postgres_exporter/tools/vendor/github.com/stripe/safesql
Will Rouesnel e2b6c973a1 Add self-contained gometalinter build tooling. 2017-06-07 00:53:19 +10:00
..
LICENSE Add self-contained gometalinter build tooling. 2017-06-07 00:53:19 +10:00
README.md Add self-contained gometalinter build tooling. 2017-06-07 00:53:19 +10:00
package15.go Add self-contained gometalinter build tooling. 2017-06-07 00:53:19 +10:00
package16.go Add self-contained gometalinter build tooling. 2017-06-07 00:53:19 +10:00
safesql.go Add self-contained gometalinter build tooling. 2017-06-07 00:53:19 +10:00

README.md

SafeSQL

SafeSQL is a static analysis tool for Go that protects against SQL injections.

Usage

$ go get github.com/stripe/safesql

$ safesql
Usage: safesql [-q] [-v] package1 [package2 ...]
  -q=false: Only print on failure
  -v=false: Verbose mode

$ safesql example.com/an/unsafe/package
Found 1 potentially unsafe SQL statements:
- /Users/alice/go/src/example.com/an/unsafe/package/db.go:14:19
Please ensure that all SQL queries you use are compile-time constants.
You should always use parameterized queries or prepared statements
instead of building queries from strings.

$ safesql example.com/a/safe/package
You're safe from SQL injection! Yay \o/

How does it work?

SafeSQL uses the static analysis utilities in go/tools to search for all call sites of each of the query functions in package database/sql (i.e., functions which accept a string parameter named query). It then makes sure that every such call site uses a query that is a compile-time constant.

The principle behind SafeSQL's safety guarantees is that queries that are compile-time constants cannot be subverted by user-supplied data: they must either incorporate no user-controlled values, or incorporate them using the package's safe placeholder mechanism. In particular, call sites which build up SQL statements via fmt.Sprintf or string concatenation or other mechanisms will not be allowed.

False positives

If SafeSQL passes, your application is free from SQL injections (modulo bugs in the tool), however there are a great many safe programs which SafeSQL will declare potentially unsafe. These false positives fall roughly into two buckets:

First, SafeSQL does not currently recursively trace functions through the call graph. If you have a function that looks like this:

func MyQuery(query string, args ...interface{}) (*sql.Rows, error) {
        return globalDBObject.Query(query, args...)
}

and only call MyQuery with compile-time constants, your program is safe; however SafeSQL will report that (*database/sql.DB).Query is called with a non-constant parameter (namely the parameter to MyQuery). This is by no means a fundamental limitation: SafeSQL could recursively trace the query argument through every intervening helper function to ensure that its argument is always constant, but this code has yet to be written.

If you use a wrapper for database/sql (e.g., sqlx), it's likely SafeSQL will not work for you because of this.

The second sort of false positive is based on a limitation in the sort of analysis SafeSQL performs: there are many safe SQL statements which are not feasible (or not possible) to represent as compile-time constants. More advanced static analysis techniques (such as taint analysis) or user-provided safety annotations would be able to reduce the number of false positives, but this is expected to be a significant undertaking.