Re: [PATCH v4 3/7] drm/fb-helper: Add fb_deferred_io support

[Daniel Vetter]

On Fri, Apr 29, 2016 at 04:47:08PM +0200, Noralf Trønnes wrote:
> 
> Den 29.04.2016 14:50, skrev Tomi Valkeinen:
> >Hi,
> >
> >On 28/04/16 18:18, Noralf Trønnes wrote:
> >>This adds deferred io support to drm_fb_helper.
> >>The fbdev framebuffer changes are flushed using the callback
> >>(struct drm_framebuffer *)->funcs->dirty() by a dedicated worker
> >>ensuring that it always runs in process context.
> >>
> >>Signed-off-by: Noralf Trønnes <noralf@xxxxxxxxxxx>
> >>Reviewed-by: Daniel Vetter <daniel.vetter@xxxxxxxx>
> >>---
> >Thanks for the series! Unfortunately I haven't been able to follow the
> >discussions properly, so I hope my questions haven't been covered earlier.
> >
> >>Changes since v3:
> >>- Don't use forward decl, move drm_fb_helper_dirty_work()
> >>- Use DIV_ROUND_UP in drm_fb_helper_deferred_io()
> >>
> >>Changes since v2:
> >>- FB_DEFERRED_IO is now always selected by DRM_KMS_FB_HELPER, ifdef removed
> >>- The drm_clip_rect utility functions are dropped, so open code it
> >>- docs: use & to denote structs
> >>
> >>Changes since v1:
> >>- Use a dedicated worker to run the framebuffer flushing like qxl does
> >>- Add parameter descriptions to drm_fb_helper_deferred_io
> >>
> >>  drivers/gpu/drm/Kconfig         |   1 +
> >>  drivers/gpu/drm/drm_fb_helper.c | 103 +++++++++++++++++++++++++++++++++++++++-
> >>  include/drm/drm_fb_helper.h     |  15 ++++++
> >>  3 files changed, 118 insertions(+), 1 deletion(-)
> >>
> >>diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
> >>index 9e4f2f1..8e6f34b 100644
> >>--- a/drivers/gpu/drm/Kconfig
> >>+++ b/drivers/gpu/drm/Kconfig
> >>@@ -52,6 +52,7 @@ config DRM_KMS_FB_HELPER
> >>  	select FB_CFB_FILLRECT
> >>  	select FB_CFB_COPYAREA
> >>  	select FB_CFB_IMAGEBLIT
> >>+	select FB_DEFERRED_IO
> >>  	help
> >>  	  FBDEV helpers for KMS drivers.
> >>
> >>diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> >>index 855108e..62f849f 100644
> >>--- a/drivers/gpu/drm/drm_fb_helper.c
> >>+++ b/drivers/gpu/drm/drm_fb_helper.c
> >>@@ -84,6 +84,15 @@ static LIST_HEAD(kernel_fb_helper_list);
> >>   * and set up an initial configuration using the detected hardware, drivers
> >>   * should call drm_fb_helper_single_add_all_connectors() followed by
> >>   * drm_fb_helper_initial_config().
> >>+ *
> >>+ * If CONFIG_FB_DEFERRED_IO is enabled and &drm_framebuffer_funcs ->dirty is
> >>+ * set, the drm_fb_helper_{cfb,sys}_{write,fillrect,copyarea,imageblit}
> >>+ * functions will accumulate changes and schedule &fb_helper .dirty_work to run
> >>+ * right away. This worker then calls the dirty() function ensuring that it
> >>+ * will always run in process context since the fb_*() function could be
> >>+ * running in atomic context. If drm_fb_helper_deferred_io() is used as the
> >Who's calling {write,fillrect,copyarea,imageblit} in an atomic context?
> >That sounds like a very bad idea to me...
> 
> I haven't verified it myself, but took it as fact based on
> commit: bcb39af4486be07e896fc374a2336bad3104ae0a

I've added a citation to this commit as the reason why we need to handle
atomic, to make sure Tomi's question is answered for posterity.
-Daniel

