[igt-dev] [RFC v2 01/43] test/i915: gem_bad_reloc: use the gem_engine_topology library

[Tvrtko Ursulin]

On 21/06/2019 11:02, Chris Wilson wrote:
> Quoting Ramalingam C (2019-06-21 11:03:03)
>> Replace the legacy for_each_engine* defines with the ones
>> implemented in the gem_engine_topology library.
>>
>> Signed-off-by: Ramalingam C <ramalingam.c at intel.com>
>> ---
>>   tests/i915/gem_bad_reloc.c | 14 +++++++-------
>>   1 file changed, 7 insertions(+), 7 deletions(-)
>>
>> diff --git a/tests/i915/gem_bad_reloc.c b/tests/i915/gem_bad_reloc.c
>> index 7624cd8e0949..26a932a3a933 100644
>> --- a/tests/i915/gem_bad_reloc.c
>> +++ b/tests/i915/gem_bad_reloc.c
>> @@ -50,7 +50,8 @@ IGT_TEST_DESCRIPTION("Simulates SNA behaviour using negative self-relocations"
>>    * than the total size of the GTT), the GPU will hang.
>>    * See https://bugs.freedesktop.org/show_bug.cgi?id=78533
>>    */
>> -static void negative_reloc(int fd, unsigned engine, unsigned flags)
>> +static void negative_reloc(int fd, const struct intel_execution_engine2 *e,
>> +                          unsigned flags)
>>   {
>>          struct drm_i915_gem_execbuffer2 execbuf;
>>          struct drm_i915_gem_exec_object2 obj;
>> @@ -60,7 +61,6 @@ static void negative_reloc(int fd, unsigned engine, unsigned flags)
>>          uint64_t *offsets;
>>          int i;
>>   
>> -       gem_require_ring(fd, engine);
>>          igt_require(intel_gen(intel_get_drm_devid(fd)) >= 7);
>>   
>>          memset(&obj, 0, sizeof(obj));
>> @@ -70,7 +70,7 @@ static void negative_reloc(int fd, unsigned engine, unsigned flags)
>>          memset(&execbuf, 0, sizeof(execbuf));
>>          execbuf.buffers_ptr = (uintptr_t)&obj;
>>          execbuf.buffer_count = 1;
>> -       execbuf.flags = engine | (flags & USE_LUT);
>> +       execbuf.flags = e->flags | (flags & USE_LUT);
>>          igt_require(__gem_execbuf(fd, &execbuf) == 0);
>>   
>>          igt_info("Found offset %lld for 4k batch\n", (long long)obj.offset);
>> @@ -185,7 +185,7 @@ static void negative_reloc_blt(int fd)
>>   
>>   igt_main
>>   {
>> -       const struct intel_execution_engine *e;
>> +       const struct intel_execution_engine2 *e;
>>          int fd = -1;
>>   
>>          igt_fixture {
>> @@ -193,12 +193,12 @@ igt_main
>>                  igt_require_gem(fd);
>>          }
>>   
>> -       for (e = intel_execution_engines; e->name; e++) {
>> +       __for_each_physical_engine(fd, e) {
> 
> uABI testing not physical engines.

Should we add something like for_each_legacy_abi_engine to use in these 
tests so it is more self-documenting? (And get rid of raw access to 
intel_execution_engines.)

Regards,

Tvrtko