Patchwork [1/3] qxl: add fb_size_mb and fb_size

login
register
mail settings
Submitter Alon Levy
Date June 11, 2012, 6:23 a.m.
Message ID <1339395841-31413-2-git-send-email-alevy@redhat.com>
Download mbox | patch
Permalink /patch/164078/
State New
Headers show

Comments

Alon Levy - June 11, 2012, 6:23 a.m.
In preperation for supporting a larger framebuffer for multiple monitors
on a single card, add a property to qxl fb_size_mb, and corresponding
byte sized fb_size, and use instead of VGA_RAM_SIZE.

Signed-off-by: Alon Levy <alevy@redhat.com>
---
 hw/qxl.c |   31 ++++++++++++++++++++-----------
 hw/qxl.h |    2 ++
 2 files changed, 22 insertions(+), 11 deletions(-)
Gerd Hoffmann - June 11, 2012, 7:42 a.m.
Hi,

> +        if (d->fb_size < fb) {
> +            d->fb_size = fb;
> +        }
> +        if (fb <= d->fb_size) {

This check doesn't make sense, it will always be true.

> +    if (qxl->fb_size_mb != -1 &&
> +        qxl->fb_size_mb * 1024 * 1024 < qxl->vga.vram_size) {
> +        qxl->fb_size = qxl->fb_size_mb * 1024 * 1024;
> +    } else {
> +        qxl->fb_size = VGA_RAM_SIZE;
> +    }

Reminds me that I have some pending work to make vga ram size
configurable which I should finish and repost ...

Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1

It probably makes sense to base this on top.

> +        DEFINE_PROP_UINT32("fb_size_mb", PCIQXLDevice, fb_size_mb, -1),

Maybe also rename this to vgamem_mb for consistency with standard vga.

cheers,
  Gerd
Alon Levy - June 11, 2012, 8:26 a.m.
On Mon, Jun 11, 2012 at 09:42:28AM +0200, Gerd Hoffmann wrote:
>   Hi,
> 
> > +        if (d->fb_size < fb) {
> > +            d->fb_size = fb;
> > +        }
> > +        if (fb <= d->fb_size) {
> 
> This check doesn't make sense, it will always be true.

My mistake. Will remove the first three lines, so fb_size setting will
be required to get larger modes.

> 
> > +    if (qxl->fb_size_mb != -1 &&
> > +        qxl->fb_size_mb * 1024 * 1024 < qxl->vga.vram_size) {
> > +        qxl->fb_size = qxl->fb_size_mb * 1024 * 1024;
> > +    } else {
> > +        qxl->fb_size = VGA_RAM_SIZE;
> > +    }
> 
> Reminds me that I have some pending work to make vga ram size
> configurable which I should finish and repost ...
> 
> Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1
> 
> It probably makes sense to base this on top.

So I can rebase on top and send the patches, with the understanding you
will do any later changes and repost the whole thing.

> 
> > +        DEFINE_PROP_UINT32("fb_size_mb", PCIQXLDevice, fb_size_mb, -1),
> 
> Maybe also rename this to vgamem_mb for consistency with standard vga.

Will change.

> 
> cheers,
>   Gerd
>
Gerd Hoffmann - June 11, 2012, 8:29 a.m.
Hi,

>> Reminds me that I have some pending work to make vga ram size
>> configurable which I should finish and repost ...
>>
>> Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1
>>
>> It probably makes sense to base this on top.
> 
> So I can rebase on top and send the patches, with the understanding you
> will do any later changes and repost the whole thing.

Yes, makes sense to put them into one series I think.

cheers,
  Gerd
Alon Levy - June 11, 2012, 8:35 a.m.
On Mon, Jun 11, 2012 at 10:29:59AM +0200, Gerd Hoffmann wrote:
>   Hi,
> 
> >> Reminds me that I have some pending work to make vga ram size
> >> configurable which I should finish and repost ...
> >>
> >> Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1
> >>
> >> It probably makes sense to base this on top.
> > 
> > So I can rebase on top and send the patches, with the understanding you
> > will do any later changes and repost the whole thing.
> 
> Yes, makes sense to put them into one series I think.

You dropped me from the to field, please don't.

Hope you didn't miss the one liner async_complete reset patch.

