| Message ID | 20190124165831.16427-5-daniel.vetter@ffwll.ch (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | some cleanups, mostly around fbdev emulation | expand |
Den 24.01.2019 17.58, skrev Daniel Vetter: > The fbdev split between fix and var information is kinda > pointless for drm drivers since everything is fixed: The fbdev > emulation doesn't support changing modes at all. > > Create a new simplified helper and use it in the generic fbdev > helper code. Follow-up patches will beef it up more and roll > it out to all drivers. > > Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> > --- > drivers/gpu/drm/drm_fb_helper.c | 27 +++++++++++++++++++++++++-- > include/drm/drm_fb_helper.h | 1 + > 2 files changed, 26 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c > index 5eaccd202f4f..34c4ed378796 100644 > --- a/drivers/gpu/drm/drm_fb_helper.c > +++ b/drivers/gpu/drm/drm_fb_helper.c > @@ -2105,6 +2105,30 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe > } > EXPORT_SYMBOL(drm_fb_helper_fill_var); > > +/** > + * drm_fb_helper_fill_info - initializes fbdev information > + * @info: fbdev instance to set up > + * @fb_helper: fb helper instance to use as template > + * > + * > + * Sets up the variable and fixed fbdev metainformation from the given fb helper > + * instance and the drm framebuffer allocated in &drm_fb_helper.fb. > + * > + * Drivers should call this (or their equivalent setup code) from their > + * &drm_fb_helper_funcs.fb_probe callback after having allocated the fbdev > + * backing storage framebuffer. > + */ > +void drm_fb_helper_fill_info(struct fb_info *info, No need to pass in fb_info it's available as fb_helper->fbdev. Set by drm_fb_helper_alloc_fbi(). > + struct drm_fb_helper *fb_helper) > +{ > + struct drm_framebuffer *fb = fb_helper->fb; > + > + drm_fb_helper_fill_fix(info, fb->pitches[0], fb->format->depth); > + drm_fb_helper_fill_var(info, fb_helper, fb->width, fb->height); AFAIU fb->width/height can be different from sizes->fb_width/fb_height when there's double/triple buffering. I belive you need to use the sizes values here. Noralf. > + > +} > +EXPORT_SYMBOL(drm_fb_helper_fill_info); > + > static int drm_fb_helper_probe_connector_modes(struct drm_fb_helper *fb_helper, > uint32_t maxX, > uint32_t maxY) > @@ -3163,8 +3187,7 @@ int drm_fb_helper_generic_probe(struct drm_fb_helper *fb_helper, > #endif > strcpy(fbi->fix.id, "DRM emulated"); > > - drm_fb_helper_fill_fix(fbi, fb->pitches[0], fb->format->depth); > - drm_fb_helper_fill_var(fbi, fb_helper, sizes->fb_width, sizes->fb_height); > + drm_fb_helper_fill_info(fbi, fb_helper); > > if (fb->funcs->dirty) { > struct fb_ops *fbops; > diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h > index bb9acea61369..e8d92724f472 100644 > --- a/include/drm/drm_fb_helper.h > +++ b/include/drm/drm_fb_helper.h > @@ -292,6 +292,7 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe > uint32_t fb_width, uint32_t fb_height); > void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch, > uint32_t depth); > +void drm_fb_helper_fill_info(struct fb_info *info, struct drm_fb_helper *fb_helper); > > void drm_fb_helper_unlink_fbi(struct drm_fb_helper *fb_helper); > >
