Fix it now
0x8009030C is SEC_E_LOGON_DENIED, and Microsoft’s published meaning is just that: “The logon attempt failed”. It names no cause. In a hardened environment one common cause is that NTLM is restricted, so the exchange is refused on policy grounds before any file or printer is touched. The NTLM operational log is what tells the two apart.
gpupdate /force
klist
- On both the client and the target, open Event Viewer and go to Applications and Services Logs, Microsoft, Windows, NTLM, Operational. Look for entries naming the server you are trying to reach.
- If that log is empty, the audit policies are not enabled. Turn on the auditing variants of the restrict NTLM settings first, reproduce the failure, then read it again. Audit mode blocks nothing.
- Check the policy itself under Computer Configuration, Windows Settings, Security Settings, Local Policies, Security Options, in the entries beginning with the network security restrict NTLM wording.
- Check whether the user is a member of Protected Users, which changes how that account may authenticate.
- Try connecting by the fully qualified name rather than by address. An address cannot be matched to a service principal name, so it forces the NTLM path and will fail wherever NTLM is blocked.
- Clear stored credentials for that target in Credential Manager, re-enter them by hand, and retry.
Before assuming policy, rule out the ordinary case. A wrong password, a wrong account or a wrong domain produces this same code, and it is much more common than an NTLM restriction on a network nobody has hardened.
If the resource opens by the name people will actually use, you are done. If not, the next section separates the two situations that produce this code.
Why it happens
Before any share is opened, the client and server negotiate through the security support provider interface. They agree a package, usually Kerberos with NTLM as the fallback, and exchange tokens until one side is satisfied. 0x8009030C is that exchange ending in a refusal. The file server never sees a request for a file, which is why permissions on the share are irrelevant to it.
Two very different situations produce the code, and the published meaning does not distinguish them. The first is an ordinary credential rejection: wrong password, wrong account, wrong domain. The second is a policy decision that the package may not be used. The NTLM operational log separates them, because it records the target server name for every attempt it blocks or audits.
The companion codes narrow things further, and one of them is routinely misreported. 0x80090311 is SEC_E_NO_AUTHENTICATING_AUTHORITY, “No authority could be contacted for authentication” – nothing could validate the credentials at all, because no domain controller was reachable or the target belongs to no domain the client trusts. 0x80090302 is SEC_E_UNSUPPORTED_FUNCTION, “The function requested is not supported”: one end lacks a capability the other asked for.
0x8009035D is the one to be careful with. It is widely described as a detected downgrade, and it is not. Microsoft publishes it as SEC_E_INVALID_PARAMETER, “One or more of the parameters passed to the function was invalid” – a malformed call rather than a security judgement. The downgrade code is a different number, 0x80090350, and even that is published as “The system cannot contact a domain controller to service the authentication request”. If you have been sent to check channel binding and extended protection because of 0x8009035D, you have been sent to the wrong place.
NTLM restriction policy is blocking the attempt
You have this one if Entries in the NTLM operational log naming the target, while other protocols to the same server work normally.
- Decide whether the target should be reachable by NTLM at all. If it should, add it to the server exception list rather than switching the restriction off.
- Apply the policy where it belongs: outgoing on the client, incoming on the server, domain-wide on controllers.
- Run
gpupdate /force, reproduce the connection and re-read the log.
Exception lists accept the server name, so every name people use has to be listed or the exception silently misses some of them.
It is an ordinary credential failure after all
You have this one if No NTLM operational log entries name the target, and typing the credential by hand behaves differently from letting Windows supply it.
- Read the security event on the machine that refused the connection, and its sub-status, before touching any policy.
- Clear stored credentials for that host in Credential Manager, including entries under other name forms.
- Retry with the credential typed by hand, so you know whether a saved one was the problem.
The connection is made in a way that cannot use Kerberos
You have this one if The share opens by fully qualified name and fails by address, by an unregistered alias, or across a boundary with no trust.
- Connect by the name that has a service principal name registered for it, normally the machine’s fully qualified name.
- For an alias, register the matching name on the account that answers so Kerberos can be used rather than falling back.
- Where a monitoring tool or script connects by address, change the script rather than relaxing the policy.
The account is in Protected Users
You have this one if One user fails against a particular target while colleagues are unaffected, and the failure follows them to any machine.
- Check membership with
Get-ADUser <name> -Properties MemberOf. - Leave the membership in place and make the target Kerberos-capable, which is the outcome the group exists to encourage.
- Only remove the account from the group if it was added without the consequences being understood, and record why.
Protected Users changes more than one thing about how an account may authenticate. Expect other behaviour to change too when somebody is added, and check what else that account does before you remove them from it.
An old device or application cannot meet the current security level
You have this one if A copier, appliance or legacy application fails while current Windows machines succeed, sometimes after a hardening update.
- Establish what the device can actually do. Many can be updated, and the vendor is the right first call.
- Where the device supports it, move it to a modern authentication method or a service account with a supported protocol.
- Do not lower the authentication level for the whole domain to accommodate one device.
