Add a generic iterator for the descriptors in an LSA, with
specializations for the descriptor type.

This introduces a callback mechanism to decouple the LSA type from the
function that operates on each of the descriptors.

It prepares for further specialization of the iterators for LSAs that
contain TLVs, while reusing the callback that operate on the descriptors
contained in the TLVs.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

…terator

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Add utility function to find the Nth TLV in an E-LSA.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Gotcha:
The E-Entra-Area-Prefix LSA has removed the field containing the number
of prefixes from the LSA, even though RFC 8362 says: "all LSA Header
fields are the same as defined for the Intra-Area-Prefix-LSA".

In order to report the number of prefixes, the iterator has to keep count.

Signed-off-by: Andrew Cooks <[email protected]>

to appease checkpatch and the CI system.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

This adds a handler and .lh_get_prefix_str callback for E-Router LSAs,
similar to ospf6_router_lsa_get_nbr_id().

Signed-off-by: Andrew Cooks <[email protected]>

This adds a handler and .lh_get_prefix_str callback for E-Network LSAs,
similar to ospf6_network_lsa_get_ar_id().

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

This adds a handler and .lh_get_prefix_str callback for E-Link LSAs,
similar to ospf6_link_lsa_get_prefix_str()

Signed-off-by: Andrew Cooks <[email protected]>

Add a handler and .lh_get_prefix_str callback for E-Intra-Area-Prefix
LSAs, similar to ospf6_intra_prefix_lsa_get_prefix_str().

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

The inline lookup doesn't handle all the necessary types, and there is
already a convenient name lookup function, so use that instead.

Signed-off-by: Andrew Cooks <[email protected]>

There are multiple functions that deal with both struct
ospf6_network_lsdesc and struct ospf6_router_lsdesc and resort to using
generic pointers and macros to cast these pointers to the expected type
or access the struct members.

The problem with that pattern is that not only is it more verbose and
harder to read the VERY_LONG_CAPITALIZED_MACRO_NAMES, but they also
break type checking when passed an incorrect pointer, hide endian
conversions and introduce another name for the struct members being
used.

Signed-off-by: Andrew Cooks <[email protected]>

Use the new struct ospf6_lsdesc *lsdesc instead of a caddr_t *lsdesc
when the code path needs to handle lsdesc as either
struct ospf6_network_lsdesc or struct ospf6_router_lsdesc.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

This applies selective formatting changes made using tools/indent.py.

ospf6d is inconsistent with the code style in numerous places, which
makes it cumbersome to use indent.py to apply consistent formatting
to any non-trivial change, due to the large number of formatting changes
it makes to otherwisse untouched code.

This commit contains a small subset of the changes indent.py proposes,
and reduces the number of irrelevant changes to consider when staging
a commit.

Signed-off-by: Andrew Cooks <[email protected]>

Add a tri state to adjust LSA origination
- legacy: originate legacy LSA only
- elsa: originate Extended-LSA types only
- both: originate legacy and elsa

Signed-off-by: Jacob Lodge <[email protected]>
Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Process either E-LSAs or legacy LSAs, or both in SPF calc.

If a neighbour is originating legacy LSA and we cant find an E-LSA when
performing spf calc then try check for a legacy LSA.

Co-authored-by: Jacob Lodge <[email protected]>
Co-authored-by: Andrew Cooks <[email protected]>
Signed-off-by: Andrew Cooks <[email protected]>

Remove unecessary indentation;
remove using the word 'type' as a variable name;
replace zero initialization with explicit assignment in the 'else'
block.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

The purpose of this test is to check that ospf6d can operate on Legacy
LSAs, without originating or receiving any of the TLV-based LSAs of
RFC 8362.

It checks for the presence of the Legacy LSAs, and when that passes, it
checks for the absence of any E-LSAs.

"Absence" means:
'not detected during a short time interval following the stable
convergence'

This obviously does not prove that an E-LSA could ever exist,
but hopefully it shows that E-LSAs can be disabled.

Signed-off-by: Andrew Cooks <[email protected]>

Improve the handling of the pos parameter, so that larger-than-expected
values are safe when calling the .lh_get_prefix_str callback.

Signed-off-by: Andrew Cooks <[email protected]>

When the lh_get_prefix_str() callback correctly handles the pos
parameter, the specialised code path is no longer needed.

Signed-off-by: Andrew Cooks <[email protected]>

Complementary to the topotest for legacy-only LSAs, the purpose of this
test is to check that an ospf6d network can operate on E-LSAs, without
originating or receiving any legacy LSAs defined in RFC 5340.

It checks for presence of E-LSAs and when that passes, checks for
absence of legacy LSAs.

"Absence" means: Not detected during a short time interval following expected
convergence.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

This change moves the origination of inter-area LSAs out,
so that summarization and origination are clearly delineated.

No functional change.

Signed-off-by: Andrew Cooks <[email protected]>

Add origination of both E-Inter-Area-Router-LSA and
E-Inter-Area-Prefix-LSA.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

remove magical caddr_t + 4 pointer
and subsequent call of FOO_LSDESC_GET_BAR macro on said pointer.

Signed-off-by: Andrew Cooks <[email protected]>

The pattern where a macro receives a pointer, immediately casts it,
accesses a struct member and also converts byte order is surely not a
recipe for success...

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Use indent.py to apply coding style changes and selectively commit the
changes to logging calls.

Not only is the newer style easier to read, but committing the style
changes separately from surrounding functional changes will make functional
changes easier to stage and review.

Signed-off-by: Andrew Cooks <[email protected]>

When calculating inter-area routes, consider TLV-based E-Intra-Area-Prefix
LSA in addition to Intra-Area-Prefix LSA.

Signed-off-by: Andrew Cooks <[email protected]>

When calculating inter-area routes, consider TLV-based
E-Inter-Area-Router LSA in addition to Inter-Area-Router LSA.

Signed-off-by: Andrew Cooks <[email protected]>

Attached Routers TLV contains multiple Router IDs. Print all of them.

In RFC8362, only a single Attached-Routers TLV is applicable to
E-Network-LSA, so for now, the end of the TLV is the end of the LSA.

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Andrew Cooks <[email protected]>

previous code was only checking lsdb for legacy inter lsa to
determine lsid

recursively check for free lsid for IAP and EIAP

find a free link state id values across both extended and legacy LSA's

Signed-off-by: Jacob Lodge <[email protected]>
Signed-off-by: Andrew Cooks <[email protected]>

Signed-off-by: Jacob Lodge <[email protected]>