From: Rodrigo Vivi <rodrigo.vivi@intel.com>
To: "Ghimiray, Himal Prasad" <himal.prasad.ghimiray@intel.com>
Cc: Tejas Upadhyay <tejas.upadhyay@intel.com>,
<intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH] drm/xe: sysfs_ops needs to be defined on parent directory
Date: Tue, 25 Mar 2025 15:14:11 -0400 [thread overview]
Message-ID: <Z-MAg1F9_2seEgYy@intel.com> (raw)
In-Reply-To: <e91f93d2-a4e2-4304-b026-eea2f1fae13d@intel.com>
On Tue, Mar 25, 2025 at 03:18:40PM +0530, Ghimiray, Himal Prasad wrote:
>
>
> On 19-03-2025 17:43, Tejas Upadhyay wrote:
> > Currently, xe_hw_engine_sysfs_kobj_type is defining sysfs_ops
> > on wrong directory. Sysfs_ops needs to be defined on immediate
> > parent directory to be able to called on each attribute set/get.
> >
> > Fixes: 3f0e14651ab0 ("drm/xe: Runtime PM wake on every sysfs call")
> > Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
> > ---
> > drivers/gpu/drm/xe/xe_hw_engine_class_sysfs.c | 67 +++++++++----------
> > 1 file changed, 33 insertions(+), 34 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_hw_engine_class_sysfs.c b/drivers/gpu/drm/xe/xe_hw_engine_class_sysfs.c
> > index b53e8d2accdb..25592f178482 100644
> > --- a/drivers/gpu/drm/xe/xe_hw_engine_class_sysfs.c
> > +++ b/drivers/gpu/drm/xe/xe_hw_engine_class_sysfs.c
> > @@ -492,39 +492,6 @@ static const struct attribute * const files[] = {
> > NULL
> > };
> > -static void kobj_xe_hw_engine_class_fini(void *arg)
> > -{
> > - struct kobject *kobj = arg;
> > -
> > - sysfs_remove_files(kobj, files);
> > - kobject_put(kobj);
> > -}
> > -
> > -static struct kobj_eclass *
> > -kobj_xe_hw_engine_class(struct xe_device *xe, struct kobject *parent, const char *name)
> > -{
> > - struct kobj_eclass *keclass;
> > - int err = 0;
> > -
> > - keclass = kzalloc(sizeof(*keclass), GFP_KERNEL);
> > - if (!keclass)
> > - return NULL;
> > -
> > - kobject_init(&keclass->base, &kobj_xe_hw_engine_type);
> > - if (kobject_add(&keclass->base, parent, "%s", name)) {
> > - kobject_put(&keclass->base);
> > - return NULL;
> > - }
> > - keclass->xe = xe;
> > -
> > - err = devm_add_action_or_reset(xe->drm.dev, kobj_xe_hw_engine_class_fini,
> > - &keclass->base);
> > - if (err)
> > - return NULL;
> > -
> > - return keclass;
> > -}
> > -
> > static void hw_engine_class_defaults_fini(void *arg)
> > {
> > struct kobject *kobj = arg;
> > @@ -611,6 +578,38 @@ static const struct kobj_type xe_hw_engine_sysfs_kobj_type = {
> > .sysfs_ops = &xe_hw_engine_class_sysfs_ops,
> > };
> > +static void kobj_xe_hw_engine_class_fini(void *arg)
> > +{
> > + struct kobject *kobj = arg;
> > +
> > + sysfs_remove_files(kobj, files);
> > + kobject_put(kobj);
> > +}
> > +
> > +static struct kobj_eclass *
> > +kobj_xe_hw_engine_class(struct xe_device *xe, struct kobject *parent, const char *name)
> > +{
> > + struct kobj_eclass *keclass;
> > + int err = 0;
> > +
> > + keclass = kzalloc(sizeof(*keclass), GFP_KERNEL);
> > + if (!keclass)
> > + return NULL;
> > +
> > + kobject_init(&keclass->base, &xe_hw_engine_sysfs_kobj_type);
> > + if (kobject_add(&keclass->base, parent, "%s", name)) {
> > + kobject_put(&keclass->base);
> > + return NULL;
> > + }
> > + keclass->xe = xe;
> > +
> > + err = devm_add_action_or_reset(xe->drm.dev, kobj_xe_hw_engine_class_fini,
> > + &keclass->base);
> > + if (err)
> > + return NULL;
> > +
> > + return keclass;
> > +}
> > static void hw_engine_class_sysfs_fini(void *arg)
> > {
> > struct kobject *kobj = arg;
> > @@ -640,7 +639,7 @@ int xe_hw_engine_class_sysfs_init(struct xe_gt *gt)
> > if (!kobj)
> > return -ENOMEM;
> > - kobject_init(kobj, &xe_hw_engine_sysfs_kobj_type);
> > + kobject_init(kobj, &kobj_xe_hw_engine_type);
>
>
> The patch makes sense if we need to use xe_pm_runtime_get and
> xe_pm_runtime_put to show and store the attributes. Are there any hardware
> read/writes associated with these attribute store/show operations? If not,
> why do we need pm_runtime just for software states?
>
> +Rodrigo, for his views here.
Better safe than sorry. The goal is to protect every entry point on the
upper (outer) levels that might interact with or trigger some memory operation.
>
> > err = kobject_add(kobj, gt->sysfs, "engines");
> > if (err)
>
next prev parent reply other threads:[~2025-03-25 19:14 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-19 12:13 [PATCH] drm/xe: sysfs_ops needs to be defined on parent directory Tejas Upadhyay
2025-03-19 12:22 ` ✓ CI.Patch_applied: success for " Patchwork
2025-03-19 12:23 ` ✓ CI.checkpatch: " Patchwork
2025-03-19 12:24 ` ✓ CI.KUnit: " Patchwork
2025-03-19 12:40 ` ✓ CI.Build: " Patchwork
2025-03-19 12:43 ` ✓ CI.Hooks: " Patchwork
2025-03-19 12:44 ` ✓ CI.checksparse: " Patchwork
2025-03-19 13:03 ` ✓ Xe.CI.BAT: " Patchwork
2025-03-19 14:05 ` ✓ Xe.CI.Full: " Patchwork
2025-03-25 9:48 ` [PATCH] " Ghimiray, Himal Prasad
2025-03-25 19:14 ` Rodrigo Vivi [this message]
2025-03-25 19:12 ` Rodrigo Vivi
2025-03-26 4:14 ` Upadhyay, Tejas
2025-03-26 14:43 ` Rodrigo Vivi
2025-03-26 14:50 ` Upadhyay, Tejas
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Z-MAg1F9_2seEgYy@intel.com \
--to=rodrigo.vivi@intel.com \
--cc=himal.prasad.ghimiray@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=tejas.upadhyay@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox