My bugs for 0.09b41 -- The AMD side

[mastercracker]

BUgs for v0.09b41 with catalyst 12.6 on HD5750 GPU under Windows 7 64 bit.

I decided to put it in a different thread since it's a different bug pattern than CUDA cards and it might be followed by different people as well. For reference on passwords, attacks done, etc refer to the first topic at: http://hashcat.net/forum/thread-1430.html

- Status bar at the bottom not always appearing (really intermittent).
- The following algos where making the driver crash: 2100, 1600, 400 and 500. Heavy algos basically. I tried also with accel and loops at 1 and it still crashed. These are the lines it crashes on, the others work:
oclhashcat-plus64.exe -a 3 -d 1 -m 2100 -o found.txt -1 ?s?d?u?l --bf-min 4 --bf-max 4 hash.txt ?u?d?s?d
oclhashcat-plus64.exe -a 3 -d 1 -m 2100 -o found.txt -1 abcd --bf-min 8 --bf-max 8 hash.txt ?1?1?1?1?1?1?1?1
oclhashcat-plus64.exe -a 6 -d 1 -m 2100 -o found.txt hash.txt dic1.txt ?u?l?l
oclhashcat-plus64.exe -a 7 -d 1 -m 2100 -o found.txt hash.txt ?u?l?l dic1.txt
oclhashcat-plus64.exe -a 7 -d 1 -m 2100 -o found.txt hash.txt ?l?l?l?l dic1.txt
oclhashcat-plus64.exe -a 7 -d 1 -m 2100 -o found.txt hash.txt ?d?d?d?d dic1.txt

In summary, it crashes on bruteforce length 4 and 8, on length 15 for -a 6 and on length 8, 12, 15 on -a 7.
Additional notes: I have noticed that for those that crashes, the rejected shows 100%.


- Algos where nothing works: 2811, 11, 2711
- For -m 900 MD4, bruteforce 2 to 12 characters does not work but length 15 works.
- For -m 2600 DoubleMD5, -a 0 dictionary and -a 3 bruteforce are not working. In Combinatory and hybrid with mask are working perfectly well.
- For -m 0 MD5, bruteforce on length 2,4 and 8 are not working while length 12 and 15 are working fine.
- For -m 1000 NTLM, bruteforce on length 2,4 and 8 are not working while length 12 and 15 are working fine.
- For -m 1400 SHA256, weird pattern. For -a 3, length 2 and 8 are not working. For -a 6, length 4 is not working. For -a 7 length 4 and 12 are not working.
- For -m 121 SMF, bruteforce mode is not working.
- For -m 2611 Vbulletin < 3.8.5, -a 0 dictionary and -a 3 bruteforce are not working. In Combinatory and hybrid with mask are working perfectly well.

- Algo tested that works 100%: 1100, 300, 21, 101, 100, 1700.

[atom]

did you remove all *.kernel files from kernels\4098\ before you tested?

[K9]

Is there any easy way to test all possible attack and hash modes? A script that outputs an error report would be handy. I think 2811, 2611, 2711 are also not working for me. I just noticed the new sha512 also doesn't work for me.

[mastercracker]

did you remove all *.kernel files from kernels\4098\ before you tested?

No. Are we supposed to do that? I just decompressed the archive after download and put my test file in. I have noticed that it was saying "building up kernel" or something like that but I did not pay attention to when it did it.

[proinside]

mastercracker, did you decompressed all the files over an already existing folder
or you did it to a new folder (whatever named you might give to it) ?
I guess, if you did it to a new folder, there's no need to delete all kernel files in
you older/previous version which are located in a diferent folder name.

[atom]

did you remove all *.kernel files from kernels\4098\ before you tested?

No. Are we supposed to do that? I just decompressed the archive after download and put my test file in. I have noticed that it was saying "building up kernel" or something like that but I did not pay attention to when it did it.

Yes. Since v0.09 kernels are cached (the .kernel files). So as a beta tester (and only as a beta-tester) you have to either unpack the beta into a clean/new directory or to remove the .kernel files from kernels/4098/

[atom]

There is a new BETA version up, please retry everything. I did several changes. The BUG with the PROMPT still exists, I will fix this later

[mastercracker]

There is a new BETA version up, please retry everything. I did several changes. The BUG with the PROMPT still exists, I will fix this later

Ok. I know that we are already 2 versions ahead but since I have made the test on v42 and to note the progress, the 2 following post will be: 1) the command lines where the driver crash with a Windows error saying: "Display driver stopped responding and has recovered.". In this case, oclhashcat-plus command window stays open like it was cracking but there are no progress even if GPU load says 99%. In post 2) I will write differences in bugs already reported + test on all the other algos except WPA.