On Thu, Jan 24, 2019 at 06:40:52PM +0100, Noralf Trønnes wrote: > > > Den 24.01.2019 17.58, skrev Daniel Vetter: > > The fbdev split between fix and var information is kinda > > pointless for drm drivers since everything is fixed: The fbdev > > emulation doesn't support changing modes at all. > > > > Create a new simplified helper and use it in the generic fbdev > > helper code. Follow-up patches will beef it up more and roll > > it out to all drivers. > > > > Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> > > --- > > drivers/gpu/drm/drm_fb_helper.c | 27 +++++++++++++++++++++++++-- > > include/drm/drm_fb_helper.h | 1 + > > 2 files changed, 26 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c > > index 5eaccd202f4f..34c4ed378796 100644 > > --- a/drivers/gpu/drm/drm_fb_helper.c > > +++ b/drivers/gpu/drm/drm_fb_helper.c > > @@ -2105,6 +2105,30 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe > > } > > EXPORT_SYMBOL(drm_fb_helper_fill_var); > > > > +/** > > + * drm_fb_helper_fill_info - initializes fbdev information > > + * @info: fbdev instance to set up > > + * @fb_helper: fb helper instance to use as template > > + * > > + * > > + * Sets up the variable and fixed fbdev metainformation from the given fb helper > > + * instance and the drm framebuffer allocated in &drm_fb_helper.fb. > > + * > > + * Drivers should call this (or their equivalent setup code) from their > > + * &drm_fb_helper_funcs.fb_probe callback after having allocated the fbdev > > + * backing storage framebuffer. > > + */ > > +void drm_fb_helper_fill_info(struct fb_info *info, > > No need to pass in fb_info it's available as fb_helper->fbdev. Set by > drm_fb_helper_alloc_fbi(). Uh, I get to edit all the patches. Still, thanks for pointing out. I also realized that I've forgotten to add the dummy functiont for !CONFIG_FB, will fix that too. > > + struct drm_fb_helper *fb_helper) > > +{ > > + struct drm_framebuffer *fb = fb_helper->fb; > > + > > + drm_fb_helper_fill_fix(info, fb->pitches[0], fb->format->depth); > > + drm_fb_helper_fill_var(info, fb_helper, fb->width, fb->height); > > AFAIU fb->width/height can be different from sizes->fb_width/fb_height > when there's double/triple buffering. I belive you need to use the sizes > values here. Yup they're bigger than the actual screen size, but so is sizes: For overallocting we adjust sizes before we call ->fb_probe, and the ->fb_probe callbacks in turn then call this helper. So I think it should be all equivalent. Agreed or do I miss something? -Daniel > > Noralf. > > > + > > +} > > +EXPORT_SYMBOL(drm_fb_helper_fill_info); > > + > > static int drm_fb_helper_probe_connector_modes(struct drm_fb_helper *fb_helper, > > uint32_t maxX, > > uint32_t maxY) > > @@ -3163,8 +3187,7 @@ int drm_fb_helper_generic_probe(struct drm_fb_helper *fb_helper, > > #endif > > strcpy(fbi->fix.id, "DRM emulated"); > > > > - drm_fb_helper_fill_fix(fbi, fb->pitches[0], fb->format->depth); > > - drm_fb_helper_fill_var(fbi, fb_helper, sizes->fb_width, sizes->fb_height); > > + drm_fb_helper_fill_info(fbi, fb_helper); > > > > if (fb->funcs->dirty) { > > struct fb_ops *fbops; > > diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h > > index bb9acea61369..e8d92724f472 100644 > > --- a/include/drm/drm_fb_helper.h > > +++ b/include/drm/drm_fb_helper.h > > @@ -292,6 +292,7 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe > > uint32_t fb_width, uint32_t fb_height); > > void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch, > > uint32_t depth); > > +void drm_fb_helper_fill_info(struct fb_info *info, struct drm_fb_helper *fb_helper); > > > > void drm_fb_helper_unlink_fbi(struct drm_fb_helper *fb_helper); > > > >
Den 25.01.2019 09.48, skrev Daniel Vetter: > On Thu, Jan 24, 2019 at 06:40:52PM +0100, Noralf Trønnes wrote: >> >> >> Den 24.01.2019 17.58, skrev Daniel Vetter: >>> The fbdev split between fix and var information is kinda >>> pointless for drm drivers since everything is fixed: The fbdev >>> emulation doesn't support changing modes at all. >>> >>> Create a new simplified helper and use it in the generic fbdev >>> helper code. Follow-up patches will beef it up more and roll >>> it out to all drivers. >>> >>> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> >>> --- >>> drivers/gpu/drm/drm_fb_helper.c | 27 +++++++++++++++++++++++++-- >>> include/drm/drm_fb_helper.h | 1 + >>> 2 files changed, 26 insertions(+), 2 deletions(-) >>> >>> diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c >>> index 5eaccd202f4f..34c4ed378796 100644 >>> --- a/drivers/gpu/drm/drm_fb_helper.c >>> +++ b/drivers/gpu/drm/drm_fb_helper.c >>> @@ -2105,6 +2105,30 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe >>> } >>> EXPORT_SYMBOL(drm_fb_helper_fill_var); >>> >>> +/** >>> + * drm_fb_helper_fill_info - initializes fbdev information >>> + * @info: fbdev instance to set up >>> + * @fb_helper: fb helper instance to use as template >>> + * >>> + * >>> + * Sets up the variable and fixed fbdev metainformation from the given fb helper >>> + * instance and the drm framebuffer allocated in &drm_fb_helper.fb. >>> + * >>> + * Drivers should call this (or their equivalent setup code) from their >>> + * &drm_fb_helper_funcs.fb_probe callback after having allocated the fbdev >>> + * backing storage framebuffer. >>> + */ >>> +void drm_fb_helper_fill_info(struct fb_info *info, >> >> No need to pass in fb_info it's available as fb_helper->fbdev. Set by >> drm_fb_helper_alloc_fbi(). > > Uh, I get to edit all the patches. Still, thanks for pointing out. > > I also realized that I've forgotten to add the dummy functiont for > !CONFIG_FB, will fix that too. > >>> + struct drm_fb_helper *fb_helper) >>> +{ >>> + struct drm_framebuffer *fb = fb_helper->fb; >>> + >>> + drm_fb_helper_fill_fix(info, fb->pitches[0], fb->format->depth); >>> + drm_fb_helper_fill_var(info, fb_helper, fb->width, fb->height); >> >> AFAIU fb->width/height can be different from sizes->fb_width/fb_height >> when there's double/triple buffering. I belive you need to use the sizes >> values here. > > Yup they're bigger than the actual screen size, but so is sizes: For > overallocting we adjust sizes before we call ->fb_probe, and the > ->fb_probe callbacks in turn then call this helper. So I think it should > be all equivalent. Agreed or do I miss something? I did a test setting drm_fbdev_overalloc = 200: [drm:drm_setup_crtcs] desired mode 320x240 set on crtc 33 (0,0) [drm:drm_fb_helper_generic_probe] surface width(320), height(480) and bpp(16) [drm:drm_fb_helper_generic_probe] fb_width(320), fb_height(240) sizes->surface_height != sizes->fb_height sizes->surface_height is used to create the framebuffer -> fb->height. My driver doesn't allow me to create such a big framebuffer so I didn't see the actual fb->height value come out the other end, but I still believe you need to use sizes->fb_height. Noralf. > -Daniel > >> >> Noralf. >> >>> + >>> +} >>> +EXPORT_SYMBOL(drm_fb_helper_fill_info); >>> + >>> static int drm_fb_helper_probe_connector_modes(struct drm_fb_helper *fb_helper, >>> uint32_t maxX, >>> uint32_t maxY) >>> @@ -3163,8 +3187,7 @@ int drm_fb_helper_generic_probe(struct drm_fb_helper *fb_helper, >>> #endif >>> strcpy(fbi->fix.id, "DRM emulated"); >>> >>> - drm_fb_helper_fill_fix(fbi, fb->pitches[0], fb->format->depth); >>> - drm_fb_helper_fill_var(fbi, fb_helper, sizes->fb_width, sizes->fb_height); >>> + drm_fb_helper_fill_info(fbi, fb_helper); >>> >>> if (fb->funcs->dirty) { >>> struct fb_ops *fbops; >>> diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h >>> index bb9acea61369..e8d92724f472 100644 >>> --- a/include/drm/drm_fb_helper.h >>> +++ b/include/drm/drm_fb_helper.h >>> @@ -292,6 +292,7 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe >>> uint32_t fb_width, uint32_t fb_height); >>> void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch, >>> uint32_t depth); >>> +void drm_fb_helper_fill_info(struct fb_info *info, struct drm_fb_helper *fb_helper); >>> >>> void drm_fb_helper_unlink_fbi(struct drm_fb_helper *fb_helper); >>> >>> >
