From: Konstantin Khlebnikov <koct9i@gmail.com>
To: Matthew Wilcox <mawilcox@linuxonhyperv.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
"Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
Andrew Morton <akpm@linux-foundation.org>,
Ross Zwisler <ross.zwisler@linux.intel.com>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
"linux-mm@kvack.org" <linux-mm@kvack.org>,
linux-fsdevel <linux-fsdevel@vger.kernel.org>,
Matthew Wilcox <mawilcox@microsoft.com>
Subject: Re: [PATCH 2/2] radix-tree: Fix optimisation problem
Date: Sat, 24 Sep 2016 11:36:27 +0300 [thread overview]
Message-ID: <CALYGNiNczCgx31dVPtdoN2kOhUEk2uCTk_Xd_28JrRTy5_Ee8g@mail.gmail.com> (raw)
In-Reply-To: <1474570415-14938-3-git-send-email-mawilcox@linuxonhyperv.com>
On Thu, Sep 22, 2016 at 9:53 PM, Matthew Wilcox
<mawilcox@linuxonhyperv.com> wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>
>
> When compiling the radix tree with -O2, GCC thinks it can optimise:
>
> void *entry = parent->slots[offset];
> int siboff = entry - parent->slots;
> void *slot = parent->slots + siboff;
>
> into
>
> void *slot = entry;
>
> Unfortunately, 'entry' is a tagged pointer, so this optimisation leads
> to getting an unaligned pointer back from radix_tree_lookup_slot().
> The test suite wasn't being compiled with optimisation, so we hadn't
> spotted it before now. Change the test suite to compile with -O2, and
> fix the optimisation problem by passing 'entry' through entry_to_node()
> so gcc knows this isn't a plain pointer.
> ---
> lib/radix-tree.c | 3 ++-
> tools/testing/radix-tree/Makefile | 2 +-
> 2 files changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/lib/radix-tree.c b/lib/radix-tree.c
> index 1b7bf73..8bf1f32 100644
> --- a/lib/radix-tree.c
> +++ b/lib/radix-tree.c
> @@ -105,7 +105,8 @@ static unsigned int radix_tree_descend(struct radix_tree_node *parent,
>
> #ifdef CONFIG_RADIX_TREE_MULTIORDER
> if (radix_tree_is_internal_node(entry)) {
> - unsigned long siboff = get_slot_offset(parent, entry);
> + unsigned long siboff = get_slot_offset(parent,
> + (void **)entry_to_node(entry));
As I see this is the only place where get_slot_offset used for
unaligned pointer.
Nobody uses "multiorder entries" so this never happens. And I have
plan to kill this code.
> if (siboff < RADIX_TREE_MAP_SIZE) {
> offset = siboff;
> entry = rcu_dereference_raw(parent->slots[offset]);
> diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
> index 3b53046..9d0919ed 100644
> --- a/tools/testing/radix-tree/Makefile
> +++ b/tools/testing/radix-tree/Makefile
> @@ -1,5 +1,5 @@
>
> -CFLAGS += -I. -g -Wall -D_LGPL_SOURCE
> +CFLAGS += -I. -g -O2 -Wall -D_LGPL_SOURCE
> LDFLAGS += -lpthread -lurcu
> TARGETS = main
> OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
> --
> 2.9.3
>
next prev parent reply other threads:[~2016-09-24 8:37 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-09-22 18:53 [PATCH 0/2] Fix radix_tree_lookup_slot() Matthew Wilcox
2016-09-22 18:53 ` [PATCH 1/2] radix tree test suite: Test radix_tree_replace_slot() for multiorder entries Matthew Wilcox
2016-09-22 18:53 ` [PATCH 2/2] radix-tree: Fix optimisation problem Matthew Wilcox
2016-09-22 18:09 ` Linus Torvalds
2016-09-23 20:16 ` Matthew Wilcox
2016-09-24 20:21 ` Linus Torvalds
2016-09-24 20:47 ` Linus Torvalds
2016-09-24 21:04 ` Kirill A. Shutemov
2016-09-24 22:54 ` Linus Torvalds
2016-09-24 8:36 ` Konstantin Khlebnikov [this message]
2016-09-24 23:35 ` Cedric Blancher
2016-09-25 0:18 ` Linus Torvalds
2016-09-25 17:59 ` Cedric Blancher
[not found] ` <CA+55aFzge97L-JLKZq0CTW1wtMOsnt8QzOw3b5qCMmzbKxZ5aw@mail.gmail.com>
2016-09-25 19:04 ` Linus Torvalds
2016-09-25 19:40 ` Cedric Blancher
2016-09-25 19:56 ` Linus Torvalds
2016-09-26 21:28 ` Matthew Wilcox
2016-09-26 21:48 ` Cedric Blancher
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=CALYGNiNczCgx31dVPtdoN2kOhUEk2uCTk_Xd_28JrRTy5_Ee8g@mail.gmail.com \
--to=koct9i@gmail.com \
--cc=akpm@linux-foundation.org \
--cc=kirill.shutemov@linux.intel.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mawilcox@linuxonhyperv.com \
--cc=mawilcox@microsoft.com \
--cc=ross.zwisler@linux.intel.com \
--cc=torvalds@linux-foundation.org \
/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).