the memory does not come from vmalloc().

Setting struct fb_info.fix.smem_start breaks systems where DMA memory is backed by vmalloc address space. An example error is shown below.

[    3.536043] ------------[ cut here ]------------
[    3.540716] virt_to_phys used for non-linear address: 000000007fc4f540 (0xffff800086001000) [    3.552628] WARNING: CPU: 4 PID: 61 at arch/arm64/mm/physaddr.c:12 __virt_to_phys+0x68/0x98 [    3.565455] Modules linked in:
[    3.568525] CPU: 4 PID: 61 Comm: kworker/u12:5 Not tainted 6.6.23-06226-g4986cc3e1b75-dirty #250 [    3.577310] Hardware name: NXP i.MX95 19X19 board (DT)
[    3.582452] Workqueue: events_unbound deferred_probe_work_func
[    3.588291] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[    3.595233] pc : __virt_to_phys+0x68/0x98
[    3.599246] lr : __virt_to_phys+0x68/0x98
[    3.603276] sp : ffff800083603990
[    3.677939] Call trace:
[    3.680393]  __virt_to_phys+0x68/0x98
[    3.684067]  drm_fbdev_dma_helper_fb_probe+0x138/0x238
[    3.689214] __drm_fb_helper_initial_config_and_unlock+0x2b0/0x4c0
[    3.695385]  drm_fb_helper_initial_config+0x4c/0x68
[    3.700264]  drm_fbdev_dma_client_hotplug+0x8c/0xe0
[    3.705161]  drm_client_register+0x60/0xb0
[    3.709269]  drm_fbdev_dma_setup+0x94/0x148

Additionally, DMA memory is assumed to by contiguous in physical address space, which is not guaranteed by vmalloc().

Resolve this by checking the module flag drm_leak_fbdev_smem when DRM allocated the instance of struct fb_info. Fbdev-dma then only sets smem_start only if required (via FBINFO_HIDE_SMEM_START). Also guarantee that the framebuffer is not located in vmalloc address space.

Quoted 4 lines
Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
Quoted 4 lines

Thanks, this fixes the issue I was seeing on R-Car Gen3/Gen4 with rcar-du.

No regressions on R-Car Gen2 (rcar-du) and R-Mobile A1 (shmobile) which didn't shown the warning in the first place.

Right, your bug report overlapped with the other one. I'll add your report to the patch tags.

Quoted 2 lines

Thanks a lot.

Best regards
Thomas

Quoted 11 lines

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
Signature · 7 lines
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)
Reviewed-by on v1 1/1
#3 lore raw
In reply to Thomas Zimmermann
Quoted 74 lines · Thomas Zimmermann

Only export struct fb_info.fix.smem_start if that is required by the user and the memory does not come from vmalloc().

Setting struct fb_info.fix.smem_start breaks systems where DMA memory is backed by vmalloc address space. An example error is shown below.

[    3.536043] ------------[ cut here ]------------
[    3.540716] virt_to_phys used for non-linear address: 000000007fc4f540 (0xffff800086001000)
[    3.552628] WARNING: CPU: 4 PID: 61 at arch/arm64/mm/physaddr.c:12 __virt_to_phys+0x68/0x98
[    3.565455] Modules linked in:
[    3.568525] CPU: 4 PID: 61 Comm: kworker/u12:5 Not tainted 6.6.23-06226-g4986cc3e1b75-dirty #250
[    3.577310] Hardware name: NXP i.MX95 19X19 board (DT)
[    3.582452] Workqueue: events_unbound deferred_probe_work_func
[    3.588291] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[    3.595233] pc : __virt_to_phys+0x68/0x98
[    3.599246] lr : __virt_to_phys+0x68/0x98
[    3.603276] sp : ffff800083603990
[    3.677939] Call trace:
[    3.680393]  __virt_to_phys+0x68/0x98
[    3.684067]  drm_fbdev_dma_helper_fb_probe+0x138/0x238
[    3.689214]  __drm_fb_helper_initial_config_and_unlock+0x2b0/0x4c0
[    3.695385]  drm_fb_helper_initial_config+0x4c/0x68
[    3.700264]  drm_fbdev_dma_client_hotplug+0x8c/0xe0
[    3.705161]  drm_client_register+0x60/0xb0
[    3.709269]  drm_fbdev_dma_setup+0x94/0x148

Additionally, DMA memory is assumed to by contiguous in physical address space, which is not guaranteed by vmalloc().

