Strange networking problem

From time to time, my copy of Vista seems to go into a "spasming" state
which affects networking. When in this state, the following symptoms
recur frequently, some of them at intervals of roughly five minutes.
(One time, the first symptom group kept recurring like clockwork almost
exactly every 2 minutes.) The state itself lasts hours; its absence also
tends to last hours and sometimes days. It can be induced by restarting
either the router (a BEFSR41) or the computer. I know of no way to
induce its absence.

Symptom group 1 recurs in a particular sequence when this state exists:
* First then mouse pointer briefly shows a busy-cursor-with-arrow, even
if the user and all applications are (supposed to be) idle.
* About five seconds later, the networking icon in the tray loses its
globe, usually replacing it with a red X and occasionally with a
yellow ! in a triangle or with nothing.
* The pointer indicates "busy" again but more briefly about a second
later.
* The networking icon returns to normal about a second later.
This full sequence takes less than ten seconds and recurs every few
minutes, sometimes less frequently, sometimes as often as every two
minutes and (in that case) with clockwork regularity.

The other symptom groups involve applications that use networking. As a
rule these symptoms involve various intermittent or sporadic erroneous
behaviors by these applications, some of which could be explained by
network timeouts or similar effects from a hypothetical brief network
outage but some of which cannot.

The timing of these symptoms generally does NOT correlate with the
timing of the events in symptom group 1; e.g. Firefox may fail to load
pages while the tray icon shows normal connectivity, and may succeed
while the tray icon shows a red X (request initiated by mouse click and
web page displayed by browser all during the two-second duration of the
red X). This and some symptoms not being explained by loss of network
response cast doubt on any theory where this is simply the network
connection itself spasming. The fact that this state can be induced
intentionally by resetting the router or computer (without touching the
DSL modem) also points away from a simple network flakiness explanation,
as does the fact that the modem lights do not exhibit any symptoms
whatsoever. The problem thus seems to exist closer to the computer than
the modem, or at a higher level of the networking stack than the wiring
or PPPoE protocol (so, the IP layer or higher), or both.

In the below symptom groups, symptoms marked with a (!) at the end
cannot be explained solely by inability of the application to contact a
remote host. If it's (!!) the application's behavior seems to be
patently incorrect no matter what -- i.e., it looks like an application
bug is involved, not just the networking, router, and/or operating
system behaving flakily.

Symptom group 2 involves Firefox:

* Pages may intermittently fail to load normally, load only part way,
timeout, or similarly, and hostname lookup may hang or fail (hanging
being particularly common).
* The browser may ignore a link click, or spin for a bit but then stop
with "Done" displayed in the status bar without having gone
anywhere.(!!)
* The browser may display "This page cannot be displayed because the
browser is in Offline Mode. Select "Work Offline" from the File menu."
or words to that effect; the user did not activate offline mode and if
the file menu is dropped down, there is no checkmark by "Work
Offline". Furthermore, again without user intervention the browser
will behave as if offline mode were toggled off again after roughly
five seconds.(!!)
(Clearly an application bug involved here; the state of the check on
the menu item and whether it considers itself to be in offline mode
for the purpose of page retrieval should never get out of synch, nor
should the mode toggle without user input. On the other hand something
external is clearly the trigger, since FF only does this when Vista/my
network/whatever is in the state being described by this news post.)
* There are problems with specific web sites above and beyond the above.
For example, some sites, including Sourceforge, will log me out every
five minutes like clockwork while my computer is in this state. This
suggests cookie destruction is occurring, but it's selective to
particular sites' cookies. No cookie blocking addons or software
are installed on this system, and Firefox's cookie policy is left
default; furthermore, this only occurs during the "spasming state" of
the machine/network.(!)
* At the same time as the red-X spasms mentioned in Symptom Group 1,
Firefox may briefly (for a few seconds) consume up to 25% CPU.(!!)
The other FF symptoms either don't correlate with SG1 at all, or
correlate but not in a way where both can be explained simply as a
brief loss of connectivity -- in particular, the "premature Done"
error tends to be followed by the spurious "offline mode" toggle,
THEN by the symptoms of SG1, with "offline mode" toggling back off
AND normal connectivity (web pages successfully retrieved) before SG1
subsides. The timing is wrong for it to be simple connectivity loss
causing it, where the red X should coincide, rather than follow,
a period of web pages being unretrievable (and the browser should
simply display timeout errors, not any of the other symptoms!).