> 
> cheers,
>   Gerd
> 
>
Alon Levy - June 11, 2012, 9:33 a.m.
On Mon, Jun 11, 2012 at 10:29:59AM +0200, Gerd Hoffmann wrote:
>   Hi,
> 
> >> Reminds me that I have some pending work to make vga ram size
> >> configurable which I should finish and repost ...
> >>
> >> Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1
> >>
> >> It probably makes sense to base this on top.
> > 
> > So I can rebase on top and send the patches, with the understanding you
> > will do any later changes and repost the whole thing.
> 
> Yes, makes sense to put them into one series I think.

git://people.freedesktop.org/~alon/qemu fb_async_fix.v2

> 
> cheers,
>   Gerd
> 
>
Gerd Hoffmann - June 11, 2012, 10:18 a.m.
On 06/11/12 11:33, Alon Levy wrote:
> On Mon, Jun 11, 2012 at 10:29:59AM +0200, Gerd Hoffmann wrote:
>>   Hi,
>>
>>>> Reminds me that I have some pending work to make vga ram size
>>>> configurable which I should finish and repost ...
>>>>
>>>> Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1
>>>>
>>>> It probably makes sense to base this on top.
>>>
>>> So I can rebase on top and send the patches, with the understanding you
>>> will do any later changes and repost the whole thing.
>>
>> Yes, makes sense to put them into one series I think.
> 
> git://people.freedesktop.org/~alon/qemu fb_async_fix.v2

http://www.kraxel.org/cgit/qemu/log/?h=vga.2

Did some small fixups [no need for the -1 dance, the reason the other
properties do that is that we have one accepting byte values and one
accepting megabyte values].

looks ok to you?

cheers,
  Gerd
Alon Levy - June 11, 2012, 10:23 a.m.
On Mon, Jun 11, 2012 at 12:18:14PM +0200, Gerd Hoffmann wrote:
> On 06/11/12 11:33, Alon Levy wrote:
> > On Mon, Jun 11, 2012 at 10:29:59AM +0200, Gerd Hoffmann wrote:
> >>   Hi,
> >>
> >>>> Reminds me that I have some pending work to make vga ram size
> >>>> configurable which I should finish and repost ...
> >>>>
> >>>> Current state pushed to http://www.kraxel.org/cgit/qemu/log/?h=vga.1
> >>>>
> >>>> It probably makes sense to base this on top.
> >>>
> >>> So I can rebase on top and send the patches, with the understanding you
> >>> will do any later changes and repost the whole thing.
> >>
> >> Yes, makes sense to put them into one series I think.
> > 
> > git://people.freedesktop.org/~alon/qemu fb_async_fix.v2
> 
> http://www.kraxel.org/cgit/qemu/log/?h=vga.2
> 
> Did some small fixups [no need for the -1 dance, the reason the other
> properties do that is that we have one accepting byte values and one
> accepting megabyte values].
> 
> looks ok to you?

Looks like you squashed and removed the -1 (you're right of course, it
was just me doing a straight copy), other then that I didn't notice
anything. Looks good.

> 
> cheers,
>   Gerd

Patch

diff --git a/hw/qxl.c b/hw/qxl.c
index b5e53ce..a9b4fd1 100644
--- a/hw/qxl.c
+++ b/hw/qxl.c
@@ -114,20 +114,16 @@  static QXLMode qxl_modes[] = {
     QXL_MODE_EX(1600, 1200),
     QXL_MODE_EX(1680, 1050),
     QXL_MODE_EX(1920, 1080),
-#if VGA_RAM_SIZE >= (16 * 1024 * 1024)
     /* these modes need more than 8 MB video memory */
     QXL_MODE_EX(1920, 1200),
     QXL_MODE_EX(1920, 1440),
     QXL_MODE_EX(2048, 1536),
     QXL_MODE_EX(2560, 1440),
     QXL_MODE_EX(2560, 1600),
-#endif
-#if VGA_RAM_SIZE >= (32 * 1024 * 1024)
     /* these modes need more than 16 MB video memory */
     QXL_MODE_EX(2560, 2048),
     QXL_MODE_EX(2800, 2100),
     QXL_MODE_EX(3200, 2400),
-#endif
 };
 
 static PCIQXLDevice *qxl0;
@@ -284,6 +280,7 @@  static inline uint32_t msb_mask(uint32_t val)
 static ram_addr_t qxl_rom_size(void)
 {
     uint32_t rom_size = sizeof(QXLRom) + sizeof(QXLModes) + sizeof(qxl_modes);
+
     rom_size = MAX(rom_size, TARGET_PAGE_SIZE);
     rom_size = msb_mask(rom_size * 2 - 1);
     return rom_size;
@@ -296,8 +293,9 @@  static void init_qxl_rom(PCIQXLDevice *d)
     uint32_t ram_header_size;
     uint32_t surface0_area_size;
     uint32_t num_pages;
-    uint32_t fb, maxfb = 0;
+    uint32_t fb;
     int i;
+    int n_modes = 0;
 
     memset(rom, 0, d->rom_size);
 
