[igt-dev] [PATCH i-g-t] tests/i915/gem_exec_faulting_reloc: Use a coherent mapping

Dixit, Ashutosh ashutosh.dixit at intel.com
Sat Feb 1 03:55:49 UTC 2020


On Fri, 31 Jan 2020 16:11:35 -0800, Antonio Argenziano wrote:
>
> With mappable aperture not being available on some platforms, use a
> different mapping. While this is changing a portion of the test, it
> should keep intact the spirit of the same by submitting relocations to
> faulting addresses.

Apart from the nit's on the comments below:

Reviewed-by: Ashutosh Dixit <ashutosh.dixit at intel.com>

>
> Signed-off-by: Antonio Argenziano <antonio.argenziano at intel.com>
> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> ---
>  tests/i915/gem_exec_faulting_reloc.c | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
>
> diff --git a/tests/i915/gem_exec_faulting_reloc.c b/tests/i915/gem_exec_faulting_reloc.c
> index c2aea039..82d67a68 100644
> --- a/tests/i915/gem_exec_faulting_reloc.c
> +++ b/tests/i915/gem_exec_faulting_reloc.c
> @@ -39,12 +39,11 @@
>  #include <sys/time.h>
>  #include "drm.h"
>
> -/* Testcase: Submit patches with relocations in memory that will fault
> +/* Testcase: Submit batches with relocations in memory that will fault
>   *
> - * To be really evil, use a gtt mmap for them.

"To be really evil, use a gtt mmap for them when available."?

> @@ -203,7 +202,7 @@ static void run(int object_size)
>
>	handle_relocs = gem_create(fd, 4096);
>	gem_write(fd, handle_relocs, 0, reloc, sizeof(reloc));
> -	gtt_relocs = gem_mmap__gtt(fd, handle_relocs, 4096,
> +	relocs = gem_mmap__device_coherent(fd, handle_relocs, 0, 4096,
>				   PROT_READ | PROT_WRITE);
>
>	exec[2].handle = handle;
> @@ -212,7 +211,7 @@ static void run(int object_size)
>	else
>		exec[2].relocation_count = len > 40 ? 4 : 2;
>	/* A newly mmap gtt bo will fault on first access. */

/* A newly mmap bo will fault on first access. */ ?


More information about the igt-dev mailing list