Bug #2518

mounting devfs at boot hangs on at least two i386 machines

Added by davshao almost 2 years ago. Updated almost 2 years ago.

Status:ResolvedStart date:02/23/2013
Priority:NormalDue date:
Assignee:-% Done:

100%

Category:-
Target version:-

Description

It appears on i386 there may be a regression after

commit 79fd1696254e1413ba2bd0b0c029e2acd9ba416a
Date: Fri Feb 22 21:57:45 2013 -0800

kernel - Implement shared namecache locks

and possibly including

commit 1327cb0bd0b371b2d8c86c95cf0f6c84880b215b
Date: Sat Feb 23 11:49:31 2013 -0800

debug - vmpageinfo changes

that results in booting hanging at

mounting devfs

Attached is a dmesg of one of the machines. So far this has been tested on an Intel and an AMD64 machine that could run x86_64 but for which I also have a partition running i386.

I am compiling world and kernel using

# make -j7 buildworld && make CCVER=gcc44 buildkernel

asrock_z77m_i386_dmesg.txt Magnifier (9.96 KB) davshao, 02/23/2013 11:10 PM

History

#1 Updated by marino almost 2 years ago

i am not sure "&& make CCVER=gcc44" is doing anything.
for buildkernel, it should be "&& make WORLD_CCVER=gcc44"
(see man 5 make.conf)

I suspect the kernel has been building with gcc47 this whole time.

#2 Updated by davshao almost 2 years ago

Thanks to the recommendation to mv / delete /include on

http://bugs.dragonflybsd.org/issues/2505

I have been able to switch all of my machines to use the default gcc 4.7 for everything.

However I have also been able to reproduce the hang after

Mounting devfs

on multiple machines running i386.

#3 Updated by ftigeot almost 2 years ago

My own bisect tests on i386 point to 1b3342693b737646f3cab0715e31ec6ab5216b38 as the first problematic commit.
That is: <malloc.h>: Restrict support for <malloc.h> to !defined(__STDC__).

#4 Updated by ftigeot almost 2 years ago

... and this was the last good commit.

The devfs mount issue appears with the next one, ce94514e6b38abad5c78107e3d1ac401117df5c4
"kernel - Implementat much deeper use of shared VM object locks"

Sorry for the confusion.

#5 Updated by davshao almost 2 years ago

commit b3371fc101500242b133de85a1c4d10f67f53655
Date: Mon Feb 25 18:31:26 2013 -0800

kernel - Fix vm.shared_fault for vkernels and 32-bit

has been successfully tested on two machines running i386, one an Asrock Z77M Intel I3 and the other an Asus M5A87 AMD64, with successful booting and operation. Thanks!

#6 Updated by ftigeot almost 2 years ago

  • Status changed from New to Resolved
  • % Done changed from 0 to 100

Mark as resolved

Also available in: Atom PDF