<feed xmlns='http://www.w3.org/2005/Atom'>
<title>imxrt-enet/src/bd, branch main</title>
<subtitle>Prototype ENET driver for i.MX RT MCUs
</subtitle>
<id>https://git.mciantyre.dev/imxrt-enet/atom?h=main</id>
<link rel='self' href='https://git.mciantyre.dev/imxrt-enet/atom?h=main'/>
<link rel='alternate' type='text/html' href='https://git.mciantyre.dev/imxrt-enet/'/>
<updated>2025-07-18T00:50:40+00:00</updated>
<entry>
<title>Rewrite descriptors with atomics</title>
<updated>2025-07-18T00:50:40+00:00</updated>
<author>
<name>Ian McIntyre</name>
<email>me@mciantyre.dev</email>
</author>
<published>2025-07-18T00:50:40+00:00</published>
<link rel='alternate' type='text/html' href='https://git.mciantyre.dev/imxrt-enet/commit/?id=6498f85de3078ad7c3206c22c690ecb3d0fa71bb'/>
<id>urn:sha1:6498f85de3078ad7c3206c22c690ecb3d0fa71bb</id>
<content type='text'>
The memory ordering ensures that operations to normal memory are
synchronized with operations on device memory, at runtime. I've seen
this play out in the transmit path: writes to the transmit buffer's
flags weren't reaching memory by the time the MAC was checking,
resulting in missing packets. Moving the fence works, but it's better to
use atomics.
</content>
</entry>
<entry>
<title>First commit</title>
<updated>2023-11-26T20:11:27+00:00</updated>
<author>
<name>Ian McIntyre</name>
<email>ianpmcintyre@gmail.com</email>
</author>
<published>2023-10-22T22:23:56+00:00</published>
<link rel='alternate' type='text/html' href='https://git.mciantyre.dev/imxrt-enet/commit/?id=c7e5123f6604fbb9ca510f01af9b60e777bf57b4'/>
<id>urn:sha1:c7e5123f6604fbb9ca510f01af9b60e777bf57b4</id>
<content type='text'>
A prototype of an i.MX RT ENET driver. There's design decisions I'm
thinking of changing. Nevertheless, the smoltcp support seems to be
working; an 1170EVK can act as a DHCP client and a TCP loopback server.
</content>
</entry>
</feed>
