[musl] [PATCH] use __getauxval in mallocng

[musl] [PATCH] use __getauxval in mallocng

[Érico Nogueira]

saves around 20 bytes of .text
---
 src/malloc/mallocng/glue.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/malloc/mallocng/glue.h b/src/malloc/mallocng/glue.h
index 151c48b8..4b988fb2 100644
--- a/src/malloc/mallocng/glue.h
+++ b/src/malloc/mallocng/glue.h
@@ -12,6 +12,7 @@
 #include "libc.h"
 #include "lock.h"
 #include "dynlink.h"
+#include "sys/auxv.h"
 
 // use macros to appropriately namespace these.
 #define size_classes __malloc_size_classes
@@ -42,9 +43,8 @@
 static inline uint64_t get_random_secret()
 {
 	uint64_t secret = (uintptr_t)&secret * 1103515245;
-	for (size_t i=0; libc.auxv[i]; i+=2)
-		if (libc.auxv[i]==AT_RANDOM)
-			memcpy(&secret, (char *)libc.auxv[i+1]+8, sizeof secret);
+	const char *at_random = (void *)__getauxval(AT_RANDOM);
+	if (at_random) memcpy(&secret, at_random+8, sizeof secret);
 	return secret;
 }
 
-- 
2.37.2

Re: [musl] [PATCH] use __getauxval in mallocng

[Rich Felker]

On Mon, Aug 15, 2022 at 03:21:10PM -0300, Érico Nogueira wrote:
> saves around 20 bytes of .text
> ---
>  src/malloc/mallocng/glue.h | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/src/malloc/mallocng/glue.h b/src/malloc/mallocng/glue.h
> index 151c48b8..4b988fb2 100644
> --- a/src/malloc/mallocng/glue.h
> +++ b/src/malloc/mallocng/glue.h
> @@ -12,6 +12,7 @@
>  #include "libc.h"
>  #include "lock.h"
>  #include "dynlink.h"
> +#include "sys/auxv.h"
>  
>  // use macros to appropriately namespace these.
>  #define size_classes __malloc_size_classes
> @@ -42,9 +43,8 @@
>  static inline uint64_t get_random_secret()
>  {
>  	uint64_t secret = (uintptr_t)&secret * 1103515245;
> -	for (size_t i=0; libc.auxv[i]; i+=2)
> -		if (libc.auxv[i]==AT_RANDOM)
> -			memcpy(&secret, (char *)libc.auxv[i+1]+8, sizeof secret);
> +	const char *at_random = (void *)__getauxval(AT_RANDOM);
> +	if (at_random) memcpy(&secret, at_random+8, sizeof secret);
>  	return secret;
>  }
>  
> -- 
> 2.37.2

Hm, this might be worth doing. I'm trying to remember if there was a
reason it wasn't done initially. It wasn't absence of a namespace-safe
version because __getauxval already existed when mallocng was merged.
And it doesn't look like there's an early-use issue before getauxval
is usable either. Might it just be more expensive in the static linked
case where __getauxval isn't otherwise used?

Rich

Re: [musl] [PATCH] use __getauxval in mallocng

[Érico Nogueira]

On Mon Aug 15, 2022 at 3:31 PM -03, Rich Felker wrote:
> On Mon, Aug 15, 2022 at 03:21:10PM -0300, Érico Nogueira wrote:
> > saves around 20 bytes of .text
> > ---
> >  src/malloc/mallocng/glue.h | 6 +++---
> >  1 file changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/src/malloc/mallocng/glue.h b/src/malloc/mallocng/glue.h
> > index 151c48b8..4b988fb2 100644
> > --- a/src/malloc/mallocng/glue.h
> > +++ b/src/malloc/mallocng/glue.h
> > @@ -12,6 +12,7 @@
> >  #include "libc.h"
> >  #include "lock.h"
> >  #include "dynlink.h"
> > +#include "sys/auxv.h"
> >  
> >  // use macros to appropriately namespace these.
> >  #define size_classes __malloc_size_classes
> > @@ -42,9 +43,8 @@
> >  static inline uint64_t get_random_secret()
> >  {
> >  	uint64_t secret = (uintptr_t)&secret * 1103515245;
> > -	for (size_t i=0; libc.auxv[i]; i+=2)
> > -		if (libc.auxv[i]==AT_RANDOM)
> > -			memcpy(&secret, (char *)libc.auxv[i+1]+8, sizeof secret);
> > +	const char *at_random = (void *)__getauxval(AT_RANDOM);
> > +	if (at_random) memcpy(&secret, at_random+8, sizeof secret);
> >  	return secret;
> >  }
> >  
> > -- 
> > 2.37.2
>
> Hm, this might be worth doing. I'm trying to remember if there was a
> reason it wasn't done initially. It wasn't absence of a namespace-safe
> version because __getauxval already existed when mallocng was merged.
> And it doesn't look like there's an early-use issue before getauxval
> is usable either. Might it just be more expensive in the static linked
> case where __getauxval isn't otherwise used?

That does seem to be the case :/

From some quick testing, a program calling only malloc and freeing it
got ~60 bytes bigger.

I don't think there's a good way to get the best outcome in both
situations, is there? And 20 bytes in a shared mapping makes more sense.

>
> Rich