Fix it now
This is an I/O failure, which is a narrower statement than it looks. It does not say a file is corrupt or missing; it says a read issued during boot did not complete. The realistic explanations are a loose or failing connection, a drive on its way out, or a storage configuration that changed under a machine that was hibernated rather than shut down.
diskpart
list volume
exit
chkdsk X: /f
- Power off completely, unplug, and reseat the drive’s data and power cables. On an M.2 drive, reseat the module and check its retaining screw.
- Remove every USB device, memory card and external drive, then start with nothing but the internal disk attached.
- In firmware setup, confirm the drive is still detected, that it is still in the boot order, and that the storage controller mode has not changed between AHCI, RAID and Intel VMD.
- Boot the recovery environment from a USB stick, find the Windows volume with
diskpartandlist volume, and runchkdskagainst that letter – not against C:, which in recovery is usually something else. - Test the drive with the manufacturer’s utility from another machine before you repair boot files. If the drive is suspect, image it first.
Only once the drive tests clean is it worth rewriting boot files with bcdboot X:\Windows /s S: /f UEFI. Repairing boot files on a failing drive produces a machine that boots twice.
If the machine starts reliably from cold and from a restart, you are done. If not, the next section explains what runs between power on and the Windows logo, and why fast startup complicates it.
Why it happens
Before Windows exists as such, the firmware hands control to the boot manager, which reads its configuration from the EFI system partition and loads the boot loader, the kernel and the boot-start drivers. All of that runs through firmware storage services and a very small driver set, with no error handling worth the name. A read either works or it does not.
That is why this code carries different weight from the corruption codes. 0xC0000102 says the data arrived and made no sense. A missing-file status says the loader found nothing where it looked. An I/O failure says the request itself did not complete, so the fault sits in the path between the processor and the media rather than in a file. Microsoft does not publish a meaning for 0xC00000E9 on any page we could find, so treat the screen as the statement it makes and diagnose the path.
0xC0000006 is the same failure seen from memory management: the required data was not placed into memory because of an I/O error. 0xC0000015 is more specific and more useful – the specified sector does not exist – which points at the media or at a partition structure that no longer describes the disk. Read those two as corroboration when they appear.
0x000000F3 does not belong in the storage story at all, despite turning up in the same searches. Microsoft publishes DISORDERLY_SHUTDOWN as Windows being unable to shut down due to lack of memory: there were no free pages left to continue, because applications were not terminated and drivers were not unloaded, and they kept touching pages after the modified page writer had stopped. If you see it, that is a memory pressure problem on shutdown, not a disk problem on startup.
Fast startup complicates the picture on client machines. What most people call a shutdown writes the kernel and drivers to a hibernation file, so the next power on is a resume. If the storage stack changed while the machine was off, that resume reads stale device state and the first real read fails.
A loose or failing physical connection
You have this one if The drive appears in firmware setup on some starts and not on others, or reseating a cable changes the behaviour even briefly.
- Reseat both ends of the data cable and the power connector, and try a different port and a different power lead.
- Replace the data cable outright. They are cheap, they fail, and a marginal one produces exactly this symptom.
- On M.2 and NVMe drives, remove and refit the module, checking the slot for dust and the standoff and screw for fit.
- On a laptop, check for a swollen battery pressing on the drive assembly.
Fast startup is resuming into a stack that has changed
You have this one if A restart from within Windows works and a full shutdown followed by a power on does not, or the failure began after adding, removing or swapping a drive or a dock.
- If you can reach Windows, run
powercfg /hibernate offfrom an elevated prompt. Microsoft documents that option as disabling the hibernate feature; fast startup depends on the hibernation file, so it stops with it. - You can also clear the tick box under Control Panel, Power Options, Choose what the power buttons do, after choosing Change settings that are currently unavailable.
- Test by shutting down fully and powering on, rather than by restarting – a restart does not use the hibernation file and will pass either way.
Deleting hiberfil.sys by hand from the recovery environment is not a documented procedure. If you resort to it, understand that the suspended session and anything unsaved in it are gone.
The storage controller mode changed in firmware
You have this one if The failure began right after a firmware update, a CMOS reset or a battery replacement, and firmware now shows a different storage mode from the one Windows was installed with.
- Set the mode back to what it was. If you do not know, try each option in turn; you want the one where Windows starts.
- To keep the new mode, make Windows load the alternative driver: start in the old mode, run
bcdedit /set {current} safeboot minimal, restart, change the mode in firmware, let it reach Safe Mode, then runbcdedit /deletevalue {current} safeboot. - Confirm Device Manager then shows the expected storage controller with no warning icons.
On a BitLocker-protected machine, firmware changes can trigger a request for the recovery key. Suspend protection first with manage-bde -protectors -disable C: and have the key to hand. Protection resumes at the next restart unless you pass -rebootcount.
The drive itself is failing
You have this one if chkdsk stalls or reports a growing number of bad sectors, drive health shows reallocated or pending sectors, or the machine hangs for long periods before failing. 0xC0000015 alongside is a strong hint.
- Stop trying to repair it. Every further attempt asks a dying drive to do more work.
- Image it to a healthy disk with a tool that can skip and retry unreadable areas, and recover from the image rather than the original.
- Replace the drive and reinstall or restore from backup.
