Swapping on RAID-5 should be 100% safe over the last two years. All the
required resources are pre-allocated to ensure that deadlocks are avoided;
the basic RAID-5 structure, the raid stripe, contains all the buffer
heads and actual buffers required for the completion of that RAID stripe,
and the stripe structures are being recycled during actual use.
We planned to do the same for RAID-1, but never actually implemented it,
so there we have the deadlock potential (although to a much lesser
extent than the very first version of the RAID-5 code -- RAID-1 allocates
only the smaller buffer head structures, rather than actual buffer space).
Gadi
On Fri, 11 Jun 1999, Pavel Machek wrote:
> Hi!
>
> > I've been told that the problem is that RAID1 and RAID5 make_request()
> > use kmalloc() and can therefore deadlock if a swap is needed to complete
> > the malloc during swapping.
> >
> > However, RAID0 does not malloc at all, so I can assume that its not
> > affected. Can I suggest Pavel's documentation patch then be changed to
> > specify that its only RAID 1 and 5 that cannot be used for swapping?
>
> Well - as there's no need to use RAID0 (and you knew that), I do not
> see urgent need for your change. But yes I updated my home tree, and
> if Linus will not accept the patch, my next tries will go with update
> similar to yours.
>
> Pavel
>
> > Suggested...
> >
> >
> > @@ -1113,6 +1119,8 @@
> ~ you need to change his number to 11 ;-)
> > Software-RAID mini-HOWTO, available via FTP (user: anonymous) from
> > ftp://metalab.unc.edu/pub/Linux/docs/HOWTO/mini. There you will also
> > learn where to get the supporting user space utilities raidtools.
> > +
> > + It is very bad idea to swap onto a RAID 1 or 5 device. Swapping over
> > + RAID 0 does work, however it may be easier to just add seperate swap
> > + partitions with equal priorities and let the kernel handle the
> > + balancing.
> >
> > If unsure, say N.
>
>
> --
> I'm really pavel@ucw.cz. Look at http://195.113.31.123/~pavel. Pavel
> Hi! I'm a .signature virus! Copy me into your ~/.signature, please!
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/
>
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/