Skip to main content

TECH VEDA

Corporate on-site training - Submit proposal Pick your modulesSharpen your kernel skills: deep dives, drivers, Yocto, CVEs, careers — updated daily. Read the blog →
Fix Guides

Unable to Mount Root fs on unknown-block(0,0): Which Stage of the Root Mount Failed?

VFS: Unable to mount root fs on unknown-block(0,0): what the two numbers mean, six causes on embedded boards, one check for each, and the fix.

Unable to Mount Root fs on unknown-block(0,0): Which Stage of the Root Mount Failed?

This panic means the kernel could not mount a root filesystem on the device named by root=. Read the two numbers first. unknown-block(0,0) means no device number was assigned, most often because the root= string is in a form the kernel does not accept or because the storage device was not registered when the kernel looked for it. The first thing to check is the partition list printed immediately above the panic line.

The kernel prints VFS: Unable to mount root fs at the end of its own boot sequence, when drivers are initialised and it cannot attach the filesystem that holds init. It has several unrelated causes, and rootwait addresses only one. Every check here is made on the failing board’s serial console or on the build host.

The message: Unable to mount root fs on unknown-block(0,0)

The panic line as the kernel formats it, with both numbers at zero:

Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)

The two numbers are the major and minor device number that the kernel derived from root=. One of two pairs of lines is printed above the Unable to mount root fs panic, and it shows which stage failed. These are the kernel’s message formats, not output captured from a board; %s and %d stand for the name and the error number:

VFS: Cannot open root device "%s" or %s: error %d
Please append a correct "root=" boot option; here are the available partitions:
List of all partitions:
No filesystem could mount root, tried: 

The first pair means the device could not be opened, or the filesystem type named in rootfstype= is unknown to the kernel. The second means the mount was attempted and every filesystem that was tried rejected the contents. In both cases the kernel then lists the block devices it can see. From v6.7 the panic after the second pair reads VFS: Unable to mount root fs on "%s" or %s.

Read the numbers as follows:

  • (0,0): no device number was assigned. The name was not understood, or the device did not exist yet.
  • A non-zero major number, such as (179,2): root= was translated to a device number. 179 is the major number of MMC and SD block devices.
  • (0,253): from v6.5, the placeholder for a root= that starts with mtd or ubi. Before v6.5 such a root shows (0,0).

Quick diagnosis

CauseHow to confirmFix
1. The root= string is not in a form the kernel resolvesNumbers are (0,0); read the Kernel command line: log lines; from v6.5, with rootwait set, look for Disabling rootwait; root= is invalid.Use /dev/<name>, PARTUUID= or PARTLABEL=
2. The storage device is registered after the kernel looks for itNumbers are (0,0); the disk is missing from the partition list above the panic; with a wait added, the log shows Waiting for root device and then VFS: Mounted rootrootwait=N from v6.6; bare rootwait before that
3. The storage controller driver never bindsThe disk is still missing after a bounded wait; boot with deferred_probe_timeout=5 rootwait=15 and read the deferred probe pending linesBuild the driver and its suppliers in; correct the device tree
4. The device is found but no built-in filesystem accepts itThe major number is not 0; the log contains No filesystem could mount root, tried:Build the filesystem driver in; correct root=; rewrite the image
5. The image expects an initramfs that was not loadedroot=UUID= or root=LABEL= on the command line; no /init in the archive the bootloader loadsLoad the initramfs, or use a root= form the kernel resolves
6. The root is on UBI or MTD and the volume is not availableroot= starts with ubi or mtd; numbers are (0,253) from v6.5Correct ubi.mtd= and rootfstype=; build the drivers in

Cause 1: the root= string is not in a form the kernel resolves

Without an initramfs, the kernel itself has to turn root= into a device, and it accepts only these forms:

  • root=/dev/mmcblk0p2: a device name with the /dev/ prefix.
  • root=PARTUUID=<uuid>: the partition UUID from the partition table.
  • root=PARTLABEL=<name>: a GPT partition label.
  • root=179:2 or a bare hexadecimal number: a device number.

root=UUID= and root=LABEL=, which refer to the filesystem and are common on desktop distributions, are not in that list. Neither is a name without the prefix, such as root=mmcblk0p2. The kernel cannot resolve any of these, and the Unable to mount root fs panic reports (0,0).