---

NOTE: I have changed some of the command line I was previously using to make them faster and doable for high iteration algorithm. Here are the ones where the crash happens:

oclhashcat-plus64.exe -a 3 -d 1 -m 400 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt ?u?d?s?d
oclhashcat-plus64.exe -a 3 -d 1 -m 500 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt ?u?d?s?d
oclhashcat-plus64.exe -a 3 -d 1 -m 1600 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt ?u?d?s?d
oclhashcat-plus64.exe -a 3 -d 1 -m 1800 -o found.txt --bf-min 2 --bf-max 2 hash.txt ?s?s
oclhashcat-plus64.exe -a 3 -d 1 -m 1800 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt ?u?d?s?d
oclhashcat-plus64.exe -a 6 -d 1 -m 1800 -o found.txt hash.txt dic1.txt ?l?l
oclhashcat-plus64.exe -a 6 -d 1 -m 1800 -o found.txt hash.txt dic1.txt A?l?l
oclhashcat-plus64.exe -a 7 -d 1 -m 1800 -o found.txt hash.txt ?s?s dic1.txt
oclhashcat-plus64.exe -a 7 -d 1 -m 1800 -o found.txt hash.txt B?l?l dic1.txt
oclhashcat-plus64.exe -a 7 -d 1 -m 1800 -o found.txt -1 1234 hash.txt ?1?1?1?1 dic1.txt
oclhashcat-plus64.exe -a 3 -d 1 -m 2100 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt ?u?d?s?d
oclhashcat-plus64.exe -a 3 -d 1 -m 2600 -o found.txt --bf-min 15 --bf-max 15 -1 a -2 b -3 c -4 d hash.txt ?1?1?1?1?2?2?2?2?3?3?3?3?4?4?4
oclhashcat-plus64.exe -a 3 -d 1 -m 2611 -o found.txt --bf-min 15 --bf-max 15 -1 a -2 b -3 c -4 d hash.txt ?1?1?1?1?2?2?2?2?3?3?3?3?4?4?4

[mastercracker]

Post 2:

New

- For -m 0 MD5, bruteforce of lenght 2,4 and 8 failed.
- For -m 120 sha1($salt.$pass), bruteforce (-a 3) fails on all length.
- For -m 121 SMF > v1.1, bruteforce (-a 3) fails on all length.
- For -m 122 OSX v10.4, v10.5, v10.6, bruteforce (-a 3) fails on all length.
- FOr -m 130 sha1(unicode($pass).$salt), bruteforce of length 2 failed, the rest all worked.
- For -m 131 MSSQL(2000), bruteforce of length 4 and -a 7 of length 2 and 4 worked. All the rest failed.
- For -m 132 MSSQL(2005), bruteforce of length 2 failed.
- For -m 900 MD4, bruteforce all failed except for length 15. All the other attack mode were fine.
- For -m 1000 NTLM, bruteforce all failed except for length 12 and 15. All the other attack mode were fine.
- For -m 1400 SHA256, -a 3 length 2, -a 6 length 4 and -a 7 length 4 failed. The rest all worked.
- For -m 1410 sha256($pass.$salt), -a 3 length 2, -a 6 length 4 and -a 7 length 4 failed. The rest all worked.
- For -m 1420 sha256($salt.$pass), -a 3 length 2, -a 6 length 4 and -a 7 length 4 failed. The rest all worked.
- For -m 1710 sha512($pass.$salt), bruteforce on length 2 and 4 failed. The rest worked.
- For -m 1800 sha512crypt, SHA512(Unix), similar to Nvidia bug. -a 1 length 15 failed. -a 3 length 2 and 4 crashed. -a 6 length 8, 12 and 15 crashed. -a 7 length 4, 8 and 15 crashed.
- For -m 2600 Double MD5, dictionary and bruteforce mode fails. The rest works.
- For -m 2611 Double MD5, dictionary and bruteforce mode fails. The rest works.

Shared with Nvidia
- For -m 10 md5($pass.$salt), bruteforce of length 2 failed.
- For -m 30 md5(unicode($pass).$salt), Length 15 does not work on all modes and length 2 does not work on bruteforce.
- For -m 110 sha1($pass.$salt), Bruteforce of length 2 is not working.
- For -m 111 nsldaps, SSHA-1(Base64), Netscape LDAP SSHA, Dictionary attack works (-a 0) but all the other attack mode fails.
- For -m 112 Oracle 11g, Bruteforce of length 2 is not working. The rest is perfect.
- For -m 141 EPiServer 6.x, Nothing worked.
- For -m 2100 DCC2, Dictionary mode for length 2 failed. Bruteforce for length 4 failed. All the rest worked.

[mastercracker]

Ok. I am also doing things to pinpoint the source of the crash on AMD. It seems to somehow be related to the mask or the amount of memory used by the mask. The password to crack was P0$7:

oclhashcat-plus64.exe -a 3 -d 1 -m 400 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt ?u?d?s?d

This attack crashes on all heavy algos.

oclhashcat-plus64.exe -a 3 -d 1 -m 400 -o found.txt --bf-min 4 --bf-max 4 -1 ?s?d?u?l hash.txt P0?s?d

This attack runs without a problem.

Edit: mask ?u0?s?d and P?d?s?d crash. mask P?d$?d does not crash. My HD5750 has 1 GB memory.