Re: [PATCH v6 1/4] drm: Use XArray instead of IDR for minors

[James Zhu <jamesz@xxxxxxx>]

On 2023-08-28 17:08, Michał Winiarski wrote:
> On Fri, Aug 25, 2023 at 12:59:26PM -0400, James Zhu wrote:
> > On 2023-07-24 17:14, Michał Winiarski wrote:
> > > IDR is deprecated, and since XArray manages its own state with internal
> > > locking, it simplifies the locking on DRM side.
> > > Additionally, don't use the IRQ-safe variant, since operating on drm
> > > minor is not done in IRQ context.
> > >
> > > Signed-off-by: Michał Winiarski<michal.winiarski@xxxxxxxxx>
> > > Suggested-by: Matthew Wilcox<willy@xxxxxxxxxxxxx>
> > > ---
> > >    drivers/gpu/drm/drm_drv.c | 63 ++++++++++++++++-----------------------
> > >    1 file changed, 25 insertions(+), 38 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
> > > index 3eda026ffac6..3faecb01186f 100644
> > > --- a/drivers/gpu/drm/drm_drv.c
> > > +++ b/drivers/gpu/drm/drm_drv.c
> > > @@ -34,6 +34,7 @@
> > >    #include <linux/pseudo_fs.h>
> > >    #include <linux/slab.h>
> > >    #include <linux/srcu.h>
> > > +#include <linux/xarray.h>
> > >    #include <drm/drm_accel.h>
> > >    #include <drm/drm_cache.h>
> > > @@ -54,8 +55,7 @@ MODULE_AUTHOR("Gareth Hughes, Leif Delgass, José Fonseca, Jon Smirl");
> > >    MODULE_DESCRIPTION("DRM shared core routines");
> > >    MODULE_LICENSE("GPL and additional rights");
> > > -static DEFINE_SPINLOCK(drm_minor_lock);
> > > -static struct idr drm_minors_idr;
> > > +static DEFINE_XARRAY_ALLOC(drm_minors_xa);
> > >    /*
> > >     * If the drm core fails to init for whatever reason,
> > > @@ -101,26 +101,23 @@ static struct drm_minor **drm_minor_get_slot(struct drm_device *dev,
> > >    static void drm_minor_alloc_release(struct drm_device *dev, void *data)
> > >    {
> > >    	struct drm_minor *minor = data;
> > > -	unsigned long flags;
> > >    	WARN_ON(dev != minor->dev);
> > >    	put_device(minor->kdev);
> > > -	if (minor->type == DRM_MINOR_ACCEL) {
> > > +	if (minor->type == DRM_MINOR_ACCEL)
> > >    		accel_minor_remove(minor->index);
> > > -	} else {
> > > -		spin_lock_irqsave(&drm_minor_lock, flags);
> > > -		idr_remove(&drm_minors_idr, minor->index);
> > > -		spin_unlock_irqrestore(&drm_minor_lock, flags);
> > > -	}
> > > +	else
> > > +		xa_erase(&drm_minors_xa, minor->index);
> > >    }
> > > +#define DRM_MINOR_LIMIT(t) ({ typeof(t) _t = (t); XA_LIMIT(64 * _t, 64 * _t + 63); })
> > > +
> > >    static int drm_minor_alloc(struct drm_device *dev, enum drm_minor_type type)
> > >    {
> > >    	struct drm_minor *minor;
> > > -	unsigned long flags;
> > > -	int r;
> > > +	int index, r;
> > >    	minor = drmm_kzalloc(dev, sizeof(*minor), GFP_KERNEL);
> > >    	if (!minor)
> > > @@ -129,24 +126,17 @@ static int drm_minor_alloc(struct drm_device *dev, enum drm_minor_type type)
> > >    	minor->type = type;
> > >    	minor->dev = dev;
> > > -	idr_preload(GFP_KERNEL);
> > >    	if (type == DRM_MINOR_ACCEL) {
> > >    		r = accel_minor_alloc();
> > > +		index = r;
> > >    	} else {
> > > -		spin_lock_irqsave(&drm_minor_lock, flags);
> > > -		r = idr_alloc(&drm_minors_idr,
> > > -			NULL,
> > > -			64 * type,
> > > -			64 * (type + 1),
> > > -			GFP_NOWAIT);
> > > -		spin_unlock_irqrestore(&drm_minor_lock, flags);
> > > +		r = xa_alloc(&drm_minors_xa, &index, NULL, DRM_MINOR_LIMIT(type), GFP_KERNEL);
> > >    	}
> > > -	idr_preload_end();
> > >    	if (r < 0)
> > >    		return r;
> > > -	minor->index = r;
> > > +	minor->index = index;
> > >    	r = drmm_add_action_or_reset(dev, drm_minor_alloc_release, minor);
> > >    	if (r)
> > > @@ -163,7 +153,7 @@ static int drm_minor_alloc(struct drm_device *dev, enum drm_minor_type type)
> > >    static int drm_minor_register(struct drm_device *dev, enum drm_minor_type type)
> > >    {
> > >    	struct drm_minor *minor;
> > > -	unsigned long flags;
> > > +	void *entry;
> > >    	int ret;
> > >    	DRM_DEBUG("\n");
> > > @@ -190,9 +180,12 @@ static int drm_minor_register(struct drm_device *dev, enum drm_minor_type type)
> > >    	if (minor->type == DRM_MINOR_ACCEL) {
> > >    		accel_minor_replace(minor, minor->index);
> > >    	} else {
> > > -		spin_lock_irqsave(&drm_minor_lock, flags);
> > > -		idr_replace(&drm_minors_idr, minor, minor->index);
> > > -		spin_unlock_irqrestore(&drm_minor_lock, flags);
> > > +		entry = xa_store(&drm_minors_xa, minor->index, minor, GFP_KERNEL);
> > > +		if (xa_is_err(entry)) {
> > > +			ret = xa_err(entry);
> > > +			goto err_debugfs;
> > > +		}
> > > +		WARN_ON(entry);
> > [JZ] would WARN_ON(entry != minor)be better?
