If broken it is, fix it you should

Using the powers of the debugger to solve the problems of the world - and a bag of chips    by Tess Ferrandez, ASP.NET Escalation Engineer (Microsoft)

A Case of Invalid Viewstate

A Case of Invalid Viewstate

  • Comments 38

Last week I was helping a colleague of mine with a viewstate case that turned out to be pretty interesting...

Scenario

The customer was getting events similar to the following in the eventlog and needed to know why they occurred

Event Type: Information
Event Source: ASP.NET 2.0.50727.0
Event Category: Web Event
Event ID: 1316
Date: 2007-06-11
Time: 09:48:02
User: N/A
Computer: MYMACHINE

Description:

Event code: 4009
Event message: Viewstate verification failed. Reason: The viewstate supplied failed integrity check.
Event time: 2007-06-11 09:48:02
Event time (UTC): 2007-06-11 07:48:02
Event ID: 14cc57c05e834de98c7df506a013a706
Event sequence: 10
Event occurrence: 1
Event detail code: 50203

Application information:

Application domain: /LM/w3svc/1/root/MyApp-3-128260216693527710
Trust level: Full
Application Virtual Path: /MyApp
Application Path: c:\inetpub\wwwroot\MyApp\
Machine name: MYMACHINE

Process information:

Process ID: 3640
Process name: w3wp.exe
Account name: NT AUTHORITY\NETWORK SERVICE
Request information:
Request URL: http://mymachine/MyApp/MyWebForm.aspx
Request path: /TestBadViewstate/WebForm1.aspx
User host address: 127.0.0.1
User:
Is authenticated: False
Authentication Type:
Thread account name: NT AUTHORITY\NETWORK SERVICE

ViewStateException information:

Exception message: Invalid viewstate.
Client IP: 127.0.0.1
Port: 14644
User-Agent: Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 5.2; WOW64; .NET CLR 2.0.50727; .NET CLR 1.1.4322; .NET CLR 3.0.04324.17; InfoPath.2)
PersistedState:
dDwxOTM2NzUxODMzO3Q8O2w8aTwxPjs+O2w8dDw7bDxpPDE+Oz47bDx0PHA8bDxfX0hlYWRpbmc7PjtsPENvdXJzZSBTZWFyY2g7Pj47bDxpPDE+Oz47bDx0PDtsPGk8MT47aTwzPjs+O2w...
Referer: http://SomeThirdPartySite.com/SomePage.htm
Path: /MyApp/MyWebForm.aspx

Custom event details:

For more information, see Help and Support Center at http://go.microsoft.com/fwlink/events.asp.

 

And the callstack reported was:

[HttpException (0x80004005): Validation of viewstate MAC failed. If this application is hosted by a Web Farm or cluster, ensure that <machineKey> 
configuration specifies the same validationKey and validation algorithm. AutoGenerate cannot be used in a cluster.]

System.Web.UI.ViewStateException.ThrowError(Exception inner, String persistedState, String errorPageMessage, Boolean macValidationError) +119
System.Web.UI.ObjectStateFormatter.Deserialize(String inputString) +252
System.Web.UI.ObjectStateFormatter.System.Web.UI.IStateFormatter.Deserialize(String serializedState) +5
System.Web.UI.Util.DeserializeWithAssert(IStateFormatter formatter, String serializedState) +37
System.Web.UI.HiddenFieldPageStatePersister.Load() +222
System.Web.UI.Page.LoadPageStateFromPersistenceMedium() +80
System.Web.UI.Page.LoadAllState() +35
System.Web.UI.Page.ProcessRequestMain(Boolean includeStagesBeforeAsyncPoint, Boolean includeStagesAfterAsyncPoint) +9041
System.Web.UI.Page.ProcessRequest(Boolean includeStagesBeforeAsyncPoint, Boolean includeStagesAfterAsyncPoint) +217
System.Web.UI.Page.ProcessRequest() +85
System.Web.UI.Page.ProcessRequestWithNoAssert(HttpContext context) +20
System.Web.UI.Page.ProcessRequest(HttpContext context) +110
ASP.MyPage_aspx.ProcessRequest(HttpContext context) +30
System.Web.CallHandlerExecutionStep.System.Web.HttpApplication.IExecutionStep.Execute() +405
System.Web.HttpApplication.ExecuteStep(IExecutionStep step, Boolean& completedSynchronously) +65

 

 

Troubleshooting

If you get viewstate errors on 1.1. and dont get events like the above, make sure that you have SP1 installed or at least a version of 1.1 later than http://support.microsoft.com/?id=831150  since this hotfix introduced the type of logging above, which is crucial to troubleshooting most viewstate errors.

