CVE-20XX-?
agent?

CVE-2026-99607

unscored No fix listed Other OS & kernel

What it is

In the Linux kernel, the following vulnerability has been resolved: ext2: preserve IOMAP_F_DIRTY for extending direct writes ext2_iomap_begin() sets IOMAP_F_DIRTY for writes that may extend i_size, but clears the flag after ext2_get_blocks(). When the write uses an existing mapped block, iomap can then use FUA and skip generic_write_sync(), even though ext2_dio_write_end_io() updates the inode size. An O_DSYNC write can therefore return without persisting the new size. For example, on a FUA-capable device, a 512-byte O_DIRECT|O_DSYNC write at offset 512 into a fsynced one-byte file returns successfully and extends the in-memory size to 1024, while an immediate read of the on-disk inode still shows size 1. An explicit fsync persists the new size. Move setting of IOMAP_F_DIRTY after the initialization.

At a glance

Fix
No patch or advisory is linked yet. Look at mitigations or restrict exposure.

Timeline

  1. PublishedDisclosed publicly on CVE.org and NVD
  2. Last updatedMost recent change to the record

Why it ranks here

Details

Published
Last modified
Record status
Received

Affected

Is it in your code?

Hand this to your coding agent. It searches your project for the affected products and checks versions. Nothing leaves your machine.

Read https://cve-list.com/agent/cve/CVE-2026-99607.md and follow it. Check whether this project is affected by CVE-2026-99607. Read only. Don't change anything and don't upload my code. Tell me yes, no or not sure, with the file as proof.

· More options · Status badge

Meme: I wake up, CVE-2026-99607
Download Square Reddit

References

NVD record · CVE.org record