[igt-dev] [PATCH i-g-t] i915: Skip two subsets when pread/pwrite are unavailable

Zbigniew Kempczyński zbigniew.kempczynski at intel.com
Fri Mar 26 06:58:53 UTC 2021


On Thu, Mar 25, 2021 at 09:33:03PM -0700, Ashutosh Dixit wrote:
> The pread/pwrite replacement implemented in
> ad5eb02eb3f1 ("lib/ioctl_wrappers: Keep IGT working without pread/pwrite
> ioctls") uses gem_set_domain which pins all pages which have to be
> read/written. When the read/write size is large this causes gem_set_domain
> to return -ENOMEM with a trace such as:
> 
> ioctl_wrappers-CRITICAL: Test assertion failure function gem_set_domain, file ../lib/ioctl_wrappers.c:563:
> ioctl_wrappers-CRITICAL: Failed assertion: __gem_set_domain(fd, handle, read, write) == 0
> ioctl_wrappers-CRITICAL: Last errno: 12, Cannot allocate memory
> ioctl_wrappers-CRITICAL: error: -12 != 0
> igt_core-INFO: Stack trace:
> igt_core-INFO:   #0 ../lib/igt_core.c:1746 __igt_fail_assert()
> igt_core-INFO:   #1 [gem_set_domain+0x44]
> igt_core-INFO:   #2 ../lib/ioctl_wrappers.c:367 gem_write()
> igt_core-INFO:   #3 ../tests/prime_mmap.c:67 test_aperture_limit()
> igt_core-INFO:   #4 ../tests/prime_mmap.c:578 __real_main530()
> igt_core-INFO:   #5 ../tests/prime_mmap.c:530 main()
> 
> Skip these tests when pread/pwrite are unavailable because they cannot pass
> with the pread/pwrite replacement.
> 
> Signed-off-by: Ashutosh Dixit <ashutosh.dixit at intel.com>
> ---
>  tests/i915/gem_exec_params.c | 2 ++
>  tests/prime_mmap.c           | 2 ++
>  2 files changed, 4 insertions(+)
> 
> diff --git a/tests/i915/gem_exec_params.c b/tests/i915/gem_exec_params.c
> index 6840cf40ce..a7bcd39d32 100644
> --- a/tests/i915/gem_exec_params.c
> +++ b/tests/i915/gem_exec_params.c
> @@ -314,6 +314,8 @@ static void test_larger_than_life_batch(int fd)
>  	 */
>  	igt_require(size < gem_aperture_size(fd));
>  	intel_require_memory(2, size, CHECK_RAM); /* batch + shadow */
> +	/* Prevent gem_set_domain -ENOMEM failures */
> +	gem_require_pread_pwrite(fd);

I have doubts regarding that test. 

What is real cause we got -ENOMEM in that subtest? And on which 
machines? I've enforced on my SKL (gen9) to use mmap/set_domain 
and I see success on that test.

Why require:

	igt_require(size < gem_aperture_size(fd));

comes after:

	uint64_t size = 1ULL << 32; /* batch_len is __u32 as per the ABI */
	struct drm_i915_gem_exec_object2 exec = {
		.handle = batch_create_size(fd, size),
	};

I think require before mmap/set_domain would catch this issue.

--
Zbigniew
 
>  
>  	__for_each_physical_engine(fd, e) {
>  		/* Keep the batch_len implicit [0] */
> diff --git a/tests/prime_mmap.c b/tests/prime_mmap.c
> index cdf2d51497..af313917da 100644
> --- a/tests/prime_mmap.c
> +++ b/tests/prime_mmap.c
> @@ -451,6 +451,8 @@ test_aperture_limit(void)
>  	uint64_t size1 = (gem_mappable_aperture_size(fd) * 7) / 8;
>  	uint64_t size2 = (gem_mappable_aperture_size(fd) * 3) / 8;
>  
> +	/* Prevent gem_set_domain -ENOMEM failures */
> +	gem_require_pread_pwrite(fd);
>  	handle1 = gem_create(fd, size1);
>  	fill_bo(handle1, BO_SIZE);
>  
> -- 
> 2.29.2
> 


More information about the igt-dev mailing list