x86/asm/crypto: Don't use RBP as a scratch register

The frame pointer (RBP) is getting clobbered in
sha1_mb_mgr_submit_avx2() before a function call, which can mess up
stack traces. Use R12 instead.

Signed-off-by: Josh Poimboeuf <[e-mail]>
Cc: Andrew Morton <[e-mail]>
Cc: Andy Lutomirski <[e-mail]>
Cc: Andy Lutomirski <[e-mail]>
Cc: Arnaldo Carvalho de Melo <[e-mail]>
Cc: Bernd Petrovitsch <[e-mail]>
Cc: Borislav Petkov <[e-mail]>
Cc: Brian Gerst <[e-mail]>
Cc: Chris J Arges <[e-mail]>
Cc: Denys Vlasenko <[e-mail]>
Cc: H. Peter Anvin <[e-mail]>
Cc: Jiri Slaby <[e-mail]>
Cc: Linus Torvalds <[e-mail]>
Cc: Michal Marek <[e-mail]>
Cc: Namhyung Kim <[e-mail]>
Cc: Pedro Alves <[e-mail]>
Cc: Peter Zijlstra <[e-mail]>
Cc: Thomas Gleixner <[e-mail]>
Cc: [e-mail]
Link: http://lkml.kernel.org/r/[e-mail]
Signed-off-by: Ingo Molnar <[e-mail]>

arch/x86/crypto/sha-mb/sha1_mb_mgr_submit_avx2.S

···
 len             = %rax
 DWORD_len	= %eax
 
-lane            = %rbp
-tmp3            = %rbp
+lane            = %r12
+tmp3            = %r12
 
 tmp             = %r9
 DWORD_tmp	= %r9d
···
 # arg 2 : rdx : job
 ENTRY(sha1_mb_mgr_submit_avx2)
 	push	%rbx
-	push	%rbp
+	push	%r12
 
 	mov     _unused_lanes(state), unused_lanes
 	mov	unused_lanes, lane
···
 	movl    DWORD_tmp, _result_digest+1*16(job_rax)
 
 return:
-	pop	%rbp
+	pop	%r12
 	pop	%rbx
 	ret