> We expect NULL here.
> The same one that was previously stored here ⌄⌄⌄
> r = xa_alloc(&drm_minors_xa, &minor->index, NULL, DRM_EXTENDED_MINOR_LIMIT, GFP_KERNEL);
> in drm_minor_alloc.
[JZ] My mistake.
> > >    	}
> > >    	DRM_DEBUG("new minor registered %d\n", minor->index);
> > > @@ -206,20 +199,16 @@ static int drm_minor_register(struct drm_device *dev, enum drm_minor_type type)
> > >    static void drm_minor_unregister(struct drm_device *dev, enum drm_minor_type type)
> > >    {
> > >    	struct drm_minor *minor;
> > > -	unsigned long flags;
> > >    	minor = *drm_minor_get_slot(dev, type);
> > >    	if (!minor || !device_is_registered(minor->kdev))
> > >    		return;
> > >    	/* replace @minor with NULL so lookups will fail from now on */
> > > -	if (minor->type == DRM_MINOR_ACCEL) {
> > > +	if (minor->type == DRM_MINOR_ACCEL)
> > >    		accel_minor_replace(NULL, minor->index);
> > > -	} else {
> > > -		spin_lock_irqsave(&drm_minor_lock, flags);
> > > -		idr_replace(&drm_minors_idr, NULL, minor->index);
> > > -		spin_unlock_irqrestore(&drm_minor_lock, flags);
> > > -	}
> > > +	else
> > > +		xa_store(&drm_minors_xa, minor->index, NULL, GFP_KERNEL);
> > >    	device_del(minor->kdev);
> > >    	dev_set_drvdata(minor->kdev, NULL); /* safety belt */
> > > @@ -238,13 +227,12 @@ static void drm_minor_unregister(struct drm_device *dev, enum drm_minor_type typ
> > >    struct drm_minor *drm_minor_acquire(unsigned int minor_id)
> > >    {
> > >    	struct drm_minor *minor;
> > > -	unsigned long flags;
> > > -	spin_lock_irqsave(&drm_minor_lock, flags);
> > > -	minor = idr_find(&drm_minors_idr, minor_id);
> > > +	xa_lock(&drm_minors_xa);
> > > +	minor = xa_load(&drm_minors_xa, minor_id);
> > >    	if (minor)
> > >    		drm_dev_get(minor->dev);
> > [JZ] why minor->dev need ca_lock here?
> We're relying on the ordering for acquire/release (previously
> guaranteed by the drm_minor_lock, now by internal XArray locking).
> xa_load doesn't take xa_lock internally:
> https://docs.kernel.org/core-api/xarray.html#locking
[JZ] I means that drm_dev_get(minor->dev); can move out of xa_lock.
> > > -	spin_unlock_irqrestore(&drm_minor_lock, flags);
> > > +	xa_unlock(&drm_minors_xa);
> > >    	if (!minor) {
> > >    		return ERR_PTR(-ENODEV);
> > > @@ -1067,7 +1055,7 @@ static void drm_core_exit(void)
> > >    	unregister_chrdev(DRM_MAJOR, "drm");
> > >    	debugfs_remove(drm_debugfs_root);
> > >    	drm_sysfs_destroy();
> > > -	idr_destroy(&drm_minors_idr);
> > [JZ] Should we call xa_destroy instead here?
> We could, if we expect the xarray to potentially not be empty.
>  From what I understand - all minors should be released at this point.
[JZ] In practice,  adding destroy here will be better.
> Thanks,
> -Michał
>
> > > +	WARN_ON(!xa_empty(&drm_minors_xa));
> > >    	drm_connector_ida_destroy();
> > >    }
> > > @@ -1076,7 +1064,6 @@ static int __init drm_core_init(void)
> > >    	int ret;
> > >    	drm_connector_ida_init();
> > > -	idr_init(&drm_minors_idr);
> > >    	drm_memcpy_init_early();
> > >    	ret = drm_sysfs_init();