Bug 2229209

Summary: entries "epo" and "map" in hostname_map section of default_mappings change the "sosreport" file name
Product: Red Hat Enterprise Linux 8 Reporter: Carlos Santos <casantos>
Component: sosAssignee: Pavel Moravec <pmoravec>
Status: NEW --- QA Contact: Supportability QE <supportability-qe>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 8.8CC: agk, jcastillo, jjansky, plambri, sbradley, theute
Target Milestone: rc   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 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 Carlos Santos 2023-08-04 16:02:29 UTC 
Description of problem:

When the hostname_map section in /etc/sos/cleaner/default_mapping contains
entries like these,



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


How reproducible:

Always

Steps to Reproduce:

1. Ensure that sos report will find hosts with name "epo" and "map"

   # echo 192.168.122.1 epo.map.example.com epo map >> /etc/hosts

2. Remove the existing default mapping

   # rm -f /etc/sos/cleaner/default_mapping

3. Generate the report

   # sos report --batch --case-id 00000000 --clean

Actual results:

[...]
sosreport-<hostname>-00000000-2023-08-04-zddhrim :   Beginning obfuscation...
sosreport-<hostname>-00000000-2023-08-04-zddhrim :   Obfuscation completed [removed 920 unprocessable files]

Successfully obfuscated 1 report(s)

Creating compressed archive...

A mapping of obfuscated elements is available at
	/var/tmp/sosrhost1rt-host140-00000000-2023-08-04-zddhrim-private_map

Your sosreport has been generated and saved in:
	/var/tmp/sosrhost141rt-host14140-00000000-2023-08-04-zddhrim-obfuscated.tar.xz

 Size	6.36MiB
 Owner	root
 sha256	3c9e8df05545a6d57c058e03ea0d474dbdf8720b6d59a8493d3985b309cf1b75

Please send this file to your support representative.

Expected results:

Your sosreport has been generated and saved in:
	/var/tmp/sosreport-host14140-00000000-2023-08-04-zddhrim-obfuscated.tar.xz

Additional info:
Comment 1 Pavel Moravec 2023-08-05 21:08:30 UTC 
This is tricky.

Sos takes /etc/hosts for hostname mapping. So whatever string is there, it treats it as a sensitive information to obfuscate. So for e.g. "rep" in /etc/hosts, "sosreport" string is obfuscated - including tarball filename (as it usually contains host's shortname).

How to behave better? Obfuscate not (sub)strings blindly, but for each string STR to camouflage, find just strings matching [^|\W]STR[$|\W] regular expression? (i.e. start of line or non-alphanumeric character before the STR, and end of line or non-alphanumeric character after the STR)?

What about *not* obfuscating "rep01", then? Is it ok? (I feel so..)


sos cleaner has a few parsers/mappers, one for hostnames+domainnames, one for IPv4, one for IPv6, one for MAC addresses, usernames and keywords (provided by user via a config). I guess this feature "dont obfuscate substrings, just strings" should be applied only to:
- hostnames+domainnames
- usernames
- keywords

As only those are alphanumeric words.

Is that improvement sound?
Comment 2 Carlos Santos 2023-08-07 12:26:34 UTC 
(In reply to Pavel Moravec from comment #1)

> sos cleaner has a few parsers/mappers, one for hostnames+domainnames, one
> for IPv4, one for IPv6, one for MAC addresses, usernames and keywords
> (provided by user via a config). I guess this feature "dont obfuscate
> substrings, just strings" should be applied only to:
> - hostnames+domainnames
> - usernames
> - keywords
> 
> As only those are alphanumeric words.
> 
> Is that improvement sound?

I will ask the customer in the associated support case if that's a satisfying
approach.
Comment 3 Carlos Santos 2023-08-08 16:34:51 UTC 
(In reply to Pavel Moravec from comment #1)

> sos cleaner has a few parsers/mappers, one for hostnames+domainnames, one
> for IPv4, one for IPv6, one for MAC addresses, usernames and keywords
> (provided by user via a config). I guess this feature "dont obfuscate
> substrings, just strings" should be applied only to:
> - hostnames+domainnames
> - usernames
> - keywords
> 
> As only those are alphanumeric words.
> 
> Is that improvement sound?

According to the customer, that seems fine, but mostly concerning are the
single and double chars masking. It might lead to unknown issues, like
decreased time of cleaning, with wrong masking of insensitive chars.

---> "e": "host22",