[LAU] further emagic woes

Clemens Ladisch clemens at ladisch.de
Wed Nov 18 02:56:23 EST 2009

David Woodhouse wrote:
> On Tue, 2009-11-17 at 09:01 +0100, Clemens Ladisch wrote:
> > David, do you have any clue what could be going on? 
> Not really

The code bytes disassemble to the following:

f80d2460: 31 db                	xor    %ebx,%ebx
f80d2462: 8b 4d e8             	mov    -0x18(%ebp),%ecx
f80d2465: 8b 55 e4             	mov    -0x1c(%ebp),%edx
f80d2468: 8b 45 e0             	mov    -0x20(%ebp),%eax
f80d246b: c7 44 24 04 a5 00 00 	movl   $0xa5,0x4(%esp)
f80d2472: 00 
f80d2473: 89 34 24             	mov    %esi,(%esp)
f80d2476: e8 94 fb ff ff       	call   f80d208a
f80d247b: 85 c0                	test   %eax,%eax
f80d247d: 0f 88 21 03 00 00    	js     f80d2827
f80d2483: 85 f6                	test   %esi,%esi
f80d2485: 0f 84 9c 00 00 00    	je     f80d25b2
f80d248b: 8b 0b                	mov    (%ebx),%ecx   <-- crash here
f80d248d: 31 f6                	xor    %esi,%esi
f80d248f: 0f c9                	bswap  %ecx
f80d2491: 85 db                	test   %ebx,%ebx
f80d2493: 89 4d e4             	mov    %ecx,-0x1c(%ebp)
f80d2496: 74 ca                	je     f80d2462
f80d2498: 0f b7 43 04          	movzwl 0x4(%ebx),%eax
f80d249c: 0f b7 d0             	movzwl %ax,%edx

The offending command is obviously line 157 of emi62.c.

> this driver was using the same type of 'hex records' before we
> touched it.

rec being NULL is how the ihex helper functions signal the end of the
firmware data; the old driver just had a harmless empty record.

emi26.c uses just "rec" as outer loop condition; that ought to work.


More information about the Linux-audio-user mailing list