On Fri, Jan 25, 2019 at 10:46 AM Noralf Trønnes <noralf@tronnes.org> wrote: > > > > Den 25.01.2019 09.48, skrev Daniel Vetter: > > On Thu, Jan 24, 2019 at 06:40:52PM +0100, Noralf Trønnes wrote: > >> > >> > >> Den 24.01.2019 17.58, skrev Daniel Vetter: > >>> The fbdev split between fix and var information is kinda > >>> pointless for drm drivers since everything is fixed: The fbdev > >>> emulation doesn't support changing modes at all. > >>> > >>> Create a new simplified helper and use it in the generic fbdev > >>> helper code. Follow-up patches will beef it up more and roll > >>> it out to all drivers. > >>> > >>> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> > >>> --- > >>> drivers/gpu/drm/drm_fb_helper.c | 27 +++++++++++++++++++++++++-- > >>> include/drm/drm_fb_helper.h | 1 + > >>> 2 files changed, 26 insertions(+), 2 deletions(-) > >>> > >>> diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c > >>> index 5eaccd202f4f..34c4ed378796 100644 > >>> --- a/drivers/gpu/drm/drm_fb_helper.c > >>> +++ b/drivers/gpu/drm/drm_fb_helper.c > >>> @@ -2105,6 +2105,30 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe > >>> } > >>> EXPORT_SYMBOL(drm_fb_helper_fill_var); > >>> > >>> +/** > >>> + * drm_fb_helper_fill_info - initializes fbdev information > >>> + * @info: fbdev instance to set up > >>> + * @fb_helper: fb helper instance to use as template > >>> + * > >>> + * > >>> + * Sets up the variable and fixed fbdev metainformation from the given fb helper > >>> + * instance and the drm framebuffer allocated in &drm_fb_helper.fb. > >>> + * > >>> + * Drivers should call this (or their equivalent setup code) from their > >>> + * &drm_fb_helper_funcs.fb_probe callback after having allocated the fbdev > >>> + * backing storage framebuffer. > >>> + */ > >>> +void drm_fb_helper_fill_info(struct fb_info *info, > >> > >> No need to pass in fb_info it's available as fb_helper->fbdev. Set by > >> drm_fb_helper_alloc_fbi(). > > > > Uh, I get to edit all the patches. Still, thanks for pointing out. > > > > I also realized that I've forgotten to add the dummy functiont for > > !CONFIG_FB, will fix that too. > > > >>> + struct drm_fb_helper *fb_helper) > >>> +{ > >>> + struct drm_framebuffer *fb = fb_helper->fb; > >>> + > >>> + drm_fb_helper_fill_fix(info, fb->pitches[0], fb->format->depth); > >>> + drm_fb_helper_fill_var(info, fb_helper, fb->width, fb->height); > >> > >> AFAIU fb->width/height can be different from sizes->fb_width/fb_height > >> when there's double/triple buffering. I belive you need to use the sizes > >> values here. > > > > Yup they're bigger than the actual screen size, but so is sizes: For > > overallocting we adjust sizes before we call ->fb_probe, and the > > ->fb_probe callbacks in turn then call this helper. So I think it should > > be all equivalent. Agreed or do I miss something? > > I did a test setting drm_fbdev_overalloc = 200: > > [drm:drm_setup_crtcs] desired mode 320x240 set on crtc 33 (0,0) > [drm:drm_fb_helper_generic_probe] surface width(320), height(480) and > bpp(16) > [drm:drm_fb_helper_generic_probe] fb_width(320), fb_height(240) > > sizes->surface_height != sizes->fb_height > > sizes->surface_height is used to create the framebuffer -> fb->height. > My driver doesn't allow me to create such a big framebuffer so I didn't > see the actual fb->height value come out the other end, but I still > believe you need to use sizes->fb_height. Hm right, I totally missed that. So the right interface would be: drm_fb_helper_fill_info(fb_helper, sizes); I also just noticed tha we could move the call to drm_fb_helper_fill_info into the helper itself, I think that would work for all drivers. -Daniel > > Noralf. > > > -Daniel > > > >> > >> Noralf. > >> > >>> + > >>> +} > >>> +EXPORT_SYMBOL(drm_fb_helper_fill_info); > >>> + > >>> static