xfs
[Top] [All Lists]

Re: Things todo before we announce

To: Steve Lord <lord@xxxxxxx>
Subject: Re: Things todo before we announce
From: "Andi Kleen" <ak@xxxxxxx>
Date: Thu, 30 Mar 2000 14:57:38 +0200
Cc: "Andi Kleen" <ak@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <200003301247.GAA12024@xxxxxxxxxxxxxxxxxxxx>; from lord@xxxxxxx on Thu, Mar 30, 2000 at 06:47:11AM -0600
References: <200003301247.GAA12024@xxxxxxxxxxxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Thu, Mar 30, 2000 at 06:47:11AM -0600, Steve Lord wrote:
> > 
> > P.S.: I compiled up a XFS tree from CVS last night. It created and mounted
> > fine, but hung the machine when I copied a kernel tree
> > to it (kdb didn't trigger, I'll need to go to the office to reset it) 
> 
> Hmm, we seem to have 'announced' already - I wonder who put this thread onto
> the oss list. Anyways, which options did you use to build XFS? There was a
> hard hang in the case where CONFIG_PAGE_BUF_META is turned on (doing a
> tlb_flush with interrupts turned off). I fixed this yesterday, but I suspect
> this would be after you checked out the cvs copy, so you may not have that
> yet - look for a spinlock inside of vfree and friends.
> 
> Let me know if you have this change or not, and what configuration you
> were running on.

Configuration: 

CONFIG_XFS_FS=m
CONFIG_PAGE_BUF=y
CONFIG_PAGE_BUF_LOCKING=y
CONFIG_AVL=y
CONFIG_PAGE_BUF_META=y
CONFIG_AVL=y
# CONFIG_XFS_ARCH_MIPS is not set
CONFIG_XFS_ARCH_NATIVE=y
# CONFIG_XFS_ARCH_MULTI is not set

I think I compiled with your change: 

ak@bert:~/lsrc/sgi/linux-2.3-xfs > cvs status linux/mm/vmalloc.c
===================================================================
File: vmalloc.c         Status: Up-to-date

   Working revision:    1.9
   Repository revision: 1.9     /cvs/linux-2.3-xfs/linux/mm/vmalloc.c,v
   Sticky Tag:          (none)
   Sticky Date:         (none)
   Sticky Options:      -ko

There is no spinlock in mm/vmalloc.c



> p.s. please don't regard this code as done - we have a ways to go yet.

I understand that, I was just curious. 

BTW, it does not boot on AMD CPUs when KDB is enabled because it uses
Intel specific MSRs unprotected. I sent a patch a few days ago to the kdb
list to fix this, but never got an response. You probably want to 
incorporate this patch before you announce to avoid lots of 
``my kernel doesn't boot with XFS'' reports.


-Andi

<Prev in Thread] Current Thread [Next in Thread>