[PATCH 6/9] drm/xe/sriov: Move VF bootstrap and query_config to vf_guc_init

[Rodrigo Vivi]

On Fri, Nov 15, 2024 at 03:21:43PM +0100, Maarten Lankhorst wrote:
> Hey,
> 
> Den 2024-11-14 kl. 17:09, skrev Rodrigo Vivi:
> > On Thu, Nov 07, 2024 at 11:01:37AM +0100, Maarten Lankhorst wrote:
> >> We want to split up GUC init to an alloc and noalloc part to keep the
> >> init path the same for VF and !VF as much as possible.
> >>
> >> Everything in vf_guc_init should be done as early as possible, otherwise
> >> VRAM probing becomes impossible.
> >>
> >> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> >> Link: https://patchwork.freedesktop.org/patch/msgid/20241105121857.17389-6-maarten.lankhorst@linux.intel.com
> >> Signed-off-by: Maarten Lankhorst,,, <dev at lankhorst.se>
> >> ---
> >>  drivers/gpu/drm/xe/xe_gt_sriov_vf.c |  1 -
> >>  drivers/gpu/drm/xe/xe_guc.c         | 17 +++++++++--------
> >>  2 files changed, 9 insertions(+), 9 deletions(-)
> >>
> >> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_vf.c b/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
> >> index cca5d57328021..997438047a037 100644
> >> --- a/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
> >> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
> >> @@ -948,7 +948,6 @@ u32 xe_gt_sriov_vf_read32(struct xe_gt *gt, struct xe_reg reg)
> >>  	struct vf_runtime_reg *rr;
> >>  
> >>  	xe_gt_assert(gt, IS_SRIOV_VF(gt_to_xe(gt)));
> >> -	xe_gt_assert(gt, gt->sriov.vf.pf_version.major);
> >>  	xe_gt_assert(gt, !reg.vf);
> >>  
> >>  	if (reg.addr == GMD_ID.addr) {
> >> diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c
> >> index 7224593c9ce9b..5a050a5379911 100644
> >> --- a/drivers/gpu/drm/xe/xe_guc.c
> >> +++ b/drivers/gpu/drm/xe/xe_guc.c
> >> @@ -301,6 +301,7 @@ static int xe_guc_realloc_post_hwconfig(struct xe_guc *guc)
> >>  
> >>  static int vf_guc_init(struct xe_guc *guc)
> >>  {
> >> +	struct xe_gt *gt = guc_to_gt(guc);
> >>  	int err;
> >>  
> >>  	xe_guc_comm_init_early(guc);
> >> @@ -313,6 +314,14 @@ static int vf_guc_init(struct xe_guc *guc)
> >>  	if (err)
> >>  		return err;
> >>  
> >> +	err = xe_gt_sriov_vf_bootstrap(gt);
> >> +	if (err)
> >> +		return err;
> >> +
> >> +	err = xe_gt_sriov_vf_query_config(gt);
> >> +	if (err)
> >> +		return err;
> >> +
> >>  	return 0;
> >>  }
> >>  
> >> @@ -753,14 +762,6 @@ static int vf_guc_min_load_for_hwconfig(struct xe_guc *guc)
> > Cc: Michal Wajdeczko <michal.wajdeczko at intel.com>
> >
> > as I believe there was a good reason for these functions to be called this early.
> 
> This is in fact called slightly earlier than before. It's now called in xe_uc_init instead of xe_uc_init_hwconfig. :-)

Oh! Indeed! one step earlier actually... that name min_for_hwconfig got me confused, I'm sorry.

The only difference that I see now is that this functions are not avoided anymore if xe.force_execlist=1...

We should probably add some skips inside the function? I know I know we likely have many other bugs
if using force_execlists, but let's at least try to avoid introducing more cases...

> 
> Cheers,
> 
> ~Maarten
>