Viewstate, as most of you know, is a Base64 encoded string containing information about the state of the controls on the webform.  To avoid tampering the viewstate is validated against the machine key, pagename etc. and if the string is either corrupt in some way such that it cant be Base64 decoded or such that it doesn't pass validation you will get an error like the above.

Typically viewstate errors will occur if

  • You're on a web farm and the machine keys are not consistent on all nodes
  • You're on a web farm and the page is slightly different on one node than ot the others, for example if one node is upgraded to 2.0 and the other is still 1.1 or if some controls used on the page are built differently

For a more comprehensive list see: http://support.microsoft.com/?id=829743

 

In this particular case they had just upgraded to 2.0 from 1.1, they were not running on a web farm.  The error occurred intermittently and only on two specific pages, but most of the time these pages were served just fine without viewstate errors.

Since it is very unusual for errors like this to occur in a non-web farm scenario, especially intermittently, this issue was very curious.   I have seen that happen sometimes when the client was a mobile device and the viewstate was very large as some mobile devices will only send x kb of data so the issue there would be that the viewstate was corrupted because not all the viewstate was sent.   In this case the client was IE 7.0 so no such luck...

I used Fritz Onion's viewstatedecoderr to decode the viewstate to verify that it wasn't corrupted and sure enough the viewstate was just fine, the contents seemed to be fairly standard textboxes etc.

Lucky break

Lucky for me, and for them:)  I have seen enough encoded viewstate to know what viewstate is supposed to look like, so when i took another look at the eventlog entry i noticed that the persisted state started with dDw...    to most people this probably just looks like mumbojumbo but I happen to know that viewstate on 2.0 is supposed to start with /wE...  in fact when I browsed their site on the internet I could indeed see that the typical viewstate for these affected pages (from the hidden viewstate field in view source) was perfectly normal 2.0 viewstate while the dDw... that was sent to the page is typical for 1.1 viewstate.

Knowing this, the question is now, where does this 1.1. viewstate come from?   This is where the eventlog entry comes in handy again... and if you have jumped ahead a bit you have probably noticed by now that the referer is a htm page http://somethirdpartysite.com/SomePage.htm, in other words. there is nothing wrong with the site itself, the problem is that somehow this 3rd party site is posting 1.1. viewstate to the page, and of course that won't validate well against the new 2.0 page.

Browsing to the 3rd party site and looking at the source for the htm file we can see what we already suspected... the page contains a form, with a hidden viewstate field and the post action for the form is our aspx page, so this 3rd party site apparently did a bit of a hack, copying the viewstate while this site was still 1.1. and posting that viewstate to the form... 

Technically, noone but the aspx page itself should really be allowed to post to the aspx page, so anytime the referer of a post to the aspx page is someone other than itself that is bad, and again, to guard against things like this the viewstate validation exists, but of course it might prove to cause a bit of a troubleshooting headache for you.

 

Conclusion

The solution in this case was to talk to the 3rd party site, let them know about the upgrade and ask them to avoid posting to the site but rather link to the site or redirect to it since there is no way of knowing how long viewstate will be valid as it can change with any type of upgrade to the page.

If there is a benefit to having the 3rd party site being able to post data and show the results in their proprietary format, a web service would be a pretty good solution.

 

Until next time

Tess





  • As a regular reader of your blog, I was interested in read about this viewstate problem, as we were experience a very, very similar problem last week. The more I read, the more it sounded like our problem. The funny thing is it turned out to be us, and we had contacted PSS over this exact problem.

    Anyway, much thanks for your assistance is solving this problem and for the tips on decoding viewstate and different prefixing of viewstate in different .NET versions.

  • :) that is funny,  I hope you don't mind me sharing the story...  

    I thought it was a really interesting case and made for a good gateway to discussing other common viewstate issues.

  • Apple Safari for Windows and Microsoft Silverlight [Via: interactive ] Refactoring Dumb, Dumber, Dumbest...

  • Last update: June 13 , 2007 Document version 0.6 Preface If you have something to add, or want to take

  • Hi Tess, I've been having a very similar issue to what you've posted except I have no 3rd party websites posting to a ASP.NET 2.0 page.

    The referer and the Request URL are both the same (and the Event Log entry matches this).

