Bug#286803: Kind of a memory leak in kernel 2.6?
- To: 286803@bugs.debian.org
- Subject: Bug#286803: Kind of a memory leak in kernel 2.6?
- From: Ingo Strüwing <ingo@mysql.com>
- Date: Tue, 09 Aug 2005 10:48:48 +0200
- Message-id: <[🔎] 1123577328.2126.38.camel@chilla.local>
- Reply-to: Ingo Strüwing <ingo@mysql.com>, 286803@bugs.debian.org
- In-reply-to: <1113466481.1155.17.camel@chilla.local>
- References: <1107964249.1184.48.camel@chilla.local> <20050210013331.GD26536@verge.net.au> <1108022820.1146.6.camel@chilla.local> <20050210092827.GA8571@verge.net.au> <1108737210.3587.44.camel@chilla.local> <20050223032838.GH11232@verge.net.au> <1109149793.3565.8.camel@chilla.local> <20050223102048.GB11232@verge.net.au> <1113466481.1155.17.camel@chilla.local>
Hi,
Am Donnerstag, den 14.04.2005, 10:14 +0200 schrieb Ingo Strüwing:
...
> - Has this problem be passed to the Linux kernel developers? What is the
> kernel developers bug list in which to find the problem?
> bugzilla.kernel.org does not seem to know of it.
>
> - Does anyone know if there are any tools that show, which memory pages
> are allocated to which process (or purpose)? That way one might get a
> hint what these "lost" pages were used for.
unfortunately I still did not get an answer to these questions yet.
I have additional information:
- The problem is independent from the "overcommit_memory" setting
(/proc/sys/vm/overcommit_memory, /usr/src/linux/Documentation/vm/overcommit-accounting).
- Allocating memory with malloc() until it fails does not show up in the
"Active" memory value. Only initializing the allocated memory does. This
means that the pages which remain in the "Active" memory after my
library build loop are pages which have been in real use.
Regards,
Ingo
--
Ingo Strüwing, Senior Software Developer
MySQL AB, www.mysql.com
Reply to: