#427 Checklist for new release

Open
opened 2 years ago by swiftgeek · 2 comments
  • Working Fallback in bootblock
  • RTC NVRAM for config
  • Document SeaVGABIOS limitations at exact point of release (SeaGRUB)
  • Full memtest86+ test with 8GiB of RAM / RAW Card F (to make sure there aren't issues with block move and other tests)
  • I_want_a_brick logic working inside grub.cfg
  • Drive preference with UUID (eg. when using 2 FDE drives)
  • H8S indicators for coreboot failure (like missing RAM)
  • Display message when loading libreboot_grub.cfg (obvious security reasons, with variable setting for that) (a wallpaper/text could be changed for brief moment to red or sth)
  • Test GELI (FreeBSD) and LUKS (Linux) FDE boot paths
  • braille line if we ever get user/test setup
  • Discard boot time EC trash from buggy EC fw (easily tested on H8S, then even more on T400s (waikiki-3) x86 cold boot). https://review.coreboot.org/c/coreboot/+/22180/28 . Reportedly rfkill and backlight control doesn't work on cold boot and works fine on subsequent warm boots of x86
  • Validity of uk dvorak layout #531
  • POST card on devices with legacy PCI slots, ThinkPads once free implementation exists. Single byte writes over LPC/PCI shouldn't cause significant slowdown.
  • Statically compiled end user tools (nvramtool, cbfstool, flashrom)
  • Make sure that OptionROMs from outside of CBFS are NOT executed, especially on targets without iGPU
  • Make sure all sata ports with default NVRAM configuration have "SATA ALPM" disabled. SATA ALPM causes many drives to glitch out to the point of SMART identifying it as errors and potentially data corruption (there is a list of entire vendors products that glitch out with this setting)
  • Make sure that all grub modules that are shipped inside (memdisk) are accessible - use ls (memdisk)/boot/grub/i386-coreboot/ and probably take a photo of it or dump over serial. Stable on T60 has at the very least ahci module not accessible despite it being loaded on start somehow (and two boot directories!)
  • In the case old grub has to be used, backport initial LUKS2 support
  • IFD~ needs to disable EAR things via OPMENU, otherwise PRx protection from libreboot may be circumvented when using 32MiB and bigger flash PRx would be then applied to every 16MiB bank, so the result would be a softbrick which isn't as important.
  • Make sure it boots correctly on device without any drives and (usb) peripherals connected - known bad: 20160907 with T60/X60 (Napa), GA-G41M-ES2L (x4x+ich7)
  • acpi_backlight keys and lid events actually doing something
  • acpi code for docking, including booting when (not) docked
  • Use current stable debian AND report version used, so there is at least one known good setup in case something fails.
  • trace all the downloads performed when building from scratch
  • make sure each downloaded file is backed in wayback and has fallback download available in upstream code (libreboot patch to add wayback as alternative mirror)
  • make sure it can be built offline on fresh install of debian (with all the required packages installed)
  • check VBTs for invalid copyright strings
- [ ] Working Fallback in bootblock - [ ] RTC NVRAM for config - [ ] Document SeaVGABIOS limitations at exact point of release (SeaGRUB) - [ ] Full memtest86+ test with 8GiB of RAM / RAW Card F (to make sure there aren't issues with block move and other tests) - [ ] I_want_a_brick logic working inside grub.cfg - [ ] Drive preference with UUID (eg. when using 2 FDE drives) - [ ] H8S indicators for coreboot failure (like missing RAM) - [ ] Display message when loading libreboot_grub.cfg (obvious security reasons, with variable setting for that) (a wallpaper/text could be changed for brief moment to red or sth) - [ ] Test GELI (FreeBSD) and LUKS (Linux) FDE boot paths - [ ] braille line if we ever get user/test setup - [ ] Discard boot time EC trash from buggy EC fw (easily tested on H8S, then even more on T400s (waikiki-3) x86 cold boot). https://review.coreboot.org/c/coreboot/+/22180/28 . Reportedly rfkill and backlight control doesn't work on cold boot and works fine on subsequent warm boots of x86 - [ ] Validity of uk dvorak layout #531 - [ ] POST card on devices with legacy PCI slots, ThinkPads once free implementation exists. Single byte writes over LPC/PCI shouldn't cause significant slowdown. - [ ] ***Statically*** compiled end user tools (nvramtool, cbfstool, flashrom) - [ ] Make sure that OptionROMs from outside of CBFS are **NOT** executed, especially on targets without iGPU - [ ] Make sure all sata ports with default NVRAM configuration have "SATA ALPM" disabled. SATA ALPM causes many drives to glitch out to the point of SMART identifying it as errors and potentially data corruption (there is a list of entire vendors products that glitch out with this setting) - [ ] Make sure that all grub modules that are shipped inside (memdisk) are accessible - use `ls (memdisk)/boot/grub/i386-coreboot/` and probably take a photo of it or dump over serial. Stable on T60 has at the very least ahci module not accessible despite it being loaded on start somehow (and two boot directories!) - [ ] In the case old grub has to be used, backport [initial LUKS2 support](https://git.savannah.gnu.org/cgit/grub.git/commit/?id=365e0cc3e7e44151c14dd29514c2f870b49f9755) - [ ] <s>IFD~ needs to disable EAR things via OPMENU, otherwise PRx protection from libreboot may be circumvented when using 32MiB and bigger flash</s> PRx would be then applied to every 16MiB bank, so the result would be a softbrick which isn't as important. - [ ] Make sure it boots correctly on device without any drives and (usb) peripherals connected - known bad: 20160907 with T60/X60 (Napa), GA-G41M-ES2L (x4x+ich7) - [ ] acpi_backlight keys and lid events actually doing something - [ ] acpi code for docking, including booting when (not) docked - [ ] Use current stable debian AND report version used, so there is at least one known good setup in case something fails. - [ ] trace all the downloads performed when building from scratch - [ ] make sure each downloaded file is backed in wayback and has fallback download available in upstream code (libreboot patch to add wayback as alternative mirror) - [ ] make sure it can be built offline on fresh install of debian (with all the required packages installed) - [ ] check VBTs for invalid copyright strings
Andrew Robbins commented 2 years ago
Collaborator

I tested Memtest86+ with 8GiB of RAM on my X200 and had no issues with test #7 (block move) or any others on a full pass. Memtest86+ revision used was 5ca4eb9544e51254254d09ae6e70f93403469ec3

I tested Memtest86+ with 8GiB of RAM on my X200 and had no issues with test #7 (block move) or any others on a full pass. Memtest86+ revision used was `5ca4eb9544e51254254d09ae6e70f93403469ec3`
Swift Geek commented 2 years ago
Collaborator

This is list is strictly for just before release, to make sure there are no regressions, or something uncommitted

This is list is strictly for just before release, to make sure there are no regressions, or something uncommitted
Sign in to join this conversation.
Loading...
Cancel
Save
There is no content yet.