Thank you for your contribution! Sashiko AI review found 1 potential issue(s) 
to consider:

Pre-existing issues:
- [High] A time-of-check to time-of-use (TOCTOU) race condition in 
`vmbus_probe()` allows `drv->probe` to be called with a NULL `dev_id`, leading 
to a NULL pointer dereference in drivers like `storvsc_probe()`.
--

commit 76d8c5c515ad429741f331c31ec380bf773956cb
Author: Runyu Xiao <[email protected]>

vmbus: use generic driver_override infrastructure

This commit switches the vmbus subsystem to use the generic driver core
driver_override infrastructure rather than bus-private storage. It removes
the private driver_override storage and updates the match path to use
device_match_driver_override() for locked reads, preventing the match
logic from inspecting freed memory during concurrent updates.

Link: https://lore.kernel.org/driver-core/[email protected]/

> diff --git a/drivers/hv/vmbus_drv.c b/drivers/hv/vmbus_drv.c
> index d28ff45d4cfdd..a81e2b0976367 100644
> --- a/drivers/hv/vmbus_drv.c
> +++ b/drivers/hv/vmbus_drv.c
[ ... ]
> @@ -708,9 +679,11 @@ static const struct hv_vmbus_device_id 
> *hv_vmbus_get_id(const struct hv_driver *
>  {
>       const guid_t *guid = &dev->dev_type;
>       const struct hv_vmbus_device_id *id;
> +     int ret;
>  
>       /* When driver_override is set, only bind to the matching driver */
> -     if (dev->driver_override && strcmp(dev->driver_override, drv->name))
> +     ret = device_match_driver_override(&dev->device, &drv->driver);
> +     if (ret == 0)
>               return NULL;

[Severity: High]
This is a pre-existing issue, but does this behavior still leave a
time-of-check to time-of-use race condition open when called later
during probing?

The driver core __driver_attach() calls driver_match_device() locklessly,
which delegates to vmbus_match() and eventually hv_vmbus_get_id(). If a
matching driver name is written to the driver_override sysfs attribute (or
a matching dynamic ID is present), it returns a match.

Before __driver_attach() acquires the device lock to call
driver_probe_device(), a concurrent sysfs write can clear the
driver_override attribute or remove the dynamic ID.

Inside vmbus_probe(), which executes with the device lock held,
hv_vmbus_get_id() is called again. Because the state was changed
concurrently, it can evaluate to NULL:

drivers/hv/vmbus_drv.c:vmbus_probe() {
    ...
    dev_id = hv_vmbus_get_id(drv, dev);
    if (drv->probe) {
        ret = drv->probe(dev, dev_id);
    ...
}

Since vmbus_probe() lacks a NULL check for dev_id, it directly calls
drv->probe(dev, dev_id) with dev_id equal to NULL.

Sub-drivers like storvsc_probe() then unconditionally dereference dev_id
(for example, checking dev_id->driver_data), which results in a kernel panic.

Could a NULL check for dev_id be added in vmbus_probe() to fully prevent
this race?

-- 
Sashiko AI review ยท 
https://sashiko.dev/#/patchset/[email protected]?part=3

Reply via email to