cbwww/site/docs/install/nvmutil.md

18 KiB

TODO: write a full list her ofe what actual PHYs are known to work.

It's probably all of them, but some newer ones might have changed the standard by which they are configured. This program actively avoids working on files that have invalid checksums, on most commands, precisely so that the user does not inadvertently use it on incompatible files; it is assumed that intel would later change the file size and/or checksum value and/or checksum location.

How to obtain the GbE file

The chip containing your BIOS/UEFI firmware (or coreboot) has it, if you have an Intel PHY for gigabit ethernet.

The sections below will teach you how to obtain the GbE file, containing your NIC's configuration. This is the part that many people will struggle with, so we will dedicated an entire next section to it:

Use flashrom

If you wish to operate on the GbE section that's already flashed, you should dump the current full ROM image. If you already have a ROM image, you do not need to dump it, so you can skip this section.

Download flashrom here:

Using recent flashrom versions, you can extract this region. If your regions are unlocked, you can run flashrom on the target system, like so:

flashrom -p internal -r rom.bin

If your system has two flash chips, the GbE region is usually stored on SPI1 (not SPI2). Otherwise, it may be that you have a single-flash setup. In that case, it's recommended to dump both chips, as spi1.rom and spi2.rom; you can then cat them together:

cat spi1.rom spi2.rom > rom.bin

If your GbE region is locked (per IFD settings), you can dump and flash it using external flashing equipment. The Canoeboot project has a handy guide for this; it can be used for reading from and writing to the chip. See: spi.md

If you're using an external programmer, the -p internal option should be changed accordingly. Read flashrom documentation, and make sure you have everything properly configured.

Use ifdtool

NOTE: This has only been tested on systems that use IFDv1 (Intel Flash Descriptor, version 1). This distinction, between v1 and v2, is made in the ifdtool source code, which you should read if you're interested. Intel`s v2 specification has more regions in it, whereas v1 systems usually defined: IFD, GbE, PD, ME and BIOS regions.

The ifdtool program is a powerful tool, allowing you to manipulate Intel Flash Descriptors. It's part of coreboot, available in the coreboot.git repository under util/ifdtool/. Just go in there and build it with make, to get an ifdtool binary.

To make internal flashing possible later on, you might do:

ifdtool --unlock rom.bin

Running this command will create a modified image, named rom.bin.new. This file will have all regions set to read-write, per configuration in the Intel Flash Descriptor.

In addition to unlocked regions, you may wish to neuter the Intel Management Engine, removing all the nasty spying features from it, using me_cleaner. See:

The me_cleaner program is outside the scope of this article, so you should read their documentation.

Now run this:

ifdtool -x rom.bin

Several files will be created, and the one you need to operate on is named flashregion_3_gbe.bin so please ensure that you have this file.

Read the notes below about how to use the nvmutil program, operating on this file. When you're done, you can insert the modified GbE file back into your ROM image, like so:

ifdtool -i gbe:flashregion_3_gbe.bin rom.bin

This will create the file rom.bin.new, which contains your modified GbE section with the NVM images inside; this includes your MAC address.

Refer to flashrom documentation. You may flash the new ROM like so, if running on the same system and the regions are read-write:

flashrom -p internal -w rom.bin.new

Newer versions of flashrom support flashing just the specified region, like so:

flashrom -p internal --ifd -i gbe -w rom.bin.new

If you're running flashrom from host CPU on the target system, and it's dual flash, you can just flash the concatenated image, which you created earlier by running the cat program; dual-IC flash configurations appear to your operating system as one large flash area, as though it were a single chip.

If you're using an external programmer, you should change the -p internal parameter to something else. In this situation, you should re-split the file accordingly, if you have a dual-IC flash set, like so:

dd if=rom.bin.new of=spi2.rom bs=1M skip=8
dd if=rom.bin.new of=spi1.rom bs=1M count=8

These files would then be flashed externally, separately, using an external programmer.

The above example (using dd) is for setups with 12MB flash, where you have 8MB as SPI1 and 4MB as SPI2. SPI1 would contain the IFD, and SPI2 is the upper flash area containing your bootblock; GbE is probably located in SPI1. You should adjust the above parameters, according to your configuration.

How to compile source code

The nvmutil source code is located under util/nvmutil/ in the cbmk repository. A makefile is included there, for you to build an executable.

The nvmutil programs will work just fine, on any modern BSD Unix operating system, or unix-like system such as GNU+Linux.

You must be sure to have toolchains installed, for building; a normal libc, C compiler and linker should be enough. GCC and LLVM have all these things included, so use whichever one you want.

If the code is compiled on OpenBSD, pledge(2) is used. This is done with an ifdef rule, so that the code still compiles on other systems. When the dump command is specified, pledge will use these promises: stdio rpath. When any other command is used, these pledge promises will be used: stdio wpath.

The nvmutil software has been build-tested on Clang, GCC and tcc. Only standard library functions (plus err.h) are used, so you don't need any extra libraries.

How to compile it

First, ensure that the current working directory is your copy of the nvmutil source code!

You may run this in your terminal:

make

This will result in a binary being created named nvm. Install this to wherever you want, such as /usr/bin (or whatever is in your $PATH for userspace programs).

TODO: Add make install to the Makefile, portably.

How to use nvmutil

You run it, passing as argument the path to a file, and you run commands on that file. This section will tell you how to perform various tasks, by using these commands.