@@ -312,11 +310,13 @@  static void init_qxl_rom(PCIQXLDevice *d)
     rom->slots_end     = NUM_MEMSLOTS - 1;
     rom->n_surfaces    = cpu_to_le32(NUM_SURFACES);
 
-    modes->n_modes     = cpu_to_le32(ARRAY_SIZE(qxl_modes));
     for (i = 0; i < modes->n_modes; i++) {
         fb = qxl_modes[i].y_res * qxl_modes[i].stride;
-        if (maxfb < fb) {
-            maxfb = fb;
+        if (d->fb_size < fb) {
+            d->fb_size = fb;
+        }
+        if (fb <= d->fb_size) {
+            n_modes++;
         }
         modes->modes[i].id          = cpu_to_le32(i);
         modes->modes[i].x_res       = cpu_to_le32(qxl_modes[i].x_res);
@@ -327,11 +327,13 @@  static void init_qxl_rom(PCIQXLDevice *d)
         modes->modes[i].y_mili      = cpu_to_le32(qxl_modes[i].y_mili);
         modes->modes[i].orientation = cpu_to_le32(qxl_modes[i].orientation);
     }
-    if (maxfb < VGA_RAM_SIZE && d->id == 0)
-        maxfb = VGA_RAM_SIZE;
+    if (d->fb_size < VGA_RAM_SIZE && d->id == 0) {
+        d->fb_size = VGA_RAM_SIZE;
+    }
+    modes->n_modes     = cpu_to_le32(n_modes);
 
     ram_header_size    = ALIGN(sizeof(QXLRam), 4096);
-    surface0_area_size = ALIGN(maxfb, 4096);
+    surface0_area_size = ALIGN(d->fb_size, 4096);
     num_pages          = d->vga.vram_size;
     num_pages         -= ram_header_size;
     num_pages         -= surface0_area_size;
@@ -1720,6 +1722,12 @@  static void qxl_init_ramsize(PCIQXLDevice *qxl, uint32_t ram_min_mb)
     if (qxl->vga.vram_size < ram_min_mb * 1024 * 1024) {
         qxl->vga.vram_size = ram_min_mb * 1024 * 1024;
     }
+    if (qxl->fb_size_mb != -1 &&
+        qxl->fb_size_mb * 1024 * 1024 < qxl->vga.vram_size) {
+        qxl->fb_size = qxl->fb_size_mb * 1024 * 1024;
+    } else {
+        qxl->fb_size = VGA_RAM_SIZE;
+    }
 
     /* vram32 (surfaces, 32bit, bar 1) */
     if (qxl->vram32_size_mb != -1) {
@@ -2052,6 +2060,7 @@  static Property qxl_properties[] = {
         DEFINE_PROP_UINT32("ram_size_mb",  PCIQXLDevice, ram_size_mb, -1),
         DEFINE_PROP_UINT32("vram_size_mb", PCIQXLDevice, vram32_size_mb, -1),
         DEFINE_PROP_UINT32("vram64_size_mb", PCIQXLDevice, vram_size_mb, -1),
+        DEFINE_PROP_UINT32("fb_size_mb", PCIQXLDevice, fb_size_mb, -1),
         DEFINE_PROP_END_OF_LIST(),
 };
 
diff --git a/hw/qxl.h b/hw/qxl.h
index a4ab7cc..1b4ec16 100644
--- a/hw/qxl.h
+++ b/hw/qxl.h
@@ -84,6 +84,7 @@  typedef struct PCIQXLDevice {
     QXLReleaseInfo     *last_release;
     uint32_t           last_release_offset;
     uint32_t           oom_running;
+    uint32_t           fb_size;
 
     /* rom pci bar */
     QXLRom             shadow_rom;
@@ -105,6 +106,7 @@  typedef struct PCIQXLDevice {
     uint32_t          ram_size_mb;
     uint32_t          vram_size_mb;
     uint32_t          vram32_size_mb;
+    uint32_t          fb_size_mb;
 
     /* qxl_render_update state */
     int                render_update_cookie_num;