How does fdisk utility use CHS addressing on SSDs? - disk

Question

In my free time I am working on a personal project for reading and writing all kinds of disk structures like MBR, GPT, EXT2/3/4, NTFS, etc. While working on the MBR part I noticed that the fdisk utility populates the CHS address fields of the MBR partition entries when formatting a device... even if the target device is an SSD or a normal file like a disk image. The lack of any actual cylinders and heads/platters in an SSD or file made me curious: What does fdisk write to those CHS address fields?
Well, after some testing (i.e. formatting a normal file multiple times with different start sectors for the first partition), I am fairly certain that fdisk actually writes CHS addresses into those fields. But how does fdisk determine the cylinder/head sizes for an SSD or for a normal file? I can't imagine the HDIO_REQ and HDIO_GETGEO ioctls working on normal files.

Answer 1

After some more testing (see below) I am confident that the CHS addresses produced by fdisk always satisfy sector % (255 * 63) == 63 * chs.head + (chs.sector - 1) and chs.cylinder == 0 with chs.head only taking values in [0,254] and chs.sector only taking values in [1,63], regardless of the underlying (device) file's size.
I used the following Python script to quickly convert from LBAs to fdisk-CHSs to generate test cases. Tested on normal files as well as device files with different sector sizes.
import ctypes, subprocess
class chs_struct (ctypes.LittleEndianStructure):
# Unused: No system modern enough to run Python uses CHS addressing
_pack_ = 1
_fields_ = [
( "head", ctypes.c_uint8 ),
( "sector", ctypes.c_uint8, 6 ),
( "cylinder_hi", ctypes.c_uint8, 2 ),
( "cylinder_lo", ctypes.c_uint8 )
]
def chs_of_sector (first_sector, filename):
"""
Formats the file pointed to by filename with a new MBR whose first and
only partition is a primary partition starting at the specified
first_sector. This function returns the corresponding CHS address the
fdisk utility calculated.
"""
with subprocess.Popen(
["fdisk", filename],
universal_newlines = True,
stdin = subprocess.PIPE,
stdout = subprocess.DEVNULL,
stderr = subprocess.DEVNULL
) as proc:
# +------------------------------------------- Create new MBR
# | +-------------------------------------- Create new partition
# | | +--------------------------------- Make it a primary partition...
# | | | +---------------------------- and make it the first partition...
# | | | | +--------------- starting at this sector
# | | | | | +----- Max out partition size
# | | | | | | + Write MBR to file
# v v v v vvvvvvvvvvvv v v
proc.communicate("\n".join(["o", "n", "p", "1", str(first_sector), "", "w"]))
with open(filename, "rb") as file:
file.seek(447) # Offset of the first partition entry's first sector CHS address
chs = chs_struct.from_buffer_copy(file.read(3))
return (chs.head, chs.cylinder_hi << 8 + chs.cylinder_lo, chs.sector)