linux-kernel - Re: [PATCH v2 1/2] ACPI/PPTT: Add support for ACPI 6.3 thread flag

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

Message-ID: <63f6c6a8-9d79-ae75-3c15-96bded9b14e4@huawei.com>
Date:   Wed, 19 Jun 2019 10:15:24 +0100
From:   John Garry <john.garry@...wei.com>
To:     Jeremy Linton <jeremy.linton@....com>,
        Valentin Schneider <valentin.schneider@....com>,
        <linux-arm-kernel@...ts.infradead.org>
CC:     <linux-acpi@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <catalin.marinas@....com>, <will.deacon@....com>,
        <rjw@...ysocki.net>, <sudeep.holla@....com>, <lenb@...nel.org>
Subject: Re: [PATCH v2 1/2] ACPI/PPTT: Add support for ACPI 6.3 thread flag

On 18/06/2019 22:28, Jeremy Linton wrote:
> Hi,
>
> On 6/18/19 12:23 PM, John Garry wrote:
>> On 18/06/2019 15:40, Valentin Schneider wrote:
>>> On 18/06/2019 15:21, Jeremy Linton wrote:
>>> [...]
>>>>>> + * Return: -ENOENT if the PPTT doesn't exist, the CPU cannot be
>>>>>> found or
>>>>>> + *       the table revision isn't new enough.
>>>>>> + * Otherwise returns flag value
>>>>>> + */
>>>>>
>>>>> Nit: strictly speaking we're not returning the flag value but its mask
>>>>> applied to the flags field. I don't think anyone will care about
>>>>> getting
>>>>> the actual flag value, but it should be made obvious in the doc:
>>>>
>>>> Or I clarify the code to actually do what the comments says. Maybe
>>>> that is what John G was also pointing out too?
>>>>
>>
>> No, I was just saying that the kernel topology can be broken without
>> this series.
>>
>>>
>>> Mmm I didn't find any reply from John regarding this in v1, but I
>>> wouldn't
>>> mind either way, as long as the doc & code are aligned.
>>>
>>
>> BTW, to me, function acpi_pptt_cpu_is_thread() seems to try to do too
>> much, i.e. check if the PPTT is new enough to support the thread flag
>> and also check if it is set for a specific cpu. I'd consider separate
>> functions here.
>

Hi,

> ? Your suggesting replacing the
>

I am not saying definitely that this should be changed, it's just that 
acpi_pptt_cpu_is_thread() returning false, true, or "no entry" is not a 
typical API format.

How about acpi_pptt_support_thread_info(cpu) and 
acpi_pptt_cpu_is_threaded(cpu), both returning false/true only?

None of this is ideal.

BTW, Have you audited which arm64 systems have MT bit set legitimately?

>
> if (table->revision >= rev)

I know that checking the table revision is not on the fast path, but it 
seems unnecessarily inefficient to always read it this way, I mean 
calling acpi_table_get().

Can you have a static value for the table revision? Or is this just how 
other table info is accessed in ACPI code?

>     cpu_node = acpi_find_processor_node(table, acpi_cpu_id);
>
> check with
>
> if (revision_check(table, rev))
>     cpu_node = acpi_find_processor_node(table, acpi_cpu_id);
>
>
> and a function like
>
> static int revision_check(acpixxxx *table, int rev)
> {
>     return (table->revision >= rev);
> }
>
> Although, frankly if one were to do this, it should probably be a macro
> with the table type, and used in the dozen or so other places I found
> doing similar checks (spcr, iort, etc).
>
> Or something else?
>
>
>
>

thanks,
John

>>
>>> [...]
>>>
>>> .
>>>
>>
>>
>
>
> .
>