Bug 1861300

Summary: Huge memory leak in probe_rpmverifyfile [rhel-7.9.z]
Product: Red Hat Enterprise Linux 7 Reporter: Renaud Métrich <rmetrich>
Component: openscapAssignee: Jan Černý <jcerny>
Status: CLOSED ERRATA QA Contact: Matus Marhefka <mmarhefk>
Severity: high Docs Contact:
Priority: urgent    
Version: 7.8CC: cbesson, ekolesni, jcerny, jreznik, kwalker, mhaicman, reg, sadas
Target Milestone: rcKeywords: AutoVerified, Regression, ZStream
Target Release: ---Flags: lcervako: mirror+
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2020-11-10 13:03:14 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Renaud Métrich 2020-07-28 09:11:08 UTC
Description of problem:

A huge memory leak is present in probe_rpmverifyfile, causing various issues on the system when executing oscap, typically OOMKiller starts killing processes and services.

On a small system with 450 RPMs, the memory consumed grows up to 900 MB!

    ==1668== 901,115,904 bytes in 219,999 blocks are definitely lost in loss
    record 77 of 77
    ==1668==    at 0x4C29F73: malloc (vg_replace_malloc.c:309)
    ==1668==    by 0x749B93D: realpath@@GLIBC_2.3 (canonicalize.c:78)
    ==1668==    by 0x10EC5D: UnknownInlinedFun (stdlib.h:48)
    ==1668==    by 0x10EC5D: rpmverify_collect (rpmverifyfile.c:249)
    ==1668==    by 0x10EC5D: probe_main (rpmverifyfile.c:546)
    ==1668==    by 0x110C4F: probe_worker (worker.c:954)
    ==1668==    by 0x11074B: probe_worker_runfn (worker.c:56)
    ==1668==    by 0x7243EA4: start_thread (pthread_create.c:307)
    ==1668==    by 0x75568DC: clone (clone.S:111)
    
    ==1668== LEAK SUMMARY:
    ==1668==    definitely lost: 901,150,675 bytes in 224,438 blocks


Version-Release number of selected component (if applicable):

openscap-1.2.17-9.el7_8 and later


How reproducible:

Always


Steps to Reproduce:
1. Start probe_rpmverifyfile under Valgrind

  # /usr/bin/valgrind --tool=memcheck --leak-check=yes /usr/libexec/openscap/probe_rpmverifyfile

2. Send the various orders on stdin one at a time, wait for result

First: (fast)
(#d8:seap.msg#d3::id#d0((#d20:rpmverifyfile_object#d3::id#d47:oval:ssg-object_files_fail_user_ownership:obj:1#d13::oval_version#d4:5.11)((#d4:name#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d5:epoch#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d7:version#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d7:release#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d4:arch#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d8:filepath#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d6:filter#d7::action#d2)#d46:oval:ssg-state_files_fail_user_ownership:ste:1)((#d9:behaviors#d6::nomd5#d4:true#d13::noghostfiles#d4:true))))

Second: (slow)
(#d8:seap.cmd#d3::id#d0#d9::reply_id#d0#d6::class#d3:usr#d5::type#d5:async#d1(((#d19:rpmverifyfile_state#d3::id#d46:oval:ssg-state_files_fail_user_ownership:ste:1#d9::operator#d1)((#d17:ownership_differs#d10::operation#d5#d10::var_check#d1#d13::entity_check#d1#d16::check_existence#d3)#d4:fail))))

Third: (fast)
(#d8:seap.msg#d3::id#d1((#d20:rpmverifyfile_object#d3::id#d48:oval:ssg-object_files_fail_group_ownership:obj:1#d13::oval_version#d4:5.11)((#d4:name#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d5:epoch#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d7:version#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d7:release#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d4:arch#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d8:filepath#d10::operation#d11#d10::var_check#d1)#d2:.*)((#d6:filter#d7::action#d2)#d47:oval:ssg-state_files_fail_group_ownership:ste:1)((#d9:behaviors#d6::nomd5#d4:true#d13::noghostfiles#d4:true))))

