Using buffered IO risks non-atomic updates to parts of the
device that we don't actually want to write to. This isn't in
general safe.
So switch to O_DIRECT for all that IO and make sure we have
properly aligned buffers.
When loading the metadata for a subarray (super_by_fd), we set
->subarray to be the name read from md/metadata_version so that
getinfo_super can return info about the correct array.
With this we can differentiate between a container and
an array within the container by looking at ->subarray[0].
Using write_init_super to add a spare to an active array is quite
different to how it is used when creating an array.
It mostly works, but if we are adding two devices to an array,
then when we add the second, there are still traces of the first
which confuse write_init_super.
So get write_init_super to ignore those traces. Longer term, we
probably want to do this differently as for DDF, hot-adding to
an active array will have to be quite different - it will want to
write to all metadata, possibly via mdmon.
It is important that dup_super always returns an 'st' with the same
->ss and ->minor_version as the st that was passed.
This wasn't happening for 0.91 metadata (i.e. in the middle of a reshape).
The current model for creating arrays involves writing
a superblock to each device in the array.
With containers (as with DDF), that model doesn't work.
Every device in the container may need to be updated
for an array made from just some the devices in a container.
So instead of calling write_init_super for each device,
we call it once for the array and have it iterate over
all the devices in the array.
To help with this, ->add_to_super now passes in an 'fd' and name for
the device. These get saved for use by write_init_super. So
add_to_super takes ownership of the fd, and write_init_super will
close it.
This information is stored in the new 'info' field of supertype.
As part of this, write_init_super now removes any old traces of raid
metadata rather than doing this in common code.
There is still a problem: If array is partially assembled and started
read-only, the last device doesn't get added properly. Probably a kernel
problem.
The new superblock needs to have a new disk.number. This is a bit of a hack...
Fix handling of negative bitmap offsets on 64bit hosts.
The bitmap offset is a signed 32bit number, so casting to (long)
isn't sufficient. We must cast to (int32_t).
Fix various problems with --grow --add for linear.
The code to add a drive to a live linear array had never
been tested properly and so was buggy. This tidies it up
and means that the new regression-test passes.
udev likes to get information about a device as key=value pairs so it
can create disk/by-id links etc. So add --export flag which causes
the output of --detail to easily parsable.
From: Kay Sievers <kay.sievers@novell.com>
Depending on the size of the array we reserve space for up to 128K
of bitmap, and we use it where possible.
When hot-adding to a version 1.0 we can still only use the 3K at the
end though - need a sysfs interface to improve that.
If a small chunksize is requested on Create, we don't auto-enlarge
the reserved space - this still needs to be fixed.
It is much better for this sort of thing to be predictable
rather than depending on what devices exist, especially as
these days entries in /dev/ often don't pre-exist.
So make it always
/dev/mdX
for version0 superblocks.
Version1 are always /dev/md/NAME
Make -assemble a bit more resilient to finding strange
information in superblocks.
Don't claim newly added spares are InSync!! (don't know why that
code was ever in there)
Apparently there are license issues with openssl, so
just use sha1.c always. This means we can get rid of
SHA1.c
Signed-off-by: Neil Brown <neilb@suse.de>
From: Luca Berra <bluca@vodka.it>
glibc 2.4 is pedantic on ignoring return values from fprintf, fwrite and
write, so now we check the rval and actually do something with it.
in the Grow.c case i only print a warning, since i don't think we can do
anything in case we fail invalidating those superblocks (is should never
happen, but then...)
Signed-off-by: Neil Brown <neilb@suse.de>
This cannot be used yet, but it is working towards auto-assembly.
When auto-assembling an array, we make a name in /dev/md/
giving a number (from the peferred minor) or name (from set-name).
Signed-off-by: Neil Brown <neilb@suse.de>
i.e. if assembling with --name or --super-minor, then if we find two
different arrays with the same apparent identity, and one was built
for 'this' host, then prefer that one instead of giving up in disgust.
Signed-off-by: Neil Brown <neilb@suse.de>
When an array is created, if the homehost is know,
the superblock gets it, either in the uuid, (via sha1)
or in the name field.
Signed-off-by: Neil Brown <neilb@suse.de>
For 'force' and 'assemble', update_super must return true
if anything was changed.
Also fix a bug with wonly handling in super0.
Signed-off-by: Neil Brown <neilb@suse.de>
### Diffstat output
./super0.c | 18 ++++++++++++++----
./super1.c | 7 +++++++
2 files changed, 21 insertions(+), 4 deletions(-)
diff ./super0.c~current~ ./super0.c
clean up 'long long' usage for size of array, so that
with v-1 superblocks a raid1 larger than 2TB is possible.
Signed-off-by: Neil Brown <neilb@suse.de>
Currently this includes
--write-behind to set level of write-behind supported
--write-mostly to flag devices as write-mostly.
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
This includes:
adding --metadata= option to choose metadata format
adding metadata= word to config file.
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Neil Brown
Kay Sievers
Kay Sievers