4.3.25 - ouch (reverting to 4.3.22) 🔗 link

wrote:

My testing of the release candidate was obviously inadequate. My phone
rang off the hook after shipping 4.3.25 to my production server. I've
reverted it to 4.3.22 while I try to get a handle on the changes.

:: svcstatus refresh ::

I didn't see any mention of this in the release notes, but the page is now
being delivered with a 30-second _header_ refresh in place of a 60-second
_meta_ refresh.

I assume the change in method is related to XSS protection changes, but I
don't know for sure. The shortening of the interval concerns me much more
than the change in method. I'd prefer the darned page didn't refresh at
all, but hard coding it to every half minute?  Ouch. Ick. What's the
business case here? Anyone who his visually monitoring a single info page
can hit F5 when they want a refresh, can't they?

:: XSS protections ::

It looks like the XSS protection bits inserted into http headers

content-security-policy: script-src 'self'; connect-src 'self';
form-action 'self'; sandbox allow-forms;
X-Content-Security-Policy: script-src 'self'; connect-src 'self';
form-action 'self'; sandbox allow-forms;
X-Webkit-CSP: script-src 'self'; connect-src 'self'; form-action 'self';
sandbox allow-forms;

has broken several functions for me. The most important things are the
notes our operators use to figure out who to call. We insert some html into
the DESCR tag of hosts.cfg. This appears on the "info" page as a link. The
links are to static content served by the same Apache web server. The tags
are of the form:

DESCR:"Important Server:<a href=/xymon/CNotes/ImportantServer
target=Notes>Reference notes</a>"

With Firefox, the links work. With InternetExploder and Chrome, the links
quietly fail to work. The Chrome console returns the error message:
  Blocked script execution in 'https://foo.bar.com/xymon-
  cgi/svcstatus.sh?HOST=bb.bar.com&SERVICE=info' because
  the document's frame is sandboxed and the 'allow-scripts'
  permission is not set.
Removing the "target" property from the A tag allows it to work. It
overwrites the info page, but at least it works :p It feels like a step
back to the 90's be forced into a single page that keeps getting replaced.

I don't know enough about the "content security policy". What are my
options to retain the named-window/tab capabilities?

:: svcstatus disable function ::

We are unable to enable/disable test from a host's "info" page. We can do
so from enadis.sh, but this is a lot harder for many of our users.

In Chrome, attempts to disable tests from the "info" page is generating
the same "frame is sandboxed" messages.

In Firefox, the attempts just don't do anything. There are messages in the
console
  Content Security Policy: The page's settings blocked the loading
  of a resource at self ("script-src https://x.foo.com";).
so maybe the information necessary to build the form isn't even being
loaded.


--
   Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

I very much disagree with the idea of having to hit F5 every so often to
see a refreshed Xymon page, rather then it refreshing automatically.

If I have a Xymon page up on a monitor and then I am working on
something else on another monitor, I don't want to have to go over to
the browser on the other monitor and hit F5 every so often to refresh
the page. I'd much rather it refresh every so often automatically, so
then I just have to look over at the monitor every so often to look for
any changes.

Where exactly are you seeing this refresh change (just curious)?

Why wouldn't you want it to refresh automatically? Isn't the idea to get
the most up-to-date info possible?

-- 
    Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

wrote:

On 2/10/2016 12:44 PM, Matt Vander Werf wrote:
- snip -

I very much disagree with the idea of having to hit F5 every so often to
see a refreshed Xymon page, rather then it refreshing automatically.

If I have a Xymon page up on a monitor and then I am working on
something else on another monitor, I don't want to have to go over to
the browser on the other monitor and hit F5 every so often to refresh
the page. I'd much rather it refresh every so often automatically, so
then I just have to look over at the monitor every so often to look for
any changes.

Where exactly are you seeing this refresh change (just curious)?

This refresh has been moved from the _meta_ content of the svchosts page
the HTTP _header_ . As such, "view source" will no longer reveal it.

Why wouldn't you want it to refresh automatically? Isn't the idea to get

the most up-to-date info possible?

Not in all cases. I want the general pages of hosts (like xymon.html) to
refresh periodically, but there is no need to have them auto-refresh faster
than I am running xymongen.

The issue I see is specifically with the "info" page produced by
svcstatus.sh. I have it open when I'm studying a host for a specific
reason. I would prefer that page not auto-refresh while I'm studying it
trying to figure out a problem. If I want new information, I can refresh
it. I might even want to keep it frozen and open a new copy in a new tab so
I could compare the past and the present.

Does anyone put a single host's "info" page on a screen and leave it there
for general monitoring? If so, why? What problem does it solve that looking
at the parent page of hosts does not?

