Ingo Oeser wrote:
Valentine Barshak schrieb:
Oliver Neukum wrote:
Am Montag 22 Oktober 2007 schrieb Valentine Barshak:
static int asix_mdio_read(struct net_device *netdev, int phy_id, int loc)
{
struct usbnet *dev = netdev_priv(netdev);
+ void *buf;
u16 res;
mutex_lock(&dev->phy_mutex);
asix_set_sw_mii(dev);
+
+ buf = kmalloc(2, GFP_KERNEL);
This is done under lock. Can you allocate the buffer once and reuse it?
I think we can use 2 bytes of the usbnet data buffer for this.
I'll submit a new patch soon.
If this cannot be done for some reason, then you can at least kmalloc() before
you do "mutex_lock(&dev->phy_mutex);" and kfree() after you did
"mutex_unlock(&dev->phy_mutex);"
The reason to can do this, is that "buf" has a life time limited to this
function.
The reason you should do this, is that kmalloc(, GFP_KERNEL) is allowed to
sleep,
which will block the mutex for that time. While this is technically ok,
since mutexes can sleep, it is not desireable, since other users of that mutex
are blocked until the allocation is done.
If you are able to implement the "2 bytes of usbnet data buffer" version,
please ignore that mail :-)
Best Regards
Ingo Oeser
Looks like we cannot use usbnet data buffer for read transfers either,
because it's just a part of the usbnet structure and we may still lose
data while invalidating cache the same way we do using buffers on stack.
Allocating a permanent buffer for phy transfer needs more driver
changes: we should add unbind finction to the device_info for all asix
devices to deallocate the buffer at exit.
And we still need to allocate buffers for other transfers dynamically,
so having just one permanent buffer for phy doesn't help much.
I've reworked the original patch a bit: moved kmalloc out of the
phy_mutex and added more clean-ups.
Will submit shortly.
Thanks,
Valentine.
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