From v6.5 the same mistake also switches rootwait off. The kernel prints this fixed line, so adding rootwait changes nothing:

Disabling rootwait; root= is invalid.

To confirm, find the line or lines beginning Kernel command line: in the boot log and read root= as the kernel received it. Then search the log for the line above. Also compare root= with the partition list above the panic: each line gives the device number in hexadecimal, the size, the name and the partition UUID.

The fix is to write one of the accepted forms. PARTUUID= does not depend on the order in which devices are detected. Write the complete UUID, because a shortened value can match another partition. Avoid the numeric forms for a device that may appear late: a number is accepted without checking that the device exists, and rootwait then does not wait.

Cause 2: the storage device is registered after the kernel looks for it

USB and MMC devices are detected asynchronously, so the controller can be ready before its card or disk has been registered. The kernel resolves root= once, and without rootwait it goes on to the mount with no device.

To confirm, read the partition list above the panic. If the root disk is absent, boot once more with a wait added to the command line: rootwait=10 from v6.6, bare rootwait before that. If the console then shows Waiting for root device followed by the name, and after it VFS: Mounted root, the device was registered late.

The fix is a bounded wait. From v6.6, rootwait=N waits at most N seconds, continues as soon as the device appears, and otherwise goes on to the same panic. Bare rootwait never expires. Kernels before v6.6 do not recognise rootwait=N and do not wait at all when it is given, so use bare rootwait there. rootdelay=N is different: a fixed sleep of N seconds on every boot, whether or not the device is present.

Warning: with bare rootwait, a device that never appears leaves the board waiting indefinitely with no panic and no reset. Make sure a watchdog covers that state in a product.

Cause 3: the storage controller driver never binds

If the partition list is empty, the kernel has no usable block device at all. A wait does not help when the controller driver is not built in, is a module stored on the root filesystem itself, or is waiting for a clock, regulator or pin controller that is never registered.

The kernel can list the devices that are still waiting for a dependency, but only when its deferred probe timeout expires, and the panic normally comes first. To confirm, keep the kernel waiting past the timeout by booting with deferred_probe_timeout=5 rootwait=15 (bare rootwait before v6.6), and read the lines containing deferred probe pending. They appear on the failing console and name each waiting device.

If nothing is pending, check in the kernel build directory that the driver is built in (=y, not =m). The host controller symbol depends on the SoC:

raghu@techveda.org:~$ grep -E 'CONFIG_MMC=|CONFIG_MMC_BLOCK=' .config

The fix is to set the storage driver and everything it depends on to y, or to correct the device tree node. The kernel documentation describes deferred_probe_timeout= as a debugging option that makes some dependencies be ignored; remove it after the test.

Cause 4: the device is found but no built-in filesystem accepts it

Without rootfstype= the kernel tries every built-in filesystem that works on a block device; with it, only the comma-separated names you gave. The filesystems normally stay quiet about contents they do not recognise, so the result is Unable to mount root fs with a non-zero major number and little else.

To confirm, read the names after No filesystem could mount root, tried:. If your image’s filesystem is absent, its driver is not built in or rootfstype= excludes it:

raghu@techveda.org:~$ grep -E 'CONFIG_EXT4_FS=|CONFIG_SQUASHFS=' .config

Two variants print Cannot open root device instead: a rootfstype= that names a driver not built in, which from v6.5 is followed by List of all bdev filesystems:, and a root= partition that does not exist on a disk that does.

If the filesystem was tried, the partition holds something else. The fix is to enable the driver as y, correct root=, or write the image again.

Warning: writing an image to a partition destroys its current contents. Confirm the target device on the host with lsblk first.

Cause 5: the image expects an initramfs that was not loaded

When the kernel is given an initramfs that contains /init, or the program named by rdinit=, it does not interpret root= at all. The program in the initramfs does, and it accepts UUID= and LABEL=. If the bootloader stops loading that initramfs, or the archive no longer contains /init, the kernel interprets root= itself and fails as in Cause 1.

To confirm, list the archive on the build host (gzip-compressed cpio assumed), and check that the boot command still passes it to the kernel:

raghu@techveda.org:~$ zcat initramfs.cpio.gz | cpio -t | grep -x -e init -e ./init -e /init

The fix is to restore the initramfs in the boot command or to use a root= form the kernel resolves.

