From: Matthew Wilcox <willy@infradead.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: stable <stable@vger.kernel.org>,
Ilias Apalodimas <ilias.apalodimas@linaro.org>,
Jesper Dangaard Brouer <brouer@redhat.com>,
Vlastimil Babka <vbabka@suse.cz>,
Matteo Croce <mcroce@linux.microsoft.com>,
Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] mm: fix struct page layout on 32-bit systems
Date: Sun, 16 May 2021 20:09:20 +0100 [thread overview]
Message-ID: <YKFt4Njj5au/JEhT@casper.infradead.org> (raw)
In-Reply-To: <CAHk-=wihKAt+Wz6=nccQAXxi_VWFJpx4JwWTJSwT0UvUs1RtZw@mail.gmail.com>
On Sun, May 16, 2021 at 11:27:10AM -0700, Linus Torvalds wrote:
> On Sun, May 16, 2021 at 11:22 AM Matthew Wilcox <willy@infradead.org> wrote:
> >
> > Nobody's been willing to guarantee that all 32-bit architectures keep the
> > top 20 bits clear for their DMA addresses. I've certainly seen hardware
> > (maybe PA-RISC? MIPS?) which uses the top few bits of the DMA address to
> > indicate things like "coherent" or "bypasses IOMMU". Rather than trying
> > to find out, I thought this was the safer option.
>
> Fair enough. I just find it somewhat odd.
>
> But I still find this a bit ugly. Maybe we could just have made that
> one sub-structure "__aligned(4)", and avoided this all, and let the
> compiler generate the split load (or, more likely, just let the
> compiler generate a regular load from an unaligned location).
>
> IOW, just
>
> struct { /* page_pool used by netstack */
> /**
> * @dma_addr: might require a 64-bit value even on
> * 32-bit architectures.
> */
> dma_addr_t dma_addr;
> } __aligned((4));
>
> without the magic shifting games?
That was the other problem fixed by this patch -- on big-endian 32-bit
platforms with 64-bit dma_addr_t (mips, ppc), a DMA address with bit 32 set
inadvertently sets the PageTail bit. So we need to store the low bits
in the first word, even on big-endian platforms.
There's an upcoming patch to move dma_addr out of the union with
compound_head, but that requires changing page_is_pfmemalloc() to
use an indicator other than page->index == -1. Once we do that,
we can have fun with __aligned(). Since we knew it would have to
be backported, this felt like the least risky patch to start with.
next prev parent reply other threads:[~2021-05-16 19:09 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-16 12:18 [PATCH] mm: fix struct page layout on 32-bit systems Matthew Wilcox (Oracle)
2021-05-16 12:31 ` Matthew Wilcox
2021-05-17 10:06 ` Greg KH
2021-05-16 16:29 ` Linus Torvalds
2021-05-16 18:22 ` Matthew Wilcox
2021-05-16 18:27 ` Linus Torvalds
2021-05-16 19:09 ` Matthew Wilcox [this message]
2021-05-16 19:22 ` Linus Torvalds
2021-05-17 2:42 ` Matthew Wilcox
2021-05-17 22:18 ` Linus Torvalds
2021-05-17 23:02 ` Matthew Wilcox
2021-05-17 23:37 ` Linus Torvalds
2021-05-18 0:02 ` Linus Torvalds
2021-05-18 0:53 ` Matthew Wilcox
-- strict thread matches above, loose matches on Subject: below --
2021-05-10 15:32 [PATCH] mm: Fix " Matthew Wilcox (Oracle)
2021-05-10 16:38 ` Matteo Croce
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=YKFt4Njj5au/JEhT@casper.infradead.org \
--to=willy@infradead.org \
--cc=akpm@linux-foundation.org \
--cc=brouer@redhat.com \
--cc=ilias.apalodimas@linaro.org \
--cc=mcroce@linux.microsoft.com \
--cc=stable@vger.kernel.org \
--cc=torvalds@linux-foundation.org \
--cc=vbabka@suse.cz \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).