    It's currently happening to two pages (as far as I can tell), and the persisted state for both don't start with /wE. The first erroring page has ViewState starting with "sod", and the second with "47x", both of which are obviously different to the two you've listed for .NET 1.1/2.0.

    Any help would be much appreciated. Thanks!

  • Hi William,

    I am not sure where you are grabbing the state from, if it is from the persistedstate that does seem a bit unusual that it would start with sod and 47x...  

    It is a bit hard to say what it could be without digging into it deeper, but have you checked to see that there is no mention of clients disconnecting?  A pretty common reason for "corrupt" viewstate, is that the page takes so long to load that clients shut down their viewstate prior to sending the whole post data, causing the viewstate not to be sent through properly.  However in such cases the client will never see the error, but it is still a concern since it is usually caused by really slow pages.

    Thanks

    Tess

  • Hi Tess, I am taking it straight from the persisted viewstate.

    On occassion there is the odd "the client has disconnected" message come up, it just doesn't come up as often as the Invalid viewstate message.

    Other details: I'm running ASP.NET Ajax on both of those pages. And on a third page that contains ASP.NET Ajax I can generate this invalid viewstate problem just by changing the selected index for a drop down box. The DropDownBox triggers an autopostback, which updates a GridView. One of the pages has a 3rd party component compiled against ASP.NET 1.1 (http://www.carlosag.net/Tools/WebChart/Default.aspx), could this be affecting it?

  • I am in the same boat as William, except that my viewstate does start with /wEP. It was a postback from one of my own pages. Interestingly enough, the viewstate from the error would only decode properly using the 2.0 version of fritz's viewstate decoder.

    I too am running Ajax, but no third party control on the page in question.

    Here is my exact viewstate, there is nothing sensitive in it, so decode away:

    /wEPDwUJNjIxMzEzMTU0D2QWAmYPZBYCAgMPZBYCAgMPZBYCAgkPZBYCZg9kFgJmDw8WAh4ISW1hZ2VVcmwFKy4uL0FwcF9UaGVtZXMvZGVmYXVsdC9pbWFnZXMvbmVlZF9mbGFzaC5wbmdkZGQyhrbmsVJZRX/8lG0EOlJrO+vVYg==

  • If you have a component compiled against the 1.1 framework, storing its own viewstate that could potentially cause viewstate errors.  I would try a modified version of the page without this control and see if you still get the viewstate errors...

  • Unfortunately I do not have one to remove. :( The only 3rd party control on this page's control tree is an atlas scriptmanager.

  • I just noticed. Every single error that this occurred on, was from Safari on a Mac. Not sure it is relevant. Just seems odd.

  • Hi Tim,

    The comment regarding compiled against 1.1. was based on that it was the same issue as william reported.

    The fact that it always occurrs from Safari on Mac is definitely interesting,  I'm not too familiar with the details of this browser but it might be that this page has very large viewstate and that there is a limit to how much Safari will post,  a lot of browsers have such a limit, especially on mobile devices (not saying that Mac is a mobile device though:)),  in any case i would say it is extremely relevant unless your site is specifically targeted to users on Mac so that the safari browser for Mac is highly represented amongst the clients.

  • You are correct Tess, I think that is what I am experiencing. This is the user agent for every single error:

    Mozilla/5.0 (Macintosh; U; PPC Mac OS X; en) AppleWebKit/522.11.1 (KHTML, like Gecko) Version/3.0.3 Safari/522.12.1

    Turns out they are using the beta 3 of Safari which we did not test against. The viewstate for the page in question is quite large. I am going to try to figure out a way around this.

  • This is a very interesting post and a good discussion.

    One more question about the logged viewstate error though... in the exception itself, what does the port number mean?

    Exception message: Invalid viewstate.

    Client IP: 127.0.0.1

    Port: 14644

    User-Agent: Mozilla/4.0...

    When I see these viewstate errors on my webfarm, I see various ports listed, all of which seem to be non-standard. Any help would be appreciated.

  • The port is the port used by the client to communicate with the server, which uniquely identifies the client.  

    For example if you have two browsers open going to the same site, each of them will open its own socket and get its own port (typically a port number over 5000) and communicate over this.

    Practically for troubleshooting viewstate errors knowing the client port is only interesting if you are grabbing network traces as well.

    The fact that the client ip is localhost is a bit curious, but I am guessing that is because you are testing from the server itself...

    Thanks

    Tess

Page 1 of 3 (38 items) 123
Leave a Comment
  • Please add 5 and 1 and type the answer here:
  • Post