crypto algoritms failing?

crypto algoritms failing?

[Jasper Spaans]

[-- Attachment #1: Type: text/plain, Size: 1200 bytes --]

Hi List,

When booting I see this in dmesg:

testing tea ECB encryption 
test 1 (128 bit key):
0a3aea4140a9ba94
fail
test 2 (128 bit key):
775d2a6af6ce9209
fail
test 3 (128 bit key):
be7abb81952d1f1edd89a1250421df95
fail
test 4 (128 bit key):
e04d5d3cb78c364794189591a9fc49f844d12dc299b8082a078973c24592c690
fail
[..]
testing xtea ECB encryption 
test 1 (128 bit key):
aa2296e56c61f345
fail
test 2 (128 bit key):
823eeb35dcddd9c3
fail
test 3 (128 bit key):
e204dbf289859eea6135aaedb5cb712c
fail
test 4 (128 bit key):
0b03cd8abe95fdb1c144910ba5c91bb4a9da1e9eb13e2a8feaa56a85d1f4a8a5
fail

CPU in that machine is an athlon xp, cpu flags according to /proc/cpuinfo
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 mmx fxsr sse pni syscall mmxext 3dnowext 3dnow

Compiler: gcc 3.3.5 (debian package 1:3.3.5-6)

Is this supposed to happen?


Jasper
-- 
Jasper Spaans                                       http://jsp.vs19.net/
 00:24:05 up 10207 days, 16:11, 0 users, load average: 6.29 6.03 6.13
         There already is an object oriented version of COBOL.
             It's called "ADD ONE TO COBOL GIVING COBOL."

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: crypto algoritms failing?

[James Morris]

On Fri, 28 Jan 2005, Jasper Spaans wrote:

> Is this supposed to happen?

No.  What is your kernel version?


- James
-- 
James Morris
<jmorris@redhat.com>

Re: crypto algoritms failing?

[Jasper Spaans]

[-- Attachment #1: Type: text/plain, Size: 767 bytes --]

On Thu, Jan 27, 2005 at 07:38:43PM -0500, James Morris wrote:
> > Is this supposed to happen?
> 
> No.  What is your kernel version?

Current bitkeeper + latest swsusp2 patches and hostap driver, however, those
two don't come near touching the crypto stuff[1] so they're not really on my
suspect shortlist, but I'll see if I can find time to build a vanilla one
tomorrow (that is, without swsusp/hostap).. right now, it's time to sleep in
my local timezone..


Groet,
-- 
Jasper Spaans                                       http://jsp.vs19.net/
 01:40:46 up 10207 days, 17:27, 0 users, load average: 6.00 6.00 6.12

[1] hostap however does use some crypto algoritms, if I'm not mistaken, but
    its modules are not loaded in that stage of booting

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: crypto algoritms failing?

[James Morris]

On Fri, 28 Jan 2005, Jasper Spaans wrote:

> On Thu, Jan 27, 2005 at 07:38:43PM -0500, James Morris wrote:
> > > Is this supposed to happen?
> > 
> > No.  What is your kernel version?
> 
> Current bitkeeper + latest swsusp2 patches and hostap driver, however, those
> two don't come near touching the crypto stuff[1] so they're not really on my
> suspect shortlist, but I'll see if I can find time to build a vanilla one
> tomorrow (that is, without swsusp/hostap).. right now, it's time to sleep in
> my local timezone..

Looks like a cleanup broke the test vectors:
http://linux.bkbits.net:8080/linux-2.5/gnupatch@41ad5cd9EXGuUhmmotTFBIZdIkTm0A

Patch below, please apply.

Signed-off-by: James Morris <jmorris@redhat.com>


---

diff -purN -X dontdiff linux-2.6.11-rc1-mm1.o/crypto/tcrypt.h linux-2.6.11-rc1-mm1.w/crypto/tcrypt.h
--- linux-2.6.11-rc1-mm1.o/crypto/tcrypt.h	2005-01-19 09:30:32.000000000 -0500
+++ linux-2.6.11-rc1-mm1.w/crypto/tcrypt.h	2005-01-27 20:28:23.312918312 -0500
@@ -1986,7 +1986,7 @@ static struct cipher_testvec arc4_dec_tv
 #define TEA_ENC_TEST_VECTORS	4
 #define TEA_DEC_TEST_VECTORS	4
 
-static struct cipher_testvec xtea_enc_tv_template[] =
+static struct cipher_testvec tea_enc_tv_template[] =
 {
 	{
 		.key    = { [0 ... 15] = 0x00 },
@@ -2080,7 +2080,7 @@ static struct cipher_testvec tea_dec_tv_
 #define XTEA_ENC_TEST_VECTORS	4
 #define XTEA_DEC_TEST_VECTORS	4
 
-static struct cipher_testvec tea_enc_tv_template[] =
+static struct cipher_testvec xtea_enc_tv_template[] =
 {
 	{
 		.key    = { [0 ... 15] = 0x00 },

Re: crypto algoritms failing?

