Messages in this thread | | | Subject | Re: Kmemleak for mips | From | Catalin Marinas <> | Date | Tue, 03 Nov 2009 09:23:47 +0000 |
| |
On Mon, 2009-11-02 at 09:07 -0800, Luis R. Rodriguez wrote: > Curious what the limitations are on restricting kmemleak to non-mips > archs. I have a user and situation [1] where this could be helpful [1] > in debugging an issue. The user reports he cannot enable it on mips.
It may just work but cannot be enabled because I cannot test kmemleak on such hardware. In general you need to make sure that the _sdata/_edata and __bss_start/__bss_stop symbols are defined. If there are other ways of allocating memory than the standard API, it would need additional hooks. Some false-positives specific to MIPS may need to be annotated (usually with kmemleak_not_leak).
(btw, you could also merge the kmemleak.git tree on git.kernel.org as it has improvements on the rate of false positives; the patches will be pushed in 2.6.33-rc1)
-- Catalin
| |