RE: [PATCH v4] video: hyperv: hyperv_fb: Use physical memory for fb on HyperV Gen 1 VMs.

[Dexuan Cui]

> From: Michael Kelley <mikelley@xxxxxxxxxxxxx>
> Sent: Monday, December 9, 2019 8:33 AM
> To: Wei Hu <weh@xxxxxxxxxxxxx>; b.zolnierkie@xxxxxxxxxxx; KY
> Srinivasan <kys@xxxxxxxxxxxxx>; Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>;
> Stephen Hemminger <sthemmin@xxxxxxxxxxxxx>; sashal@xxxxxxxxxx;
> hch@xxxxxx; m.szyprowski@xxxxxxxxxxx; mchehab+samsung@xxxxxxxxxx;
> sam@xxxxxxxxxxxx; gregkh@xxxxxxxxxxxxxxxxxxx;
> alexandre.belloni@xxxxxxxxxxx; info@xxxxxxxxx; arnd@xxxxxxxx;
> dri-devel@xxxxxxxxxxxxxxxxxxxxx; linux-fbdev@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx; linux-hyperv@xxxxxxxxxxxxxxx; Dexuan Cui
> <decui@xxxxxxxxxxxxx>
> Cc: kbuild test robot <lkp@xxxxxxxxx>
> Subject: RE: [PATCH v4] video: hyperv: hyperv_fb: Use physical memory for
> fb on HyperV Gen 1 VMs.
> 
> From: Wei Hu <weh@xxxxxxxxxxxxx> Sent: Sunday, December 8, 2019 11:58
> PM
> >
> > On Hyper-V, Generation 1 VMs can directly use VM's physical memory for
> > their framebuffers. This can improve the efficiency of framebuffer and
> > overall performance for VM. The physical memory assigned to framebuffer
> > must be contiguous. We use CMA allocator to get contiguous physicial
> > memory when the framebuffer size is greater than 4MB. For size under
> > 4MB, we use alloc_pages to achieve this.
> >
> > To enable framebuffer memory allocation from CMA, supply a kernel
> > parameter to give enough space to CMA allocator at boot time. For
> > example:
> >     cma=130m
> > This gives 130MB memory to CAM allocator that can be allocated to
> > framebuffer. If this fails, we fall back to the old way of using
> > mmio for framebuffer.
> >
> > Reported-by: kbuild test robot <lkp@xxxxxxxxx>
> > Signed-off-by: Wei Hu <weh@xxxxxxxxxxxxx>
> > ---
> >     v2: Incorporated review comments form hch@xxxxxx, Michael Kelley
> and
> >     Dexuan Cui
> >     - Use dma_alloc_coherent to allocate large contiguous memory
> >     - Use phys_addr_t for physical addresses
> >     - Corrected a few spelling errors and minor cleanups
> >     - Also tested on 32 bit Ubuntu guest
> >     v3: Fixed a build issue reported by kbuild test robot and incorported
> >     some review comments from Michael Kelley
> >     - Add CMA check to avoid link failure
> >     - Fixed small memory leak introduced by alloc_apertures
> >     - Cleaned up so code
> >     v4: Removed request_pages variable as it is no longer needed
> >
> >  drivers/video/fbdev/Kconfig     |   1 +
> >  drivers/video/fbdev/hyperv_fb.c | 182
> +++++++++++++++++++++++++-------
> >  2 files changed, 144 insertions(+), 39 deletions(-)
> >
> 
> Reviewed-by: Michael Kelley <mikelley@xxxxxxxxxxxxx>

Tested-by: Dexuan Cui <decui@xxxxxxxxxxxxx>

For a Gen-1 VM running on recent Hyper-V hosts, this patch can greatly 
reduce the CPU utilization because it avoids the slow data copy from the 
shadow framebuffer to the MMIO framebuffer, and hence it resolves the
"blurred screen" issue when we output a lot of characters on the text-mode
ternimal (e.g. "dmesg").