* [PATCH v2] usbip: vhci_hcd: let the driver core manage the sysfs attributes
@ 2026-08-16 1:50 Deepanshu Kartikey
2026-08-16 4:55 ` Greg KH
0 siblings, 1 reply; 4+ messages in thread
From: Deepanshu Kartikey @ 2026-08-16 1:50 UTC (permalink / raw)
To: valentina.manea.m, shuah, i, gregkh
Cc: yuyang.du, linux-usb, linux-kernel, Deepanshu Kartikey,
syzbot+8753715f05759f1a10de
The vhci attribute group is created in vhci_start() and removed in
vhci_stop(), guarded by usb_hcd_is_primary_hcd(). Both run from
usb_add_hcd()/usb_remove_hcd(), which are called once per hcd, so the
attach attribute is live while only one of the two hcds exists: it is
created during the first usb_add_hcd() before vhci_hcd_ss is set, and it
survives the first usb_put_hcd() on removal. A concurrent write to attach
can therefore reach a NULL or freed vhci_hcd_ss.
Register the group as dev_groups on the platform driver instead, so the
driver core creates the files before probe and removes them after remove
returns, and drop the sysfs_create_group()/sysfs_remove_group() calls from
the driver. The attributes have always been created only on vhci_hcd.0;
an is_visible() callback keeps them there.
vhci_init_attr_group() is moved into vhci_hcd_init() ahead of
platform_driver_register() so the group is populated before the core
reads it. The attribute array is still built at runtime because the
number of status attributes comes from CONFIG_USBIP_VHCI_NR_HCS and the
preprocessor cannot emit a variable number of __ATTR() declarations. If
statically declaring all 32 and hiding the unused ones with is_visible()
is preferred, I can respin that way.
Reported-by: syzbot+8753715f05759f1a10de@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=8753715f05759f1a10de
Fixes: 1c9de5bf4286 ("usbip: vhci-hcd: Add USB3 SuperSpeed support")
Link: https://lore.kernel.org/all/20260815143241.19047-1-kartikey406@gmail.com/T/ [v1]
Tested-by: syzbot+8753715f05759f1a10de@syzkaller.appspotmail.com
Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
---
v2:
- use dev_groups instead of moving sysfs_create_group() into probe (Greg)
- add is_visible() to keep the attributes on vhci_hcd.0 only
- move vhci_init_attr_group() into vhci_hcd_init()
---
drivers/usb/usbip/vhci.h | 1 +
drivers/usb/usbip/vhci_hcd.c | 37 ++++++++++------------------------
drivers/usb/usbip/vhci_sysfs.c | 19 +++++++++++++++++
3 files changed, 31 insertions(+), 26 deletions(-)
diff --git a/drivers/usb/usbip/vhci.h b/drivers/usb/usbip/vhci.h
index 5659dce1526e..844500e3b0ea 100644
--- a/drivers/usb/usbip/vhci.h
+++ b/drivers/usb/usbip/vhci.h
@@ -121,6 +121,7 @@ struct vhci_hcd {
extern int vhci_num_controllers;
extern struct vhci *vhcis;
extern struct attribute_group vhci_attr_group;
+extern const struct attribute_group *vhci_groups[];
/* vhci_hcd.c */
void rh_port_connect(struct vhci_device *vdev, enum usb_device_speed speed);
diff --git a/drivers/usb/usbip/vhci_hcd.c b/drivers/usb/usbip/vhci_hcd.c
index 39e8faf4c18c..4bace5b8b879 100644
--- a/drivers/usb/usbip/vhci_hcd.c
+++ b/drivers/usb/usbip/vhci_hcd.c
@@ -1199,7 +1199,6 @@ static int vhci_start(struct usb_hcd *hcd)
{
struct vhci_hcd *vhci_hcd = hcd_to_vhci_hcd(hcd);
int id, rhport;
- int err;
usbip_dbg_vhci_hc("enter vhci_start\n");
@@ -1230,40 +1229,17 @@ static int vhci_start(struct usb_hcd *hcd)
return -EINVAL;
}
- /* vhci_hcd is now ready to be controlled through sysfs */
- if (id == 0 && usb_hcd_is_primary_hcd(hcd)) {
- err = vhci_init_attr_group();
- if (err) {
- dev_err(hcd_dev(hcd), "init attr group failed, err = %d\n", err);
- return err;
- }
- err = sysfs_create_group(&hcd_dev(hcd)->kobj, &vhci_attr_group);
- if (err) {
- dev_err(hcd_dev(hcd), "create sysfs files failed, err = %d\n", err);
- vhci_finish_attr_group();
- return err;
- }