Symptom group 3 involves Thunderbird:

* Attempting to view a news post may hang with the throbber going and
no progress being made.
* When the above occurs, the "Stop" button may fail to abort the
connection so as to enable a retry; Thunderbird then has to be closed
and restarted.(!!)
* News posts may hang while sending.
* These posts are eventually reported as successful, but never appear
on the server.(!)
* Canceling a news post send that's hung and then manually recreating
the post (aided by copy/paste) and resubmitting tends to get
interrupted by a pair of dialogs complaining about some error moving
something to the "sent" folder (despite the fact that the user is
currently editing, not sending, a news post).(!!)
* Thunderbird may hang with unstoppable, no-results network activity
on startup and have to be restarted three or four times (stop
button doesn't work).(!!)

Symptom group 4 involves the BEFSR41:

* During one of these seizures, the BEFSR41 may stop functioning
correctly in a manner requiring it to be powered down and back
up again.(!!) The symptoms of this are:
* No network connectivity behind the router.
* Router may display normal status (connected, normal IP/DNS/etc.
data) over its web interface, or said interface may become
unreachable.
* Using the web interface (if functional) to Disconnect and then
Connect rarely fixes the problem; usually causes the web
interface to hang when Connect is clicked with no restoration
of connectivity.
* Windows shows a networking icon with no X or triangle or globe
once this has occurred. Power cycling the BEFSR41 invariably
fixes it, but I suspect that it prolongs, just as it may
cause to begin with, the spasming.

Other applications that use the network may display symptoms of
intermittent connectivity, but by and large do not display any symptoms
not explainable solely by their connectivity being intermittent.

There's an XP box behind the same router and it gets affected at the
same times, but in a milder way: all network-using applications behave
as if connectivity is intermittent, and "Local Area Connection is now
connected" balloons pop up randomly from time to time from the tray,
probably XP's equivalent of Symptom Group 1.

My suspicion at this time is that the root problem lies with the BEFSR41
but has symptoms elsewhere.

1. That the BEFSR41 gets affected (Symptom Group 4) and the XP box gets
affected proves the problem is not confined to the Vista box; the
lack of any abnormality with the DSL modem strongly implies that it
isn't sited further away, either.
Even if it is sited further away (likely at the DSL hub) the router's
reaction to it indicates bugs in the router.
2. Symptom Groups 2 and 3 indicate that there are a plethora of bugs and
warty behaviors in Mozilla's products that are provoked by
intermittent connectivity. The cookie crunchage may, however, be the
operating system's fault, since the OS can obviously delete files,
though it seems unlikely given the specific targeting of not only
cookies, but non-Internet Explorer cookies *from particular web
sites*.
3. Symptom Group 1 is inconsistent, however, with simple brief
connectivity losses. Such would cause the globe icon to come and go
only, and should not cause busy cursors. Furthermore, I occasionally
get connectivity losses unrelated to these "spasm states"; in these, the
DSL modem's external-connectivity light goes off for a while. The
following symptoms occur:

* Globe icon vanishes from tray for the duration.
* All network-using apps act like connectivity has been lost.
* TB may exhibit any of the bugs in Symptom Group 3, particularly stop
button failures and startup failures, at onset; this suggests that
SG3 is due to bugs in TB triggered by any loss of connectivity.
* FF does NOT display any of the bugs in SG2 other than the simple
inability to load web pages. In particular none of the bugs marked
with a (!) or (!!) occurs. It does not spontaneously toggle "offline
mode", it does not claim a page load is "done" with neither a page
load nor an error message having occurred, and it does not lose any
login cookies or any other cookies.
This suggests that the bugs in FF are NOT provoked by any old loss
of connectivity; there's something "special" about these "spasms".

This points back to the BEFSR41. Let's take one of the SG1 symptoms at
face value: the red X indicating that there's no even *local* network
for a short time. This suggests that the BEFSR41 is in some kind of
failure mode independently of whether it has lost its connection
upstream. No problem upstream (e.g. at my ISP, or with the DSL modem)
should cause the BEFSR41 to stop functioning even as a local router
between the XP box and the Vista box. Perhaps a problem in one of those
two places does exist and is the trigger, but what it triggers is a
buggy behavior in the router, not merely a loss in connectivity, and
moreover loss of connectivity is not by itself sufficient to cause the
buggy behavior in the router.

When the router does exhibit the buggy behavior, my theory is that it
triggers some buggy behavior in Vista and in Firefox, in turn. Probably
the first domino is Vista deciding there's no network connection at all
rather than one that temporarily isn't working, and doing some kind of
busy work (busy cursor) to unload (and later load again) a bunch of
drivers or something. Plus it tells all applications that the network
doesn't even exist anymore, rather than merely is down. This in turn I
theorize triggers the buggy behaviors in FF that don't occur with any
old loss in connectivity.

So we have bugs in Thunderbird caused by any loss in connectivity, bugs
in Firefox caused by the OS telling FF the network's been physically
unplugged (even if you grant that aborting a page load without any error
message, or spontaneously toggling on "offline mode", is a legitimate
response to such, a) not keeping the "offline mode" menu item checkmark
in synch with the actual mode state is a bug and b) crunching cookies in
response is a bug, and chewing 25% CPU for 2 seconds whenever this
happens and even if idle at the time is questionable at best), bugs in
the router, and an unknown cause.