> 
> drm/udl: make usage as a console safer
> Okay you don't really want to use udl devices as your console, but if
> you are unlucky enough to do so, you run into a lot of schedule while atomic
> due to printk being called from all sorts of funky places. So check if we
> are in an atomic context, and queue the damage for later, the next printk
> should cause it to appear. This isn't ideal, but it is simple, and seems to
> work okay in my testing here.
> 
> (dirty area idea came from xenfb)
> 
> fixes a bunch of sleeping while atomic issues running fbcon on udl devices.
> 
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Dave Airlie <airlied@xxxxxxxxxx>
> 
> >If this is only for accumulating changes, I think it may be better to
> >leave that to the driver as it may have better idea of how to accumulate.
> >
> >But, of course, this is a helper, so if all the drivers use this kind of
> >accumulation, it makes sense =).
> 
> qxl already accumulates damage this way and upcoming tinydrm also.
> udl has handled this damage inline except when in_atomic() which results
> in damage being deferred to the next fb_*() call.
> 
> It is possible for drivers to have their own fb_*() handling and still
> use drm_fb_helper_deferred_io().
> 
> 
> For those who likes details, this is fbcon unblanking which results in a
> full display update on a 320x240 display, shell with blinking cursor on
> the last line of the console:
> 
> [ 5505.164150] drm_fb_helper_dirty: x=152,width=8,y=224,height=16
> [ 5505.164186] [drm:drm_atomic_state_init] Allocated atomic state b859e400
> [...more drm atomic msgs...]
> [ 5505.164713] drm_fb_helper_dirty: x=152,width=8,y=224,height=16
> [ 5505.164746] [drm:drm_atomic_state_init] Allocated atomic state b859e440
> [...more drm atomic msgs...]
> [ 5505.165086] drm_fb_helper_dirty: x=0,width=320,y=0,height=16
> [ 5505.165153] drm_fb_helper_dirty: x=0,width=320,y=16,height=16
> [ 5505.165220] drm_fb_helper_dirty: x=0,width=320,y=32,height=16
> [ 5505.165287] drm_fb_helper_dirty: x=0,width=320,y=48,height=16
> [ 5505.165354] drm_fb_helper_dirty: x=0,width=320,y=64,height=16
> [ 5505.165420] drm_fb_helper_dirty: x=0,width=320,y=80,height=16
> [ 5505.165487] drm_fb_helper_dirty: x=0,width=320,y=96,height=16
> [ 5505.165553] drm_fb_helper_dirty: x=0,width=320,y=112,height=16
> [ 5505.165619] drm_fb_helper_dirty: x=0,width=320,y=128,height=16
> [ 5505.165686] drm_fb_helper_dirty: x=0,width=320,y=144,height=16
> [ 5505.165752] drm_fb_helper_dirty: x=0,width=320,y=160,height=16
> [ 5505.165818] drm_fb_helper_dirty: x=0,width=320,y=176,height=16
> [ 5505.165884] drm_fb_helper_dirty: x=0,width=320,y=192,height=16
> [ 5505.165949] drm_fb_helper_dirty: x=0,width=320,y=208,height=16
> [ 5505.165978] drm_fb_helper_dirty: x=0,width=112,y=224,height=16
> [ 5505.165988] drm_fb_helper_dirty: x=112,width=8,y=224,height=16
> [ 5505.166002] drm_fb_helper_dirty: x=120,width=24,y=224,height=16
> [ 5505.166041] drm_fb_helper_dirty: x=144,width=176,y=224,height=16
> [ 5505.166058] drm_fb_helper_dirty: x=152,width=8,y=224,height=16
> [ 5505.166079] drm_fb_helper_dirty: x=152,width=8,y=224,height=16
> [ 5505.166424] drm_fb_helper_dirty_work: x1=0,x2=320,y1=0,y2=240
> [ 5505.166452] adafruit-tft spi0.0: mipi_dbi_dirtyfb: vmem=bac40000, x1=0,
> x2=320, y1=0, y2=240
> 
> Cursor blinking:
> 
> [ 5505.363478] drm_fb_helper_dirty: x=152,width=8,y=224,height=16
> [ 5505.363509] drm_fb_helper_dirty_work: x1=152,x2=160,y1=224,y2=240
> [ 5505.363539] adafruit-tft spi0.0: mipi_dbi_dirtyfb: vmem=bac40000, x1=152,
> x2=160, y1=224, y2=240
> 
> [ 5505.563488] drm_fb_helper_dirty: x=152,width=8,y=224,height=16
> [ 5505.563514] drm_fb_helper_dirty_work: x1=152,x2=160,y1=224,y2=240
> [ 5505.563542] adafruit-tft spi0.0: mipi_dbi_dirtyfb: vmem=bac40000, x1=152,
> x2=160, y1=224, y2=240
> 
> 
> Noralf.
> 
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch