Date: Thu, 24 May 2001 09:37:15 -0700 (PDT) From: John Baldwin <jhb@FreeBSD.org> To: Michael Harnois <mdharnois@home.com> Cc: freebsd-current@FreeBSD.org, The Hermit Hacker <scrappy@hub.org> Subject: Re: Tonights panic: free_newdirblk+0x73: movl %eax,0x10(%e Message-ID: <XFMail.010524093715.jhb@FreeBSD.org> In-Reply-To: <86vgmqvlkb.fsf@mharnois.workgroup.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 24-May-01 Michael Harnois wrote: > On Wed, 23 May 2001 22:10:09 -0300 (ADT), The Hermit Hacker <scrappy@hub.org> > said: > > > Oops ... neat, now it just locked up solid and ctl-alt-esc > > doesn't even get me to debugger :( > > This has been my experience for about a week on the -current kernel > ... I held out high hopes that these changes by John and Alfred over > the past 24 hours or so would fix it, but no go ... I had a epiphany this morning, and will be committing a bandaid to get current workign again on SMP systems. The problem occurs when we take a fault while hold the vm lock already. In that case, the VM was in an inconsistent state when we faulted. That's why we were holding the lock when we faulted. Well, there are places in the fault handler where we release the vm lock to sleep or acquire Giant or some such. When we do so, we allow another thread to try and dink with the VM, which can lead to threads trying to use free'd pages, etc. on SMP systems. The bandaid is to stick all the VM related syscalls that were marked MPSAFE back under Giant for the time being, which I will do shortly. -- John Baldwin <jhb@FreeBSD.org> -- http://www.FreeBSD.org/~jhb/ PGP Key: http://www.baldwin.cx/~john/pgpkey.asc "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.010524093715.jhb>