Werner Almesberger wrote:
Michael Trimarchi wrote:
Subject: [PATCH 3/4] Little clean up and move the do_irq on a deferrable work
Hmm, why do you call it a tasklet if it's a work queue ? :)
It's the old name, I can fix it. But the patches are in early stage again
Second, why do you need that change in the first place ? Does the
old mechanism have performance issues ? Or are you running into a
locking conflict somewhere ?
I verified it, it contains a warning if a request arrive during a pending
request and I don't have any warning during tests.
Making the interrupt processing deferrable looks dangerous to me.
If you need this to make things work, this usually suggests that
the logic is upside-down somewhere else.
Seems to be that is not dangerues and clean up the code that now is
readable. But it's
only a cleanup, and I don't see any regression in performance
Michael
- Werner
