View script | License | Download script | History | Other scripts by: anton · brianh |
7-Sep 23:42 UTC
[0.056] 21.029k
[0.056] 21.029k
Archive version of: secure-clean-path.r ... version: 1 ... anton 20-Oct-2005Amendment note: new script || Publicly available? Yes REBOL [ Title: "secure-clean-path" File: %secure-clean-path.r Date: 19-Sep-2002 Version: 1.0.1 Needs: [] Author: ["Brian Hawley" "Anton Rolls"] Rights: { Copyright (C) Brian Hawley and Anton Rolls 2002. License for redistribution, use and modification is granted only if this copyright notice is included, and does not in any way confer ownership. It is requested, but not required, that the authors be notified of any use or modification, for quality-control purposes. } Library: [ level: 'intermediate platform: 'all type: 'function domain: [cgi file-handling files] tested-under: [view 1.3.1.3.1 on "Windows XP"] support: none license: 'bsd see-also: none ] Language: 'English Purpose: {Cleans up parent markers in a path, whilst restricting the output to a sandbox directory} Usage: { For help, type: help secure-clean-path Example: do %secure-clean-path.r ; now some examples secure-clean-path %my-dir/ ;== %my-dir/ secure-clean-path %my-dir/../ ;== %"" secure-clean-path %my-dir/../../ ;== %"" Use the inbuilt tests do/args %secure-clean-path.r "test" } ToDo: { - add more secure tests - implement limit refinement within parse rule like Brian's version (done 1.0.1) - /intolerant refinement - parenting beyond the root should make an error, eg. leading ../ should not just be lopped off - or parenting beyond the root should leave the parent-marker intact ? } History: [ 1.0.0 [12-Sep-2002 {First version, chopped secure-clean-path from simple-clean-path.r, implemented nocopy & limit refinements, added secure-tests, changed testing a bit} "Anton"] 1.0.1 [19-Sep-2002 {Applied simple-clean-path 1.0.6 bug fix to a refined version of Brian's secure-clean-path. Added test wrapper: do/args filename "test". Fixed some security holes and a misconception of limit in the tests.} "Brian"] ] Notes: {} ] comment { ; Previous versions ; Brian's first renamed version of simple-clean-path secure-clean-path: func [ {Cleans-up '.' and '..' in path; returns the cleaned path. Does not convert to absolute path from current directory. Considers multiple '/' to be erroneous, removes them. Only cleans relative to root value. Works on strings as well.} target [any-string!] {The path to be cleaned} /limit {Limit paths relative to this root} root [any-string!] {The root path (Default "", not applied if "")} /nocopy {Modify target instead of copy} /local root-rule a b c x ] [ ; Make copy of target if required (preserves any offset) if not nocopy [target: at copy head target index? target] ; Set root rule based on nature of root root-rule: either all [root not empty? root] [ ; Root rule applies, set to match root dir with / either (#"/" = pick root length? root) [root] [[root "/"]] ] [ ["/" | none] ; No root, just skip leading / if any ] ; Return cleaned target path starting with root-rule if parse/all target [ root-rule limit: any [ ; Quickly remove ./ a: "./" (remove/part a 2) :a | ; Remove (considered erroneous) multiple / a: some "/" b: (remove/part a b) :a | ; Apply any relative dots within limit a: some "." b: "/" c: ( ; Set x to the number of dots x: (index? b) - (index? a) ; For every . set 'a back one / within limit while [all [ positive? x: x - 1 b: find/reverse back a "/" (index? limit) < (index? b) ]] [a: next b] ; Remove the marked portion remove/part a c ) :a | thru "/" ; Regular directory, skip ] to end ] [target] ] ; End secure-clean-path ; Anton's version based on Anton's simple-clean-path version 5 secure-clean-path: func [{Cleans-up '.' and '..' in path; returns the cleaned path. Does not take absolute path from current directory (like clean-path does).} target [any-string!] "The path to be cleaned" /nocopy {Modify target instead of copy} /limit {Limit paths relative to this root} root [any-string! none!] {The root path (Default "", not applied if "")} /local blk result start pos prev post ][ ; Make copy of target if required (preserves any offset) if not nocopy [target: at copy head target index? target] ; clean the path if parse/all target [ ["/" |] ; leading slash or nothing start: pos: any [ end break | ["/" (remove pos) :pos] | [["./" :pos (remove/part pos 2)] | ["." end :pos (remove pos)]] | [["../" | ".." end] post: ( ; find previous slash prev: either prev: find/reverse back pos "/" [next prev][start] ) :prev ; set internal pointer back so it isn't left outside the string by the remove ( ; Remove the previous path section and set pos pos: remove/part prev post )] | [thru "/" pos:] | to end ] ][ ; handle limit refinement ;?? target ;?? limit ;?? root ;print ["all [limit root not empty? root] =" all [limit root not empty? root]] either all [limit root not empty? root][ if not-equal? #"/" last root [append root "/"] ; ensure root has a final slash find/match target root ; return portion of target path left over after the root ][target] ] ] } ; End previous versions ; Brian's new version, refined while trying to solve the crash, ; with the crash fix from Anton's simple-clean-path v5 adopted. ; Further refined after examining (and fixing) tests. secure-clean-path: func [ {Cleans-up '.', '..' and so on in path; returns cleaned path. Does not convert to absolute path from current directory. Considers multiple '/' to be erroneous, removes them. Only cleans relative to root value. Works on strings as well.