[Jasper Spaans]

[-- Attachment #1: Type: text/plain, Size: 407 bytes --]

On Thu, Jan 27, 2005 at 08:43:18PM -0500, James Morris wrote:

> Looks like a cleanup broke the test vectors:
> http://linux.bkbits.net:8080/linux-2.5/gnupatch@41ad5cd9EXGuUhmmotTFBIZdIkTm0A
> 
> Patch below, please apply.

That fixes it, thanks.

-- 
Jasper Spaans                                       http://jsp.vs19.net/
 10:13:13 up 10208 days,  2:00, 0 users, load average: 6.00 6.00 6.12

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: crypto algoritms failing?

[Adrian Bunk]

On Thu, Jan 27, 2005 at 08:43:18PM -0500, James Morris wrote:
> On Fri, 28 Jan 2005, Jasper Spaans wrote:
> 
> > On Thu, Jan 27, 2005 at 07:38:43PM -0500, James Morris wrote:
> > > > Is this supposed to happen?
> > > 
> > > No.  What is your kernel version?
> > 
> > Current bitkeeper + latest swsusp2 patches and hostap driver, however, those
> > two don't come near touching the crypto stuff[1] so they're not really on my
> > suspect shortlist, but I'll see if I can find time to build a vanilla one
> > tomorrow (that is, without swsusp/hostap).. right now, it's time to sleep in
> > my local timezone..
> 
> Looks like a cleanup broke the test vectors:
> http://linux.bkbits.net:8080/linux-2.5/gnupatch@41ad5cd9EXGuUhmmotTFBIZdIkTm0A
> 
> Patch below, please apply.
>...

Ops, yes, sorry.
Where are the broen paperbags?

Signed-off-by: Adrian Bunk <bunk@stusta.de>

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: crypto algoritms failing?

[Nigel Cunningham]

Hi.

You normally test cryptoapi functionality while booting?

Anyway, I can confirm that if suspend2 touches anything remotely related
to this, it's unintentional and I'll fix it :>

Nigel

On Fri, 2005-01-28 at 10:30, Jasper Spaans wrote:
> Hi List,
> 
> When booting I see this in dmesg:
> 
> testing tea ECB encryption 
> test 1 (128 bit key):
> 0a3aea4140a9ba94
> fail
> test 2 (128 bit key):
> 775d2a6af6ce9209
> fail
> test 3 (128 bit key):
> be7abb81952d1f1edd89a1250421df95
> fail
> test 4 (128 bit key):
> e04d5d3cb78c364794189591a9fc49f844d12dc299b8082a078973c24592c690
> fail
> [..]
> testing xtea ECB encryption 
> test 1 (128 bit key):
> aa2296e56c61f345
> fail
> test 2 (128 bit key):
> 823eeb35dcddd9c3
> fail
> test 3 (128 bit key):
> e204dbf289859eea6135aaedb5cb712c
> fail
> test 4 (128 bit key):
> 0b03cd8abe95fdb1c144910ba5c91bb4a9da1e9eb13e2a8feaa56a85d1f4a8a5
> fail
> 
> CPU in that machine is an athlon xp, cpu flags according to /proc/cpuinfo
> flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
> cmov pat pse36 mmx fxsr sse pni syscall mmxext 3dnowext 3dnow
> 
> Compiler: gcc 3.3.5 (debian package 1:3.3.5-6)
> 
> Is this supposed to happen?
> 
> 
> Jasper
-- 
Nigel Cunningham
Software Engineer
Cyclades Corporation

http://cyclades.com

Re: crypto algoritms failing?

[James Morris]

On Fri, 28 Jan 2005, Nigel Cunningham wrote:

> You normally test cryptoapi functionality while booting?

This happens if you link tcrypt statically into the kernel.


- James
-- 
James Morris
<jmorris@redhat.com>

Re: crypto algoritms failing?

[Nigel Cunningham]

Hi.

On Fri, 2005-01-28 at 13:25, James Morris wrote:
> On Fri, 28 Jan 2005, Nigel Cunningham wrote:
> 
> > You normally test cryptoapi functionality while booting?
> 
> This happens if you link tcrypt statically into the kernel.

Yes, but why would you? Oh well. Doesn't matter ;>

Nigel
-- 
Nigel Cunningham
Software Engineer
Cyclades Corporation

http://cyclades.com