diff mbox

[RFC] crypto: sha1/arm - fix build with CONFIG_THUMB2_KERNEL

Message ID 20121030100742.17910.70241.stgit@localhost6.localdomain6 (mailing list archive)
State New, archived
Headers show

Commit Message

Jussi Kivilinna Oct. 30, 2012, 10:07 a.m. UTC
Building cryptodev-2.6 tree with CONFIG_THUMB2_KERNEL=y and
CONFIG_CRYPTO_SHA1_ARM=y give following error:

   AS      arch/arm/crypto/sha1-armv4-large.o
arch/arm/crypto/sha1-armv4-large.S: Assembler messages:
arch/arm/crypto/sha1-armv4-large.S:197: Error: r13 not allowed here --
`teq r14,sp'
arch/arm/crypto/sha1-armv4-large.S:377: Error: r13 not allowed here --
`teq r14,sp'
arch/arm/crypto/sha1-armv4-large.S:469: Error: r13 not allowed here --
`teq r14,sp'

Build can be fixed with adding '.arm' at head of sha1-armv4-large.S.
However I'm not sure if this correct fix as this is my first stab
at arm assembler and I could not get CONFIG_THUMB2_KERNEL boot up
on 'qemu-system-arm -M vexpress-a9' to verify correctness. So this
is only build tested.

Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-crypto@vger.kernel.org
Cc: David McCullough <ucdevel@gmail.com>
---
 arch/arm/crypto/sha1-armv4-large.S |    2 ++
 1 file changed, 2 insertions(+)

Comments

David McCullough Oct. 31, 2012, 12:25 a.m. UTC | #1
Jussi Kivilinna wrote the following:
> Building cryptodev-2.6 tree with CONFIG_THUMB2_KERNEL=y and
> CONFIG_CRYPTO_SHA1_ARM=y give following error:
> 
>    AS      arch/arm/crypto/sha1-armv4-large.o
> arch/arm/crypto/sha1-armv4-large.S: Assembler messages:
> arch/arm/crypto/sha1-armv4-large.S:197: Error: r13 not allowed here --
> `teq r14,sp'
> arch/arm/crypto/sha1-armv4-large.S:377: Error: r13 not allowed here --
> `teq r14,sp'
> arch/arm/crypto/sha1-armv4-large.S:469: Error: r13 not allowed here --
> `teq r14,sp'
> 
> Build can be fixed with adding '.arm' at head of sha1-armv4-large.S.
> However I'm not sure if this correct fix as this is my first stab
> at arm assembler and I could not get CONFIG_THUMB2_KERNEL boot up
> on 'qemu-system-arm -M vexpress-a9' to verify correctness. So this
> is only build tested.

That fixes the problem for the assembler,  in that you have forced it to
treat the code as 32 bit (non-thumb).

I know you can mix thumb and non-thumb code,  and it seems that as long as
all parties know what code is in what mode it should work.

Unfortunately I don't know enough about it be sure this is a valid fix.
I think we need a long term arm expert to pass judgement on this one,
unless of source you manage to get it tested.  I don't have any thumb
targets I can try it on, sory,

Cheers,
Davidm

> Cc: linux-arm-kernel@lists.infradead.org
> Cc: linux-crypto@vger.kernel.org
> Cc: David McCullough <ucdevel@gmail.com>
> ---
>  arch/arm/crypto/sha1-armv4-large.S |    2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/arch/arm/crypto/sha1-armv4-large.S b/arch/arm/crypto/sha1-armv4-large.S
> index 7050ab1..e8a55eb 100644
> --- a/arch/arm/crypto/sha1-armv4-large.S
> +++ b/arch/arm/crypto/sha1-armv4-large.S
> @@ -53,6 +53,8 @@
>  
>  .text
>  
> +.arm
> +
>  .global	sha1_block_data_order
>  .type	sha1_block_data_order,%function
>  
>
diff mbox

Patch

diff --git a/arch/arm/crypto/sha1-armv4-large.S b/arch/arm/crypto/sha1-armv4-large.S
index 7050ab1..e8a55eb 100644
--- a/arch/arm/crypto/sha1-armv4-large.S
+++ b/arch/arm/crypto/sha1-armv4-large.S
@@ -53,6 +53,8 @@ 
 
 .text
 
+.arm
+
 .global	sha1_block_data_order
 .type	sha1_block_data_order,%function