CVE-2022-33103 is an out-of-bounds write in the SquashFS driver of the U-Boot bootloader, versions 2020.10 to 2022.07-rc3: sqfs_readdir() copied directory names of up to 65535 bytes into a 256-byte buffer. A crafted SquashFS image can corrupt bootloader memory. NVD scores it 7.8; U-Boot 2022.07 fixes it, and only builds with SquashFS enabled are exposed.
Vulnerability summary
CVSS base score
7.8
High severity
EPSS score (FIRST)
0.5%
Probability of exploitation in the wild in the next 30 days
CISA KEV catalog
No
Not listed
Published
Vulnerability ID: CVE-2022-33103 · Source: NVD
Affected components
| Component | CPE | Version range |
|---|---|---|
| Das U-Boot bootloader (SquashFS driver, sqfs_readdir) | cpe:2.3:a:denx:u-boot:*:*:*:*:*:*:*:* | 2020.10 up to (excluding) 2022.07, plus 2022.07-rc1 to rc3 (NVD); first fixed release 2022.07 (commit 2ac0baab, in v2022.07-rc5). Only builds with CONFIG_FS_SQUASHFS |
| Debian u-boot package | pkg:deb/debian/u-boot | buster not affected; bullseye fixed in 2021.01+dfsg-5+deb11u1 (DLA-4150-1, 2025-05-01); unstable fixed in 2022.07+dfsg-1; bookworm and later fixed (Debian security tracker, as of 2026-11-20) |
| Buildroot boot/uboot (default 'latest version') | — | U-Boot 2020.10 (Buildroot 2020.11) to 2022.04 (Buildroot 2022.08 to 2023.11 and the 2023.02.x branch, no patches); 2024.01 from Buildroot 2024.02 (commit 083b03ae). Board-pinned U-Boot versions need a separate check |
| Yocto Project poky meta/recipes-bsp/u-boot | — | kirkstone (4.0) builds 2022.01 with the backported fix; langdale (4.1) builds 2022.07; dunfell (3.1) builds 2020.01, which predates SquashFS support (as of 2026-11-20) |
What does this mean for affected devices?
NVD's description is one line: "Das U-Boot from v2020.10 to v2022.07-rc3 was discovered to contain an out-of-bounds write via the function sqfs_readdir()." sqfs_readdir() lists directory entries in U-Boot's SquashFS driver, the code behind commands such as ls and load on a SquashFS partition. The record was published on 2022-07-01 (NVD, CVE record).
What does the fix commit change?
The fix is commit 2ac0baab by Miquel Raynal of Bootlin ("fs/squashfs: sqfs_read: Prevent arbitrary code execution"), crediting Jincheng Wang as reporter and tester. The commit message explains the bug: "an out-of-band write leading to arbitrary code execution is possible because on one side the squashfs logic accepts directory names up to 65535 bytes (u16), while U-Boot fs logic accepts directory names up to 255 bytes long." Before the fix, sqfs_readdir() copied the entry name with strncpy() using the length stored in the image, into the 256-byte name field of struct fs_dirent. The patch adds FS_DIRENT_NAME_LEN (256) to include/fs.h and caps the copy at 255 bytes plus the terminator.
The commit is in tag v2022.07-rc5 but not in v2022.07-rc4, so the first release with the fix is U-Boot 2022.07. SquashFS support itself arrived in v2020.10, which is where the affected range starts.
How severe is CVE-2022-33103?
NVD's Primary score is CVSS v3.1 7.8 (CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H); v2 is 4.6. The CNA, MITRE, published no CVSS score. NVD lists the weakness as CWE-787 (out-of-bounds write). NVD's CPE range is U-Boot 2020.10 up to (excluding) 2022.07, plus 2022.07-rc1 to rc3.
The practical attacker is someone who can put a crafted SquashFS image where the bootloader will list it: removable media, a writable partition or an update image. The driver only exists in builds with CONFIG_FS_SQUASHFS, which has no default and is selected by CONFIG_CMD_SQUASHFS. On a device with verified boot, memory corruption in the bootloader before the kernel signature is checked is the case that matters.
