Re: [PATCH] serial: Initialize spinlocks in 8250 and don't clobber them.

[David Daney <ddaney@caviumnetworks.com>]

Alan Cox wrote:
> > But yes, copying a spinlock by value is quite wrong.  Perhaps we could
> > retain the struct assigment and then run spin_lock_init() to get the
> > spinlock into a sane state?
> Kind of irrelevant now however, the split of patches that caused the
> original bug is over and the NR_IRQ removal patch half of it hit Linus
> tree.
My original patch fixed *two* problems.  As you note here, you already fixed 
the first one.

As far as I know, the second problem is still present, and that is what akpm 
was referring to above.  Several days ago I posted a revised patch for this 
here:

http://marc.info/?l=linux-serial&m=122408950013741&w=2

The question is:  What is the best way to initialize some (or all) fields of a 
structure *except* a single lock field that was previously initialized?

We can just copy field by field as my patch does, or you could do something 
ugly using memcpy on portions of the structure.  In this case we know which 
structure elements will be used by the early console, so I just copied them.

Any comments about that patch are certainly most welcome.

Thanks,
David Daney