} target [any-string!] {The path to be cleaned} /limit {Limit paths relative to this root} root [any-string!] {The root path (Default "", not applied if "")} /nocopy {Modify target instead of copy} /local root-rule a b c ] [ ; Make copy of target if required (preserves any offset) if not nocopy [target: at copy head target index? target] ; Set root rule based on nature of root root-rule: either all [root not empty? root] [ ; Root rule applies, set to match root dir with / either (#"/" = pick root length? root) [root] [[root "/"]] ] [ ["/" | none] ; No root, just skip leading / if any ] ; Return cleaned target path that starts with root-rule if parse/all target [ root-rule limit: any [ ; Quickly remove ./ or ending . a: "." ["/" | end] (remove/part a 2) :a | ; Remove (considered erroneous) multiple / a: some "/" b: (remove/part a b) :a | ; Apply any relative dots within limit a: some "." b: ["/" | end] c: ( ; For every extra . set 'a back one / within limit loop ((index? b) - (index? a)) - 1 [ either all [ b: find/reverse back a "/" (index? limit) <= (1 + index? b) ] [a: next b] [a: limit break] ] ) :a ( ; Set new position of a, or else crash! ; Remove the marked portion remove/part a c ) | thru "/" ; Regular directory, skip ] to end ; Skip any trailing stuff ] [target] ] ; End secure-clean-path ; Test cases! if system/script/args = "test" [ use [ simple-tests secure-tests do-test-block limit path correct-result result err ] [ ; tests without limit refinement simple-tests: [ %"" %"" %one %one %one/ %one/ %one/two %one/two %one/two/ %one/two/ %../ %"" %.. %"" %./ %"" %. %"" %one/../ %"" %one/.. %"" %one/./../ %"" %one/./.. %"" %one/.././../ %"" %one/.././.. %"" %one/two/../ %one/ %one/two/.. %one/ %one/../two %two %one/../two/ %two/ %one/.././two/three/../ %two/ %one/two/../../.././/../../hello////../././there/path %there/path ; these are the same as above, except beginning with a slash %/ %/ %/one %/one %/one/ %/one/ %/one/two %/one/two %/one/two/ %/one/two/ %/../ %/ %/.. %/ %/./ %/ %/. %/ %/one/../ %/ %/one/.. %/ %/one/./../ %/ %/one/./.. %/ %/one/.././../ %/ %/one/.././.. %/ %/one/two/../ %/one/ %/one/two/.. %/one/ %/one/../two %/two %/one/../two/ %/two/ %/one/.././two/three/../ %/two/ %/one/two/../../.././/../../hello////../././there/path %/there/path %// %/ ; Multiple dots, to deal with REBOL's behavior on Brian's system %/////one/two/../../../.././././../../.../....////five/six/seven?/?/ %/five/six/seven?/?/ %/one/two/three/.../ %/one/ %/one/two/three/.../four/ %/one/four/ "risqué%filename" "risqué%filename" "risqué//%filename" "risqué/%filename" ] ; Tests with limit refinement, fixed to reflect proper behavior. secure-tests: compose [ %one/ %one/ %one/ (none) %one/ %one/ %one/ %one/two/three//.. %one/two/ %one/ %one/two/three//../ %one/two/ %one/two/ %one/two/three %one/two/three %one/two %one/two/three %one/two/three ; these are the same as above, except beginning with a slash %/one/ %/one/ %/one/ (none) %/one/ %/one/ %/one/ %/one/two/three//.. %/one/two/ %/one/ %/one/two/three//../ %/one/two/ %/one/two/ %/one/two/three %/one/two/three %/one/two %/one/two/three %/one/two/three ; Testing for improperly late limit application - allowing ; such would let an attacker get info about your system! %one/two/ %one/../one/two/ (none) %/one/two/ %/one/../one/two/ (none) %one/../two/ %two/ (none) %/one/../two/ %/two/ (none) %one/../two/ %one/../two/ %one/../two/ %/one/../two/ %/one/../two/ %/one/../two/ %one/../two/ %one/../two/three/ %one/../two/three/ %/one/../two/ %/one/../two/three/ %/one/../two/three/ ; Testing for proper failure %/one/ %one/ (none) ] do-test-block: [ if error? set/any 'err try [ result: either none? limit [ secure-clean-path path ] [ secure-clean-path/limit path limit ] ][ print [ "Error" newline tab "limit:" mold limit newline tab "path:" mold path newline tab "correct-result:" mold correct-result newline mold disarm err ] err break ] either not-equal? result correct-result [ print [ "[Wrong] " mold limit " ^-" mold path "^- -> ^-" mold result " " "should be " mold correct-result ] ;break ][ print ["[OK] " mold limit " ^-" mold path "^- -> ^-" mold result] ] ] print "Simple Tests:" limit: none foreach [path correct-result] simple-tests do-test-block print "Tests with /Limit:" foreach [limit path correct-result] secure-tests do-test-block ] ; End use ] ; End if "tests" |