linux-kernel - Re: [PATCH v9 00/21] Generic page walk and ptdump

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]

Date:   Wed, 31 Jul 2019 12:18:34 +0100
From:   Steven Price <steven.price@....com>
To:     Sven Schnelle <svens@...ckframe.org>
Cc:     Mark Rutland <Mark.Rutland@....com>,
        Peter Zijlstra <peterz@...radead.org>,
        Catalin Marinas <catalin.marinas@....com>,
        Dave Hansen <dave.hansen@...ux.intel.com>, linux-mm@...ck.org,
        "H. Peter Anvin" <hpa@...or.com>, Will Deacon <will@...nel.org>,
        "Liang, Kan" <kan.liang@...ux.intel.com>,
        Helge Deller <deller@....de>, x86@...nel.org,
        Ingo Molnar <mingo@...hat.com>, Arnd Bergmann <arnd@...db.de>,
        Anshuman Khandual <anshuman.khandual@....com>,
        Jérôme Glisse <jglisse@...hat.com>,
        Borislav Petkov <bp@...en8.de>,
        Andy Lutomirski <luto@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        linux-arm-kernel@...ts.infradead.org,
        Ard Biesheuvel <ard.biesheuvel@...aro.org>,
        linux-kernel@...r.kernel.org, James Morse <james.morse@....com>,
        Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH v9 00/21] Generic page walk and ptdump

On 31/07/2019 10:27, Sven Schnelle wrote:
> Hi Steven,
> 
> On Mon, Jul 29, 2019 at 12:32:25PM +0100, Steven Price wrote:
>>
>> parisc is more interesting and I'm not sure if this is necessarily
>> correct. I originally proposed a patch with the line "For parisc, we
>> don't support large pages, so add stubs returning 0" which got Acked by
>> Helge Deller. However going back to look at that again I see there was a
>> follow up thread[2] which possibly suggests I was wrong?
> 
> I just started a week ago implementing ptdump for PA-RISC. Didn't notice that
> you're working on making it generic, which is nice. I'll adjust my code
> to use the infrastructure you're currently developing.

Great, hopefully it will make it easier to implement.

>> Can anyone shed some light on whether parisc does support leaf entries
>> of the page table tree at a higher than the normal depth?
>>
>> [1] https://lkml.org/lkml/2019/2/27/572
>> [2] https://lkml.org/lkml/2019/3/5/610
> 
> My understanding is that PA-RISC only has leaf entries on PTE level.

Yes, that's my current interpretation.

>> The intention is that the page table walker would be available for all
>> architectures so that it can be used in any generic code - PTDUMP simply
>> seemed like a good place to start.
>>
>>> Now that pmd_leaf() and pud_leaf() are getting used in walk_page_range() these
>>> functions need to be defined on all arch irrespective if they use PTDUMP or not
>>> or otherwise just define it for archs which need them now for sure i.e x86 and
>>> arm64 (which are moving to new generic PTDUMP framework). Other archs can
>>> implement these later.
> 
> I'll take care of the PA-RISC part - for 32 bit your generic code works, for 64Bit
> i need to learn a bit more about the following hack:
> 
> arch/parisc/include/asm/pgalloc.h:15
> /* Allocate the top level pgd (page directory)
>  *
>  * Here (for 64 bit kernels) we implement a Hybrid L2/L3 scheme: we
>  * allocate the first pmd adjacent to the pgd.  This means that we can
>  * subtract a constant offset to get to it.  The pmd and pgd sizes are
>  * arranged so that a single pmd covers 4GB (giving a full 64-bit
>  * process access to 8TB) so our lookups are effectively L2 for the
>  * first 4GB of the kernel (i.e. for all ILP32 processes and all the
>  * kernel for machines with under 4GB of memory)
>  */

As far as I understand this, the page table tree isn't any different
here. It's just that there's a PMD which is allocated at the same time
as the PGD. The PGD's first entry then points to the PMD (P4D/PUD are
folded). There are then some tricks which means that for addresses < 4GB
the PGD stage can be skipped because you already know where the relevant
PMD is.

However, nothing should stop a simple walk from PGD down - it's just an
optimisation to remove the pointer fetch from PGD in the usual case for
accesses < 4GB.

> I see that your change clear P?D entries when p?d_bad() returns true, which - i think -
> would be the case with the PA-RISC implementation.

The only case where p?d_bad() is checked is at the PGD and P4D levels
(unless I'm missing something?). I have to admit I'm a little unsure
about this. Basically the code as it stands doesn't allow leaf entries
at PGD or P4D. I'm not aware of any architectures that do this though.

Thanks,

Steve