linux-kernel - Re: [PATCH] kbuild, sparseirq: work around GCC bug with _

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: <20081227105944.GA13198@elte.hu>
Date:	Sat, 27 Dec 2008 11:59:44 +0100
From:	Ingo Molnar <mingo@...e.hu>
To:	Paul Mackerras <paulus@...ba.org>
Cc:	Sam Ravnborg <sam@...nborg.org>, Yinghai Lu <yinghai@...nel.org>,
	David Howells <dhowells@...hat.com>,
	Kamalesh Babulal <kamalesh@...ux.vnet.ibm.com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Stephen Rothwell <sfr@...b.auug.org.au>,
	linux-next@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>,
	mel@....ul.ie
Subject: Re: [PATCH] kbuild, sparseirq: work around GCC bug with __weak
	aliases


* Paul Mackerras <paulus@...ba.org> wrote:

> Ingo Molnar writes:
> 
> > * Ingo Molnar <mingo@...e.hu> wrote:
> > 
> > > > I recall David Howells had a similar issue with the bootparamter patch 
> > > > set. The workaround he used was to add a barrier(); call in the weak 
> > > > function to avoid the inline.
> > > 
> > > could we add some extra attribute to __weak that would have a similar 
> > > effect? Something like __attribute__((noinline)), or something silly 
> > > like __attribute__((deprecated)) - just to keep gcc from screwing up 
> > > __weak functions? Perhaps adding a section attribute would have a 
> > > similar effect? (putting weak definitions into an extra section is 
> > > probably helpful anyway)
> > 
> > I've applied the patch below to tip/irq/sparseirq - could someone with an 
> > affected GCC version please check whether this solves the crash?
> 
> I recall from discussions earlier that noinline doesn't fix the problem, 
> and I just tested a similar case and verified that adding noinline 
> doesn't stop some versions of gcc from inlining them.  The empty weak 
> functions in kernel/perf_counter.c were getting inlined by the cross-gcc 
> (gcc 4.1.1) I use for compiling powerpc kernels on my laptop, and adding 
> noinline doesn't help there.

hm, does Yinghai's patch below do the trick? (also in perfcounters/core 
and tip/master)

	Ingo

-------------------->
>From 01ea1ccaa24dea3552e103be13b7897211607a8b Mon Sep 17 00:00:00 2001
From: Yinghai Lu <yinghai@...nel.org>
Date: Fri, 26 Dec 2008 21:05:06 -0800
Subject: [PATCH] perf_counter: more barrier in blank weak function

Impact: fix panic possible panic

Some versions of GCC inline the weak global function if it's empty.
Add a barrier() to work it around.

Signed-off-by: Yinghai Lu <yinghai@...nel.org>
Signed-off-by: Ingo Molnar <mingo@...e.hu>
---
 kernel/perf_counter.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index d7a79f3..37f7716 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -45,8 +45,8 @@ hw_perf_counter_init(struct perf_counter *counter)
 }
 
 u64 __weak hw_perf_save_disable(void)		{ return 0; }
-void __weak hw_perf_restore(u64 ctrl)		{ }
-void __weak hw_perf_counter_setup(void)		{ }
+void __weak hw_perf_restore(u64 ctrl)		{ barrier(); }
+void __weak hw_perf_counter_setup(void)		{ barrier(); }
 
 static void
 list_add_counter(struct perf_counter *counter, struct perf_counter_context *ctx)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/