snapshot of april 12th wont boot at all

[Beeblebrox]

I have (possibly) similar problem, running kernel from March 29 as
fall-back (older kernel, latest world).

* OS: HardenedBSD-12 on amd64 Athlon II X3 460
* Kernels built yesterday: MYKERN & HARDENEDBSD (GENERIC with Debug).
Both kernels fail. No errors encountered during buildworld/kernel.
World installs normally.
* Boot process: Grub -> /boot/loader -> Kernel Selection. No EFI (MBR).
Grub has own partition (ext2fs), /boot/loader on ZFS.
* zpool: Striped on 2 HDD's, scrub returns clean pool.

FAILURE POINT: System freeze, completely unresponsive, requires hard
reset.
* Initially seems like failure is at "Mount Root from zpool/ROOT/bsd".
Further inspection of dmesg shows garbled text at services stage
(start reading /etc/rc.conf)
* /etc/rc.conf: Everything disabled for bug hunting. Cause does not
seem related to linux / radeonkms either.
* Selecting "Safe Mode" (kern.smp.disabled ?) gets me to single-user
mode, and later multi-user but system freezes as soon as I login.
Not 100% sure about this behaviour though, as I have been at it since
early morning yesterday.
* Have not tried with ACPI disabled. Dmesg: Firmware Warning (ACPI):
32/64X length mismatch in FADT/Gpe0Block: 64/32 (20180313/tbfadt-748)

There was a recent bug involving databases failing to start due to
"limits: NO: No such file or directory". I have Mariadb-Server-100
installed with the new patch for this bug. I wonder whether these two
could be related?

Would prefer not posting my verbose boot log dmesg file here, can email
it (or download link) to developers (at FreeBSD.org) if needed though.

Regards.

[Shawn Webb]

Post by Beeblebrox
I have (possibly) similar problem, running kernel from March 29 as
fall-back (older kernel, latest world).
* OS: HardenedBSD-12 on amd64 Athlon II X3 460
* Kernels built yesterday: MYKERN & HARDENEDBSD (GENERIC with Debug).
Both kernels fail. No errors encountered during buildworld/kernel.
World installs normally.
* Boot process: Grub -> /boot/loader -> Kernel Selection. No EFI (MBR).
Grub has own partition (ext2fs), /boot/loader on ZFS.
* zpool: Striped on 2 HDD's, scrub returns clean pool.
FAILURE POINT: System freeze, completely unresponsive, requires hard
reset.
* Initially seems like failure is at "Mount Root from zpool/ROOT/bsd".
Further inspection of dmesg shows garbled text at services stage
(start reading /etc/rc.conf)
* /etc/rc.conf: Everything disabled for bug hunting. Cause does not
seem related to linux / radeonkms either.
* Selecting "Safe Mode" (kern.smp.disabled ?) gets me to single-user
mode, and later multi-user but system freezes as soon as I login.
Not 100% sure about this behaviour though, as I have been at it since
early morning yesterday.
32/64X length mismatch in FADT/Gpe0Block: 64/32 (20180313/tbfadt-748)
There was a recent bug involving databases failing to start due to
"limits: NO: No such file or directory". I have Mariadb-Server-100
installed with the new patch for this bug. I wonder whether these two
could be related?
Would prefer not posting my verbose boot log dmesg file here, can email
it (or download link) to developers (at FreeBSD.org) if needed though.

What happens when you set vm.pmap.pti=0 at the loader prompt?

[Beeblebrox]

Hey Shawn,

Post by Shawn Webb
What happens when you set vm.pmap.pti=0 at the loader prompt?

Yep, that fixed it. Booted into MYKERN without problem.
Assuming Page Table Isolation still a bit buggy?

FreeBSD 12.0-CURRENT #3 5b8586c22aa(hardened/current/master):
Tue Apr 17 08:11:53 EDT 2018

Regards.

[Shawn Webb]

Post by Beeblebrox
Hey Shawn,

Post by Shawn Webb
What happens when you set vm.pmap.pti=0 at the loader prompt?

Yep, that fixed it. Booted into MYKERN without problem.
Assuming Page Table Isolation still a bit buggy?
Tue Apr 17 08:11:53 EDT 2018

HardenedBSD enables PTI regardless of underlying CPU by default. We've
found that some older AMD CPUs have issues with PTI as currently
implemented.

