The legacy behavior adds the pgid seed and pool together as the input for
CRUSH.  That is problematic because each pool's PGs end up mapping to the
same OSDs: 1.5 == 2.4 == 3.3 == ...

Instead, if the HASHPSPOOL flag is set, we has the ps and pool together and
feed that into CRUSH.  This ensures that two adjacent pools will map to
an independent pseudorandom set of OSDs.

Advertise our support for this via a protocol feature flag.

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

Instead of using the old ceph_object_layout struct, update our internal
ceph_calc_object_layout method to use the ceph_pg type.  This allows us to
pass the full 32-bit precision of the pgid.seed to the callers.  It also
allows some callers to avoid reaching into the request structures for the
struct ceph_object_layout fields.

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

Support (and require) the PGID64, PGPOOL3, and OSDENC protocol features.
These have been present in ceph.git since v0.42, Feb 2012.  Require these
features to simplify support; nobody is running older userspace.

Note that the new request and reply encoding is still not in place, so the new
code is not yet functional.

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

Always decode data into our cpu-native ceph_pg type that has the correct
field widths.  Limit any remaining uses of ceph_pg_v1 to dealing with the
legacy protocol.

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

Rename the old version this type to distinguish it from the new version.

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

The variable "str" is used as both the source and destination in
function snprintf(), which is undefined behavior based on C11. The
original description in C11 is:
	"If copying takes place between objects that
	overlap, the behavior is undefined."

And, the function of ceph_osdmap_state_str() is to return the osdmap
state, so it should return "doesn't exist" when all the conditions
are not satisfied. I fix it in this patch.

[elder@inktank.com: shortened the commit message]

Signed-off-by: Cong Ding <dinggnu@gmail.com>
Reviewed-by: Alex Elder <elder@inktank.com>

ceph_calc_file_object_mapping() takes (among other things) a "file"
offset and length, and based on the layout, determines the object
number ("bno") backing the affected portion of the file's data and
the offset into that object where the desired range begins.  It also
computes the size that should be used for the request--either the
amount requested or something less if that would exceed the end of
the object.

This patch changes the input length parameter in this function so it
is used only for input.  That is, the argument will be passed by
value rather than by address, so the value provided won't get
updated by the function.

The value would only get updated if the length would surpass the
current object, and in that case the value it got updated to would
be exactly that returned in *oxlen.

Only one of the two callers is affected by this change.  Update
ceph_calc_raw_layout() so it records any updated value.

Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>

Add libceph support for a new CRUSH tunable recently added to Ceph servers.

Consider the CRUSH rule
  step chooseleaf firstn 0 type <node_type>

This rule means that <n> replicas will be chosen in a manner such that
each chosen leaf's branch will contain a unique instance of <node_type>.

When an object is re-replicated after a leaf failure, if the CRUSH map uses
a chooseleaf rule the remapped replica ends up under the <node_type> bucket
that held the failed leaf.  This causes uneven data distribution across the
storage cluster, to the point that when all the leaves but one fail under a
particular <node_type> bucket, that remaining leaf holds all the data from
its failed peers.

This behavior also limits the number of peers that can participate in the
re-replication of the data held by the failed leaf, which increases the
time required to re-replicate after a failure.

For a chooseleaf CRUSH rule, the tree descent has two steps: call them the
inner and outer descents.

If the tree descent down to <node_type> is the outer descent, and the descent
from <node_type> down to a leaf is the inner descent, the issue is that a
down leaf is detected on the inner descent, so only the inner descent is
retried.

In order to disperse re-replicated data as widely as possible across a
storage cluster after a failure, we want to retry the outer descent. So,
fix up crush_choose() to allow the inner descent to return immediately on
choosing a failed leaf.  Wire this up as a new CRUSH tunable.

Note that after this change, for a chooseleaf rule, if the primary OSD
in a placement group has failed, choosing a replacement may result in
one of the other OSDs in the PG colliding with the new primary.  This
requires that OSD's data for that PG to need moving as well.  This
seems unavoidable but should be relatively rare.

This corresponds to ceph.git commit 88f218181a9e6d2292e2697fc93797d0f6d6e5dc.

Signed-off-by: Jim Schutt <jaschut@sandia.gov>
Reviewed-by: Sage Weil <sage@inktank.com>

Define and export function ceph_pg_pool_name_by_id() to supply
the name of a pg pool whose id is given.  This will be used by
the next patch.

Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>

Ensure that we set the err value correctly so that we do not pass a 0
value to ERR_PTR and confuse the calling code.  (In particular,
osd_client.c handle_map() will BUG(!newmap)).

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>