linux-kernel - Re: [RFC PATCH 1/5] treewide: replace __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, 29 Aug 2019 09:29:23 -0700
From:   Joe Perches <joe@...ches.com>
To:     Rasmus Villemoes <linux@...musvillemoes.dk>, x86@...nel.org,
        linux-kernel@...r.kernel.org
Cc:     Ingo Molnar <mingo@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        "H. Peter Anvin" <hpa@...or.com>, Nadav Amit <namit@...are.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Miguel Ojeda <miguel.ojeda.sandonis@...il.com>,
        ndesaulniers@...gle.com
Subject: Re: [RFC PATCH 1/5] treewide: replace __inline__ by inline

On Thu, 2019-08-29 at 10:32 +0200, Rasmus Villemoes wrote:
> Currently, compiler_types.h #defines __inline__ as inline. However,
> that defeats the purpose of gcc providing __inline__ as an alternate
> spelling of that keyword - namely, that it is always accessible under
> that name, even if one chooses to #define inline, which we also do in
> order to attach attribute(gnu_inline), and sometimes imply
> attribute(always_inline), etc.
> 
> Note that it is quite possible that some header file defines a static
> inline function before the include chain has reached compiler_types.h,
> but in that case both the existing __inline__ as well as the new
> inline spelling refer to gcc's keyword, and the redefinitions in
> compiler_types.h have no effect anyway.
> 
> For those static inline definitions that appear after compiler_types.h
> has been processed, this is obviously a no-op due to the #define
> __inline__ inline.
> 
> We will need to be able to use the __inline__ keyword to make use of
> the "asm inline()" feature from gcc 9. This is preparation for
> removing the #define.
> 
> Generated by
> 
>   git grep --files-with-matches -w __inline__ | \
>     grep -vE '^(usr|Documentation|scripts)/' | \
>     grep -v /uapi/ | \
>     xargs sed -i -e 's/static *__inline__/static inline/'

bikeshed: Perhaps better to use

	xargs sed -i -e 's/\b__inline__\b/inline/g'

As static and __inline__ do not have to be on the same line.

for instance: (and I didn't look thoroughly)

This misses one of the uses of __static__ in
drivers/scsi/qla2xx/qla_os.c

The patch has:

> diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
[]
> @@ -342,7 +342,7 @@ qla2x00_restart_timer(scsi_qla_host_t *vha, unsigned long interval)
>  	mod_timer(&vha->timer, jiffies + interval * HZ);
>  }
>  
> -static __inline__ void
> +static inline void
>  qla2x00_stop_timer(scsi_qla_host_t *vha)
>  {
>  	del_timer_sync(&vha->timer);

But the code is:

drivers/scsi/qla2xxx/qla_os.c-318-/* TODO Convert to inlines
drivers/scsi/qla2xxx/qla_os.c-319- *
drivers/scsi/qla2xxx/qla_os.c-320- * Timer routines
drivers/scsi/qla2xxx/qla_os.c-321- */
drivers/scsi/qla2xxx/qla_os.c-322-
drivers/scsi/qla2xxx/qla_os.c:323:__inline__ void
drivers/scsi/qla2xxx/qla_os.c-324-qla2x00_start_timer(scsi_qla_host_t *vha, unsigned long interval)
drivers/scsi/qla2xxx/qla_os.c-325-{
drivers/scsi/qla2xxx/qla_os.c-326-      timer_setup(&vha->timer, qla2x00_timer, 0);
drivers/scsi/qla2xxx/qla_os.c-327-      vha->timer.expires = jiffies + interval * HZ;
drivers/scsi/qla2xxx/qla_os.c-328-      add_timer(&vha->timer);
--
drivers/scsi/qla2xxx/qla_os.c-340-      }
drivers/scsi/qla2xxx/qla_os.c-341-
drivers/scsi/qla2xxx/qla_os.c-342-      mod_timer(&vha->timer, jiffies + interval * HZ);
drivers/scsi/qla2xxx/qla_os.c-343-}
drivers/scsi/qla2xxx/qla_os.c-344-
drivers/scsi/qla2xxx/qla_os.c:345:static __inline__ void
drivers/scsi/qla2xxx/qla_os.c-346-qla2x00_stop_timer(scsi_qla_host_t *vha)
drivers/scsi/qla2xxx/qla_os.c-347-{
drivers/scsi/qla2xxx/qla_os.c-348-      del_timer_sync(&vha->timer);
drivers/scsi/qla2xxx/qla_os.c-349-      vha->timer_active = 0;
drivers/scsi/qla2xxx/qla_os.c-350-}

The possible impacted files using 's/\b__inline__\b/inline/'

$ git grep -w '__inline__' -- '*.[ch]' | grep -vP 'static\s+__inline__'
(only hand selected matches)
arch/alpha/include/asm/compiler.h:#undef __inline__
arch/alpha/include/asm/floppy.h:__inline__ void virtual_dma_init(void)
arch/ia64/hp/common/sba_iommu.c:#define SBA_INLINE	__inline__
drivers/parisc/sba_iommu.c:#define SBA_INLINE	__inline__
drivers/scsi/qla2xxx/qla_os.c:__inline__ void
drivers/video/fbdev/intelfb/intelfbdrv.c:__inline__ int intelfb_var_to_depth(const struct fb_var_screeninfo *var)