There are four plausible locations for the cause.

1. The router itself. Since it exhibits indisputably wrong behavior at
times, the simplest explanation is this one. The router sometimes
gets into a wonky state (causing SG1-3), which may resolve after a
while or may cause it ultimately to hang (SG4). It starts up in this
state and possibly rebooting the computer (either computer?) sends
some signal down the line (a DHCP request?) to it that may trigger
it.
That it starts up in this state points away from overheating being
the cause, as does its being well ventilated, with the holes in its
case unobstructed, and the fact that this has been observed to happen
when the room ambient temp was only 18 deg C.
One candidate trigger is the DHCP host in the router. If the router
got kicked into the spasming state by its DHCP requesting a new IP
address from my ISP's network, then spasming would be triggered by
any reboot of the router AND whenever the DHCP lease from my ISP
expired. I've several times witnessed spastic behavior start right at
7 in the morning on the dot; perhaps my ISP's DHCP leases all expire
at that time of day.
So my #1 suspect site for the bug is the router's DHCP host
functionality.
2. The DSL modem. This is the third device directly connected to the
router. Weighing against this is the absence of any overt symptoms
involving the modem itself, such as changes in its status lights.
3. My ISP's network/config. Perhaps something the ISP does from time
to time (some ICMP message involved in keeping their network
running and keeping track of connected endpoints, so they can free
up an IP address early if a user shuts their computer down or
whatnot?) triggers a bug in the BEFSR41.
4. Something further out there; there's some type of network packet that
any random Internet host can send that sends BEFSR41s into a spastic
state for hours. (This is unlikely, but scary if true; it means the
routers have a moderately nasty denial-of-service vulnerability.)

Does anyone here have any insight? Perhaps some of the symptoms can be
mitigated, e.g. some Vista setting that prevents it from ever treating
the network as actually gone rather than just temporarily down?

If anyone here knows of a way to fix a BEFSR41 to be immune to this,
that would of course be even better.

 

Well?

 

Check the crosspost list.

 

I guess I need to spell it out explicitly for you.

I selected TWO Vista newsgroups to post to. I found one in microsoft.*
and posted to that so my post would appear where Microsoft was most
likely to notice it, and additionally I posted to a Vista newsgroup in
alt.* because that one seemed to be the highest-traffic Vista newsgroup.

If you're meaning to suggest that there's a third newsgroup that would
be a good place to post it, then please stop beating around the bush and
simply name the newsgroup. Then I'll see if my server carries it, and if
so post a copy of my original post to that group.

If you have no intention of being constructive, EITHER with regard to
the original post's queries OR by naming a newsgroup where you think
they would get more attention (of a constructive kind!), then please
don't waste time and bandwidth by posting again to this thread.

 

I guess I need to spell it out explicitly for you.

I selected TWO Vista newsgroups to post to. I found one in microsoft.*
and posted to that so my post would appear where Microsoft was most
likely to notice it, and additionally I posted to a Vista newsgroup in
alt.* because that one seemed to be the highest-traffic Vista newsgroup.

If you're meaning to suggest that there's a third newsgroup that would
be a good place to post it, then please stop beating around the bush and
simply name the newsgroup. Then I'll see if my server carries it, and if
so post a copy of my original post to that group.

If you have no intention of being constructive, EITHER with regard to
the original post's queries OR by naming a newsgroup where you think
they would get more attention (of a constructive kind!), then please
don't waste time and bandwidth by posting again to this thread.

 

Then why were you rude? Twice?

I don't notice any newsgroup names, real information about the problem,
or silence from you. Shouldn't there have been at least one of those three?

 

Sure there is, unless you for some reason believe the problem can never
be solved by anyone other than Microsoft.

Perhaps there's a newsgroup focused on Linksys routers?

How fortunate then that I didn't do so -- I merely ASKED you to EITHER
be silent OR do one of two other things (none of which you have so far
done).

I am, by attempting to elicit a more constructive post from you. However
I begin to suspect that that, sadly, isn't going to happen.

 

Updating the firmware on the BEFSR41 appears to have fixed it. If anyone
suggested that amid all the flaming between Speed Dial and whoever else,
thanks.

 

I read newsgroups to understand more about Windows Vista Professional. From
the discussion, it seemed to be a hardware problem. That was cleared up, and
I'm glad it was. Vista seems overall to have many peculiar behaviors. e.g.,
taking eight minutes to launch Internet Explorer (that happened once). I
feel that the more information I can get the better; newsgroups are a goods
means to that end.