Board:lenovo/x60/Installation

From coreboot
Jump to navigation Jump to search

The wiki is being retired!

Documentation is now handled by the same processes we use for code: Add something to the Documentation/ directory in the coreboot repo, and it will be rendered to https://doc.coreboot.org/. Contributions welcome!

Flashing on the laptop instructions.

Lenovo X60, X60s, T60 and T60p flashing instructions.

These Lenovo laptops have a register that must be flipped before coreboot can be flashed.

For those/some models with SPI flash chips you have also to modify flashrom. Because the chipset locks down the available commands that flashrom can send to the flash chip, you also need to change the flashrom source in a way that is not suitable to upstream. Flash chips can be identified by various commands (REMS*, RDID etc.). Some of them reply with an ID for the vendor and the exact chip model; others just reply with a single byte which is fine if there is only a small number of chips to distinguish, but won't work for the huge number of flash chips known to flashrom. The problem with the vendor BIOS is that it forbids the higher quality identification commands, so you need to force flashrom to use the lower quality opcode for the chip in your Thinkpad. You have to know the chip model beforehand (e.g. by inspection). Known models on the x60s are SST25VF016B, MX25L1605D and maybe others.

You will need: the flashrom source (at least r1613 to make sure the laptops are whitelisted to work with flashrom), a small modification of it (as explained below in detail), the datasheet of your flash chip retrieve the right values for the modification, and the bucts utility.

  1. Patch flashrom to use RES SPI identification and spi_chip_write_1 for your flash chip, as well as change the flash chip model id to fit the RES opcode.
    • Find the definition of your flash chip in flashrom's flashchips.c
      • Optionally, you can copy the existing definition as it is done in this patch. This will allow to switch between the two definitions with the -c parameter. Be sure to change the .name field in that case (e.g. .name = "SST25VF016B-RES",).
    • Change the .probe field to probe_spi_resN where N equals the number of ID bytes the flash replies to the RES ID command (opcode 0xAB, look it up in the chip's datasheet) (e.g. .probe = probe_spi_res2, if the chip replies with one byte vendor ID and one byte model ID)
    • Change the .model_id field to the RES model ID given in the datasheet of the flash chip (e.g. .model_id = 0x14,)
    • Change the .write field to spi_chip_write_1 (i.e. .write = spi_chip_write_1,)
  2. Run flashrom -p internal -r factory.bin
    This step is IMPORTANT since the factory BIOS in your machine is tied to your particular system board (or "planar" in IBM FRU terms) with a unique ID not present in factory BIOS updates.
  3. Run dd if=coreboot.rom of=top64k.bin bs=1 skip=$[$(stat -c %s coreboot.rom) - 0x10000] count=64k
  4. Run dd if=coreboot.rom bs=1 skip=$[$(stat -c %s coreboot.rom) - 0x20000] count=64k | hexdump
    Verify that the complete range is filled with ff bytes before proceeding! The above command must output:
    0000000 ffff ffff ffff ffff ffff ffff ffff ffff
    *
    0010000
    If this is not the case, the coreboot image needs to be rebuilt with the second-to-last 64kbyte block unused.
  5. Run dd if=top64k.bin of=coreboot.rom bs=1 seek=$[$(stat -c %s coreboot.rom) - 0x20000] count=64k conv=notrunc
  6. Run bucts 1
  7. Run flashrom -p internal -w coreboot.rom
    This will be slow, it will output errors for addresses 0x0 and 0x1f0000 when working with a 2 Mbyte flash chip, and it will say "FAILED!" at the end, see Peter's mail before you panic.
  8. Power cycle the machine (i.e. a cold boot, not just a reboot), now starting with coreboot
  9. Revert all changes made to flashrom (maybe backup the binary for later experiments)
  10. Run flashrom -p internal -w coreboot.rom.
    This will successfully overwrite the entire flash chip, including the last 64k that were write protected with the factory BIOS.
  11. Run bucts 0


See also http://thread.gmane.org/gmane.linux.bios/69354 http://thread.gmane.org/gmane.linux.bios.flashrom/575

Recovery

If you had a bad flash you will need a recovery method.

If you only set bucts, then rebooted without doing any flash writes, things might be easier: bucts sets a register that lives on the RTC well, ie. it is powered by the same source that keeps the clock alive. Usually that's a battery on the mainboard, and often there's some way to cut the source (by removing the battery, a jumper, or pads that can be shorted). After doing that (for a few seconds, there might be some capacitors in the way that keep power stable), the register should be reset and the system should boot as normal.

On the x60x, bucts issues might also be solved by "discarging RTC", which is done by pressing the power button 5 times for 10 seconds.

Required/advised hardware and informations

Howto

0. wire the pomona clip to a programmer that way:

From the #coreboot IRC Channel on FreeNode servers:

Oct 01 15:35:48 <CareBear\>     one important thing is that when you connect the clip to the X60 you should not connect all pins
[...]
Oct 01 15:36:22 <CareBear\>     only connect these pins: 1, 2, 4, 5, 6
[...]
Oct 01 15:37:21 <CareBear\>     also important: first connect charger to laptop, then connect the clip
[...]
Oct 01 17:49:41 <CareBear\>     GNUtoo-desktop : the mainboard must be powered off, but with the charger connected
[...]
Oct 01 17:50:39 <CareBear\>     um, that way there is no way anything will break
[...]
Oct 01 17:51:00 <CareBear\>     it is important not to connect 3v3 from the outside
Oct 01 17:51:39 <CareBear\>     because the correct power sequencing is not known, and if any other rail must come on before the standby 3v3 then the machine may well break when 3v3 is applied from the outside
[...]
Oct 01 17:52:48 <CareBear\>     it may also be fine - but it is unknown what happens
[...]
Oct 01 17:53:47 <CareBear\>     not supplying 3v3 from the outside is safer
Oct 01 17:54:25 <CareBear\>     and because the machine is powered off, there is no risk of the chipset accessing the flash chip

In another hand I didn't follow that and wired it without powering the mainboard(mainboard disconnected from power plug, no battery in) and with all pins and it worked...

  1. Disassemble carefully the laptop, the SO-8 chip is on the bottom of the mainboard...
  2. connect the pomona clip to the BIOS chip
  3. flash coreboot or the BIOS
  4. remount the laptop

Coreboot standard configuration

  • It's advised to make SeaBios(instead of coreboot) run the VGA option rom by disabling CONFIG_VGA_ROM_RUN:
[ ] Run VGA Option ROMs

in make menuconfig. Note that you still need to include the option rom in coreboot:

[*] Add a VGA BIOS image

See VGA_support for details on how to include the VGA BIOS image.

  • Also disable CONFIG_S3_VGA_ROM_RUN which is for really old linux kernels(2.4) (which is disabled automatically if you don't select CONFIG_VGA_ROM_RUN).

From the #coreboot IRC Channel on FreeNode servers:

Oct 04 13:47:09 <patrickg>      that's about running vga init on s3 wakeup - required for some older linux kernels
[...]
Oct 04 13:47:25 <patrickg>      BIOSes call it "POST on wakeup" or sth like that
Oct 04 13:47:30 <patrickg>      older ~ 2.4 class ;)


VGA Option ROM / VBIOS replacement (native graphics)

The VGA option ROM (see above) is proprietary. If you have an X60 or X60s, GNUtoo has developed a free replacement which is under review, but you can use it right now:


Problems in native graphics code exposed by recent kernels

It was discovered that while this code seems to work fine on kernels older than linux 3.12.

On 3.12 and later kernels the i915 driver became unstable. Causing broken/glitchy 3D. and performance bugs.

The current suspicion is that linux had a bug which caused it to miss stolen memory altogether on this machine and hence disable parts of code. This bug was corrected in: 17fec8a08698bcab98788e1e89f5b8e7502ababd

Which was identified by git bisect, to be the cause of the driver crashes using native vga on the x60.

The native vga init code in coreboot needs to be corrected to have more engineered memory management. In the interim users have attempted running patched linux 3.12 kernels which revert commit 17fec. To apparent success.

Recently tested revisions on the X60

8496c4dbec41b3a9284fc29b0dcd97fc8946223b coreboot-4.0-5045-g9bf05de

d7813436b325551604b13f8cefc30a6f92be4534 coreboot-4.0-5047-g7e2f226

Status