- dev_info(hcd_dev(hcd), "created sysfs %s\n", hcd_name(hcd));
- }
-
return 0;
}
static void vhci_stop(struct usb_hcd *hcd)
{
struct vhci_hcd *vhci_hcd = hcd_to_vhci_hcd(hcd);
- int id, rhport;
+ int rhport;
usbip_dbg_vhci_hc("stop VHCI controller\n");
- /* 1. remove the userland interface of vhci_hcd */
- id = hcd_name_to_id(hcd_name(hcd));
- if (id == 0 && usb_hcd_is_primary_hcd(hcd)) {
- sysfs_remove_group(&hcd_dev(hcd)->kobj, &vhci_attr_group);
- vhci_finish_attr_group();
- }
-
- /* 2. shutdown all the ports of vhci_hcd */
+ /* shutdown all the ports of vhci_hcd */
for (rhport = 0; rhport < VHCI_HC_PORTS; rhport++) {
struct vhci_device *vdev = &vhci_hcd->vdev[rhport];
@@ -1513,6 +1489,7 @@ static struct platform_driver vhci_driver = {
.resume = vhci_hcd_resume,
.driver = {
.name = driver_name,
+ .dev_groups = vhci_groups,
},
};
@@ -1541,6 +1518,10 @@ static int __init vhci_hcd_init(void)
if (vhcis == NULL)
return -ENOMEM;
+ ret = vhci_init_attr_group();
+ if (ret)
+ goto err_init_attr_group;
+
ret = platform_driver_register(&vhci_driver);
if (ret)
goto err_driver_register;
@@ -1565,9 +1546,12 @@ static int __init vhci_hcd_init(void)
return 0;
+
err_add_hcd:
platform_driver_unregister(&vhci_driver);
err_driver_register:
+ vhci_finish_attr_group();
+err_init_attr_group:
kfree(vhcis);
return ret;
}
@@ -1576,6 +1560,7 @@ static void __exit vhci_hcd_exit(void)
{
del_platform_devices();
platform_driver_unregister(&vhci_driver);
+ vhci_finish_attr_group();
kfree(vhcis);
}
diff --git a/drivers/usb/usbip/vhci_sysfs.c b/drivers/usb/usbip/vhci_sysfs.c
index a7ede6fb3da9..8c0bff2ddabb 100644
--- a/drivers/usb/usbip/vhci_sysfs.c
+++ b/drivers/usb/usbip/vhci_sysfs.c
@@ -497,8 +497,27 @@ static void finish_status_attrs(void)
kfree(status_attrs);
}
+static umode_t vhci_attr_is_visible(struct kobject *kobj,
+ struct attribute *attr, int n)
+{
+ struct platform_device *pdev = to_platform_device(kobj_to_dev(kobj));
+ /*
+ * The attributes control every controller and have always lived on
+ * vhci_hcd.0 only. Keep them there now that the driver core creates
+ * the group for each device.
+ */
+ return pdev->id == 0 ? attr->mode : 0;
+}
+
+
struct attribute_group vhci_attr_group = {
.attrs = NULL,
+ .is_visible = vhci_attr_is_visible,
+};
+
+const struct attribute_group *vhci_groups[] = {
+ &vhci_attr_group,
+ NULL,
};
int vhci_init_attr_group(void)
--
2.43.0
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH v2] usbip: vhci_hcd: let the driver core manage the sysfs attributes
2026-08-16 1:50 [PATCH v2] usbip: vhci_hcd: let the driver core manage the sysfs attributes Deepanshu Kartikey
@ 2026-08-16 4:55 ` Greg KH
2026-08-16 6:11 ` Deepanshu Kartikey
0 siblings, 1 reply; 4+ messages in thread
From: Greg KH @ 2026-08-16 4:55 UTC (permalink / raw)
To: Deepanshu Kartikey
Cc: valentina.manea.m, shuah, i, yuyang.du, linux-usb, linux-kernel,
syzbot+8753715f05759f1a10de
On Sun, Aug 16, 2026 at 07:20:51AM +0530, Deepanshu Kartikey wrote:
> The vhci attribute group is created in vhci_start() and removed in
> vhci_stop(), guarded by usb_hcd_is_primary_hcd(). Both run from
> usb_add_hcd()/usb_remove_hcd(), which are called once per hcd, so the
> attach attribute is live while only one of the two hcds exists: it is
> created during the first usb_add_hcd() before vhci_hcd_ss is set, and it
> survives the first usb_put_hcd() on removal. A concurrent write to attach
> can therefore reach a NULL or freed vhci_hcd_ss.
>
> Register the group as dev_groups on the platform driver instead, so the
> driver core creates the files before probe and removes them after remove
> returns, and drop the sysfs_create_group()/sysfs_remove_group() calls from
> the driver. The attributes have always been created only on vhci_hcd.0;
> an is_visible() callback keeps them there.
