[igt-dev] [PATCH i-g-t v4] i915/i915_hangman: fail only after freeing spinners

[Zbigniew Kempczyński]

On Tue, Aug 30, 2022 at 10:48:36AM +0200, Kamil Konieczny wrote:
> Failed checks may cause following tests fail, so check for
> errors only after all spinners are released.
> 
> v2: be verbose about what failed (Chris)
> v3: simplify declaration, add separate wait var (Zbigniew)
> v4: removed bitshift (Zbigniew), added storing last not
>     expected error for fence stat
> 
> Cc: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> Signed-off-by: Kamil Konieczny <kamil.konieczny at linux.intel.com>
> ---
>  tests/i915/i915_hangman.c | 34 ++++++++++++++++++++++++++++------
>  1 file changed, 28 insertions(+), 6 deletions(-)
> 
> diff --git a/tests/i915/i915_hangman.c b/tests/i915/i915_hangman.c
> index c7d69fdd..ddbe1067 100644
> --- a/tests/i915/i915_hangman.c
> +++ b/tests/i915/i915_hangman.c
> @@ -295,6 +295,20 @@ static void context_unban(int fd, unsigned ctx)
>  	gem_context_set_param(fd, &param);
>  }
>  
> +static void chk_err(int *dst, int err, int expected)
> +{
> +	if (err == expected)
> +		return;
> +
> +	*dst = err;
> +}

Looks good for me. Single unexpected value will overwrite the dst status.

Reviewed-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>

--
Zbigniew

> +
> +#define ERR_HANG_WAIT  0
> +#define ERR_HANG_STAT  1
> +#define ERR_FENCE_BUSY 2
> +#define ERR_FENCE_END  3
> +#define ERR_FENCE_STAT 4
> +
>  static void
>  test_engine_hang(const intel_ctx_t *ctx,
>  		 const struct intel_execution_engine2 *e, unsigned int flags)
> @@ -305,6 +319,7 @@ test_engine_hang(const intel_ctx_t *ctx,
>  	IGT_LIST_HEAD(list);
>  	uint64_t ahnd = get_reloc_ahnd(device, ctx->id), ahndN;
>  	int num_ctx;
> +	int err[ERR_FENCE_STAT + 1];
>  
>  	igt_skip_on(flags & IGT_SPIN_INVALID_CS &&
>  		    gem_engine_has_cmdparser(device, &ctx->cfg, e->flags));
> @@ -340,18 +355,20 @@ test_engine_hang(const intel_ctx_t *ctx,
>  				      flags));
>  
>  	/* Wait for the hangcheck to terminate the hanger */
> -	igt_assert(sync_fence_wait(spin->out_fence, 30000) == 0); /* 30s */
> -	igt_assert_eq(sync_fence_status(spin->out_fence), -EIO);
> +	err[ERR_HANG_WAIT] = sync_fence_wait(spin->out_fence, 30000); /* 30s */
> +	err[ERR_HANG_STAT] = sync_fence_status(spin->out_fence); /* -EIO */
>  	igt_spin_free(device, spin);
>  
>  	/* But no other engines/clients should be affected */
> +	err[ERR_FENCE_BUSY] = -ETIME;
> +	err[ERR_FENCE_END] = 0;
> +	err[ERR_FENCE_STAT] = 1;
>  	igt_list_for_each_entry_safe(spin, next, &list, link) {
>  		ahndN = spin->opts.ahnd;
> -		igt_assert(sync_fence_wait(spin->out_fence, 0) == -ETIME);
> +		chk_err(err+ERR_FENCE_BUSY, sync_fence_wait(spin->out_fence, 0), -ETIME);
>  		igt_spin_end(spin);
> -
> -		igt_assert(sync_fence_wait(spin->out_fence, 500) == 0);
> -		igt_assert_eq(sync_fence_status(spin->out_fence), 1);
> +		chk_err(err+ERR_FENCE_END, sync_fence_wait(spin->out_fence, 500), 0);
> +		chk_err(err+ERR_FENCE_STAT, sync_fence_status(spin->out_fence), 1);
>  		igt_spin_free(device, spin);
>  		put_ahnd(ahndN);
>  	}
> @@ -360,6 +377,11 @@ test_engine_hang(const intel_ctx_t *ctx,
>  	while (num_ctx)
>  		intel_ctx_destroy(device, local_ctx[--num_ctx]);
>  
> +	igt_assert_f(err[ERR_HANG_WAIT] == 0, "hanged spinner wait failed\n");
> +	igt_assert_f(err[ERR_HANG_STAT] == -EIO, "hanged spinner failed\n");
> +	igt_assert_f(err[ERR_FENCE_BUSY] == -ETIME, "background spinner not busy\n");
> +	igt_assert_f(err[ERR_FENCE_END] == 0, "background spinner not terminated\n");
> +	igt_assert_f(err[ERR_FENCE_STAT] == 1, "background fence not signalled\n");
>  	check_alive();
>  }
>  
> -- 
> 2.34.1
>