david/ipxe
Archived
1
0
Commit Graph

33 Commits

Author SHA1 Message Date
Michael Brown
663904a7bc [infiniband] Split subnet management agent client out into ib_smc.c
Not all Infiniband cards have embedded subnet management agents.
Split out the code that communicates with such an embedded SMA into a
separate ib_smc.c file, and have drivers call ib_smc_update()
explicitly when they suspect that the answers given by the embedded
SMA may have changed.
2008-11-11 05:31:07 +00:00
Michael Brown
830e19eb54 [infiniband] Pass address vector in receive completions
Receive completion handlers now get passed an address vector
containing the information extracted from the packet headers
(including the GRH, if present), and only the payload remains in the
I/O buffer.

This breaks the symmetry between transmit and receive completions, so
remove the ib_completer_t type and use an ib_completion_queue_operations
structure instead.

Rename the "destination QPN" and "destination LID" fields in struct
ib_address_vector to reflect its new dual usage.

Since the ib_completion structure now contains only an IB status code,
("syndrome") replace it with a generic gPXE integer status code.
2008-11-11 05:31:07 +00:00
Michael Brown
0de5f7af6d [infiniband] Maintain queue fill level as a property of a work queue
Both queue owners and drivers often need to keep track of the fill
level, so let's make it a generic property.
2008-11-11 05:31:06 +00:00
Michael Brown
d9751edafa [infiniband] Flush uncompleted work queue entries at QP teardown
Avoid leaking I/O buffers in ib_destroy_qp() by completing any
outstanding work queue entries with a generic error code.  This
requires the completion handlers to be available to ib_destroy_qp(),
which is done by making them static configuration parameters of the CQ
(set by ib_create_cq()) rather than being provided on each call to
ib_poll_cq().

This mimics the functionality of netdev_{tx,rx}_flush().  The netdev
flush functions would previously have been catching any I/O buffers
leaked by the IPoIB data queue (though not by the IPoIB metadata
queue).
2008-11-11 05:31:06 +00:00
Michael Brown
35a5836677 [Infiniband] Move event-queue process from driver to Infiniband core 2008-04-21 13:23:11 +01:00
Michael Brown
a176a24ac0 [Infiniband] Add preliminary multiple port support for Hermon cards
Infiniband devices no longer block waiting for link-up in
register_ibdev().

Hermon driver needs to create an event queue and poll for link-up events.

Infiniband core needs to reread MAD parameters when link state changes.

IPoIB needs to cope with Infiniband link parameters being only partially
available at probe and open time.
2008-04-18 02:50:48 +01:00
Michael Brown
bc3f0a25c6 [Infiniband] Add preliminary support for multi-port devices.
Arbel and Hermon cards both have multiple ports.  Add the
infrastructure required to register each port as a separate IB
device.  Don't yet register more than one port, since registration
will currently fail unless a valid link is detected.

Use ib_*_{set,get}_{drv,owner}data wrappers to access driver- and
owner-private data on Infiniband structures.
2008-02-27 13:59:56 +00:00
Michael Brown
6570203571 [Infiniband] Centralise MAD operations
Pull out common code for handling management datagrams from arbel.c
and hermon.c into infiniband.c.

Add port number to struct ib_device.

Add open(), close() and mad() methods to struct ib_device_operations.
2008-02-26 22:56:19 +00:00
Michael Brown
bdac591726 Some interesting packet corruption happening now. 2007-09-17 13:31:13 +01:00
Michael Brown
2ed1acb9e9 Broadcast GID is now calculated by IPoIB layer. 2007-09-17 11:50:25 +01:00
Michael Brown
ab191e45db Unicasts seem to be working. :) 2007-09-17 10:39:30 +01:00
Michael Brown
b3d3814c17 Obtains a response to the get path record! 2007-09-17 08:36:56 +01:00
Michael Brown
e05a8cd4de Use pkey table access to determine broadcast GID directly. 2007-09-17 06:35:21 +01:00
Michael Brown
440e7926fb Dead code removal 2007-09-17 05:12:47 +01:00
Michael Brown
4e78a53cf2 IPoIB code separated out to ipoib.c. 2007-09-17 05:04:58 +01:00
Michael Brown
67836430e6 Read port GID directly using MAD IFC. 2007-09-17 02:54:15 +01:00
Michael Brown
3c6a6bdc5d Multicast join now works. 2007-09-17 00:24:44 +01:00
Michael Brown
7e85f0d296 create_qp() and destroy_qp() now written (but not tested). 2007-09-16 20:54:21 +01:00
Michael Brown
251cc84ed6 Started implementing create_qp() and destroy_qp(). 2007-09-16 19:03:24 +01:00
Michael Brown
b21d4ca21e Revert to dev_priv/owner_priv scheme, rather than container_of; it
makes it easier to put the generic allocation code into infiniband.c
2007-09-16 17:25:15 +01:00
Michael Brown
e238bb1e43 destroy_cq() now implemented (not tested). 2007-09-16 14:39:51 +01:00
Michael Brown
156b409ccc Rearrange data structures to maximise embedding (and hence minimise
the number of separate allocations that need to be done).
2007-09-16 04:02:20 +01:00
Michael Brown
37fc40bc8c post_recv() now works, and we can pass data on the IPoIB queue pair
using entirely our own code.
2007-09-15 23:33:25 +01:00
Michael Brown
8deef093d9 Direct polling of TX completion queue now works. 2007-09-15 20:58:29 +01:00
Michael Brown
6a791649f0 Updated MLX_* accessor macros to use implicit type information. 2007-09-15 18:44:09 +01:00
Michael Brown
a3a91fedc1 Started added poll_cq() verb.
Started reworking MLX_EXTRACT(), MLX_POPULATE() etc. to automatically
determine type information.
2007-09-15 15:40:35 +01:00
Michael Brown
970951666f arbel_post_send() has been observed to transmit a packet! 2007-09-15 02:53:05 +01:00
Michael Brown
8b27da9de1 Gets a response out of the hardware. (An error completion, to be precise.) 2007-09-15 01:35:07 +01:00
Michael Brown
38a73b55c4 Now at least compiles 2007-09-15 00:19:38 +01:00
Michael Brown
9d08b7c692 Starting to introduce an Infiniband device abstraction 2007-09-14 20:29:44 +01:00
Michael Brown
08e8dfd801 Now handling TX completions in our poll loop. 2007-09-14 11:10:25 +01:00
Michael Brown
30a19c3f1c Can now both send and receive packets. LL header format not yet
fixed; still using a quick hack-up just to be able to pass through
data.
2007-09-13 14:43:12 +01:00
Michael Brown
7b6d11e713 Started IB driver rewrite 2007-09-12 22:17:43 +01:00