--
   Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

On Wed, February 10, 2016 2:02 pm, John Thurston wrote:

On 2/10/2016 12:44 PM, Matt Vander Werf wrote:
- snip -

I very much disagree with the idea of having to hit F5 every so often to
see a refreshed Xymon page, rather then it refreshing automatically.

If I have a Xymon page up on a monitor and then I am working on
something else on another monitor, I don't want to have to go over to
the browser on the other monitor and hit F5 every so often to refresh
the page. I'd much rather it refresh every so often automatically, so
then I just have to look over at the monitor every so often to look for
any changes.

Where exactly are you seeing this refresh change (just curious)?

This refresh has been moved from the _meta_ content of the svchosts page
the HTTP _header_ . As such, "view source" will no longer reveal it.

Why wouldn't you want it to refresh automatically? Isn't the idea to get
the most up-to-date info possible?

Not in all cases. I want the general pages of hosts (like xymon.html) to
refresh periodically, but there is no need to have them auto-refresh
faster than I am running xymongen.

The issue I see is specifically with the "info" page produced by
svcstatus.sh. I have it open when I'm studying a host for a specific
reason. I would prefer that page not auto-refresh while I'm studying it
trying to figure out a problem. If I want new information, I can refresh
it. I might even want to keep it frozen and open a new copy in a new tab
so I could compare the past and the present.

Does anyone put a single host's "info" page on a screen and leave it
there for general monitoring? If so, why? What problem does it solve
that looking at the parent page of hosts does not?

--
    Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

On 2/10/2016 1:14 PM, J.C. Cleaver wrote:

On the refresh value, this was an unintentionally broad change.

CSP sadly catches the META HTTP-EQUIV "Refresh" tag as well as something
that is no longer honored, which requires moving it up into the actual
HTTP headers. This updated version of the CSP patch (from the last email)
does two things:
1) separate out info and trends pages from "regular" svcstatus pages. The
former won't be auto-refreshed
2) adds a previously-referenced XYMWEBREFRESH variable, which can be used
to configure this (default: 60s)

Going from 60s to 30s was an error on my part. I'd actually thought that
was the value for some reason...

On info pages not allowing _targets, that's also something caught by CSP.
The patch should fix this as well. Please verify if you can.

-- 
    Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

On Wed, February 10, 2016 3:25 pm, John Thurston wrote:

On 2/10/2016 1:14 PM, J.C. Cleaver wrote:

1) separate out info and trends pages from "regular" svcstatus pages.
The
former won't be auto-refreshed
2) adds a previously-referenced XYMWEBREFRESH variable, which can be
used
to configure this (default: 60s)

Going from 60s to 30s was an error on my part. I'd actually thought that
was the value for some reason...

These patches are helping. Thank you!

On info pages not allowing _targets, that's also something caught by
CSP.
The patch should fix this as well. Please verify if you can.

To let the "target=_blank" option work, I needed to add "allow-popups"
to line 269 of lib/cgi.c

    269         else if (strncmp(str, "svcstatus-info", 14) == 0) csppol
= strdup("script-src 'self' 'unsafe-inline'; connect-src 'self';
form-action 'self'; sandbox allow-forms allow-scripts
allow-popups;");

On 2/10/2016 6:52 PM, J.C. Cleaver wrote:

Can you test with the following two patches? I believe these should take
care of the outstanding issues. It seems like some versions of Chrome need
the allow-modals permission as well.

-- 
    Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

On 2/11/2016 9:48 AM, John Thurston wrote:

On 2/10/2016 6:52 PM, J.C. Cleaver wrote:

Can you test with the following two patches? I believe these should take
care of the outstanding issues. It seems like some versions of Chrome
need the allow-modals permission as well.

I've taken a fresh extract of the .25 tar and applied these two patches.

-- 
    Do things because you should, not just because you can.

John Thurston    XXX-XXX-XXXX
user-ce4d79d99bab@xymon.invalid
Enterprise Technology Services
Department of Administration
State of Alaska

On Thu, February 11, 2016 11:02 am, John Thurston wrote:

On 2/11/2016 9:48 AM, John Thurston wrote:

On 2/10/2016 6:52 PM, J.C. Cleaver wrote:

Can you test with the following two patches? I believe these should
take
care of the outstanding issues. It seems like some versions of Chrome
need the allow-modals permission as well.

I've taken a fresh extract of the .25 tar and applied these two patches.

It looks like links (such as I embed in the DESCR tags) on
'svcstatus.sh?SERVICE=info' are not getting a referrer set on Chrome or
Exploder. The referrer is present with FireFox.

I suspect it is back with the "content-security-policy" again. I'll see
if I can identify the required pixie dust.