delta/python-packages/pycrypto.git - github.com: dlitz/pycrypto.git

diff options


context:
space:
mode:

Diffstat (limited to 'pipermail/pycrypto/2013q4/000713.html')

-rw-r--r--

pipermail/pycrypto/2013q4/000713.html

176

1 files changed, 176 insertions, 0 deletions

diff --git a/pipermail/pycrypto/2013q4/000713.html b/pipermail/pycrypto/2013q4/000713.html
new file mode 100644
index 0000000..21dff87
--- /dev/null
+++ b/pipermail/pycrypto/2013q4/000713.html

@@ -0,0 +1,176 @@

+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">

+<HTML>

+ <HEAD>

+ <TITLE> [pycrypto] Crypto.Random crashes due to unaligned access

+ </TITLE>

+ <LINK REL="Index" HREF="index.html" >

+ <LINK REL="made" HREF="mailto:pycrypto%40lists.dlitz.net?Subject=Re%3A%20%5Bpycrypto%5D%20Crypto.Random%20crashes%20due%20to%20unaligned%20access&In-Reply-To=%3CCAFCC3ev7TbeY_JzdQ6zdgd5eN8MTPDPqdbxZCMvpGEKN-BhP%3Dg%40mail.gmail.com%3E">

+ <META NAME="robots" CONTENT="index,nofollow">

+ <style type="text/css">

+ pre {

+ white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */

+ }

+ </style>

+ <META http-equiv="Content-Type" content="text/html; charset=us-ascii">

+ <LINK REL="Previous" HREF="000716.html">

+ <LINK REL="Next" HREF="000717.html">

+ </HEAD>

+ <BODY BGCOLOR="#ffffff">

+ <H1>[pycrypto] Crypto.Random crashes due to unaligned access</H1>

+ Greg Price

+ <A HREF="mailto:pycrypto%40lists.dlitz.net?Subject=Re%3A%20%5Bpycrypto%5D%20Crypto.Random%20crashes%20due%20to%20unaligned%20access&In-Reply-To=%3CCAFCC3ev7TbeY_JzdQ6zdgd5eN8MTPDPqdbxZCMvpGEKN-BhP%3Dg%40mail.gmail.com%3E"

+ TITLE="[pycrypto] Crypto.Random crashes due to unaligned access">gnprice at gmail.com

+ </A>

+ Sun Oct 27 16:00:12 PDT 2013

+ <UL>

+ <LI>Previous message: <A HREF="000716.html">[pycrypto] Crypto.Random crashes due to unaligned access

+</A></li>

+ <LI>Next message: <A HREF="000717.html">[pycrypto] Need your input: Major modernization; dropping legacy Python support?

+</A></li>

+ <LI> Messages sorted by:

+ <a href="date.html#713">[ date ]</a>

+ <a href="thread.html#713">[ thread ]</a>

+ <a href="subject.html#713">[ subject ]</a>

+ <a href="author.html#713">[ author ]</a>

+ </LI>

+ </UL>

+ <HR>

+

+<PRE>And the other message that didn't make it before.

+Greg

+---------- Forwarded message ----------

+From: Greg Price <<A HREF="http://lists.dlitz.net/cgi-bin/mailman/listinfo/pycrypto">gnprice at gmail.com</A>>

+Date: Thu, Oct 24, 2013 at 12:53 PM

+Subject: Re: Crypto.Random crashes due to unaligned access

+To: Dwayne Litzenberger <<A HREF="http://lists.dlitz.net/cgi-bin/mailman/listinfo/pycrypto">dlitz at dlitz.net</A>>

+Cc: <A HREF="http://lists.dlitz.net/cgi-bin/mailman/listinfo/pycrypto">pycrypto at lists.dlitz.net</A>

+GCC 4.6.3, and same with 4.8.1 (see my later message). "setup.py

+test" crashes too.

+See my follow-up last night for more of a diagnosis -- the problem is

+that 'rk', aka st->ek, has a type that implies 16-byte alignment, and

+we don't get that from PyObject_New. That allocation comes (I think)

+straight from malloc(), which with glibc on a 32-bit x86 system gives

+only 8-byte alignment.

+One fix would be to move st->ek and st->dk to a separate buffer we

+allocate with something like posix_memalign(). This wouldn't require

+any copying, as we fill those buffers ourselves in the first place.

+It'd just add a layer of indirection in the AESNI.c block_state

+struct.

+Alternatively we could try to eliminate the 16-byte alignment, but I

+don't immediately see a way of doing that without making the code much

+messier, and it'd probably also slow the code down.

