On 2019-04-02 10:29 a.m., Paul E. McKenney wrote:
> Having DEFINE_SRCU() or DEFINE_STATIC_SRCU() in a loadable module
> requires that the size of the reserved region be increased, which is
> not something we really want to be doing.  This commit therefore removes
> the DEFINE_STATIC_SRCU() from drivers/gpu/drm/amd/amdkfd/kfd_process.c in
> favor of defining kfd_processes_srcu as a simple srcu_struct, initializing
> it in amdgpu_amdkfd_init(), and cleaning it up in amdgpu_amdkfd_fini().
>
> Reported-by: kbuild test robot <lkp@intel.com>
> Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
> Tested-by: kbuild test robot <lkp@intel.com>
> Cc: Oded Gabbay <oded.gabbay@gmail.com>
> Cc: Alex Deucher <alexander.deucher@amd.com>
> Cc: "Christian König" <christian.koenig@amd.com
> Cc: "David (ChunMing) Zhou" <David1.Zhou@amd.com>
> Cc: David Airlie <airlied@linux.ie>
> Cc: Daniel Vetter <daniel@ffwll.ch>
> Cc: Tejun Heo <tj@kernel.org>
> Cc: <dri-devel@lists.freedesktop.org>
> Cc: <amd-gfx@lists.freedesktop.org>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 +++++
>   drivers/gpu/drm/amd/amdkfd/kfd_process.c   | 2 +-
>   2 files changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> index fe1d7368c1e6..eadb20dee867 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> @@ -28,6 +28,8 @@
>   #include <linux/module.h>
>   #include <linux/dma-buf.h>
>   
> +extern struct srcu_struct kfd_processes_srcu;
> +
>   static const unsigned int compute_vmid_bitmap = 0xFF00;
>   
>   /* Total memory size in system memory and all GPU VRAM. Used to
> @@ -40,6 +42,8 @@ int amdgpu_amdkfd_init(void)
>   	struct sysinfo si;
>   	int ret;
>   
> +	ret = init_srcu_struct(&kfd_processes_srcu);
> +	WARN_ON(ret);

kfd_processes_srcu only exists if kfd_process.c is compiled in. That 
depends on CONFIG_HSA_AMD. So this should at least move into #ifdef a 
few lines below.

However, it would be cleaner to move this initialization into kfd_init 
in kfd_module.c, or better yet, into kfd_process_create_wq in 
kfd_process.c. Then kfd_process_create_wq should be renamed to something 
more generic, such as kfd_process_init.


>   	si_meminfo(&si);
>   	amdgpu_amdkfd_total_mem_size = si.totalram - si.totalhigh;
>   	amdgpu_amdkfd_total_mem_size *= si.mem_unit;
> @@ -57,6 +61,7 @@ int amdgpu_amdkfd_init(void)
>   void amdgpu_amdkfd_fini(void)
>   {
>   	kgd2kfd_exit();
> +	cleanup_srcu_struct(&kfd_processes_srcu);

Similarly, this would be cleaner in kfd_exit in kfd_module.c or 
kfd_process_destroy_wq in kfd_process.c, with that function similarly 
renamed to kfd_process_fini.

I'm attaching a revised patch. It's only compile tested.

Regards,
   Felix


>   }
>   
>   void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
> index 4bdae78bab8e..98b694068b8a 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
> @@ -47,7 +47,7 @@ struct mm_struct;
>   DEFINE_HASHTABLE(kfd_processes_table, KFD_PROCESS_TABLE_SIZE);
>   static DEFINE_MUTEX(kfd_processes_mutex);
>   
> -DEFINE_SRCU(kfd_processes_srcu);
> +struct srcu_struct kfd_processes_srcu;
>   
>   /* For process termination handling */
>   static struct workqueue_struct *kfd_process_wq;