Re: Firefox hangs in state 'kqread'
:...
:I have also recompiled -Stable dating from 02/22. But after the usual "make
:buildworld" and "make buildkernel", I only did "make installkernel", and
:rebooted, getting a -Stable kernel on a -CURRENT world.
:(I know that is not something to do, but I could not resist...)
:I checked if Opera would hang: it did.
:Then, I did a "make installworld", rebooted, checked Opera: it did not hang,
:whatever I tried.
:
:I am currently recompiling the latest -Stable to try the same thing, but I
:will not be able post the results before tomorrow. I will try with the latest
:-CURRENT as well.
:
:So I guess this is a userspace issue, and the problem is in the difference
:between -CURRENT and -Stable...
:
:Regards,
:
:Laurent
That's good sleuthing. Clearly you are correct, it has to be a userspace
issue. Hiten had a thought which I will include:
: (Hiten)
:Matt, I have not touched libc in a very long time, but if you
:remember that once when I modified the locking code in libc it
:made Mozilla give weird timeouts on DNS lookups because the
:lockup hierarchy was messed.
:
:Maybe you want to check the recent libc related commits that
:are to do with some form of serialisation, etc.
Considering the amount of work that has gone into libc recently,
this is a very real possibility. Now the question is finding what was
changed in the (locking code probably) that created the problem.
-Matt
Matthew Dillon
<dillon@backplane.com>
討論串 (同標題文章)
完整討論串 (本文為第 17 之 27 篇)：