Re: sparc patches commited
Nick Holgate wrote:
>
> Hi Eric,
>
> I see you've just committed some Spark fixes to the boot-floppies CVS tree.
> I'm not sure what they actually where as I don't have CVS access, but it
> appears as though you are working on a similar installation goals to me,
> i.e. diskless/serial console. I noticed my name in one of the log messages
> on the boot-floppies list and assume that it was to do with my posting on
> fixing inittab and my inadequate knowledge of 'sed'. If so I will put down
> my book of 'sed' and concentrate on other matters.
Happy to see I'm not the only one interested in diskless & serial console
support in dbootstrap :-))
I tried to improve the diskless support by displaying adequat messages when no
local disks are found. Now dbootstrap doesn't tell you about partitionning &
formating the local disk(s). It starts with configuring the network before
doing anything else, then allows you to mount any NFS partition as the rootfs.
It skips the bootloader configuration step too.
> I also noticed another log message about a fix to select_not_mounted.c
> causing a seg fault when trying top mount an NFS partition. I too have fixed
> a bug in that file to do with mounting NFS partitions, perhaps it's the same
> one? In addition to that I also found there is another bug in
> libfdisk/fdisk.c that also causes a seg fault with NFS partitions, I have
> attached a patch that I sent to Enrique Friday evening, just a little late
> for 2.1.6, that you might be interested in.
The first patch (in select_not_mounted.c) is the same ;-)
About the second one, I guess you can commit it (I did'nt find it but I don't
say it was not a bug ;-) ).
> I have also posed a suggestion to Enrique about having an NFSROOT variable
> in /etc/default/rcS that could be set by dinstall and checked by the init
> scripts so that an NFS root filesystem is not fsck'd at boot. Have you any
> comments on this? I have attached a patch showing my suggested changes to
> the init scripts.
Well, in fact fsck of / is no good with rootfs on NFS, but actually it doesn't
hurt because fsck aborts & returns 0 on NFS filesystems.
Don't know if your second patch in mountall.sh is required because no binaries
outside / (/sbin & /bin) are accessed between the execution of mountall.sh &
mountnfs.sh. Moreover, not mounting other NFS partitions before the network
is fully configured allows you to mount any NFS partitions (other than rootfs)
even beyond a gateway.
Regards.
--
Eric Delaunay | "La guerre justifie l'existence des militaires.
delaunay@lix.polytechnique.fr | En les supprimant." Henri Jeanson (1900-1970)
Reply to: