2010/11/4 Marc Kleine-Budde <m...@pengutronix.de>:
> this is the current version of Andre's async mcp2515 driver. It contains
> the following fixes/improvements since Andre's last posting:

Great! Thanks.

> On my test system (ARM i.MX35) the driver performs much better than the
> synchronous one with regards to receiving frames in the correct order and
> total frame loss. Also I was not able to reproduce the repeated-tx-messages
> problem that has been observed with the synchronous driver.

Indeed. I finally had some time to capture the oscilloscope
screenshots I promised, comparing the mcp251x driver in the current
net-2.6 git and the mcp2515 driver I sent a couple of weeks ago, and
they also show that.

In this one, we can see the overhead introduced by the threaded
interrupt model used by the mcp251x driver:
http://clientes.netvisao.pt/anbadeol/mcp2515-1frame.png

At 500kbps, the mcp2515 driver can handle frame reception interrupts
faster than the transmission of the next successive frame. Not so with
the mcp215x driver, which is already struggling to keep up:
http://clientes.netvisao.pt/anbadeol/mcp2515-500kbps.png

At 1Mbps, the mcp251x driver quickly starts losing frames:
http://clientes.netvisao.pt/anbadeol/mcp2515-1mbps.png

My test system is a Gumstix Verdex board with a 400MHz PXA270 ARM
processor, which has a hardware SPI controller, DMA capable.

> Andre, can you please make the license of the driver consistent.
> GPLv2+ in the comments vs. MODULE_LICENSE("GPL").

I did, and I think the second version I sent, the one in your git, is
already consistent: I used MODULE_LICENSE("GPL"), like most drivers in
the kernel, which according to include/linux/module.h means "GNU
Public License v2 or later", which is what I used in the header
comments.

> For those who're interested, the individual patches are in this branch:
> http://git.pengutronix.de/?p=mkl/linux-2.6.git;a=shortlog;h=refs/heads/can/mcp2515-async

Thanks!

> The driver still lacks CAN error handling, though.

Ok, I'll try to implement that.

> Please review the driver,

I will. Looks very good. Thanks!

-- 
Andre B. Oliveira
http://clientes.netvisao.pt/anbadeol/
_______________________________________________
Socketcan-core mailing list
Socketcan-core@lists.berlios.de
https://lists.berlios.de/mailman/listinfo/socketcan-core

Reply via email to