Messages in this thread | | | Subject | Re: [PATCH] 2.6.16-rc1-mm1 - produce useful info for kzalloc with DEBUG_SLAB | From | Valdis.Kletnieks@vt ... | Date | Thu, 19 Jan 2006 04:47:03 -0500 |
| |
On Thu, 19 Jan 2006 10:58:13 +0200, Pekka Enberg said: > On 1/19/06, Valdis.Kletnieks@vt.edu <Valdis.Kletnieks@vt.edu> wrote: > > The following patch makes a few minor changes so the CONFIG_DEBUG_SLAB > > statistics report the actual caller for kzalloc() - otherwise its call to > > kmalloc() just points at kzalloc(). Basically, we force __always_inline on > > several routines, so the __builtin_return_address calls point where we > > want them to point, even if gcc wouldn't otherwise do it. > > Couldn't we use this [1] trick Steven came up with for this? > > 1. http://article.gmane.org/gmane.linux.kernel/362494
I posted the basic idea of this patch back on Dec 18, Steven came up with his stuff about 2 weeks later, and it's a bit too creative with its use of the preprocessor for my tastes, so I didn't retrofit the idea.
On the other hand, I'm not *that* attached to my solution - if somebody wants to code a patch Steven's way and toss it to Andrew, they're welcome to do so, and we'll let Andrew decide. ;) [unhandled content-type:application/pgp-signature] | |