Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 20 Jun 2016 15:58:32 -0400
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: build musl for armv7m

On Thu, Jun 16, 2016 at 11:34:28AM -0700, Zhao, Weiming wrote:
> I tried to build for armv6m (cortex-m0) and I got other build issues
> with .S and inline asms.
> 
> Below are the changes for building armv7m:

I thought I'd already replied to this but I don't see my reply so I'm
doing it [again?] now.

> diff --git a/src/setjmp/arm/longjmp.s b/src/setjmp/arm/longjmp.s
> index e28d8f3..e9b9b32 100644
> --- a/src/setjmp/arm/longjmp.s
> +++ b/src/setjmp/arm/longjmp.s
> @@ -8,7 +8,9 @@ longjmp:
>      mov ip,r0
>      movs r0,r1
>      moveq r0,#1
> -    ldmia ip!, {v1,v2,v3,v4,v5,v6,sl,fp,sp,lr}
> +    ldmia ip!, {v1,v2,v3,v4,v5,v6,sl,fp}
> +    ldr sp, [ip]!
> +    ldr lr, [ip]!

I think changes like this are ok. They could be conditional on
__thumb__ if they hurt performance measurably on arm but I doubt it
matters.

> diff --git a/src/string/arm/memcpy_le.S b/src/string/arm/memcpy_le.S
> index 4db4844..2517d15 100644
> --- a/src/string/arm/memcpy_le.S
> +++ b/src/string/arm/memcpy_le.S
> @@ -241,7 +241,8 @@ non_congruent:
>      beq     2f
>      ldr     r5, [r1], #4
>      sub     r2, r2, #4
> -    orr     r4, r3, r5,             lsl lr
> +    lsl     r4, r5, lr
> +    orr     r4, r3, r4
>      mov     r3, r5,                 lsr r12
>      str     r4, [r0], #4
>      cmp     r2, #4

If this is in a hot path it may need to be conditional.

> diff --git a/src/thread/arm/atomics.s b/src/thread/arm/atomics.s
> index 673fc03..a4bd03a 100644
> --- a/src/thread/arm/atomics.s
> +++ b/src/thread/arm/atomics.s
> @@ -6,7 +6,8 @@
>  .type __a_barrier,%function
>  __a_barrier:
>      ldr ip,1f
> -    ldr ip,[pc,ip]
> +    add ip,pc,ip
> +    ldr ip,[ip]
>      add pc,pc,ip
>  1:    .word __a_barrier_ptr-1b
>  .global __a_barrier_dummy

As far as I can tell, this does not work at all. The arithmetic on pc
is assuming the particular offset between the instruction using pc and
the following code as arm opcodes.

There's also the matter of the cp15 register load in this file that
doesn't exist on cortex-m. IMO the kernel (or bare-metal trap handler)
needs to trap and emulate it.

Rich

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.