[PATCH 1/5] drm/xe: Move part of xe_file cleanup to a helper
Matthew Brost
matthew.brost at intel.com
Mon Jul 8 22:18:14 UTC 2024
On Mon, Jul 08, 2024 at 03:12:33PM -0700, Umesh Nerlige Ramappa wrote:
> On Mon, Jul 08, 2024 at 09:21:01PM +0000, Matthew Brost wrote:
> > On Mon, Jul 08, 2024 at 01:20:59PM -0700, Umesh Nerlige Ramappa wrote:
> > > In order to make xe_file ref counted, move destruction of xe_file
> > > members to a helper.
> > >
> >
> > I don't see where xe_file_close is called in this patch but appears to
> > called in .postclose later in the series. I'd move that change into this
> > patch.
>
> I don't follow this point. The series is not calling
> xe_file_close/.postclose, so not sure what change I need to move.
>
> Thanks,
> Umesh
Yea my bad, disregard this comment. I though xe_file_close was a new
function but misread the diff.
Comment below about moving xe_vm_close_and_put to xe_file_close is valid
though.
Matt
>
> >
> > Also I think the VM should ref count the xref too to be uniform and it
> > appears this series is not doing that either. With a VM ref counting
> > xref, the 'xe_vm_close_and_put' is going to need to be in xe_file_close
> > too.
> >
> > Matt
> >
> > > Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
> > > ---
> > > drivers/gpu/drm/xe/xe_device.c | 36 +++++++++++++++++++++-------------
> > > 1 file changed, 22 insertions(+), 14 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
> > > index cfda7cb5df2c..babb697652d5 100644
> > > --- a/drivers/gpu/drm/xe/xe_device.c
> > > +++ b/drivers/gpu/drm/xe/xe_device.c
> > > @@ -90,24 +90,12 @@ static int xe_file_open(struct drm_device *dev, struct drm_file *file)
> > > return 0;
> > > }
> > >
> > > -static void xe_file_close(struct drm_device *dev, struct drm_file *file)
> > > +static void xe_file_destroy(struct xe_file *xef)
> > > {
> > > - struct xe_device *xe = to_xe_device(dev);
> > > - struct xe_file *xef = file->driver_priv;
> > > + struct xe_device *xe = xef->xe;
> > > struct xe_vm *vm;
> > > - struct xe_exec_queue *q;
> > > unsigned long idx;
> > >
> > > - /*
> > > - * No need for exec_queue.lock here as there is no contention for it
> > > - * when FD is closing as IOCTLs presumably can't be modifying the
> > > - * xarray. Taking exec_queue.lock here causes undue dependency on
> > > - * vm->lock taken during xe_exec_queue_kill().
> > > - */
> > > - xa_for_each(&xef->exec_queue.xa, idx, q) {
> > > - xe_exec_queue_kill(q);
> > > - xe_exec_queue_put(q);
> > > - }
> > > xa_destroy(&xef->exec_queue.xa);
> > > mutex_destroy(&xef->exec_queue.lock);
> > > mutex_lock(&xef->vm.lock);
> > > @@ -125,6 +113,26 @@ static void xe_file_close(struct drm_device *dev, struct drm_file *file)
> > > kfree(xef);
> > > }
> > >
> > > +static void xe_file_close(struct drm_device *dev, struct drm_file *file)
> > > +{
> > > + struct xe_file *xef = file->driver_priv;
> > > + struct xe_exec_queue *q;
> > > + unsigned long idx;
> > > +
> > > + /*
> > > + * No need for exec_queue.lock here as there is no contention for it
> > > + * when FD is closing as IOCTLs presumably can't be modifying the
> > > + * xarray. Taking exec_queue.lock here causes undue dependency on
> > > + * vm->lock taken during xe_exec_queue_kill().
> > > + */
> > > + xa_for_each(&xef->exec_queue.xa, idx, q) {
> > > + xe_exec_queue_kill(q);
> > > + xe_exec_queue_put(q);
> > > + }
> > > +
> > > + xe_file_destroy(xef);
> > > +}
> > > +
> > > static const struct drm_ioctl_desc xe_ioctls[] = {
> > > DRM_IOCTL_DEF_DRV(XE_DEVICE_QUERY, xe_query_ioctl, DRM_RENDER_ALLOW),
> > > DRM_IOCTL_DEF_DRV(XE_GEM_CREATE, xe_gem_create_ioctl, DRM_RENDER_ALLOW),
> > > --
> > > 2.38.1
> > >
More information about the Intel-xe
mailing list