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: <87k0v7mrrd.fsf@nanos.tec.linutronix.de>
Date:   Fri, 30 Oct 2020 20:35:18 +0100
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Matthew Wilcox <willy@...radead.org>
Cc:     LKML <linux-kernel@...r.kernel.org>, linux-arch@...r.kernel.org,
        Linus Torvalds <torvalds@...uxfoundation.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Paul McKenney <paulmck@...nel.org>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        Ard Biesheuvel <ardb@...nel.org>,
        Herbert Xu <herbert@...dor.apana.org.au>,
        Christoph Hellwig <hch@....de>,
        Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Ingo Molnar <mingo@...nel.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        x86@...nel.org, Vineet Gupta <vgupta@...opsys.com>,
        linux-snps-arc@...ts.infradead.org,
        Russell King <linux@...linux.org.uk>,
        Arnd Bergmann <arnd@...db.de>,
        linux-arm-kernel@...ts.infradead.org, Guo Ren <guoren@...nel.org>,
        linux-csky@...r.kernel.org, Michal Simek <monstr@...str.eu>,
        Thomas Bogendoerfer <tsbogend@...ha.franken.de>,
        linux-mips@...r.kernel.org, Nick Hu <nickhu@...estech.com>,
        Greentime Hu <green.hu@...il.com>,
        Vincent Chen <deanbo422@...il.com>,
        Michael Ellerman <mpe@...erman.id.au>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        linuxppc-dev@...ts.ozlabs.org,
        "David S. Miller" <davem@...emloft.net>,
        sparclinux@...r.kernel.org, Chris Zankel <chris@...kel.net>,
        Max Filippov <jcmvbkbc@...il.com>,
        linux-xtensa@...ux-xtensa.org
Subject: Re: [patch V2 00/18] mm/highmem: Preemptible variant of kmap_atomic & friends

On Fri, Oct 30 2020 at 13:06, Matthew Wilcox wrote:
> On Thu, Oct 29, 2020 at 11:18:06PM +0100, Thomas Gleixner wrote:
>> This series provides kmap_local.* iomap_local variants which only disable
>> migration to keep the virtual mapping address stable accross preemption,
>> but do neither disable pagefaults nor preemption. The new functions can be
>> used in any context, but if used in atomic context the caller has to take
>> care of eventually disabling pagefaults.
>
> Could I ask for a CONFIG_KMAP_DEBUG which aliases all the kmap variants
> to vmap()?  I think we currently have a problem in iov_iter on HIGHMEM
> configs:

For kmap() that would work, but for kmap_atomic() not so much when it is
called in non-preemptible context because vmap() might sleep.

> copy_page_to_iter() calls page_copy_sane() which checks:
>
>         head = compound_head(page);
>         if (likely(n <= v && v <= page_size(head)))
>                 return true;
>
> but then:
>
>                 void *kaddr = kmap_atomic(page);
>                 size_t wanted = copy_to_iter(kaddr + offset, bytes, i);
>                 kunmap_atomic(kaddr);
>
> so if offset to offset+bytes is larger than PAGE_SIZE, this is going to
> work for lowmem pages and fail miserably for highmem pages.  I suggest
> vmap() because vmap has a PAGE_SIZE gap between each allocation.

On 32bit highmem the kmap_atomic() case is easy: Double the number of
mapping slots and only use every second one, which gives you a guard
page between the maps.

For 64bit we could do something ugly: Enable the highmem kmap_atomic()
crud and enforce an alias mapping (at least on the architectures where
this is reasonable). Then you get the same as for 32bit.

> Alternatively if we could have a kmap_atomic_compound(), that would
> be awesome, but probably not realistic to implement.  I've more
> or less resigned myself to having to map things one page at a time.

That might be horribly awesome on 32bit :)

Thanks,

        tglx

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