Date: Tue, 28 Jan 1997 00:00:34 -0800 (PST) From: Simon Shapiro <Shimon@i-Connect.Net> To: Michael Smith <msmith@atrad.adelaide.edu.au> Cc: hackers@freebsd.org Subject: Re: 2.2-BETA Questions Message-ID: <XFMail.970128004129.Shimon@i-Connect.Net> In-Reply-To: <199701280316.NAA06348@genesis.atrad.adelaide.edu.au>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi Michael Smith; On 28-Jan-97 you wrote:
> > 1. Does anyone care? Coming from (too much) Linux, and seeing 2.1.6,
> > 2.2-BETa, 3.0... it is not a stupid question.
>
> Er, yes. Lots of people care. 2.2-BETA is the leadup to 2.2-RELEASE,
> the next production-level version. 3.0 is the current 'development'
> version, which will lead to a release probably sometime late this
> year. Exposing the devlopment process like this means that everyone
> can see where things are going.
Good. Coming from another freely distributed ``O/S'' over the last couple
of years, I am used to bi-weekly versions, etc. This is why I asked.
> > # mkdir /NewStuff
> > # mount -t nfs -o ro nomis;/usr/src/FreeBSD /NewStuff
> > # ls -al /NewStuff
> > ls: /NewStuff: Permission denied
>
> What are the permissions on "NewStuff" on the server? Try "ls" without
> any other flags first.
Permissions matter not. Can be 777, 1777 755, ...
The mount point simply disappears.
> It looks to me like the server is being _very_ weird. Someone else
> (Doug R.?) might have a better idea about that.
Yea, I figured this much, but the Linux ``gurus'' insist, theirs is the
only ``correct'' NFS server... As I said, it works (very well!) the other
way around...
>
> > 3. Made a kernel with sound, etc... Worked fine until some days ago.
> > Now, all of the sudden, without me doing anything (really :-):
> >
> > # xmcd -debug
> > ....
> > Lock file: /tmp/.cdaudio/lock.f02
> > Cannot open /dev/rcd0c: errno=6
>
> Is there a CD in the drive? 6 is "not configured", which xmcd should be
> telling you. A list of the boot-time probe messages (output of 'dmesg')
> would be handy here, as I suspect that your CD wasn't found.
Yup. So much so that sysinstall tells me it is not a current installation
CD! (Yes, it IS music, not iso9660 - Kitaro ``Ten Years'').
Actually, the problem may arise because there was a GNU distribution CD
in the drive at boot time. Works for data, does not for music, until the
next boot. Ugly.
Just checked. I can mount an is09660 CD with no problem (even a Linux
one :-), but all the CD music tools I found cannot open /dev/rcd0c
crw------- 1 root wheel 15, 536870912 Jan 24 16:01/dev/rcd0.ctl
crw-r----- 1 root operator 15, 0 Jan 24 16:01 /dev/rcd0a
crw-r----- 1 root operator 15, 2 Jan 24 16:01 /dev/rcd0c
crw-r----- 1 root operator 15, 3 Sep 15 18:16 /dev/rcd0d
What is the kernel config option for CD audio support?
>
> > 4. Shutdown questions:
> >
> > a. When init goes to single user, prompts, asking for a shell.
> > You press ENTER and it sits on ``(.???msg - Cannot exactly
> > remember) not found''
> > ^C will get you a prompt, most of the time. Sometimes you get
> > a fast roll talking about some malloc() failure. Sometimes a
> > ^C will stop it, sometimes it will not.
>
> Er "init goes to single user"? How are you shutting down?
# shutdown now
>
> > b. umount -a will leave things not in /etc/fstab mounted.
> > It always leaves root mounted RW, only to fsck it at boot.
> > Seems lie an unnecessary risk.
>
> You're _definitely_ not shutting down correctly. 'man shutdown'.
Thanx. But ``shutdown now'' is a valid option, generates no complaints
and gives the results indicated.
> > 5. More CD fun. Once a music CD is played, you cannot mount a data
> > cd because ``device is busy''. Reboot cures.
>
> Try exiting the CD-playing program first, if you aren't already.
Yup. Done that, been there. To no avail. Something IS wron.
> > 8. Education Question: What is the logic in assigning slice ID's?
> > I understand c to be the entire disk
> > (why `c'? Why not?)
> > Why does sysinstall assign 'e', 'f',
> > but (almost) never 'd'?
>
> You mean partition names. Tradition, mostly. 'a' is traditionally
> used for a root filesystem, 'b' for swap, 'c' for the whole disk, and
> d-h for 'other' partitions. For a while, 'd' was used by various
> 386 BSD's to deal with the disparity between "the whole disk" and
> "the whole part of the disk that BSD uses"; this is obsoleted by
> the 'slice' paradigm.
Thanx! now I know.
> > 9. Some safety checks in disklabel and newfs and/or kernel slice-
> > partition handling could be nice. If you create an 'a' partition
> > which is exactly an overlap of a 'c' in a slice that dominates
> > the disk, newfs will FREEZE the system.
>
> Novel. I've never seen that, and I've done it many times.
I checked some more. It has to do with the manner in which disklabel is
called. If you have it use /etc/disktab, all is well. The -e is only safe
if you initialized with the (-rwB?) /etc/disktab option.
> > 10. Kernel Question: On an i386 PC, how does one make sure that
> > another driver does not use the same ISA ports as you do?
> > You are trying to be nice and NOT use something someone else is
> > already using. There is a Linux thing to do that...
>
> ISA resource allocation is a particularly noisome can of worms.
> Currently, if your driver is configured with a base address in a
> region previously claimed by another driver, your probe routine won't
> be called. That can obviously cause problems if you plan to probe
> several possible port ranges in a single probe routine.
Perfect! Linux drivers seem to explicitly call some routine that registers
the addresses. I like hte BSD solution better.
> If you have any particular ideas or requests here, please raise them,
> as we're always open to suggestions on cleaning this up.
A blow torch and a stick of dynamite :-) Reason for asking is a PCI
controller that can be configured (via BIOS) to ``sit'' in ISA adress
range. The Linux driver probes to see that it does not (essentially)
overlap with an IDE controller which it can emulate...
> > 11. Another Kernel question: A device driver for a controller that
> > is available in ISA, EISA and PCI. How do you split the code?
> > We put the PCI part in pci, the ISA/EISA parts in i386/{isa,eisa}?
> > But the code is NOT i386 dependant. We are putting it in dev/dpt.
> > Is that a good choice?
>
> Perhaps:
> - have three seperate drivers (bad idea).
> - look at the 'ahc' and 'bt' drivers; the former is pci/eisa, the latter
> is pci/isa. The 'ahc' driver also has code in dev/.
Yes, but pieces still sit in i386. We will avoid that. Thanx!
Simon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.970128004129.Shimon>