Full reference
The four codes, with their published meanings
| Code | Symbolic name | Published meaning |
|---|---|---|
0x8009030C |
SEC_E_LOGON_DENIED | The logon attempt failed |
0x80090302 |
SEC_E_UNSUPPORTED_FUNCTION | The function requested is not supported |
0x8009035D |
SEC_E_INVALID_PARAMETER | One or more of the parameters passed to the function was invalid |
0x80090311 |
SEC_E_NO_AUTHENTICATING_AUTHORITY | No authority could be contacted for authentication |
Two of those are worth reading twice. 0x8009030C says nothing about NTLM: it says the logon attempt failed, and everything else in this article is about how to find out why. And 0x8009035D is not the downgrade code, whatever you have read – it means a parameter was invalid. Chasing channel binding or extended protection on the strength of that number is chasing the wrong thing.
Narrowing it down
| What you see | Where to look |
|---|---|
| The client’s NTLM log records the target, the server’s log is empty | Outgoing NTLM is restricted on the client |
| The target’s NTLM log records a blocked inbound attempt | Incoming NTLM is restricted on the server or in the domain |
| Neither NTLM log names the target at all | Not a policy problem. Read the security event and its sub-status |
| It fails only for one user, everywhere | That account’s own restrictions, including Protected Users membership |
| It works by name and fails by address | Kerberos handles the name; the address falls back to NTLM, which is blocked |
| It began after a security update on the target | Something the target now requires that this client cannot meet |
Turning the NTLM log on before you need it
The NTLM operational log fills only when the audit or restriction policies are enabled, which is why it is empty on the day you first need it. The audit variants block nothing and cost nothing, and having them on across the estate turns this entire class of problem from an argument into a log query.
- Enable the audit variants of the restrict NTLM settings under Security Options, on clients, servers and domain controllers.
- Leave them running for a week and read what the estate is actually doing. Most people are surprised.
- Build the exception list from what the log shows rather than from what people remember.
- Only then move from audit to restriction, one scope at a time.
- Keep the audit setting on afterwards, because it is what tells you what a restriction broke.
Adding an exception properly
- Add the specific server, not a wildcard, and add every name form people use to reach it.
- Put the exception in the same policy object as the restriction, so the two are found together by whoever inherits this.
- Record why the exception exists and what would let it be removed. An exception with no reason attached becomes permanent.
- Prefer making the target Kerberos-capable. An exception keeps a weaker protocol alive; a registered service principal name removes the need for it.
Switching an NTLM restriction off to make one connection work removes the protection for every connection that machine makes. Those settings exist because NTLM is the weaker protocol and the route many credential attacks take. An exception for a named server keeps the protection everywhere else and leaves a record of what you allowed and why.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0x8009030C |
SEC_E_LOGON_DENIED: the logon attempt failed. That is the whole published meaning – it names no cause, and NTLM policy is one situation among several that produces it | Microsoft Learn |
0x80090302 |
SEC_E_UNSUPPORTED_FUNCTION: the function requested is not supported | Microsoft Learn |
0x8009035D |
SEC_E_INVALID_PARAMETER: one or more of the parameters passed to the function was invalid. This is not the downgrade code, despite being widely described as one | Microsoft Learn |
0x80090311 |
SEC_E_NO_AUTHENTICATING_AUTHORITY: no authority could be contacted for authentication | Microsoft Learn |
Confirm the fix worked
- Reproduce the connection and confirm the NTLM operational log records no new block for that target.
- Confirm the resource opens by the name people will actually use, not only by a workaround name.
- Run
kliston the client and confirm a Kerberos ticket for the target was issued, where Kerberos is the intended outcome. - Test from a second machine, to prove the fix was to policy or to the target rather than to one client.
- Confirm any exception you added is recorded with a reason, in whatever holds your policy documentation.
Questions people ask about this
Does anything here need buying?
No. Policy settings, the NTLM operational log and Credential Manager are all part of Windows. The only cost is the time to work out whether the target should be reachable by NTLM at all.
Does 0x8009030C mean NTLM is blocked?
Not by itself. Microsoft publishes it as SEC_E_LOGON_DENIED, “The logon attempt failed”, and that is all it means. An NTLM restriction is one thing that produces it; a wrong password is another, and on most networks the second is far more likely. Read the NTLM operational log before you conclude anything.
Should I just switch the NTLM restriction off?
No. Those settings exist because NTLM is the weaker protocol and is the route many credential attacks take. Adding a specific exception for a specific server keeps the protection everywhere else and leaves a record of what you allowed and why.
Why is the NTLM log empty?
Because it only fills when the audit or restriction policies are enabled. Turn on the audit variants first, reproduce the failure, then read the log. Audit mode blocks nothing and is safe to leave on.
I was told 0x8009035D means a downgrade was detected.
It does not. Microsoft publishes it as SEC_E_INVALID_PARAMETER: one or more of the parameters passed to the function was invalid. The downgrade code is 0x80090350, and Microsoft publishes that one as being about not reaching a domain controller. Neither of them sends you to channel binding.
