linux-kernel - Re: [PATCH 02/11] Add "handle page fault" PV helper.

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: <4AF28B49.7000509@redhat.com>
Date:	Thu, 05 Nov 2009 10:22:33 +0200
From:	Avi Kivity <avi@...hat.com>
To:	"Tian, Kevin" <kevin.tian@...el.com>
CC:	Rik van Riel <riel@...hat.com>, Ingo Molnar <mingo@...e.hu>,
	Gleb Natapov <gleb@...hat.com>,
	"kvm@...r.kernel.org" <kvm@...r.kernel.org>,
	"linux-mm@...ck.org" <linux-mm@...ck.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"H. Peter Anvin" <hpa@...or.com>,
	Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH 02/11] Add "handle page fault" PV helper.

On 11/05/2009 08:44 AM, Tian, Kevin wrote:
>> From: Avi Kivity
>> Sent: 2009年11月3日 12:57
>>
>> On 11/03/2009 01:35 AM, Rik van Riel wrote:
>>     
>>>> We can't add an exception vector since all the existing 
>>>>         
>> ones are either
>>     
>>>> taken or reserved.
>>>>         
>>>
>>> I believe some are reserved for operating system use.
>>>       
>> Table 6-1 says:
>>
>>   9 |  | Coprocessor Segment Overrun (reserved)  |  Fault |  No  | 
>> Floating-point instruction.2
>>   15 |  ― |  (Intel reserved. Do not use.) |   | No |
>>   20-31 |  ― | Intel reserved. Do not use.  |
>>   32-255 |  ―  | User Defined (Non-reserved) Interrupts |  Interrupt  
>> |   | External interrupt or INT n instruction.
>>
>> So we can only use 32-255, but these are not fault-like 
>> exceptions that 
>> can be delivered with interrupts disabled.
>>
>>     
> would you really want to inject a fault-like exception here? Fault
> is architurally synchronous event while here apf is more like an 
> asynchronous interrupt as it's not caused by guest itself. If 
> guest is with interrupt disabled, preemption won't happen and 
> apf path just ends up "wait for page" hypercall to waste cycles.
>   

An async page fault is, despite its name, synchronous, since it is
associated with an instruction. It must either be delivered immediately
or not at all.

It's true that in kernel mode you can't do much with an apf if
interrupts are disabled, but you still want to receive apfs for user
mode with interrupts disabled (for example due to interrupt shadow).

-- 
error compiling committee.c: too many arguments to function

--
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/