Ssa transform #713

serpilliere · 2018-04-11T08:39:01Z

No description provided.

serpilliere · 2018-04-11T08:44:08Z

Hi!
@mrphrazer, news for the SSA: It's really an key feature and we really want to include it in Miasm. As your PR is old (2 year ago!) I reworked it and rebased it on the current master. Old comments from the original PR have been taken in account as well.

We have used it (for example in constant propagation), and it really gives instesting results 😄
As I modified your commit, but keep your name in the commit, tell me if you are ok with this, because we are burning to release the next parts 😄

Thanks to you and Niko Schmidt for the job!

mrphrazer · 2018-04-11T09:15:25Z

Hi!

We are happy to hear this :) Go for it!

serpilliere · 2018-06-15T15:56:13Z

PR seems ready to me

serpilliere · 2018-07-05T15:58:27Z

Rebase done

commial · 2018-07-06T12:22:21Z

miasm2/analysis/ssa.py

+    Handling of
+    - variable generation
+    - variable renaming
+    - conversion of an DiGraphIR block into SSA


It's now an "IRCFG" ;)

commial · 2018-07-06T12:23:24Z

miasm2/analysis/ssa.py

+    def __init__(self, ircfg):
+        """
+        Initialises generic class for SSA
+        :param ircfg: instance of DiGrpahIR


Same remark here

commial · 2018-07-06T12:25:38Z

miasm2/analysis/ssa.py

+        self.ssa_to_location = {}
+
+        # Don't SSA architecture's instruction pointer
+        self.immutable_ids = set([self.ircfg.IRDst])


From what I understand, the architecture's instruction pointer (PC / RIP / ...) will be SSA. Is it wanted? If so, the comment is maybe misleading

I reflected the comment

commial · 2018-07-06T12:26:38Z

miasm2/analysis/ssa.py

+        # Don't SSA architecture's instruction pointer
+        self.immutable_ids = set([self.ircfg.IRDst])
+
+    @classmethod


This could even be a staticmethod

I removed this code and used native expr function

commial · 2018-07-06T12:32:23Z

miasm2/analysis/ssa.py

+        return expr
+
+
+class SSAPath(SSABlock):


Why is SSAPath inheriting from SSABlock and not from SSA?

Because you inherit reassemble_expr. Understanding a path as a sequence of block, this makes sense. A similar function on CFG-level would work differently since it has to consider all the different branches/phi nodes.

commial · 2018-07-06T12:34:37Z

Thanks for the PR and for the work on re-integration to the current codebase!
Apart the few tiny comments, the patch is OK for me

serpilliere · 2018-07-09T07:20:24Z

miasm2/analysis/ssa.py

+
+    def __init__(self, ircfg):
+        """
+        Initialises SSA class for directed acyclic graphs


A last question for @mrphrazer
The acyclic here seems strange to me here.
I think it's possible to have cylces in even in ssa form, passing through a phi operator.
Did I miss something here?

You are right, the implemention works as it is also for cycles. The comment is a relic from a project where I had to enforce acylcic graphs via loop unrolling.

Joint work with Niko Schmidt

commial · 2018-07-09T11:01:10Z

Thanks again @mrphrazer for this contribution, and @serpilliere for refactoring it. We were using it internally for some time now, and it's quite a nice tool!

Ssa transform

serpilliere added the WIP label Apr 11, 2018

serpilliere force-pushed the ssa_transform branch from 2482b3f to 328c6f1 Compare April 11, 2018 08:51

serpilliere force-pushed the ssa_transform branch from 328c6f1 to b79ef9f Compare April 15, 2018 20:15

serpilliere force-pushed the ssa_transform branch from b79ef9f to ab6f9fc Compare May 15, 2018 08:45

serpilliere force-pushed the ssa_transform branch from ab6f9fc to be76045 Compare June 12, 2018 18:00

serpilliere removed the WIP label Jun 12, 2018

serpilliere force-pushed the ssa_transform branch from be76045 to 5c708f8 Compare July 5, 2018 15:58

commial reviewed Jul 6, 2018

View reviewed changes

serpilliere force-pushed the ssa_transform branch 2 times, most recently from 4f80bfd to afb5d04 Compare July 9, 2018 07:16

serpilliere commented Jul 9, 2018

View reviewed changes

mrphrazer and others added 2 commits July 9, 2018 12:31

Analysis: Add SSA transformation

bb40c1a

Joint work with Niko Schmidt

SSA: add tests

953676d

serpilliere force-pushed the ssa_transform branch from afb5d04 to 953676d Compare July 9, 2018 10:34

commial merged commit 17d48de into cea-sec:master Jul 9, 2018

serpilliere deleted the ssa_transform branch July 9, 2018 11:02

w4kfu pushed a commit to w4kfu/miasm that referenced this pull request Jun 16, 2020

Merge pull request cea-sec#713 from serpilliere/ssa_transform

60ff3db

Ssa transform

Ssa transform #713

Ssa transform #713

Uh oh!

Conversation

serpilliere commented Apr 11, 2018

Uh oh!

serpilliere commented Apr 11, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mrphrazer commented Apr 11, 2018

Uh oh!

serpilliere commented Jun 15, 2018

Uh oh!

serpilliere commented Jul 5, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

commial commented Jul 6, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

commial commented Jul 9, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

serpilliere commented Apr 11, 2018 •

edited

Loading