[<prev] [next>] [day] [month] [year] [list]
Message-ID: <11814119-0e5c-3e4f-29c4-d2e9b8dfb622@suse.cz>
Date: Tue, 22 Mar 2022 16:10:46 +0100
From: Vlastimil Babka <vbabka@...e.cz>
To: Yinan Zhang <zhangyinan2019@...il.szu.edu.cn>,
Matthew Wilcox <willy@...radead.org>
Cc: akpm@...ux-foundation.org, william.kucharski@...cle.com,
linux-kernel@...r.kernel.org, linux-mm@...ck.org,
yejiajian2018@...il.szu.edu.cn, hanshenghong2019@...il.szu.edu.cn,
caoyixuan2019@...il.szu.edu.cn, zhaochongxi2019@...il.szu.edu.cn,
yuhongf@....edu.cn
Subject: Re: [PATCH 1/2] mm/page_owner.c: introduce vmalloc allocator for
page_owner
On 3/22/22 15:48, Yinan Zhang wrote:
> on 2022/3/22 11:28, Matthew Wilcox wrote:
>
>> On Tue, Mar 22, 2022 at 11:22:24AM +0800, Yinan Zhang wrote:
>>> An application's memory consumption is high and keeps increasing,
>>> then it is suspected of having memory leak. There are several
>>> commonly used memory allocators: slab, cma, vmalloc, etc. The memory
>>> leak identification can be speed up if page information allocated
>>> by an individual allocator are analyzed individually. This patch
>>> introduce vmalloc allocator for page_owner.
>> Why is /proc/vmallocinfo not enough?
>
> There is only one owner information in /proc/vmallocinfo: caller.
> While there are plenty of owner information in page owner,
> besides caller, it includes task id (pid), process id (tgid),
> and other useful information, which enables us for more comprehensive
> page allocation/release relevant statistics and analysis.
>
> Adding allocator infomation can augment page owner for tracking about
> who allocates each page. And this patch adds vmalloc allocator infomation.
Hm but page_owner has the whole stack trace of allocation so it should be
easy to determine vmalloc from slab etc? How exactly would this extra field
help?
Powered by blists - more mailing lists