int drm_fb_helper_probe_connector_modes(struct drm_fb_helper *fb_helper, > >>> uint32_t maxX, > >>> uint32_t maxY) > >>> @@ -3163,8 +3187,7 @@ int drm_fb_helper_generic_probe(struct drm_fb_helper *fb_helper, > >>> #endif > >>> strcpy(fbi->fix.id, "DRM emulated"); > >>> > >>> - drm_fb_helper_fill_fix(fbi, fb->pitches[0], fb->format->depth); > >>> - drm_fb_helper_fill_var(fbi, fb_helper, sizes->fb_width, sizes->fb_height); > >>> + drm_fb_helper_fill_info(fbi, fb_helper); > >>> > >>> if (fb->funcs->dirty) { > >>> struct fb_ops *fbops; > >>> diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h > >>> index bb9acea61369..e8d92724f472 100644 > >>> --- a/include/drm/drm_fb_helper.h > >>> +++ b/include/drm/drm_fb_helper.h > >>> @@ -292,6 +292,7 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe > >>> uint32_t fb_width, uint32_t fb_height); > >>> void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch, > >>> uint32_t depth); > >>> +void drm_fb_helper_fill_info(struct fb_info *info, struct drm_fb_helper *fb_helper); > >>> > >>> void drm_fb_helper_unlink_fbi(struct drm_fb_helper *fb_helper); > >>> > >>> > >
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 5eaccd202f4f..34c4ed378796 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -2105,6 +2105,30 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe } EXPORT_SYMBOL(drm_fb_helper_fill_var); +/** + * drm_fb_helper_fill_info - initializes fbdev information + * @info: fbdev instance to set up + * @fb_helper: fb helper instance to use as template + * + * + * Sets up the variable and fixed fbdev metainformation from the given fb helper + * instance and the drm framebuffer allocated in &drm_fb_helper.fb. + * + * Drivers should call this (or their equivalent setup code) from their + * &drm_fb_helper_funcs.fb_probe callback after having allocated the fbdev + * backing storage framebuffer. + */ +void drm_fb_helper_fill_info(struct fb_info *info, + struct drm_fb_helper *fb_helper) +{ + struct drm_framebuffer *fb = fb_helper->fb; + + drm_fb_helper_fill_fix(info, fb->pitches[0], fb->format->depth); + drm_fb_helper_fill_var(info, fb_helper, fb->width, fb->height); + +} +EXPORT_SYMBOL(drm_fb_helper_fill_info); + static int drm_fb_helper_probe_connector_modes(struct drm_fb_helper *fb_helper, uint32_t maxX, uint32_t maxY) @@ -3163,8 +3187,7 @@ int drm_fb_helper_generic_probe(struct drm_fb_helper *fb_helper, #endif strcpy(fbi->fix.id, "DRM emulated"); - drm_fb_helper_fill_fix(fbi, fb->pitches[0], fb->format->depth); - drm_fb_helper_fill_var(fbi, fb_helper, sizes->fb_width, sizes->fb_height); + drm_fb_helper_fill_info(fbi, fb_helper); if (fb->funcs->dirty) { struct fb_ops *fbops; diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h index bb9acea61369..e8d92724f472 100644 --- a/include/drm/drm_fb_helper.h +++ b/include/drm/drm_fb_helper.h @@ -292,6 +292,7 @@ void drm_fb_helper_fill_var(struct fb_info *info, struct drm_fb_helper *fb_helpe uint32_t fb_width, uint32_t fb_height); void drm_fb_helper_fill_fix(struct fb_info *info, uint32_t pitch, uint32_t depth); +void drm_fb_helper_fill_info(struct fb_info *info, struct drm_fb_helper *fb_helper); void drm_fb_helper_unlink_fbi(struct drm_fb_helper *fb_helper);
The fbdev split between fix and var information is kinda pointless for drm drivers since everything is fixed: The fbdev emulation doesn't support changing modes at all. Create a new simplified helper and use it in the generic fbdev helper code. Follow-up patches will beef it up more and roll it out to all drivers. Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> --- drivers/gpu/drm/drm_fb_helper.c | 27 +++++++++++++++++++++++++-- include/drm/drm_fb_helper.h | 1 + 2 files changed, 26 insertions(+), 2 deletions(-)