Re: [PATCH v2] mm: Restructure kmem_cache_create() to move debug cache integrity checks into a new function

[Shuah Khan <shuah.khan@hp.com>]

On Thu, 2012-08-09 at 14:08 -0500, Christoph Lameter (Open Source)
wrote:
> On Thu, 9 Aug 2012, Shuah Khan wrote:
> 
> > Moving these checks into kmem_cache_sanity_check() would mean return
> > path handling will change. The first block of sanity checks for name,
> > and size etc. are done before holding the slab_mutex and the second
> > block that checks the slab lists is done after holding the mutex.
> > Depending on which one fails, return handling is going to be different
> > in that if second block fails, mutex needs to be unlocked and when the
> > first block fails, there is no need to do that. Nothing that is too
> > complex to solve, just something that needs to be handled.
> 
> Right. The taking of the mutex etc is not depending on the parameters at
> all. So its possible. Its rather simple.
> 
> > Comments, thoughts on
> >
> > 1. just remove size from kmem_cache_sanity_check() parameters
> > or
> > 2. move first block sanity checks into kmem_cache_sanity_check()
> >
> > Personally I prefer the first option to avoid complexity in return path
> > handling. Would like to hear what others think.
> 
> We already have to deal with the return path handling for other failure
> cases.

Thanks for the feedback. I will send v3 patch with the changes we
discussed.

-- Shuah