Getting nvme disk corruptions when issuing IDENTIFY commands via ioctl()

Issue

• Getting nvme disk corruptions when issuing IDENTIFY commands via ioctl()

• Filesystem created on nvme disk gets corrupt when the application sends NVME_IOCTL_ADMIN_CMD / IDENTIFY command to the disk:

  EXT4-fs error (device nvme0n1): ext4_ext_find_extent:885: inode #2: comm cp: pblk 138 bad header/extent: invalid magic - magic a8, entries 0, max 20(0), depth 267(0)
  EXT4-fs warning: 571 callbacks suppressed
  EXT4-fs warning (device nvme0n1): __ext4_read_dirblock:676: error reading directory block (ino 2, block 0)
  EXT4-fs error (device nvme0n1): ext4_ext_find_extent:885: inode #2: comm cksum: pblk 138 bad header/extent: invalid magic - magic a8, entries 0, max 20(0), depth 267(0)
  EXT4-fs warning (device nvme0n1): __ext4_read_dirblock:676: error reading directory block (ino 2, block 0)
  EXT4-fs error (device nvme0n1): ext4_ext_find_extent:885: inode #2: comm cp: pblk 138 bad header/extent: invalid magic - magic a8, entries 0, max 20(0), depth 267(0)
  EXT4-fs warning (device nvme0n1): __ext4_read_dirblock:676: error reading directory block (ino 2, block 0)
  [...]
  

  Similar corruption is seen across the filesystems created directly on nvme disk as well as lvm volumes, mdraid devices created on top of nvme disk.