linux-kernel - Re: [2.6 patch] blackfin: "extern inline" -> "static inline"

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 25 Oct 2007 22:46:31 +0200
From:	Adrian Bunk <bunk@...nel.org>
To:	Mike Frysinger <vapier.adi@...il.com>
Cc:	"H. Peter Anvin" <hpa@...or.com>,
	Jie Zhang <jzhang.linux@...il.com>, bryan.wu@...log.com,
	linux-kernel@...r.kernel.org
Subject: Re: [2.6 patch] blackfin: "extern inline" -> "static inline"

On Thu, Oct 25, 2007 at 04:20:20PM -0400, Mike Frysinger wrote:
> On 10/25/07, Adrian Bunk <bunk@...nel.org> wrote:
> > On Thu, Oct 25, 2007 at 04:07:45PM -0400, Mike Frysinger wrote:
> > > On 10/25/07, Adrian Bunk <bunk@...nel.org> wrote:
> > > > On Thu, Oct 25, 2007 at 12:16:40PM -0400, Mike Frysinger wrote:
> > > > > On 10/25/07, Adrian Bunk <bunk@...nel.org> wrote:
> > > > > > On Wed, Oct 24, 2007 at 08:06:14PM -0700, H. Peter Anvin wrote:
> > > > > > > Mike Frysinger wrote:
> > > > > > >> we'll have to either use the gcc attributes to force old inline
> > > > > > >> behavior or use the gcc flag to force it
> > > > > > >
> > > > > > > We should probably have an extern_inline define then, assuming this is a
> > > > > > > function that does exist in a linkable version already -- otherwise "static
> > > > > > > inline" is correct.
> > > > > >
> > > > > > Since we #define inline to be __attribute__((always_inline))
> > > > > > "extern inline" with the old semantics would only behave differently
> > > > > > if someone took the address of one of these string functions.
> > > > >
> > > > > that isnt what we intended ;)
> > > > >
> > > > > > Does this happen anywhere in the blackfin port?
> > > > >
> > > > > gcc is also free to ignore the optimized inline in favor of an
> > > > > external reference
> > > >
> > > > It is not since we #define inline to be __attribute__((always_inline)).
> > >
> > > as i said, that was not what we intended ... and actually,
> > > always_inline does not mean always inline ... it is still possible to
> > > get gcc to not inline things when building debug versions.
> >
> > Do you have any example for your claim "to get gcc to not inline things
> > when building debug versions"?
> 
> $ cat test.c
> __attribute__((always_inline)) int foo(void) { return 0; }
> int main(void){ return foo(); }
> $ gcc -g test.c -o test
> $ readelf -s test | grep FUNC | grep -v _
>     61: 00000000004004b8    11 FUNC    GLOBAL DEFAULT   13 foo
>     68: 00000000004004c3    11 FUNC    GLOBAL DEFAULT   13 main
> 
> looks pretty straightforward to me

That a function gets emitted for a non-static inline function is 
expected (and unrelated to the static inline cases).

But if you replace "__attribute__((always_inline))" with
"inline __attribute__((always_inline))" as we do in the kernel,
foo() is no longer called from main().

And if you make it "static inline __attribute__((always_inline))" (which 
is what "static inline" in the kernel expands to), there's no longer a 
function "foo" in the binary.

> -mike

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

-
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/