Re: [Qemu-devel] [PATCH] bcm2835_property: use cached values when querying framebuffer

[Andrew Baumann]

> From: Qemu-devel [mailto:qemu-devel-
> address@hidden On Behalf Of Sylvain
> Garrigues
> Sent: Thursday, 21 April 2016 12:42
> 
> As the framebuffer settings are copied into the result message before it is
> reconfigured, inconsistent behavior can happen when, for instance, you set
> with a sinle message the width, height, and depth, and ask at the same time to
> allocate the buffer and get the pitch and the size.
> 
> In this case, the reported pitch and size would be incorrect as they were
> computed with the initial values of width, height and depth, not the ones the
> client requested.
> 
> Signed-off-by: Sylvain Garrigues <address@hidden>
> ---
>  hw/misc/bcm2835_property.c | 31 ++++++++++++++++++++++---------
>  1 file changed, 22 insertions(+), 9 deletions(-)
> 
> diff --git a/hw/misc/bcm2835_property.c b/hw/misc/bcm2835_property.c
> index 530411f..0102144 100644
> --- a/hw/misc/bcm2835_property.c
> +++ b/hw/misc/bcm2835_property.c
> @@ -21,6 +21,7 @@ static void
> bcm2835_property_mbox_push(BCM2835PropertyState *s, uint32_t value)
>      int n;
>      uint32_t offset, length, color;
>      uint32_t xres, yres, xoffset, yoffset, bpp, pixo, alpha;
> +    uint32_t tmp_xres, tmp_yres, tmp_xoffset, tmp_yoffset, tmp_bpp,
> + tmp_pixo, tmp_alpha;
>      uint32_t *newxres = NULL, *newyres = NULL, *newxoffset = NULL,
>          *newyoffset = NULL, *newbpp = NULL, *newpixo = NULL, *newalpha =
> NULL;
> 
> @@ -139,7 +140,10 @@ static void
> bcm2835_property_mbox_push(BCM2835PropertyState *s, uint32_t value)
> 
>          case 0x00040001: /* Allocate buffer */
>              stl_le_phys(&s->dma_as, value + 12, s->fbdev->base);
> -            stl_le_phys(&s->dma_as, value + 16, s->fbdev->size);
> +            tmp_xres = newxres != NULL ? *newxres : s->fbdev->xres;
> +            tmp_yres = newyres != NULL ? *newyres : s->fbdev->yres;
> +            tmp_bpp = newbpp != NULL ? *newbpp : s->fbdev->bpp;
> +            stl_le_phys(&s->dma_as, value + 16, tmp_xres * tmp_yres *
> + (tmp_bpp >> 3));

Personal style nit: I prefer * 8 rather than >> 3, because it's more 
immediately obvious what you're computing, a trivial optimisation for the 
compiler, and in this specific example wouldn't need the brackets to ensure the 
correct operator precedence.

But in any case,
Reviewed-by: Andrew Baumann <address@hidden>

Thanks,
Andrew