Full reference
Matching the pattern to the fault
| Pattern | Most likely cause |
|---|---|
| Restart works, shutdown then power on fails | Fast startup resuming into a changed storage stack |
| Sometimes starts, sometimes does not, with no pattern | A marginal cable, connector or drive |
| Started right after a firmware update or CMOS battery change | Controller mode reset in firmware |
| The drive is missing from firmware setup entirely | The drive, its connection or its power |
| 0xC0000015 appears alongside | The media or the partition structure, not the cable |
The codes, and how much each one is really saying
| Code | Status | What to take from it |
|---|---|---|
0xC00000E9 |
Not published by Microsoft | A read issued during boot did not complete. Diagnose the path, not a file |
0xC0000006 |
STATUS_IN_PAGE_ERROR | The required data was not placed into memory because of an I/O error |
0x000000F3 |
DISORDERLY_SHUTDOWN | Windows was unable to shut down due to lack of memory. Not a storage problem |
0xC00000B5 |
Not published by Microsoft | Read it as a device that did not answer in time, and confirm with the System log |
0xC0000015 |
STATUS_NONEXISTENT_SECTOR | The specified sector does not exist. Media or partition structure |
Two of the five are not published in any form we could verify against a Microsoft page, and that is worth knowing rather than papering over: it means nobody can tell you authoritatively what they mean, and any page that gives you a confident one-line definition for them is guessing.
Suspending BitLocker before firmware work
manage-bde -protectors -disable C: suspends protection by making the encryption key available unsecured on the drive, without removing any key protector. Microsoft documents that protection resumes automatically the next time Windows starts unless you specify a reboot count: -rebootcount <0-15>, where 0 suspends indefinitely. Re-enable it deliberately afterwards with manage-bde -protectors -enable C: rather than relying on a restart you might not get.
Getting the data off first
- Stop powering the machine on repeatedly. Each attempt is another set of reads the drive may not survive.
- Remove the drive and attach it to a working machine with a USB adapter, as a secondary device.
- If it mounts, copy what you need immediately, largest and most important first.
- If it does not mount, image it with a tool that retries and skips, then work from the image.
- Only then decide whether the original is worth a professional recovery attempt.
If the drive is suspect, image it before running chkdsk or any repair tool. Repair operations write to the disk, and writing to failing media is the fastest way to turn a recoverable problem into an unrecoverable one.
When the drive tests clean
Then the boot files are the remaining candidate. Assign a letter to the EFI system partition in diskpart, run bcdboot X:\Windows /s S: /f UEFI with X as the Windows volume, and confirm with bcdedit /enum that a boot loader entry now points at the right volume. If that changes nothing, go back and check the connection again: an I/O failure is a hardware statement by default, and a clean self-test is not the same as a healthy cable.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0xC00000E9 |
Reported by the boot manager when a read issued during startup fails. Microsoft publishes no meaning for this value on any page we could reach, so treat it as a statement about the storage path rather than about a file | not published by the vendor |
0xC0000006 |
STATUS_IN_PAGE_ERROR: the required data was not placed into memory because of an I/O error – the same storage failure seen from memory management | Microsoft Learn |
0x000000F3 |
DISORDERLY_SHUTDOWN: Windows was unable to shut down due to lack of memory. No free pages remained because applications were not terminated and drivers were not unloaded. It is not a storage failure | Microsoft Learn |
0xC00000B5 |
Seen alongside stalling storage. Microsoft publishes no meaning for this value on any page we could reach; read it as a device that did not answer within the driver’s timeout and corroborate from the System log | not published by the vendor |
0xC0000015 |
STATUS_NONEXISTENT_SECTOR: {Sector Not Found} The specified sector does not exist, which points at the media or the partition structure | Microsoft Learn |
Confirm the fix worked
- Shut the machine down fully, then power it on cold. Then restart it from within Windows. Both paths must reach the desktop.
chkdsk C: /scanfrom an elevated prompt finds no problems.- The drive vendor’s utility reports no reallocated or pending sectors, and no growth in them a week later.
- No new warnings from the disk or Ntfs sources appear in the System log over the following days.
- If fast startup was the cause, confirm the machine still starts correctly after you decide whether to leave hibernation off.
Questions people ask about this
Can I get my data off before replacing the drive?
Usually, if you stop making the drive work. Take it out, attach it to a working machine with a USB adapter, and copy what you need. If it will not mount, image it first with a tool that retries and skips bad areas.
Does this ever mean Windows is corrupt?
Rarely, and it is the last thing to consider rather than the first. If the drive tests clean, the cables are sound and the firmware settings are right, rebuilding the boot files is worth trying. An I/O failure is a hardware statement by default.
Is it worth running chkdsk straight away?
Only when you are confident the drive is healthy. Microsoft notes that a /r pass reads every sector and takes a considerable time; on a failing drive that is hours of stress on hardware with very little left to give. Image first.
Why does a restart work but a cold start fail?
Because they are different operations. On a client machine, what looks like a shutdown writes the kernel and drivers to a hibernation file and the next power on is a resume, while a restart is a genuine restart. If the storage stack changed while the machine was off, only the resume path fails.
Does fixing this involve buying anything?
No licence and no software. Everything here is built into Windows or is a free download from the drive manufacturer. If the drive is dead you will be buying a drive, and reinstalling Windows on it normally reactivates from the machine’s existing digital licence.
