[PATCH i-g-t] tests/amd/amd_fuzzing: Fix compilation error on fedora

[PATCH i-g-t] tests/amd/amd_fuzzing: Fix compilation error on fedora

[Karthik B S]

Fix compilation error:

../tests/amdgpu/amd_fuzzing.c: In function ‘amd_kgd_multi_ioctl_field_fuzzing’:
../tests/amdgpu/amd_fuzzing.c:1103:13: error: ‘have_amdgpu_dev’ undeclared (first use in this function); did you mean ‘amdgpu_dev’?
 1103 |         if (have_amdgpu_dev) {
      |             ^~~~~~~~~~~~~~~
      |             amdgpu_dev

Cc: Vitaly Prosyak <vitaly.prosyak@amd.com>
Cc: Kamil Konieczny <kamil.konieczny@linux.intel.com>
Fixes: d512e54bbace ("tests/amdgpu/amd_fuzzing: Remove unused variable")
Signed-off-by: Karthik B S <karthik.b.s@intel.com>
---
 tests/amdgpu/amd_fuzzing.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/amdgpu/amd_fuzzing.c b/tests/amdgpu/amd_fuzzing.c
index 00717e593..9d7468ea4 100644
--- a/tests/amdgpu/amd_fuzzing.c
+++ b/tests/amdgpu/amd_fuzzing.c
@@ -1100,7 +1100,7 @@ amd_kgd_multi_ioctl_field_fuzzing(int fd, amdgpu_device_handle amdgpu_dev)
 	 * Note: USERQ_WAIT has no timeout field; it is not a blocking wait.
 	 * It only registers the dependency.  We therefore only test ret==0.
 	 */
-	if (have_amdgpu_dev) {
+	if (amdgpu_dev) {
 		uint32_t timeline_handle = 0;
 		uint32_t timeline_handles[1];
 		uint64_t timeline_points[1];
@@ -1142,7 +1142,7 @@ amd_kgd_multi_ioctl_field_fuzzing(int fd, amdgpu_device_handle amdgpu_dev)
 	 * Note: USERQ_WAIT has no timeout field and is not a blocking call; it
 	 * only registers a dependency.  We therefore only verify ret==0 here.
 	 */
-	if (have_amdgpu_dev) {
+	if (amdgpu_dev) {
 		uint32_t timeline_handles[2] = { 0, 0 };
 		uint64_t signal_points[2] = { 8, 8 };
 		uint64_t wait_points[2]   = { 8, 8 }; /* both already signaled */
-- 
2.43.0

Re: [PATCH i-g-t] tests/amd/amd_fuzzing: Fix compilation error on fedora

[Kamil Konieczny]

Hi Karthik,
On 2026-04-02 at 20:54:39 +0530, Karthik B S wrote:
> Fix compilation error:
> 
> ../tests/amdgpu/amd_fuzzing.c: In function ‘amd_kgd_multi_ioctl_field_fuzzing’:
> ../tests/amdgpu/amd_fuzzing.c:1103:13: error: ‘have_amdgpu_dev’ undeclared (first use in this function); did you mean ‘amdgpu_dev’?
>  1103 |         if (have_amdgpu_dev) {
>       |             ^~~~~~~~~~~~~~~
>       |             amdgpu_dev
> 
> Cc: Vitaly Prosyak <vitaly.prosyak@amd.com>
> Cc: Kamil Konieczny <kamil.konieczny@linux.intel.com>
> Fixes: d512e54bbace ("tests/amdgpu/amd_fuzzing: Remove unused variable")
> Signed-off-by: Karthik B S <karthik.b.s@intel.com>

LGTM
Reviewed-by: Kamil Konieczny <kamil.konieczny@linux.intel.com>

btw using
	bool b = (ptr != NULL);
is strange and should be catched by checkpatch.pl. In this case just
use 'ptr', no point in using helper var.

This also shows that code inside ifdef/endif should be very minimal
and all code should compile.

Vitalyi, what about only placing one var per one ifdef? For example:

#ifdef AMDGPU_USERQ_ENABLED
	bool amdgpu_userq_enabled = true;
#else
	bool amdgpu_userq_enabled = false;
#endif

and have you code always compile?

Regards,
Kamil 

> ---
>  tests/amdgpu/amd_fuzzing.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/amdgpu/amd_fuzzing.c b/tests/amdgpu/amd_fuzzing.c
> index 00717e593..9d7468ea4 100644
> --- a/tests/amdgpu/amd_fuzzing.c
> +++ b/tests/amdgpu/amd_fuzzing.c
> @@ -1100,7 +1100,7 @@ amd_kgd_multi_ioctl_field_fuzzing(int fd, amdgpu_device_handle amdgpu_dev)
>  	 * Note: USERQ_WAIT has no timeout field; it is not a blocking wait.
>  	 * It only registers the dependency.  We therefore only test ret==0.
>  	 */
> -	if (have_amdgpu_dev) {
> +	if (amdgpu_dev) {
>  		uint32_t timeline_handle = 0;
>  		uint32_t timeline_handles[1];
>  		uint64_t timeline_points[1];
> @@ -1142,7 +1142,7 @@ amd_kgd_multi_ioctl_field_fuzzing(int fd, amdgpu_device_handle amdgpu_dev)
>  	 * Note: USERQ_WAIT has no timeout field and is not a blocking call; it
>  	 * only registers a dependency.  We therefore only verify ret==0 here.
>  	 */
> -	if (have_amdgpu_dev) {
> +	if (amdgpu_dev) {
>  		uint32_t timeline_handles[2] = { 0, 0 };
>  		uint64_t signal_points[2] = { 8, 8 };
>  		uint64_t wait_points[2]   = { 8, 8 }; /* both already signaled */
> -- 
> 2.43.0
> 

