osdir.com
mailing list archive F.A.Q. -since 2001!



Subject: [ANNOUNCE] mthca: alternative Mellanox HCA driver
- msg#00377

List: linux.drivers.openib

Mail Archive Navigation:
by Date: Prev Next Date Index by Thread: Prev Next Thread Index

Because of the large number of requests for a simpler HCA driver, I
decided to clean up my "science project" driver for the Mellanox HCA
and release it. The driver is extremely immature but does enough for
IPoIB to work. I created a branch of the subversion repository in
https://openib.org/svn/gen1/users/roland/mthca, and this new "mthca"
driver code is in the directory


https://openib.org/svn/gen1/users/roland/mthca/src/linux-kernel/infiniband/hw/mthca

This driver currently cannnot be used as a full replacement for
Mellanox's driver, as it is quite incomplete and does not support many
critical features (more details below).

However, this driver was written from scratch as a fully native Linux
2.6 driver (in fact, because CLEAR_BITMAP was for some reason renamed
to bitmap_zero, you will need to build against Linux 2.6.6 or newer).
I believe that the driver can serve at least as a good proof of
concept and should clarify the OpenIB driver architecture. The stack
with the mthca low-level driver looks like:

ib_core etc. (ib_mad, ib_client_query etc. will be gradually
merged into a single ib_core module):
The core IB midlayer.
ib_mthca:
A standard Linux PCI driver for the Mellanox HCA, which plugs
into ib_core.
ib_ipoib, ib_sdp, ib_srp, etc.:
Upper-level protocol drivers that use the ib_core API to
provide useful protocols over IB.

The mthca driver currently does _far_ less than Mellanox's mature HCA
driver. The following list of limitations is probably incomplete but
should give some idea of the level of completeness:

No userspace support: ib_mthca provides _no_ facilities for
userspace access to the HCA, beyond the generic MAD services
provided by the ib_useraccess module.
Firmware commands are polled: ib_mthca does not implement
event-driven firmware commands (which limits it to one
outstanding command at a time).
Only UD transport: ib_mthca does not handle transports other than
UD correctly, so the only functional ULP is IPoIB.
No FMR support: ib_mthca does not support Mellanox-style "fast
memory regions" as used by SDP and SRP.
No SRQ support: ib_mthca does not support shared receieve queues.
Only handles successful completions: the code to process
completions with an error status is not written yet.
Limited support for asynchronous events: only port change
asynchronous events will be dispatched.
Only physical memory registration: the register virtual memory
region verb is not implemented.
UD address vectors are in host memory: putting address vectors in
HCA memory increases performance; ib_mthca does not implement
this yet.

Comments, questions and discussion are very welcome. I'm happy to get
even the most nitpicking suggestions on the code.

- Roland

--
To unsubscribe send an email with subject unsubscribe to
openib-general@xxxxxxxxxxx
Please contact moderator@xxxxxxxxxx for questions.



Thread at a glance:

Previous Message by Date:

Re: OpenSM problem

Eitan, > Ports normally go through : > DOWN -> INIT (automatically by the hardware) > INIT -> ARMED (by OpenSM) > ARMED -> ACTIVE (by OpenSM or other traffic going through the port). > The Port can go back to INIT if the cable is not connected very well or > damaged. > Ports can stay in ARMED state if for some reason OpenSM failed to take them > to ACTIVE. > > Anyway if you use -V you should get a full detailed log of what is going on > in /tmp/osm.log. I can have a look if you want me to ... OpenSM behaves quite differently in each run. Sometimes cards get ACTIVE, sometimes not. In any case, OpenSM never stops itself (although it announces sometimes to do so), and it does not end after pressing ctrl-c. Here is a log of a run, where all cards stayed in ARMED state forever. -Thomas osm.log.bz2 Description: BZip2 compressed data pgpkKYWFURDc8.pgp Description: signature

Next Message by Date:

Re: kernel code from svn and 2.4

Thomas> Hi, just in case someone is interested, I attach a patch Thomas> that adds a patch for linux-2.4.21-15.EL. Thanks for the patch. I'm not sure I want to check it into our kernel patch directory as is, because I'm not sure that the approach of modifying files in the drivers/infiniband directory with the patch is the best way to go. As it stands there's not a requirement for a particular order between copying openib source into drivers/infiniband and applying the kernel patch; also, it is easy to update files in drivers/infiniband with newer versions without having to worry about whether they have been patched after being copied. What do other people think? - Roland -- To unsubscribe send an email with subject unsubscribe to openib-general@xxxxxxxxxxx Please contact moderator@xxxxxxxxxx for questions.

Previous Message by Thread:

Recall: openib kernel inclusion and performance.

Merav Dean would like to recall the message, "[openib-general] openib kernel inclusion and performance.". -- To unsubscribe send an email with subject unsubscribe to openib-general@xxxxxxxxxxx Please contact moderator@xxxxxxxxxx for questions.

Next Message by Thread:

Re: [ANNOUNCE] mthca: alternative Mellanox HCA driver

On Mon, May 17, 2004 at 10:16:17AM -0700, Roland Dreier wrote: > Because of the large number of requests for a simpler HCA driver, I > decided to clean up my "science project" driver for the Mellanox HCA > and release it. The driver is extremely immature but does enough for > IPoIB to work. I created a branch of the subversion repository in > https://openib.org/svn/gen1/users/roland/mthca, and this new "mthca" > driver code is in the directory > > > https://openib.org/svn/gen1/users/roland/mthca/src/linux-kernel/infiniband/hw/mthca any chance to upload a tarball and/or install viewcvs? The current svn web-frontend is pretty unusable, and viewcvs in addition to providing tarball snaphots of the repository also has lots of really nifty history/annotation/etc features. -- To unsubscribe send an email with subject unsubscribe to openib-general@xxxxxxxxxxx Please contact moderator@xxxxxxxxxx for questions.
blog comments powered by Disqus

Home | News | Sitemap | FAQ | advertise | OSDir is an Inevitable website. GBiz is too!