>
> vhci_init_attr_group() is moved into vhci_hcd_init() ahead of
> platform_driver_register() so the group is populated before the core
> reads it. The attribute array is still built at runtime because the
> number of status attributes comes from CONFIG_USBIP_VHCI_NR_HCS and the
> preprocessor cannot emit a variable number of __ATTR() declarations. If
> statically declaring all 32 and hiding the unused ones with is_visible()
> is preferred, I can respin that way.
>
> Reported-by: syzbot+8753715f05759f1a10de@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=8753715f05759f1a10de
> Fixes: 1c9de5bf4286 ("usbip: vhci-hcd: Add USB3 SuperSpeed support")
> Link: https://lore.kernel.org/all/20260815143241.19047-1-kartikey406@gmail.com/T/ [v1]
> Tested-by: syzbot+8753715f05759f1a10de@syzkaller.appspotmail.com
> Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
> ---
> v2:
> - use dev_groups instead of moving sysfs_create_group() into probe (Greg)
> - add is_visible() to keep the attributes on vhci_hcd.0 only
> - move vhci_init_attr_group() into vhci_hcd_init()
> ---
> drivers/usb/usbip/vhci.h | 1 +
> drivers/usb/usbip/vhci_hcd.c | 37 ++++++++++------------------------
> drivers/usb/usbip/vhci_sysfs.c | 19 +++++++++++++++++
> 3 files changed, 31 insertions(+), 26 deletions(-)
>
> diff --git a/drivers/usb/usbip/vhci.h b/drivers/usb/usbip/vhci.h
> index 5659dce1526e..844500e3b0ea 100644
> --- a/drivers/usb/usbip/vhci.h
> +++ b/drivers/usb/usbip/vhci.h
> @@ -121,6 +121,7 @@ struct vhci_hcd {
> extern int vhci_num_controllers;
> extern struct vhci *vhcis;
> extern struct attribute_group vhci_attr_group;
Is this still needed here? And can't it be static now?
> @@ -1565,9 +1546,12 @@ static int __init vhci_hcd_init(void)
>
> return 0;
>
> +
Extra line?
thanks,
greg k-h
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] usbip: vhci_hcd: let the driver core manage the sysfs attributes
2026-08-16 4:55 ` Greg KH
@ 2026-08-16 6:11 ` Deepanshu Kartikey
2026-08-16 6:56 ` Greg KH
0 siblings, 1 reply; 4+ messages in thread
From: Deepanshu Kartikey @ 2026-08-16 6:11 UTC (permalink / raw)
To: Greg KH
Cc: valentina.manea.m, shuah, i, yuyang.du, linux-usb, linux-kernel,
syzbot+8753715f05759f1a10de
On Sun, Aug 16, 2026 at 10:26 AM Greg KH <gregkh@linuxfoundation.org> wrote:
>
> > extern struct vhci *vhcis;
> > extern struct attribute_group vhci_attr_group;
>
> Is this still needed here? And can't it be static now?
>
> > @@ -1565,9 +1546,12 @@ static int __init vhci_hcd_init(void)
> >
> > return 0;
> >
> > +
>
> Extra line?
>
> thanks,
>
> greg k-h
I have sent patch v3
Thanks
Deepanshu
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] usbip: vhci_hcd: let the driver core manage the sysfs attributes
2026-08-16 6:11 ` Deepanshu Kartikey
@ 2026-08-16 6:56 ` Greg KH
0 siblings, 0 replies; 4+ messages in thread
From: Greg KH @ 2026-08-16 6:56 UTC (permalink / raw)
To: Deepanshu Kartikey
Cc: valentina.manea.m, shuah, i, yuyang.du, linux-usb, linux-kernel,
syzbot+8753715f05759f1a10de
On Sun, Aug 16, 2026 at 11:41:41AM +0530, Deepanshu Kartikey wrote:
> On Sun, Aug 16, 2026 at 10:26 AM Greg KH <gregkh@linuxfoundation.org> wrote:
> >
>
> > > extern struct vhci *vhcis;
> > > extern struct attribute_group vhci_attr_group;
> >
> > Is this still needed here? And can't it be static now?
> >
> > > @@ -1565,9 +1546,12 @@ static int __init vhci_hcd_init(void)
> > >
> > > return 0;
> > >
> > > +
> >
> > Extra line?
> >
> > thanks,
> >
> > greg k-h
>
> I have sent patch v3
Please slow down, there is no rush here, you shouldn't be sending
patches more than one every few days.
greg k-h
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-08-16 6:57 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-08-16 1:50 [PATCH v2] usbip: vhci_hcd: let the driver core manage the sysfs attributes Deepanshu Kartikey
2026-08-16 4:55 ` Greg KH
2026-08-16 6:11 ` Deepanshu Kartikey
2026-08-16 6:56 ` Greg KH
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox