Re: [Qemu-devel] [PATCH V2 5/5] vga-cirrus: Workaround during restore when using Xen.

[Jan Kiszka <jan.kiszka@web.de>]

[-- Attachment #1: Type: text/plain, Size: 2387 bytes --]

On 2011-12-09 22:54, Anthony PERARD wrote:
> During the initialisation of the machine at restore time, the access to the
> VRAM will fail because QEMU does not know yet the right guest address to map,
> so the vram_ptr is NULL.
> 
> So this patch avoid using a NULL pointer during initialisation, and try to get
> another vram_ptr if the call failed the first time.
> 
> Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
> ---
>  hw/cirrus_vga.c |   16 +++++++++++++---
>  1 files changed, 13 insertions(+), 3 deletions(-)
> 
> diff --git a/hw/cirrus_vga.c b/hw/cirrus_vga.c
> index c7e365b..2e049c9 100644
> --- a/hw/cirrus_vga.c
> +++ b/hw/cirrus_vga.c
> @@ -32,6 +32,7 @@
>  #include "console.h"
>  #include "vga_int.h"
>  #include "loader.h"
> +#include "sysemu.h"
>  
>  /*
>   * TODO:
> @@ -2696,6 +2697,13 @@ static int cirrus_post_load(void *opaque, int version_id)
>      s->vga.gr[0x01] = s->cirrus_shadow_gr1 & 0x0f;
>  
>      cirrus_update_memory_access(s);
> +    if (!s->vga.vram_ptr) {
> +        /* At this point vga.vram_ptr can be invalid on Xen because we need to
> +         * know the position of the videoram in the guest physical address space in
> +         * order to be able to map it. After cirrus_update_memory_access we do know
> +         * where the videoram is, so let's map it now. */
> +        s->vga.vram_ptr = memory_region_get_ram_ptr(&s->vga.vram);
> +    }
>      /* force refresh */
>      s->vga.graphic_mode = -1;
>      cirrus_update_bank_ptr(s, 0);
> @@ -2784,9 +2792,11 @@ static void cirrus_reset(void *opaque)
>      }
>      s->vga.cr[0x27] = s->device_id;
>  
> -    /* Win2K seems to assume that the pattern buffer is at 0xff
> -       initially ! */
> -    memset(s->vga.vram_ptr, 0xff, s->real_vram_size);
> +    if (!runstate_check(RUN_STATE_PREMIGRATE)) {
> +        /* Win2K seems to assume that the pattern buffer is at 0xff
> +           initially ! */
> +        memset(s->vga.vram_ptr, 0xff, s->real_vram_size);
> +    }
>  
>      s->cirrus_hidden_dac_lockindex = 5;
>      s->cirrus_hidden_dac_data = 0;

Is there really no way to fix this properly in the Xen layer? This looks
highly fragile as specifically the second hunk appears unrelated to Xen.

Also, is this the only device affected by the shortcoming? What about
other VGA adapters e.g.?

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]