Re: [PATCH v3 3/3] mm: kmemleak: Use the memory pool for early allocations

[Qian Cai <cai@lca.pw>]

On Mon, 2019-08-12 at 17:06 +0100, Catalin Marinas wrote:
> Currently kmemleak uses a static early_log buffer to trace all memory
> allocation/freeing before the slab allocator is initialised. Such early
> log is replayed during kmemleak_init() to properly initialise the
> kmemleak metadata for objects allocated up that point. With a memory
> pool that does not rely on the slab allocator, it is possible to skip
> this early log entirely.
> 
> In order to remove the early logging, consider kmemleak_enabled == 1 by
> default while the kmem_cache availability is checked directly on the
> object_cache and scan_area_cache variables. The RCU callback is only
> invoked after object_cache has been initialised as we wouldn't have any
> concurrent list traversal before this.
> 
> In order to reduce the number of callbacks before kmemleak is fully
> initialised, move the kmemleak_init() call to mm_init().
> 
> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
> ---
>  init/main.c       |   2 +-
>  lib/Kconfig.debug |  11 +-
>  mm/kmemleak.c     | 267 +++++-----------------------------------------
>  3 files changed, 35 insertions(+), 245 deletions(-)
> 
> diff --git a/init/main.c b/init/main.c
> index 96f8d5af52d6..ca05e3cd7ef7 100644
> --- a/init/main.c
> +++ b/init/main.c
> @@ -556,6 +556,7 @@ static void __init mm_init(void)
>  	report_meminit();
>  	mem_init();
>  	kmem_cache_init();
> +	kmemleak_init();
>  	pgtable_init();
>  	debug_objects_mem_init();
>  	vmalloc_init();
> @@ -740,7 +741,6 @@ asmlinkage __visible void __init start_kernel(void)
>  		initrd_start = 0;
>  	}
>  #endif
> -	kmemleak_init();
>  	setup_per_cpu_pageset();
>  	numa_policy_init();
>  	acpi_early_init();
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 4d39540011e2..39df06ffd9f4 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -592,17 +592,18 @@ config DEBUG_KMEMLEAK
>  	  In order to access the kmemleak file, debugfs needs to be
>  	  mounted (usually at /sys/kernel/debug).
>  
> -config DEBUG_KMEMLEAK_EARLY_LOG_SIZE
> -	int "Maximum kmemleak early log entries"
> +config DEBUG_KMEMLEAK_MEM_POOL_SIZE
> +	int "Kmemleak memory pool size"
>  	depends on DEBUG_KMEMLEAK
>  	range 200 40000
>  	default 16000

Hmm, this seems way too small. My previous round of testing with
kmemleak.mempool=524288 works quite well on all architectures.