Thanks,

[Ed Maste]

Post by Shawn Webb
HardenedBSD enables PTI regardless of underlying CPU by default. We've
found that some older AMD CPUs have issues with PTI as currently
implemented.

The oldest AMD CPU I have is a FX-6100, and FreeBSD-CURRENT works fine
there with PTI enabled. If this is reproducible on stock FreeBSD can
you submit a PR with CPU details?

[Beeblebrox]

Hello Ed.

If this is reproducible on stock FreeBSD can you submit a PR with CPU details?

I don't have a cloned repo of FreeBSD, but AFAIK it's the "options PAX_*" related knobs in the kernel that, if eliminated, would revert the system to a FreeBSD one for all intents and purposes? I can just build MINIMAL or GENERIC and try if that's the case.

Separate finding I have for this CPU is that with an unlocked 4th core (gets detected as Phenom II X4 B60 instead of Athlon II X3) observable oddities are:
* Some ports have runaway builds under poudriere, end up filling swap and fail. www/webkit-gtk2 is an example. This could also be due to separate src.conf settings used to build the poudriere jail world.
* Xorg desktop becomes jittery, mouse control is difficult.
Mentioning these to ask whether you prefer them to be included in PR?

Regards.

[Beeblebrox]

My proposed plan did not work anyway, as both MINIMAL and GENERIC fail to build.
I'll have to clone the FreeBSD repo unless Shawn has something to comment on this.

[Shawn Webb]

Post by Beeblebrox
My proposed plan did not work anyway, as both MINIMAL and GENERIC fail to build.
I'll have to clone the FreeBSD repo unless Shawn has something to comment on this.

Due to our changing of certain kernel structures, GENERIC likely won't
build anymore. FreeBSD's kernel source code contains compile-time
assertions that detect when offsets of values within certain key
structures are different than expected (example commit
f9f820f2e418f9b68f9d94631171135bfb79989f in hardened/current/master).

I'm unsure why PTI causes issues in these cases. I would love to
attempt to make an effort to debug it myself, but I would need some
way to replicate the issue on my end.

Thanks,

[Kyle Evans]

Post by Beeblebrox
I have (possibly) similar problem, running kernel from March 29 as
fall-back (older kernel, latest world).
* OS: HardenedBSD-12 on amd64 Athlon II X3 460
* Kernels built yesterday: MYKERN & HARDENEDBSD (GENERIC with Debug).
Both kernels fail. No errors encountered during buildworld/kernel.
World installs normally.
* Boot process: Grub -> /boot/loader -> Kernel Selection. No EFI (MBR).
Grub has own partition (ext2fs), /boot/loader on ZFS.
* zpool: Striped on 2 HDD's, scrub returns clean pool.
FAILURE POINT: System freeze, completely unresponsive, requires hard
reset.
* Initially seems like failure is at "Mount Root from zpool/ROOT/bsd".
Further inspection of dmesg shows garbled text at services stage
(start reading /etc/rc.conf)
* /etc/rc.conf: Everything disabled for bug hunting. Cause does not
seem related to linux / radeonkms either.
* Selecting "Safe Mode" (kern.smp.disabled ?) gets me to single-user
mode, and later multi-user but system freezes as soon as I login.
Not 100% sure about this behaviour though, as I have been at it since
early morning yesterday.
32/64X length mismatch in FADT/Gpe0Block: 64/32 (20180313/tbfadt-748)
There was a recent bug involving databases failing to start due to
"limits: NO: No such file or directory". I have Mariadb-Server-100
installed with the new patch for this bug. I wonder whether these two
could be related?

I can't speak much to the rest of your e-mail, but as far as this last
bit goes: the bug you're referencing would've just failed to start the
database server because `limits` was passed an invalid argument. If
you have filesystem access, you can confirm this by setting
mariadb_limits="" in your rc.conf(5)

[wes wroten]

I am sorry i sent a blank email, I didn't know better so i changed it to text only.

the problem i have is during the boot from the cd-rom is I just get a big black screen with a cursor
and that's all, not even an attempt to install, nothing

my machine is running freebsd 11.1-STABLE currently but i want to switch to 12-CURRENT to help
out with the project in ANY way I can.

i have an amd64
4gb ram (but 12g swap)

any other machine questions? ill happily answer if i know how

~wes