linux-kernel - Re: [PATCH 6/7] BC: kernel memory (core)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <44FC4A2B.5050004@in.ibm.com>
Date:	Mon, 04 Sep 2006 21:15:47 +0530
From:	Balbir Singh <balbir@...ibm.com>
To:	Kirill Korotaev <dev@...ru>
Cc:	Andrew Morton <akpm@...l.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Alan Cox <alan@...rguk.ukuu.org.uk>,
	Christoph Hellwig <hch@...radead.org>,
	Pavel Emelianov <xemul@...nvz.org>,
	Andrey Savochkin <saw@...ru>, devel@...nvz.org,
	Rik van Riel <riel@...hat.com>, Andi Kleen <ak@...e.de>,
	Oleg Nesterov <oleg@...sign.ru>,
	Alexey Dobriyan <adobriyan@...l.ru>,
	Matt Helsley <matthltc@...ibm.com>,
	CKRM-Tech <ckrm-tech@...ts.sourceforge.net>
Subject: Re: [PATCH 6/7] BC: kernel memory (core)

Kirill Korotaev wrote:
> Balbir Singh wrote:
>> Kirill Korotaev wrote:

>>> +#ifdef CONFIG_BEANCOUNTERS
>>> +    union {
>>> +        struct beancounter    *page_bc;
>>> +    } bc;
>>> +#endif
>>> };
>>>
>>> +#define page_bc(page)            ((page)->bc.page_bc)
>>
>>
>> Minor comment - page->(bc).page_bc has too many repititions of page 
>> and bc - see
>> the Practice of Programming by Kernighan and Pike
>>
>> I missed the part of why you wanted to have a union (in struct page 
>> for bc)?
> because this union is used both for kernel memory accounting and user 
> memeory tracking.

Ok.. that's good. I remember seeing a user_bc sometime back in the code.
I had some idea about allowing tasks to migrate across resources (bean
counters), which I think can be easily done for user space pages, if the
user limits are tracked separately.

> 
>>> const char *bc_rnames[] = {
>>> +    "kmemsize",    /* 0 */
>>> };
>>>
>>> static struct hlist_head bc_hash[BC_HASH_SIZE];
>>> @@ -221,6 +222,8 @@ static void init_beancounter_syslimits(s
>>> {     int k;
>>>
>>> +    bc->bc_parms[BC_KMEMSIZE].limit = 32 * 1024 * 1024;
>>> +
>>
>>
>> Can't this be configurable CONFIG_XXX or a #defined constant?
> This is some arbitraty limited container, just to make sure it is not
> created unlimited. User space should initialize limits properly after 
> creation
> anyway. So I don't see reasons to make it configurable, do you?

May be its not very important now but configurable limits will help a confused
user. Even if we decide to use this number for now, a constant like
BC_DEFAULT_MEM_LIMIT is easier to read.

>> I wonder if bc_page_charge() should be called bc_page_charge_failed()?
>> Does it make sense to atleast partially start reclamation here? I know 
>> with
>> bean counters we cannot reclaim from a particular container, but for now
>> we could kick off kswapd() or call shrink_all_memory() inline (Dave's 
>> patches do this to shrink memory from the particular cpuset). Or do 
>> you want to leave this
>> slot open for later?
> yes. my intention is to account correctly all needed information first.
> After we agree on accounting, we can agree on how to do reclamaition.
> 

That sounds like a good plan.

-- 

	Balbir Singh,
	Linux Technology Center,
	IBM Software Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/