[lkml]   [2002]   [Feb]   [14]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] __free_pages_ok oops
On Thu, 14 Feb 2002, Andrea Arcangeli wrote:
> On Thu, Feb 14, 2002 at 12:10:37PM +0100, Gerd Knorr wrote:
> >
> > I've recently changed the code to make it *not* call unmap_kiobuf/vfree
> > from irq context. Instead bttv 0.8.x doesn't allow you to close the
> > device with DMA xfers in flight. If you try this the release() fops
> > handler will block until the transfer is done, then unmap_kiobuf from
> > process context, then return.
> perfect, that's the right fix for 2.4 (waiting DMA to complete at
> ->release looks also much saner). unmap_kiobuf wasn't supposed to be run
> from irq handlers. Everything dealing with userspace mappings cannot run
> from irq handlers, tlb flushes, VM, swapping etc... everything must run
> from normal kernel context. If you obey this rule, my previous email to
> this thread will still apply. I wasn't aware of bttv running
> unmap_kiobuf from irq.

It's good that Gerd has made his change, but we don't know who else
might have been doing similar. unmap_kiobuf does not involve unmapping
virtual address space: it used to be safe run from irq handlers, now not.

We don't have to make a change for 2.4.18, but we really should add some
kind of safety check there in 2.4.soon: either of the "it's a BUG()"
kind I first suggested (which may embarrass us by firing too often),
or of the "we can handle that" kind which I last suggested.

I don't disagree with Andrew's and your count-LRU approach,
but it does have slight drawbacks, as you noted.

> As said this should be a matter only for 2.5, now that Gerd recalls
> unmap_kiobuf from normal kernel context.

That's just a hope: you may be right, we simply don't know.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:24    [W:0.073 / U:0.300 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site