Sorry Dave,
The delay on that issue is my fault. While Dan is an Intel employee,
he's in a totally separate part of the company (and a different state)
from me and the other networking folks.
I was trying to determine if reordering so that the spin_lock was inside
the lock_cpu_hotplug would work, or if something else was needed, Dan
suggested making the net_dma_event_lock a mutex, but then as Andrew
pointed out as long as a lock is being held preemption is disabled and
we're safe from cpu hotplug.
Simple patch to remove the lock_cpu_hotplug/unlock_cpu_hotplug calls as
Andrew suggested coming up.
- Chris
> -----Original Message-----
> From: David Miller [mailto:[email protected]]
> Sent: Monday, July 31, 2006 11:12 PM
> To: Williams, Dan J
> Cc: [email protected]; [email protected];
> [email protected]; Leech, Christopher;
> [email protected]
> Subject: Re: [PATCH rev2 1/4] dmaengine: enable mutliple
> clients and operations
>
>
>
> Can I ask that the known bugs in the I/O AT DMA code be fixed
> before we start adding new features to it?
>
> Specifically, the lock_cpu_hotplug() call in net_dma_rebalance()
> is still there and being invoked with a spinlock held. The
> spinlock is grabbed by the caller, netdev_dma_event() which
> grabs the net_dma_event_lock spinlock.
>
> You cannot invoke lock_cpu_hotplug() while holding a spinlock
> because lock_cpu_hotplug(), as seen in kernel/cpu.c, takes
> a semaphore which can sleep. Sleeping while holding a spinlock
> is not allowed.
>
> This is the second time I have tried to make the Intel developers
> aware of this bug. So please fix this problem.
>
> Thanks a lot.
>