Cause 6: the root is on UBI or MTD and the volume is not available

rootwait has no effect on a root that starts with mtd or ubi. If the UBI device is not attached or the filesystem type is not offered, the Unable to mount root fs panic follows.

To confirm, read the command line. The kernel documentation for UBIFS gives the form ubi.mtd=0 root=ubi0:rootfs rootfstype=ubifs. rootfstype=ubifs is required: without it the kernel tries only filesystems that work on a block device, and UBIFS is not one of them. Then check the drivers:

raghu@techveda.org:~$ grep -E 'CONFIG_MTD_UBI=|CONFIG_UBIFS_FS=' .config

The fix is to correct those parameters and to set both symbols to y.

If none of these

Capture one complete boot log on a serial console with ignore_loglevel and initcall_debug on the command line.

Check also that the message is Unable to mount root fs and not No working init found., which is printed after the root filesystem has mounted and means that no init program could be executed. For the order of these steps, see the TECH VEDA article From start_kernel() to Userspace.

Then ask with the full log: the linux-mmc list for MMC and SD, the linux-mtd list for raw flash, or the vendor support channel for a vendor kernel.

Versions

Releases read: v5.10.274, v5.15.225, v6.1.192, v6.6.162, v6.12.118, v6.18.64, v7.3.5 and v7.4-rc3.

  • Panic text. Before v6.7 both cases print VFS: Unable to mount root fs on %s. v7.3.5 and v7.4-rc3 also have VFS: Unable to mount root fs: not enough memory for a failed allocation.
  • Invalid root= with rootwait. On v6.1.192 and the older longterm releases, Disabling rootwait; root= is invalid. is printed only for a malformed PARTUUID=. There, root=mmcblk0p2 with rootwait waits indefinitely and does not panic.
  • Deferred probe report. The default timeout is 0 in v5.10.274 and v5.15.225. From v6.1.192 it is 10 seconds when CONFIG_MODULES is set and 0 otherwise; v7.3.5 and v7.4-rc3 make that default configurable as CONFIG_DRIVER_DEFERRED_PROBE_TIMEOUT. With 0, no list is printed.

What was checked and what could not be verified

Every message, parameter and configuration symbol here was read in the kernel source and documentation at the releases above on 19 November 2026. Nothing was run on hardware or in an emulator. The behaviour of root=UUID= without an initramfs, the order of the deferred probe report and the panic, and the two variants in Cause 4 are readings of the code, not observed boots. No error number or device timing is given, because none was measured.

Key takeaways

  • Read the numbers first. (0,0) is a naming or availability problem, or a flash root before v6.5; (0,253) is a flash root; with a non-zero major number, the lines above the Unable to mount root fs panic say whether the device could not be opened or no filesystem accepted it.
  • The kernel resolves /dev/ names, PARTUUID=, PARTLABEL= and device numbers. UUID= and LABEL= need an initramfs.
  • From v6.6, prefer rootwait=N to bare rootwait and to rootdelay=; older kernels have only the bare form.
Was this worth your time?

Frequently asked questions

What does unknown-block(0,0) mean in Unable to mount root fs?

The two numbers are the major and minor device number that the kernel derived from root=. Both are zero when no device number was assigned, because the root= string is not in a form the kernel accepts or the device was not yet registered. Before v6.5 a root on UBI or MTD also shows (0,0).

Why does adding rootwait not change the panic?

From v6.5, a root= string that the kernel cannot resolve makes it print “Disabling rootwait; root= is invalid.” and switch the wait off. The wait also has no effect when root= begins with mtd or ubi or is a device number.

What does “No filesystem could mount root, tried:” tell me?

The mount was attempted and every filesystem that was tried rejected the contents. If your image’s filesystem is not among the names after the colon, its driver is not built in or rootfstype= excludes it.

References

RB
Raghu Bharadwaj

Founder, TECH VEDA — 20+ years teaching the Linux kernel, device drivers and embedded systems.

Follow on LinkedIn

Get new posts by email

Kernel, embedded Linux and AI-era engineering — a few sharp reads a month. No spam.

We email occasionally and never share your address.