linux-kernel - Re: [PATCH v3 3/5] x86/sgx: Optimize the free_cnt count in sgx_epc

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 for Android: free password hash cracker in your pocket

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

Message-ID: <483f1ff9-9022-b7fc-5931-79c0c83f1289@linux.alibaba.com>
Date:   Tue, 16 Feb 2021 11:30:13 +0800
From:   Tianjia Zhang <tianjia.zhang@...ux.alibaba.com>
To:     Jarkko Sakkinen <jarkko@...nel.org>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        "H. Peter Anvin" <hpa@...or.com>,
        Sean Christopherson <seanjc@...gle.com>,
        Shuah Khan <shuah@...nel.org>, x86@...nel.org,
        linux-sgx@...r.kernel.org, linux-kselftest@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Jia Zhang <zhang.jia@...ux.alibaba.com>
Subject: Re: [PATCH v3 3/5] x86/sgx: Optimize the free_cnt count in
 sgx_epc_section



On 2/12/21 8:19 PM, Jarkko Sakkinen wrote:
> On Thu, Feb 11, 2021 at 02:04:12PM +0800, Tianjia Zhang wrote:
>> Hi,
>>
>> Sorry for the late reply.
>>
>> On 1/28/21 1:40 AM, Jarkko Sakkinen wrote:
>>> I could bet some money that this does not bring any significant
>>> performance gain.
>>>
>>
>> Yes, this does not bring performance gains. This is not a change for
>> performance, mainly to make the value of free_cnt look more accurate.
>>
>>> On Sun, Jan 24, 2021 at 02:29:05PM +0800, Tianjia Zhang wrote:
>>>> `section->free_cnt` represents the free page in sgx_epc_section,
>>>> which is assigned once after initialization. In fact, just after the
>>>> initialization is completed, the pages are in the `init_laundry_list`
>>>> list and cannot be allocated. This needs to be recovered by EREMOVE
>>>> of function sgx_sanitize_section() before it can be used as a page
>>>> that can be allocated. The sgx_sanitize_section() will be called in
>>>> the kernel thread ksgxd.
>>>>
>>>> This patch moves the initialization of `section->free_cnt` from the
>>>> initialization function `sgx_setup_epc_section()` to the function
>>>> `sgx_sanitize_section()`, and then accumulates the count after the
>>>
>>> Use single quotes instead of hyphens.
>>>>> successful execution of EREMOVE. This seems to be more reasonable,
>>>> free_cnt will also truly reflect the allocatable free pages in EPC.
>>>>
>>>> Sined-off-by: Tianjia Zhang <tianjia.zhang@...ux.alibaba.com>
>>>> Reviewed-by: Sean Christopherson <seanjc@...gle.com>
>>>> ---
>>>>    arch/x86/kernel/cpu/sgx/main.c | 2 +-
>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c
>>>> index 4465912174fd..e455ec7b3449 100644
>>>> --- a/arch/x86/kernel/cpu/sgx/main.c
>>>> +++ b/arch/x86/kernel/cpu/sgx/main.c
>>>> @@ -48,6 +48,7 @@ static void sgx_sanitize_section(struct sgx_epc_section *section)
>>>>    		if (!ret) {
>>>>    			spin_lock(&section->lock);
>>>>    			list_move(&page->list, &section->page_list);
>>>> +			section->free_cnt++;
>>>>    			spin_unlock(&section->lock);
>>>
>>> Someone can try to allocate a page while sanitize process is in progress.
>>>
>>> I think it is better to keep critical sections in the form that when you
>>> leave from one, the global state is legit.
>>>
>>
>> Do you mean to move the critical section to protect the entire while loop?
>> Of course, this is also possible, sanitize is a process only needed for
>> initialization, and the possibility of conflict is very small.
>>
>> Best regards,
>> Tianjia
> 
> The big picture of this change to me, to be frank is that it's completely
> useless.
> 
> Please start with the picture.
> 
> /Jarkko
> 

I carefully considered your suggestion, and I will delete 2/5 and 3/5 in 
the next version.

Best regards,
Tianjia