ext2fs seems to get swapped out

2010-07-02 Thread Samuel Thibault
Hello,

Still having a glance at memleaks:

9 S 545M 35.6M 29.6  6.5 /hurd/ext2fs -A /dev/hd2

This seems a lot. And actually, killing it frees a couple of 100MiB swap
space.  AIUI, this means that some pages of ext2fs get swapped instead
of written back by ext2fs.  I can understand how that could happen if
ext2fs is quite busy and can't answer to gnumach requests fast enough,
but shouldn't there be a mechanism to clear this up when more memory is
available to perform it?

Samuel



Re: News 2010-06

2010-07-02 Thread Samuel Thibault
Arne Babenhauserheide, le Fri 02 Jul 2010 21:23:52 +0200, a écrit :
> I just wrote the Month of the Hurd for last month. 
> 
> A month of the Hurd: *initrd*, *clustered page-in* and *auth server fixes*. 
> 
> > This month Jérémie Koenig wrote a 
> > [preliminary 
> > patch](http://lists.gnu.org/archive/html/bug-hurd/2010-06/msg00047.html)
> > to add initrd support in Mach for the Debian installer. 

It is worth mentioning that with all his various fixes, he ended up
installing a hurd system from d-i!  There were still hacks here and
there, but that means the d-i project will most probably be successful.

It might also be worth mentioning that I fixed a leak in the pfinet
server which would usually push buildds on their knees after 3-4
days building packages. Since then, bach (on which I could just
restart pfinet without even interrupting the builds going on) has been
continuously building stuff for 7 full days now, and still going on.

Samuel



News 2010-06

2010-07-02 Thread Arne Babenhauserheide
Hi, 

I just wrote the Month of the Hurd for last month. 

A month of the Hurd: *initrd*, *clustered page-in* and *auth server fixes*. 

> This month Jérémie Koenig wrote a 
> [preliminary 
> patch](http://lists.gnu.org/archive/html/bug-hurd/2010-06/msg00047.html)
> to add initrd support in Mach for the Debian installer. 
> 
> Also Karim Allah Amed added the 
> [first 
> patch](http://lists.gnu.org/archive/html/bug-hurd/2010-06/msg00023.html) 
> to port the clustered paging-in from OSF Mach to GNU Mach which should 
> improve the virtual memory performance of the Hurd. 
> 
> And Carl Fredrik Hammar added patches to improve the stability of the auth 
> server (
> [port 
> death](http://lists.gnu.org/archive/html/bug-hurd/2010-06/msg3.html) and 
> [null 
> port](http://lists.gnu.org/archive/html/bug-hurd/2010-06/msg00019.html)). 

Please tell me what I missed! 

Best wishes, 
Arne