+Greg

+On Thu, Oct 24, 2013 at 9:59 AM, Dwayne Litzenberger <<A HREF="http://lists.dlitz.net/cgi-bin/mailman/listinfo/pycrypto">dlitz at dlitz.net</A>> wrote:

+> Hi Greg!

+>

+> What version/build of GCC is this? Does "setup.py test" crash for you as well?

+>

+> I'd rather figure out how to fix the problem than to start making copies of the key.

+>

+> Greg Price <<A HREF="http://lists.dlitz.net/cgi-bin/mailman/listinfo/pycrypto">gnprice at gmail.com</A>> wrote:

+>>I get the following crash in a PyCrypto built from the current master,

+>>af058ee (aka v2.6.1-136-gaf058ee):

+>>

+>>>>> import Crypto.Random

+>>>>> Crypto.Random.new().read(1)

+>>Segmentation fault (core dumped)

+>>

+>>This is on i686. I compiled with GCC 4.6.3 (or "Ubuntu/Linaro

+>>4.6.3-1ubuntu5".)

+>>

+>>GDB shows the crash is here:

+>>

+>>Program received signal SIGSEGV, Segmentation fault.

+>>aes_key_setup_enc (keylen=32, cipherKey=

+>>0x841b1bc

+>>"L\fB2\244\225\235\206^\242\305\305b\201\200\335ņ{d\240\343\262;m\361\243\276u~\337&",

+>>rk=

+>> 0x84900a8) at src/AESNI.c:122

+>>122 rk[0] = _mm_loadu_si128((const __m128i*) cipherKey);

+>>

+>>at which the instruction is

+>>

+>>(gdb) x/i $pc

+>>=> 0xb78f2600 <ALGnew+2160>: movdqa %xmm0,0x40(%esi)

+>>

+>>This is an aligned store. The documentation of MOVDQA says it should

+>>be 16-byte aligned. The value of rk (aka %esi + 0x40) is only 8-byte

+>>aligned:

+>>

+>>(gdb) p rk

+>>$5 = (__m128i *) 0x84900a8

+>>(gdb) p/x $esi

+>>$9 = 0x8490068

+>>

+>>It's not clear to me why GCC generated an aligned instruction here --

+>>in fact, the definition of _mm_loadu_si128 in my emmintrin.h appears

+>>to be

+>>

+>>extern __inline __m128i __attribute__((__gnu_inline__,

+>>__always_inline__, __artificial__))

+>>_mm_loadu_si128 (__m128i const *__P)

+>>{

+>> return (__m128i) __builtin_ia32_loaddqu ((char const *)__P);

+>>}

+>>

+>>and the name of that builtin sure sounds more like MOVDQU than MOVDQA.

+>> Perhaps GCC somehow decides that it can prove the pointer is aligned

+>>here.

+>>

+>>I don't know why GCC makes this mistake, or (since it's never the

+>>compiler's fault) which code is lying to it about something being

+>>aligned. Anyone know how to investigate this kind of question?

+>>

+>>A workaround would be to make sure that the cipherKey argument to

+>>aes_key_setup_enc() in src/AESNI.c is always 16-byte aligned. At

+>>present, that argument comes straight from the first Python-level

+>>argument to _AESNI.new(); see the PyArg_ParseTupleAndKeywords() call

+>>in src/block_template.c. I guess to implement this workaround we'd

+>>copy the key to a new, aligned buffer if it's not aligned.

+>>

+>>I can send a patch for that workaround if it seems like the best

+>>approach. Happy to hear alternatives, and of course it'd be most

+>>satisfying if we can understand why the compiler is emitting this

+>>output in the first place.

+>>

+>>Greg

+>

+> --

+> Sent from my Android device with K-9 Mail. Please excuse my brevity.

+</PRE>

+

+ <HR>

+ <UL>

+

+ <LI>Previous message: <A HREF="000716.html">[pycrypto] Crypto.Random crashes due to unaligned access

+</A></li>

+ <LI>Next message: <A HREF="000717.html">[pycrypto] Need your input: Major modernization; dropping legacy Python support?

+</A></li>

+ <LI> Messages sorted by:

+ <a href="date.html#713">[ date ]</a>

+ <a href="thread.html#713">[ thread ]</a>

+ <a href="subject.html#713">[ subject ]</a>

+ <a href="author.html#713">[ author ]</a>

+ </LI>

+ </UL>

+<hr>

+<a href="http://lists.dlitz.net/cgi-bin/mailman/listinfo/pycrypto">More information about the pycrypto

+mailing list</a>

+</body></html>