Resolve this by checking the module flag drm_leak_fbdev_smem when DRM allocated the instance of struct fb_info. Fbdev-dma then only sets smem_start only if required (via FBINFO_HIDE_SMEM_START). Also guarantee that the framebuffer is not located in vmalloc address space.

Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
---
 drivers/gpu/drm/drm_fb_helper.c | 6 +++---
 drivers/gpu/drm/drm_fbdev_dma.c | 5 ++++-
 2 files changed, 7 insertions(+), 4 deletions(-)
drivers/gpu/drm/drm_fb_helper.c +3 -3
index e2e19f49342e1..18565ec684517 100644
@@ -517,6 +517,9 @@ struct fb_info *drm_fb_helper_alloc_info(struct drm_fb_helper *fb_helper)
 	if (!info)
 		return ERR_PTR(-ENOMEM);
 
+	if (!drm_leak_fbdev_smem)
+		info->flags |= FBINFO_HIDE_SMEM_START;
+
 	ret = fb_alloc_cmap(&info->cmap, 256, 0);
 	if (ret)
 		goto err_release;
@@ -1853,9 +1856,6 @@ __drm_fb_helper_initial_config_and_unlock(struct drm_fb_helper *fb_helper)
 	info = fb_helper->info;
 	info->var.pixclock = 0;
 
-	if (!drm_leak_fbdev_smem)
-		info->flags |= FBINFO_HIDE_SMEM_START;
-
 	/* Need to drop locks to avoid recursive deadlock in
 	 * register_framebuffer. This is ok because the only thing left to do is
 	 * register the fbdev emulation instance in kernel_fb_helper_list. */
drivers/gpu/drm/drm_fbdev_dma.c +0 -0
index 97ef6300d47e0..4b0527ddeb5a7 100644

drivers/gpu/drm/drm_fbdev_dma.c static int drm_fbdev_dma_helper_fb_probe(struct drm_fb_helper *fb_helper, · line 140

Quoted 5 lines
drivers/gpu/drm/drm_fbdev_dma.c +1 -1
@@ -137,7 +137,10 @@ static int drm_fbdev_dma_helper_fb_probe(struct drm_fb_helper *fb_helper,
 		info->flags |= FBINFO_READS_FAST; /* signal caching */
 	info->screen_size = sizes->surface_height * fb->pitches[0];
 	info->screen_buffer = map.vaddr;
-	info->fix.smem_start = page_to_phys(virt_to_page(info->screen_buffer));
+	if (!(info->flags & FBINFO_HIDE_SMEM_START)) {

I guess since the WARN is conditional on the explicit opt-in with drm_leak_fbdev_smem it should be fine.

Quoted 9 lines
drivers/gpu/drm/drm_fbdev_dma.c +3 -0
+		if (!drm_WARN_ON(dev, is_vmalloc_addr(info->screen_buffer)))
+			info->fix.smem_start = page_to_phys(virt_to_page(info->screen_buffer));
+	}
 	info->fix.smem_len = info->screen_size;
 
 	/* deferred I/O */
Signature · 1 line
2.45.2
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
author maintainer
#4 lore raw

Hi

Am 18.06.24 um 11:35 schrieb Daniel Vetter:

Quoted 74 lines · Thomas Zimmermann
Quoted 73 lines

Only export struct fb_info.fix.smem_start if that is required by the user and the memory does not come from vmalloc().

Setting struct fb_info.fix.smem_start breaks systems where DMA memory is backed by vmalloc address space. An example error is shown below.