Which firmware builds carry it?
Any product with SquashFS enabled in a U-Boot from 2020.10 to before 2022.07, unless the patch is backported. The Siemens ProductCERT advisory SSA-577017 that lists 23 U-Boot CVEs for RUGGEDCOM ROX devices (republished by CISA as ICSA-26-134-16) does not include this CVE, and we found no device vendor or CISA advisory naming products for it as of 2026-11-20. Downstream status:
- Debian marked buster not affected (SquashFS support added in 2020.10), fixed bullseye in 2021.01+dfsg-5+deb11u1 with DLA-4150-1 of 2025-05-01, and fixed unstable in 2022.07+dfsg-1; bookworm, trixie and later ship fixed packages (Debian security tracker).
- Buildroot's default U-Boot ("latest version") was inside the affected range from Buildroot 2020.11 (U-Boot 2020.10) to 2023.11, which still defaulted to U-Boot 2022.04 with no patches in
boot/uboot; the 2023.02.x LTS branch also stayed on 2022.04. The bump to 2024.01 shipped in Buildroot 2024.02; its commit message notes that "All in-tree configs with the uboot bootloader use a custom version", so check the version your board config pins, not only the default. - Yocto Project: the poky kirkstone recipe builds U-Boot 2022.01 with the backported fix (patch header
CVE: CVE-2022-33103), alongside patches for CVE-2022-33967 and the 2024 SquashFS and ext4 CVEs. Langdale builds 2022.07, which includes the fix; dunfell's 2020.01 predates SquashFS support.
How do you fix CVE-2022-33103?
Move to U-Boot 2022.07 or later, or cherry-pick commit 2ac0baab into your fork. If the product never reads SquashFS in the bootloader, disabling CONFIG_CMD_SQUASHFS and CONFIG_FS_SQUASHFS removes the code. Later SquashFS bugs such as CVE-2024-57254 and the ext4 flaw CVE-2024-57256 need U-Boot 2025.01 or backports, so a 2022.07 base is not enough. Our U-Boot security guide lists this CVE in its version-matching table; other U-Boot flaws on this site include CVE-2022-34835, CVE-2022-30552 and CVE-2022-30790.
Is CVE-2022-33103 exploited?
It is not in the CISA KEV catalog as of 2026-11-20 (catalog version 2026.11.18, 1,786 entries). FIRST EPSS gives it 0.00459, the 40.5th percentile, in the 2026-11-19 dataset (CISA KEV, FIRST EPSS).
Evidence
- NVD (read 2026-11-20): status Modified; published 2022-07-01T12:15:08.223; NVD Primary CVSS v3.1 7.8 CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, v2 4.6; NVD Primary CWE-787; CPE denx:u-boot from 2020.10 up to (excluding) 2022.07 plus 2022.07-rc1, rc2, rc3; references DLA-4150-1 and two lore.kernel.org messages. — NVD
- CVE record (checked 2026-11-20): state PUBLISHED, CNA MITRE, published 2022-07-01; no CNA CVSS metric and no CNA CWE. — CVE.org
- U-Boot commit 2ac0baab (Miquel Raynal, authored 2022-06-09, committed 2022-06-16; Reported-by and Tested-by Jincheng Wang), read 2026-11-20: squashfs accepts directory names up to 65535 bytes (u16) while U-Boot fs logic accepts 255; adds FS_DIRENT_NAME_LEN 256 and caps the strncpy in sqfs_readdir(). GitHub compare shows the commit in v2022.07-rc5 and v2022.07, not in v2022.07-rc4. SquashFS commit c5100613037 is in v2020.10, not v2020.07. FS_SQUASHFS Kconfig has no default; CMD_SQUASHFS selects it. — U-Boot repository
