On 2/13/07, Arjan van de Ven <[email protected]> wrote:
Hi,
while working on the last pieces of the file_ops constantification, DVB
is the small village in France that is holding the Romans at bay... but
I think I found the final flaw in it now:
*pdvbdev = dvbdev = kmalloc(sizeof(struct dvb_device), GFP_KERNEL);
if (!dvbdev) {
mutex_unlock(&dvbdev_register_lock);
return -ENOMEM;
}
memcpy(dvbdev, template, sizeof(struct dvb_device));
dvbdev->type = type;
dvbdev->id = id;
dvbdev->adapter = adap;
dvbdev->priv = priv;
dvbdev->fops->owner = adap->module;
this is the place in DVB that is writing to a struct file_operations.
But as with almost all such cases in the kernel, this one is buggy:
While the code nicely copies a template dvbdev, that template only has a
pointer to a *shared* fops struct, the copy doesn't help that. So this
code is overwriting the fops owner field for ALL active devices, not
just the ones the copy of the template is for....
While working on a new device node, i stumbled across a similar issue
where attaching the frontend failed due to some sort of memory
corruption. This was seen as all the callbacks suddenly just vanished,
eventhough it existed in the driver
At that point, i figured it could be something due to an error at my
side , since the device that i was working was a bit complex and had
API changes as well.
Thanks for pointing it out. It looks like the issue sounds similar.
regards,
Manu
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]