Re: [PATCH i-g-t] tests/amd/amd_fuzzing: Fix compilation error on fedora

[vitaly prosyak]

[-- Attachment #1: Type: text/plain, Size: 3073 bytes --]

Hi Kamil,

Thanks for pointing this out, including the checkpatch.pl angle.

I also found quite a few checkpatch/style warnings in this area, and I will send a cleanup patch for those formatting/style issues soon.

Thanks, Vitaly

On 2026-04-02 11:36, Kamil Konieczny wrote:
> Hi Karthik,
> On 2026-04-02 at 20:54:39 +0530, Karthik B S wrote:
>> Fix compilation error:
>>
>> ../tests/amdgpu/amd_fuzzing.c: In function ‘amd_kgd_multi_ioctl_field_fuzzing’:
>> ../tests/amdgpu/amd_fuzzing.c:1103:13: error: ‘have_amdgpu_dev’ undeclared (first use in this function); did you mean ‘amdgpu_dev’?
>>  1103 |         if (have_amdgpu_dev) {
>>       |             ^~~~~~~~~~~~~~~
>>       |             amdgpu_dev
>>
>> Cc: Vitaly Prosyak <vitaly.prosyak@amd.com>
>> Cc: Kamil Konieczny <kamil.konieczny@linux.intel.com>
>> Fixes: d512e54bbace ("tests/amdgpu/amd_fuzzing: Remove unused variable")
>> Signed-off-by: Karthik B S <karthik.b.s@intel.com>
> LGTM
> Reviewed-by: Kamil Konieczny <kamil.konieczny@linux.intel.com>
>
> btw using
> 	bool b = (ptr != NULL);
> is strange and should be catched by checkpatch.pl. In this case just
> use 'ptr', no point in using helper var.'
You are right about |bool b = (ptr != NULL)| here — in this case the helper variable is unnecessary.
>
> This also shows that code inside ifdef/endif should be very minimal
> and all code should compile.
>
> Vitalyi, what about only placing one var per one ifdef? For example:
>
> #ifdef AMDGPU_USERQ_ENABLED
> 	bool amdgpu_userq_enabled = true;
> #else
> 	bool amdgpu_userq_enabled = false;
> #endif
>
> and have you code always compile?

For the broader #ifdef point, I agree in principle and will try to keep those blocks minimal, but in amd_fuzzing.c some USERQ paths still need to stay under #ifdef AMDGPU_USERQ_ENABLED because they use USERQ-specific types/ioctls.

>
> Regards,
> Kamil 
>
>> ---
>>  tests/amdgpu/amd_fuzzing.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/tests/amdgpu/amd_fuzzing.c b/tests/amdgpu/amd_fuzzing.c
>> index 00717e593..9d7468ea4 100644
>> --- a/tests/amdgpu/amd_fuzzing.c
>> +++ b/tests/amdgpu/amd_fuzzing.c
>> @@ -1100,7 +1100,7 @@ amd_kgd_multi_ioctl_field_fuzzing(int fd, amdgpu_device_handle amdgpu_dev)
>>  	 * Note: USERQ_WAIT has no timeout field; it is not a blocking wait.
>>  	 * It only registers the dependency.  We therefore only test ret==0.
>>  	 */
>> -	if (have_amdgpu_dev) {
>> +	if (amdgpu_dev) {
>>  		uint32_t timeline_handle = 0;
>>  		uint32_t timeline_handles[1];
>>  		uint64_t timeline_points[1];
>> @@ -1142,7 +1142,7 @@ amd_kgd_multi_ioctl_field_fuzzing(int fd, amdgpu_device_handle amdgpu_dev)
>>  	 * Note: USERQ_WAIT has no timeout field and is not a blocking call; it
>>  	 * only registers a dependency.  We therefore only verify ret==0 here.
>>  	 */
>> -	if (have_amdgpu_dev) {
>> +	if (amdgpu_dev) {
>>  		uint32_t timeline_handles[2] = { 0, 0 };
>>  		uint64_t signal_points[2] = { 8, 8 };
>>  		uint64_t wait_points[2]   = { 8, 8 }; /* both already signaled */
>> -- 
>> 2.43.0
>>

[-- Attachment #2: Type: text/html, Size: 4334 bytes --]