- Debian security tracker and DLA-4150-1 (read 2026-11-20): buster not-affected (SquashFS support added in 2020.10); bullseye fixed in 2021.01+dfsg-5+deb11u1 (DLA-4150-1, Daniel Leidert, 2025-05-01); unstable fixed in 2022.07+dfsg-1 (bug 1014528); bookworm 2023.01+dfsg-2+deb12u3, trixie 2025.01-3+deb13u1 fixed; tracker cites fix 2ac0baab (v2022.07-rc5). — Debian
- Buildroot (read 2026-11-20): boot/uboot latest version 2020.10 in tag 2020.11, 2021.01 in 2021.02, 2022.01 in 2022.02/2022.05, 2022.04 in 2022.08, 2022.11, 2023.02, 2023.02.x, 2023.05, 2023.08, 2023.11, with no patches in boot/uboot from 2022.08; commit 083b03ae 'boot/uboot: bump to 2024.1' (2024-01-10); tag 2024.02 latest version 2024.01. — Buildroot
- Yocto poky (read 2026-11-20): kirkstone u-boot_2022.01.bb applies 0001-fs-squashfs-sqfs_read-Prevent-arbitrary-code-executi.patch with header 'CVE: CVE-2022-33103', Upstream-Status Backport 2ac0baab, plus patches for CVE-2022-33967 and CVE-2024-57254..57259; langdale ships u-boot_2022.07.bb; dunfell ships u-boot_2020.01.bb. — Yocto Project poky
- CISA CSAF ICSA-26-134-16 (Siemens SSA-577017, RUGGEDCOM ROX), read 2026-11-20: lists U-Boot CVEs including CVE-2022-30552, CVE-2022-30790, CVE-2022-34835 and CVE-2024-57256, but not CVE-2022-33103. — CISA CSAF
- Not in the CISA KEV catalog, version 2026.11.18 (1,786 entries), checked 2026-11-20. — CISA KEV
- EPSS 0.00459, percentile 0.40503, dataset 2026-11-19, read 2026-11-20. — FIRST EPSS

Related advisories
Sources
- NVD
- U-Boot commit 2ac0baab: fs/squashfs: sqfs_read: Prevent arbitrary code execution
- Debian security tracker: CVE-2022-33103
- Debian LTS Advisory DLA-4150-1: u-boot security update
- Buildroot commit 083b03ae: boot/uboot: bump to 2024.1
- Buildroot 2023.11 boot/uboot/Config.in (default U-Boot 2022.04)
- Yocto Project poky (kirkstone): u-boot_2022.01.bb with CVE patches
- CISA ICSA-26-134-16 (Siemens SSA-577017) CSAF
- CVE.org record CVE-2022-33103 (CNA: MITRE)
- CISA KEV
- FIRST EPSS
Frequently asked questions
Which U-Boot version fixes CVE-2022-33103?
U-Boot 2022.07 is the first release with the fix; commit 2ac0baab is in tag v2022.07-rc5. Debian backported it to bullseye in 2021.01+dfsg-5+deb11u1 (DLA-4150-1), and Yocto kirkstone carries it as a patch on U-Boot 2022.01.
Is my device exposed if U-Boot does not read SquashFS?
No. The bug is in the SquashFS driver, which is only built with CONFIG_FS_SQUASHFS (selected by CONFIG_CMD_SQUASHFS) and has no default. Check the board defconfig or the .config of the shipped bootloader; if SquashFS is off, the vulnerable code is not in the image.
How would an attacker use CVE-2022-33103?
By getting U-Boot to list a directory in a crafted SquashFS image, for example from an SD card, USB drive, writable partition or update image. An overlong entry name is copied past the 256-byte name buffer. NVD rates it local with low privileges required.
Is a Buildroot or Yocto build affected?
Buildroot's default U-Boot stayed in the affected range up to Buildroot 2023.11 (U-Boot 2022.04, no patches) and moved to 2024.01 in Buildroot 2024.02. Yocto kirkstone patches its 2022.01 recipe; langdale's 2022.07 includes the fix. Board-specific U-Boot forks need their own check.
Is CVE-2022-33103 exploited in the wild?
It is not in the CISA KEV catalog as of 2026-11-20 (catalog version 2026.11.18). FIRST EPSS gives it 0.00459, the 40.5th percentile, in the 2026-11-19 dataset.
Questions about this page?