In these examples, it is assumed that you have installed the nvm binary to somewhere in your $PATH. If you haven't done that, you could still run it in cwd for instance:

./nvm bla bla bla

Exit status

The nvmutil program uses errno extensively. The best error handling is done this way, the Unix way. Error handling is extremely strict, in nvmutil; on program exit, the errno message is printed (if not zero) and the value of errno is returned (upon exit from int main).

The main function always returns errno, no matter what. This style of programming (set errno and return) is a very old fashioned way of doing things, and in many cases it is the most correct way.

This is why we say zero status and non-zero status in Unix programs, when we talk about exit status. Zero is success, and anything above zero is fail; errno is zero by default, unless set, and it will always be set to a value above zero (if set).

All commands (except dump) require read and write access. The dump command only requires read access on files. Where sufficient permission is not given (read and/or write), nvmutil will exit with non-zero status.

Non-zero status will also be returned, if the target file is not of size 8KB.

Additional rules regarding exit status shall apply, depending on what command you use. Commands are documented in the following sections:

Change MAC address

The nvm program lets you change the MAC address. It sets a valid checksum, after changing the MAC address. This program operates on both NVM parts, but it will only modify a given part if the existing checksum is correct. It will exit with zero status if at least one part is modified; otherwise, it will exit with non-zero status.

The following rules are enforced in code:

  • User cannot specify multicast addresses
  • User cannot specify 00:00:00:00:00:00
  • When generating random addresses, if the right most nibble of the left-most byte is ? (random), nvmutil will (in code) force the generated MAC address to be local (not global), and will prevent a multicast address from being generated.

A multicast address is invalid because it represents multiple devices; you must specify a unicast address. A global address is one uniquely assigned by the vendor, and a local address is an overridden one. You can set global MAC addresses in nvmutil, for example if you are simply copying what was officially assigned to your NIC, you can do that. For example, if your MAC address was 00:de:ad:be:ef:69 as assigned by the manufacturer, which is a global unicast MAC address, you would type:

nvm gbe.bin setmac 00:de:ad:be:ef:69

How to use (the MAC address in just an example):

nvm gbe.bin setmac 00:de:ad:be:ef:00

You can also set random MAC addresses:

nvm gbe.bin setmac ??:??:??:??:??:??

In this example, every character is random. However, you can mix and match random characters with static ones. For example:

nvm gbe.bin setmac 00:1f:16:??:??:??

You can also pass it without a MAC address:

nvm gbe.bin setmac

If you only type setmac without specifying a MAC address, it will do the same thing as setmac ??:??:??:??:??:??.

This will set the last three bytes randomly, while the MAC address would begin with 00:1f:16.

The reason nvmutil doesn't alter a part with an existing invalid checksum, is precisely so that if the algorithm changes in future Intel PHYs, nvmutil will just fail and not modify your file. This is because the checksum would then be invalid, at all times. However, correct NVM parts with otherwise invalid checksums do exist, and can be corrected if you use the setchecksum command in nvmutil. It is common for vendor gbe files to contain one valid part and one invalid part, per checksum rules.

Verify checksums (and show MAC addresses)

This command only requires read access on files.

The nvm program can show a hexdump of both NVM parts, and tell you whether each one is valid (as per checksum calculation). It also prints the MAC address from each part.

How to use:

nvm gbe.bin dump

NOTE: This will exit with zero status if at least one part contains a valid checksum. If both parts are invalid, nvmutil will exit with non-zero status.

Copy part

This command requires read and write access on files.

The nvm program can copy one NVM part to another. It copies the entire 4KB part, within the 8KB file.

Overwrite part 0 with the contents of part 1:

nvm gbe.bin copy 1

Overwrite part 1 with the contents of part 0:

nvm gbe.bin copy 0

NOTE: If the part to be copied has a bad checksum, no operation will be performed, and nvmutil will exit with non-zero status. Otherwise, it will (if all other conditions are met) exit with zero status.

Swap parts

This command requires read and write access on files.

The nvm program can swap both 4KB parts in the GbE file. It does this, via simple XOR swaps.

How to use:

nvm gbe.bin swap

NOTE: This operation will be aborted if BOTH checksums are invalid. This is to guard against accidentally using nvmutil on the wrong file.

If at least one part is valid, nvmutil will return with zero exit status. If both parts are invalid, it will return non-zero.

Set valid checksum

This command requires read and write access on files.

The nvm program can calculate and sets a valid checksum, on the desired NVM part. Usage:

Fix part 0:

nvm gbe.bin setchecksum 0

Fix part 1:

nvm gbe.bin setchecksum 1

WARNING: NO validity checks are performed. This will simply set the checksum. There is no feasible way to guard against use on the wrong file, unlike with the other commands. Please make SURE you're running this on the correct file!

Set invalid checksum

This command requires read and write access on files.

The nvm program can intentionally set an invalid checksum, on the desired NVM part. Usage:

Invalidate part 0:

nvm gbe.bin brick 0

Invalidate part 1:

nvm gbe.bin brick 1

NOTE: If the part already has an invalid checksum, no operation will be performed, and nvmutil will exit with non-zero status. This is to guard against nvmutil being used on the wrong file.

This may be desirable, if you've made modifications to both parts but you want to guarantee that only one of them is used. Also, the setmac command will only operate on parts that already have a valid checksum, so you could run brick before running setmac (or run it afterwards).

The GNU+Linux kernel's e1000 driver will refuse to initialise Intel gigabit NICs that don't have a valid checksum. This is software-defined, and not enforced by the hardware.

LICENSE

This page is released under different copyright terms than most other pages on this website.

The nvmutil software and documentation are released under the following terms:

Copyright 2022 Leah Rowe

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.