incorrect tagging of the project memory #1125
Assignees
Labels
Comments
ivg
added a commit
to ivg/bap
that referenced
this issue
Jun 12, 2020
ivg
added a commit
to ivg/bap
that referenced
this issue
Jun 24, 2020
ivg
added a commit
to ivg/bap
that referenced
this issue
Jun 29, 2020
ivg
added a commit
to ivg/bap
that referenced
this issue
Jun 29, 2020
ivg
added a commit
to ivg/bap
that referenced
this issue
Jul 9, 2020
ivg
added a commit
to ivg/bap
that referenced
this issue
Jul 15, 2020
ivg
added a commit
that referenced
this issue
Jul 15, 2020
* creates program modules in the current package or in the user package if one is specified. Also, demistifies the program objects and documents explicitly how they are formed. * adds package and set_package functions to the knowledge interface * adds `?package` to corresponding function in the Tid module * adds the package parameter to the Project.creat function * fixes stub-resolver, it was adding bogus slots to the program class * properly qualifies the start and exit nodes of the Tid graph * uses the new package field in the disassemble command * interns contexts in the current package * adds the collator extension points * adds the collate command * adds the callgraph collators * adds scoped promises and proposals to knowledge * switches to scoped promises for arches * enables lazy processing of the projects in the collate command * improves the diagnostic message on non-monotonic update error * tweaks the callgraph collator output * introduces rules and their introspection A rule is a static description of a promise. This commit introduces a DSL from rule specification, as well as a documentation view on the rule for its introspection. * adds `bap list rules` command It will list all rules known to the system. * adds declaration for some of the existing rules except objdump, it will be added later in a new form * adds path and possible name to the core-theory this commit should be pushed down, before the changes with rules * adds more rule definitions and more rules when a project is created the knowledge about the filename path and architecture is reflected into the knowledge base. * declares BIL rules * a bulk rename of the package bap.std to bap This .std is totally superfluous here, we need only one package for bap. * moves logging facilities to Bap_main * adds more of_string* parsers to Bitvec. * rewrites objdump plugin so it no longer depends on bap.std * moves bap attributes (aka value tags) to bap namespace But keep uuids for typeid and serialization, otherwise we will break presistence. While `package:name` denotes the name of a property, `uuid:name` denotes its representation. * provides the filename using scoped promises * adds paths to symbolizers to prevent symbolizers registered for one file provide names for other files. * adds the Stream.Variadic documentation it is really hard to grok how to use it without documentation, even for an author :) * adds paths to rooters * updates the byteweight rooter * updates read-symbols to use the path * updates IDA service to provide rooter and symbolizer with path * drops debugging info * use the path as the package not the digest Rationale: in a multi-project analysis we will need to know the package to obtain the information relative to a project and getting the digest is a bit long road. Moreover, it is less readable. Concerning the possible conflicts from the different files with the same name, it is quite unlikely, since the caching mechanism separates knowledge bases by the digests and other systems that we or anyone else will develop will manage their own separation schemes. * adds path constraint to branchers also * updates the ida brancher to respect the paths * adds `bap list collators` command * code clean up, adds docs and the rest * fixes the symbolizer block parsing function Instead of giving the name to the entry address it was naming all addresses belonging to a function. * adds the missing base binary to the comparison * adds filters and custom output destination to callgraph collator also normalizes names and fixes a few bugs. * cleans up documentation * stops objdump from providing addresses from ambiguous names * temprorary reverts the fix in stub resolver see #1125 * makes radare2 plugin respect the paths * introduces code units A new class of Core Theory objects that represents compilation units, libaries, executables, and other aggregations of programs. The path property now belongs to the unit, not the the program itself, so we basically introduced an extra layer of indirection. One of the important properties of units is that they could be shifted in memory, which paves the road to the llvm-base fix. * passes through bias from llvm up to the knowledge base fixes (hot) objdump (we still need to know the modulus of the address), radare2 and other symbolizers will follow. * treat all old rooters and symbolizers as unbiased sources * adds more target-specific information in the llvm backend output and did some refactoring * adds bits and endianessness to the target properties * adds a rule that provides bap:arch from unit's target defs * improves signature mismatch error in the OGRE parser * fixes abi printer * adds the newly added fields to the image specification * enables propagation of spec into KB and arch into spec * initializes the unit for the low-level disassembler interface * enables biased information sources We assumed by default that all our information sources (rooters, symbolizers, and branchers) are unbiased but it wasn't true for the sources that we created of images (using corresponding of_image) that were already operating using the biased information from the loader. To fix this issue we added a hidden parameter to mark information sources as biased or unbiased and perform bias substraction (and also addition in case when the destinations are provided) based on this parameter. We may later make it public, but so far it is only set for information sources that are derived from images. Also the common code between information sources were factored out into the Bap_disasm_source module. * fixes the hardcoded modulus in the objdump plugin We are now able to query the bitness without having to pull in the Bap.Std interface so we can implement everything neatly. * scopes the information sources obtained from the image * enables scopes promises for information sources and uses it for the rec, as far as I can grep, I don't see any information sources that are not properly scoped, either by limiting them to the path or to a function. * updates the project documentation, documents the toplevel interface also exports the full toplevel interface * documents and tweaks some function in the unit module The [for_file] function now also sets the path (the same as the corresponding [for_addr] function sets the address). The [for_region] now interns the boundaries in the current package and builds the finaly symbol from their concatenation to enable intersecting regions from different files in the same knowledge base. * flushes the formatter in the callgraph collator. * documents the Bap_main_event module also establishes equalities between it and its re-export in Bap.Std and adds an convenience alias in the Bap_main module for the loggers. * adds the compare command tests
ivg
added a commit
to ivg/bap
that referenced
this issue
Nov 2, 2020
Introduces memory (Memmap.t) to the knowledge base accessible via the `Project.memory_slot` and uses it for computing stubs in the stub-resolver plugin. Also fixes BinaryAnalysisPlatform#1125.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
We have a case with an even number of bugs. The first bug, is in the stub-resolver plugin which searches for the sections named
.pltin the data segments (while.pltand.plt.gotare clearly executable, moreover, we're using these addresses to classify disassembled functions as stubs or not). Another bug is that the.pltentry is not present in the providedcodesegments and indeed we have to usedatahere otherwise the component won't work. The second bug has to be fixed and once it is the first one could also be fixed.The text was updated successfully, but these errors were encountered: