Matěj Volf <hele...@mvolfik.com> wrote:
> On 21. 04. 25 22:03, Miroslav Cimerman wrote:
> 
> > > Bisecting this shows that:
> > > 
> > > c3d9aaf561afb8b493c3460b1898f33b5499c920 is the first bad commit
> > > commit c3d9aaf561afb8b493c3460b1898f33b5499c920
> > > Author: Jiri Svoboda jiri@wiwaxia
> > > 
> > > Date: Mon Feb 3 16:52:08 2025 +0100
> > > 
> > > Determine when device (sub)tree is stable.
> > > 
> > > Devman will only return value when the entire device tree is stable.
> > > 
> > > uspace/srv/devman/dev.c | 15 +++++++++++++++
> > > uspace/srv/devman/dev.h | 3 ++-
> > > uspace/srv/devman/devman.h | 7 ++++++-
> > > uspace/srv/devman/devtree.c | 13 +++++++++++++
> > > uspace/srv/devman/devtree.h | 3 ++-
> > > uspace/srv/devman/driver.c | 42 ++++++++++++++++++++++++++++++++++++++++--
> > > uspace/srv/devman/main.c | 5 ++++-
> > > 7 files changed, 82 insertions(+), 6 deletions(-)
> > 
> > Thanks, this commit probably also broke a lot more stuff,
> > on amd64 I get every ~25 boots a freezed kernel console.
> > 
> > I am inlining the dmesg, there is also HelenRAID (comming soon :D)
> > server started but it doesn't touch anything.
> 
> Hm, I don't recall seeing this. Are you sure it's not from your raid patches? 
> :D

My internal tree doesn't touch anything outside HelenRAID, when
I need a system wide change I open a pull request.

In the dmesg you can see that devman deadlocked. I run tmpfs on RAM disk,
but that probably doesn't change much (still there is strange ENOMEM on root fs
extraction).

Also I almost always have some drives connected to qemu, that can do something
as well - see the pci-ide crash in the end (and also isa-ide). If they weren't
there, nothing would have happened probably, devman might deadlock but you
might not notice, as your boot finishes.

And regarding the ~25 boots, its just a random number, its probably much much
higher - I have been hacking most of the past few days, and I got 2 crashes.
So that can be per ~100 boots?

--
mc

_______________________________________________
HelenOS-devel mailing list
HelenOS-devel@lists.modry.cz
http://lists.modry.cz/listinfo/helenos-devel

Reply via email to