Fourth: (slow)
(#d8:seap.cmd#d3::id#d1#d9::reply_id#d1#d6::class#d3:usr#d5::type#d5:async#d1(((#d19:rpmverifyfile_state#d3::id#d47:oval:ssg-state_files_fail_group_ownership:ste:1#d9::operator#d1)((#d13:group_differs#d10::operation#d5#d10::var_check#d1#d13::entity_check#d1#d16::check_existence#d3)#d4:fail))))


Actual results:

High memory consumption

Expected results:

Small memory consumption (~200MB)


Additional info:

This is a critical issue, system can get down due to this.

Comment 4 Jan Černý 2020-07-28 11:56:43 UTC
The bug is reproducible also with current upstream maint-1.3 branch as of HEAD 802eb801dd083a67bd64d4f9c711ba1426e626cb.

It seems to be triggered by any rule that uses rpmverifyfile probe, eg. xccdf_org.ssgproject.content_rule_rpm_verify_permissions, xccdf_org.ssgproject.content_rule_rpm_verify_ownership.

For example, on Fedora 32:

oscap xccdf eval --profile standard --rule xccdf_org.ssgproject.content_rule_rpm_verify_permissions /usr/share/xml/scap/ssg/content/ssg-fedora-ds.xml

Valgrind reports the same leak:

==400443== 
==400443== 352,256 bytes in 86 blocks are possibly lost in loss record 352 of 353
==400443==    at 0x483A809: malloc (vg_replace_malloc.c:307)
==400443==    by 0x51F23A6: realpath@@GLIBC_2.3 (in /usr/lib64/libc-2.31.so)
==400443==    by 0x489F8EA: oscap_realpath (util.c:251)
==400443==    by 0x495E766: rpmverify_collect (rpmverifyfile_probe.c:248)
==400443==    by 0x495F4D8: rpmverifyfile_probe_main (rpmverifyfile_probe.c:543)
==400443==    by 0x493560F: probe_worker (worker.c:1090)
==400443==    by 0x4932F87: probe_worker_runfn (worker.c:81)
==400443==    by 0x4CDA431: start_thread (in /usr/lib64/libpthread-2.31.so)
==400443==    by 0x52A99D2: clone (in /usr/lib64/libc-2.31.so)
==400443== 
==400443== 2,359,775,232 bytes in 576,117 blocks are definitely lost in loss record 353 of 353
==400443==    at 0x483A809: malloc (vg_replace_malloc.c:307)
==400443==    by 0x51F23A6: realpath@@GLIBC_2.3 (in /usr/lib64/libc-2.31.so)
==400443==    by 0x489F8EA: oscap_realpath (util.c:251)
==400443==    by 0x495E766: rpmverify_collect (rpmverifyfile_probe.c:248)
==400443==    by 0x495F4D8: rpmverifyfile_probe_main (rpmverifyfile_probe.c:543)
==400443==    by 0x493560F: probe_worker (worker.c:1090)
==400443==    by 0x4932F87: probe_worker_runfn (worker.c:81)
==400443==    by 0x4CDA431: start_thread (in /usr/lib64/libpthread-2.31.so)
==400443==    by 0x52A99D2: clone (in /usr/lib64/libc-2.31.so)

Comment 10 Marek Haicman 2020-08-19 12:15:10 UTC
For posterity - as the rpmverifyfile probe is used in very limited number of SCAP rules, the workaround for now is to use tailoring, and remove those rules:

* file_permissions_unauthorized_sgid
* rpm_verify_hashes
* rpm_verify_permissions
* rpm_verify_ownership

This will prevent the crashes. With that, I am slightly reducing the severity rating of the BZ.

Comment 23 errata-xmlrpc 2020-11-10 13:03:14 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (openscap bug fix and enhancement update), and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2020:5017