[    3.536043] ------------[ cut here ]------------
[    3.540716] virt_to_phys used for non-linear address: 000000007fc4f540 (0xffff800086001000)
[    3.552628] WARNING: CPU: 4 PID: 61 at arch/arm64/mm/physaddr.c:12 __virt_to_phys+0x68/0x98
[    3.565455] Modules linked in:
[    3.568525] CPU: 4 PID: 61 Comm: kworker/u12:5 Not tainted 6.6.23-06226-g4986cc3e1b75-dirty #250
[    3.577310] Hardware name: NXP i.MX95 19X19 board (DT)
[    3.582452] Workqueue: events_unbound deferred_probe_work_func
[    3.588291] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[    3.595233] pc : __virt_to_phys+0x68/0x98
[    3.599246] lr : __virt_to_phys+0x68/0x98
[    3.603276] sp : ffff800083603990
[    3.677939] Call trace:
[    3.680393]  __virt_to_phys+0x68/0x98
[    3.684067]  drm_fbdev_dma_helper_fb_probe+0x138/0x238
[    3.689214]  __drm_fb_helper_initial_config_and_unlock+0x2b0/0x4c0
[    3.695385]  drm_fb_helper_initial_config+0x4c/0x68
[    3.700264]  drm_fbdev_dma_client_hotplug+0x8c/0xe0
[    3.705161]  drm_client_register+0x60/0xb0
[    3.709269]  drm_fbdev_dma_setup+0x94/0x148

Additionally, DMA memory is assumed to by contiguous in physical address space, which is not guaranteed by vmalloc().

Resolve this by checking the module flag drm_leak_fbdev_smem when DRM allocated the instance of struct fb_info. Fbdev-dma then only sets smem_start only if required (via FBINFO_HIDE_SMEM_START). Also guarantee that the framebuffer is not located in vmalloc address space.

Fixes b79fe9abd58b ("drm/fbdev-dma: Implement fbdev emulation for GEM DMA helpers")
---
  drivers/gpu/drm/drm_fb_helper.c | 6 +++---
  drivers/gpu/drm/drm_fbdev_dma.c | 5 ++++-
  2 files changed, 7 insertions(+), 4 deletions(-)
drivers/gpu/drm/drm_fb_helper.c +3 -3
index e2e19f49342e1..18565ec684517 100644
@@ -517,6 +517,9 @@ struct fb_info *drm_fb_helper_alloc_info(struct drm_fb_helper *fb_helper)
  	if (!info)
  		return ERR_PTR(-ENOMEM);
 
+	if (!drm_leak_fbdev_smem)
+		info->flags |= FBINFO_HIDE_SMEM_START;
+
  	ret = fb_alloc_cmap(&info->cmap, 256, 0);
  	if (ret)
  		goto err_release;
@@ -1853,9 +1856,6 @@ __drm_fb_helper_initial_config_and_unlock(struct drm_fb_helper *fb_helper)
  	info = fb_helper->info;
  	info->var.pixclock = 0;
 
-	if (!drm_leak_fbdev_smem)
-		info->flags |= FBINFO_HIDE_SMEM_START;
-
  	/* Need to drop locks to avoid recursive deadlock in
  	 * register_framebuffer. This is ok because the only thing left to do is
  	 * register the fbdev emulation instance in kernel_fb_helper_list. */
drivers/gpu/drm/drm_fbdev_dma.c +0 -0
index 97ef6300d47e0..4b0527ddeb5a7 100644
Quoted 8 lines

drivers/gpu/drm/drm_fbdev_dma.c static int drm_fbdev_dma_helper_fb_probe(struct drm_fb_helper *fb_helper, · line 140

Quoted 6 lines
drivers/gpu/drm/drm_fbdev_dma.c +1 -1
@@ -137,7 +137,10 @@ static int drm_fbdev_dma_helper_fb_probe(struct drm_fb_helper *fb_helper,
  		info->flags |= FBINFO_READS_FAST; /* signal caching */
  	info->screen_size = sizes->surface_height * fb->pitches[0];
  	info->screen_buffer = map.vaddr;
-	info->fix.smem_start = page_to_phys(virt_to_page(info->screen_buffer));
+	if (!(info->flags & FBINFO_HIDE_SMEM_START)) {

I guess since the WARN is conditional on the explicit opt-in with drm_leak_fbdev_smem it should be fine.

Yes, that's the intention. If someone enables this flag, it warns if that's not supported. By default, smem_start is clear.

Quoted 2 lines

Thanks

Best regards
Thomas

Quoted 10 lines
Quoted 10 lines
drivers/gpu/drm/drm_fbdev_dma.c +3 -0
+		if (!drm_WARN_ON(dev, is_vmalloc_addr(info->screen_buffer)))
+			info->fix.smem_start = page_to_phys(virt_to_page(info->screen_buffer));
+	}
  	info->fix.smem_len = info->screen_size;
 
  	/* deferred I/O */
Signature · 1 line
2.45.2
Signature · 7 lines
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)