2021-08-18 13:04:04

by Nadezda Lutovinova

Subject: [PATCH] usb: gadget: mv_u3d: Change functon call in mv_u3d_probe()

If IRQ occurs between calling request_irq() and mv_u3d_eps_init(),
then null pointer dereference occurs since u3d->eps[] wasn't
initialized yet but used in mv_u3d_nuke().

The patch puts registration of the interrupt handler after
initializing of neccesery data.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Nadezda Lutovinova <[email protected]>
---
drivers/usb/gadget/udc/mv_u3d_core.c | 19 ++++++++++---------
1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/drivers/usb/gadget/udc/mv_u3d_core.c b/drivers/usb/gadget/udc/mv_u3d_core.c
index ce3d7a3eb7e3..a1057ddfbda3 100644
--- a/drivers/usb/gadget/udc/mv_u3d_core.c
+++ b/drivers/usb/gadget/udc/mv_u3d_core.c
@@ -1921,14 +1921,6 @@ static int mv_u3d_probe(struct platform_device *dev)
goto err_get_irq;
}
u3d->irq = r->start;
- if (request_irq(u3d->irq, mv_u3d_irq,
- IRQF_SHARED, driver_name, u3d)) {
- u3d->irq = 0;
- dev_err(&dev->dev, "Request irq %d for u3d failed\n",
- u3d->irq);
- retval = -ENODEV;
- goto err_request_irq;
- }

/* initialize gadget structure */
u3d->gadget.ops = &mv_u3d_ops; /* usb_gadget_ops */
@@ -1941,6 +1933,15 @@ static int mv_u3d_probe(struct platform_device *dev)

mv_u3d_eps_init(u3d);

+ if (request_irq(u3d->irq, mv_u3d_irq,
+ IRQF_SHARED, driver_name, u3d)) {
+ u3d->irq = 0;
+ dev_err(&dev->dev, "Request irq %d for u3d failed\n",
+ u3d->irq);
+ retval = -ENODEV;
+ goto err_request_irq;
+ }
+
/* external vbus detection */
if (u3d->vbus) {
u3d->clock_gating = 1;
@@ -1964,8 +1965,8 @@ static int mv_u3d_probe(struct platform_device *dev)

err_unregister:
free_irq(u3d->irq, u3d);
-err_request_irq:
err_get_irq:
+err_request_irq:
kfree(u3d->status_req);
err_alloc_status_req:
kfree(u3d->eps);
--
2.17.1

2021-08-18 13:16:36

by Felipe Balbi

Subject: Re: [PATCH] usb: gadget: mv_u3d: Change functon call in mv_u3d_probe()

Hi,

(first of all, your subject could be a little more descriptive,
something like:

usb: gadget: mv_u3d: request_irq() after initializing UDC

as that would better detail what you're doing)

Nadezda Lutovinova <[email protected]> writes:

> If IRQ occurs between calling request_irq() and mv_u3d_eps_init(),
> then null pointer dereference occurs since u3d->eps[] wasn't
> initialized yet but used in mv_u3d_nuke().
>
> The patch puts registration of the interrupt handler after
> initializing of neccesery data.
>
> Found by Linux Driver Verification project (linuxtesting.org).

this looks like an important bug fix, it probably deserves a stable tag
here. Which commit introduce this problem? Otherr than that, commits
looks good.

--
balbi

2021-08-18 14:20:33

by Nadezda Lutovinova

Subject: [PATCH v2] usb: gadget: mv_u3d: request_irq() after initializing UDC

If IRQ occurs between calling request_irq() and mv_u3d_eps_init(),
then null pointer dereference occurs since u3d->eps[] wasn't
initialized yet but used in mv_u3d_nuke().

The patch puts registration of the interrupt handler after
initializing of neccesery data.

Found by Linux Driver Verification project (linuxtesting.org).

Fixes: 90fccb529d24 ("usb: gadget: Gadget directory cleanup - group UDC drivers")
Signed-off-by: Nadezda Lutovinova <[email protected]>
---
v2: fix subject and add a stable tag
---
drivers/usb/gadget/udc/mv_u3d_core.c | 19 ++++++++++---------
1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/drivers/usb/gadget/udc/mv_u3d_core.c b/drivers/usb/gadget/udc/mv_u3d_core.c
index ce3d7a3eb7e3..a1057ddfbda3 100644
--- a/drivers/usb/gadget/udc/mv_u3d_core.c
+++ b/drivers/usb/gadget/udc/mv_u3d_core.c
@@ -1921,14 +1921,6 @@ static int mv_u3d_probe(struct platform_device *dev)
goto err_get_irq;
}
u3d->irq = r->start;
- if (request_irq(u3d->irq, mv_u3d_irq,
- IRQF_SHARED, driver_name, u3d)) {
- u3d->irq = 0;
- dev_err(&dev->dev, "Request irq %d for u3d failed\n",
- u3d->irq);
- retval = -ENODEV;
- goto err_request_irq;
- }

/* initialize gadget structure */
u3d->gadget.ops = &mv_u3d_ops; /* usb_gadget_ops */
@@ -1941,6 +1933,15 @@ static int mv_u3d_probe(struct platform_device *dev)

mv_u3d_eps_init(u3d);

+ if (request_irq(u3d->irq, mv_u3d_irq,
+ IRQF_SHARED, driver_name, u3d)) {
+ u3d->irq = 0;
+ dev_err(&dev->dev, "Request irq %d for u3d failed\n",
+ u3d->irq);
+ retval = -ENODEV;
+ goto err_request_irq;
+ }
+
/* external vbus detection */
if (u3d->vbus) {
u3d->clock_gating = 1;
@@ -1964,8 +1965,8 @@ static int mv_u3d_probe(struct platform_device *dev)

err_unregister:
free_irq(u3d->irq, u3d);
-err_request_irq:
err_get_irq:
+err_request_irq:
kfree(u3d->status_req);
err_alloc_status_req:
kfree(u3d->eps);
--
2.17.1

2021-08-18 14:36:30

by Felipe Balbi

Subject: Re: [PATCH v2] usb: gadget: mv_u3d: request_irq() after initializing UDC

Nadezda Lutovinova <[email protected]> writes:

> If IRQ occurs between calling request_irq() and mv_u3d_eps_init(),
> then null pointer dereference occurs since u3d->eps[] wasn't
> initialized yet but used in mv_u3d_nuke().
>
> The patch puts registration of the interrupt handler after
> initializing of neccesery data.
>
> Found by Linux Driver Verification project (linuxtesting.org).
>
> Fixes: 90fccb529d24 ("usb: gadget: Gadget directory cleanup - group UDC drivers")
> Signed-off-by: Nadezda Lutovinova <[email protected]>

Thanks for updating so quickly:

Acked-by: Felipe Balbi <[email protected]>

--
balbi